The copy in libdrm has not been maintained so much since mesa (being the
only consumer) copied this into the mesa tree in ~2018)
Signed-off-by: Rob Clark <robclark@freedesktop.org>
drmAllocCpy() and many resource-fetching functions took kernel-returned
counts as int or __u32, multiplied them by sizeof(...) as ints, and
passed the product to drmMalloc(int size). On LP64 systems this
truncates any size_t result down to a signed 32-bit quantity. Products
at or above 2^31 become negative and are sign-extended when passed to
calloc(), so the allocation either fails (yielding a NULL that is
subsequently used as a write target) or, combined with an overflowing
multiply, succeeds with an undersized buffer.
drmAllocCpy() compounded this: it multiplied count and entry_size as
ints before handing the result to drmMalloc(), and iterated with an int
index. A large count could wrap the product to a small positive value,
producing an undersized allocation filled by a full-count memcpy.
Fix by replacing drmMalloc(count * sizeof(T)) with calloc(count,
sizeof(T)) at all call sites where count originates from the kernel.
calloc's two-argument form performs the multiplication in size_t and
rejects overflowing products, and drmAllocCpy now takes size_t
parameters and iterates with a size_t index so all arithmetic stays in
an unsigned 64-bit domain on LP64.
Affected functions: drmAllocCpy, drmModeGetResources,
drmModeGetConnector, drmModeGetProperty, drmModeObjectGetProperties,
drmModeAtomicDuplicate, drmModeAtomicCommit, drmMarkBufs, drmGetBufInfo,
drmMapBufs, drmDMA, drmGetReservedContextList.
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
In the /proc/dri name lookup path, the return value of read() was
tested with a plain truthiness check before being used to index the
stack buffer: "if (retcode) buf[retcode-1] = '\0';". read() returns
-1 on error, which is truthy, so the code would write to buf[-2],
corrupting memory before the start of the buffer.
Test "retcode > 0" instead so the NUL terminator is only written when
at least one byte was actually read.
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
_drmModeGetConnector() allocates arrays for the connector's props,
prop values, modes and encoders, then re-runs the GETCONNECTOR ioctl.
If a hotplug event between the two ioctls raises any of the counts, the
allocated buffers are too small, so they are freed and the function
loops back to retry with the larger counts.
On that retry path the freed pointers were left dangling in the local
conn struct. During the retry each pointer is only re-assigned when its
corresponding count is non-zero. If the subsequent ioctl reports a count
of zero for one of those fields, the stale (already freed) pointer
survives all the way down to err_allocs, where it is freed a second
time, resulting in a double free.
NULL out each pointer immediately after freeing it on the retry path so
that a zero count on the next iteration leaves the pointer as NULL and
the cleanup in err_allocs becomes a harmless drmFree(NULL).
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Generated using make headers_install
Generated from drm-next branch commit 565626e74db94686f2389cbdea845891f468f044
Signed-off-by: Simon Ser <contact@emersion.fr>
Since glibc-2.43:
For ISO C23, the functions bsearch, memchr, strchr, strpbrk, strrchr,
strstr, wcschr, wcspbrk, wcsrchr, wcsstr and wmemchr that return pointers
into their input arrays now have definitions as macros that return a
pointer to a const-qualified type when the input argument is a pointer
to a const-qualified type.
fixes:
tests/modetest/modetest.c: In function 'parse_connector':
tests/modetest/modetest.c:1997:22: warning: assignment discards 'const' qualifier from pointer target type [-Wdiscarded-qualifiers]
1997 | endp = strpbrk(p, ",@:");
| ^
Signed-off-by: Rudi Heitbaum <rudi@heitbaum.com>
Since glibc-2.43:
For ISO C23, the functions bsearch, memchr, strchr, strpbrk, strrchr,
strstr, wcschr, wcspbrk, wcsrchr, wcsstr and wmemchr that return pointers
into their input arrays now have definitions as macros that return a
pointer to a const-qualified type when the input argument is a pointer
to a const-qualified type.
fixes:
amdgpu/amdgpu_asic_id.c: In function 'find_asic_id_table':
amdgpu/amdgpu_asic_id.c:249:19: warning: assignment discards 'const' qualifier from pointer target type [-Wdiscarded-qualifiers]
249 | file_name = strrchr(AMDGPU_ASIC_ID_TABLE, '/');
| ^
Signed-off-by: Rudi Heitbaum <rudi@heitbaum.com>
This fixes a build error with Tegra.
Fixes: 8cc2d09c ("amdgpu: reserve half of the addr space to implement a workaround for PRT")
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
On GFX6-GFX12, except GFX9, SMEM loads on NULL PRT pages would just
fault and hang the GPU. The better workaround that we found is to split
the address space in two equal parts "LOW"/"HIGH", here's more details:
To workaround that, the driver splits the total VA space in half, so
that a single bit controls whether it's the "HIGH" or the "LOW"
address space. Every sparse residency buffer allocations that might
be used with SMEM get two allocations:
- the "HIGH" address space is mapped normally and its VA is returned
to the application.
- the "LOW" address space is explicitly mapped to a zero-initialized
buffer when it's allocated or when it's unmapped.
Other buffer allocations are always allocated in the "LOW" address.
The driver also creates a zero-allocated BO that will be used to map
partially resident buffers at creation and when explicitly unmapped.
The size is arbitrary but it seems that allocating 8MiB is perfectly
reasonable and fast enough in most cases. This BO is marked as
read-only for the GPU because the backend compilers don't use SMEM
stores.
For example this makes the "LOW" half of the HIGH VA range like
[0xffff800100000000,0xffffbfffffbfe000], and the "HIGH" half is left
for PRT.
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Check the return value of asprintf() in drmGetFormatModifierNameFromNvidia(),
drmGetFormatModifierNameFromAmlogic(), and drmGetFormatModifierNameFromVivante()
to avoid leaving pointers uninitialized on allocation failure.
asprintf() returns -1 on failure and does not guarantee the output pointer
is set to NULL, which could lead to undefined behavior if the pointer is
used subsequently. Set the pointer to NULL explicitly on failure.
Fixes: -Wunused-result warnings from GCC
Signed-off-by: Rudi Heitbaum <rudi@heitbaum.com>
The modifiers for DRM_FORMAT_MOD_VENDOR_APPLE are simple and cna be
represented in a static table. Add "APPLE" as pattern.
Signed-off-by: Janne Grunau <j@jannau.net>
Synchronize drm.h, drm_mode.h and drm_fourcc.h to drm-next.
Generated using make headers_install
Generated from drm-next branch commit 0692602defb0c273f80dec9c564ca50726404aca
Signed-off-by: Simon Ser <contact@emersion.fr>
The header shipped with libdrm was missing blob resources, context_init,
syncobj support, and capset definitions.
Update to the current version from Linux, which was last modified in 6.16.
Signed-off-by: Val Packett <val@invisiblethingslab.com>
The buffer size is calculated using pixels, not bytes as it should. The
result is often correct, though, as the stride is aligned to page size,
but there are still many cases where the size ends up being wrong.
Fix this by not calculating the size at all, as in that case
DRM_OMAP_GEM_INFO ioctl is used to get the correct size from the kernel.
This is better in any case as then the userspace library doesn't need to
know how the tiled buffers need to be aligned.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ti.com>
Signed-off-by: Andrew Davis <afd@ti.com>
After a change in the patch, the environment variable name for
the `amdgpu.ids` file paths changed to plurar form, but the
documentation wasn't updated.
This patch fixes it.
This patch allows to specify several colon-separated paths where
to search for the `amdgpu.ids` file in the AMDGPU_ASIC_ID_TABLE_PATH
environment variable.
In some cases, like when building a Snap application that uses
libdrm, the `amdgpu.ids` file isn't directly available at the
compiling place, but inside a mounted folder. This forces each
application to link/bind the file from the current place
(usually at the $SNAP/gnome-platform/usr/share/libdrm/amdgpu.ids)
which is cumbersome.
This patch allows to set an environment variable, called
AMDGPU_ASIC_ID_TABLE_PATH, where the file will be also searched
if it isn't located in the default, meson-configured, path.
Synchronize drm.h, drm_mode.h and drm_fourcc.h to drm-next.
Generated using make headers_install
Generated from drm-next branch commit 2a084f4ad727244768b919455aa9dc1c04630487
Signed-off-by: Simon Ser <contact@emersion.fr>
Some older Vega APUs don't provide a very useful string. If we have
a string in amdgpu.ids use that, but fallback to /proc/cpuinfo.
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
The string will generally have a CPU and GPU component, so if both
are found split it up. Make sure that it starts with AMD to be
consistent.
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Mario Limonciello (AMD) <superm1@kernel.org>
This commit adds essential parameter validation to several key
functions in amdgpu_bo.c to prevent null pointer dereferences
that were causing segmentation faults and improve overall code
robustness.
The changes address the following crash scenario:
Received signal SIGSEGV.
Stack trace:
#0 [fatal_sig_handler+0x17b]
#1 [__sigaction+0x50]
#2 [amdgpu_bo_alloc+0x37]
#3 [__igt_unique____real_main461+0x7d5]
#4 [main+0x2d]
#5 [__libc_init_first+0x90]
#6 [__libc_start_main+0x80]
#7 [_start+0x25]
Changes made:
1. amdgpu_bo_alloc():
- Validate alloc_buffer and buf_handle parameters
- Return -EINVAL if either is NULL
- Prevents null pointer dereference in memset and subsequent operations
2. amdgpu_bo_set_metadata():
- Validate info parameter
- Return -EINVAL if info is NULL
- Prevents accessing invalid metadata structure
3. amdgpu_bo_query_info():
- Validate info parameter in addition to existing bo->handle check
- Return -EINVAL if info is NULL
- Prevents writing to invalid info pointer
4. amdgpu_bo_list_create():
- Validate resources parameter
- Return -EINVAL if resources is NULL when number_of_resources > 0
- Prevents invalid memory access during resource array processing
These changes ensure proper error handling when callers pass invalid
null pointers, preventing potential segmentation faults and making
the API more robust against programming errors. The validation occurs
early in each function to minimize performance impact.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Jesse Zhang <Jesse.Zhang@amd.com>
The correct marketing name is encoded in the model name field
that is read from the hardware on an APU. Try to read from /proc/cpuinfo
when an APU is found to identify such hardware.
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Its contents is the same as drmFauxBusInfo. While it is technically an
API breaking change, libdrm with the relevant changes has been released
only recently and there are issues with quering DRM nodes that are on
the faux bus, which result in crashes, so it is unlikely that there are
users that depend on the new API.
Signed-off-by: Vlad Zahorodnii <vlad.zahorodnii@kde.org>
The name string is allocated on the heap, so the memcmp() function will
not return 0 even though two nodes share the same device name.
Also, the device info will not be populated when counting the available
drm devices. For example, using drmGetDevices2(0, NULL, 0). In that
case, drmGetDevices2() will eventually crash in drmDevicesEqual() while
merging nodes belonging to the same device due to passing null pointers
to memcmp().
This change adds faux device name to businfo. The businfo is populated
regardless whether device info is needed, which is needed to make the
drmGetDevices2(0, NULL, 0) case work. drmFauxBusInfo can also be used
with the memcmp() function to match the other cases in drmDevicesEqual().
Signed-off-by: Vlad Zahorodnii <vlad.zahorodnii@kde.org>
This prevents text editors that trim trailing whitespace automatically
from making unrelated changes.
Signed-off-by: Vlad Zahorodnii <vlad.zahorodnii@kde.org>
Linux 6.14 included a new type of bus, the "faux" bus [1].
The next version of Linux (v6.16) will move the VKMS driver to the faux
bus. See kernel commit 5686601908d8 ("drm/vkms: convert to use
faux_device") for more details.
Add support for the faux bus so drmGetDeviceFromDevId(), drmGetDevices()
and drmGetDevices2() return the devices on it.
[1] https://lore.kernel.org/all/2025021023-sandstorm-precise-9f5d@gregkh/
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@oss.qualcomm.com>
Signed-off-by: José Expósito <jose.exposito89@gmail.com>
The inline function is_power_of_two should be set as static since it's
only called in pattern.c, otherwise it will encounter below build error.
[114/136] Linking target tests/modetest/modetest
FAILED: tests/modetest/modetest
cc -o tests/modetest/modetest tests/modetest/modetest.p/buffers.c.o tests/modetest/modetest.p/cursor.c.o tests/modetest/modetest.p/modetest.c.o -Wl,--as-needed -Wl,--no-undefined -fno-omit-frame-pointer '-Wl,-rpath,$ORIGIN/../..' -Wl,-rpath-link,/home/ray/repo/drm/build/ -Wl,--start-group libdrm.so.2.125.0 tests/util/libutil.a -lm -pthread /usr/lib/x86_64-linux-gnu/libcairo.so -Wl,--end-group
/usr/bin/ld: tests/util/libutil.a.p/pattern.c.o: in function `check_yuv':
/home/ray/repo/drm/build/../tests/util/pattern.c:1989: undefined reference to `is_power_of_two'
/usr/bin/ld: /home/ray/repo/drm/build/../tests/util/pattern.c:1990: undefined reference to `is_power_of_two'
/usr/bin/ld: /home/ray/repo/drm/build/../tests/util/pattern.c:1991: undefined reference to `is_power_of_two'
collect2: error: ld returned 1 exit status
Signed-off-by: Huang Rui <ray.huang@amd.com>
Add seed argument for modetest noise patterns. If no seed is provided
the current time will be used as seed.
The used seed will be printed so the noise pattern can be reproduced.
Signed-off-by: Emil Svendsen <emas@bang-olufsen.dk>
Add three new test patterns:
- noise: random black and white pixels
- noise-color: random color pixels
- black-white: alternate between black and white pixels
These patterns are useful for measuring radiated emissions for EMC
compliance tests.
Signed-off-by: Emil Svendsen <emas@bang-olufsen.dk>
Synchronize drm.h, drm_mode.h and drm_fourcc.h to drm-next.
Generated using make headers_install.
Generated from drm-next branch commit 04c8970771b4f1f39bb8453a2eeb188c4d5edbd6
Signed-off-by: Simon Ser <contact@emersion.fr>
buildah tried to pull the arm64 variant of the arm32v7/debian:bookworm
container:
* choosing an image from manifest list docker://arm32v7/debian:bookworm: no image found in image index for architecture "arm64", variant "v8", OS "linux"
Signed-off-by: Simon Ser <contact@emersion.fr>
py39-docutils doesn't exist in the package repositories, likely
because the Python version has been upgraded. Let's not hardcode
the Python version in the package name.
Signed-off-by: Simon Ser <contact@emersion.fr>
Add queue id support to the user queue wait IOCTL
drm_amdgpu_userq_wait structure.
This is required to retrieve the wait user queue and maintain
the fence driver references in it so that the user queue in
the same context releases their reference to the fence drivers
at some point before queue destruction.
Otherwise, we would gather those references until we
don't have any more space left and crash.
Signed-off-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Suggested-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
This patch adds new IOCTL functions to support
userqueue create, remove, signal and wait etc.
v2:(Marek)
- Add csa support for SDMA queue.
- Addressed's review comments.
- Removed raw2/op2 ioctl.
- Added syncobj_timeline_handles in amdgpu_userq_wait IOCTL.
v3:(Yogesh)
- Rename timeline* objects as per UAPI review (Arvind).
v4: (Marek)
- Drop AMDGPU_USERQ_BO_WRITE as this should not be a global option
of the IOCTL, It should be option per buffer. Hence adding separate
array for read and write BO handles. (Arun)
- Modify num_fences to __u16, flags changed to __u16 and placed
the num_fences next to flags for optimal padding and size. (Arun)
v5:(Marek/Pierre-Eric)
- add more detail params description for signal and wait IOCTL calls.
- Remove the unused structure fields in signal and wait structs.
- Add separate array of read and write for BO handles. (Arun)
- Removes the unused flags parameter from the
amdgpu_create_userqueue IOCTL. (Arvind)
v6:(Pierre-Eric)
- Remove unused headers. (Arvind)
- Modify the function parameter names and struct
field names as per the review comments. (Arun)
v7:(Marek)
- Modify the structure field name and comments. (Arun)
- Rename vm_timeline_syncobj and add comment for
vm_timeline_point.
- Remove GDS buffer support from MQD. (Arvind)
v8:(Pierre-Eric)
- Modify the function parameter names.
- Added new function in amdgpu-symbols.txt (Arvind)
v9:(Marek)
- Use the drm signal/wait structure as the parameter. (Arun)
Cc: Deucher, Alexander <alexander.deucher@amd.com>
Cc: Koenig, Christian <christian.koenig@amd.com>
Cc: Sharma, Shashank <shashank.sharma@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Signed-off-by: Arvind Yadav <arvind.yadav@amd.com>
Signed-off-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
This patch adds UAPI interface changes for AMDGPU usermode
queues, semaphore and new AMDGPU GEM domain for doorbells.
Usermode queues allow a userspace process to create
and submit its graphics/compute/sdma work directly to the GPU.
v2:(Marek)
- Add csa support for SDMA queue.
- Rename UAPI objects and struct as per UAPI review. (Shashank)
v3:(Yogesh)
- Rename UAPI timeline* objects as per UAPI review. (Arvind)
v4: (Marek)
- Drop AMDGPU_USERQ_BO_WRITE as this should not be a global option
of the IOCTL, It should be option per buffer. Hence adding separate
array for read and write BO handles. (Arun)
- Modify num_fences to __u16, flags changed to __u16 and placed
the num_fences next to flags for optimal padding and size. (Arun)
- Fix 32-bit holes issue in sturct drm_amdgpu_gem_va as per
UAPI review (Arvind).
v5:(Marek/Pierre-Eric)
- add more detail params description for signal and wait IOCTL calls.
- Remove the unused structure fields in signal and wait structs.
- Add separate array of read and write for BO handles. (Arun)
- Removes the unused flags parameter from the usermode queue UAPI structure
- Clarify comments on top of drm_amdgpu_userq_in
- Clarify comment for queue_id (in)
- Clarify comment for mqd
- Clarify comment for compute MQD size
- Clarify comment for queue_id (out)
- Adds padding variables in userqueue in/out structures. (Shashank)
v6:(Pierre-Eric)
- Modify the function parameter names and struct
field names as per the review comments. (Arun)
v7:(Marek)
- Modify the structure field name and comments. (Arun)
- Rename vm_timeline_syncobj and add comment for
vm_timeline_point.
- Remove GDS buffer support from MQD. (Arvind)
v8:(Pierre-Eric)
- Modify the function parameter names.
Cc: Koenig, Christian <christian.koenig@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Signed-off-by: Shashank Sharma <shashank.sharma@amd.com>
Signed-off-by: Arvind Yadav <arvind.yadav@amd.com>
Signed-off-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
This patch:
- adds a new subquery (AMDGPU_INFO_UQ_FW_AREAS) in AMDGPU_INFO_IOCTL
to get the size and alignment of shadow and csa objects from the
kernel. This information is required for a userqueue consumer (like
MESA/libdrm) to create the userqueue metadata objects properly.
- also adds supporting metadata structures and a high level wrapper
function (amdgpu_query_uq_metadata_info) to the query, to make it
easy to use.
The corresponding kernel changes for this UAPI extension can be found
in amd-gfx mailing list, link:
https://patchwork.freedesktop.org/patch/621390/?series=139715&rev=2
This patch adds support only for the GFX IP, and the other engines may
be supported in subsequent development.
V2: fix the build error due to exporting of helper function
V3: make an entry for amdgpu_query_uq_metadata_info in
amdgpu-symbols.txt
V4: Rename the subquery to AMDGPU_INFO_UQ_FW_AREAS (Marek, Pierre-Eric)
V5: Addressed review comments (Pierre-Eric):
- Fix the API comment to match the new IOCTL name
- remove the unused uq_metadata parameter
Cc: Marek Olsak <marek.olsak@amd.com>
Cc: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian Koenig <christian.koenig@amd.com>
Cc: Arvind Yadav <arvind.yadav@amd.com>
Reviewed-by: Marek Olsak <marek.olsak@amd.com>
Signed-off-by: Shashank Sharma <shashank.sharma@amd.com>
Just as it is already handled in the caller, `drmGetVersion`.
I'm not sure what the offending driver is, but the Ubuntu incidents
seem to be coming from a dual Intel/Nvidia machine. And they show
it is `card1` so I'm guessing `nvidia-drm` is the offender.
Related: https://bugs.launchpad.net/bugs/2104352
Make modetest available to vendors on Android. libdrm_util and
libdrm_test_headers is also made available to vendors since these are
depenencies of modetest. This results in the module target
modetest.vendor being availble to vendor modules.
Signed-off-by: Rob Barnes <robbarnes at google.com>
Fixes the following building error:
external/libdrm/xf86drm.c:158:10: fatal error: 'generated_static_table_fourcc.h' file not found
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1 error generated.
Change-Id: I1b0cac498ed63ebec6e8c03629bbf4a1b6a9618d
Reviewed-by: Mauro Rossi <issor.oruam@gmail.com>
Currently modetest supports only the YUV420 and YVU420 planar YCbCr
plane formats (aka YV12 and YU12). Extend the code to add support for
YUV422 / YVU422 and YUV444 / YVU444 plane formats.
Reviewed-by: Marijn Suijten <marijn.suijten@somainline.org>
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
In preparation to adding more planar YUV formats, introduce X, Y
subsampling ratios and use them to calculate plane offsets and buffer
size.
Reviewed-by: Marijn Suijten <marijn.suijten@somainline.org>
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Whenever `util_open()` is called to open a device for the first matching
module, it will skip devices for the `nvidia_drm` kernel module which
is not in the list. We could add this module for now, but keeping this
list of DRM modules up to date is cumbersome.
At the same time walking a list of modules and calling `drmOpen()` for
each of them is incredibly expensive (when the user doesn't explicitly
specify one with `-M`), as each each call opens every DRM node just
to see if they are associated to the requested module. And for no
good reason: all we want is the first `DRM_NODE_PRIMARY` (which is what
`drmOpen()` also returns) to use by default.
For example on the `"msm"` driver, which used to be the 9th in the
modules list, all nodes are opened for the 9th time before e.g.
`modetest` returns a useful result, which takes ages unless the user
painstakingly provides the module for the currently known device on
the cmdline.
This is very simply solved by calling `drmGetDevices(2)()`, which
iterates through all DRM nodes only once and allows us to immediately
find + `open()` the first device that has a PRIMARY node. A random
search for the error shows that this was also attempted in (a fork of?)
kmscube:
https://git.ti.com/cgit/glsdk/kmscube/commit/?id=456cabc661caac5c60729751d45efd668faa8e97
Finally we add a `drmIsKMS()` check to make sure we only include primary
nodes that actually support rendering, and also print the values from
`drmGetVersion()` on success to make it easier to identify the device.
In the future we could extrapolate this feature by letting query
commands like `modetest -c` list connectors for every device/module, not
just the first PRIMARY node that we found.
Bionic libc ships with `ioctl` that has two signatures, one with an
unsigned `request` parameter and one with a signed request parameter.
This leads to compilation failing due to `__typeof__(ioctl)` being used
by DRM which fails to resolve which overload to use, this has been fixed
by defining `BIONIC_IOCTL_NO_SIGNEDNESS_OVERLOAD` on Android.
Signed-off-by: Mark Collins <mark@igalia.com>
There is an excellent writeup explaining this requirement here:
https://gitlab.freedesktop.org/wayland/wayland/-/issues/175
In short, for mixed environments such as the Steam Runtime and other
container-like environments, choosing which libdrm to link into the
client's address space is a hard problem. If the runtime has a newer
libdrm than the host, then it should be preferred, because the client
may be using newly-added symbols. But if the host has a newer libdrm,
then that should be used, because drivers may be depending on those.
Bumping the DSO minor version is transparent to all users because apps
only link against the major version, e.g. DT_NEEDED libdrm.so.2; the
fact that libdrm.so.2 is a link to libdrm.so.2.122.0 is a detail known
only to the loader, but it does let a smart runtime make better
decisions.
Signed-off-by: Daniel Stone <daniels@collabora.com>
When the GPU feature is turned on in crosvm, these modules are added as
dependencies. Since crosvm is included in the virt APEX, add the APEX to
the apex_available properties of the modules to make them available in
the APEX.
To do so, crosvm and its dependencies have the apex_available property
set to "//apex_available:platform", "com.android.virt" to explicitly
acknowledge the joining.
drm_property_type_is function in xf86drmMode.h file can cause compiler
error because it performs unsigned to signed conversion. Some Android.mk
modules have been avoiding this by adding "-isystem external/libdrm"
flag, because warnings from system headers are suppressed.
This changes exported_include_dirs to export_system_include_dirs
to workaround the potential error with the same manner above.
... to make drm format header visible for host wayland server.
Adds -Wno-implicit-function-declaration for vasprintf() on
xf86drm.c:2965.
Adds -Wno-int-conversion for drm_mmap() on libdrm_macros.h:60.
Also, drive-by alphabetize cflags.
See build/soong/README.md for more information about Soong.
Removes BOARD_GPU_DRIVERS, which wasn't affecting anything, since none
of the HAVE_* macros are defined. Even if they were, we'd prefer to
compile all of them so that a single library can support multiple
boards.
Some platforms (eg. SunOS) explicitly need extra symbols in order to define
those functions. There're many files needing the __EXTENSIONS__ symbol,
so doing this on a global scale.
Signed-off-by: Enrico Weigelt, metux IT consult <info@metux.net>
Synchronize drm_fourcc.h and drm_mode.h to drm-next.
Generated using make headers_install.
Generated from drm-next branch commit 541b1b0a8fc235bca355921eb7f3f59a8efa3e9a
Signed-off-by: Simon Ser <contact@emersion.fr>
The <linux/fb.h> header only exists on Linux, but isn't used anymore,
so drop it entirely.
Signed-off-by: Enrico Weigelt, metux IT consult <info@metux.net>
Several drivers still including <linux/stddef.h>, but not using anything
from it, thus breaking build on non-Linux platforms (eg. FreeBSD).
Since not needed at all, just stop including it.
Signed-off-by: Enrico Weigelt, metux IT consult <info@metux.net>
../xf86drm.c:4622:9: error: use of undeclared identifier 'written'; did you mean 'write'?
if (written + 1 > max_node_length)
^~~~~~~
write
../xf86drm.c:4624:21: error: use of undeclared identifier 'sbuf'; did you mean 'sbrk'?
if (stat(node, &sbuf))
^~~~
Signed-off-by: Enrico Weigelt, metux IT consult <info@metux.net>
fix commit cc3c80c6("tests/amdgpu: refactor dispatch/draw test")
../../SOURCES/drm/tests/amdgpu/shader_code.h:113:2: error: initializer
element is not constant
ps_##_ps##_shader_patchinfo_code_size_gfx##_n, \
^
Signed-off-by: Flora Cui <flora.cui@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
../../drm/tests/amdgpu/shader_test_util.c: In function
‘amdgpu_dispatch_init’:
../../drm/tests/amdgpu/shader_test_util.c:296:9: warning: enumeration
value ‘AMDGPU_TEST_GFX_MAX’ not handled in switch [-Wswitch]
296 | switch (test_priv->info->version) {
| ^~~~~~
Signed-off-by: Flora Cui <flora.cui@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
The issue was found by a static analysis tool:
Error: LOCK_EVASION (CWE-543):
libdrm-2.4.115/amdgpu/amdgpu_cs.c:596: thread1_checks_field:
Thread1 uses the value read from field "context" in the
condition "sem->signal_fence.context". It sees that the
condition is false. Control is switched to Thread2.
libdrm-2.4.115/amdgpu/amdgpu_cs.c:596: thread2_checks_field:
Thread2 uses the value read from field "context" in the
condition "sem->signal_fence.context". It sees that the
condition is false.
libdrm-2.4.115/amdgpu/amdgpu_cs.c:598: thread2_acquires_lock:
Thread2 acquires lock "amdgpu_context.sequence_mutex".
libdrm-2.4.115/amdgpu/amdgpu_cs.c:599: thread2_modifies_field:
Thread2 sets "context" to a new value. Note that this write can
be reordered at runtime to occur before instructions that do
not access this field within this locked region. After Thread2
leaves the critical section, control is switched back to
Thread1.
libdrm-2.4.115/amdgpu/amdgpu_cs.c:598: thread1_acquires_lock:
Thread1 acquires lock "amdgpu_context.sequence_mutex".
libdrm-2.4.115/amdgpu/amdgpu_cs.c:599: thread1_overwrites_value_in_field:
Thread1 sets "context" to a new value. Now the two threads have
an inconsistent view of "context" and updates to fields of
"context" or fields correlated with "context" may be lost.
libdrm-2.4.115/amdgpu/amdgpu_cs.c:596: use_same_locks_for_read_and_modify:
Guard the modification of "context" and the read used to decide
whether to modify "context" with the same set of locks.
# 597| return -EINVAL;
# 598| pthread_mutex_lock(&ctx->sequence_mutex);
# 599|-> sem->signal_fence.context = ctx;
# 600| sem->signal_fence.ip_type = ip_type;
# 601| sem->signal_fence.ip_instance = ip_instance;
Check `sem->signal_fence.context` in the locked region to avoid a race
condition.
Reviewed-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Signed-off-by: José Expósito <jexposit@redhat.com>
I always need to double-check what the return value means when
using that function (since it's not a bool).
Signed-off-by: Simon Ser <contact@emersion.fr>
str.format used to allow any type as an argument, which often resulted
in using an internal string representation. This is considered broken
behavior, and is deprecated since Meson 1.3.0.
Signed-off-by: Joaquim Monteiro <joaquim.monteiro@protonmail.com>
The devices weren't removed from dev_list.
Instead of just fixing the issue by adding:
if (*node) *node = dev->next;
after the while loop, use this opportunity to use a clearer
control flow.
Fixes: 7275ef8e ("amdgpu: add amdgpu_device_initialize2")
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Tested-by: Mike Lothian <mike@fireburn.co.uk>
Allows to opt-out from the device deduplication logic. This is not the
recommended way of using dev handles, but it's necessary for native context:
in this situation one process (eg: Qemu) will init many devices and we
want independent devices to make sure guest applications are isolated from
each other.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
This is the same functionnality that amdgpu_va_range_alloc offers,
except it's now usable without a device handle.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
This will allow applications to use this feature without a device.
The first use case will be native context: we want VA address to
be managed by the guest (to avoid a round-trip to the host to only
generate a VA) but the amdgpu_device only exist on the host.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Until now VA management was tied to a device handle, but there's no
reason for this.
As a first step to export VA management outside of amdgpu_device,
this commit adds a new structure type holding the 4 va_mgr.
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Device will be removed from dev_list when refcount reaches 0, so the
dev_mutex must be locked before decreasing reference otherwise there's
a race where this device is still in dev_list with refcount 0 which will
assert or crash in amdgpu_device_initialize trying to use this device
instead of creating new one.
Fixes issue reported in https://gitlab.freedesktop.org/drm/amd/-/issues/2156#note_2268110
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Add an entry for the "tidss" driver, so that the test utilities work
with this driver without passing the -M argument.
Signed-off-by: Francesco Valla <valla.francesco@gmail.com>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
This fixes the logic that decides if a node name is valid to use the
same length restrictions that are used in drmDeviceAlloc, which expects
node names to conform to a specific naming scheme (On OSes except
OpenBSD this means `/dev/dri/renderD123`). This addresses the problem of
node names that are longer than expected, while still allowing symlinks
to work.
I've also applied the same fix to the OpenBSD path, while bringing the
check that `snprintf` didn't error from OpenBSD to the main path.
Signed-off-by: Dylan Baker <dylan.c.baker@intel.com>
Tested-by: Mark Janes <markjanes@swizzler.org>
Tested-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
If the user has some UDev rules in place that creates symlinks for
one of the card or render nodes, and the name of the symlink is
too long, then drmDeviceAlloc() ends up truncating the name of
the node.
This in turn results in chaos in different subsystems. E.g.
vulkaninfo dies early with this:
Code 0 : failed to stat DRM primary node /dev/dri/my-favorite- (VK_ERROR_INITIALIZATION_FAILED)
(if the symlink is called /dev/dri/my-favorite-card-node)
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Synchronize drm.h and drm_mode.h to drm-next.
Generated using make headers_install.
Generated from drm-next branch commit a60501d7c2d3e70b3545b9b96576628e369d8e85
Signed-off-by: Simon Ser <contact@emersion.fr>
Acked-by: Simon Zeni <simon.zeni@collabora.com>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Synchronize drm.h, drm_mode.h and drm_fourcc.h to drm-next.
Generated using make headers_install.
Generated from drm-next branch commit c79b972eb88b077d2765e7790d0902b3dc94d55c
Signed-off-by: Simon Ser <contact@emersion.fr>
This is useful to figure out whether the dev_t refers to a primary
node or a render node. Indeed, drmGetDeviceFromDevId returns a
drmDevice, which holds both the primary and render nodes.
Signed-off-by: Simon Ser <contact@emersion.fr>
Add support for creating buffers using big-endian formats.
For now this is limited to XRGB1555 and RGB565, which are the most
common big-endian formats.
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
---
v5:
- Add Reviewed-by,
v4:
- No changes,
v3:
- No changes,
v2:
- New.
Add support for rendering the crosshairs in a buffer using the
big-endian RGB565 format.
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
---
v5:
- Add Reviewed-by,
v4:
- No changes,
v3:
- No changes,
v2:
- New.
Cairo always uses native byte order for rendering.
Hence if the byte order of the frame buffer differs from the byte order
of the CPU, the frame buffer contents need to be byteswapped twice: once
before rendering, to convert to native byte order, and a second time
after rendering, to restore the frame buffer format's byte order.
Note that byte swapping is not done for ARGB32 formats, as for these
formats, byte order only affects the order of the red, green, and blue
channels, which we do not care about here.
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
---
v5:
- Add Reviewed-by,
v4:
- No changes,
v3:
- Wrap byteswap_buffer{16,32}() implementation inside #if HAVE_CAIRO
to avoid defined-but-not-used compiler warnings,
v2:
- RGB30 is untested.
Add support for drawing the SMPTE and tiles test patterns in buffers
using big-endian formats.
For now this is limited to XRGB1555 and RGB565, which are the most
common big-endian formats.
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
---
v5:
- Add Reviewed-by,
v4:
- No changes,
v3:
- Increase indentation after definition of cpu_to_be16(),
v2:
- New.
When specifying a frame buffer format like "RG16_BE" (big-endian RG16),
modetest still uses the little-endian variant, as the format string is
truncated to four characters.
Fix this by increasing the format string size to 8 bytes (7 characters +
NUL terminator).
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
---
v5:
- Add Reviewed-by,
v4:
- No changes,
v3:
- Update for suffix change from "be" to "_BE", cfr. commit
ffb9375a50 ("xf86drm: handle DRM_FORMAT_BIG_ENDIAN in
drmGetFormatName()"),
- Replace hardcoded numbers in code by sizeof(),
v2:
- New.
DRM formats are defined to be little-endian, unless the
DRM_FORMAT_BIG_ENDIAN flag is set. Hence writes of multi-byte pixel
values need to take endianness into account.
Introduce a swap16() helper to byteswap 16-bit values, and a
cpu_to_le16() helper to convert 16-bit values from CPU-endian to
little-endian, and use the latter in the various pattern fill functions
for 16-bit formats.
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
---
v5:
- Add Reviewed-by,
v4:
- No changes,
v3:
- Increase indentation after definition of cpu_to_le16(),
v2:
- New.
DRM formats are defined to be little-endian, unless the
DRM_FORMAT_BIG_ENDIAN flag is set. Hence writes of multi-byte pixel
values need to take endianness into account.
Introduce a swap32() helper to byteswap 32-bit values, and a
cpu_to_le32() helper to convert 32-bit values from CPU-endian to
little-endian, and use the latter in the various pattern fill functions
for 32-bit formats.
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
---
v5:
- Add Reviewed-by,
v4:
- Use new HAVE_BIG_ENDIAN symbol,
v3:
- Increase indentation after definition of cpu_to_le32(),
v2:
- Add Acked-by,
- Add swap32() intermediate helper,
- Add __ARM_BIG_ENDIAN and __s390__.
The endianness of the target is currently determined based on
preprocessor symbols. Unfortunately some symbols checked are wrong
(sparc64-linux-gnu-gcc does not define __BIG_ENDIAN__ or SPARC), and
several checks for big-endian architectures are missing.
Fix this by introducing a new preprocessor symbol HAVE_BIG_ENDIAN, which
is set based on meson's knowledge of the target endianness.
Android.common.mk does not need an update, as Android is always
little-endian (https://developer.android.com/ndk/guides/abis.html).
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
---
v5:
- Add Reviewed-by,
v4:
- Replace explicit #ifdef checks by a define set by meson,
v3:
- No changes,
v2:
- Add arm, aarch64, microblaze, s390, and sh.
Add smpte and tiles pattern for 10-bit NV15, NV20 and NV30 pixel formats
based on the existing pattern for NV12 with colors simply scaled from
8-bit to 10-bit.
These pixel formats are typically used by video decoder and display
pipeline on Rockchip SoCs, e.g. on RK322X, RK3288, RK3328 and RK3399
the video decoder produce 10-bit video frames in NV15 and NV20 format.
NV20 and NV30 pixel formats was added in drm-misc commit 728c15b4b5f3
("drm/fourcc: Add NV20 and NV30 YUV formats").
This can be tested/validated on Rockchip SoCs with drm-misc commit
d4b384228562 ("drm/rockchip: vop: Add NV15, NV20 and NV30 support").
Signed-off-by: Jonas Karlman <jonas@kwiboo.se>
Reviewed-by: Christopher Obbard <chris.obbard@collabora.com>
Tested-by: Christopher Obbard <chris.obbard@collabora.com>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Replace system() with cpu_family() for libdrm_intel
This restore libdrm_intel to be built by default
Closes: #93
Signed-off-by: David Jagu <marav8@free.fr>
Add support for drawing the SMPTE pattern in buffers using a
color-indexed frame buffer formats with two, four, or sixteen colors.
Note that this still uses 256 as the CLUT size, as
DRM_IOCTL_MODE_SETGAMMA enforces that the size matches against the
(fixed) gamma size, while the CLUT size depends on the format.
Move clearing the color LUT entries from util_smpte_index_gamma() to its
caller, as only the caller knows how many entries there really are
(currently DRM always assumes 256 entries).
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
---
v5:
- Add Reviewed-by,
v4:
- Add missing C[12] to oneline-summary,
- Do not remove memset() of full lut, else some entries may stay
uninitialized,
v3:
- Add Acked-by,
v2:
- Split off changes to tests/modetest/modetest.c,
- Add C1 and C2 support.
The linuxdoc comments say userspace can query the gamma size:
* drm_mode_gamma_set_ioctl - set the gamma table
*
* Set the gamma table of a CRTC to the one passed in by the user. Userspace can
* inquire the required gamma table size through drm_mode_gamma_get_ioctl.
* drm_mode_gamma_get_ioctl - get the gamma table
*
* Copy the current gamma table into the storage provided. This also provides
* the gamma table size the driver expects, which can be used to size the
* allocated storage.
but the code doesn't seem to support that in an easy way (like setting
red/green/blue to NULL on input, retrieving gamma_size on output), only
by providing big enough buffers for red/green/blue, and looping over
gamma_size until -EINVAL is no longer returned.
Add support for creating buffers using the new color-indexed frame
buffer formats with two, four, and sixteen colors.
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
---
v5:
- Add Reviewed-by,
v4:
- No changes,
v3:
- Add Acked-by,
v2:
- Split off changes to tests/modetest/buffers.c.
Add support for drawing the SMPTE pattern in a buffer using the C2
indexed format.
As only four colors are available, resolution is halved, and the pattern
is drawn in a PenTile RG-GB matrix, using Floyd-Steinberg dithering.
The magnitude of the green subpixels is reduced, as there are twice as
many green subpixels as red or blue subpixels.
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
---
Dithering example at https://drive.google.com/file/d/1g5O8XeacrjrC8rgaVENvR65YeI6QvmtO/view
v5:
- Add Reviewed-by,
v4:
- Replace FILL_COLOR() use by pentile_color_lut[],
v3:
- Add Acked-by,
v2:
- New.
Add support for drawing the SMPTE pattern in a buffer using the C4
indexed format.
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
---
v5:
- Add Reviewed-by,
v4:
- No changes,
v3:
- Add Acked-by,
v2:
- Use new smpte_top[],
- Split off changes to tests/util/pattern.c.
Store the number of available colors for color-indexed frame
buffer formats in the format_info[] array. This avoids the need of test
code for having to use switch statements all the time to obtain the
number of colors, or to check if a mode is color-indexed or not.
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
---
v5:
- Add Reviewed-by,
v4:
- No changes,
v3:
- Add Acked-by,
v2:
- New.
Add support for creating buffers using the new color-indexed frame
buffer formats with two, four, and sixteen colors.
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
---
v5:
- Add Reviewed-by,
v4:
- No changes,
v3:
- Add Acked-by,
v2:
- Split off changes to tests/util/format.c.
The color LUT for the SMPTE pattern in indexed mode contains 22 entries,
although only 13 are non-unique.
Reduce the size of the color LUT by dropping duplicate entries, so it
can be reused for formats supporting e.g. 16 colors. Rename the
function util_smpte_c8_gamma() to util_smpte_fill_lut(), and its first
parameter size to ncolors, to match their actual use.
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
---
v5:
- Add Reviewed-by,
v4:
- Rename util_smpte_index_gamma() to util_smpte_fill_lut(), and its
first parameter from size to ncolors,
- Move smpte_color_lut[] down,
- Kill FILL_COLOR() macro,
- Add and use EXPAND_COLOR() macro,
v3:
- Add Acked-by,
v2:
- Factor out smpte color LUT.
Fill in the LSB when converting color components from 8-bit to 16-bit.
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
---
v5:
- Add Reviewed-by,
v4:
- No changes,
v3:
- Add Acked-by,
v2:
- New.
To reduce the size and complexity of checks. require() allows combining
auto and enabled checks(), so that something like
```meson
x = get_option('feature')
y = false
if x.enabled()
if not condition
error(...)
endif
y = condition
endif
```
can be rewritten as:
```meson
y = get_option('feature').require(condition, error_message : ...).allowed()
```
require checks the condition, then if the feature is required it emits
an error with the given message otherwise it returns a disabled feature.
allowed then returns whether the feature is not disabled, and returns
that (ie, .allowed() == not .disabled()). This is especially helpful for
longer more complex conditions
Signed-off-by: Dylan Baker <dylan.c.baker@intel.com>
Intel requires libpciaccess and an x86/x86_64 host, so if those
aren't found and it's enabled we need to error
Signed-off-by: Dylan Baker <dylan.c.baker@intel.com>
Reviewed-by: Simon Ser <contact@emersion.fr>
Print modifiers in hex in addtion to in strings returned by
drmGetFormatModifierName. In some cases, hex numbers can be more easily
compared visually.
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Currently the atomic request is only assigned after `set_property()` is
called, leaving `dev.req` in its uninitialized state causing
`drmModeAtomicAddProperty()` to return an error code, which is printed
as `"Success"` because `errno` is not set by `libdrm` (but it would have
been when non-atomic `drmModeObjectSetProperty()` called an IOCTL
immediately):
sony-akatsuki-row ~ $ modetest -M msm -a -w 81:ACTIVE:0
failed to set CRTC 81 property ACTIVE to 0: Success
Solve this by assigning a new atomic request object before calling
`set_property()`, when there are properties to set. Likewise, commit
these properties after `set_property()` even if there is no other
operation (setting modes or planes) specified.
Furthermore `drmModeObjectSetProperty()` is implemented in terms of
`DRM_IOCTL()` which already returns `-errno` when `ioctl()` returns
`-1`, so we should instead pass `ret` to `strerror()` and get an
accurate error string out of `drmModeAtomicAddProperty()` too.
Fixes: 93220283 ("tests/modetest: Add atomic support")
Signed-off-by: Marijn Suijten <marijn.suijten@somainline.org>
Let's permit testing vsync with the default mode, this returns
back the pipe content and count when calling set_mode() so the
vsync test can also be used.
Signed-off-by: Neil Armstrong <neil.armstrong@linaro.org>
As found and discussed in [MR 58] a blob is not created in the else arm
because adding the GAMMA_LUT property with a NULL/0 blob_id causes it
to be reset to a default linear / pass-thru gamma table. The values
in the gamma_lut table might still be consumed in the legacy API path
below though, so it has to be initialized to a linear table.
[MR 58]: https://gitlab.freedesktop.org/mesa/drm/-/merge_requests/58#note_466972
Signed-off-by: Marijn Suijten <marijn.suijten@somainline.org>
It is useful to be able to specify mode parameters manually. Add support
for setting user-supplied modes. This patch is based on the original
idea by Rohit and Jessica, but implemented from scratch.
Suggested-by: Rohith Iyer <quic_rohiiyer@quicinc.com>
Suggested-by: Jessica Zhang <quic_jesszhan@quicinc.com>
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
These are unused since commit edcef53685 ("modetest: Add test
pattern support for missing RGB formats").
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
---
v2:
- Add Reviewed-by.
mode_vrefresh() does not take into account interlaced, doublescan, and
multiscan modes, leading to incorrect refresh rates.
Fix this, based on drm_mode_vrefresh() in Linux.
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
The YUV SMPTE patterns use RGB 191/192/192 instead of 192/192/192 for
the grey color in the top color bar.
Change it to 192/192/192, to match the RGB SMPTE patterns.
Fixes: a94ee62429 ("modetest: Add SMPTE test pattern")
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
---
v2:
- Add Reviewed-by.
Add support for drawing the SMPTE and tiles patterns in buffers using
semi-planar YUV formats with non-subsampled chroma planes.
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
---
v2:
- Add Reviewed-by.
Add writeback support to modetest with the below options:
- Passing in -a -c will now also show the writeback connector
- Dump the writeback output buffer to bitstream
Usage: "./modetest -M msm -s <connector_id>:<widthxheight>
-a -o <filepath>
-P <plane_id>@<crtc_id>:<widthxheight>+0+0@RG24"
This currently supports single writeback connector.
Co-developed-by: Rohith Iyer <quic_rohiiyer@quicinc.com>
Signed-off-by: Jessica Zhang <quic_jesszhan@quicinc.com>
[DB: dropped custom mode support, fixed segfault]
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
On 32-bit:
../tests/amdgpu/amdgpu_stress.c: In function ‘alloc_bo’:
../tests/amdgpu/amdgpu_stress.c:178:49: warning: format ‘%lx’ expects argument of type ‘long unsigned int’, but argument 4 has type ‘uint64_t’ {aka ‘long long unsigned int’} [-Wformat=]
fprintf(stdout, "Allocated BO number %u at 0x%lx, domain 0x%x, size %lu\n",
~~^
%llx
num_buffers++, addr, domain, size);
~~~~
../tests/amdgpu/amdgpu_stress.c:178:72: warning: format ‘%lu’ expects argument of type ‘long unsigned int’, but argument 6 has type ‘uint64_t’ {aka ‘long long unsigned int’} [-Wformat=]
fprintf(stdout, "Allocated BO number %u at 0x%lx, domain 0x%x, size %lu\n",
~~^
%llu
num_buffers++, addr, domain, size);
~~~~
../tests/amdgpu/amdgpu_stress.c: In function ‘submit_ib’:
../tests/amdgpu/amdgpu_stress.c:276:54: warning: format ‘%lx’ expects argument of type ‘long unsigned int’, but argument 5 has type ‘uint64_t’ {aka ‘long long unsigned int’} [-Wformat=]
fprintf(stdout, "Submitted %u IBs to copy from %u(%lx) to %u(%lx) %lu bytes took %lu usec\n",
~~^
%llx
count, from, virtual[from], to, virtual[to], copied, delta / 1000);
~~~~~~~~~~~~~
../tests/amdgpu/amdgpu_stress.c:276:65: warning: format ‘%lx’ expects argument of type ‘long unsigned int’, but argument 7 has type ‘uint64_t’ {aka ‘long long unsigned int’} [-Wformat=]
fprintf(stdout, "Submitted %u IBs to copy from %u(%lx) to %u(%lx) %lu bytes took %lu usec\n",
~~^
%llx
count, from, virtual[from], to, virtual[to], copied, delta / 1000);
~~~~~~~~~~~
../tests/amdgpu/amdgpu_stress.c:276:70: warning: format ‘%lu’ expects argument of type ‘long unsigned int’, but argument 8 has type ‘uint64_t’ {aka ‘long long unsigned int’} [-Wformat=]
fprintf(stdout, "Submitted %u IBs to copy from %u(%lx) to %u(%lx) %lu bytes took %lu usec\n",
~~^
%llu
count, from, virtual[from], to, virtual[to], copied, delta / 1000);
~~~~~~
../tests/amdgpu/amdgpu_stress.c:276:85: warning: format ‘%lu’ expects argument of type ‘long unsigned int’, but argument 9 has type ‘uint64_t’ {aka ‘long long unsigned int’} [-Wformat=]
fprintf(stdout, "Submitted %u IBs to copy from %u(%lx) to %u(%lx) %lu bytes took %lu usec\n",
~~^
%llu
count, from, virtual[from], to, virtual[to], copied, delta / 1000);
~~~~~~~~~~~~
../tests/amdgpu/amdgpu_stress.c: In function ‘parse_size’:
../tests/amdgpu/amdgpu_stress.c:296:24: warning: format ‘%li’ expects argument of type ‘long int *’, but argument 3 has type ‘uint64_t *’ {aka ‘long long unsigned int *’} [-Wformat=]
if (sscanf(optarg, "%li%1[kmgKMG]", &size, ext) < 1) {
~~^ ~~~~~
%lli
../tests/amdgpu/amdgpu_stress.c: In function ‘main’:
../tests/amdgpu/amdgpu_stress.c:378:45: warning: format ‘%lu’ expects argument of type ‘long unsigned int’, but argument 3 has type ‘uint64_t’ {aka ‘long long unsigned int’} [-Wformat=]
fprintf(stderr, "Buffer size to small %lu\n", size);
~~^ ~~~~
%llu
Fix this by using the proper "PRI?64" format specifiers.
Fixes: d77ccdf3ba ("amdgpu: add amdgpu_stress utility v2")
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
---
On Linux/amd64, the format strings in the resulting binary are
unchanged.
v3:
- Add Reviewed-by,
v2:
- Use PRI?64 to unbreak 64-bit build.
On 32-bit:
../amdgpu/amdgpu_bo.c: In function ‘amdgpu_find_bo_by_cpu_mapping’:
../amdgpu/amdgpu_bo.c:554:13: warning: cast to pointer from integer of different size [-Wint-to-pointer-cast]
cpu < (void*)((uintptr_t)bo->cpu_ptr + bo->alloc_size))
^
Indeed, as amdgpu_bo_info.alloc_size is "uint64_t", the sum is
always 64-bit, while "void *" can be 32-bit or 64-bit.
Fix this by casting bo->alloc_size to "size_t", which is either
32-bit or 64-bit, just like "void *".
Fixes: c6493f360e ("amdgpu: Eliminate void* arithmetic in amdgpu_find_bo_by_cpu_mapping")
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
---
v2:
- Add Reviewed-by.
This is useful for when GEM handles are exported and may be shared
between multiple buffer objects without going through other libdrm
interfaces.
Signed-off-by: Dor Askayo <dor.askayo@gmail.com>
Reviewed-by: Karol Herbst <git@karolherbst.de>
Currently drm node's minor range is used to identify node's type.
Since kernel drm uses node type name and minor to generate drm
device name, It will be more general to use drm device name to
identify drm node type.
Signed-off-by: James Zhu <James.Zhu@amd.com>
Reviewed-by: Simon Ser <contact@emersion.fr>
Current DRM_NODE_NAME_MAX only can support up to 999 nodes,
Update to support up to 2^MINORBITS nodes.
Signed-off-by: James Zhu <James.Zhu@amd.com>
Reviewed-by: Simon Ser <contact@emersion.fr>
Synchronize drm.h, drm_mode.h and drm_fourcc.h to drm-next.
Generated using make headers_install.
Generated from drm-next branch commit 52920704df878050123dfeb469aa6ab8022547c1
Signed-off-by: Simon Ser <contact@emersion.fr>
Drop support for control nodes. The kernel never returns such
nodes. Stop trying to detect and handle them, and always return
an error when a caller tries to open them.
The header is left untouched to avoid breaking libdrm's API.
Signed-off-by: Simon Ser <contact@emersion.fr>
issue:
in vcn1-vcn3, session buffer was not truly added, it shows
decoding creation commands cannot be sent multiple times.
problem:
session buffer has to be the first buffer sending out.
Otherwise, system could assume session buffer doesn't
exist.
solution:
move session buffer sending sequence to be the first one.
Reviewed-by: Veerabadhran Gopalakrishnan <Veerabadhran.Gopalakrishnan@amd.com>
Signed-off-by: Ruijing Dong <ruijing.dong@amd.com>
Signed-off-by: Saleemkhan Jamadar <saleemkhan.jamadar@amd.com>
update decoder unit test with session context buffer for VCN v1 to v4
v2: remove multiple checks for vcn4 (Ruijing Dong)
Signed-off-by: Saleemkhan Jamadar <saleemkhan.jamadar@amd.com>
Reviewed-by: Ruijing Dong <ruijing.dong@amd.com>
1. fixed an issue that drm test vcn3/4 encoding test
could cause VCN engine stuck.
2. adding missing or errous encoding ib package members.
Reviewed-by: Boyuan Zhang <Boyuan.Zhang@amd.com>
Signed-off-by: Ruijing Dong <ruijing.dong@amd.com>
Add the command line argument -e s[.t] to exclude (disable) suite s, or to
exclude suite s test t.
This is useful for instance to run the Basic Suite, but disable the GPU reset
test, on the command line, like this:
amdgpu_tests -s 1 -e 1.13
This option can be specified more than once on the command line, in order to
exclude more than one suite and/or suite and test combination from being run.
Cc: Alex Deucher <Alexander.Deucher@amd.com>
Signed-off-by: Luben Tuikov <luben.tuikov@amd.com>
Add -s and -t to the help output, as well as sort
the options output alphabetically.
v1: Fix a spelling in the subject of this commit.
Cc: Alex Deucher <Alexander.Deucher@amd.com>
Signed-off-by: Luben Tuikov <luben.tuikov@amd.com>
Fix the Usage: string on -h (help) in amdgpu_tests.c,
so brackets match, and remove mismatched angle brackets.
Cc: Alex Deucher <Alexander.Deucher@amd.com>
Signed-off-by: Luben Tuikov <luben.tuikov@amd.com>
For apu, it is integrated with cpu.
So hotplug test should be unnecessary for it.
Signed-off-by: Jesse Zhang <Jesse.Zhang@amd.com>
Reviewed-by: Flora Cui <flora.cui@amd.com>
pthread-stubs >= 0.4 simply passes -pthread which is similar to what
dependency('threads') returns. And make it a private dependency
for subprojects even on Linux.
Reviewed-by: Emmanuel Vadot <manu@FreeBSD.org>
Add an entry for the "vkms" driver, so that the test utilities work with
the vkms driver without passing the -M argument.
Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
Signed-off-by: Yi Xie <yixie@google.com>
See kernel commit 543036a2de71 ("drm/amd: Add GFX11 modifiers support
to AMDGPU (v3)").
Signed-off-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Fixes the following warning:
../xf86drm.c: In function ‘drmGetFormatModifierNameFromVivante’:
../xf86drm.c:614:14: warning: passing argument 1 of ‘asprintf’ from incompatible pointer type [-Wincompatible-pointer-types]
614 | asprintf(&mod_vivante, "%s%s%s", color_tiling, tile_status, compression);
| ^~~~~~~~~~~~
| |
| const char **
In file included from ../xf86drm.c:34:
/usr/include/stdio.h:396:40: note: expected ‘char ** restrict’ but argument is of type ‘const char **’
396 | extern int asprintf (char **__restrict __ptr,
| ~~~~~~~~~~~~~~~~~~^~~~~
../xf86drm.c:615:12: warning: return discards ‘const’ qualifier from pointer target type [-Wdiscarded-qualifiers]
615 | return mod_vivante;
| ^~~~~~~~~~~
Signed-off-by: Simon Ser <contact@emersion.fr>
Update drm_fourcc.h to include latest changes from drm-next branch.
This brings in sub-8bpp formats, AVUY and XVUY 8:8:8:8, and
Vivante tile-status and compression modifiers.
Generated using make headers_install.
Generated from drm-next branch commit 077bd80083ab
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
The Intel-specific code in libdrm is continually updated with new PCI
IDs for each new platform so that we can recognize the IP version
properly. However this is mostly a pointless exercise; none of the
Intel code in libdrm is conditional on IP versions above 8. If we just
treat any future unrecognized Intel platforms as IP version 8, we should
get the same behavior without the need for continued PCI ID updates.
Note that the intel_decode tool probably _should_ have conditions on
newer IP versions, but it was last updated for gen8 and has been
bitrotting from gen9 onward. This change won't make the tool behave any
more incorrectly than it already does today.
Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Up until now, DRM clients had to hand-roll their code to create,
destroy and map dumb buffers. This is slightly inconvenient,
a bit error-prone, and not easily discoverable.
Introduce wrappers for these operations, just like we have for
other KMS IOCTLs.
Signed-off-by: Simon Ser <contact@emersion.fr>
The compiler isn't smart enough to tell that this can't happen:
[30/74] Compiling C object amdgpu/libdrm_amdgpu.so.1.0.0.p/amdgpu_bo.c.o
In file included from ../amdgpu/amdgpu_internal.h:32,
from ../amdgpu/amdgpu_bo.c:39:
../xf86atomic.h: In function ‘amdgpu_find_bo_by_cpu_mapping’:
../xf86atomic.h:47:54: warning: ‘bo’ may be used uninitialized [-Wmaybe-uninitialized]
47 | # define atomic_inc(x) ((void) __sync_fetch_and_add (&(x)->atomic, 1))
| ^
../amdgpu/amdgpu_bo.c:536:27: note: ‘bo’ was declared here
536 | struct amdgpu_bo *bo;
| ^~
Signed-off-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
This makes the test utilities work with the i.MX LCDIFv3 driver
without the necessity of using the -M argument.
Signed-off-by: Marco Felsch <m.felsch@pengutronix.de>
Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
Wire up the pciaccess dep to the intel option. This automatically
skips the dep if intel is explicitly disabled, fails if intel is
explicitly enabled and it's not found, and disables intel if it's
set to auto and the dep is not found.
Signed-off-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Fixes the following warning:
[65/74] Compiling C object intel/libdrm_intel.so.1.0.0.p/intel_bufmgr_gem.c.o
../intel/intel_bufmgr_gem.c: In function âdrm_intel_gem_bo_unreferenceâ:
../intel/intel_bufmgr_gem.c:1388:9: warning: ISO C90 forbids mixed declarations and code [-Wdeclaration-after-statement]
1388 | drm_intel_bufmgr_gem *bufmgr_gem =
| ^~~~~~~~~~~~~~~~~~~~
Signed-off-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Generated from the Linux v6.0-rc7 tag with a sha1 of
f76349cf41451c5c42a99f18a9163377e4b364ff.
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Generated from the Linux v6.0-rc7 tag with a sha1 of
f76349cf41451c5c42a99f18a9163377e4b364ff.
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
It's not worth even attempting to configure anything on OSes where there
is no DRM to have a userspace library for.
This failure message can be useful in e.g. the case where libdrm is an
optional wrap fallback in another project.
Signed-off-by: Eli Schwartz <eschwartz93@gmail.com>
Reviewed-by: Dylan Baker <dylan@pnwbakers.com>
In commit 98794e2a0d ("lib: sync i915_pciids.h with kernel") we
resynchronized the PCI header with the kernel to bring in the
definitions for several new platforms. But before those IDs will be
recognized, we still need to hook them up in the libdrm chipset code as
well.
References: https://gitlab.freedesktop.org/drm/intel/-/issues/5416
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
There is a Vivante GC1000 gpu in LS2K1000 and LS7A1000.
LS7A1000 is a bridge chip made by Loongson corporation
which act as north and/or south bridge of loongson's
desktop and server level processor. It is equivalent
to RS780E or something like that. In fact, the company
use RS780E as bridge of LS3A3000 at its early stage,
but as RS780E is out of stock long long time ago, the
company have to made one by themself. More details can
be read from its user manual[1].
This bridge chip typically use with LS3A3000, LS3A4000
and LS3A5000.
LS3A3000 is 4 core 1.45gHz mips64r2 compatible cpu.
LS3A4000 is 4 core 1.8gHz mips64r5 compatible cpu.
LS3A5000 is 4 core 2.5gHz loongarch cpu, the company
acclaim that loongarch a new archtecture with its
instruction set is released[2].
LS2K1000 is a double core 1.0Ghz mips64r2 compatible SoC[3].
we need to enable it to test and developing driver on above
listed archtecture.
[1] https://loongson.github.io/LoongArch-Documentation/Loongson-7A1000-usermanual-EN.html
[2] https://loongson.github.io/LoongArch-Documentation/Loongson-3A5000-usermanual-EN.html
[3] https://wiki.debian.org/InstallingDebianOn/Lemote/Loongson2K1000
Signed-off-by: Sui Jingfeng <15330273260@189.cn>
[Eric: rebase over meson changes, add ARM & ARC architectures, and drop
"experimental" from the description]
Signed-off-by: Eric Engestrom <eric@engestrom.ch>
../intel/test_decode.c: In function ‘compare_batch’:
../intel/test_decode.c:109:39: error: dereferencing type-punned pointer might break strict-aliasing rules [-Werror=strict-aliasing]
109 | out = open_memstream((char **)&ptr, &size);
| ^~~~
cc1: some warnings being treated as errors
The fix is simple: just declare `ptr` as a `char *` to begin with.
Allows users to easily enable everything (eg. packagers), or select just
the drivers they want with something like:
-D auto-features=disabled -D amdgpu=enabled
Signed-off-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Simon Ser <contact@emersion.fr>
You can't have an error if your driver is requested by you're missing
a dep, but then happily build that driver without the dep in `auto`.
Signed-off-by: Eric Engestrom <eric@engestrom.ch>
Building the project as a meson subproject, meson inherits the warning level
from the parent project. Making the tests optional bypasses that issue and
reduces build time.
Signed-off-by: Simon Zeni <simon@bl4ckb0ne.ca>
Reviewed-by: Simon Ser <contact@emersion.fr>
Issue slow copy linear for sdma to trigger SDMA hang test.
Signed-off-by: Likun Gao <Likun.Gao@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
This synchronizes with kernel commit 7835303982d1 ("drm/i915/mtl: Add
MeteorLake PCI IDs") to bring in the missing PCI IDs for several recent
platforms.
These days adding PCI IDs to libdrm doesn't really matter for real-world
system usage. However there are still a few driver testing situations
where they're needed (such as the IGT dma-buf tests that still rely on
libdrm's bufmgr code). At some point we should probably break that
final IGT dependency on libdrm so that these PCI ID resyncs won't be
necessary anymore, but that hasn't happened yet.
References: https://gitlab.freedesktop.org/drm/intel/-/issues/5416
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Add an entry for the "simpledrm" driver, so that the test utilities
work with the simpledrm driver without passing the -M argument.
Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
Signed-off-by: Eric Curtin <ecurtin@redhat.com>
From VCN4, AMDGPU_HW_IP_VCN_ENC is re-used to support both encoding
and decoding jobs.
Reviewed-by: Boyuan Zhang <Boyuan.Zhang@amd.com>
Signed-off-by: Ruijing Dong <ruijing.dong@amd.com>
Nowadays, users don't really care about encoders except for retrieving
the list of CRTCs compatible with a connector. Introduce a new function
so that users no longer need to deal with encoders.
This is a re-do of [1], but with a slightly different API.
Signed-off-by: Simon Ser <contact@emersion.fr>
[1]: https://gitlab.freedesktop.org/mesa/drm/-/merge_requests/102
This is the only modifier printed with a "_MODIFIER" suffix. It
looks inconsistent when callers already print this word (e.g.
"modifier: INVALID_MODIFIER").
Signed-off-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Marius Vlad <marius.vlad@collabora.com>
Replaced the type PRId64 with PRIu64 in a printf as the argument was
unsigned to fix the related compiler warning.
Signed-off-by: Eleni Maria Stea <elene.mst@gmail.com>
Reviewed-by: Simon Ser <contact@emersion.fr>
It seems that __u64 values are defined differently across systems. In
glibc it's defined as unsigned long, in Linux kernel headers
(int-ll64.h) as unsigned long long, and on FreeBSD as uint64_t so it
matches glibc. A temporal solution is to cast all __u64 values to
uint64_t to avoid warnings on Linux, but ideally we'd like a better fix
in the future.
See also: https://gitlab.freedesktop.org/mesa/drm/-/merge_requests/212
for discussion.
Signed-off-by: Eleni Maria Stea <elene.mst@gmail.com>
Moved declaration to the top to resolve C99 compliance warning.
Signed-off-by: Eleni Maria Stea <elene.mst@gmail.com>
Reviewed-by: Simon Ser <contact@emersion.fr>
libkms was a very early attempt at a KMS management library, that only
got as far as handling requests to create buffers. It has since been
superseded by GBM in doing this, which everyone uses, unlike libkms
which no-one uses.
Remove it from the tree to avoid any confusion.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Switch the logic to only disable the tests for asics which don't
have GPU reset support. This way we don't need to update it
every time we add a new asic which does support it.
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
I tested with latest amd-staging-drm-next and after minor
fix for me all the testys pass. I bumped libdrm minor version
for this.
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
This test will attempt to use the VIC to blit one surface to another
and perform a vertical flip.
Reviewed-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
This test will attempt to use the VIC to blit from one surface to
another.
Reviewed-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
This test will attempt to use VIC to clear a surface.
Reviewed-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
The Video Image Composer (VIC) 4.2 can be found on NVIDIA Tegra194 SoCs.
It uses a different class (C5B6) that is slightly incompatible with the
class found on earlier generations, although it is backwards compatible
with the class implemented on Tegra186 (B1B6).
Reviewed-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
The Video Image Composer (VIC) 4.1 can be found on NVIDIA Tegra186 SoCs.
It uses a different class (B1B6) that is slightly incompatible with the
class found on earlier generations.
Reviewed-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
The Video Image Composer (VIC) 4.0 can be found on NVIDIA Tegra210 SoCs.
It uses a different class (B0B6) that is slightly incompatible with the
class found on earlier generations.
Reviewed-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
The Video Image Composer (VIC) 3.0 can be found on NVIDIA Tegra124 SoCs.
Reviewed-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Implement a small abstraction interface to allow different versions of
VIC to be used transparently. An implementation will be chosen based on
the VIC version number reported by the DRM_TEGRA_IOCTL_OPEN_CHANNEL
IOCTL.
Reviewed-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
This test can be used to purposefully trigger a job timeout.
Reviewed-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
This is a very simple sanity test to check whether or not a syncpt can
be incremented by a host1x client. This uses gr2d on Tegra20 through
Tegra114 and VIC on Tegra124 and later.
Reviewed-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
This test uses the IOCTLs for job submission and fences to fill a sub-
region of the screen to a specific color using gr2d.
Reviewed-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
This library provides helpers for common functionality needed by test
programs.
Reviewed-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
These new functions can be used to allocate and free syncpoints, as well
as wait for a syncpoint threshold to be reached. Jobs can also be waited
on if a syncpoint was attached to them.
Reviewed-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
These new functions can be used to create a job on a given channel, add
commands to the job using its push buffer and submit the job.
Reviewed-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
These new functions can be used to open a channel to a given engine, map
and unmap buffer objects to that channel, and close the channel.
Reviewed-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
This makes sure that the proper dependencies are created and that the
file is distributed.
Reviewed-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
This new UABI is a more modern version that works better with both old
and recent chips.
Reviewed-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Allow this simple test to be installed so that it can easily be run on a
target device.
Reviewed-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Most functions in libdrm_tegra take as first parameter the object that
they operate on. Make the device and buffer object creation functions
follow the same scheme.
Reviewed-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
These helpers facilitate exporting and importing buffer objects to and
from PRIME file descriptors.
Reviewed-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Add helpers to export and import buffer objects via flink names.
Reviewed-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Store 64-bit offset values and use libdrm's built-in drm_mmap() function
instead of mmap() to ensure the full 64-bit offset is used.
Reviewed-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
All of the buffer object allocation functions use the same boilerplate
code. Move that code into a separate function that can be reused.
Reviewed-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
The DRM_TEGRA_GEM_{GET,SET}_FLAGS and DRM_TEGRA_GEM_{GET,SET}_TILING
IOCTLs were badly designed and have since been obsoleted by framebuffer
modifiers. Remove these implementations to make it clear their usage is
discouraged.
Reviewed-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Reindent the sources according to the settings found in the newly added
.editorconfig.
Reviewed-by: Mikko Perttunen <mperttunen@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Atomic request property lists are defined to be de-duplicated: an atomic
request can contain multiple sets for the same property on the same
object, and only the last one will take effect.
drmModeAtomicCommit already sorts the property set by object and
property ID. We were relying on qsort to also sort by cursor - i.e.
pointer value - when object and property ID are equal, however whilst
glibc does this, the sort order is explicitly undefined when the
comparator is equal. Using the pointer is also not stable on all
implementations.
Add an explicit 'cursor' member to each property set which is used as
the tie-breaker comparator.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Fixes: #46
It's cleaner, it's nicer looking, and it's a nice builtin.
Signed-off-by: Dylan Baker <dylan@pnwbakers.com>
Reviewed-by: Simon Ser <contact@emersion.fr>
This produces no differences in the generated output. I've had to
manually add `requires : 'libdrm'` to libdrm_intel, otherwise libdrm
ends up in `Requires.private` instead of `Requires`.
Signed-off-by: Dylan Baker <dylan@pnwbakers.com>
Reviewed-by: Simon Ser <contact@emersion.fr>
It's less code, and also allows meson to short circuit for compilers is
knows don't support this.
Signed-off-by: Dylan Baker <dylan@pnwbakers.com>
Reviewed-by: Simon Ser <contact@emersion.fr>
This is generally faster, as meson is able to parallelize the checks for
us.
This also removes the workaround for checking gcc/clang -Wno-*
arguments, which meson now handles internally so we don't need to handle
it ourselves.
Signed-off-by: Dylan Baker <dylan@pnwbakers.com>
Reviewed-by: Simon Ser <contact@emersion.fr>
This allows meson to check if the compiler supports gnu style symbol
visibility, and apply the appropriate flags as necessary, rather than us
adding them by hand
Signed-off-by: Dylan Baker <dylan@pnwbakers.com>
Reviewed-by: Simon Ser <contact@emersion.fr>
And switch to c_std=c99. This simplifies using libdrm as a meson
subproject for mesa.
v2: (dylan)
- switch to c99 as the standard
- Fix amdgpu security tests as well
Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Signed-off-by: Dylan Baker <dylan@pnwbakers.com>
Reviewed-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Emma Anholt <emma@anholt.net>
This allows consumers of libdrm as a subproject to use the simpler
`dependency('libdrm', fallback : 'libdrm')` syntax, as the libdrm build
files already tell meson that they override a dependency called
"libdrm".
Signed-off-by: Dylan Baker <dylan@pnwbakers.com>
Reviewed-by: Simon Ser <contact@emersion.fr>
So we don't have to duplicate the libdrm library call just to not set
the version keyword for android
Reviewed-by: Simon Ser <contact@emersion.fr>
Signed-off-by: Dylan Baker <dylan@pnwbakers.com>
v2:
- remove dec create/destroy msg as its not relevant to jpeg (Leo)
- enable the test for jpeg2/jpeg3 (Leo)
- validate checksum of result (Leo)
- add appropriate comments (Leo)
v3:
- linux style function definition indent (James)
- use multiline comment delimiter (Leo)
Signed-off-by: Sathishkumar S <sathishkumar.sundararaju@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Used casting to fix warnings about assigning different enum types to
variables. Used error checks in places where snprintf is called and
output might be truncated to fix gcc format-truncation warnings.
v2: Removed a change in drm.h (Simon Ser)
v3, v4: Removed unecessary braces in snprintf (Simon Ser)
Signed-off-by: Eleni Maria Stea <elene.mst@gmail.com>
Dual purpose:
- The drm fd dedupe functionality confuses the radeonsi
amdgpu winsys if radeonsi isn't the first thing opening
the device. By exposing the fd we can detect this case.
- For a common mesa Vulkan sync objects implementation
with syncobj. (notable: no buffer allocation)
Both shouldn't interferece with libdrm_amdgpu functionality
though it does somewhat piece the abstraction of the library.
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Gitlab: https://gitlab.freedesktop.org/mesa/mesa/-/issues/3424
Gitlab: https://gitlab.freedesktop.org/mesa/mesa/-/issues/5630
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Use qemu to do CI on FreeBSD.
Not everything is compiled as all arm aren't supported on FreeBSD.
Same thing for Nouveau.
The tests aren't enable for now as they are all failing.
Signed-off-by: Emmanuel Vadot <manu@FreeBSD.org>
This adds a function to get a drmDevicePtr from a dev_t identifier
of a device. This is useful for Wayland that uses these to identify
devices over the protocol.
This is done by taking the implementation of drmGetDevice2, and removing
the call to fstat to find the dev_t.
Signed-off-by: Scott Anderson <scott.anderson@collabora.com>
Signed-off-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Add support for parsing IN_FORMATS property blobs. Providing libdrm
with this functionality helps to standardise how user-space reads
kernel blobs and decreases duplication on the client side.
drmModeFormatModifierBlobIterNext() allows the caller to view
formats and associated modifiers given a valid property blob.
An example is available inside the libdrm unit test, modetest.c.
Signed-off-by: Luigi Santivetti <luigi.santivetti@imgtec.com>
Reviewed-by: Simon Ser <contact@emersion.fr>
On 64-bit FreeBSD targets uint64_t is generally defined as `unsigned long`
and not `unsigned long long`. Use the PRI macros to fix -Wformat.
Signed-off-by: Alex Richardson <Alexander.Richardson@cl.cam.ac.uk>
Reviewed-by: Simon Ser <contact@emersion.fr>
Simple yet useful command line utility to simulate memory pressure.
Already found quite a number of problems in TTM with that.
v2: replace spaces with tabs
Signed-off-by: Christian König <christian.koenig@amd.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Instead of manually calling drmIoctl, use the equivalent function
from libdrm core.
Signed-off-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Instead of manually calling drmIoctl, use the equivalent function
from libdrm core.
Signed-off-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Instead of manually calling drmIoctl, use the equivalent function
from libdrm core.
Signed-off-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Instead of manually calling drmIoctl, use the equivalent function
from libdrm core.
Signed-off-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Instead of manually calling drmIoctl, use the equivalent function
from libdrm core.
Signed-off-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Instead of manually calling drmIoctl, use the equivalent function
from libdrm core.
Signed-off-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Instead of manually calling drmIoctl, use the equivalent function
from libdrm core.
Signed-off-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Instead of manually calling drmIoctl, use the equivalent function
from libdrm core.
Signed-off-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Instead of using a hand-rolled amdgpu_close_kms_handle function,
use the function from libdrm core, which does exactly the same
thing.
Signed-off-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Update drm_fourcc.h to include latest changes from drm-next branch.
This brings in AFRC (Arm Fixed-Rate Compression) modifiers.
Generated using make headers_install.
Generated from drm-next branch commit 6880fa6
Signed-off-by: Dennis Tsiang <dennis.tsiang@arm.com>
No need to have two branches depending on DRM_MODE_PROP_EXTENDED_TYPE.
We can just use drmModeGetPropertyType instead.
This does introduce a slight change: previously, drm_property_type_is()
could be called with non-type flags such as IMMUTABLE. However no user
seems to do this (checked KWin/Mutter/Sway/Weston/Xorg).
Signed-off-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
This function only needs read-only access to the property. This is
not a breaking ABI change.
Signed-off-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Execbuffer2 support was introduced to libdrm in b50964027b, 10 years
ago, and no driver has used the old execbuf path since. There's no need
to support 10-year-old kernels.
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
We have wrappers for PRIME_HANDLE_TO_FD and PRIME_FD_TO_HANDLE,
but not for GEM_CLOSE. Add it so that callers don't need to
manually call drmIoctl.
Signed-off-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Acked-by: Daniel Stone <daniels@collabora.com>
There is this long standing nit of igt/tools/intel_error_decode asserting
when you feed it an error state from a GPU the local libdrm does not know
of.
To fix this I need a tweak in drm_intel_decode_context_alloc to make it
not assert but just return NULL (which seems an already possible return
value).
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Zbigniew Kempczyński <zbigniew.kempczynski@intel.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Sync libdrm with kernel, a new DG1 pci was added.
Commit 5f0d4214938d ("drm/i915/dg1: Add new PCI id")
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
`sysfs_uevent_get()` returns a `strndup()`ed string, which must be `free()`d.
Fixes: bf63f8acdc ("libdrm: Handle usb_interface devices for usb parsing")
Reviewed-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
To update your local repository to use the new default branch, these
commands may help:
$ git fetch origin
$ git checkout master
$ git branch -m main
$ git branch --set-upstream-to=origin/main
$ git remote set-head origin --auto
Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Generated using make headers_install from the drm-next
tree - git://anongit.freedesktop.org/drm/drm
branch - drm-next
commit - 8a02ea42bc1d4c448caf1bab0e05899dad503f74
Some changes were omitted, e.g., to nouveau_drm.h, i915_drm.h, and
msm_drm.h, as the nouveau and i915 changes looked to me as if they
could break compatibility or require other compatibility fixes to
libdrm which i can not judge. msm_drm.h broke the build, as there
are definitely changes needed to libdrm's msm support code.
The shortlog below is edited to only list what corresponds to files
that are included here, because it looked safe to me.
The changes were as follows (shortlog from
b10733527bfd864605c33ab2e9a886eec317ec39..HEAD):
Aaron Liu (1):
drm/amdgpu: add uapi to define yellow carp series
Alex Deucher (1):
drm/amdgpu: add INFO ioctl support for querying video caps (v4)
Christian Gmeiner (1):
drm/etnaviv: provide more ID values via GET_PARAM ioctl.
Felix Kuehling (1):
drm/amdgpu: Add new placement for preemptible SG BOs
Jiawei Gu (1):
drm/amdgpu: Add vbios info ioctl interface
Lionel Landwerlin (1):
drm: fix drm_mode_create_blob comment
Mario Kleiner (1):
drm/fourcc: Add 16 bpc fixed point framebuffer formats.
Nirmoy Das (1):
drm/amdgpu: remove AMDGPU_GEM_CREATE_SHADOW flag
Noralf Trønnes (1):
drm/uapi: Add USB connector type
Radhakrishna Sripada (1):
drm/framebuffer: Format modifier for Intel Gen 12 render compression with Clear Color
Simon Ser (13):
drm: improve kernel-docs in drm_mode.h
drm: document drm_mode_get_connector
drm: document drm_mode_modeinfo
drm: document that user-space should force-probe connectors
drm/doc: atomic implicitly enables other caps
drm/doc: re-format drm.h file comment
drm/doc: demote old doc-comments in drm.h
drm/fourcc: fix Amlogic format modifier masks
drm/uapi: document kernel capabilities
drm/connector: demote connector force-probes for non-master clients
drm: reference mode flags in DRM_CLIENT_CAP_* docs
drm: clarify and linkify DRM_CLIENT_CAP_WRITEBACK_CONNECTORS docs
drm: document minimum kernel version for DRM_CLIENT_CAP_*
Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com>
Android vendor libraries don't have sonames, and libdrm.so shouldn't
either. This lets a Mesa built against a libdrm.so built for Android
be copied directly to a Chrome OS ARC installation.
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
SDMA copy from Alice to Bob is in TMZ mode. Therefore
SDMA copy back from Bob to Alice should be in TMZ mode too.
Signed-off-by: Aaron Liu <aaron.liu@amd.com>
Reviewed-by: Luben Tuikov <luben.tuikov@amd.com>
Fix the TMZ secure bounce test, in that Bob's
buffer has to be created encrypted (with the
encrypted flag set), so that when we copy from
Alice's buffer, which is also encrypted, to Bob's
buffer, the copy can be successful and the data
actually copied.
This fixes the test and it no longer fails. Tested
on Sienna Cichlid.
Cc: Alex Deucher <Alexander.Deucher@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Aaron Liu <aaron.liu@amd.com>
Cc: Huang Rui <ray.huang@amd.com>
Signed-off-by: Luben Tuikov <luben.tuikov@amd.com>
Reviewed-by: Alex Deucher <Alexander.Deucher@amd.com>
As format modifiers can be encoded in quite complex forms, the static
table previously added is not sufficient to retrieve, extract and decode
the token formats to a human-readable string. This patch introduces a
vendor specific callback which could be used to perform an additional
search to match up with vendor encoding scheme, which, will be used
first, before resorting to searching the static table.
With it, add support for decoding the ARM format modifiers.
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
Introduces two new methods to retrieve a human readable representation of a
format modifier:
drmGetFormatModifierName() - returns a format modifier as a string,
from a token modifier
drmGetFormatModifierVendor() - returns the vendor as a string, from a
token modifier
and the fourcc_mod_get_vendor macro that returns the vendor.
New format modifiers added in drm_fourcc.h uapi kernel header should be
sync'ed up with libdrm and should include a human readable
representation for that format modifier, in order to display it
correctly as a string.
That happens with the help of a python script that reads up drm_fourcc
header file and outputs a static table comprised of token modifiers
alongside a vendor table (Suggested-by Simon Ser <contact@emersion.fr>).
The reason for doing it in libdrm is to have a unified place instead of each
user of libdrm having a way to keep track of the format modifiers.
With this patch, modetest has also been modified to make use of it.
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
<linux/limits.h> should be included conditionally for Linux only, also
SPECNAMELEN used conditionally when the OS is FreeBSD requires to
include <sys/params.h>.
Signed-off-by: Eleni Maria Stea <elene.mst@gmail.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
In FreeBSD's errno.h ERESTART is not defined by default, only when the
user requests the pseudo-errors returned inside the kernel to be
enabled. As a result the block where drmCommandWriteRead is
called returns compile error. Defined _WANT_KERNEL_ERRNO to fix it (see
FreeBSD's /usr/include/errno.h)
Signed-off-by: Eleni Maria Stea <elene.mst@gmail.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
<sys/types.h> need to be included in xf86drmMode.c for type u_int in
<sys/sysctl.h> (that is included when OS is FreeBSD) to be recognized.
Signed-off-by: Eleni Maria Stea <elene.mst@gmail.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
In Vulkan we have extensions to assist with capture in replay in a
world where addresses are returned to the application. This involves
creating buffers at the same VA during replay as they were during
capture.
By itself libdrm_amdgpu already has support for this, but there is
the obvious failure mode that if another buffer is already allocated
at that VA things fail spectacularly. This is an actual issue as
internal buffers, like winsys images or shader binaries also
participate in the same VA allocation.
To avoid this problem applications have to create buffers which
are going to be captured with a flag, and the implementation is to
separate VA allocation for those buffers to reduce the collision risk:
"Implementations are expected to separate such buffers in the GPU address
space so normal allocations will avoid using these addresses. Apps/tools
should avoid mixing app-provided and implementation-provided addresses for
buffers created with VK_BUFFER_CREATE_DEVICE_ADDRESS_CAPTURE_REPLAY_BIT,
to avoid address space allocation conflicts."
This patch implements that by adding a flag for these buffers and allocating
address space from the top of the address range instead of the bottom.
Signed-off-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Reviewed-by: Christian König <christian.koenig@amd.com>
Disconnect device while fence is exported.
Also disable this test for sytem with single GPU.
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed-by: Alex Deucher alexander.deucher@amd.com
Add plug/unplug device and open/close device file
infrastructure.
Add basic test - unplug device while device file still
open. Close device file afterwards and replug the device.
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed-by: Alex Deucher alexander.deucher@amd.com
nvc0 sets the NVC0_IB_ENTRY_1_NO_PREFETCH bit on some pushbuffers
Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>
This reverts commit b362850689.
This breaks when the kernel driver does not support modifiers and the
application properly zeroes the modifiers.
Acked-by: Simon Ser <contact@emersion.fr>
In syncobj test, 3 threads will be created. Sometimes
the first gfx IB and the third sdma IB will use same
physical page. There will be risk that sdma engine will
read gfx IB in the same physical page. So better to flush
the cache before commit the sdma IB.
Signed-off-by: Jinzhou Su <Jinzhou.Su@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
buffer_load/store_format_xyzw require 64bit vgpr_a[2].
The original parameter is one u32. Modify the shader binary to
fit the 64bit parameter.
Signed-off-by: Feifei Xu <Feifei.Xu@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Tested-by: Gang Long <Gang.Long@amd.com>
The kernel will always return EINVAL if modifiers are supplied but
the flag DRM_MODE_FB_MODIFIERS isn't set. That's a pretty nice
footgun.
Be a little more helpful and set the flag if the user has supplied
a modifier array.
Signed-off-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
On Raven2/Picasso, the default VRAM size is 2048M,
and the default GTT size is 3072M. If max_allocation
of VRAM exceeds half of GTT size, GTT memory can't
hold evicted bo from VRAM and bo in itself at the
same time. Then amdgpu_cs_list_validate will failed
with "Not enough memory for command submission" error.
NOTE:
The installed DRAM should be larger than 8GB,
if the VRAM size is 2048M.
Signed-off-by: Lang Yu <Lang.Yu@amd.com>
Signed-off-by: Marek Olšák <marek.olsak@amd.com>
The unit of size_metadata is one byte not four bytes.
Enable Metadata test.
Signed-off-by: Lang Yu <Lang.Yu@amd.com>
Signed-off-by: Marek Olšák <marek.olsak@amd.com>
We already have drm_property_type_is, but it's needlessly complicated
and doesn't cover all use-cases (requires the caller to provide a
type).
Signed-off-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
In syncobj test, wait thread and signal thread create
simultaneously. The ptr for GFX IB and SDMA IP should be
operated separately. With static, there will be risk that
GFX NOP is in SDMA IB or SDMA NOP is in GFX IB, then GFX or
SDMA hang caused.
Signed-off-by: Jinzhou Su <Jinzhou.Su@amd.com>
Signed-off-by: Marek Olšák <marek.olsak@amd.com>
The general direction at this time is to phase out pread/write ioctls and
not support them in future products. The ioctls have already been disabled
in i915 for future products. This means libdrm must handle the absence of
these ioctls. This patch does this by modifying drm_intel_gem_bo_subdata()
and drm_intel_gem_bo_get_subdata() to do the read/write using the
pread/pwrite ioctls first but when these ioctls are unavailable fall back
to doing the read/write using a combination of mmap and memcpy.
A similar solution was added to igt-gpu-tools in commit
ad5eb02eb3 ("lib/ioctl_wrappers: Keep IGT working without pread/pwrite
ioctls").
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Signed-off-by: Ashutosh Dixit <ashutosh.dixit@intel.com>
Do as the documentation says - when devices non NULL, cap the reported
devices to max_devices. Otherwise we risk out-of-bound access
for users of the API.
v2:
- Fix this w/o breaking the API
v3:
- Drop local variables, flip inverted conditional (Simon)
Issue: https://gitlab.freedesktop.org/mesa/drm/-/issues/56
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Simon Ser <contact@emersion.fr>
This reverts commit 8cb12a2528.
The commit fixed the OOB, yet it broke drmDevices2(0, NULL, 0) - aka we
did not return the total devices list.
Reviewed-by: Simon Ser <contact@emersion.fr>
What are these headers?
Adding currently missing stuff from https://cgit.freedesktop.org/drm/drm/tree/include/uapi/drm/amdgpu_drm.h based on
the latest commit there:
commit f730f39eb981af249d57336b47cfe3925632a7fd (HEAD -> drm-next, tag: drm-next-2021-02-19, origin/drm-next, origin/HEAD)
Merge: 4f8ad4045b38 81ce8f04aa96
Author: Dave Airlie <airlied@redhat.com>
Date: Fri Feb 19 13:54:29 2021 +1000
Merge tag 'drm-intel-next-fixes-2021-02-18' of git://anongit.freedesktop.org/drm/drm-intel into drm-next
Which headers go where?
From https://cgit.freedesktop.org/drm/drm/tree/include/uapi/drm/amdgpu_drm.h to
https://cgit.freedesktop.org/mesa/drm/tree/include/drm/amdgpu_drm.h
When and which headers to update?
If the kernel uapi drm header changes, the header here should be sync-ed.
When and how to update these files
The steps for generating this patch:
- Switch to freedesktop drm-next kernel branch (https://cgit.freedesktop.org/drm/drm);
- Install the headers via `make headers_install';
- Copy from kernel "include/uapi/drm/amdgpu_drm.h" to libdrm "include/drm/amdgpu_drm.h";
- generate the patch;
The commits from drm-next (https://cgit.freedesktop.org/drm/drm) are:
Mauro Carvalho Chehab (1)
c45dd3bda1c809eb120452597097e14a96b58c1f drm/amdgpu: fix some kernel-doc markups
Huang Rui(3)
6fbcb00c7984fa7d49af2c361453c0397cdea400 drm/amdgpu: add TOC firmware definition
1e483203965bdab466af0739c1edf7da07da241d drm/amdgpu: add uapi to define van gogh memory type
f7b2cdb23abf62bc3d33c2e0b0009a09412ff475 drm/amdgpu: add uapi to define van gogh series
Pierre-Eric Pelloux-Prayer(1)
16c642ec3fe9a144fbe1e97dc56f13a6308f1381 drm/amdgpu: new ids flag for tmz (v2)
Yong Zhao(1)
130c88931f6cbdb4513d307b4a13fcfff08a8041 drm/amdgpu: Improve the MTYPE comments
Signed-off-by: Leo Liu <leo.liu@amd.com>
- Remove one unused variable.
- Convert two int-s into 'unsigned int'.
Motivated by a failed build of Chromium.
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Victor Hugo Vianna Silva <victor.vianna10@gmail.com>
Add an entry for the "mxsfb-drm" driver, so that the test utilities
work with the mxsfb driver without passing the -M argument.
Signed-off-by: Fabio Estevam <festevam@gmail.com>
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
The amdgpu implementation uses `clock_gettime` so it needs to check whether it needs to link
against `-librt`.
Signed-off-by: Valentin Churavy <v.churavy@gmail.com>
Generated using make headers_install from the drm-next
tree - git://anongit.freedesktop.org/drm/drm
branch - drm-next
commit - b10733527bfd864605c33ab2e9a886eec317ec39
The changes were as follows (shortlog from
14d2bd53a47a7e1cb3e03d00a6b952734cf90f3f):
core: (drm_mode.h)
Alexander A. Klimov (1):
drm: Replace HTTP links with HTTPS ones
Noralf Trønnes (1):
drm: Add SPI connector type
Oleg Vasilev (1):
drm: report dp downstream port type as a subconnector property
Simon Ser (1):
drm: document that blobs are ref'counted
Uma Shankar (3):
drm: Add HDR source metadata property
drm: Fixed doc warnings in drm uapi header
drm: Fix docbook warnings in hdr metadata helper structures
core: (drm_fourcc.h)
Adam Jackson (1):
drm/fourcc: Fix undefined left shift in DRM_FORMAT_BIG_ENDIAN macros
Bas Nieuwenhuizen (2):
drm/fourcc: Add AMD DRM modifiers.
drm/fourcc: Fix modifier field mask for AMD modifiers.
Ben Davis (2):
drm: drm_fourcc: add NV15, Q410, Q401 YUV formats
drm: drm_fourcc: Add uncompressed AFBC modifier
Brian Starkey (1):
drm: drm_fourcc: Add generic alias for 16_16_TILE modifier
Dave Airlie (1):
Merge tag 'amd-drm-next-5.11-2020-11-05' of git://people.freedesktop.org/~agd5f/linux into drm-next
Dhinakaran Pandiyan (2):
drm/framebuffer: Format modifier for Intel Gen-12 render compression
drm/framebuffer: Format modifier for Intel Gen-12 media compression
James Jones (1):
drm: Generalized NV Block Linear DRM format mod
Maarten Lankhorst (1):
Backmerge remote-tracking branch 'drm/drm-next' into drm-misc-next
Matteo Franchin (1):
drm/fourcc: Add AXBXGXRX106106106106 format
Mika Kahola (1):
uapi/drm/drm_fourcc.h: Note on platform specificity for format modifiers
Neil Armstrong (2):
drm/fourcc: Add modifier definitions for describing Amlogic Video Framebuffer Compression
drm/fourcc: fix Amlogic Video Framebuffer Compression macro
Raymond Smith (1):
drm/fourcc: Add Arm 16x16 block modifier
Simon Ser (4):
drm/fourcc: document modifier uniqueness requirements
drm: deprecate DRM_FORMAT_MOD_NONE
drm/fourcc: add table describing AMD modifiers bit layout
drm/fourcc: fix AMD modifiers PACKERS field doc
Signed-off-by: Antonin Décimo <antonin.decimo@gmail.com>
We include stdint.h unconditionally in the header. We don't require
users to include it manually before xf86drmMode.h.
Signed-off-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
I don't exactly know why these were duplicated before. Maybe libdrm
didn't always vendored drm_mode.h from the kernel? In any case, we now
do, so instead of having copy-pasted definitions, just include our
vendored version which cannot be outdated.
Contrary to what the comment says, drm.h doesn't include drm_mode.h, so
we need to add the include.
Signed-off-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
DocBook makes it hard to write and maintain docs. Hopefully
reStructuredText can make this less painful.
The man pages were converted from DocBook to reStructuredText via
Pandoc:
pandoc -s -f docbook -t rst -o man/drm.7.rst man/drm.xml
And then manual editing to fixup e.g. references to other man pages. To
compare the result with the DocBook version, this command was used:
rst2man man/drm-kms.7.rst | man -l -
Signed-off-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Instead of always dumping the rendered picture, check whether it matches
the expectations. This makes more sense for automated testing.
Retain the ability to dump the picture instead of checking it when a
file name is given as an argument. This also removes use of a hardcoded
file name in a world writable directory, which is an unsafe thing to
do anyway.
Signed-off-by: Lubomir Rintel <lkundrak@v3.sk>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Run the test on a core capable of 2D rendering instead of hardcoding to
core zero.
Signed-off-by: Lubomir Rintel <lkundrak@v3.sk>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
This makes the test utilities work with the i.MX DCSS driver without the
necessity of using the -M argument.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
log() is affected by FP control word and can provide inaccurate result.
Fixes Killer Instinct under Wine not being able to find AMD vulkan
device.
Reviewed-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Paul Gofman <pgofman@codeweavers.com>
komeda is one of the supported GPUs in the kernel tree so this adds it
to libdrm modules to look for in tests.
Signed-off-by: Carsten Haitzler <carsten.haitzler@arm.com>
Implement secure bounce test. Steps implemented
as outlined by Christian K.
v2: Remove gpu_info; add comment describing
the purpose and steps of the test.
v3: Parameterize "secure" in amdgpu_bo_lcopy() and
amdgpu_bo_move(). Set them both to 0.
Allocate buffer Bob to be non-TMZ.
v4: Fix an off-by-one bug which was causing
the test to segfault.
Acked-by: Huang Rui <ray.huang@amd.com>
Acked-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Luben Tuikov <luben.tuikov@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
DMA's atomic behavir is unlike GFX,If the comparing data is not
equal to destination data,
For GFX, loop again till gfx timeout(system hang).
For DMA, loop again till timer expired and then send interrupt.
So testcase can't use interrupt mechanism.
We take another way to verify. When the comparing data is not
equal to destination data, overwrite the source data to the destination
buffer. Otherwise, original destination data unchanged.
So if the bo_cpu data is overwritten, the result is passed.
Steps:
1. use linear write packet to write 0xdeadbeaf to secure buffer,
2. use atmoic packet and ATOMIC_CMPSWAP_RTN_32 opcode to compare
the cmp_data(0xdeadbeaf) to the written data which has been encrypted.
v2: add the case of (dest_data != cmp_data).
Acked-by: Huang Rui <ray.huang@amd.com>
Acked-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Aaron Liu <aaron.liu@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Secure buffer is only able to be read with trusted ip block.
So we need use GFX ip to read it back instead of CPU.
Steps:
1. use write_data packet to write 0xdeadbeaf to secure buffer,
2. use atmoic_mem packet and ATOMIC_CMPSWAP_RTN_32 opcode to compare
the cmp_data(0xdeadbeaf) to the written data which has been encrypted.
If the result is equal, then overwrite the src_data(0x12345678) to the
secure buffer and return directly. Otherwise loop again until gfx timeout
and the secure buffer data unchanged.
Acked-by: Huang Rui <ray.huang@amd.com>
Acked-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Aaron Liu <aaron.liu@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This patch is to add add device handle as input param for exec_cs_helper
and write_linear_helper.
Because they are needed in security tests.
v2: fix typo that basic tests should be un-secure.
v3: refine the function implementation.
v4: remove amdgpu_cs_ctx_create3 calling.
Acked-by: Huang Rui <ray.huang@amd.com>
Acked-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Aaron Liu <aaron.liu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This patch is to add a new test suite to store security tests.
In Raven+ asics, it will support TMZ (trust memory zone), and it is
page-based protection feature.
v2: remove tests/amdgpu/Makefile.am and update to
tests/amdgpu/meson.build
Acked-by: Huang Rui <ray.huang@amd.com>
Acked-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Aaron Liu <aaron.liu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Since Arcturus has no gfx pipeline(CPG), cases below is not suitable:
- Command submission Test (GFX)
- Command submission Test (Multi-Fence)
- Sync dependency Test
Acked-by: Huang Rui <ray.huang@amd.com>
Acked-by: Leo Liu <leo.liu@amd.com>
Signed-off-by: Le Ma <le.ma@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This enables drm_intel_bufmgr on DG1 and allows us to pass dmabuf
import/export related tests with Piglit.
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Missing RKL PCI ID preventing below test cases to succeed on RKL Platform.
igt@kms_frontbuffer_tracking
igt@kms_draw-crc
igt@kms_big_fb
Signed-off-by: sunil kumar dora sermsity <sunilx.kumar.dora.sermsity@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Two new patches landed in kernel adding new PCI ids:
123f62de419f ("drm/i915/rkl: Add RKL platform info and PCI ids")
52797a8e8529 ("drm/i915/ehl: Add new PCI ids")
Cc: Matt Roper <matthew.d.roper@intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
This is workaround of firmware issue, and the change has no impact
on the legacy HW.
Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Thong Thai <thong.thai@amd.com>
Reviewed-by: James Zhu <James.Zhu@amd.com>
It is not used for VCN from VCN1, but VCN3 use it
for other feature, so clear it, because we don't
use the feature for now
Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Thong Thai <thong.thai@amd.com>
Reviewed-by: James Zhu <James.Zhu@amd.com>
./xf86drm.c: In function 'drmNodeIsDRM':
../xf86drm.c:2825:7: error: "__FreeBSD__" is not defined [-Werror=undef]
#elif __FreeBSD__
^
../xf86drm.c: In function 'drmGetMinorNameForFD':
../xf86drm.c:2938:7: error: "__FreeBSD__" is not defined [-Werror=undef]
#elif __FreeBSD__
^
../xf86drm.c: In function 'drmParsePciBusInfo':
../xf86drm.c:3258:7: error: "__FreeBSD__" is not defined [-Werror=undef]
#elif __FreeBSD__
^
../xf86drm.c: In function 'drmParsePciDeviceInfo':
../xf86drm.c:3427:7: error: "__FreeBSD__" is not defined [-Werror=undef]
#elif __FreeBSD__
^
../xf86drm.c: In function 'drmGetDeviceNameFromFd2':
../xf86drm.c:4305:7: error: "__FreeBSD__" is not defined [-Werror=undef]
#elif __FreeBSD__
^
Signed-off-by: Heiko Thiery <heiko.thiery@gmail.com>
This option finds all connected connector and then sets its preferred
mode on it. If no preferred mode is available, first mode is used.
This option must be set w/o any mode or plane.
This allows for a quick test on all connected outputs.
Loosely based on the work by Ezequiel Garcia <ezequiel@collabora.com>
Changes since Ezequiel's work:
- implement atomic codepath
- set all connectors
- pick correct crtc
- don't set -r by default
- nearly identical output in atomic and non-atomic codepaths
v2:
- Use the crtc->crtc_id, instead of the plane's current crtc_id
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Ezequiel Garcia <ezequiel@collabora.com>
Tested-by: Ezequiel Garcia <ezequiel@collabora.com>
The very final drmModeAtommicCommit tears down the existing mode/plane
setup. Following it we clean up other misc state laying around.
Chances are that it will not fail, but in the extremely unlikely case it
does, there's nothing one can do.
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Ezequiel Garcia <ezequiel@collabora.com>
Tested-by: Ezequiel Garcia <ezequiel@collabora.com>
Move the hunk of code into a function, making the overall flow easier to
follow and providing some symmetry to the non-atomic path.
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Ezequiel Garcia <ezequiel@collabora.com>
Tested-by: Ezequiel Garcia <ezequiel@collabora.com>
Instead of duplicating the exact same code across the two functions,
fold them into one.
For some strange reason git diff may show atomic_clear_mode() as changed
The function in untouched, despite the misleading output.
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Ezequiel Garcia <ezequiel@collabora.com>
Tested-by: Ezequiel Garcia <ezequiel@collabora.com>
Makes the code shorter and easier to read.
Currently if the user has not set the crtc_id, we fetch the crtc yet do
not "bother" setting the id - do so.
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Ezequiel Garcia <ezequiel@collabora.com>
Tested-by: Ezequiel Garcia <ezequiel@collabora.com>
Let's make the code shorter, this avoid crashes (when drmModeGetCrtc()
fails) by using a couple of helpers. As get_resources() considers the
drmModeGetCrtc() fail non-fatal, we might as well handle it properly.
v2: Add a comment above the unreachable abort() (Eze)
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Ezequiel Garcia <ezequiel@collabora.com>
Tested-by: Ezequiel Garcia <ezequiel@collabora.com>
There's no point in keeping these around since we already fetch the
complete data set. Add respective count_ variables and greatly simplify
the existing code.
Extra brownie points for:
- using the inverse order in free_resources()
- don't memory leak the connector properties
- free the properties themselves, instead of only the objects
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Ezequiel Garcia <ezequiel@collabora.com>
Tested-by: Ezequiel Garcia <ezequiel@collabora.com>
Flesh out the bo_create + drmModeAddFB2 dance into a helper and use it.
Currently we're duplicating that in 4 places, many of which leaking et
al.
As a bonus point this highlights that the atomic_set_plane() seems tad
buggy. That'll be fixed with separate commit.
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Ezequiel Garcia <ezequiel@collabora.com>
Tested-by: Ezequiel Garcia <ezequiel@collabora.com>
The two functions have been stubs for ages. The alluded generic ioctls
never came to be, assumingly because all new drivers support those.
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Ezequiel Garcia <ezequiel@collabora.com>
Tested-by: Ezequiel Garcia <ezequiel@collabora.com>
Freedreno uses VALGRIND_ENABLE_ADDR_ERROR_REPORTING_IN_RANGE that was
introduced in Valgrind 3.10.0
Raspbian Buster includes Valgrind 3.7.0, so when valgrind is installed
as freedreno is build by default the build becomes broken. So lets
require 3.10 to enable valgrind when freedreno is built.
v2: Keep the arguments listed in the same order (Emil Velikov)
Closes: https://gitlab.freedesktop.org/mesa/drm/-/issues/37
Signed-off-by: Jose Maria Casanova Crespo <jmcasanova@igalia.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Since we now always returns the /dev/dri/ node for
drmGet<nodetype>DeviceNameFromFd, be consistant with the names returned
in drmGetDeviceNameFromFd.
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Emmanuel Vadot <manu@FreeBSD.org>
Fix the FreeBSD variant by getting the node type represented by fd to deduce
the target minor name.
We then return the full /dev/dri/<minorname><id> version.
Fix: #41
Fixes: 6818a50b12
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Emmanuel Vadot <manu@FreeBSD.org>
The sed was incorrectly modifying e.g. "nicholasbishop" to
"nicholasbop". The updated pattern will only match `.sh` at the end of
the string.
Signed-off-by: Nicholas Bishop <nicholasbishop@gmail.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Renoir is the same family as Raven, but it's with VCN2.0,
so it has to use VCN2.0 reg set
Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-By: Thong Thai <thong.thai@amd.com>
Currently the code expects that the device found at
/sys/char/$maj:$min/device for USB devices is a "usb_device". However,
at least for some devices, such as for the udl driver, they are instead
a "usb_interface".
A usb_interface is a child of the usb_device we're interested in, so we
walk up one in the /sys path to get there.
For example, with a USB device I have, trimmed to show the relevant
information:
```
$ udevadm info /dev/dri/card1
P: /devices/pci0000:00/0000:00:01.3/0000:02:00.0/usb1/1-4/1-4:1.0/drm/card1
E: DEVTYPE=drm_minor
$ udevadm info /sys/devices/pci0000:00/0000:00:01.3/0000:02:00.0/usb1/1-4/1-4:1.0
E: DEVTYPE=usb_interface
E: DRIVER=udl
$ udevadm info /sys/devices/pci0000:00/0000:00:01.3/0000:02:00.0/usb1/1-4
E: DEVTYPE=usb_device
E: DRIVER=usb
E: BUSNUM=001
E: DEVNUM=009
```
Signed-off-by: Scott Anderson <scott@anderso.nz>
This fixes bug in drmParseSubsystemType() that cases situation when
subsequent call to readlink() from get_subsystem_type() will result in
EACCESS.
Signed-off-by: Mikhail Golubev <mikhail.golubev@opensynergy.com>
Just like the other tests (tools really) install drmdevice. It is a
simple tool which is useful for basic check/testing.
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Acked-by: Eric Engestrom <eric@engestrom.ch>
The FreeBSD kernel expose a pseudo-device /dev/pci to obtain information
about present PCI device.
Uee the PCIOCGETCONF ioctl on this device to look up the desired device
information.
Signed-off-by: Emmanuel Vadot <manu@FreeBSD.org>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
The FreeBSD drm driver expose a sysctl hw.dri.%d.busid which contain
the busid.
Use this sysctl to parse the busid information based on the major/minor
that allow us to implement FreeBSD support for drmParsePciBusInfo.
Signed-off-by: Emmanuel Vadot <manu@FreeBSD.org>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
The FreeBSD variant of drmGetDeviceNameFromFd can already handle
the different node type so just call it.
Signed-off-by: Emmanuel Vadot <manu@FreeBSD.org>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
FreeBSD have some support for DRM on !PCI device but no code is currently
upstream. Default to PCI for now.
Signed-off-by: Emmanuel Vadot <manu@FreeBSD.org>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Use the FreeBSD variant function to obtain the minor name and the
device node.
Return the correct path based on where the node is (drm/ versus dri/).
Signed-off-by: Emmanuel Vadot <manu@FreeBSD.org>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Get the major/minor via fstat and after checking that this is a drm node
construct the full device node name using devname.
Note that we should be able to use fdevname to avoid calling fstat + devname
but for some reason it doesn't work on drm node (probably due to how the device
node are created in the linux compat code for drm on FreeBSD).
Signed-off-by: Emmanuel Vadot <manu@FreeBSD.org>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Resolve the minor type based on the device node path.
The minor type is either in /dev/drm/X where X is the type or
in a Linux-compatible device node in /dev/dri/
This means we need the major number on FreeBSD so add it to the function
arguments.
Signed-off-by: Emmanuel Vadot <manu@FreeBSD.org>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
FreeBSD devfs only provides on the fly generated major/minor.
The major number is irrelevant for FreeBSD so remove the special case.
Signed-off-by: Emmanuel Vadot <manu@FreeBSD.org>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
FreeBSD devfs have on the gly generated major minor so we cannot use them
to test if the device is a drm node.
Instead get the devfs node name and test if it is in a subdirectory "drm/"
or "dri/".
Historycally DRM device on FreeBSD are created in /dev/drm/ and link are
present in /dev/dri/ for compatibility reason.
Signed-off-by: Emmanuel Vadot <manu@FreeBSD.org>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Mesa expects to be able to map the same buffer, without unmapping it.
This leads to problem on long-running program.
On the other hand, libdrm uses cpu_map_count as a refcount and expects
its value to decrease so it can unmap buffers.
The previoulsy proprosed fix (https://patchwork.freedesktop.org/patch/258005/)
stopped increased the counter when it went past INT_MAX.
This commit instead proposes to use a larger type to store cpu_map_count.
The outcome is the same: long running apps will not crash, only the
implementation differs.
Closes: https://gitlab.freedesktop.org/mesa/mesa/issues/1423
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
xf86drmMode.h currently duplicates the connector-type definitions from
drm_mode.h. Add DRM_MODE_CONNECTOR_WRITEBACK, which is only visible
through a client cap, from drm_mode.h.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
Currently the code assumes that a virtio based device is always located
on the PCI bus.
Modify the parser to make it check the device's parent directory to
determine on which bus it is located.
Output for virtio-pci is the PCI bus.
Output for virtio-mmio is the Platform bus.
Signed-off-by: Vasyl Vavrychuk <vasyl.vavrychuk@opensynergy.com>
Signed-off-by: Mikhail Golubev <Mikhail.Golubev@opensynergy.com>
Reviewed-by: Gurchetan Singh <gurchetansingh@chromium.org>
Tested-by: Gurchetan Singh <gurchetansingh@chromium.org>
Suggested by Emil [1]:
> Feel free to drop the random test altogether. It's an old public API
> no active users (pretty ancient code uses it) and the in-tree users
> drmSL and drmHash already have respective tests.
This test takes minutes to run, while all the other tests combined take
barely more than a second.
Dropping it also helps the CI by avoiding random timeouts when `random`
takes more than the 4 minutes (!) we've allowed for it.
[1] https://gitlab.freedesktop.org/mesa/drm/merge_requests/26#note_390066
Suggested-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
This patch is to use generic variables as the input of amdgpu_cs_submit_raw2.
Because amdgpu_cs_submit_one won't handle IOCTL directly.
Signed-off-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
So far, amdgpu_cs_submit_raw2 is mainly used for upper layer (Mesa), however,
amdgpu_cs_submit is used for current all unit tests. Our intention is that the
unit tests can actually verify the API which is really used.
Signed-off-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Use alloca instead of malloc, then we don't need free them at the end of this
function.
Signed-off-by: Huang Rui <ray.huang@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Add a wrapper around the getfb2 ioctl, which returns extended
framebuffer information mirroring addfb2, including multiple planes and
modifiers.
Changes since v7:
- add new symbols to core-symbol.txt (Eric Engestrom)
Changes since v5:
- style change
Changes since v4:
- Set fb_id at init instead of memclear() and set (Eric Engestrom)
Changes since v3:
- remove unnecessary null check in drmModeFreeFB2 (Daniel Stone)
Changes since v2:
- getfb2 ioctl has been merged upstream
- sync include/drm/drm.h in a seperate patch
Changes since v1:
- functions should be drm_public
- modifier should be 64 bits
- update ioctl number
Signed-off-by: Juston Li <juston.li@intel.com>
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
a) delta: Adds DRM_IOCTL_MODE_GETFB2
b) Generated using make headers_install
c) Taken from drm-next-misc:
commit 3ff4c24bdb1f494c217c80348f9db4896043ed81
Author: Lyude Paul <lyude@redhat.com>
Date: Fri Jan 17 17:47:48 2020 -0500
drm/dp_mst: Fix indenting in drm_dp_mst_topology_mgr_set_mst()
Signed-off-by: Juston Li <juston.li@intel.com>
Acked-by: Eric Engestrom <eric@engestrom.ch>
Platforms without a HW detiler doesn't support the get_tiling IOCTL.
Fix the drm_intel_bo_gem_create_from_* functions assuming the default
no-tiling, no-swizzling setting for the GEM buffer in this case.
v2:
- Add the missing gem handle IOCTL parameter. (Eric)
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Juha-Pekka Heikkila <juhapekka.heikkila@gmail.com>
When a mode is set with just a connector "-s foo",
we get a nasty segmentation fault. Fix it.
Signed-off-by: Ezequiel Garcia <ezequiel@collabora.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
This patch fixes the following warning:
-Wformat-overflow=
v2: Use the correct strlcat(3).
v3: Use strncat(3) and remove libbsd dependency.
Signed-off-by: Luben Tuikov <luben.tuikov@amd.com>
This patch fixes the following warnings:
-Wformat=
-Wmaybe-uninitialized
-Wmisleading-indentation
-Wstringop-truncation
-Wunused-function
-Wunused-variable
It also removes forward declarations and moves
global functions to the bottom, keeping locals
at the top, in ras_tests.c.
v2: Fix compilation.
Signed-off-by: Luben Tuikov <luben.tuikov@amd.com>
Proper format for command line option "-l",
listing the supported and unsupported tests:
1) Add an aligned column header.
2) Align all fields into columns.
3) Fixed length fields, come before the last
column, which is a variable length field.
4) Variable length field, which is the name of the
test, goes in the last column.
5) If a suite is disabled, do not iterate over its
tests, as they'd naturally be all disabled.
Now the output looks like this:
$sudo ./amdgpu_test -l
What: ID: Status: Name
Suite: 1: ENABLED: Basic Tests
Test: 1: ENABLED: Query Info Test
Test: 2: ENABLED: Userptr Test
Test: 3: DISABLED: bo eviction Test
Test: 4: ENABLED: Command submission Test (GFX)
Test: 5: ENABLED: Command submission Test (Compute)
Test: 6: ENABLED: Command submission Test (Multi-Fence)
Test: 7: ENABLED: Command submission Test (SDMA)
Test: 8: ENABLED: SW semaphore Test
Test: 9: DISABLED: Sync dependency Test
Test: 10: DISABLED: Dispatch Test (Compute)
Test: 11: DISABLED: Dispatch Test (GFX)
Test: 12: DISABLED: Draw Test
Test: 13: DISABLED: GPU reset Test
Suite: 2: ENABLED: BO Tests
Test: 1: ENABLED: Export/Import
Test: 2: DISABLED: Metadata
Test: 3: ENABLED: CPU map/unmap
Test: 4: ENABLED: Memory alloc Test
Test: 5: ENABLED: Memory fail alloc Test
Test: 6: ENABLED: Find bo by CPU mapping
Suite: 3: DISABLED: CS Tests
Suite: 4: DISABLED: VCE Tests
Suite: 5: ENABLED: VCN Tests
Test: 1: ENABLED: VCN DEC create
Test: 2: ENABLED: VCN DEC decode
Test: 3: ENABLED: VCN DEC destroy
Test: 4: ENABLED: VCN ENC create
Test: 5: ENABLED: VCN ENC decode
Test: 6: ENABLED: VCN ENC destroy
Suite: 6: DISABLED: UVD ENC Tests
Suite: 7: DISABLED: Deadlock Tests
Suite: 8: ENABLED: VM Tests
Test: 1: ENABLED: resere vmid test
Test: 2: ENABLED: unaligned map
Test: 3: ENABLED: vm mapping test
Suite: 9: DISABLED: RAS Tests
Suite: 10: ENABLED: SYNCOBJ TIMELINE Tests
Test: 1: ENABLED: syncobj timeline test
$_
Signed-off-by: Luben Tuikov <luben.tuikov@amd.com>
This patch fixes the following warning:
-Wformat-overflow=
v2: Use the correct strlcat(3).
v3: Use strncat(3) and remove libbsd dependency.
Signed-off-by: Luben Tuikov <luben.tuikov@amd.com>
This patch fixes the following warnings:
-Wformat=
-Wmaybe-uninitialized
-Wmisleading-indentation
-Wstringop-truncation
-Wunused-function
-Wunused-variable
It also removes forward declarations and moves
global functions to the bottom, keeping locals
at the top, in ras_tests.c.
Signed-off-by: Luben Tuikov <luben.tuikov@amd.com>
The computer ring test name mis-match in different files,
thus may be set with TRUE on wrong platforms.
Change-Id: I0b918ff8faf08c9c9f1ad55f4dcd18f66b956901
Signed-off-by: Yifan Zhang <yifan1.zhang@amd.com>
Signed-off-by: Marek Olšák <marek.olsak@amd.com>
This fixes an issue with libdrm failing to build when used as a meson
subproject. Using 'config.h' directly will cause it to possibly refer to
the wrong file.
By using `@0@.format(config_file)`, it will be transformed into the
correct relative path, e.g. `./config.h` in normal build,
`./subprojects/libdrm/config.h` in subproject build.
Signed-off-by: Scott Anderson <scott@anderso.nz>
Acked-by: Eric Engestrom <eric.engestrom@intel.com>
Add function to derive floating value of vertical
refresh rate from drm mode using pixel clock,
horizontal total size and vertical total size.
Use this function to find suitable mode having vrefresh
value which is matching with user provided vrefresh value.
If user doesn't provide any vrefresh value in args then
update vertical refresh rate value in pipe args using this
function.
Also use this function for printing floating vrefresh while
dumping all available modes.
This will give more accurate picture to user for available modes
differentiated by floating vertical refresh rate and help user
select more appropriate mode using suitable refresh rate value.
V4:
1) While setting mode, print mode name and vrefresh using struct
drmModeModeInfo instead of struct pipe_args.
2) Revert back to using a float value instead of float *
for vrefresh arg in connector_find_mode().
V3:
1) Change name of function used to derive refresh rate.
V2:
1) Don't use inline function for deriving refresh rate from mode.
2) If requested mode not found, print refresh rate only
if user had provided it in args.
Signed-off-by: Devarsh Thakkar <devarsh.thakkar@xilinx.com>
Reviewed-by: Neil Armstrong <narmstrong@baylibre.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
It can run dispatch/draw tests on new renoir chips. So it needs to
enable dispatch/draw tests for Renoir again.
Change-Id: I3a72a4bbfe0fc663ee0e3e58d8e9c304f513e568
Signed-off-by: changzhu <Changfeng.Zhu@amd.com>
Reviewed-by: Flora Cui <flora.cui@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Huang Rui <ray.huang@amd.com>
Since AO_compare_and_swap_full() is used by libdrm, AO_REQUIRE_CAS
must be defined before including <atomic_ops.h> so that we are sure
that CAS support will be provided. This is necessary to make sure that
the AO_compare_and_swap_full() function will be provided on all
architectures, including the ones that don't have built-in CAS support
such as SPARCv8.
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Seiderer <ps.report@gmx.net>
Acked-by: Eric Engestrom <eric.engestrom@intel.com>
This provides support for Xorg interface. Without this the vivante
samples will hang during close requiring a reboot
[Adapted from yocto project]
Upstream-Status: Pending
Signed-off-by: Lauren Post <lauren.post@freescale.com>
Signed-off-by: Evan Kotara <evan.kotara@freescale.com>
[Thomas: change CAS code to only be used on ARMv6/ARMv7, and not
ARMv4/ARMv5, which don't support ldrex/strex. If no CAS implementation
is provided libdrm falls back to a system call for locking/unlocking.]
Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Peter Seiderer <ps.report@gmx.net>
Acked-by: Eric Engestrom <eric.engestrom@intel.com>
---
Changes v1 -> v2:
- add comment explaining exclusion of ARMv4/ARMv5 and lower
Use the stronger compiler.link() test (instead of the weaker
compiler.compile()) to fix the intel atomics detection.
Fixes false positive in case of sparc compile (buildroot toolchain).
Signed-off-by: Peter Seiderer <ps.report@gmx.net>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>
Userptr Test will fail on PAGE_SIZE bigger than BUFFER_SIZE(8 * 1024)
Signed-off-by: xinxu <xinxu@loongson.cn>
(cherry picked from commit fb7dfdc5fb58795365b70117c3eb625f2edb8f06)
For the scenario where user may require to modeset with a mode
supporting a fractional value for vertical refresh-rate,
appropriate mode can be selected by searching for mode
having matching fractional vertical refresh rate using
below equation.
vrefresh = (1000 * pixel clock) / (htotal * vtotal) Hz.
We do this way since driver doesn't return float value of vrefresh
as it use int for vrefresh in struct drm_mode_info, but we can derive
the actual value using pixel clock, horizontal total size and
vertical total size values.
So for e.g. if user want to select mode having 59.94 Hz as refresh rate
then with this patch it be can done as shown in below command,
given there is an appropriate mode is available :
modetest -M xlnx -s 39:1920x1080-59.94@BG24 -v
NOTE: Above command was tested on xilinx DRM driver with DP
monitor which was supporting mode having 59.94 Hz refresh rate.
V2: Update commit message
V3: Update with below changes as per review comments :
1) Use epsilon for vrefresh comparison
2) Use implicit type-casting wherever possible
V4: Keep patch version history on main commit message
Signed-off-by: Devarsh Thakkar <devarsh.thakkar@xilinx.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Including xf86drmMode.h results in undefined references to uint32_t
and ssize_t. Include the stdlib headers that define them to allow the
file to be included without xf86drm.h.
Signed-off-by: Simon Ser <contact@emersion.fr>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>
a) delta: drm: use pad as flags in drm_syncobj_timeline_array.
b) Generated using make headers_install.
c) Generated from origin/drm-misc-next commit 949561eb85bcee10248e7da51d44a0325d5e0d1b"
Signed-off-by: Chunming Zhou <david1.zhou@amd.com>
Shaves about 50% off the build time on both debian and arch builds.
(yeah, I know, it's very small anyway compared to mesa, but we might
want to add more things in the future)
This also makes the build no longer dependent on external websites:
once the image is build, only fdo-internal services are used.
Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Michel Dänzer <mdaenzer@redhat.com>
FreeBSD requires sys/types.h for sys/sysctl.h, add it as part of the
includes when checking for headers.
Instead of splitting out the check for sys/sysctl.h from the other
header checks, just add sys/types.h to all header checks.
v2 [Emil]
- add inline comment
- drop bash/sh hunk
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Niclas Zeising <zeising@daemonic.se>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>
On Solaris, sys/sysmacros.h has long-deprecated copies of major() & minor()
but not makedev().
sys/mkdev.h has all three and is the preferred choice.
Let's make sure we check for all 3 major(), minor() and makedev().
Fixes build failure with error:
../xf86drm.c: In function ‘drmOpenMinor’:
../xf86drm.c:454:30: error: implicit declaration of function ‘makedev’ [-Werror=implicit-function-declaration]
454 | return drmOpenDevice(makedev(DRM_MAJOR, minor), minor, type);
| ^~~~~~~
Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Tested-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Add the new CML PCI IDS.
Align with kernel commit:
bfc4c359b2822 ("drm/i915/cml: Add Missing PCI IDs")
This is in sync with kernel header as of:
0747590267e7 ("drm-tip: 2019y-08m-30d-18h-03m-18s UTC integration manifest")
Cc: José Roberto de Souza <jose.souza@intel.com>
Signed-off-by: Anusha Srivatsa <anusha.srivatsa@intel.com>
This configuration file will be picked up when
running gfx ras inject tests by amdgpu_test tool.
For the time being, only add those tests that are
successfully trafficked. In addition, this file
can also be modified by user to add or delete ras
inject unit tests for different IP blocks/subblocks.
Signed-off-by: Dennis Li <dennis.li@amd.com>
Signed-off-by: Guchun Chen <guchun.chen@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
for better clarification
v2: accordingly change dispatch_test caller in gpu_reset test
Signed-off-by: Le Ma <le.ma@amd.com>
Reviewed-by: Flora Cui <flora.cui@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Straight copy from the kernel file, aligned with drm-intel-next-queued
commit cb823ed9915b ("drm/i915/gt: Use intel_gt as the primary object
for handling resets")
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: José Roberto de Souza <jose.souza@intel.com>
1. perform gpu reset
2. perform dispatch test to verify gpu reset to a good state
Signed-off-by: Flora Cui <flora.cui@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
drm render nodes have the same major as drm primary devices but offset
the minor by a base of 128.
I expected the name of the device to have numbering starting at 0 when
these non-linux codepaths were added (before OpenBSD had render nodes).
Signed-off-by: Jonathan Gray <jsg@jsg.id.au>
Acked-by: Eric Engestrom <eric.engestrom@intel.com>
Unlike Linux the OpenBSD primary "drm" device name is substring of the
"drmR" render node device name and strncmp() tests resulted in render
nodes being flagged as primary nodes.
Signed-off-by: Jonathan Gray <jsg@jsg.id.au>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>
Acked-by: Eric Engestrom <eric.engestrom@intel.com>
Simplifies its callers.
dev->bo_table_mutex is now always held when amdgpu_bo_create is called
(this was already the case in amdgpu_bo_import).
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
And propagate drmIoctl's return value.
This allows replacing all remaining open-coded DRM_IOCTL_GEM_CLOSE
ioctl calls with amdgpu_close_kms_handle calls.
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
The following situation can happen in a multithreaded OpenGL application.
A BO is submitted from etna_cmd_stream #1 with flags set for read.
A BO is submitted from etna_cmd_stream #2 with flags set for write.
This triggers a flush on stream #1 and clears the BO's current_stream
pointer. If at this point, stream #2 attempts to queue BO again, which
does happen, the BO will be added to the submit list twice. The Linux
kernel driver correctly detects this and warns about it with "BO at
index %u already on submit list" kernel message.
However, when cleaning the BO cache in etna_bo_cache_free(), the BO
which was submitted twice will also be free()d twice, this triggering
a glibc double free detector.
The fix is easy, even if the BO does not have current_stream set,
iterate over current streams' list of BOs before adding the BO to it
and verify that the BO is not yet there.
Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Christian Gmeiner <christian.gmeiner@gmail.com>
Cc: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Reviewed-by: Wladimir J. van der Laan <laanwj@gmail.com>
Avoids compiler warning:
../../amdgpu/amdgpu_cs.c: In function 'amdgpu_cs_ctx_override_priority':
../../amdgpu/amdgpu_cs.c:155:2: warning: ISO C90 forbids mixed declarations and code [-Wdeclaration-after-statement]
union drm_amdgpu_sched args;
^~~~~
Reviewed-by: Christian König <christian.koenig@amd.com>
There is no implementation and also no users, so there is no point
in keeping it in the API.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
As new features are added and others are declared to be legacy, it's
nice to be able to implement fallbacks. As such, create a
property-setting variant that does not generate errors which can very
well be entirely expected.
Will be used for gamma control in a future change.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
The idea is to have a horizontal pattern split into two with the top and
bottom halves having different precision. This allows one to see whether
10bpc support is working properly or not, as there are many pieces to
the puzzle beyond the basic format support (gamma ramps, bpc encodings,
etc).
This is really only useful on 10bpc formats, but we also add support for
8bpc formats to ease testing. In the future, this could be applied to
16bpc formats as well.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
We need to shift the values up, otherwise we'd end up with a negative
shift. This works for up-to 16-bit components, which is fine for now.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
This also adds a helper to generate a color LUT, which has to be used in
conjunction with the C8 indexed format.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Acked-by: Emil Velikov <emil.velikov@collabora.com>
Feature is controlled by DRM_CAP_SYNCOBJ_TIMELINE drm capability.
Signed-off-by: Chunming Zhou <david1.zhou@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
a) delta: only DRM_CAP_SYNCOBJ_TIMELINE
b) Generated using make headers_install.
c) Generated from origin/drm-misc-next commit 982c0500fd1a8012c31d3c9dd8de285129904656"
Signed-off-by: Chunming Zhou <david1.zhou@amd.com>
Suggested-by: Michel Dänzer <michel@daenzer.net>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
v2: drop DRM_SYNCOBJ_CREATE_TYPE_TIMELINE, fix timeout calculation,
fix some warnings
v3: add export/import and cpu signal testing cases
Signed-off-by: Chunming Zhou <david1.zhou@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Acked-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
In drmModeGetPropertyPtr(), from upper error path, it calls free
but with just next error path, it does not call. Fix the possible
memory leak.
Signed-off-by: Seung-Woo Kim <sw0312.kim@samsung.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>
The pointer p aquired with drmModeGetPlane() is not free in error
path. Fix possible memory leak by calling drmModeFreePlane() in
the error path.
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Signed-off-by: Seung-Woo Kim <sw0312.kim@samsung.com>
When calling drmModeAtomicAddProperty allocation of memory
happens as needed in increments of 16 elements. This can be very
slow if there are multiple properties to be updated in an Atomic
Commit call.
Increase this to as many as can fit in a memory PAGE to avoid
having to reallocate memory too often.
Also this patch has a small one line perf tweak in
drmModeAtomicDuplicate() to only memcpy items to the cursor
position in order avoid copying the entire item array if its
mostly empty.
Cc: Sean Paul <seanpaul@chromium.org>
Cc: Alistair Strachan <astrachan@google.com>
Cc: Marissa Wall <marissaw@google.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
[jstultz: Expanded commit message]
Signed-off-by: John Stultz <john.stultz@linaro.org>
This way we can override the priority of a single context using a
master fd.
Since we cannot usefully create an amdgpu device of a master fd
without the fd deduplication kicking in this takes a plain fd.
This can be used by e.g. radv to get high priority contexts using
a master fd from the primary node or a lease.
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Andres Rodriguez <andresx7@gmail.com>
Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
This makes the test utilities work with the Armada driver without the
necessity of using the -M argument.
Signed-off-by: Lubomir Rintel <lkundrak@v3.sk>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
This reverts commit 56c21f877b.
There were issues pointed out during review that were not addressed.
Would love to have this re-land, once those are addressed.
Generated using make headers_install from the drm-next
tree - git://anongit.freedesktop.org/drm/drm
branch - drm-next
commit - 14d2bd53a47a7e1cb3e03d00a6b952734cf90f3f
The changes were as follows :-
core: (drm.h, drm_fourcc.h, drm_mode.h)
- Added 'struct drm_syncobj_transfer', 'struct drm_syncobj_timeline_wait' and 'struct drm_syncobj_timeline_array'
- Added various DRM_IOCTL_SYNCOBJ_ ioctls
- Added some new RGB and YUV formats
- Added 'DRM_FORMAT_MOD_VENDOR_ALLWINNER'
- Added 'SAMSUNG' and Arm's 'AFBC' and 'ALLWINNER' format modifiers
- Added 'struct drm_mode_rect'
i915:
- Added struct 'struct i915_user_extension' and various 'struct drm_i915_gem_context_'
- Added different modes of per-process Graphics Translation Table
Changes from v1:-
- Removed the changes to 'msm_drm.h' as it breaks the build for 'freedreno' platform.
Signed-off-by: Ayan Kumar halder <ayan.halder@arm.com>
Acked-by: Eric Engestrom <eric.engestrom@intel.com>
From drm-next commit b4e4538a0ab5079ae5dc401970e11f0ff2ba13a7
Adds support for:
- RAS queries
- context priority updates
- CS chunks support for scheduled dependencies
- IB flag for GDS max wave id
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Generated using make headers_install.
This brings in the in/out fence support for explicit
synchronization.
v2: don't use experimental kernel branch
Signed-off-by: Gurchetan Singh <gurchetansingh@chromium.org>
Reviewed-by: Robert Foss <robert.foss@collabora.com>
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Add CML and EHL PCI IDs, and one more for ICL. This is in sync with
kernel header as of b024ab9b2d3a ("drm/i915/bios: iterate over child
devices to initialize ddi_port_info")
Signed-off-by: Anusha Srivatsa <anusha.srivatsa@intel.com>
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
deadlock test for sdma will cause gpu recoverty.
disable the test for now until GPU reset recovery could survive at least
1000 times test.
v2: add modprobe parameter
Reviewed-and-tested-by: Evan Quan <evan.quan@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Flora Cui <flora.cui@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This fixes a bug, which was introduced with commit ee798b98
"xf85drm: de-duplicate drmParse{Platform.Host1x}{Bus,Device}Info".
where accessing *compatible[i] with i>0 results in a segfault.
Signed-off-by: Andreas Baierl <ichgeh@imkreisrum.de>
Fixes: ee798b9847 "xf85drm: de-duplicate drmParse{Platform.Host1x}{Bus,Device}Info"
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>
[Eric: add the same fix to the free() below]
Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
After the commit 0926f0af54 ("meson,configure: include config.h
automatically"), there is build error for autotools because
config.h is not included. Fix the error by adding "-include
config.h" to CPPFLAGS instead of CFLAGS from configure.ac.
Reference: https://bugs.freedesktop.org/show_bug.cgi?id=106561
Signed-off-by: Seung-Woo Kim <sw0312.kim@samsung.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>
Fixes: 0926f0af54 "meson,configure: include config.h automatically"
For multiple GPUs which has the same BDF, but has different domain ID,
the drmOpenByBusid will return the wrong fd when startx.
The reproduce sequence as below:
1. Call drmOpenByBusid to open Card0, then will return the right fd0, and the
fd0 is master privilege;
2. Call drmOpenByBusid to open Card1. In function drmOpenByBusid, it will
open Card0 first, this time, the fd1 for opening Card0 is not master
privilege, and will call drmSetInterfaceVersion to identify the
domain ID feature, as the fd1 is not master privilege, then drmSetInterfaceVersion
will fail, and then won't compare domain ID, then return the wrong fd for Card1.
Solution:
First loop search the best match fd about drm 1.4.
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Emily Deng <Emily.Deng@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
base/archlinux has been replaced with archlinux/base, which is
maintained directly by the archlinux community.
Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
This reverts 6d2379857b "xf86atomic: #undef internal define"
and b541d21a0a "freedreno: remove always-defined #ifdef".
I didn't realise at the time that freedreno/freedreno_ringbuffer.h gets
installed, and then used by Mesa for instance. These two commits were
fine in libdrm, but broke Mesa which needs to use struct fd_ringbuffer
but doesn't need to access ::refcnt. The hack that I removed serves to
keep the struct at the correct size while only exposing the ::refcnt
member within libdrm.
Fixes: 6d2379857b "xf86atomic: #undef internal define"
Fixes: b541d21a0a "freedreno: remove always-defined #ifdef"
Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
We can't use drmSetMaster to query whether or not a drm fd is master
because it requires CAP_SYS_ADMIN, even if the fd *is* a master fd.
Pick DRM_IOCTL_MODE_ATTACHMODE as a long-deprecated ioctl that is
DRM_MASTER but not DRM_ROOT_ONLY as the probe by which we can detect
whether or not the fd is master.
This is useful for code that might get master by open()ing the drm device
while no other master exists, but can't call drmSetMaster itself because
it's not running as root or is in a container, where container-root isn't
real-root.
v2: Use the AUTH_MAGIC request rather than MODE_ATTACHMODE, as it's more
clearly related to master status.
v3: [Emil] Don't expose internals, check for -EACCES.
Signed-off-by: Christopher James Halse Rogers <christopher.halse.rogers@canonical.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch> (v2)
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Straight copy from the kernel file.
Add more PCI Device IDs for Coffee Lake, Ice Lake,
and Amber Lake. It also include a reorg on Whiskey Lake IDs.
Align with kernel commits:
5e0f5a58b167 ("drm/i915/cfl: Adding another PCI Device ID.")
03ca3cf8e9aa ("drm/i915/icl: Adding few more device IDs for Ice Lake")
c0c46ca461f1 ("drm/i915/aml: Add new Amber Lake PCI ID")
c1c8f6fa731b ("drm/i915: Redefine some Whiskey Lake SKUs")
Cc: José Roberto de Souza <jose.souza@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Acked-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Some devices can lack OF data or it may not be available in the uevent
file. Fallback to the MODALIAS data in those cases.
We strip any leading "MODALIAS=.*:" thus the resulting information is
compatible with existing code in Mesa.
v2: foo -> tmp_name
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Tested-by: Lucas Stach <l.stach@pengutronix.de> (v1)
Also, move the sentence about "who would use libdrm" into its own paragraph,
as it is something people discovering libdrm will want to know.
Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Thanks to the #error just above, any file including this header can only
see one state for this macro: defined, with the value `1`.
Let's just #undef it once we're done using it in here so that other
files don't misconstrue any meaning to it.
Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
While at it, let's include xf86atomic.h explicitly, instead of relying
on some other file accidentally including it before including this file.
Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Seemingly the 64-bit int is always aligned to 8 in LP64.
But this is not hold in LP32.
Consequently sizeof(gralloc_drm_handle_t) are different
between LP64 (which is 18 ints) and LP32 (which is 16 ints).
As a result, 32-bit apps will crash in 64-bit OS since the
checking handle->base.numInts != GRALLOC_GBM_HANDLE_NUM_INTS
is true.
Fix it by always aligning 64-bit int to 8. Besides, to avoid
additional padding, just exchange the order of data_owner
and modifier. It aligns modifier to 8 natually.
This makes gralloc_drm_handle_t fit in 16 ints perfectly.
(v2) gralloc_drm_handle.h patch now applied in gralloc_handle.h
and GRALLOC_HANDLE_VERSION updated to 4
Reported-by: Mauro Rossi <issor.oruam@gmail.com>
Signed-off-by: Chih-Wei Huang <cwhuang@linux.org.tw>
Reviewed-by: Robert Foss <robert.foss@collabora.com>
This is the directory used by meson/autotools (at least in the
.gitlab-ci configuration) so ignore the whole dir.
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Acked-by: Emil Velikov <emil.velikov@collabora.com>
LANG=C sort -u .gitignore | sponge .gitignore
This way it's easier to keep track of the entries.
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Acked-by: Emil Velikov <emil.velikov@collabora.com>
It is a cleaner and less fragile way to get PCI IDs than the one
currently used by local DPorts patches.
Signed-off-by: François Tigeot <ftigeot@wolfpond.org>
Make sure the kernel doesn't crash if we map something at the minimum/maximum address.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com>
For ARM systems with tinydrm displays attached to SPI, the bus name is
/spi but we have platform device info for the rest. Fixes
eglInitialize() failures on hx8357d since the EGL_EXT_device_drm
changes.
Acked-by: Eric Engestrom <eric.engestrom@intel.com>
Error message was invalid too, negative values aren't the number of
devices, they're errno error codes.
Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Since only for those ASICs gpu reset is enabled by deafult.
Also update disable message and fix identation .
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Illegal access will cause CP hang followed by job timeout and
recovery kicking in.
Also, disable the suite for all APU ASICs until GPU
reset issues for them will be resolved and GPU reset recovery
will be enabled by default.
v2:
Add KV to deasbled APUs list and add comments regarding
necessary kernel amdgpu paramteres to run the tests.
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
With streaming stateobjs to upload uniforms, the submit.cmds table gets
much larger, and iterating over it for each ring to ring reloc starts
getting expensive.
TODO if we have flag to pass when constructing parent rb, we could
avoid dynamically allocating this and bo_table in get_cmd() or bo2idx
Signed-off-by: Rob Clark <robclark@freedesktop.org>
The msm_cmd isn't refcount'd, so with stateobj rb's that have
independent lifecycle, this is no longer a safe thing to do.
Really, now that there is a bo-cache for rb's, fd_ringbuffer_reset()
should be deprecated because it adds a bunch of pointless complexity.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
It's usage in mesa was removed more than two years ago. And it stands
in the way of some optimizations needed to reduce the overhead of hw
stateobjs (ie. CP_SET_DRAW_STATE, where the # of cmds in the submit
ioctl goes up significantly).
Signed-off-by: Rob Clark <robclark@freedesktop.org>
For now, we want a way for gallium to be able to provide hints for the
upcoming rb suballocation. But could be useful for other things down
the road.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Seems like AI and RV requires uncashed memory mapping to be able
to pickup value written to memory by CPU after the WAIT_REG_MEM
command was already launched.
.
Enable the test for AI and RV.
v2:
Update commit description.
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
v2:
Call amdgpu_bo_va_op_raw directly in amdgpu_bo_alloc_and_map_raw
Move amdgpu_bo_alloc_and_map_raw into C file to avoid including
unistd.h in amdgpu_test.h
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
To determine whether a device node is a drm device node or not, the code
currently compares the node's major number to the static drm major device
number.
This breaks the standalone vmwgfx driver on XWayland dri clients,
https://cgit.freedesktop.org/mesa/vmwgfx
and any future attempt to introduce dynamic device numbers for drm.
So instead of checking for the device major, instead check for the presence
of the /sys/dev/char/<major>:<minor>/device/drm directory.
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>
Acked-by: Dave Airlie <airlied@redhat.com>
Generated using make headers_install from the drm-next
tree - git://anongit.freedesktop.org/drm/drm
branch - drm-next
commit - 2dc7bad71cd310dc94d1c9907909324dd2b0618f
The changes were as follows :-
core: (drm.h, drm_fourcc.h, drm_mode.h)
- Added client capabilities for ASPECT_RATIO and WRITEBACK_CONNECTORS
- Added Arm AFBC modifiers
- Added BROADCOM's SAND and UIF modifiers
- Added Qualcomm's modifiers
- Added some picture aspect ratio and content type options
- Added some drm mode flags
- Added writeback connector id
amdgpu:
- Added GEM domain mask
- Added some GEM flags
- Added some hardware ip flags
- Added chunk id and IB fence.
- Added some query ids
i915:
-Added an IOCTL (I915_PARAM_MMAP_GTT_COHERENT)
qxl:
- Minor changes
tegra:
- Added some comments about struct drm_tegra* members
- Modified DRM_IOCTL_TEGRA_CLOSE_CHANNEL
vc4:
- Added some members for 'struct drm_vc4_submit_cl'
Changes in v2:
- Mentioned 'libdrm' in the commit header.
Changes in v3:
- Removed the changes to radeon_drm.h, sis_drm.h and via_drm.h as suggested by
Emil Velikov <emil.l.velikov@gmail.com>
Changes in v4:
- Removed the changes to vmwgfx_drm.h as it caused a build break ie
'make check' failed.
Change-Id: I018a06f65bf4a6a68400ab252b9cd05d041299b3
Signed-off-by: Ayan Kumar halder <ayan.halder@arm.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Use libutil to lookup connector type names and state. This also
makes sure that the latest connector type addition "DPI" gets
printed correctly.
Signed-off-by: Stefan Agner <stefan@agner.ch>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Now that symbols that should be exported are annotated accordingly, make
all the rest hidden by default.
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Acked-by: Eric Engestrom <eric.engestrom@intel.com>
while read sym; do
read f func line _ <<<$(cscope -d -L -1 $sym)
if [ ! -z "$f" ]; then
sed -i "${line}s/^/drm_public /" $f
fi
done < /tmp/a.txt
In which /tmp/a.txt contains the public symbols from
exynos-symbol-check. The idea here will be to switch the default
visibility to hidden so we don't export symbols we shouldn't.
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Acked-by: Eric Engestrom <eric.engestrom@intel.com>
while read sym; do
read f func line _ <<<$(cscope -d -L -1 $sym)
if [ ! -z "$f" ]; then
sed -i "${line}s/^/drm_public /" $f
fi
done < /tmp/a.txt
In which /tmp/a.txt contains the public symbols from
tegra-symbol-check. The idea here will be to switch the default
visibility to hidden so we don't export symbols we shouldn't.
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Acked-by: Eric Engestrom <eric.engestrom@intel.com>
while read sym; do
read f func line _ <<<$(cscope -d -L -1 $sym)
if [ ! -z "$f" ]; then
sed -i "${line}s/^/drm_public /" $f
fi
done < /tmp/a.txt
In which /tmp/a.txt contains the public symbols from
radeon-symbol-check. The idea here will be to switch the default
visibility to hidden so we don't export symbols we shouldn't.
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Acked-by: Eric Engestrom <eric.engestrom@intel.com>
while read sym; do
read f func line _ <<<$(cscope -d -L -1 $sym)
if [ ! -z "$f" ]; then
sed -i "${line}s/^/drm_public /" $f
fi
done < /tmp/a.txt
In which /tmp/a.txt contains the public symbols from
omap-symbol-check. The idea here will be to switch the default
visibility to hidden so we don't export symbols we shouldn't.
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Acked-by: Eric Engestrom <eric.engestrom@intel.com>
while read sym; do
read f func line _ <<<$(cscope -d -L -1 $sym)
if [ ! -z "$f" ]; then
sed -i "${line}s/^/drm_public /" $f
fi
done < /tmp/a.txt
In which /tmp/a.txt contains the public symbols from
freedreno-symbol-check. The idea here will be to switch the default
visibility to hidden so we don't export symbols we shouldn't.
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Acked-by: Eric Engestrom <eric.engestrom@intel.com>
while read sym; do
read f func line _ <<<$(cscope -d -L -1 $sym)
if [ ! -z "$f" ]; then
sed -i "${line}s/^/drm_public /" $f
fi
done < /tmp/a.txt
In which /tmp/a.txt contains the public symbols from
etnaviv-symbol-check. The idea here will be to switch the default
visibility to hidden so we don't export symbols we shouldn't.
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Acked-by: Eric Engestrom <eric.engestrom@intel.com>
This was done with:
nm --dynamic --defined-only build/libdrm.so | \
grep " T " | \
grep -v _fini | grep -v _init | \
cut -d' ' -f3 > /tmp/a.txt
while read sym; do
read f func line _ <<<$(cscope -d -L -1 $sym)
if [ ! -z "$f" ]; then
sed -i "${line}s/^/drm_public /" $f
fi
done < /tmp/a.txt
Then the alignment of function arguments were manually fixed all over.
The idea here will be to switch the default visibility to hidden so we
don't export symbols we shouldn't.
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>
This was done with:
nm --dynamic --defined-only build/amdgpu/libdrm_amdgpu.so | \
grep amdgpu_ | \
cut -d' ' -f3 > /tmp/a.txt
while read sym; do
read f func line _ <<<$(cscope -d -L -1 $sym)
if [ ! -z "$f" ]; then
line=$((line-1))
sed -i "${line}s/^/drm_public /" $f
fi
done < /tmp/a.txt
Then the alignment of function arguments were manually fixed all over.
The idea here will be to switch the default visibility to hidden so we
don't export symbols we shouldn't.
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Acked-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
This was done with:
nm --dynamic --defined-only build/nouveau/libdrm_nouveau.so | \
grep nouveau_ | \
cut -d ' ' -f3 > /tmp/a.txt
while read sym; do
read f func line _ <<<$(cscope -d -L -1 $sym)
if [ ! -z "$f" ]; then
line=$((line-1))
sed -i "${line}s/^/drm_public /" $f
fi
done < /tmp/a.txt
Then some corner cases were manually fixed. The idea here will be to
switch the default visibility to hidden so we don't export symbols we
shouldn't.
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Acked-by: Eric Engestrom <eric.engestrom@intel.com>
This was done with:
nm --dynamic --defined-only build/libkms/libkms.so | \
grep kms_ | \
cut -d' ' -f3 > /tmp/a.txt
while read sym; do
read f func line _ <<<$(cscope -d -L -1 $sym)
if [ ! -z "$f" ]; then
sed -i "${line}s/^/drm_public /" $f
fi
done < /tmp/a.txt
The idea here will be to switch the default visibility to hidden so we
don't export symbols we shouldn't.
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Acked-by: Eric Engestrom <eric.engestrom@intel.com>
This was done with:
while read sym; do
read f func line _ <<<$(cscope -d -L -1 $sym)
if [ ! -z "$f" ]; then
line=$((line-1))
sed -i "${line}s/^/drm_public /" $f
fi
done < /tmp/a.txt
Then some corner cases were manually fixed. "a.txt" above contains the
symbols collected from intel/intel-symbol-check. The idea here will be
to switch the default visibility to hidden so we don't export symbols we
shouldn't.
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>
Otherwise the calling function won't notice that something is wrong.
v2: check map result as well
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
The heap is checked by the kernel and not libdrm, to make it even worse
it prevented allocating resources other than VRAM and GTT.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
We don't need to call IS_GEN() for each gen >= 9: we can rather use the
new intel_is_genx() helper to iterate the pciids array once.
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
The 2 PCI IDs that are used for the command line overrid mechanism
were left defined. The rest can be gone and then we just use the kernel
defines.
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
This will allow platforms to reuse kernel IDs instead of manually
keeping them in sync. In most of the cases we only need to extend
IS_9XX(). Current platforms that fit this requirement can be ported
over to use this macro. Right now it's a nop since it doesn't have any
PCI ID added.
The i915_pciids.h header is in sync with kernel tree on
drm-tip 2018y-08m-20d-21h-41m-11s.
v2: - move to a separate .c so we can have the array in a single
compilation unit
- use a single array for all gens
- add real functions to get or check gen by pciid
- define our own pci device struct rather than inherit the one
kernel uses: we can throw away most of the fields
v3: - add comment to keep ids sorted by gen
- remove misleading comment about all gens
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
typeof() is a GNU extension that will only work when the compiler is passed
-std=gnu*. __typeof__() works with -std=c*, however.
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>
It currently does 4 builds: 2 using Meson and 2 using Autotools, 2 using
the latest dependencies on ArchLinux and 2 using very old dependencies
on Debian (including manually building libpciaccess to have the oldest
version supported, to make sure it keeps being supported).
All the build options are turned on for both Meson and Autotools.
Signed-off-by: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Daniel Stone <daniels@collabora.com>
I picked up a bunch of the pieces from wayland's version:
https://gitlab.freedesktop.org/wayland/wayland/blob/master/CONTRIBUTING.md
The weston one is fairly similar. Then I rather massively trimmed it
down since in reality libdrm is a bit a dumping ground with very few
real rules. The commit rights and CoC sections I've copied verbatim
from igt respectively drm-misc. Weston/Wayland only differ in their
pick of how many patches you need (10 instead of 5). I think for
libdrm this is supremely relevant, since most everyone will get their
commit rights by contributing already to the kernel or mesa and having
commit rights there already.
Anyway, I figured this is good to get the rules documented, even if
there's mostly not many rules.
Note: This references maintainers in a MAINTAINERS file, which needs
to be created first.
Note: With the gitlab migration the entire commit rights process is
still a bit up in the air. But gitlab commit rights and roles are
hierarchical, so we can do libdrm-only maintainer/commiter roles
("Owner" and "Developer" in gitlab-speak). This should avoid
conflating libdrm roles with mesa roles, useful for those pushing to
libdrm as primarily kernel contributors.
v2: Comments from Emil:
- Recommend subject prefix.
- Fix copypaste fumbles, this isn't igt/wayland ...
v3: Comments from Marek:
- libdrm moved to mesa, update the document. Atm the entire account
request situation is entirely not clear for gitlab and mesa
projects, so that's a bit up in the air. Also, should probably send
an announcement to dri-devel@, which didn't happen.
- amd folks don't submit their patches to dri-devel, document that.
Probably applies to other drivers too.
v4: Comments from Rob:
- Also include kernel/userspace in the commit counts criteria, due to
libdrm's special role as a glue library.
v5: Summarize the irc discussion on gitlab roles in the commit message
a bit.
v6: Some grammer stuff from Eric E.
v7: Use --local in git config (Eric E.)
Cc: Dave Airlie <airlied@gmail.com>
Cc: Michel Dänzer <michel@daenzer.net>
Cc: Emil Velikov <emil.velikov@collabora.com>
Cc: Marek Olšák <marek.olsak@amd.com>
Cc: Rob Clark <robdclark@gmail.com>
Cc: Eric Engestrom <eric.engestrom@intel.com>
Reviewed-by: Rob Clark <robdclark@gmail.com> (v4)
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com> (v6)
Acked-by: Emil Velikov <emil.l.velikov@gmail.com> (v6)
Acked-by: Marek Olšák <marek.olsak@amd.com> (v5)
Acked-by: Dave Airlie <airlied@redhat.com>
References: https://gitlab.freedesktop.org/wayland/weston/blob/master/CONTRIBUTING.md
References: https://01.org/linuxgraphics/gfx-docs/maintainer-tools/drm-misc.html#commit-rights
References: https://cgit.freedesktop.org/drm/igt-gpu-tools/tree/CONTRIBUTING#n54
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
If nothing is found, error should be returned.
v2: udpate the error value different from parameter check
Signed-off-by: Junwei Zhang <Jerry.Zhang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Earlier commit reworked our sysfs handling to use realpath.
Sadly that backfired since the Firefox sandboxing mechanism rejects
that. Despite the files/folders being in the allowed list, of the
sandboxing mechanism.
Oddly enough, the Chromium sandboxing doesn't complain about any of
this.
Since there are no Firefox releases with the fix, add a temporary
solution which falls back to the original handling.
Sadly, this won't work for virgl.
v2: drop return type - function cannot return NULL (Eric)
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=107516
Fixes: a02900133b ("xf86drm: introduce a get_real_pci_path() helper")
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Tested-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>
msm_drm.h file Generated using make headers_install.
Generated from
tree - git://people.freedesktop.org/~airlied/linux
branch - drm-next
commit - 6d08b06e67cd117f6992c46611dfb4ce267cd71e
Remove freedreno/msm/msm_drm.h to maintain only
one copy of msm_drm.h and change freedreno Makefile
and meson.build file accordingly.
v2: Remove private freedreno/msm/msm_drm.h
v3: meson.build update
v3: README update (by anholt)
Signed-off-by: Tanmay Shah <tanmay@codeaurora.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
a helper function to create and initialize amdgpu bo
v2: update error handling: add label and free bo
v3: update error handling: separate each error label
v4: update error handling and rebase
Signed-off-by: Junwei Zhang <Jerry.Zhang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Fix potential memory leak when handle flink bo in bo import.
Free the flink bo after bo import and in error handling.
Signed-off-by: Junwei Zhang <Jerry.Zhang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Arithmetic using void* pointers isn't defined by the C standard, only as
a GCC extension. Avoids compiler warnings:
../../amdgpu/amdgpu_bo.c: In function ‘amdgpu_find_bo_by_cpu_mapping’:
../../amdgpu/amdgpu_bo.c:554:48: warning: pointer of type ‘void *’ used in arithmetic [-Wpointer-arith]
if (cpu >= bo->cpu_ptr && cpu < (bo->cpu_ptr + bo->alloc_size))
^
../../amdgpu/amdgpu_bo.c:561:23: warning: pointer of type ‘void *’ used in subtraction [-Wpointer-arith]
*offset_in_bo = cpu - bo->cpu_ptr;
^
v2: Use uintptr_t instead of char*, don't change function signature
(Junwei Zhang)
Fixes: 4d454424e1 ("amdgpu: add a function to find bo by cpu mapping
(v2)")
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com>
The compiler points out that an int doesn't work as intended if
dev->bo_handles.max_key > INT_MAX:
../../amdgpu/amdgpu_bo.c: In function ‘amdgpu_find_bo_by_cpu_mapping’:
../../amdgpu/amdgpu_bo.c:550:16: warning: comparison of integer expressions of different signedness: ‘int’ and ‘uint32_t’ {aka ‘unsigned int’} [-Wsign-compare]
for (i = 0; i < dev->bo_handles.max_key; i++) {
^
../../amdgpu/amdgpu_bo.c:558:8: warning: comparison of integer expressions of different signedness: ‘int’ and ‘uint32_t’ {aka ‘unsigned int’} [-Wsign-compare]
if (i < dev->bo_handles.max_key) {
^
Fixes: 4d454424e1 ("amdgpu: add a function to find bo by cpu mapping
(v2)")
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Junwei Zhang <Jerry.Zhang@amd.com>
One more CFL ID added to spec.
Align with kernel commit d0e062ebb3a4 ("drm/i915/cfl:
Add a new CFL PCI ID.")
v2: fix commit subject.
Cc: José Roberto de Souza <jose.souza@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: José Roberto de Souza <jose.souza@intel.com>
We could be dropping last reference in ->flush(), so clear the entry in
the parent rb's table to avoid deref'ing after free'd.
Also, ring_bo_del()'s use of ring_cache expects that it is dropping the
last reference. So drop our ref to the stateobj's ring_bo first.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
One stateobj can be emitted multiple times in a single cmdstream, but
only the first time is a cmd entry added to the parent. Since it will
be only unref'd once after flush, we should only ref it the first time
it is emitted (ie. the time it is added to cmd table).
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Userspace needs to know if the user memory is from BO or malloc.
v2: update mutex range and rebase
Signed-off-by: Junwei Zhang <Jerry.Zhang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
When create bo from user memory, add it to handle table
for future query.
Signed-off-by: Junwei Zhang <Jerry.Zhang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
If "-a" option is set this make modetest use atomic API instead
of legacy API.
Test the frame rate ("-v") it does a loop and swap between two
framebuffer for each active planes.
Signed-off-by: Benjamin Gaignard <benjamin.gaignard@linaro.org>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
The kernel handles are dense and the kernel always tries to use the
lowest free id. Use this to implement a more efficient handle table
by using a resizeable array instead of a hash.
v2: add handle_table_fini function, extra key checks,
fix typo in function name
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-and-Tested-by: Junwei Zhang <Jerry.Zhang@amd.com>
Adds support for "state object" cmdstream buffers which can be
constructed once, and re-used many times. This enables the use
for CP_SET_DRAW_STATE packets on newer hardware, to lower the
CPU overhead.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Splitting code-motion out from next patch. Once we add stateobj rb's
this loop could add new entries to bos table, so it needs to be before
we set req.bos/req.nr_bos.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Add new API for reusable "state objects" which can be re-used multiple
times. Backend implementation for msm will follow. (Probably not
needed to support this for any device that uses kgsl backend, since this
is mostly useful for a5xx+.)
Signed-off-by: Rob Clark <robclark@freedesktop.org>
In drmGetDevice2 when no local device is found or when
drm_device_has_rdev filters out all devices, *device might be left
uninitialized causing drmGetDevice2 to not return error - since
it's only returned when *device == NULL.
Above leads to crash in the firefox in system with amdgpu.
With this change firefox displays:
libGL error: MESA-LOADER: failed to retrieve device information
libGL error: unable to load driver: amdgpu_dri.so
libGL error: driver pointer missing
libGL error: failed to load driver: amdgpu
libGL error: MESA-LOADER: failed to retrieve device information
libGL error: unable to load driver: amdgpu_dri.so
libGL error: driver pointer missing
libGL error: failed to load driver: amdgpu
and doesn't crash.
Bugzilla: https://bugs.freedesktop.org/107384
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Mariusz Ceier <mceier+mesa-dev@gmail.com>
The GPU almost exclusively lives on the PCI bus, so we expose it as a
normal PCI one.
This allows all existing drmDevice users to work without any changes.
One could wonder why a separate typeset is not introduced, alike say
host1x. Unlike host1x the PCI/platform distinction for virtio provides
no extra information. Plus if needed we can add the separate set at a
later stage.
Here are a few 'features' that virtio seems to be missing:
- provides extra information on top the plaform devices
- supports a range of GPU devices
- is considered hardware description (DT)
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Tested-by: Robert Foss <robert.foss@collabora.com>
Reviewed-by: Robert Foss <robert.foss@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
Introduce a helper which gets the real sysfs path for the given pci
device.
In other words, instead opening the /sys/dev/char/*/device symlink, we
opt for the actual /sys/devices/pci*/*/
It folds three (nearly identical) snprintf's and paves the way of adding
extra devices (see next patch) a piece of pie.
v2: use a caller (on stack) provided real_path (Eric)
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Tested-by: Robert Foss <robert.foss@collabora.com> (v1)
Reviewed-by: Robert Foss <robert.foss@collabora.com> (v1)
Reviewed-by: Eric Engestrom <eric@engestrom.ch> (v1)
Currently we dynamically allocate 16 pointers and reallocate more as
needed.
Instead, allocate the maximum number (256) on stack - the number is
small enough and is unlikely to change in the foreseeable future.
This allows us to simplify the error handling and even shed a few bytes
off the final binary.
v2:
- add a define & description behind the magic 256
- report error to strerr and skip when over 256 device nodes
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Tested-by: Robert Foss <robert.foss@collabora.com> (v1)
Reviewed-by: Robert Foss <robert.foss@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch> (v1)
Currently we match the opened drmDevice fd with each drmDevice we
process.
Move that after all the devices are processed and folded, via the
drm_device_has_rdev(). This makes the code easier to follow and allows
us to unify the massive process loop across drmGetDevice2 and
drmGetDevices2. That in itself is coming with a later commit.
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Tested-by: Robert Foss <robert.foss@collabora.com>
Reviewed-by: Robert Foss <robert.foss@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
This reverts commit fe0488aa13.
It caused messages like
amdgpu 0000:23:00.0: bo 000000007dce0b3e va 0x0000101800-0x000010181f conflict with 0x0000101800-0x0000101820
in dmesg, and eventually a Xorg crash while running piglit.
Evidently, such BOs can actually be re-imported by other means than via
a KMS handle.
Amber Lake uses the same gen graphics as Kaby Lake, including a id
that were previously marked as reserved on Kaby Lake, but that now is
moved to AML page.
So, let's just move it to AML macro that will feed into KBL macro
just to keep it better organized to make easier future code review
but it will be handled as a KBL.
This is a copy of merged i915's
commit e364672477a1 ("drm/i915/aml: Introducing Amber Lake platform")
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Whiskey Lake uses the same gen graphics as Coffe Lake, including some
ids that were previously marked as reserved on Coffe Lake, but that
now are moved to WHL page.
So, let's just move them to WHL macros that will feed into CFL macro
just to keep it better organized to make easier future code review
but it will be handled as a CFL.
This is a copy of merged i915's
commit b9be78531d27 ("drm/i915/whl: Introducing Whiskey Lake platform")
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Something for users of fd_ringbuffer to use as they see fit. (For now,
just so mesa can add some debugging state.)
Signed-off-by: Rob Clark <robclark@freedesktop.org>
removed in commit bb45ce4e3a
Adding it back as it is still needed in the case where we don't find a
match.
Signed-off-by: Kevin Strasser <kevin.strasser@intel.com>
Fixes: bb45ce4e3a "libdrm: Use readdir instead of readdir_r to
avoid build warnings"$
Reviewed-by: Eric Engestrom <eric.engestrom@intel.com>
In mesa/gallium, a pipe_fence can outlive the pipe_context it was
created from. But to wait on the fence we need to know the submit-
queue (ie. the fd_pipe).
The most straightforward way to fix this is to add reference counting
to the fd_pipe and let the fence hold a reference to the pipe (rather
than hanging on to the context, which might have been destroyed before
the fence).
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Add the PCI IDs and the basic code to enable ICL. This is the current
PCI ID list in our documentation.
Kernel commit: d55cb4fa2cf0 ("drm/i915/icl: Add the ICL PCI IDs")
v2: Michel provided a fix to IS_9XX that was broken by rebase bot.
v3: Fix double definition of PCI IDs, update IDs according to bspec
and keep them in the same order and rebase (Lucas)
Cc: Michel Thierry <michel.thierry@intel.com>
Reviewed-by: Michel Thierry <michel.thierry@intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
128 is the maximum number of cards that the kernel can support
at the moment.
Change-Id: I155b7b21635306d8ecc440b85fb8954501ab5599
Signed-off-by: Xiaojie Yuan <Xiaojie.Yuan@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
In trying to integrate the new gralloc_handle.h with the
drm_hwcomposer, I started seeing the following compilation
errors:
In file included from external/drm_hwcomposer/platformdrmgeneric.cpp:28:
external/libdrm/android/gralloc_handle.h:108:9: error: cannot initialize return object of type 'native_handle_t *' (aka 'native_handle *') with an lvalue of type 'struct gralloc_handle_t *'
return handle;
^~~~~~
1 error generated.
This seems to be due to the gralloc_handle_create() definition
needs to return a native_handle_t * type, rather then a
gralloc_handle_t *, which is what the code actually returns.
After talking w/ Rob Herring, having the code return the
native handle should be the proper fix, so that is what
this patch changes.
Cc: Chih-Wei Huang <cwhuang@android-x86.org>
Cc: Stefan Schake <stschake@gmail.com>
Cc: Robert Foss <robert.foss@collabora.com>
Cc: Sean Paul <seanpaul@google.com>
Cc: Rob Herring <robh@kernel.org>
Signed-off-by: John Stultz <john.stultz@linaro.org>
Reviewed-by: Robert Foss <robert.foss@collabora.com>
Nouveau has made a very deliberate choice to hide its actual kernel ABI
behind libdrm. i915 is no longer out of date.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Taken from the drm-next pull for 4.17-rc1 (694f54f680f7), and manually
reconciled:
core:
- Dropped DRM_MODE_TYPE_ALL and DRM_MODE_FLAG_ALL; these are purely
internal details of the bits accepted by the currently running
kernel, and can not be generally relied on by userspace
- Add HDCP flags
- Note CTM entry representation is sign-magnitude format, not
two's-complement
amdgpu:
- Add QUERY_STATE2 context op
- Add VCN firmware version query
etnaviv:
- Add more GPU feature flags
i915:
- Add caps, params and ioctls for PMU / perf-stream
- Add support for explicit fencing
nouveau:
- Add TILE_COMP layout
vc4:
- Add perfmon ioctls
virtgpu:
- Add capset-fix param
vmware:
- Add handle-close ioctl and explicit-fencing support
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Add sensor_info type
AMDGPU_INFO_SENSOR_STABLE_PSTATE_GFX_MCLK
AMDGPU_INFO_SENSOR_STABLE_PSTATE_GFX_SCLK
Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Rex Zhu <Rex.Zhu@amd.com>
This reverts commit ed07718ae7.
The commit added a guard since libpciaccess may be missing on some
setups. As of last commit there are no traces of the project, from
Android POV.
Hence, we can revert this workaround - which caused similar breakage to
the one it's trying to fix. This time in Mesa.
Cc: Rob Herring <rob.herring@linaro.org>
Acked-by: John Stultz <john.stultz@linaro.org>
This patch makes the code not rely anymore on libpciaccess when compiled
for Android to eliminate ioperm() and iopl() syscalls required by that
library. As a side effect, the mappable aperture size is hardcoded to 64
MiB on Android, however nothing seems to rely on this value anyway, as
checked be grepping relevant code in drm_gralloc and Mesa.
Cc: Rob Herring <rob.herring@linaro.org>
Signed-off-by: Tomasz Figa <tfiga@google.com>
[Emil Velikov: rebase against master. add missing __func__, Eric]
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Acked-by: John Stultz <john.stultz@linaro.org>
This will prevent any more missing `#include "config.h"` bug, at the
cost of having to recompile some files that didn't need to be when
changing build options.
Signed-off-by: Eric Engestrom <eric.engestrom@imgtec.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Fixed memory leak issue to drmModeRes and drmModePlaneRes objects.
These objects were allocated by drmModeGetResources and
drmModeGetPlaneResources functions but not freed properly.
So this patch frees them by calling drmModeFreeResources
drmModeFreePlaneResources functions at failure case.
Signed-off-by: Inki Dae <inki.dae@samsung.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
When building AOSP after updating libdrm project to the
freedesktop/master branch, I've seen the following build errors:
external/libdrm/intel/Android.mk: error: libdrm_intel
(SHARED_LIBRARIES android-arm64) missing libpciaccess
(SHARED_LIBRARIES android-arm64) You can set
ALLOW_MISSING_DEPENDENCIES=true in your environment if this is
intentional, but that may defer real problems until later in the
build.
Using ALLOW_MISSING_DEPENDENCIES=true when building allows
things to function properly, but is not ideal.
So basically, while I'm not including the libdrm_intel package
into the build, just the fact that the Android.mk file references
libpciaccess which isn't a repo included in AOSP causes the build
failure.
So it seems we need some sort of conditional filter in the
Android.mk to skip over it if we're not building for intel.
Cc: Chad Versace <chad.versace@linux.intel.com>
Cc: Marissa Wall <marissaw@google.com>
Cc: Sean Paul <seanpaul@google.com>
Cc: Dan Willemsen <dwillemsen@google.com>
Cc: Tomasz Figa <tfiga@google.com>
Cc: Robert Foss <robert.foss@collabora.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Signed-off-by: John Stultz <john.stultz@linaro.org>
Signed-off-by: Rob Herring <robh@kernel.org>
There is already condition checking input values between 2 and 4096
so condition checking 0 is always false. Remove the dead condition.
Signed-off-by: Seung-Woo Kim <sw0312.kim@samsung.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
Meson's compiler.has_header is completely useless, it only checks that a
header exists, not whether it's usable. This creates problems if a
header contains a conditional #error declaration, like so:
> #if __x86_64__
> # error "Doesn't work with x86_64!"
> #endif
Compiler.has_header will return true in this case, even when compiling
for x86_64. This is useless.
Instead, we'll do a compile check so that any #error declarations will
be treated as errors, and compilation will work.
Fixes compilation on x32 architecture.
Gentoo Bugzilla: https://bugs.gentoo.org/show_bug.cgi?id=649746
meson bug: https://github.com/mesonbuild/meson/issues/2246
CC: Matt Turner <mattst88@gmail.com>
Signed-off-by: Dylan Baker <dylan.c.baker@intel.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
Object and property IDs cannot be zero. Prevent them from being added to
the request stream at all, rather than breaking at commit time.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
if some system has no swap space and memory is less, than the test
could fail.
And bo eviction test takes much more time in some system, which effects
automation test result and efficiency.
So disable it by default now, only be used by developer manually.
Signed-off-by: Chunming Zhou <david1.zhou@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
The range is stored as exclusive, not inclusive. Subtracts one to get
the inclusive interval for the calculation. This fixes crashes when 32bit
addresses are in use.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-and-Tested-by: Michel Dänzer <michel.daenzer@amd.com>
This reverts commit 07ea20d5be.
Unfortunately it turned out that this change broke some corner cases in
Mesa.
Revert it for now, but keep the high range in separate VA managers.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Needed for clover/OpenCL. Fortunately the kernel interface is already
in place.
Include a stub _put_iova() so mesa can tell us when it no longer needs
the buffer to be pinned. There is no kernel interface for this (yet),
but at least if we want to unpin buffers we won't need mesa changes.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
There's a number of problems with gralloc_handle_create starting with it
doesn't even compile. More importantly, it doesn't really create (i.e.
allocate) a handle. It allocates a native_handle_t, copies it to a
struct gralloc_handle_t on the stack and returns the struct (not a ptr).
So the caller still has to allocate a struct gralloc_handle_t to hold
the returned struct.
Rework gralloc_handle_create() to allocate a new handle and return the
pointer to the allocated handle. Callers should free the handle with
native_handle_close() and native_handle_delete(). In the interest of
making gralloc_handle_t opaque, return a native_handle_t ptr instead.
Reviewed-by: Robert Foss <robert.foss@collabora.com>
Signed-off-by: Rob Herring <robh@kernel.org>
Clients frequently need to convert a buffer_handle_t (aka
native_handle_t *) to a gralloc_handle_t ptr. This is a simple cast, but
add an inline function to do the conversion.
Reviewed-by: Robert Foss <robert.foss@collabora.com>
Signed-off-by: Rob Herring <robh@kernel.org>
Fix a typo where alloc_handle_t should be gralloc_handle_t. One still
remains in gralloc_handle_create, but a subsequent commit will fix that
along with other problems in gralloc_handle_create.
Reviewed-by: Robert Foss <robert.foss@collabora.com>
Signed-off-by: Rob Herring <robh@kernel.org>
Const members are problematic for dynamically allocating struct
gralloc_handle_t, so just drop the const modifier.
Reviewed-by: Robert Foss <robert.foss@collabora.com>
Signed-off-by: Rob Herring <robh@kernel.org>
-Dcairo-tests=false currently results into enabling cairo support if it
was found. Same for valgrind.
v2:
* Use underscore-prefixed variables to not change type of variable
* Use empty array for "fake" dependency instead of real empty object
v3:
* Fix typo
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
Signed-off-by: Igor Gnatenko <ignatenko@redhat.com>
The existing format modifier definitions were merged prematurely, and
recent work has unveiled that the definitions are suboptimal in several
ways:
- The format specifiers, except for one, are not Tegra specific, but
the names don't reflect that.
- The number space is split into two, reserving 32 bits for some
"parameter" which most of the modifiers are not going to have.
- Symbolic names for the modifiers are not using the standard
DRM_FORMAT_MOD_* prefix, which makes them awkward to use.
- The vendor prefix NV is somewhat ambiguous.
Fortunately, nobody's started using these modifiers, so we can still fix
the above issues. Do so by using the standard prefix. Also, remove TEGRA
from the name of those modifiers that exist on NVIDIA GPUs as well. In
case of the block linear modifiers, make the "parameter" smaller (4
bits, though only 6 values are valid) and don't let that leak into any
of the other modifiers.
Finally, also use the more canonical NVIDIA instead of the ambiguous NV
prefix.
This is based on commit 5843f4e02fbe86a59981e35adc6cabebee46fdc0 from
Linux v4.16-rc1 and also updates modetest to use the new defines.
Acked-by: Emil Velikov <emil.velikov@collabora.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Avoid compiler warnings when the val parameter is an expression.
This is based on commit 5843f4e02fbe86a59981e35adc6cabebee46fdc0 from
Linux v4.16-rc1.
Acked-by: Emil Velikov <emil.velikov@collabora.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
In order to lessen future alignment issues, lets switch to
fixed width integers where possible.
This excludes the data_owner since it is a pid_t which
in theory could be larger than 32 bits.
Signed-off-by: Robert Foss <robert.foss@collabora.com>
Reviewed-by: Rob Herring <robh@kernel.org>
The name member of gralloc_handle_t is no longer needed and has been removed.
The version field has also been bumped.
Signed-off-by: Robert Foss <robert.foss@collabora.com>
Reviewed-by: Rob Herring <robh@kernel.org>
The version variable will be used for versioning of this
struct and the corresponding accessor functions.
Signed-off-by: Robert Foss <robert.foss@collabora.com>
Reviewed-by: Rob Herring <robh@kernel.org>
This struct is used in mesa and drm_hwcomposer.
Versions of if have been implemented in several grallocs:
drm_gralloc, gbm_gralloc, minigbm and intel-minigbm.
Other than the 1:1 move of the struct a new generic name
has been chosen and variables have had comments added to them.
Signed-off-by: Robert Foss <robert.foss@collabora.com>
Reviewed-by: Rob Herring <robh@kernel.org>
for(( i=1; i < 100; i++))
do
echo "Hello, Welcome $i times "
sudo ./amdgpu_test -s 1 -t 5
done
with above stricpt, run in two terminals, will reproduce Felix's swap leeking issue.
Signed-off-by: Chunming Zhou <david1.zhou@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
issue: UMD allocates top 4GB, but don't do anything, just reserve top 4GB space,
but the performance of VP13 drops from 162fps to 99fps.
root cause:
our va hole list of vamgr is too long by time going.
fix:
reusing old hole as much as possible can make the list shortest.
result:
performance recovers as non-list path, next patch will remove non-list code path.
Signed-off-by: Chunming Zhou <david1.zhou@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Otherwise we get an EFAULT, at least on a 64-bit kernel.
Fixes: 2048a9e7 "drm: add drmSyncobjWait wrapper"
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
In case of a kernel that is new enough to support multiple submit-
queues, but with an adreno generation which doesn't support multiple
prioritized ringbuffers, we'd attempt to open a submit-queue with
prio=1 (medium), which is rejected by the kernel.
This could happen either w/ an older mesa (which uses fd_pipe_new())
or a newer mesa which defaults to prio=1 if no pipe context priority
flags are set.
The simple answer to fix both cases is to clamp the requested priority
according to the number of rings. This might not do exactly what you
want, if we hypothetically had 2 rings (it would result in requested
medium priority being high priority instead of low priority). But the
number of rings (for hw gen's that support this) is purely a software
construct, so the easy answer there is to have the kernel advertise at
least 3 rings if it supports more than one. There isn't really any
reason to do otherwise.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
If amdgpu_cs_query_fence_status terminates prematurely the BO
sometimes is unmapped before helper thread writes a vlaue
into it causing a segfault.
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Kernel will abort jobs for guilty (causing GPU hang) context
with -ECANCELED don't assert if that the case.
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Currently we ask for 0.42, but we actually require 0.43 because we pass
file objects as arguments to tests. If someone needs version 0.42 it
wouldn't be hard, just a lot of replacing files() with strings.
Signed-off-by: Dylan Baker <dylan.c.baker@intel.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
There is warning about ignoring return value of 'asprintf'. Fix to
check return value of asprintf().
Signed-off-by: Seung-Woo Kim <sw0312.kim@samsung.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
That constant needs to be 64bits.
Fixes: amdgpu: use the high VA range if possible v2
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Fixes use after free:
==2537== Invalid read of size 4
==2537== at 0x1162C9: suite_deadlock_tests_enable (deadlock_tests.c:101)
==2537== by 0x10B157: amdgpu_disable_suits (amdgpu_test.c:421)
==2537== by 0x10B157: main (amdgpu_test.c:560)
==2537== Address 0x5e44f24 is 452 bytes inside a block of size 1,016 free'd
==2537== at 0x4C2BE1B: free (vg_replace_malloc.c:530)
==2537== by 0x504CD8B: amdgpu_device_reference (amdgpu_device.c:164)
==2537== by 0x504CD8B: amdgpu_device_deinitialize (amdgpu_device.c:307)
==2537== by 0x1162BB: suite_deadlock_tests_enable (deadlock_tests.c:97)
==2537== by 0x10B157: amdgpu_disable_suits (amdgpu_test.c:421)
==2537== by 0x10B157: main (amdgpu_test.c:560)
==2537== Block was alloc'd at
==2537== at 0x4C2CC05: calloc (vg_replace_malloc.c:711)
==2537== by 0x504CA5E: amdgpu_device_initialize (amdgpu_device.c:212)
==2537== by 0x116298: suite_deadlock_tests_enable (deadlock_tests.c:93)
==2537== by 0x10B157: amdgpu_disable_suits (amdgpu_test.c:421)
==2537== by 0x10B157: main (amdgpu_test.c:560)
Reviewed-by: Christian König <christian.koenig@amd.com>
It means it just didn't find an entry for the GPU in the amdgpu.ids file.
Fixes spurious
amdgpu_parse_asic_ids: Cannot parse ASIC IDs: Resource temporarily unavailable
error messages in that case.
Reported-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
This patch adds a complete meson build system, including tests and
install. It has the necessary hooks to allow it be used as a subproject
for other meson based builds such as mesa.
Signed-off-by: Dylan Baker <dylan.c.baker@intel.com>
Reviewed-and-tested-by: Igor Gnatenko <i.gnatenko.brain@gmail.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
Retire the low range on Vega10 this frees up everything below 0xffff800000000000 for HMM.
v2: keep the 32bit range working.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
The amdgpu.ids is only required when building with amdgpu support.
Fix not to add it without amdgpu.
Signed-off-by: Seung-Woo Kim <sw0312.kim@samsung.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
These provide a crtc-id based interface to get the current sequence
(frame) number and to queue an event to be delivered at a specific sequence.
v2: Remove FIRST_PIXEL_OUT flag. This has been removed from the
proposed kernel API
Signed-off-by: Keith Packard <keithp@keithp.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
drmModeCreateLease
drmModeListLessees
drmModeGetLease
drmModeRevokeLease
Changes for v2:
Remove lessee id from GetLease
Remove lessor_id from ListLeases
Add revoke
Renumber to track kernel rebase on drm-next
Signed-off-by: Keith Packard <keithp@keithp.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Currently if the oldest BO in a bucket has different flags than what we
look for we'll miss the cache.Fix this by iterating over the cached BOs
until we find the oldest one with matching flags. This improves the hit
ratio for some of the buckets.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Add etna_cmd_stream_perf(..) to submit perform requests.
Userspace can submit pmrs via submit ioctl to sample perfmon
signals.
v3:
- mark perfmon bos as RW
Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
Query all domains and their signals and provide it this information
via struct etna_perfmon and the corresponding api functions.
v2:
- code style changes
- etna_perfmon_create(..): add missing clean up in error case
Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
Import the etnaviv header changes from kernel commit 05916bed1 (drm-next)
The drm_etnaviv_gem_submit structure was extended to include performance
monitor requests. Also two new ioctls got added to be able to readout
performance monitor domains and their signals.
Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Acked-by: Lucas Stach <l.stach@pengutronix.de>
There's no point in keeping around the full table of marketing names,
when amdgpu_get_marketing_name only ever returns the device's marketing
name.
Acked-by: Slava Abramov <slava.abramov@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
* Move empty/commented line check before the strdup and return -EAGAIN
directly
* Initialize r = -EAGAIN and remove redundant assignments
* Set r = -ENOMEM if last strdup fails, and remove redundant goto
Acked-by: Slava Abramov <slava.abramov@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
* Move error message printing into amdgpu_parse_asic_ids and make it
return void
* Print only "Invalid format" error message if parse_one_line returns
-EINVAL
* Use strerror instead of printing the (negative) error code in hex
Acked-by: Slava Abramov <slava.abramov@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
The test is as following:
1) Create context A & B
2) Send a command submission using context A which fires up a compute shader.
3) The shader wait a bit and then write a value to a memory location.
4) Send a command submission using context B which writes another value to the same memory location, but having an explicit dependency on the first command submission.
5) Wait with the CPU for both submissions to finish and inspect the written value.
Test passes if the value seen in the memory location after both submissions is from command B.
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
AOSP master has changed the build default to -Werror making all the
warnings errors. Override that with -Wno-error.
Signed-off-by: Rob Herring <robh@kernel.org>
The test stalls the CP, until RCA is done the test is
disabled to not disrupt regression testing.
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Allocates 1 TB of memory. Test is disabled by default
since it's triggers OOM killer.
v2:
FIx the test to only alloc the BO and assert if return value
not equal to -ENOMEM and remove test disable on start.
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Suits are diasbled based on hooks they provide (e.g incompatible
ASIC or missing blocks). Single tests are diasbled explicitly.
Suit or test can be forced to execute even if disabled by adding -f
flag after specifying suit [test] ids.
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
This pulls in pieces of drm-next d65d31388a23 ("Merge tag
'drm-misc-next-fixes-2017-11-07' of
git://anongit.freedesktop.org/drm/drm-misc into drm-next")
Signed-off-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
The kernel started using the struct from drm.h instead of the typedef.
Signed-off-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
This pulls in pieces of drm-next d65d31388a23
("Merge tag 'drm-misc-next-fixes-2017-11-07' of
git://anongit.freedesktop.org/drm/drm-misc into drm-next")
Signed-off-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
This pulls in pieces of drm-next d65d31388a23
("Merge tag 'drm-misc-next-fixes-2017-11-07' of
git://anongit.freedesktop.org/drm/drm-misc into drm-next")
Signed-off-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
This updates the header with the contents from drm-next d65d31388a23
("Merge tag 'drm-misc-next-fixes-2017-11-07' of
git://anongit.freedesktop.org/drm/drm-misc into drm-next")
Signed-off-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Adding the extra reservation of the 32bit space to the 64bit manager is
complete nonsense and just a waste of memory and CPU cycles.
Signed-off-by: Christian König <christian.koenig@amd.com>
With a new-enough kernel to support prioritized submit-queues, we can
expose priority level support to mesa. Open a submit queue associated
with the fd_pipe and pass it's id back to SUBMIT ioctl.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Rmove amdgpu_context_handle from the interface and use
amdgpu_device_handle instead. Uupdate VMID reservation test
accordingly.
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
The test will Reserve a VMID, submit a command and
unreserve the VMID.
v2:
Wrappers names were changed.
Signed-off-by: Andrey Grodzovsky <Andrey.Grodzovsky@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
v2:
Rename wrappers to match the IOCTL naming, fix
identation and fix make check error.
Signed-off-by: Andrey Grodzovsky <Andrey.Grodzovsky@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Both drmModeAddFB2() and drmModeAddFB2WithModifiers() have some
arguments that are just pointers to uint32_t in disguise. These
are not modified (just copied) in the function, so we can add a
const qualifier here.
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
Add a new context creation function that allows specifying the context
priority.
A high priority context has the potential of starving lower priority
contexts. The current kernel driver implementation allows only apps
that hold CAP_SYS_NICE or DRM_MASTER to acquire a priority above
AMDGPU_CTX_PRIORITY_NORMAL.
v2: corresponding changes for kernel patch v2
v3: Fixed 'make check' symbol error
Signed-off-by: Andres Rodriguez <andresx7@gmail.com>
Acked-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Generated using make headers_install from:
airlied/drm-next 282dc83 Merge tag 'drm-intel-next-2017-10-12' ...
Signed-off-by: Andres Rodriguez <andresx7@gmail.com>
Acked-by: Dave Airlie <airlied@redhat.com>
Acked-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Allow the user to override the default configuration set by setcrtc
for the primary plane. On some hardware primary planes can be freely
positioned/sized, and it'd be nice if we can actually test that feature.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
This teaches modetest about the new IN_FORMATS blob and decodes the
blob to show supported formats and modifiers.
Signed-off-by: Kristian H. Kristensen <hoegsberg@chromium.org>
As part of Treble project in Android O, all the device specific files have
to be located in a separate vendor partition. This is done by setting
LOCAL_PROPRIETARY_MODULE (the name is misleading). This change will not
break existing platforms without a vendor partition as it will just move
files to /system/vendor.
Signed-off-by: Rob Herring <robh@kernel.org>
In uvd encode parameter package, parameters input_pic_luma_pitch and
input_pic_chroma_pitch should be picture width align with hardware alignment.
The hardware alignment is 16 for amdgpu family earlier than AMDGPU_FAMILY_AI,
and 256 for later than and including AMDGPU_FAMILY_AI.
Signed-off-by: James Zhu <James.Zhu@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Query hardware IP information to find out if there are uvd encode rings
ready for use in kernel driver.
Signed-off-by: James Zhu <James.Zhu@amd.com>
Reviewed-by: Leo Liu <leo.liu@amd.com>
Adding initial tests for locks detection when SW
scheduler FIFO is full.
The test works by submitting a batch of identical commands which make the CP
stall waiting for condition to become true. The condition is later satisfied
form a helper thread. Other events that happen during this time
might create deadlock situations. One such example is GPU reset
triggered by this stall when amdgpu_lockup_timeout != 0.
v2:
Increase the delay from 2 to 100 ms.
Comment out the compute test until it's working.
Typos fix.
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Fixes this warning:
freedreno/kgsl/kgsl_ringbuffer.c: In function ‘kgsl_ringbuffer_flush’:
freedreno/kgsl/kgsl_ringbuffer.c:149:19: warning: cast from pointer to integer of different size [-Wpointer-to-int-cast]
req.timestamp = (uint32_t)kgsl_ring->bo->hostptr;
^
Signed-off-by: Eric Engestrom <eric.engestrom@imgtec.com>
Reviewed-by: Rob Clark <robdclark@gmail.com>
If drmPrimeFDToHandle fails in etna_bo_from_dmabuf, the function must
not return with the table_lock mutex held. Unlock the mutex in the error
path.
Based on ceb70a6b10 ("freedreno: prevent deadlock in error path").
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
Define two macros to avoid building errors.
Fixes: 7e6bf88cac (amdgpu: move asic id table to a separate file)
Signed-off-by: Chih-Wei Huang <cwhuang@linux.org.tw>
there is race issue between two threads on amdgpu_bo_reference and
amdgpu_bo_import, this patch tends to fix it by moving the
pthread_mutex_lock out of bo_free_internal and move to bo_reference
to cover the update_reference part.
The mutex_unlock in bo_import should also cover bo refcount
increasement.
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Monk Liu <monk.liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
`ring` cannot be non-null, so the label reduces to a simple return.
Then, there is no point initialising `ring` just to overwrite it before
anyone reads it.
Signed-off-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Rob Clark <robdclark@gmail.com>
`pipe` cannot be non-null, so the label reduces to a simple return.
Then, there is no point initialising `pipe` just to overwrite it before
anyone reads it.
Signed-off-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Rob Clark <robdclark@gmail.com>
GCC 7 started warning when a switch case has neither a `break` nor
a "fallthrough" comment.
Let's be explicit that we meant to fall through here.
Signed-off-by: Eric Engestrom <eric@engestrom.ch>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Analogous to previous commit (and the rest of the codebase), simply
discard the device if we cannot parse it.
Fixes: f8484ccbd1 ("xf86drm: Add USB support")
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
On ChromeOS devices, readdir() processes the directory in
the following order:
-NAME- -TYPE-
. n/a
.. n/a
vgem n/a
card1 DRM_BUS_PLATFORM
renderD129 DRM_BUS_PLATFORM
card0 DRM_BUS_PCI
renderD128 DRM_BUS_PCI
controlD64 DRM_BUS_PCI
In drmGetDevices2, after drmProcessPlatformDevice fails for
/dev/dri/card1, we don't process the remaining directory entries.
As such, Vulkan fails to initialize since Mesa uses drmGetDevices2.
To fix this, continue if drmProcessPlatformDevice fails.
Fixes: 7b1f37f474 ("xf86drm: Add platform and host1x bus support")
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
[Emil: correct the host1x platforms as well]
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
This just sends chunks to the kernel API for a single command
stream.
This should provide a more future proof and extensible API
for command submission.
v2: use amdgpu_bo_list_handle, add two helper functions to
access bo and context internals.
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
It is safe to submit the full struct even on older kernels as such
kernels do not process the full struct. Without this change it
becomes quite challenging to extned the submit struct.
Freedreno has no special treatment too. See git commits
- freedreno: sync uapi header
- freedreno: add fence fd support
Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
Tested-by: Philipp Zabel <p.zabel@pengutronix.de>
No functional change. Just organizing the code
so it gets clear for future platforms.
Paulo deserves credits becuase he was the one
that just noticed this IS_9XX was in the wrong position
after CNL patches got introduced.
Cc: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
By the Spec all CNL Y skus are 2+2, i.e. GT2.
This is a copy of merged i915's
commit 95578277cbdb ("drm/i915/cnl: Add Cannonlake PCI IDs for Y-skus.")
v2: Add kernel commit id for reference.
Cc: Anusha Srivatsa <anusha.srivatsa@intel.com>
Cc: Clinton Taylor <clinton.a.taylor@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Clinton Taylor <clinton.a.taylor@intel.com>
Platform enabling and its power-on are organized in different
skus (U x Y x S x H, etc). So instead of organizing it in
GT1 x GT2 x GT3 let's also use the platform sku.
This is a copy of merged i915's
commit e918d79a5d0a ("drm/i915/cnl: Add Cannonlake PCI IDs for U-skus.")
v2: Remove PCI IDs for SKU not mentioned in spec.
v3: Add kernel commit id for reference.
Cc: Anusha Srivatsa <anusha.srivatsa@intel.com>
Cc: Clinton Taylor <clinton.a.taylor@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Reviewed-by: Clinton Taylor <clinton.a.taylor@intel.com>
These ioctls are now in drm next so add the first set of libdrm APIs.
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>
This syncs the drm.h header with my drm-next branch as of
6d61e70ccc21606ffb8a0a03bd3aba24f659502b.
It brings over the semaphore API changes.
Generated using make headers_install.
Generated from git://people.freedesktop.org/~airlied/linux drm-next commit 6d61e70ccc2.
[airlied: I split patch in two, split reviewed by across both]
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Dave Airlie <airlied@redhat.com>
This syncs the amdgpu_drm header with my drm-next branch as of
6d61e70ccc21606ffb8a0a03bd3aba24f659502b.
It brings over the VM and semaphore API changes.
Generated using make headers_install.
Generated from git://people.freedesktop.org/~airlied/linux drm-next commit 6d61e70ccc2.
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
v2: fix an off by one error and leading white spaces
v3: use thread safe strtok_r(); initialize len before calling getline();
change printf() to drmMsg(); add initial amdgpu.ids
v4: integrate some recent internal changes, including format changes
v5: fix line number for empty/commented lines; realloc to save memory;
indentation changes
v6: remove a line error
v7: [Michel Dänzer]
* Move amdgpu.ids to new data directory
* Remove placeholder entries from amdgpu.ids
* Set libdrmdatadir variable in configure.ac instead of Makefile.am
[Emil Velikov]
* Use isblank() instead of open-coding it [Emil Velikov]
* Don't leak asic_id_table memory if realloc fails [Emil Velikov]
* Check and bump table_max_size at the beginning of the while loop [Emil
Velikov]
* Initialize table_max_size to the number of entries in data/amdgpu.ids
v8: [Michel Dänzer]
* Make sure amdgpu_asic_id.c gets rebuilt when amdgpu.ids changes
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Samuel Li <Samuel.Li@amd.com>
The device_id option [-d] was badly broken. This commit fixes
the width (was 8 is now 16 bits) as well as enables searches
without specifying a bus id. It was also comparing "dev" from
the bus field which is not the PCI device id.
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Tom St Denis <tom.stdenis@amd.com>
Adds the C++ extern guards from
ebbb0e5cfd2ceb1150b1 drm: add extern C guard for the UAPI headers
Generated using `make headers_install` from airlied/drm-next
at commit 2a1720376adda5ecf8e636fbfb05339c7dad1c55
Signed-off-by: Eric Engestrom <eric.engestrom@imgtec.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
This adds a bunch of modifiers stuff, as well as a few new formats.
Includes the following changes:
ebbb0e5cfd2ceb1150b1 drm: add extern C guard for the UAPI headers
b9fb2a21ac8058965a6b drm_fourcc: Document linear modifier
af913418261d6d3e7a29 drm_fourcc: Fix DRM_FORMAT_MOD_LINEAR #define
fd056f05b9fcba35b77e drm: add fourcc codes for 16bit R and RG
73f1a5858bf82f3bf232 drm/fourcc: add vivante tiled layout format modifiers
ba2b5277dc52cc96944d drm: add RGB formats with separate alpha plane
5e91144dd702d068b22a drm/tegra: Add tiling FB modifiers
Generated using `make headers_install` from airlied/drm-next
at commit 2a1720376adda5ecf8e636fbfb05339c7dad1c55
Signed-off-by: Eric Engestrom <eric.engestrom@imgtec.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
In commit bbe998791d ("Android's major/minor/makedev live in
<sys/sysmacros.h>"), it didn't apply cleanly and I missed the trailing
\, so add it here.
Signed-off-by: Rob Herring <robh@kernel.org>
AC_HEADER_MAJOR only defines MAJOR_IN_SYSMACROS if major() is _not_
defined by <sys/types.h> alone. It is, but it warns, and that's ugly.
To fix this, push -Werror into CFLAGS when invoking AC_HEADER_MAJOR so
the warning makes the compilation test fail.
Reviewed-by: Emil Velikov <emil.velilkov@collabora.com>
Signed-off-by: Adam Jackson <ajax@redhat.com>
drmCompareBusInfo was almost this already, but it wasn't exported, its
name didn't match its functionality, and while it almost looks like it
was usable for sorting due to memcmp it wouldn't work if you had
multiple bus types. I don't really want to think about defining a
sensible sort order for bus types, so let's at least make it less of a
trap for the caller.
Invert its boolean sense to be 'true if equal', rename it to describe
the types it actually operates on, and export.
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Emil Velikov <emil.velilkov@collabora.com>
Signed-off-by: Adam Jackson <ajax@redhat.com>
fourcc is not a string, it's a packed integer. This happens to work out
on LE, but gets reversed on BE.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Be consistent and use the canonical form while sanity checking
null pointers, also combine a few branches for brevity.
v2: rebase on top of 'add amdgpu_cs_wait_fences' series.
Signed-off-by: Edward O'Callaghan <funfunctor@folklore1984.net>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Add etna_cmd_stream_flush2 with in-fence fd and out-fence fd support for
explicit fencing.
v3: added etna_cmd_stream_flush2 to etnaviv/etnaviv-symbol-check
v2: renamed etna_cmd_stream_flush_explicit to etna_cmd_stream_flush2
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Import the etnaviv header changes from kernel commits 9ad59fea162c
("drm/etnaviv: submit support for in-fences") and 78ec187f64fa
("drm/etnaviv: submit support for out-fences") for fence fd support.
The drm_etnaviv_gem_submit structure was extended to include a flags
field, new flags for in-fence and out-fence fds and an input/output
fence fd field.
This is one-way backwards compatible because old userspace code passing
a short structure not including the flags field to new kernels will
cause the remaining fields to be zero-filled. New userspace code must
make sure to only pass the short structure to old kernels, though.
Not generated using make headers_install, since the drm/etnaviv_drm.h
uapi header is not installed yet by the kernel.
Copied from the airlied/drm-next commit 78ec187f64fa.
v2: improved commit message
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Reviewed-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
This exposes amdgpu_query_sensor_info().
v2: - add amdgpu_query_sensor_info() to the symbols list
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
When using the atomic API, one request can span multiple CRTCs, however
one event is generated per CRTC. As we cannot disambiguate the CRTC with
user data (since we only have one piece of user data to pass in), newer
kernels can include the CRTC ID in the page flip event.
Add a new vfunc to dispatch vblank events carrying a CRTC ID to clients
who negotiate a higher interface version.
[daniels: Rebased, include new cap, call page_flip_handler if it is set
but page_flip_handler2 isn't even on newer contexts, write a
commit message.]
v2: Split into separate commit.
Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@intel.com>
Generated using make headers_install, based on drm-misc-next commit
5db06a8a98f515f67446a69c57577c4c363ec65d.
This clarifies the comments around modifiers such that they are
per-framebuffer rather than per-plane, adds the beginnings of aspect
ratio mode flags, link status properties, and updates the 'reserved'
field from vblank events to include the CRTC ID.
v2: Split into separate patch, pull in full kernel changes.
v3: Undo revert of connector-type enums, since it is not actually
harmful.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
/me derped, component libdrm doesn't exist under Mesa because it's in DRI.
While at it, give the full URL and make it https.
Signed-off-by: Eric Engestrom <eric.engestrom@imgtec.com>
Accidentally removed with earlier commit. Although rather than simply
adding the check for all platforms, use the same heuristics as we do in
mesa.
Namely: pthread-stubs is a thing only for BSD platforms. Everyone else
should have lightweight pthread API in their C runtime library.
Fixes: 4ecd1ef010 ("xf86drm: fix null termination of string buffer")
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>
I get a few more symbols in my build tegra-libraries, so let's
include these in the whitelist as well.
While we're at it, update the comment at the top.
Signed-off-by: Erik Faye-Lund <kusmabite@gmail.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Disable some more warnings from clang. These don't appear to be warnings
worth fixing.
Signed-off-by: Rob Herring <robh@kernel.org>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
These tests depend on tests/util/ headers, but expect the include path
to be tests/.
Signed-off-by: Rob Herring <robh@kernel.org>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
In error path of drmGetBusid() and drmGetReservedContextList(),
there are memory leaks for error path. So this removes them.
Signed-off-by: Seung-Woo Kim <sw0312.kim@samsung.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
This patch fixes invalid code of error path including NULL
deference and leak in g2d test.
Signed-off-by: Seung-Woo Kim <sw0312.kim@samsung.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
The function g2d_blend_test() is blocked to call because of
feature unsafety. This patch blocks with proper feature name
and also blocks the function itself to remove build warning.
Signed-off-by: Seung-Woo Kim <sw0312.kim@samsung.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
As like the commit ecc2a09729 ("xf86drm: Fix type-punned pointer
build warning"), this fixes following build warning.
exynos_drm.c: In function 'exynos_handle_event':
exynos_drm.c:420:15: warning: dereferencing type-punned pointer will break strict-aliasing rules [-Wstrict-aliasing]
e = (struct drm_event *) &buffer[i];
^
Signed-off-by: Seung-Woo Kim <sw0312.kim@samsung.com>
Reviewed-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Otherwise the make check will rightfully fail.
Fixes: 4e369f25a9 ("amdgpu: add amdgpu_bo_va_op_raw")
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Changes include: PRT and preemption flags, sensor info, and some more
changes for Vega10.
Generated using make headers_install from airlied/drm-next commit
320d8c3d38739fa8e31a076b86cbdafcf8897d5e.
Signed-off-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Acked-by: Emil Velikov <emil.velikov@collabora.com>
Follow up to 'drm: don't access deprecated register on Vega10'.
The same information is available in enabled_rb_pipes_mask and reading that
register can cause GRBM bus problems.
Signed-off-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Marek Olšák <marek.olsak@amd.com>
swizzle mode needs reference and input picture luma and
chroma pitch aligned with 256
Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Move closing the fd to after subclass ->destroy() (since it might want
to delete gem bo's, etc), and actually free() the fd_device object.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
If you don't, then the first place the m4 expands is:
if test "x$INTEL" != "xno"; then
PKG_CHECK_MODULES(PCIACCESS, [pciaccess >= 0.10])
fi
So on non-Intel architectures we never find it in the path, and all
subsequent PKG_CHECK_MODULESes fail. Boo autoconf.
Signed-off-by: Adam Jackson <ajax@redhat.com>
Something that valgrind spotted:
==8441== Invalid read of size 4
==8441== at 0x5DEE168: msm_ringbuffer_emit_reloc (msm_ringbuffer.c:506)
==8441== by 0x5B48F0F: OUT_RELOCW (freedreno_util.h:241)
==8441== by 0x5B48F0F: fd5_emit_blit (fd5_emit.h:131)
==8441== by 0x5B48F0F: emit_gmem2mem_surf.isra.12 (fd5_gmem.c:450)
==8441== by 0x5B4910F: fd5_emit_tile_gmem2mem (fd5_gmem.c:477)
==8441== by 0x5B14943: render_tiles (freedreno_gmem.c:342)
==8441== by 0x5B14943: fd_gmem_render_tiles (freedreno_gmem.c:416)
==8441== by 0x5B0FBA7: batch_flush (freedreno_batch.c:281)
==8441== by 0x5B0FBA7: fd_batch_flush (freedreno_batch.c:306)
==8441== by 0x5B11FE7: fd_context_flush (freedreno_context.c:52)
==8441== by 0x58AD783: st_glFlush (st_cb_flush.c:121)
==8441== by 0x5751EE7: _mesa_make_current (context.c:1652)
==8441== by 0x58E6A97: st_api_make_current (st_manager.c:811)
==8441== by 0x5A2CE43: dri_unbind_context (dri_context.c:207)
==8441== by 0x5A2C77F: driUnbindContext (dri_util.c:589)
==8441== by 0x4AC8A67: MakeContextCurrent (glxcurrent.c:214)
==8441== Address 0x6f5eb1c is 204 bytes inside a block of size 240 free'd
==8441== at 0x4868F44: realloc (vg_replace_malloc.c:785)
==8441== by 0x5DEE143: msm_ringbuffer_emit_reloc (msm_ringbuffer.c:502)
==8441== by 0x5B48F0F: OUT_RELOCW (freedreno_util.h:241)
==8441== by 0x5B48F0F: fd5_emit_blit (fd5_emit.h:131)
==8441== by 0x5B48F0F: emit_gmem2mem_surf.isra.12 (fd5_gmem.c:450)
==8441== by 0x5B4910F: fd5_emit_tile_gmem2mem (fd5_gmem.c:477)
==8441== by 0x5B14943: render_tiles (freedreno_gmem.c:342)
==8441== by 0x5B14943: fd_gmem_render_tiles (freedreno_gmem.c:416)
==8441== by 0x5B0FBA7: batch_flush (freedreno_batch.c:281)
==8441== by 0x5B0FBA7: fd_batch_flush (freedreno_batch.c:306)
==8441== by 0x5B11FE7: fd_context_flush (freedreno_context.c:52)
==8441== by 0x58AD783: st_glFlush (st_cb_flush.c:121)
==8441== by 0x5751EE7: _mesa_make_current (context.c:1652)
==8441== by 0x58E6A97: st_api_make_current (st_manager.c:811)
==8441== by 0x5A2CE43: dri_unbind_context (dri_context.c:207)
==8441== by 0x5A2C77F: driUnbindContext (dri_util.c:589)
==8441== Block was alloc'd at
==8441== at 0x4868F44: realloc (vg_replace_malloc.c:785)
==8441== by 0x5DEE08B: msm_ringbuffer_emit_reloc (msm_ringbuffer.c:481)
==8441== by 0x5B48F0F: OUT_RELOCW (freedreno_util.h:241)
==8441== by 0x5B48F0F: fd5_emit_blit (fd5_emit.h:131)
==8441== by 0x5B48F0F: emit_gmem2mem_surf.isra.12 (fd5_gmem.c:450)
==8441== by 0x5B4909F: fd5_emit_tile_gmem2mem (fd5_gmem.c:465)
==8441== by 0x5B14943: render_tiles (freedreno_gmem.c:342)
==8441== by 0x5B14943: fd_gmem_render_tiles (freedreno_gmem.c:416)
==8441== by 0x5B0FBA7: batch_flush (freedreno_batch.c:281)
==8441== by 0x5B0FBA7: fd_batch_flush (freedreno_batch.c:306)
==8441== by 0x5B11FE7: fd_context_flush (freedreno_context.c:52)
==8441== by 0x58AD783: st_glFlush (st_cb_flush.c:121)
==8441== by 0x5751EE7: _mesa_make_current (context.c:1652)
==8441== by 0x58E6A97: st_api_make_current (st_manager.c:811)
==8441== by 0x5A2CE43: dri_unbind_context (dri_context.c:207)
==8441== by 0x5A2C77F: driUnbindContext (dri_util.c:589)
Signed-off-by: Rob Clark <robclark@freedesktop.org>
drm_intel_gem_bo_free() unconditionally attempts to remove the handle
from the hashtable. This goes horribly wrong if we haven't already added
the bo to the hashtable.
Reported-by: Michael Thayer <michael.thayer@oracle.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Move the dereference after the null check.
Fixes: 028715ee70 ("intel: Avoid the need for most overflow
checks by using a scratch page.")
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
Use obj->kflags to set EXEC_OBJECT_PINNED when the object is softpinned,
and so remember to clear the softpin status when the object is freed
(and reused).
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Implicit synchronisation is the default behaviour of the kernel when
rendering with an execobject. It may be disabled with
drm_intel_gem_bo_disable_implicit_sync(), and then to restore it use
drm_intel_gem_bo_enable_implicit_sync().
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Allow the caller to pass in an fd to an array of fences to control
serialisation of the execbuf in the kernel and on the GPU, and in return
allow creation of a fence fd for signaling the completion (and flushing)
of the batch. When the returned fence is signaled, all writes to the
buffers inside the batch will be complete and coherent from the cpu, or
other consumers. The return fence is a sync_file object and can be
passed to other users (such as atomic modesetting, or other drivers).
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
The kernel allows implicit synchronisation to be disabled on individual
buffers. Use at your own risk.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Tested:
1. As root, tests passed on primary.
2. As root, tests passed on render node.
BO export/import test was skipped
3. As non-privileged user, tests failed on primary as expected.
4. As non-privileged user, tests passed on render node.
BO export/import test was skipped
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Alex Xie <AlexBin.Xie@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This can be used to test multiple GPUs
v2: Use PCI bus ID and optional PCI device ID to choose device
Add an option to display information of AMDGPU devices
Tested:
./amdgpu_test -p
./amdgpu_test
./amdgpu_test -b 1 #fail as expected
./amdgpu_test -b 6 #pass
./amdgpu_test -b -d 1 #fail as expected
./amdgpu_test -b -d 0 #pass
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Alex Xie <AlexBin.Xie@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Verify the vender ID and driver name.
Open all AMDGPU devices.
Provide an option to open render node.
Tested as root: PASS
Tested as non-privileged user:
All tests failed as expected
v2: Return value in the ene of function amdgpu_open_devices.
Check the return value of amdgpu_open_devices.
amdgpu_test is not for USB device for the time being.
Get the name of node from function drmGetDevices2.
Drop the legacy drmAvailable() from the test.
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Alex Xie <AlexBin.Xie@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
-lpthread is not always a valid flag to pull pthread support, especially
on Android it will fail to link due to a missing libpthread.so. The more
generic way to build-in pthread support is to use the -pthread CFLAG, so
let's use it instead.
Signed-off-by: Tomasz Figa <tfiga@chromium.org>
[Emil Velikov: rebase on top of previous commit]
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Since we don't have any C++ souces this should be a no-op. Folding the
two seems to be the common practise throughout the repo.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Currently only libdrm.so properly annotates its internal/private
symbols. By setting the macro every binary produced will be in the same
boat. This should give is smaller and more secure files
Cc: Chih-Wei Huang <cwhuang@android-x86.org>
Cc: Rob Herring <robh@kernel.org>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Rob Herring <robh@kernel.org>
... across the makefiles. Currently this isn't much but that will change
shortly.
As an added bonus this fixes all present and future cases where we've
forgotten to strip out the headers from LOCAL_SRC_FILES.
In a couple of cases (the tests) we start setting
LOCAL_EXPORT_C_INCLUDE_DIRS, which shouldn't be an issue.
Cc: Chih-Wei Huang <cwhuang@android-x86.org>
Cc: Rob Herring <robh@kernel.org>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Recent patches for USB, platform and host1x bus support introduced the
sysfs_uevent_get() function that provides a generic way of parsing the
sysfs uevent file that is associated with each device in Linux.
Open-coded variants of this still exist in other places, so make those
reuse the new function to remove some code duplication.
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Extend the drmdevice test with support for the newly added USB, platform
and host1x busses.
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
ARM SoCs usually have their DRM/KMS devices on the platform bus, so add
support for that to enable these devices to be used with the drmDevice
infrastructure.
NVIDIA Tegra SoCs have an additional level in the hierarchy and DRM/KMS
devices can also be on the host1x bus. This is mostly equivalent to the
platform bus.
v4:
- continue on error to process platform or host1x device
v3:
- guard Linux-specific sysfs parsing code with #ifdef __linux__
v2:
- be careful not to overflow the full name
- read compatible strings into device info
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Allow DRM/KMS devices hosted on USB to be detected by the drmDevice
infrastructure.
v4:
- continue on error to process USB devices
v3:
- guard Linux-specific sysfs parsing code with #ifdef __linux__
v2:
- make sysfs_uevent_get() more flexible using a format string
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Thierry Reding <thierry.reding@gmail.com>
Subsequent patches will add support for other bus types to drmDevice and
they will duplicate a lot of the code to allocate a drmDevice. Factor
out the common code so it can be reused.
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Add support for Amlogic Meson DRM driver merged for Linux 4.10.
Signed-off-by: Neil Armstrong <narmstrong@baylibre.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
This brings xf86drmMode.h in sync with include/drm/drm_mode.h.
Eventually we really should only have a single set of definitions rather
than duplicating this in two files.
v2: add DPI encoder and connector types introduced in Linux v4.7
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Fix two grammar issues:
- "standard autotools packages ---> "standard autotools package"
- "If you are install" ---> "If you are installing"
Signed-off-by: Fabio Estevam <festevam@gmail.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
The string written to the buffer by read() is not null-terminated,
but currently drmParsePciBusInfo() places null character only at the end of the buffer, not at the end of the
string.
As a result, the string passed to sscanf() contains an uninitialized value.
This patch changes to places null character at the end of the string.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=99045
Signed-off-by: Taro Yamada <archer_ame@yahoo.co.jp>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
bo->global_name should be updated first before a hash value
for the entry is calculated with it by HASH_ADD macro.
Signed-off-by: Dongwon Kim <dongwon.kim@intel.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Before this change, the error message is:
"WARNING - Suite initialization failed..."
People might think this is a driver problem.
Tested with non-privileged user. Now the error message is like:
...
Error:Permission denied. Hint:Try to run this test program as root.
WARNING - Suite initialization failed for 'Basic Tests'.
...
Tested as root with no regression.
amdgpu_test uses CUnit. CUnit outputs warning message to stdout.
To be consistent, this commit outputs error message to stdout.
v2: Use strerror instead of %m. %m is a GNU C Library extension.
v3: Limit code and commit message within 80 characters per line.
Update commit message.
Remove a space before starting parenthesis in function call.
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Xie <AlexBin.Xie@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
libdrm uses spaces for indentation. Fix the two inconsistent lines in
this file.
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
When iterating over all the device nodes if drmProcessPciDevice()
returned an error for any node the function would return an error,
ignoring any valid nodes.
The result of this on OpenBSD where drmProcessPciDevice() results in
device nodes being opened to issue ioctls to get pci data
was that data obtained from /dev/drm0 would be ignored if /dev/drm1
could not be opened.
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Jonathan Gray <jsg@jsg.id.au>
Implement a generic drmGetDeviceNameFromFd2() to use on non-linux
systems without sysfs.
v2: remove min < base test as requested by Emil
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Jonathan Gray <jsg@jsg.id.au>
When constructing a path to a device node the minor number retrieved
from fstat needs to have the offset of the node type subtracted from it.
Control and render node types have the same major as the primary node
but each has their own block of minor types at fixed offsets.
v2: remove min < base test as requested by Emil
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Jonathan Gray <jsg@jsg.id.au>
Just set the rules automatically rather than asking each contributor to
update thing locally.
v2: Silence errors if run outside of git repo. (Eric)
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>
xf86drm.c:3601:21: warning: comparison between signed and unsigned integer expressions [-Wsign-compare]
while (expected < sizeof(match)) {
^
Signed-off-by: Grazvydas Ignotas <notasas@gmail.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
DRI devices on OpenBSD are not in their own directory. They reside in
/dev with a large number of statically generated /dev nodes.
Avoid stat'ing all of /dev on OpenBSD by implementing this custom path.
v2:
- use drmGetMinorType to get node type
- adapt to drmProcessPciDevice changes
- verify drmParseSubsystemType type is PCI
- add a comment describing why this was added
Signed-off-by: Jonathan Gray <jsg@jsg.id.au>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Implement drmParsePciBusInfo for OpenBSD by using the new
DRM_IOCTL_GET_PCIINFO ioctl.
v2: use drmGetMinorType to get node type instead of always
using DRM_NODE_PRIMARY.
Signed-off-by: Jonathan Gray <jsg@jsg.id.au>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Implement drmParsePciDeviceInfo for OpenBSD by using the new
DRM_IOCTL_GET_PCIINFO ioctl.
v2: adapt to drmParsePciDeviceInfo changes and use drmOpenMinor
Signed-off-by: Jonathan Gray <jsg@jsg.id.au>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Implement drmParseSubsystemType for OpenBSD by always returning
DRM_BUS_PCI. No non-pci drm drivers are in the kernel and this is
unlikely to change anytime soon as the existing ones aren't permissively
licensed.
Signed-off-by: Jonathan Gray <jsg@jsg.id.au>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Implement drmGetMinorNameForFD for systems without sysfs by
adapting drm_get_device_name_for_fd() from the Mesa loader.
v2: use type parameter to select dev name instead of always
using DRM_DEV_NAME
Signed-off-by: Jonathan Gray <jsg@jsg.id.au>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
All of these 'tests' cover UMS functionality which is neither being
worked on or actively maintained.
The only cases where developers touch UMS code is to unwrap it from the
KMS codepaths and ensure that those are secure.
Anyone who feels strong about having these around can revive them, but
in all honestly do consider _seriously_ what you're doing ;-)
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Pass along DRM_DEVICE_GET_PCI_REVISION only when the individual nodes
are opened and update the printed messages accordingly.
v2: Attribute for the flag rename, call drmGetDevices2 w/o the flag.
v3: Keep drmParsePciDeviceInfo() hunk in previous patch.
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Relative to the original version, here one can provide a flags bitmask.
Currently only DRM_DEVICE_IGNORE_PCI_REVISION is supported.
Implementation detail:
If it's set, we will only parse the separate sysfs files and we won't
touch the config one. The latter awakes the device (causing delays)
which is the core reason why this API was introduced.
v2:
- Initialize revision to 0xff if it's unread.
- Change DRM_DEVICE_IGNORE_PCI_REVISION to DRM_DEVICE_GET_PCI_REVISION
- Add explicit note that drmGetDevice[s]2 does not retrieve the
revision by default.
v3:
- Correctly fold drmParsePciDeviceInfo() hunk in this patch.
Cc: Michel Dänzer <michel@daenzer.net>
Cc: Nicolai Hähnle <nhaehnle@gmail.com>
Cc: Mauro Santos <registo.mailling@gmail.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=98502
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Up-to recently (patch should land in 4.10) the kernel did not expose the
PCI device revision field as a separate sysfs file.
Thus one needed too parse the config file to retrieve it. This in
itself wakes up the device, which in some cases can be quite slow.
To avoid that, just check for the separate files and fall-back to the
original if kernel is not new enough.
v3: rework alongside drmGetDevice[s]2
Cc: Michel Dänzer <michel@daenzer.net>
Cc: Nicolai Hähnle <nhaehnle@gmail.com>
Cc: Mauro Santos <registo.mailling@gmail.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=98502
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Be consistent with drmParsePciBusInfo() and use solely the device
major/minor pair.
Cc: Jonathan Gray <jsg@jsg.id.au>
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
a5xx and later are 64bit devices.. make reloc's handle that. A new
public symbol is introduced to avoid silent problems with new mesa and
old libdrm (since on 64b reloc consumes two dwords).
Signed-off-by: Rob Clark <robclark@freedesktop.org>
We need to pass through a timeout parameter to implement
pipe->fence_finish() properly. The new fxn accepts a timeout
in nanoseconds. Simplify etna_pipe_wait(..) by using
etna_pipe_wait_ns(..).
Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
There is no need to maintain the value in the shadowed variable from
what I can see.
Signed-off-by: Grazvydas Ignotas <notasas@gmail.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
The original version considered only card devices, while this will pick
the device/node name regardless - card, control, renderD, other...
Current implementation is "linux" specific, in such that it relies on
sysfs/uevent file. At the same time this gives us the flexibility to
support any nodes even future ones, as long as they're within DRM_MAJOR.
Shamelessly copied from mesa, latter by: Gary Wong <gtw@gnu.org>
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Having the "Entering|Leaving directory X" messages it not required nor
useful in vast majority of the cases.
One can always have them printed by `make -w' or by overriding the
AM_MAKEFLAGS variable.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Exposing the u32 context ID makes it possible to define new drm kernel
interfaces based on the same IDs that e.g. execbuf uses to identify a
gem context, that aren't themselves abstracted by libdrm but need to be
used by libdrm/drm_intel_context based clients such as (parts of) i-g-t
or Mesa.
For example this can be used to configure an i915-perf stream to collect
metrics for a specific context.
v2: s/drm_intel_gem_context_get_context_id/drm_intel_gem_context_get_id/
Signed-off-by: Robert Bragg <robert@sixbynine.org>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Like etna_device_new() but creates it's own private dup() of the fd
which is close()d when the device is finalized.
Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Acked-by: Eric Anholt <eric@anholt.net>
Since we're trying to standardise and make things more consistent in
the area, add a basic README which covers some of the more popular
topics.
v2:
- Drop drm-misc (Daniel Vetter)
v3:
- Elaborate on when and which headers to update
- Add a list of headers and the respective "issues"
- Add file to EXTRA_DIST
Cc: Dave Airlie <airlied@redhat.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch> (v1)
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
As well as allowing a hexadecimal PCI ID number, the
INTEL_DEVID_OVERRIDE environment variable can now contain one of a few
short codenames. The codenames are stored in a small table to map them
to a corresponding PCI ID. This makes it easier to use without having
to look up the PCI IDs manually.
The PCI IDs used are the same as those chosen for the -p option of
run.c in shader-db but SKL has been added as well.
Reviewed-by: Matt Turner <mattst88@gmail.com>
This function is used to look up the marking name
for a specific board.
v2: agd: Squash in subsequent updates to the table.
v3: [Michel Dänzer]
* Make amdgpu_asic_id_table static, so it's not exported from
libdrm_amdgpu.so.1
* Add amdgpu_get_marketing_name to amdgpu-symbols-check
* Fix indentation of second line of if statement
* Squash in another change removing redundant entries
* Change spelling of "RADEON" -> "Radeon"
* Remove "(TM)" from a minority of entries
v4: [Michel Dänzer]
* Use const char* instead of fixed size array for marketing_name (Emil
Velikov)
Signed-off-by: Junwei Zhang <Jerry.Zhang@amd.com>
Reviewed-by: Flora Cui <Flora.Cui@amd.com>
Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Rather than cut/pasting these couple ioctl wrappers everywhere, just
stuff them as static-inline into a header.
This is probably mostly used from mesa, but some drivers, test apps, etc
may also want to use it from libdrm.
v2: handle EINTR, add sync_accumulate() based on #dri-devel discussion,
etc
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
A slightly confused copy'n'paste from the open path where we pass in
handle but use it as a global name, in the prime handle-from-fd pass we
pass in handle and do mean handle!
References: https://bugs.freedesktop.org/show_bug.cgi?id=98416
Fixes: 2f23bf1b7b89 ("intel: Migrate handle/name lookups from linear lists...")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Walking a linear list to find a matching PRIME handle or flinked name
does not scale and becomes a major burden with just a few objects.
That said, the fixed size hash is not much better, it just buckets the
look into a few separate chains rather than one long one.
References: https://bugs.freedesktop.org/show_bug.cgi?id=94631
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Export a set of interfaces to allow the caller to have precise control
over mapping the buffer - but still provide caching of the mmaps between
callers.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
glxgears was spamming this 12 times at startup because of Mesa's
probing of the DRM device code, which doesn't support platform
devices.
Signed-off-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Fixes crashes in Mesa on platform devices, which expected *device to
have a device when 0 was returned.
(code from a paste by Rob, commit message by anholt)
Signed-off-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Adapt to recent firmware update, it's also compatible with previous
firmware version
Signed-off-by: Leo Liu <leo.liu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
The only other user of this feature open codes the ioctl. Let's add an
entry point for this to libdrm.
Signed-off-by: Kristian H. Kristensen <hoegsberg@chromium.org>
Reviewed-by: Rob Clark <robdclark@gmail.com>
This adds the following basic unit tests:
- etnaviv_2d_test
Let the 2D core render a defined pattern into a bo
and store it as bmp.
- etnaviv_bo_cache_test
Basic tests to validate the bo-cache behavior.
- etnaviv_cmd_stream_test
Tests for the etna_cmd_stream API.
Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com>
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Update kernel interface with new I915_GETPARAM ioctl entries for
pooled EU and min no. of eus in a pool. Add a wrapping function
for each parameter. Userspace drivers need these values when decide
the thread count. This kernel enabled pooled eu by default for BXT
and for fused down 2x6 parts it is advised to turn it off.
But there is another HW issue in these parts (fused
down 2x6 parts) before C0 that requires Pooled EU to be enabled as a
workaround. In this case the pool configuration changes depending upon
which subslice is disabled and the no. of eus in a pool is different,
So userspace need to know min no. of eus in a pool.
V2: use return value as the query results.
ret < 0 when error, ret = 0 when not support, and ret > 0 indicate
query results.(Chris)
V3: Correct V2 errors.
Signed-off-by: Yang Rong <rong.r.yang@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Sorry, I don't understand the android build system enough to say *which*
version of clang this effects, but either "clang-2812033" or
"clang-3016494" (probably the later).
But when 'sample' is undefined (ie. unitialized variable), the result is
not as well defined as it is with gcc. Instead use a typeof() cast with
a defined value (ie. zero).
This fixes a crash that was reported on android.
Reported-by: John Stultz <john.stultz@linaro.org>
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Since release.sh creates and pushes a libdrm-$VERSION tag for us,
there's no need to also have the user manually generating a $VERSION
tag as well.
I also dropped the "optional" part of distcheck. You shouldn't have
pushed master with a version bump that hasn't passed distcheck.
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
The plan is to use this version of virtgpu_drm.h in mesa and drop mesa's
local copy.
To actually use this header it needs to be shipped in the tarball.
This was missed in c745e541a9
Signed-off-by: Andreas Boll <andreas.boll.dev@gmail.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Generated using make headers_install.
This brings the C++ guard, proper include path for drm.h and the
switching to kernel types for fixed-with integers.
Generated from drm-next commit c11dea5b0290984fa48111957ba3fdc5b3bdae5a
Suggested-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Andreas Boll <andreas.boll.dev@gmail.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Currently only some Android Makefiles are included in the release tarball.
To be more consistent one could either add the remaining files or don't
ship Android Makefiles altogether.
According to Emil the Android folk doesn't use our release tarballs.
Thus it makes sense to remove those files from distribution which also
means less work for maintenance in the future.
Suggested-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Andreas Boll <andreas.boll.dev@gmail.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
The 'deprecated' #define was causing problems with bionic system headers
which used __attribute__((deprecated)).
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Tested-by: Rob Herring <robh@kernel.org>
Note: cache the last ring the bo was emitted on, to avoid excess
hashtable lookups. We do this by tracking ring seqno to avoid
problems with dangling pointers.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
First step towards supporting a single logical ringbuffer mapping to
multiple physical cmd buffers, which will enable dynamically growing
ringbuffers.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Not actually needed. It just needs to ensure that there is a
corresponding entry in the submit's cmds table.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Since they get vmap'd on the kernel side, they are a bit more costly.
Don't let them mingle with the riffraff.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
If user has emit'd reloc's, and then resets or deletes the ring, we want
to drop the ref's that the ring holds to the bo's to avoid a leak.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
The normal bo cache uses some intermediate steps between power of two
jumps to reduce memory wastage. But for a ringbuffer bo cache, we do
not need this.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
gallium needs to know if the kernel is new enough to support explicit
fencing, dynamically grown ringbuffers, etc.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
With a new enough drm/msm, we can let the kernel know about buffers that
are in the bo cache, so the kernel can free them under memory pressure.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Split out interface to allocate from and release to bo-cache, and get
rid of direct usage of bucket level API from fd_bo/etc.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Eventually we'll want a separate bo-cache for ringbuffer bo's, since
ringbuffer bo's get vmap'd on the kernel side, it is preferrable to
re-use them as ringbuffers rather than something else. Plus should
help to add madvise support if it is a bit better decoupled from bo
allocation (next patch).
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Provide a way to insert a reference (ie. OUT_IB()) to a target ring,
executing all the cmds in the target ring from the start.
Sometimes the ringmarker stuff is just overkill. And it will won't
really work properly once we support multiple physical cmdstream buffers
per fd_ringbuffer. So in the future the old ringmarker related APIs
will be deprecated in a few releases.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
When in multi GPU case, devices array may have some
NULL "hole" in between two devices. So check all
array elements and free non-NULL device.
Signed-off-by: Qiang Yu <Qiang.Yu@amd.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Currently drmGetDevice always returns the first device it finds under
/dev/dri/.
Move the target device to the start of the list during iteration. This
way during deduplication it'll preserve its place and will be returned
to the user.
v2: Keep the memory leak separate.
v3: Move the drmFoldDuplicatedDevices description
Signed-off-by: Qiang Yu <Qiang.Yu@amd.com>
[Emil Velikov: move drmFoldDuplicatedDevices description, add
changelog, reword commit message]
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
This header provides major/minor/makedev funcs under most Linux C
libs. Pull it in to fix building with newer versions that drop the
implicit include via sys/types.h.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=94231
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
AOSP master now errors if LOCAL_SRC_FILES contains headers, so filter
out header files from the source lists.
Signed-off-by: Rob Herring <robh@kernel.org>
This is unusual. Usually IDs listed on early stages of platform
definition are kept there as reserved for later use.
However these IDs here are not listed anymore in any of steppings
and devices IDs tables for Kabylake on configurations overview
section of BSpec.
So it is better removing them before they become used in any
other future platform.
v2: Rebase.
Reviewed-by: Dhinakaran Pandiyan <dhinakaran.pandiyan@intel.com>
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Previously, (*device)->businfo.pci would end up misaligned, which results
in undefined behavior.
Signed-off-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
When multi GPU present, after drmFoldDuplicatedDevices
merge same busid deveces, two different devices may be
seperated by zero in local_devices[]. The for loop
should check all local_devices instead of exit when
meet a zero.
Reviewed-by: Jim Qu <Jim.Qu@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Qiang Yu <Qiang.Yu@amd.com>
Make sure that this function is defined (even empty/dummy) when KGSL support is
disabled, since it's part of the driver i/f and it was reported to cause symbols
issues when building against musl libc implementation.
Reported-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Nicolas Dechesne <nicolas.dechesne@linaro.org>
Signed-off-by: Rob Clark <robclark@freedesktop.org>
This needs the kernel patch to make sure the C++ protection is in place.
Otherwise just new defines, using the right fixed-width types and
some shuffling in where stuff is defined (the DRM_CAP list moved, but
the #defines are the same).
Generated using make headers_install.
Generated fromd drm-misc commit 249c4f538b1aae55d41699f8bafc6cb762a7f48f
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Generated using make headers_install.
Only cosmetics&new definitions here now.
Generated fromd drm-misc commit 249c4f538b1aae55d41699f8bafc6cb762a7f48f
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Only real difference is switching to kernel types for fixed-width
integers, like we should.
Generated fromd drm-misc commit 249c4f538b1aae55d41699f8bafc6cb762a7f48f
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Generated using make headers_install. Only difference is a new value
of SAREA_MAX for mips. Not that we ever shipped a dri1 driver on that
platform probably ...
Generated fromd drm-misc commit 249c4f538b1aae55d41699f8bafc6cb762a7f48f
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Generated using make header_install.
Generated fromd drm-intel-next-queued commit
55c561a708eec328822721233b1148119e80f5c3
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
This was purely a kernel-internal type used in a early patch version
to add GEM bo mmap support. It was never used in-kernel in merged code
nor in userspace. Nuke it to align with kernel headers.
For reference the kernel patch:
commit 05f51722a154e73019434bd020e50ddb941046c5
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date: Wed Dec 11 11:34:32 2013 +0100
drm/bufs: remove handling of _DRM_GEM mappings
Gone with the new gem vma offset manager from David.
We can also ditch the uapi header definition from the enum since
userspace never used this. It ended up in there purely for historical
reasons (for reusing the old drm mmap code essentially), not because
userspace ever needed it.
Cc: David Herrmann <dh.herrmann@gmail.com>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Update the UAPI header to the latest version in the Linux kernel. This
changes the struct drm_tegra_gem_mmap to properly handle offsets on 64-
bit architectures.
See commit bdf765071a8b ("drm/tegra: gem: Return 64-bit offset for
mmap(2)") in the Linux kernel (as of v4.1).
Signed-off-by: Thierry Reding <treding@nvidia.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Equivalent to the amdgpu commit before. Additionally, when libdrm is
installed to a 'non-default' location, users of libdrm_radeon will fail
to build, as radeon_cs.h (and maybe others) won't have their
dependencies (drm.h radeon_drm.h) fulfilled.
Cc: Christian König <christian.koenig@amd.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
The headers were originally written in Mesa, imported to the kernel,
and improved upon in vc4-gpu-tools. These come from the v-g-t copies
and will replace the Mesa and v-g-t copies, and hopefully be used from
new tests in igt, as well.
v2: Fix linking against libdrm_intel instead of libdrm.
v3: Drop Libs and Cflags since they'll be inherited from libdrm.
v4: Switch to Requires.private. I was wrong about standard practice,
apparently only Intel was doing plain Requires (sorry to all
involved).
Signed-off-by: Eric Anholt <eric@anholt.net>
I'll build some libdrm C code soon, but for now this lets libdrm users
use vc4 ioctls. Produced from headers_install of
1df59b8497f47495e873c23abd6d3d290c730505 (drm-next) in the kernel.
Signed-off-by: Eric Anholt <eric@anholt.net>
The utils library depends on libdrm. Flip the order, orderwise we might
error during link stage like below:
CC main.o
CCLD kmstest
/usr/bin/ld: ../../tests/util/.libs/libutil.a(libutil_la-kms.o):
undefined reference to symbol 'drmOpen'
Reported-by: Tom Stellard <thomas.stellard@amd.com>
Tested-by: Tom Stellard <thomas.stellard@amd.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Used in compliance with POSIX 2001/2008
Fixes errors e.g.
error: implicit declaration of function 'select'
and helps with missing definitions of FD_* defines
v2: conditionally include sys/select.h, include in every test where
needed.
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Reviewed-by: Thierry Reding <thierry.reding@gmail.com> (v1)
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
util_open() takes a device parameter, followed by a module parameter.
The existing tests used the drmOpen() function, which uses a different
ordering of the parameters, and the old ordering was accidentally kept
during the conversion.
Signed-off-by: Thierry Reding <treding@nvidia.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Use the new util_open() helper instead of open-coding the method for
finding a usable device. While at it, make the command-line interface
more consistent with that of modetest by adding the -D and -M options.
Signed-off-by: Stefan Agner <stefan@agner.ch>
v2: correctly use util_open() - swap device, module
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Currently with --disable-amdgpu --disable-valgrind --disable-cairo-tests
cunit, valgrind and cairo are still detected.
Signed-off-by: Marcin Ślusarz <marcin.slusarz@gmail.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
We are reading at most sizeof(data) bytes, but then data may not contain
a terminating '\0', at least in theory, so strstr() may overflow the
stack allocated array.
Make sure that data always contains at least one '\0'.
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Not doing so makes it impossible for radeon_bo_open callers to set any
RADEON_GEM_* flags for the newly created BO.
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
the semaphore is a binary semaphore. the work flow is:
1. create sem
2. signal sem
3. wait sem, reset sem after signalled
4. destroy sem.
Signed-off-by: Chunming Zhou <david1.zhou@amd.com>
Reviewed-by: Jammy Zhou <Jammy.Zhou@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
drmModeGetConnectorCurrent() must provide temporary storage for the
kernel to fill in at least one mode (asking for !=0 modes is how
you prevent the heavyweight probe in the kernel). Currently we malloc
that temp storage but we fail to free it before overwriting the
pointer with the address of the actual storage we use to store the
real mode list we get from the kernel in the second ioctl call.
Let's just keep the temporary storage on the stack and thus we avoid the
leak and also eliminate some pointless mallocs.
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Fixes: 5ed5fa1060 ("mode: Retrieve only the current information for a Connector")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Because NVIF intentionally lacks some of the paths necessary to be
compatible with various mistakes we've made over the years, libdrm
needs to know whether a client has been updated and that it's safe
to make use of the new kernel interfaces.
Clients still using nouveau_device_open()/wrap() will be forced to
make use of ABI16 instead of NVIF.
v2.
- remove lib_version, nothing used it
- leave client-provided pointer unmodified on failure
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Tested-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
This will expose functionality supported by newer kernel interfaces.
Current userspace uses the chipset to determine which classes are likely
exposed, which generally works pretty well, but isn't as flexible as it
could be.
Unfortunately, the G98:GF100 video code in Mesa is still relying on the
kernel exposing incorrect vdec classes on some chipsets. The ABI16
kernel interfaces have a workaround for this in place, but that will no
longer be available once libdrm supports NVIF.
To prevent a regression when NVIF support is added, if there's no kernel
support for NVIF, libdrm will magic up a class list containing correct
vdec classes anyway instead of failing with -ENODEV.
v2.
- add description of abi16/vdec workaround
- add description of sclass/mclass
- leave client-provided pointer unmodified on abi16_sclass() failure
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Tested-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
This will expose functionality supported by newer kernel interfaces,
giving access to things such as ZBC controls, perfmon, etc.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Tested-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
v2.
- add a comment about the (ab)use of nouveau_object::length
- add a comment about abi16_object() return values
v3.
- handle new client + old kernel for sw classes
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Tested-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
This commit also modifies the install path of the main libdrm_nouveau
header to be under a nouveau/ subdirectory.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Tested-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
The comment has been copied from modetest and is not applicable
for vbltest.
Signed-off-by: Stefan Agner <stefan@agner.ch>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Use the new util_open() helper instead of open-coding the method for
finding a usable device. While at it, this adds -D and -M command-line
options to vbltest to make its usage more consistent with its siblings
modetest and proptest.
Signed-off-by: Thierry Reding <treding@nvidia.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Use the new util_open() helper instead of open-coding the method for
finding a usable device. While at it, make the command-line interface
more consistent with that of modetest by adding the -D and -M options.
Signed-off-by: Thierry Reding <treding@nvidia.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Use the new util_open() helper instead of open-coding the method for
finding a usable device.
Signed-off-by: Thierry Reding <treding@nvidia.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
The new function util_open() encapsulates the standard method employed
by tests to open a device or module. There is a verbatim copy of this in
almost all test programs, with slight variations in the list of modules.
Moving this code into a common helper allows code reuse and makes tests
more consistent.
Signed-off-by: Thierry Reding <treding@nvidia.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
This small program allows universal planes to be tested. Currently this
isn't very flexible because it allows only the first plane of a given
type to be tested on the first CRTC. However it should be simple to
extend this with some additional command-line arguments.
Signed-off-by: Thierry Reding <treding@nvidia.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
This test program sets a mode and framebuffer on a connector and cycles
through all CRTCs, moving the connector to each of them in turn. This is
useful to verify that CRTC stealing is properly handled in the DRM core
and drivers.
Signed-off-by: Thierry Reding <treding@nvidia.com>
This library contains abstractions for KMS that help remove the need for
a lot of boilerplate in KMS test programs.
Signed-off-by: Thierry Reding <treding@nvidia.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Some of the helpers, such as the pattern drawing helpers or the format
lookup helpers, have potential to be reused. Move them into a separate
library to make it easier to share them.
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
To check if g2d_move() works properly we create a small checkerboard
pattern in the center of the screen and then shift this pattern
around with g2d_move(). The pattern should be properly preserved
by the operation (but not the surrounding area).
Tested-by: Hyungwon Hwang <human.hwang@samsung.com>
Reviewed-by: Hyungwon Hwang <human.hwang@samsung.com>
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
[Emil Velikov: add g2d_move to the symbol check]
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
We already have g2d_copy() which implements G2D copy
operations from one buffer to another. However we can't
do a overlapping copy operation in one buffer.
Add g2d_move() which acts like the standard memmove()
and properly handles overlapping copies.
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
This allows setting the two direction registers, which specify how
the engine blits pixels. This can be used for overlapping blits,
which happen e.g. when 'moving' a rectangular region inside a
fixed buffer.
Reviewed-by: Hyungwon Hwang <human.hwang@samsung.com>
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Signed-off-by: Hyungwon Hwang <human.hwang@samsung.com>
This matches the G2D color mode that is used in the entire code.
The previous (incorrect) RGBA8888 would only work since the
Exynos mixer did its configuration based on the bpp, and not
based on the actual pixelformat.
Reviewed-by: Hyungwon Hwang <human.hwang@samsung.com>
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
This tests async processing of G2D jobs. A separate thread is spawned
to monitor the DRM fd for events and check whether a G2D job was
completed.
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Signed-off-by: Hyungwon Hwang <human.hwang@samsung.com>
This enables us to pass command buffers to the kernel which
trigger an event on the DRM fd upon completion.
The final goal is to enable asynchronous operation of the
G2D engine, similar to async page flips.
Passing the event userdata pointer through the G2D context
was chosen to not change the current API (e.g. by adding
a userdata argument to each public functions).
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Currently only fast solid color clear performance is measured.
A large buffer is allocated and solid color clear operations
are executed on it with randomly chosen properties (position
and size of the region, clear color). Execution time is
measured and output together with the amount of pixels
processed.
The 'simple' variant only executes one G2D command buffer at
a time, while the 'multi' variant executes multiple ones. This
can be used to measure setup/exec overhead.
The test also serves a stability check. If clocks/voltages are
too high or low respectively, the test quickly reveals this.
Tested-by: Hyungwon Hwang <human.hwang@samsung.com>
Reviewed-by: Hyungwon Hwang <human.hwang@samsung.com>
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Used to handle kernel events specific to the Exynos platform.
Currently only G2D events are handled.
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Signed-off-by: Hyungwon Hwang <human.hwang@samsung.com>
Softpin allows userspace to take greater control of GPU virtual address
space and eliminates the need of relocations. It can also be used to
mirror addresses between GPU and CPU (shared virtual memory).
Calls to drm_intel_bo_emit_reloc are still required to build the list of
drm_i915_gem_exec_objects at exec time, but no entries in relocs are
created. Self-relocs don't make any sense for softpinned objects and can
indicate a programming errors, thus are forbidden. Softpinned objects
are marked by asterisk in debug dumps.
Cc: Thomas Daniel <thomas.daniel@intel.com>
Cc: Kristian Høgsberg <krh@bitplanet.net>
Cc: Zou Nanhai <nanhai.zou@intel.com>
Cc: Michel Thierry <michel.thierry@intel.com>
Cc: Ben Widawsky <ben@bwidawsk.net>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
Signed-off-by: Kristian Høgsberg <krh@bitplanet.net>
Signed-off-by: Michał Winiarski <michal.winiarski@intel.com>
Gen8+ supports 48-bit virtual addresses, but some objects must always be
allocated inside the 32-bit address range.
In specific, any resource used with flat/heapless (0x00000000-0xfffff000)
General State Heap (GSH) or Instruction State Heap (ISH) must be in a
32-bit range, because the General State Offset and Instruction State Offset
are limited to 32-bits.
The i915 driver has been modified to provide a flag to set when the 4GB
limit is not necessary in a given bo (EXEC_OBJECT_SUPPORTS_48B_ADDRESS).
48-bit range will only be used when explicitly requested.
Callers to the existing drm_intel_bo_emit_reloc function should set the
use_48b_address_range flag beforehand, in order to use full ppgtt range.
v2: Make set/clear functions nops on pre-gen8 platforms, and use them
internally in emit_reloc functions (Ben)
s/48BADDRESS/48B_ADDRESS/ (Dave)
v3: Keep set/clear functions internal, no-one needs to use them directly.
v4: Don't set 48bit-support flag in emit reloc, check for ppgtt type
before enabling set/clear function, print full offsets in debug
statements, using port of lower_32_bits and upper_32_bits from linux
kernel (Michał)
References: http://lists.freedesktop.org/archives/intel-gfx/2015-July/072612.html
Cc: Ben Widawsky <ben@bwidawsk.net>
Cc: Michał Winiarski <michal.winiarski@intel.com>
Signed-off-by: Michel Thierry <michel.thierry@intel.com>
Reviewed-by: Kristian Høgsberg Kristensen <kristian.h.kristensen@intel.com>
Signed-off-by: Kristian Høgsberg Kristensen <kristian.h.kristensen@intel.com>
It defines the prototype of ffs that fixes the building error
on Android 6.0 64-bit image.
Signed-off-by: Chih-Wei Huang <cwhuang@linux.org.tw>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Unlikely that we'll hit a case where __sync_fetch_and_add is present
while __sync_add_and_fetch isn't. Regardless let's keep things sane and
consistent.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
The libdrm autoconf test for atomics uses __sync_val_compare_and_swap with
the address of a function argument which triggers a gcc ICE on sparc64
with the OpenBSD system compiler.
Mark Kettenis pointed out that while other architectures probably spill the
argument onto the stack this is likely not the case on register window
architectures like SPARC and suggested passing a pointer as an argument
instead which avoids the ICE and allows the drm libraries requiring
atomics to build on sparc64 with the autoconf build.
Reported-by: Christian Weisgerber <naddy@openbsd.org>
Signed-off-by: Jonathan Gray <jsg@jsg.id.au>
Reviewed-by: Mark Kettenis <kettenis@openbsd.org>
Reviewed-by: Matthieu Herrb <matthieu@openbsd.org>
Move the allocation of result prior to the IOCTL so we can cleanly
backtrack if the allocation fails.
Signed-off-by: Tom St Denis <tom.stdenis@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
In the function amdgpu_vamgr_find_va() the function would return
without unlocking the mutex if the base_required offset was below
the va managers base offset.
Signed-off-by: Tom St Denis <tom.stdenis@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Both drmGetDevice() and drmGetDevices() currently print a warning when
they encounter an unknown (non-PCI) subsystem type for a device node,
but they still proceed to assume that the drmDevicePtr was initialized
and try to add it to the local device array. Add a 'continue' to the
error case handling to bypass the rest of the processing for devices we
can't handle.
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Some of the error conditions in drmGetDevice() can lead to us calling
closedir(NULL) or leaking memory. Fix these conditions the same way we
did for drmGetDevices() in commit:
commit 8c4a1cbd98
Author: Matt Roper <matthew.d.roper@intel.com>
Date: Wed Sep 30 09:30:51 2015 -0700
xf86drm: Fix error handling for drmGetDevices()
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Arithmetic on void pointers is a GCC extension.
CC libdrm_la-xf86drm.lo
../xf86drm.c: In function 'drmProcessPciDevice':
../xf86drm.c:3017:10: warning: pointer of type 'void *' used in arithmetic [-Wpointer-arith]
addr += sizeof(drmDevice);
^
../xf86drm.c:3020:10: warning: pointer of type 'void *' used in arithmetic [-Wpointer-arith]
addr += DRM_NODE_MAX * sizeof(void *);
^
../xf86drm.c:3023:14: warning: pointer of type 'void *' used in arithmetic [-Wpointer-arith]
addr += max_node_str;
^
../xf86drm.c:3035:14: warning: pointer of type 'void *' used in arithmetic [-Wpointer-arith]
addr += sizeof(drmPciBusInfo);
^
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
If the opendir() call in drmGetDevices() returns failure, we jump to an
error label that calls closedir() and then returns. However this means
that we're calling closedir(NULL) which may not be safe on all
implementations. We are also leaking the local_devices array that was
allocated before the opendir() call.
Fix both of these issues by jumping to an earlier error label (to free
local_devices) and guarding the closedir() call with a NULL test.
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
[Emil Velikov: make the teardown symmetrical, remove the NULL check]
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Back when this was introduced commit 569da5a42eb(Merged glxmisc-3-0-0)
sys/sysmacros.h was used instead of the respecive headers (as per the
manual).
We've been handling it correctly for a little while now - in Linux, BSD
and Solaris. Thus we can drop this workaround.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
All functions from the public API only operation on
struct g2d_context*, so this shouldn't break too much.
Make the context private since we don't want the
user to modify its content directly. Also remove
the defines that were only used for fields of
g2d_context.
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Add a prefix to the messages printed to the console via
printf() and fprintf() so that one can easily see where
the message comes from.
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
The function currently checks for each added command
if an overflow of the corresponding command buffers
occurs, but none of the callers ever checks the
return value.
Since all callers are now converted to use
g2d_check_space() simplify the function.
(1) The overflow checks become asserts, so they're only
active for debug builds. This is fine since
g2d_add_cmd() is not part of the public API.
(2) Switch the return value to void.
(3) Explicitly state that the caller has to check
buffer space before calling g2d_add_cmd().
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
The g2d_point_val union consists of two coordinates of 16
bits. Whenever this union is used though, both coordinates
are explicitly set. Hence prior initialization is unnecessary.
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
We now validate the blending mode via g2d_validate_mode()
prior to feeding it to g2d_get_blend_op().
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
The G2D headers define a number of modes through enums
(like e.g. color, select, repeat, etc.).
This introduces g2d_validate_select_mode() and
g2d_validate_blending_op() which validate a
select mode or blending operation respectively.
Use this together with g2d_check_space() in
g2d_{blend,scale_and_blend}().
For this we move parameter validation to the top and
also validate the select mode of the source image and
the requested blending operation before starting
command submission.
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
This is going to be used to check if the command buffers have
enough space left prior to actual submission of the commands.
Use this in g2d_{solid_fill,copy,copy_with_scale}().
For this the parameter validation before buffer space
checking so that we can exit early if it fails.
Also don't reset the G2D context in this situation since
the buffers are not partially submitted anymore.
The repeat mode in g2d_copy_with_scale() is checked first
to make computation of space easier.
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Similar interface to the *Devices() ones but they obtain/free the
information of the opened device (as given by its fd).
Note there is a fair bit of duplication between the two Get functions,
and anyone interested is more than welcome to consolidate it.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Do a once off memory allocation for each drmDevice.
This allows us to ease the error handling and simplify the
de-duplication loop. As part of this we need to rework drmFreeDevice()
such so that it frees the relevant hunks, rather than leaving that to
the caller.
Some memory stats from the drmdevice test
before: 22 allocs, 22 frees, 66,922 bytes allocated
after: 9 allocs, 9 frees, 66,436 bytes allocated
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Move away form the boolean name, change the return value
appropriately and check if either argument is NULL.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
This will allow one to reuse the core drmGetDevices implementation on
other platforms. Keeping all the platform specifics in ParseFoo.
On the plus side this saves a bit of code :)
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
As going through the modetest patches for atomic support I've noticed
that if we pass NULL for the drmModeAtomicReqPtr argument we'll crash.
So let's handle things appropriately if the user forgot to check the
return value of drmModeAtomicAlloc and drmModeAtomicDuplicate or made a
typo somewhere along the way.
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Cc: Rob Clark <robclark@freedesktop.org>
Cc: Daniel Stone <daniels@collabora.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Fixes "error: implicit declaration of function 'alloca'" failures
when building on Solaris
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
We're about to remove the -Wno flag from configure.ac which will lead
to a lot of unnecessary spam.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
In the latest version of CUnit the fourth parameter of the CU_SuiteInfo
struct is pSetUpFunc rather than *pTests.
Seems like the CUnit ABI broke at some point, so let's the the robust
thing and use c99 designated initializers to correctly populate the
struct(s).
Cc: Leo Liu <leo.liu@amd.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
...to minimise misuse of bo_gem.
If the variable is declared at the top of the function and then used
for two (or more) different contexts this can cause confusion and errors.
Just introduce a wrapper, which can be used in a once off situations.
Suggested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by Chris Wilson <chris@chris-wilson.co.uk>
For android / drm_gralloc, we want to hook up our own debug_print()
without bothering with the reset of it.
Signed-off-by: Rob Clark <robdclark@gmail.com>
EGL_EXT_image_dma_buf_import specifies that the importer retains
ownership of the fd, rather then the importee.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
EBADMSG is a streams errno. OpenBSD does not implement streams and does
include the streams errnos, this commit fixes the build on OpenBSD.
None of the callers of this function check the return value for -EBADMSG.
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Jonathan Gray <jsg@jsg.id.au>
fixes the prime sharing race condition described by
"intel: Serialize drmPrimeFDToHandle with struct_mutex".
we inline fd_bo_from_handle() into fd_bo_from_dmabuf() and allow locking.
Signed-off-by: Varad Gautam <varadgautam@gmail.com>
Signed-off-by: Rob Clark <robclark@freedesktop.org>
don't call drmIoctl() directly for prime bo's, use the wrappers instead.
v3: remove struct drm_prime_handle and split locking
Signed-off-by: Varad Gautam <varadgautam@gmail.com>
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Commit c86dabfc9f ("omap: zero is a valid fd number, treat it as
such") corrected checks for valid file descriptors, but the OMAP buffer
object code initializes the DMA-BUF file descriptor to 0 (as a result of
calloc()'ing the structure). Obviously this isn't going to work because
subsequent code will try to use file descriptor 0 (most likely stdin at
that point) as a DMA-BUF. It may also try and close stdin when a buffer
object is destroyed.
Fix this by initializing the DMA-BUF file descriptor to -1, properly
marking it as an invalid file descriptor.
Fixes: c86dabfc9f ("omap: zero is a valid fd number, treat it as such")
Reported-by: Robert Nelson <robertcnelson@gmail.com>
Tested-by: Robert Nelson <robertcnelson@gmail.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Fixes the same problem as "intel: Serialize drmPrimeFDToHandle with struct_mutex".
Reviewed-by: Jammy Zhou <Jammy.Zhou@amd.com>
Signed-off-by: Christian König <christian.koenig@amd.com>
Last commit (b556ea127e) introduced
use of log2 which require -lm flag for the linker on quite few
distribution. Just add that flag to fix build.
Signed-off-by: Jérôme Glisse <jglisse@redhat.com>
For mutiple GPU support, the devices on the system should be enumerated
to get necessary information about each device, and the drmGetDevices
interface is added for this. Currently only PCI devices are supported for
the enumeration.
Typical usage:
int count;
drmDevicePtr *foo;
count = drmGetDevices(NULL, 0);
foo = calloc(count, sizeof(drmDevicePtr));
count = drmGetDevices(foo, count);
/* find proper device, open correct device node, etc */
drmFreeDevices(foo, count);
free(foo);
v2: [Jammy Zhou]
- return a list of devices, rather than nodes
v3: [Jammy Zhou]
- fix the signed extension for PCI device info
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Jammy Zhou <Jammy.Zhou@amd.com>
Importing a bo whose handle is still in the bo cache crashes during cleanup.
Remove bo from cache when importing.
Signed-off-by: Varad Gautam <varadgautam@gmail.com>
Signed-off-by: Rob Clark <robclark@freedesktop.org>
It is possible to hit a race condition in create_from_prime, when trying
to import a BO that's currently being freed. In case of prime sharing
we'll succesfully get a handle, but fail on get_tiling call, potentially
confusing the caller (and requiring different locking scheme than with
sharing using flink). Wrap fd_to_handle with struct_mutex to force
a more consistent behaviour between prime/flink, convert fprintf to DBG
when handling errors.
(From Chris:
The race is that the kernel returns us the same file-private handle as
the first thread, but that first thread is about to call gem_close
(thereby removing the handle from the file completely) and does so
between us acquiring the handle and taking the mutex. If we take
the mutex, then we acquire the refcnt on the bo prior to the first
thread completing its unref (and so preventing the early close). Or we
acquire the handle after the earlier close, in which case we are the new
owner.
)
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Testcase: igt/drm_import_export/import-close-race-prime
Signed-off-by: Rafał Sapała <rafal.a.sapala@intel.com>
Signed-off-by: Michał Winiarski <michal.winiarski@intel.com>
Use tabs and spaces consistently to align function arguments on
subsequent lines with those of the first line.
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Usage of blank lines can be a matter of taste, of course, but for these
we can surely all agree that they're not needed and inconsistent.
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Each device can have its own vamgr, so make it per device now.
This can fix the failure with multiple GPUs used in one single
process.
v2: rebase
Signed-off-by: Jammy Zhou <Jammy.Zhou@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
The AMDGPU_VA_RANGE_32_BIT flag is added to request VA range in the
32bit address space for amdgpu_va_range_alloc.
The 32bit address space is reserved at initialization time, and managed
with a separate VAMGR as part of the global VAMGR. And if no enough VA
space available in range above 4GB, this reserved range can be used as
fallback.
v2: add comment for AMDGPU_VA_RANGE_32_BIT, and add vamgr to va_range
v3: rebase to Emil's drm_private series
v4: fix one warning
Signed-off-by: Jammy Zhou <Jammy.Zhou@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
User should only see these with LIBGL_DEBUG=verbose. But in case you
are hitting issues like "handle X at index Y already on submit list"
errors from the kernel, this gives some useful visibility for debug.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
It should be a less common case, but it is possible for a single bo to
be on multiple rings, for example when sharing a buffer across multiple
pipe_context's created from same pipe_screen.
So rather than completely fall over in this case, fallback to slow-path
of looping over all bo's in the ring's bo-table (but retain the fast-
path of constant-lookup for the first ring the buffer is on).
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Group the parts related to building out submit ioctl into their own
sub-struct. Split out from next commit since it is just boring churn.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Allow connector names to be used in the specification of the -s option.
This requires storing the string passed on the command-line so that it
can later be resolved to a connector ID (after the DRM device has been
opened).
Connector names are constructed from the connector type name and
connector type ID using the same format as used internally in the
Linux kernel.
Signed-off-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
We need this include in porting changes for the OpenGL ES
conformance suite.
v2: remove c_plusplus usage
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
The amdgpu_device for a device node needs its own dup'ed fd, instead
of using the original fd passed in for a screen, to make multi-x-screen
ZaphodHeads configurations work on amdgpu.
The original fd's lifetime differs from that of the amdgpu_device, and from the
one stored in the hash. The hash key is the fd, and in order to compare hash
entries we fstat them, so the fd must be around for as long as the amdgpu_device
is.
This patch for libdrm/amdgpu is a translation of the radeon-winsys ZaphodHeads
fix for mesa's radeon-winsys, from mesa commit 28dda47ae4d974e3e032d60e8e0965c8c068c6d8
"winsys/radeon: Use dup fd as key in drm-winsys hash table to fix ZaphodHeads."
Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com>
Acked-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
v2: Check for valid fd's being >= 0, because fd == 0 is in theory
a valid, although unlikely, fd and fd == -1 would denote an
invalid fd. Thanks to William Lewis for pointing this out.
Reported-by: William Lewis <minutemaidpark@hotmail.com>
Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com>
The following interfaces are changed accordingly:
- amdgpu_bo_alloc
- amdgpu_create_bo_from_user_mem
v2: update the interfaces
v3: remove virtual_mc_base_address from amdgpu_bo
Signed-off-by: Jammy Zhou <Jammy.Zhou@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
The flags is added for extensibility to cover some special requirements
in the future, i.e, request VA range in the first 4GB of address space
Signed-off-by: Jammy Zhou <Jammy.Zhou@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
If the value is 64bit, but the alignment is 32bit type, the high 32bit
will be clamped with previous definition
Signed-off-by: Jammy Zhou <Jammy.Zhou@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
This is safe now because the ioctl uses an absolute timeout.
This prevents amdgpu_cs_query_fence_status from returning early e.g.
when a signal is delivered, which in turn caused Mesa winsys code to
assume a BO was idle when it actually wasn't yet.
Reviewed-by: Christian König <christian.koenig@amd.com>
amdgpu_va_range_query interface is added so that client can query va
range supported by specific device.
Signed-off-by: Sabre Shao <Sabre.Shao@amd.com>
Reviewed-by: Jammy Zhou <Jammy.Zhou@amd.com>
Two new interfaces are added to support client request for
allocate virtual address without physical memory committed to.
The virtual address space can be managed by client itself.
Signed-off-by: Sabre Shao <Sabre.Shao@amd.com>
Signed-off-by: Ken Wang <Qingqing.Wang@amd.com>
Reviewed-by: Jammy Zhou <Jammy.Zhou@amd.com>
base_preferred parameter is added to amdgpu_vamgr_find_va
so UMD can specify preferred va address when allocating.
Signed-off-by: Ken Wang <Qingqing.Wang@amd.com>
Reviewed-by: Jammy Zhou <Jammy.Zhou@amd.com>
Useful when Mesa wants to wait for a lot of fences at the same time and
doesn't want to recalculate the relative timeout after every call.
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Fix some style problems, adjust to a common indentation, reorder two
function definitions and remove stale comments.
No intended functional change.
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Jammy Zhou <Jammy.Zhou@amd.com>
Alternative solution to Mareks patch to stop causing trouble with render nodes.
v2: rebased
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
v1: by Jammy Zhou
v2: remove bo wait when destroy IB by Jammy Zhou
v3: more cleanups by Marek
Reviewed-by: Jammy Zhou <Jammy.Zhou@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
This is the first sub-patch of va interface task, the va task is
about adding more va management interfaces for UMD, by design, the
vamgr should be per-process rather than per-device.
Signed-off-by: Ken Wang <Qingqing.Wang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
A new 'offset' flag is added to the amdgpu_cs_ib_info structure,
which can be used to specify location of PM4 packets to execute
in the IB buffer object
v2: remove the shared IB flag, etc
Signed-off-by: Jammy Zhou <Jammy.Zhou@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Less code and also provides the map_size parameter.
v2: Also set offset_in_bo, use *_handle defines.
Signed-off-by: Christian König <christian.koenig@amd.com>
use heap memory instead of stack memory to avoid potential stack overflow
when a large number of resources are used for the bo_list.
v2: some minor improvement
Signed-off-by: Jammy Zhou <Jammy.Zhou@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
va+size will overflow if va equals to AMDGPU_INVALID_VA_ADDRESS, just
return when hit that condition
Signed-off-by: monk.liu <monk.liu@amd.com>
Reviewed-by: Jammy Zhou <jammy.zhou@amd.com>
0xffffffff was appended to the higher 32bit with (intptr_t) when use
32bit libdrm_amdgpu.so with 64bit kernel, and it caused segmentation
fault for 32bit application.
v2: switch to uintptr_t
Signed-off-by: Jammy Zhou <Jammy.Zhou@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
The active RB pipes can be retrieved from enabled_rb_pipes_mask,
for which each bit indicates one active pipe if it is '1'.
Signed-off-by: Jammy Zhou <Jammy.Zhou@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Instead of taking a look at the device major/minor we
just compare the primary device name to figure out if
two fds are pointing to the same device.
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Christian König <christian.koenig@amd.com>
original method is just totally wrong, it loses the bo_list handler at
all after command stream accomplished
Signed-off-by: Monk.Liu <monk.liu@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This adds some basic unit tests for the new amdgpu driver.
v2: use common util_math.h
v3: implement suggestions from Emil
replace malloc/memset with calloc
make header guards all caps
use posix_memalign rather than mem_align
replace malloc with calloc for pm4 allocations
make CU_SuiteInfo static
fix Makefile.am
fix fd numbering
use drmGetVersion/drmFreeVersion rather than open coding it
close fd, clean up CU registry on error
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This is the new ioctl wrapper used by the new admgpu driver.
It's primarily used by xf86-video-amdgpu and mesa.
v2: fix amdgpu_drm.h install
v3: Integrate some of the sugestions from Emil:
clean up Makefile.am, configure.ac
capitalize header guards
fix _FILE_OFFSET_BITS with config.h
use drm_mmap/drm_munmap
Remove unused ARRAY_SIZE macro
use shared list implementation
use shared math implementation
use drmGetNodeTypeFromFd helper
v4: remove unused tiling defines
v5: include amdgpu.h in Makefile.am
v6: update amdgpu_drm.h
v7: libdrm.h -> libdrm_macros.h
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This is used by radeon and freedreno and will be used
by amdgpu. I looked at switching to libdrm_lists.h,
but it's pretty horrible. E.g., DRMLISTFOREACHENTRYSAFE.
v2: remove missed list.h from tests/radeon/Makefile.am
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
We now have a separate tool for this in intel-gpu-tools and we don't
need to clutter up libdrm with this feature. We leave the entry points
in there to avoid breaking API/ABI.
Install intel-gpu-tools, then run (for example)
$ intel_aubdump --output=trace.aub glxgears -geometry 500x500
See the intel_aubdump man page for more details.
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Kristian Høgsberg Kristensen <kristian.h.kristensen@intel.com>
Add defines for the device minor names and make use of them
in drmGetMinorName() so the correct paths will be used on OpenBSD.
v2: don't add new defines to xf86drm.h to keep them out of the API
as requested by Emil.
Signed-off-by: Jonathan Gray <jsg@jsg.id.au>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
As far as I can tell no OpenBSD platform ever used 81
for a drm major. While the value was added to libdrm in 2003
or earlier drm didn't appear in OpenBSD till 2007.
Of the OpenBSD platforms that support drm amd64/macppc/sparc64
use a major of 87, i386 uses 88.
v2: rearrange ifdefs as suggested by Emil.
Signed-off-by: Jonathan Gray <jsg@jsg.id.au>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
This is implemented with kms ioctls so it could also be used as a
generic fallback.
Signed-off-by: Jonathan Gray <jsg@jsg.id.au>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
The whole thing is quite messy - the file is used to indicate that the
man pages were correctly generated prior to applying the "fixup" (alias)
At the same time we use a rule with the same name, to create the same
file if the generation has failed.
In other words - it attempts to create the file either way. So there is
little point in it and we can remove it.
Spotted while attempting to build with bmake which kindly blocked on the
following (non compliant construct)
.man_fixup: | $(miscman_DATA)
Cc: Jonathan Gray <jsg@jsg.id.au>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Spotted by looking for similar "let's assume fd == 0 is invalid" bugs.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Thierry Reding <treding@nvidia.com>
Abeit quite unlikely to get hit by this bug here, let just fix it.
v2: Correct conditional (do not call ioctl(DRM_IOCTL_PRIME_HANDLE_TO_FD)
when we already have the fd).
v3: Fix kgsl_pipe.c, suggested by Thierry.
Cc: freedreno@lists.freedesktop.org
Cc: Rob Clark <robdclark@gmail.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Thierry Reding <thierry.reding@gmail.com>
Linux seems to pick these up via another header, but Solaris needs them
explicitly included, or we get undefined symbol errors for major & minor.
v2: use headers documented in makedev(3C) man page instead of sysmacros.h
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
v3 [Emil Velikov]: include sys/mkdev.h only when available.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Add sys/sysctl.h to get sysctlbyname declaration on kFreeBSD
Updated by Thorsten “mirabilos” Glaser <t.glaser@tarent.de>
to add autoconf check and only include <sys/sysctl.h> if it
is detected by configure as it’s unusable on Linux/x32 (and
others, e.g. other new architectures).
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
In case of YF/YS tiled buffers libdrm need not know about the tiling
format because these buffers don't have hardware support to be tiled
or detiled through a fenced region. But, libdrm still need to know
about buffer alignment restrictions because kernel uses it when
resolving the relocation.
Mesa uses drm_intel_gem_bo_alloc_for_render() to allocate Yf/Ys buffers.
So, use the passed alignment value in this function to initialize the
align variable in drm_intel_bo. Note that we continue ignoring the
alignment value passed to drm_intel_gem_bo_alloc() to follow the
previous behavior.
V2: Add a condition to avoid allocation from cache. (Ben)
V3: Make no changes in cache allocation strategy. Just update the alignment.
Update the aperture size estimate including the alignment. (Ben, Chris)
V4: Move aperture size adjustments inside drm_intel_bo_gem_set_in_aperture_size()
Don't split sentences across the one-line header and the changelog. (Chris)
Signed-off-by: Anuj Phogat <anuj.phogat@gmail.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
This fixes a compiler warning about missing handling of enum
values in the switch statements.
Also remove the silent mapping to G2D_IMGBUF_GEM when an
unknown buffer type is encountered. We have full control
about the type here, and if it's unknown then we obviously
have a bug in the code.
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Acked-by: Emil Velikov <emil.l.velikov@gmail.com>
No test uses DRM planes at the moment so this function
is never called. Inspection of the git history shows
that DRM planes were also never used in these tests
in the past.
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Acked-by: Emil Velikov <emil.l.velikov@gmail.com>
It doesn't make sense to keep this structure, since we
can just call all tests directly. An inspection of the
git history shows that no code ever used this
abstraction in the past.
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Acked-by: Emil Velikov <emil.l.velikov@gmail.com>
Remove all unused struct members. An inspection of the
git history shows that these members were also never
used in the past.
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Acked-by: Emil Velikov <emil.l.velikov@gmail.com>
The 'usage' function already does exit(0), so that this
'return -EINVAL' is never called. Just put a break there
to avoid confusion.
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Acked-by: Emil Velikov <emil.l.velikov@gmail.com>
Even if flushing the command buffer doesn't succeed, the
G2D calls would still return zero. Fix this by just passing
the flush return code.
In fact error handling currently ignores the fact that
g2d_add_cmd() can fail. This is going to be handled
in a later patch.
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Acked-by: Emil Velikov <emil.l.velikov@gmail.com>
Add support for the atomic modesetting ioctl through a property-set API.
v1: Squashed intermediate patches from Ville, Rob and myself. Updated
for current kernel interface (no blobs).
v2: Rewrite user-facing API to provide transactional/cursor interface.
Use memclear to zero out ioctl.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Signed-off-by: Daniel Stone <daniels@collabora.com>
v3 [Emil Velikov]: Remove DRM_CAP_ATOMIC - superseded by
DRM_CLIENT_CAP_ATOMIC.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
This reverts commit fde4969176.
The commit adds an API that does not seem flexible enough to be used in
current open-source projects. Additionally it adds a hidden dependency
of libudev, which when used in mesa caused grief when combined with
Steam('s runtime).
Let's revert this for now and add a tweaked API later on that can be
used in mesa/xserver.
Cc: Frank Min <frank.min@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Jammy Zhou <Jammy.Zhou@amd.com>
This should help catch odd bugs at the callsites rather than much later
on with completely bogus bo indices.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Don't assume that a plane supports any kind of pixelformat
but do a check first.
v2: Simplify the format check.
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
This enables us to check for overlay planes which are located
'below' the primary plane.
Since the alpha value only has an effect when creating surfaces
with an alpha-pixelformat this doesn't affect the regular
XRGB8888 primary surface.
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Acked-by: Emil Velikov <emil.l.velikov@gmail.com>
xf86drm.c:356:2: warning: comparison of unsigned expression >= 0 is always true [-Wtype-limits]
group = (serv_group >= 0) ? serv_group : DRM_DEV_GID;
^
v2: do 'int' cast to fix the warning
Signed-off-by: Jammy Zhou <Jammy.Zhou@amd.com>
Reviewed-by: Jan Vesely <jan.vesely@rutgers.edu>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Add a flag allowing Nouveau to specify that an object should be coherent
at allocation time. This is required for some class of objects like
fences which are randomly-accessed by both the CPU and GPU. This flag
instructs the kernel driver to make sure the object remains coherent
even on architectures for which coherency is not guaranteed by the bus.
Reviewed-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Alexandre Courbot <acourbot@nvidia.com>
Lost in 5ea6f1c326, triggering fdo#89842.
Unlike the PRIME fd->handle interfaces, the GEM_OPEN interface doesn't
do anything at the kernel level to prevent this situation occuring,
and we end up with multiple GEM handles for a single kernel buffer.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
... in order to limit the exported symbols only to the required ones.
Both compilers used with Android (GCC and LLVM) support this, so set it
unconditionally.
Cc: Chih-Wei Huang <cwhuang@linux.org.tw>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
In order to use userptr, the kernel tracks the owner's mm with a
mmu_notifier. Setting that is very expensive - it involves taking all
mm_locks and a stop_machine(). This tracking lives only for as long as
the client is using userptr objects - so if the client allocates then
frees a userptr in a loop, we will be executing that heavyweight setup
everytime. To ammoritize this cost, just leak the test bo and the single
backing page we use for detecting userptr.
v2: Free the object and memory when bufmgr is destroyed.
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Currently we are missing the bo_destroy() when modetest terminates.
Signed-off-by: Joonyoung Shim <jy0922.shim@samsung.com>
[Emil Velikov: Tweak the commit message.]
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Currently we don't destroy buffer and remove framebuffer for
planes when closing modetest.
Signed-off-by: Joonyoung Shim <jy0922.shim@samsung.com>
[Emil Velikov: Tweak the commit message. fb_id = 0 is unused]
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Remove the framebuffer and destroy the bo when error occurs on set_mode
and test_page_flip.
Signed-off-by: Joonyoung Shim <jy0922.shim@samsung.com>
[Emil Velikov: Tweak the commit message.]
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
We should remove the framebuffer before destroying the buffer.
Signed-off-by: Joonyoung Shim <jy0922.shim@samsung.com>
[Emil Velikov: Tweak the commit message. fb_id = 0 is unused]
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
If use -P option without -s option, the program segfaults due to
dev.mode.bo being NULL.
Signed-off-by: Joonyoung Shim <jy0922.shim@samsung.com>
[Emil Velikov: Tweak the commit message.]
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Only the 'offsets' array was initialized to zero.
Since bo_create only sets the handles which are
necessary, were we passing garbage data to the
kernel when calling drmModeAddFB2 later.
The issue only seems to appear when passing e.g.
NV12 data to the kernel, a case where not only
handles[0] is used. I therefore also removed the
corresponding comment.
v2: Do the same for set_mode(), set_cursors()
and test_page_flip().
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
A couple of files use ffs() without explicitly including strings.h.
Some systems will pull in ffs()'s declaration through another header
anyway, but not when compiling against bionic in AOSP master.
Signed-off-by: Greg Hackmann <ghackmann@google.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Remove GNU make specific constructs and take into consideration that
Solaris man 7 is not the same as Linux man 7.
This commit introduces a dependency of xorg-macros 1.12 (released 4+
years ago) which is used to handle the above man section discrepancies.
Cc: Niveditha Rau <niveditha.rau@oracle.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Required by intel and drmstat at least. Considering that every compiler
used to build libdrm is C99 compatible, just enable it for the whole
build.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
The function was never part of the public API and a release or so back
was hidden from the global name-space (list of exported symbols).
According to git log this function was never used internally.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Added with commit 57b4c4c32d3(Move the renaming of mm.c symbols to
symbol duplication/collision with ones that are available elsewhere.
As the public/private symbols of libdrm are properly annotated neither
one of the symbols will end up in the global name-space, thus should no
longer be required.
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
They are less and easier to track than the public ones. The macro
drm_public will be going away by the end of the series.
Cc: Rob Clark <robdclark@gmail.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Will allow us to catch when the library exports more symbols than
the ones in the public headers.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
The functions(files) are used if one explicitly modifies radeon_cs_gem.c
by setting CS_BOF_DUMP to 1. As bof.[ch] is used (copied) to other
out-of-tree projects, keep them around in the distribution tarball.
Cc: Jerome Glisse <jglisse@redhat.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
The former does not imply the latter and vice-versa. One such example is
the Sun compiler.
v2: Add missing closing brakets. (Alan)
Cc: Alan Coopersmith <alan.coopersmith@oracle.com>
Cc: Thierry Reding <treding@nvidia.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Running __mmu_notifier_register() is surprisingly expensive, so let's
not do that unless we have to.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Add a new API that allows the caller to skip any forced probing, which
may require slow i2c to a remote display, and only report the currently
active mode and encoder for a Connector. This is often the information
of interest and is much, much faster than re-retrieving the link status
and EDIDs, e.g. if the caller only wishes to count the number of active
outputs.
v2: Fix error path to avoid double free after a failed GETCONNECTOR
ioctl.
v3: Daniel strongly disapproved of my disjoint in behaviour between
GetConnector and GetConnectorCurrent, and considering how best to make a
drop in replacement for drmmode_output_init() convinced me keeping the
API as consistent as possible was the right approach.
v4: Avoid probing on the second calls to GETCONNECTOR for unconnected
outputs.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.com>
Cc: Damien Lespiau <damien.lespiau@intel.com>
Cc: David Herrmann <dh.herrmann@googlemail.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Unfortunately, there are some users of libdrm installed headers that like
to be built with -std=c89 -pedantic, which does not like "inline".
However, __inline works.
Signed-off-by: Daniel Kurtz <djkurtz@chromium.org>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
The valies are unsigned long, thus we should use %lu.
v2: Drop old printf statement. (Jan)
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Jan Vesely <jan.vesely@rutgers.edu>
... and remove the useless SL_DEBUG and RANDOM_DEBUG
v2: Rebase on earlier changes.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Jan Vesely <jan.vesely@rutgers.edu>
... and wire it up to make check
v2: s/rand - state->check/rand != state->check/. (Jan)
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Jan Vesely <jan.vesely@rutgers.edu>
With follow up commits we can clear it up and wire to
make check
v2:
- Use xf86drmRandom.h for common struct.(Jan)
- Add test to .gitignore.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Jan Vesely <jan.vesely@rutgers.edu>
... and wire up to `make check' now that it's useful.
v2: Really return non-zero on failure.
v3: Initialise ret.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Jan Vesely <jan.vesely@rutgers.edu> (v2)
Get the test from completely broken to working like a charm.
- Use the same variable type for both HashInsert and HashLookup.
- Use correct storage type for the HashLookup return value.
- Remove useless backward iteration of HashLookup(i).
v2:
- Use void * instead of unsigned long.
- Change value to key << 16 | key.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Jan Vesely <jan.vesely@rutgers.edu>
This way with follow up commits we can fix it and wire it up to
make check
v2:
- Use xf86drmHash.h for common structs.(Jan)
- Add test to .gitignore.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Jan Vesely <jan.vesely@rutgers.edu>
v2: merge tests creation and xf86drmSL cleanup
rename tests/drmsltest -> tests/drmsl
move the test out of libudev test block
v3: run test even on noudev builds
Signed-off-by: Jan Vesely <jan.vesely@rutgers.edu>
Acked-by: Emil Velikov <emil.l.velikov@gmail.com>
Commit e4a519635f:
Tidy up compile warnings by cleaning up types.
removed call to SLLocate which gutted the function of all functionality.
This patch restores the original behavior, with an additional fix
that zeros the update array in case SLLocate bails early.
v2: zero the update array instead of checking the return value.
SLLocate returns NULL both on failure and if the element is greater
than everything in the list
v3: Improve commit message
Signed-off-by: Jan Vesely <jan.vesely@rutgers.edu>
Acked-by: Emil Velikov <emil.l.velikov@gmail.com>
Both android-x86 and android-ia versions of libpciacccess correctly
"export" the include. If anyone else is wrapping up their own version
they should do so as well.
Remove this fixed location hack from the build.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Chih-Wei Huang <cwhuang@linux.org.tw>
v2: Remove the handler function instead of commenting out
split debugmsg function removal to a separate patch
Signed-off-by: Jan Vesely <jan.vesely@rutgers.edu>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
There is a rockchip drm kms driver.
Add "rockchip" to the static lists of driver names in the the standard
set of tests.
Signed-off-by: Daniel Kurtz <djkurtz@chromium.org>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
[Emil Velikov: Resolve trivial conflicts.]
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
--enable-install-test-programs allows tests to be installed in $bindir.
This is disabled by default, but very useful when cross compiling.
Signed-off-by: Daniel Kurtz <djkurtz@chromium.org>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Use android build system functions to include Android.mk
of subdirs.
Signed-off-by: Chih-Wei Huang <cwhuang@linux.org.tw>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
[Emil Velikov: Resolve trivial conflicts.]
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
With earlier changes we've implicitly add the relevant directories
to the includes list, via LOCAL_EXPORT_C_INCLUDES_DIRS.
v2: Update the top Android.mk as well.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Each of the libdrm_${hw} modules pull libdrm for linking as such:
libdrm's LOCAL_EXPORT_C_INCLUDE_DIRS are added to the includes list.
The former of which is already set to ${top} and ${top}/include/drm.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
- Don't add ${hw}/${hw}, but ${hw} to the includes path. The former
does not exist.
- Set the variable for libkms.
Inspired by the work of from Chih-Wei from the Android-x86 project.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
It seems that the tests don't need DRM master anymore? at least make
distcheck passes when X is running.
release.sh is also invoked with just the path to the libdrm git checkout
and we don't want to pass additional arguments that will be treated as
additional modules we want to release.
Also, make a note that release.sh will run make distcheck for you, so we
don't strickly need to run it beforehand.
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
The main incentive to do this is to get I915_PARAM_REVISION.
v2: Rebase on top of some changes that were made to the header without
copying the whole file from the kernel source.
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Neil Roberts <neil@linux.intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Update kernel interface with new I915_GETPARAM ioctl entries for
subslice total and EU total. Add a wrapping function for each
parameter. Userspace drivers need these values when constructing
GPGPU commands. This kernel query method is intended to replace
the PCI ID-based tables that userspace drivers currently maintain.
The kernel driver can employ fuse register reads as needed to
ensure the most accurate determination of GT config attributes.
This first became important with Cherryview in which the config
could differ between devices with the same PCI ID.
The kernel detection of these values is device-specific. Userspace
drivers should continue to maintain ID-based tables for older
devices which return ENODEV when using this query.
v2: remove unnecessary include of <stdbool.h> and increment the
I915_GETPARAM indices to match updated kernel patch.
For: VIZ-4636
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Jeff McGee <jeff.mcgee@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
The former is a subset of the latter. Error out early so the user is
aware that they are doing something very wrong.
Cc: Rob Clark <robdclark@gmail.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Rob Clark <robdclark@gmail.com>
The message "enabled on x86" was meant for the intel libdrm.
Take the opportunity to mention how libkms is autodetected.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Previous code was busted, as it wasn't checking directly for what it was
meant to, and at the end changing the user's selection if host_cpu
heuristics were involved.
Simplify things by adding a macro that does the long message printing
for us, and check for only what we need.
This fixes commit 36cff14bb03(configure: omap, freedreno and tegra
require atomics) which incorrectly assumed that the code was working
fine, and effectively made impossible to enable freedreno due to it's
host_cpu detection.
Cc: Rob Clark <robdclark@gmail.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
AM_MAINTAINER_MODE can be used to disable generation of rebuild rules.
This is not something we want to condone/support, considering it can
cause greater problems than the perceived benefits. Additionally the
Automake manual leans towards avoiding the use of AM_MAINTAINER_MODE.
http://www.gnu.org/software/automake/manual/html_node/maintainer_002dmode.html
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
This is useful when the default repeat mode, which is 'repeat'
produces artifacts at the borders of the copied image.
Choose the 'pad' mode to make use of the color of the destination
image.
In my usage case the destination is the framebuffer, which is
solid filled with a background color. Scaling with 'pad' mode
would then just do the right thing and also produces nice
borders on the output.
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Reviewed-by: Inki Dae <inki.dae@samsung.com>
Tested-by: Joonyoung Shim <jy0922.shim@samsung.com>
This is a combination of g2d_copy_with_scale and g2d_scale.
It is a pretty common operation to scale one buffer and then
blend it on top of another, so provide a direct way to that
operation.
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Reviewed-by: Inki Dae <inki.dae@samsung.com>
Tested-by: Joonyoung Shim <jy0922.shim@samsung.com>
This makes it easier to spot memory corruptions which don't become
visible when using a plain buffer filled with a solid color (so
corruptions that are just a permutation of the bytes in the buffer).
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Reviewed-by: Inki Dae <inki.dae@samsung.com>
Tested-by: Joonyoung Shim <jy0922.shim@samsung.com>
With commit d7c0a08bc57(modetest: Allocate dumb buffers with the correct
bpp) we moved away from the libkms dependency. As such we are safe with
including the Makefile/subdir, even as we opt out of building the
library.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Needed on Solaris for the definitions of major() & minor() used in
drmGetNodeTypeFromFd() and makedev() used in drmOpenMinor()
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Only add wrapped bo's and bo's that have been exported through flink or dma-buf.
This avoids a lock in the common case, and decreases traversal needed for importing
a dma-buf or flink.
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@ubuntu.com>
Tested-By: Emil Velikov <emil.l.velikov@gmail.com>
While I've closed off most races in a previous patch, a small race still existed
where importing then unreffing cound cause an invalid bo. Add a test for this case.
Racing sequence fixed:
- thread 1 releases bo, refcount drops to zero, blocks on acquiring nvdev->lock.
- thread 2 increases refcount to 1.
- thread 2 decreases refcount to zero, blocks on acquiring nvdev->lock.
At this point the 2 threads will clean up the same bo.
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@ubuntu.com>
Reviewed-By: Emil Velikov <emil.l.velikov@gmail.com>
The fimg2d header was defining TRUE and FALSE, but actually
these defines are just used once. Remove them, since they
don't make the code better readable/understandable.
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Currently getchar() is used to pause execution after each test.
The user isn't informed if one is supposed to do anything for
the tests to continue, so print a simple message to make this
more clear.
v3: Compactify printf calls as pointed out by
Emil Velikov <emil.l.velikov@gmail.com>.
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Tested-by: Joonyoung Shim <jy0922.shim@samsung.com>
This also avoids the floating point conversion steps and just
uses pure integer arithmetic.
Since the G2D hardware scaling approach is a bit unintuitive,
document it in the function as well.
v2: Explicitly mention the normalization constant.
v3: Use common commenting style as pointed out by
Emil Velikov <emil.l.velikov@gmail.com>.
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Tested-by: Joonyoung Shim <jy0922.shim@samsung.com>
Currently most places assume reliable primary(master) <> render node
mapping. Although this may work in some cases, it is not correct.
Add a couple of helpers that hide the details and provide the name of
the master or render device name, given an fd. The latter may belong to
either the master, control or render node device.
v2:
- Rename Device and Primary to Master (aka the /dev/dri/cardX device).
- Check for the file via readdir_r() rather than stat().
- Wrap the check into a single function.
- Return NULL for non-linux platforms.
v3:
- Don't segfault if name is NULL.
- Update function names, as suggested by Frank Binns.
v4:
- Update commit message to reflect the function name changes.
Cc: Frank Binns <frank.binns@imgtec.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: David Herrmann <dh.herrmann@googlemail.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Currently the static archive libdrmtest.la links against libdrm.la.
Only to have both added to the executables' LDADD. Simplify things a
bit, by doing the link in the final module (the executables/tests in
this case).
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
None of the subdirs require the modules built in the local makefile, so
moving them at the top makes things a hell lot easier to read.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Mention that freedreno and intel are both auto-detected (depending on
the presence of atomics), plus host_cpu "defined".
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
They have used them since day one, but the check was never there.
v2:
- Update the freedreno message (API is no longer experimental).
- Move the freedreno host_cpu detection next to the intel one.
Cc: Rob Clark <robdclark@gmail.com>
Cc: Thierry Reding <treding@nvidia.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
This reverts commit 080b4929b7.
Chris noticed that "negative values wait forever" is indeed intended
behaviour and the issue is just that we didn't have a testcase (fixed
now) and that a regression slipped through (fixed and on track for all
stable kernels).
So lets undo the documentation change for consistency, since working
around kernel regressions isn't good. Practical impact is nil anyway.
v2: Add a note to docs that some kernels have been broken.
v3: Remove the random garbage included by accident.
Cc: Kristian Høgsberg <krh@bitplanet.net>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Kristian Høgsberg <krh@bitplanet.net>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
The kernel doesn't actually wait indefinately when passed a negative,
timeout, it returns immediately. Document this and suggest using INT64_MAX
for indefinite waits.
Signed-off-by: Kristian Høgsberg <krh@bitplanet.net>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
... minus test/ttmtest. The latter is not really hooked up with the
actual build.
This will give us 66 warnings on a distribution build of which
- 12 -Wunused-variable
- 11 -Wunused-function
- 19 -Wmissing-prototypes
and a few -Wswitch-enum, -Wtype-limits etc.
Adding the CFLAGS gives some exposure to these so that we can fix them.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Jan Vesely <jan.vesely@rutgers.edu>
As one adds WARN_CFLAGS to the build the compiler throws a couple of
lovely error messages. Add the relevant includes to fix them.
error: implicit declaration of function ‘time’
error: implicit declaration of function ‘getopt’
Cc: Inki Dae <inki.dae@samsung.com>
Cc: Kyungmin Park <kyungmin.park@samsung.com>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Jan Vesely <jan.vesely@rutgers.edu>
To make sure that the release/distribution tarball is not broken for all
the targets. Currently the experimental APIs are disabled by default
amongst others.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Jan Vesely <jan.vesely@rutgers.edu>
Since imx-drm has graduated from staging it seems to be a good idea to
recognize it by default in the libdrm tests.
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
With 64bit bionic mmap now handles 64bit offset, thus we no longer
need the __mmap2 trick.
Fix from Chih-Wei Huang, over at the google forums.
Cc: Chih-Wei Huang <cwhuang@android-x86.org>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
v2: call drmOpenOnceWithType in drmOpenOnce, and drop unused param
for drmOpenOnceWithType
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Jammy Zhou <Jammy.Zhou@amd.com>
v2: Add drmGetMinorBase, and call drmOpenWithType in drmOpen
v3: Pass 'type' to drmOpenByBusid and drmOpenDevice in drmOpenByName
v4: Renumber node type definitions, and return -1 for unsupported type
Signed-off-by: Jammy Zhou <Jammy.Zhou@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com> (v3)
Reviewed-by: Frank Binns <frank.binns@imgtec.com>
Since the atomic_set() macro references _ATOMIC_TYPE, undefining it
broke that macro, resulting in build errors of:
"intel_bufmgr_gem.c", line 775: undefined symbol: _ATOMIC_TYPE
Now that we're leaving the macro in the namespace, renamed it to start
with LIBDRM_ instead of _ to avoid polluting others namespace.
Regression introduced in 87fdd32c87 Add NetBSD atomic ops support.
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-By: Thomas Klausner <wiz@NetBSD.org>
We really have to do this to avoid surprises when extending the ABI
later on. Especially when growing the structures.
A bit overkill to update all the old legacy ioctl wrappers, but can't
hurt really either.
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Well just core drm. All the other callers in there that still use
direct calls to ioctl have some custom retry logic already, so should
be good already.
All the other offenders (tests, freedreno/kgsl, ...) don't really
matter (e.g. kgsl is the blob library and so not a drm thing) or are
again special exceptions with their own retry loops.
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
We really have to do this to avoid surprises when extending the ABI
later on. Especially when growing the structures.
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
We really have to do this to avoid surprises when extending the ABI
later on. Especially when growing the structures.
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
v2: Remove unrelated change in main()
This is more consistent with the rest, and avoids potential undefined
behavior (signed overflow) ind drmRandom()
Signed-off-by: Jan Vesely <jan.vesely@rutgers.edu>
Reviewed-by: Ian Romanick <idr@freedesktop.org>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
The ioctl DRM_EXYNOS_GEM_MAP_OFFSET and DRM_EXYNOS_GEM_MMAP are removed from
the linux kernel. This patch modifies libdrm and libkms to use drm generic
ioctls instead of the removed ioctls.
v2: The original patch was erroneous. In case the MODE_MAP_DUMB ioctl failed
it would return the retvalue as a void-pointer. Users of libdrm would then
happily use that ptr, eventually leading to a segfault. Change this to
return NULL in that case and also restore the previous behaviour of logging
to stderr.
The other error was that 'bo->vaddr' was never filled with the mapped
buffer address. Hence exynos_bo_map still returned NULL even if the
buffer mapping succeeded.
Signed-off-by: Hyungwon Hwang <human.hwang@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Add a new function, drmOpenRender, that can be used to open render nodes. This
can be used in the same way that drmOpenControl is used to open control nodes.
Signed-off-by: Frank Binns <frank.binns@imgtec.com>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Now that there are render nodes it doesn't seem appropriate for the type of
the card nodes to be DRM_NODE_RENDER. For this reason, rename this type to
DRM_NODE_PRIMARY as this name better represents the purpose of these nodes.
Signed-off-by: Frank Binns <frank.binns@imgtec.com>
Reviewed-by: Rob Clark <robdclark@gmail.com>
After much searching and empricial testing, and reading of
things I've no justifcation for this fix, other than it really
appears this is what the hw is doing or close enough.
It makes sense that each entry in the FMASK corresponds to
an entry in the CMASKm and the CMASK is organised into 128x128
blocks, but I can't find anything in any of the docs/info from AMD.
But I've spent a lot of time on this, and this seems to be the
simplest fix, in that we don't over allocate things too much,
once this fix in place we can nuke the extra multiplier in mesa.
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Now that enabling freedreno doesn't automatically enable kgsl, lets
enable it by default. The drm/msm driver has been upstream for a
while now, and it's ABI is locked. So I don't think we need to keep
calling it experimental.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
libdrm_freedreno currently supports two backends, 'msm' for the upstream
drm/msm driver, and 'kgsl' which supports (to some extent), the android/
downstream kgsl driver plus a sort of drm shim nonsense to get flink
names.
However, kgsl support is strictly on a best-effort basis. Different
android devices with different versions of kgsl may have different
abi's. And the existing kgsl interface (at least the parts of it that
we use) is completely broken for 64bit. Lets disable it by default lest
anyone actually try to use it.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
On recent emulator GTT entry setup for aub dump needs mem type as
GTT_ENTRY instead of NONLOCAL. NONLOCAL would write data in main
memory space which is wrong with new memory layout. GTT_ENTRY write
would setup GTT memory pool and other required internal buffers. With
this I can run aub dump on latest release without crash.
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
The modetest application uses libkms to allocate dumb buffers, leading
to overallocation due to the hardcoded 32 bpp value. This can even cause
failures in drivers when the resulting pitch is too large for the
hardware to handle and gets rejected by the driver when creating the
frame buffer.
Fix this by computing the required bpp value and allocating dumb buffers
directly without going through libkms.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Update drm_mode.h defines from kernel upstream for connector and
encoder types to expose DSI and other newly defined types.
Signed-off-by: Adam Cheney <acheney@nvidia.com>
Signed-off-by: Rob Clark <robclark@freedesktop.org>
These two functions are simple wrappers around the corresponding IOCTLs
and operate on drm_tegra_bo objects.
Signed-off-by: Thierry Reding <treding@nvidia.com>
These two functions are simple wrappers around the corresponding IOCTLs
and operate on drm_tegra_bo objects.
Signed-off-by: Thierry Reding <treding@nvidia.com>
The DRM_TEGRA_GEM_SET_FLAGS IOCTL can be used to set the flags of a
buffer object after it has been allocated or imported. Flags associated
with a buffer object can be queried using the DRM_TEGRA_GEM_GET_FLAGS
IOCTL.
Signed-off-by: Thierry Reding <treding@nvidia.com>
Currently the tiling parameters of buffer objects can only be set at
allocation time, and only a single tiled mode is supported. This new
DRM_TEGRA_GEM_SET_TILING IOCTL allows more modes to be set and also
allows the tiling mode to be changed after the allocation. This will
enable the Tegra DRM driver to import buffers from a GPU and directly
scan them out by configuring the display controller appropriately.
To complement this, the DRM_TEGRA_GEM_GET_TILING IOCTL can query the
current tiling mode of a buffer object. This is necessary when importing
buffers via handle (as is done in Mesa for example) so that userspace
can determine the proper parameters for the 2D or 3D engines.
Signed-off-by: Thierry Reding <treding@nvidia.com>
This test opens a device, dumps the version information and checks that
a Tegra DRM context can be opened on it.
Signed-off-by: Thierry Reding <treding@nvidia.com>
Add the libdrm_tegra helper library to encapsulate Tegra-specific
interfaces to the DRM.
Furthermore, Tegra is added to the list of supported chips in the
modetest and vbltest programs.
Signed-off-by: Thierry Reding <thierry.reding@avionic-design.de>
Signed-off-by: Erik Faye-Lund <kusmabite@gmail.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
This fixes an issue when trying to use -v and -C together. When trying
to read the page flip event, we are interrupted by the SIGALRM that
comes in, and so we think we timed out when we simply got EINTR. While
we could just loop checking for EINTR, SIGALRM is just bad idea to
begin with, so just rewrite it to use a thread.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
We also need to check handle_table in the _from_handle() path and the
_from_dmabuf() (which goes through _from_handle()) to avoid duplicate
imports.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
slice_pt is tileb[0] / tile_split, which isn't directly related to the
pitch.
This caused pitch_bytes to be too large in some cases.
[0] Tile size in bytes
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
4c2766b (drm_mmap/drm_unmap) brought this error for every .c file that
was not #including config.h:
In file included from private.h:4:0,
from abi16.c:29:
../libdrm.h: In function 'drm_munmap':
../libdrm.h:81:4: error: size of unnamed array is negative
Signed-off-by: Rob Clark <robdclark@gmail.com>
Unlike Linux, Android (bionic) has separate functions - mmap & mmap64.
Add a local wrapper (inspired by gallium) that will be used throughout
libdrm to combat this bionic feature.
v2:
- Drop the undefined unlikely macro.
- Use macro for munmap under Android.
- Wrap long lines.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Jakob Bornecrantz <jakob@vmware.com> (v1)
Autotools is already smart enough to pick the *.pc.in files but it
needs some help with the Android.mk ones.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Jakob Bornecrantz <jakob@vmware.com>
Just have the check once, and let new drivers opt-in if they want
to use them. Move the block further up the script, to tie nicely
with the actual usage of $HW defines.
Move the final $HW users to be alongside their brethren.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Jakob Bornecrantz <jakob@vmware.com>
Not too long ago the in-kernel drm public headers were moved. Since then
we could no longer fetch/update the ones in libdrm using the command.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Jakob Bornecrantz <jakob@vmware.com>
Rather than having two extra makefiles in order to ship ~10 headers
just fold its 5 lines of code into the top one makefile.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Jakob Bornecrantz <jakob@vmware.com>
When handling the error on GEM_CLOSE, we weren't freeing the allocated
page. Plug that.
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
The changes make sure that members of the bufmgr_gem and bo_gem
name lists are sychronized between threads
when using the create from prime and create from name methods.
Signed-off-by: Rafal Sapala <rafal.a.sapala@intel.com>
Testcase: igt/drm_import_export
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Allow userptr objects to be created and used via libdrm_intel.
At the moment tiling and mapping to GTT aperture is not supported
due hardware limitations across different generations and uncertainty
about its usefulness.
v2: Improved error handling in feature detection per review comments.
v3: Rebase on top of the drm_public addition, minor whitespace addition.
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com> (v3)
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> (v1,v2)
When using Mesa and LibVA in the same process, one would like to be
able bind buffers from the output of the decoder to a GL texture
through an EGLImage.
LibVA can reuse buffers allocated by Gbm through a file descriptor. It
will then wrap it into a drm_intel_bo with
drm_intel_bo_gem_create_from_prime().
The problem at the moment is that both library get a different
drm_intel_bufmgr object when they call drm_intel_bufmgr_gem_init()
even though they're using the same drm file descriptor. As a result,
instead of manipulating the same buffer object for a given file
descriptor, they get 2 different drm_intel_bo objects and 2 different
refcounts, leading one of the library to get errors from the kernel on
invalid BO when one of the 2 library is done with a shared buffer.
This patch modifies drm_intel_bufmgr_gem_init() so, given a file
descriptor, it will look for an already existing drm_intel_bufmgr
using the same file descriptor and return that object.
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
The file contains rules that are executed on incremental builds. This
way one can avoid doing a full clean and ensure that the new object
(library) is correctly build. Let's also cleanup the headers.
Inspired by the work of Chih-Wei Huang, from the Android-x86 project.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Contains the following patches squashed in:
commit 99247a5bd724ddcf0f06a5518baad207c53f1e2b
Author: Haitao Huang <haitao.huang@intel.com>
Date: Fri, 27 Apr 2012 13:20:53 -0500
Android.mk: use LOCAL_COPY_HEADERS to export headers.
Export necessary header files used by other components for
Android, such as libva intel-driver, gralloc, hwcomposer, etc.
Change-Id: I2feabf6941379ef4d756e942f30eba059de641f1
Signed-off-by: Haitao Huang <haitao.huang@intel.com>
[chad: Fixed inconsistent indentation.]
Signed-off-by: Chad Versace <chad.versace@linux.intel.com>
commit 7d0b528cb69995d7ea4e29b2daa1e3b28a362f42
Author: Emil Velikov <emil.l.velikov@gmail.com>
Date: Sun, 27 Jul 2014 18:22:41 +0100
android: reuse headers lists, separate libdrm from intel headers
Rather than having a duplicate copy of the headers list(s),
reuse the existing one(s). Distinguish that the intel headers
should be copied when libdrm_intel is used.
v2 Rename the headers variable(s) to *_H_FILES.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
commit 361de3ba4cadd5357596d1537bb3f216d281532b
Author: Piotr Luc <piotr.luc@intel.com>
Date: Fri, 14 Jun 2013 13:00:39 +0200
Export include dir from libdrm
BZ: 116218
Google introduced new method of specifying include path(s)
between modules. This allows a module to include header from a
library without directly specifyining by includer the path where
headers are located.
The method requires from library that holds headers to export
include path(s) in LOCAL_EXPORT_C_INCLUDE_DIRS variable.
These exported include path(s) are automatically added to
include path(s) of modules that have name of the library in the
LOCAL_SHARED_LIBRARIES or LOCAL_STATIC_LIBRARIES list.
This change sets LOCAL_EXPORT_C_INCLUDE_DIRS to folders that
contain headers file that used by other modules in order to
export these paths.
Change-Id: Id1ac885b31ef2efe194e0289fbcaecd9eb533df0
Signed-off-by: Piotr Luc <piotr.luc@intel.com>
Reviewed-on: http://android.intel.com:8080/113562
Reviewed-by: cactus <cactus@intel.com>
Reviewed-by: Luc, Piotr <Piotr.Luc@intel.com>
Reviewed-by: Purushothaman, Vijay A <vijay.a.purushothaman@intel.com>
Reviewed-by: Stimson, Dale B <dale.b.stimson@intel.com>
Tested-by: Stimson, Dale B <dale.b.stimson@intel.com>
Reviewed-by: buildbot <buildbot@intel.com>
Tested-by: buildbot <buildbot@intel.com>
commit 2bf22fcbd4cbb9e7c7764d5eff0bb4e75ab1a005
Author: Emil Velikov <emil.l.velikov@gmail.com>
Date: 27 Jul 2014 18:27:21 +0100
android: Separate libdrm and intel LOCAL_EXPORT_C_INCLUDE_DIRS
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Contains the following patches squashed in:
commit f340a8b9f2b84d5762553bef046914e0bde20795
Author: Chad Versace <chad.versace@linux.intel.com>
Date: Wed, 21 Dec 2011 11:43:57 -0800
libdrm,intel: Add Android makefiles (v2)
This enables libdrm.so and libdrm_intel.so to build on Android
IceCreamSandwich.
v2: Link libdrm_intel to libpciaccess.
Change-Id: Ie5ed4bc0e6b4f9f819e3ec44488e385c35e97128
Signed-off-by: Chad Versace <chad.versace@linux.intel.com>
commit 8fb3f42389dea34218ed1fe59550ec2abb4d6953
Author: Andrew Boie <andrew.p.boie@intel.com>
Date: Wed, 26 Sep 2012 13:32:05 -0700
libdrm, libdrm_intel: Skip driver name checks
These libraries have 'optional' tags, which means they won't get
built unless something else depends on them or they are added to
PRODUCT_PACKAGES. There's no need for additional filtering.
Change-Id: I5d90969f38671f8144c0dc27d47144b3f09a15ce
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>
Will be used to consolidate the required sources lists as well as the
install-able headers. This is turn will help us to avoid the
duplication with the upcoming Android build support.
v2: Rename the headers variable to *_H_FILES.
v3: Rebase on top of symbol visibility patches.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
configure.ac has AC_SYS_LARGEFILE which provides the define and/or
approapriate magic when required.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
For the sake of simplicity, KMS support can always be considered
present on DragonFly.
If some particular version doesn't support KMS yet, appropriate
checks are already done in Dports's x11-drivers/ Makefiles and
KMS-enabled driver packages don't get built.
Signed-off-by: François Tigeot <ftigeot@wolfpond.org>
Signed-off-by: Maarten Lankhorst <dev@mblankhorst.nl>
This hides all the abi16_* functions and the nouveau_debug variable,
they should have been private to begin with.
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Checks whether or not the compiler supports the -fvisibility option. If
so it sets the VISIBILITY_CFLAGS variable which can be added to the per
directory AM_CFLAGS where appropriate.
By default all symbols will be hidden via the VISIBILITY_CFLAGS. The
drm_public macro can be used to mark symbols that should be exported.
Reviewed-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Use "drm.h" instead of "drm/drm.h" as used in the other header files.
Fixes xserver-xorg-video-qxl build with KMS support on Debian, where this
file is installed in /usr/include/libdrm.
Fixes Debian bug #746807
Reported-by: Bastian Blank <waldi@debian.org>
Signed-off-by: Andreas Boll <andreas.boll.dev@gmail.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
When division of source and destination width yields the
scaling factor for the x-coordinate, then it should be
source/destination _height_ for y.
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
The hardware accepts scaling factors formatted in a
fixed-point format. The current macro casts to integer
first, then multiplies by the fp conversion factor.
This does not make any sense. In particular, truly
'fractional' inputs, like 1.5, won't work that way.
Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
drmOpenByName() is a static function that is only called by drmOpen().
drmOpen() already checks drmAvailable(), so the check in
drmOpenByName() is redundant.
Signed-off-by: Daniel Kurtz <djkurtz@chromium.org>
Signed-off-by: Thierry Reding <treding@nvidia.com>
The documentation says fd holds the fd from prime import/export.
However, it isn't actually used, nor is it necessary, so let's just remove
it.
Signed-off-by: Daniel Kurtz <djkurtz@chromium.org>
Acked-by: Inki Dae <inki.dae@samsung.com>
Reuse the common drmPrime() helper functions rather than reinventing them.
Signed-off-by: Daniel Kurtz <djkurtz@chromium.org>
Acked-by: Inki Dae <inki.dae@samsung.com>
The first parameter should be the drm fd, second param is the fb id.
Signed-off-by: Daniel Kurtz <djkurtz@chromium.org>
Acked-by: Inki Dae <inki.dae@samsung.com>
This looks like it was copied from kmstest, but isn't needed, and doesn't
actually work since exynos_fimg2d_test requires parameters.
Signed-off-by: Daniel Kurtz <djkurtz@chromium.org>
Acked-by: Inki Dae <inki.dae@samsung.com>
warning: assignment makes pointer from integer without a cast [enabled by default]
warning: initialization makes integer from pointer without a cast [enabled by default]
Signed-off-by: Daniel Kurtz <djkurtz@chromium.org>
Acked-by: Inki Dae <inki.dae@samsung.com>
This exynos test was added just before HAVE_INSTALL_TESTS, and so didn't
get this annotation.
Signed-off-by: Daniel Kurtz <djkurtz@chromium.org>
Acked-by: Inki Dae <inki.dae@samsung.com>
DRM_IOCTL_MODE_SETPLANE crtc_x, crtc_y are s32.
This is to allow a destination location that is partially off screen.
Make this more obvious to users of libdrm by using signed crtc_x/_y
parameters for drmModeSetPlane() as well.
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Daniel Kurtz <djkurtz@chromium.org>
drmFreeVersion() frees the memory allocated for the name, date and desc
fields in addition to that for the struct _drmVersion.
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Thierry Reding <treding@nvidia.com>
The debug message's format string doesn't contain any conversion
specifiers, therefore making the fd argument unused.
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Some of the format strings for debug messages use the wrong modifier to
print sizes.
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Thierry Reding <treding@nvidia.com>
These functions all take a format string and either a list of variable
arguments or a va_list. Use the new DRM_PRINTFLIKE macro to tell the
compiler about it so that the arguments can be checked against the
format string.
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Thierry Reding <treding@nvidia.com>
I cannot make nouveau_bo_wrap thread-safe (by design), but it seems to be used to convert
drm fb's to nouveau_bo's and to get a notify handle from fifo->notify in nv30_screen.c
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@ubuntu.com>
Don't even bother trying DRM_KGSL_GEM_CREATE_FD. It hasn't worked since
(afaict) 2.6.35 kernels. And in some cases seems to cause some
problems. Instead just allocate a minimum size dummy object (just for
purposes of having a handle) and then mmap the framebuffer as user-mem
(which is deprecated, but seems to still work.. and as far as I can tell
is the best option for now).
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Need to update timestamp on all ring's associated with a submit (ie.
both the binning pass and main ring). Also, make sure nr_reloc's
in particular gets cleared if the rb is reset.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
The existing 'offset' field is unfortunately typed as 'unsigned long',
which is unfortunately only 4 bytes with a 32-bit userspace.
Traditionally, the hardware has only supported 32-bit virtual addresses,
so even though the kernel uses a __u64, the value would always fit.
However, Broadwell supports 48-bit addressing. So with a 64-bit kernel,
the card virtual address may be too large to fit in the 'offset' field.
Ideally, we would change the type of 'offset' to be a uint64_t---but
this would break the libdrm ABI. Instead, we create a new 'offset64'
field to hold the full 64-bit value from the kernel, and store the
32-bit truncation in the existing 'offset' field, for compatibility.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
I've seen a number of apps spending unreasonable amounts of time in
drm_intel_bo_busy during the buffer mapping process.
We can't track idleness in general, in the case of buffers shared
across processes. But this should significantly reduce our overhead
for checking for busy on things like VBOs.
Improves (unoptimized) glamor x11perf -f8text by 0.243334% +/-
0.161498% (n=1549), which has formerly been spending about .5% of its
time hitting the kernel for drm_intel_gem_bo_busy().
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Fixes valgrind complaints in the modesetting driver. I tried to
follow each ioctl's pattern for whether it was initializing just the
in values, or both in and out values.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
the drmServerInfo member, debug_print, takes a printf format string
and varargs list. Tell the compiler about it.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
There seem to be some cases (I've noticed this switching resolution in
some games, for example) where the fd can get closed() before the device
and all it's bo's are destroyed. Which, if the drm device is opened
again and bo's are allocated with the same handles, results that when
the first pipe_screen/pipe_context is destroyed causes the first dev to
close handles for bo's allocated by the second device.
The easy solution to that is to add a mode where the fd_device creates
it's own private fd (a dup()).
Signed-off-by: Rob Clark <robclark@freedesktop.org>
When mode is selected we only give the name of the mode as parameter.
But sometime, two different modes have the same name but not
the same vrefresh frequency.
This patch give the possibility to select a mode by its name
and optionally by its refresh frequency.
Signed-off-by: Vincent Abriou <vincent.abriou@st.com>
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Rob Clark <robclark@freedesktop.org>
In this patch, to support exynos for KMS, Exynos KMS driver is newly added.
Also, Exynos is added to the list of kmstest supported modules.
Signed-off-by: Hyungwon Hwang <human.hwang@samsung.com>
Signed-off-by: Rob Clark <robclark@freedesktop.org>
The previous code would just use the potentially unallocated variable,
which is probably okay most of the time, but not very nice to the user
of the library.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Allow IB to different ringbuffer in addition to just different part of
same ringbuffer. In particular, we need to add bo's to the parent (ie.
one passed to flush) bo table, since the bo table applies to all the
cmd buffers in submit ioctl.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Some users report hitting a divide by 0 with the tile split in
certain apps. Tile_split shouldn't ever be 0 unless the surface
structure was not properly initialized. I think there may be some
cases where mesa uses an improperly initialized surface struct,
but I haven't had time to track it down.
Bug:
https://bugs.freedesktop.org/show_bug.cgi?id=72425
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Workloads which create many transient buffers cause significant CPU
overhead in buffer allocation, zeroing, cache maint, and mmap setup.
By caching and re-using existing buffers, the CPU overhead drops
significantly. See:
http://bloggingthemonkey.blogspot.com/2013/09/freedreno-update-moar-fps.html
A simple time based policy is used for purging the cache. Once the
kernel supports it, we could use madvise style API to handle memory
pressure scenarios a bit better.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
- *.log/*.trs are generated by make check
- TAGS are generated by make tags
- build-aux, config.h.in~ by autoconf
- *.sw? are temporary files create by vim
- name_from_fd wasn't ignored yet for some reason
v2: Ignore *~ as well (Daniel Vetter)
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
If the application sends us a file descriptor pointing at a prime
buffer that we've already got, we have to re-use the same bo_gem
structure or chaos will result.
Track the set of all known prime objects and look to see if the kernel
has returned one of those for a new file descriptor.
Also checks for prime buffers in the flink case.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Bug fixes and simplification by Marek.
We have to use the tile index of 0 for non-MSAA depth-stencil after all.
Signed-off-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-and-Tested-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
I would have just used the drmIoctl interface directly in Mesa, but the
ioctl needs some data from the drm_intel_context that is not exposed
outside libdrm.
This ioctl is in the drm-intel-next tree as b635991.
v2: Update based on Mika's kernel work.
v3: Fix compile failures from last-minute typos. Sigh.
v4: Import the actual changes from the kernel i915_drm.h. Only comments
on some fields of drm_i915_reset_stats differed. There are still some
deltas between the kernel i915_drm.h and the one in libdrm, but those
can be resolved in other patches.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org> [v3]
Reviewed-by: Damien Lespiau <damien.lespiau@intel.com>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
This reverts commit 6335e1d28c.
No taxation without representation, in other words no userspace without kernel
stuff being in a stable location, either drm-next but I'll accept drm-intel-next
for intel specific stuff.
I would have just used the drmIoctl interface directly in Mesa, but the
ioctl needs some data from the drm_intel_context that is not exposed
outside libdrm.
v2: Update based on Mika's kernel work.
v3: Fix compile failures from last-minute typos. Sigh.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Cc: Mika Kuoppala <mika.kuoppala@intel.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
The command now takes a 48bits address and is thus 1 dword longer.
v2 (Ben): commit message: s/byte/dword (Eric)
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Since our aub file dumping's GTT handling is totally fake, we always put
everything in the low 4GB anyway and shouldn't ever need to set
AddressHigh to anything other than 0.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
[ben: slight commit message change]
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
This exposes the kernel API for performing asynchronous flips
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
The various create and open functions set the buffer size, but
drm_intel_bo_gem_create_from_prime() is an exception. In the 3.12 kernel
we can now use lseek on the prime fd to determine the size of the bo.
Use that and override the userprovided size. If the kernel doesn't
support this, we get an error and fall back to the user provided size.
Signed-off-by: Kristian Høgsberg <krh@bitplanet.net>
This fixes VM protection faults.
I have a new piglit test which can iterate over all possible widths, heights,
and depths (including NPOT) and tests mipmapping with various texture targets.
After this is committed, I'll make a new release of libdrm and bump
the libdrm version requirement in Mesa.
Documentation states that AC_*_IFELSE has to use AC_LANG_SOURCE or
friends in order to generate the source code to compile.
AC_LINK_IFELSE already handles this, thus convert AC_COMPILE_IFELSE
to silence the final autoconf warnings.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Signed-off-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Daniel Kurtz <djkurtz@chromium.org>
The following minor changes were needed to these headers:
* Convert // comments to /* */
* No , after final member of enum
With these changes, these header files can be included by a program that
is built with gcc options:
-std=c89 -Werror -pedantic
Signed-off-by: Daniel Kurtz <djkurtz@chromium.org>
Signed-off-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Eric Anholt <eric@anholt.net>
This is helpful for differentiating between multiple devices that use
the same module.
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Aligns the IB to 8 DWs. The aligns the IB to the
CP fetch size. r6xx also require at least 4 DW
alignment to avoid a hw bug.
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Handles automake complains about lack of forward-compatibility, due to the
source files in the kgsl and msm backends/folders.
Cc: Rob Clark <robclark@freedesktop.org>
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
The compiler is unaware of that we have at least one crts/connector/plane
thus it complains that some of our variables will be used uninitialised.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Prodives memset() and strlen(), used in tests/setversion
tests/getversion respectively.
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Split out common code and backend. Current backend is for 'kgsl'
android driver, but a new backend will provide support for the
upstream msm drm/kms driver.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Multiple the image height by 1.5 for NV12/NV21 and by 2 for NV16/NV61 to
make room for the chroma plane.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
The line stride passed to the function is expressed in bytes, there's no
need to multiply it by 2.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
When building the pipeline, instead of using only the encoders attached
to a connector, take all possible encoders into account to locate a
CRTC.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
The -s argument can now take a list of connectors. Configure all of them
in cloned mode using a single CRTC.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
This prepares the code for handling multiple connectors in a single
pipeline in a cloned configuration.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
There's not reason to require setting a mode to test planes. Split the
two operations.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Planes are associated with CRTCs, not connectors. Don't try to be too
clever, use the CRTC ID in the -P option. This prepares for splitting
CRTC and planes setup.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
This prepares the code for the split in separate functions of CRTC and
planes setup.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
This prepares the code for the split in separate functions of CRTC and
planes setup.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Instead of passing the device fd and resources as global variables group
them in a device structure and pass it explictly to all functions that
need it.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
As modetest automatically selects an unused plan, providing the plane ID
allows modifying plane properties for the selected planes.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Extend the -P option to allow specifying the plane x and y offsets. The
position is optional, if not specified the plane will be positioned at
the center of the screen as before.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
The -w parameter can be used to set a property value from the command
line, using the target object ID and the property name.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Configuring mode on more than two connectors or two planes is perfectly
valid. Support it.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Instead of retrieving resources as they are needed, retrieve them all
(except property blobs) in one go at startup.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
If the -d parameter is specified, modetest will drop master permissions
after setting the mode.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
If the -M parameter is specified, modetest will use the requested device
name instead of trying its builtin list of device names.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
The current mostly random sort order hinders code readability. Sort the
options alphabetically in the code, and by group in the help message.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Enable all standard automake warnings except for -Wpointer-arith (as the
test pattern generation code uses void pointer arithmetics) and fix
them.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Mark the address ranges as accessible with VALGRIND_MAKE_MEM_DEFINED.
Signed-off-by: Chia-I Wu <olvaffe@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
No need to prepare the .aub header and dump in that case, it'll be
done with the next call with true.
Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
RADEON_GEM_WAIT_IDLE is declared DRM_IOW but libdrm
uses it with drmCommandWriteRead instead of drmCommandWrite
which leads to the ioctl being unmatched and returning an
error on at least OpenBSD.
Signed-off-by: Mark Kettenis <kettenis@openbsd.org>
Signed-off-by: Jonathan Gray <jsg@jsg.id.au>
When publishing first HSW ids we weren't allowed to use "GT3" codname.
But this is the correct codname and Mesa is using it already.
So to avoid people getting confused why in Mesa it is called GT3 and here
it is called GT2_PLUS let's fix this name in a standard and correct way.
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
When adding the name tracking, I missed removing from the name table
when the bo was deleted, leaving a dangling pointer.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Things are worse if we issueibcmds with bogus gpu ptrs, so it is better
to just make userspace crash when things go pear shaped.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Due to the evil userspace buffer tracking we have to do, and hacks for
creating GEM buffer from fbdev/scanout, "evil-twin" fd_bo objects are
problematic. So introduce hashtable tracking of bo's and dev's, to
avoid getting duplicate fd_bo ptrs for the same underlying gem object,
in particular when importing via flink name.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Super-cheezy way to synchronization between mesa and ddx.. the
SET_ACTIVE ioctl gives us a way to stash a 32b # w/ a GEM bo, and
GET_BUFINFO gives us a way to retrieve it. We use this to stash
the timestamp of the last ISSUEIBCMDS on the buffer.
To avoid an obscene amount of syscalls, we:
1) Only set the timestamp for buffers w/ an flink name, ie.
only buffers shared across processes. This is enough to
catch the DRI2 buffers.
2) Only set the timestamp for buffers submitted to the 3d ring
and only check the timestamps on buffers submitted to the
2d ring. This should be enough to handle synchronizing of
presentation blit. We could do synchronization in the other
direction too, but that would be problematic if we are using
the 3d ring from DDX, since client side wouldn't know this.
The waiting on timestamp happens before flush, and setting of
timestamp happens after flush. It is transparent to the user
of libdrm_freedreno as all the tracking of buffers happens via
_emit_reloc()..
Signed-off-by: Rob Clark <robclark@freedesktop.org>
YUV420 support is trivial to add since the code already supports
YVU420.
But it looks like the YVU420 support is a bit broken. The chroma
planes are passed in the wrong order to the fill functions, so
fix that while were at it.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
The fourcc is inside the format_info structure, so if we want to use
it inside the various fill_tiles functions, we need to pass down the
whole format_info, not just the rgb/yuv infos.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
libkms only has the xrgb8888 format, so we're overallocating the bo by
quite a lot in some cases. But we still need to get the pitch from the
libkms since it's the driver that decides how to align it.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
The function returns returns 0 on success or a negative value in case of an
error, except when given invalid attributes in which case it returns the
positive EINVAL value. Replace that with -EINVAL to allow the caller to detect
errors with a < 0 check.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Jakob Bornecrantz <wallbraker@gmail.com>
v2: Only writte tile index if flags for it is set
v3: Remove useless allow2d scanout flags
v4: Split radeon_drm.h update to its own patch
v5: update against lastest next tree for radeon
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
It accidentally used the cmd id for the gen7 command and had an
outdated lenght field. Spotted while trying to make sense of an ivb
error_state from mesa 7.11 ...
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The second digit was off by one, which meant we accidentally treated
GT(n) as GT(n-1). This also meant no support for GT1 at all.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
If clock_gettime did fail, it would return -1 and set errno.
What we really want to strerror() is the errno.
Signed-off-by: Daniel Kurtz <djkurtz@chromium.org>
Protect the macro argument evaluations with parens.
This is already touching most lines, so while at it, fix up all white
space to uniform style throughout the file.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Currently single pushbuffer can take up to 80% of VRAM and 80% of GART.
As this value seems to be arbitrary (and user may need to set it differently)
this patch adds support for 2 environment variables:
NOUVEAU_LIBDRM_VRAM_LIMIT_PERCENT (default 80)
NOUVEAU_LIBDRM_GART_LIMIT_PERCENT (default 80)
which will let users override pushbuffer VRAM/GART limits.
Signed-off-by: Marcin Slusarz <marcin.slusarz@gmail.com>
Install test programs is useful in cross compilation case. By default
the behavior is the same and test programs aren't installed in $bindir.
If --enable-install-test-programs is set then test programs are
installed in $bindir.
Signed-off-by: Benjamin Gaignard <benjamin.gaignard@linaro.org>
Signed-off-by: Rob Clark <robdclark@gmail.com>
This patch adds library and test application for g2d gpu(fimg2d).
The fimg2d hardware is a 2D graphics accelerator(G2D) that
supports Bit Block Transfer(BitBLT).
The library includes the following primitive drawing operations:
.solid fill - This operation fills the given buffer with
the given color data.
.copy - This operation copies contents in source buffer to
destination buffer.
.copy_with_scale - This operation copies contents in source buffer
to destination buffer scaling up or down properly.
.blend - This operation blends contents in source buffer with
the ones in destination buffer.
And the above operations uses gem handle or user space address
allocated by malloc() as source or destination buffer.
And the test application includes just simple primitive drawing
tests with the above library.
And the guide to test is as the following,
"#exynos_fimg2d_test -s connector_id@crtc_id:mode"
With this above simple command, four primitive drawing operations
would be called step by step and also rendered on the output device
to the given connector and crtc id.
Signed-off-by: Inki Dae <inki.dae@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
Signed-off-by: Rob Clark <robdclark@gmail.com>
The libdrm_freedreno helper layer for use by xf86-video-freedreno,
fdre (freedreno r/e library and tests for driving gpu), and eventual
gallium driver for the Adreno GPU. This uses the msm gpu driver
from QCOM's android kernel tree.
Note that current msm kernel driver is a bit strange. It provides a
DRM interface for GEM, which is basically sufficient to have DRI2
working. But it does not provide KMS. And interface to 2d and 3d
cores is via different other devices (/dev/kgsl-*). This is not
quite how I'd write a DRM driver, but at this stage it is useful for
xf86-video-freedreno and fdre (and eventual gallium driver) to be
able to work on existing kernel driver from QCOM, to allow to
capture cmdstream dumps from the binary blob drivers without having
to reboot. So libdrm_freedreno attempts to hide most of the crazy.
The intention is that when there is a proper kernel driver, it will
be mostly just changes in libdrm_freedreno to adapt the gallium
driver and xf86-video-freedreno (ignoring the fbdev->KMS changes).
So don't look at freedreno as an example of how to write a libdrm
module or a DRM driver.. it is just an attempt to paper over a non-
standard kernel driver architecture.
v1: original
v2: hold ref's to pending bo's (because qcom's kernel driver doesn't),
various bug fixes, add ringbuffer markers so we can emit IB's to
portion of ringbuffer (so that gallium driver can use a single
ringbuffer for both tile cmds and draw cmds.
Signed-off-by: Rob Clark <robclark@freedesktop.org>
Intel GPU Tools is newer and arguably better. This change doesn't
completely merge the files because it's a bit simpler if we move the
I9XX macro over to Intel GPU Tools, and don't move over a few macros
from IGT that libdrm doesn't care about.
It has been discussed, and would seem even easier if Intel GPU Tools
simply used the libdrm header files. Whether or not we move to that,
this should help that effort.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Acked-by: Jesse Barnes <jbarnes@virtuousgeek.org>
On Gen6, bit 15 is now `Depth Clear Value Valid`. This was being treated
as part of the length, and failing the rest of the batchbuffer decode.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Chris Forbes <chrisf@ijw.co.nz>
We didn't set the ring flag for BLT batches, so they got run on the
render ring. Shenanigans ensued, especially when we sent commands that
were only valid on the BLT ring.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
I'm fairly sure the extra flags it specifies no longer exist, so
there is no point in keeping it. It only adds a warning when
running make distcheck.
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Due to the typo, none of the .xml files would end up in the release
tarball and cause make distcheck as well as builds from the tarball to
fail.
Using $() isn't strictly necessary but other variables and expressions
use that variant already so it makes the usage consistent.
Reviewed-by: David Herrmann <dh.herrmann@googlemail.com>
Signed-off-by: Thierry Reding <thierry.reding@avionic-design.de>
Very similar to Evergreen, but slightly different rules for tile / slice
alignment. Fortunately, these map quite naturally onto the previous fixes for
linear aligned layout on SI.
2D tiling still needs more work here and possibly in the kernel.
Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
This fixes all the out-of-tree build-failures with manpages and uses a
.man_fixup file to avoid overriding man-pages on every build.
Manpages are only built if xsltproc is found and the stylesheets are
available locally. You can disable building manpages with
--disable-manpages so the quite expensive xsltproc procedure can be
skipped.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
v2: Take Maarten Lankhorst's suggestion of nesting the struct to prevent
sizeof() issues due to padding on older revisions.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
As we clear the relocs from the bo, we also need to clear the
contribution of the reloc_target_bo from the fence count. Otherwise they
are leaked and prevent any further relocations being added to the bo.
This adds an overview page that describes Dumb-Buffers, TTM and GEM. It
does not describe chipset-specific features. You should do that in the
driver-manpages.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
This is an overview page for KMS. It is again targeted at novice users
that need redirection to the correct function man-pages.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
The drm.xml file compiles to drm.7 and is meant as a global overview page
for libdrm. It is targeted to new users of libdrm and redirects to all
other main man-pages.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
If we want to use the manpages in external documentation other than normal
manpages, we should rather use XML. Furthermore, almost no-one knows troff
today, anyway, and XML allows others to easily add more pages without
having to learn troff.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
Under certain circumstances it's possible for libdrm to decide to move
a GART|VRAM pushbuf to be VRAM-only. This causes the kernel to reject
the command submission on GF8 and up, due to a stricter policy where
buffers are only allowed to move to memory types that were specified
at creation time.
The simplest fix for this is to force the creation-time memory type for
the lifetime of the push buffer.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
To avoid kernel rejecting cs if we return different global name
for same bo keep track of global name and always return the same.
Seems to fix issue with suspend/resume failing and repeatly printing
following message :
[drm:radeon_cs_ioctl] *ERROR* Failed to parse relocation -35!
There might still be way for a rogue program to trigger this issue.
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
If we have valid timings, we can at least set width/height to
*something*, which is I think at least less confusing than always
seeing width/height of zero. At least modeprint and modetest
seem to expect width/height to mean something.
Signed-off-by: Rob Clark <rob@ti.com>
We don't want to build libdrm tests with Cairo support under Poky, since
they're never used and also cause a build loop from libdrm -> cairo ->
mesa-dri -> libdrm.
To avoid variance in build results, introduce a --disable-cairo-tests
switch.
Signed-off-by: Daniel Stone <daniel@fooishbar.org>
Signed-off-by: Dave Airlie <airlied@redhat.com>
intel_bufmgr_gem.c: In function 'drm_intel_bo_gem_export_to_prime':
intel_bufmgr_gem.c:2477:6: warning: unused variable 'ret' [-Wunused-variable]
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
commit 92fd0ce4f6
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date: Fri Aug 31 11:16:53 2012 +0200
intel: properly test for HAS_LLC
missed slightly and in effect had no effect on the outcome of checking
whether the kernel/chipset supported LLC.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
This allows texturing with depth-stencil buffers directly without the copy
to CB. The separate miptree description for stencil is added, because
the stencil mipmap offsets are not really depth offsets/4 (at least
for the texture units).
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
It's the same situation as flink and we need take the same precautions.
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Kristian Høgsberg <krh@bitplanet.net>
If the kernel supports the test, we need to check the param.
Copy&pasta from the above checks that only look at the return value.
Interesting how much one can get such a simple interface wrong.
Issue created in
commit 151cdcfe68
Author: Eugeni Dodonov <eugeni.dodonov@intel.com>
Date: Tue Jan 17 15:20:19 2012 -0200
intel: query for LLC support
Patch even claims to have fixed this in v2, but is actually unchanged
from v1.
Reported-by: Xiang, Haihao <haihao.xiang@intel.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
omap_drm.h uses data type defined in stdint.h, but that header was
not included.
omap_drm.h includes drm.h as a local file when it is part of the
compiler c flags.
This two issues are fixed. New code can include omap_drm.h alone.
Signed-off-by: Víctor Manuel Jáquez Leal <vjaquez@igalia.com>
Signed-off-by: Rob Clark <rob@ti.com>
Otherwise pad appears uninitialized and valgrind grumbles.
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Eric Anholt <eric@anholt.net>
It warns about totally sensible things done in intel_decode.c. I've
never seen this warn do anything useful, and apparently I was the one
to introduce it when I added the giant pile of warning flags back in
2008.
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>
If we don't need stencil, don't allocate it.
If we need only stencil (like PIPE_FORMAT_S8_UINT), don't allocate depth.
v2: actually do it correctly
Reviewed-by: Christian König <christian.koenig@amd.com>
Otherwise we end up with X hitting a fail-loop as the embedded libGL
stacks asserts whilst initialising.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
If color format for CRTC layer is not specified on commandline, then
c->fourcc is unintialized resulting in addfb call failing.
Signed-off-by: Rob Clark <rob@ti.com>
As the modeset test application is often referred to as an example of
the KMS API usage, move test pattern generation and buffer allocation to
a separate file to keep it simple and clear.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
This will make it easier to add additional parameters to the connector
and plane arguments.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Implement tiles and SMPTE test pattern generation for the RGB565,
BGR888, RGB888, ARGB8888, BGRA8888 and BGRX8888 formats.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Implement tiles and SMPTE test pattern generation for the NV12, NV21,
NV16 and NV61 formats.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Implement tiles and SMPTE test pattern generation for the UYVY, VYUY and
YVYU formats.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Merge the create_test_buffer() and create_grey_buffer() functions into a
single buffer allocation function that takes the pixel format and fill
pattern as parameters.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
CCLD test_decode
./.libs/libdrm_intel.so: undefined reference to `drmPrimeHandleToFD'
./.libs/libdrm_intel.so: undefined reference to `drmPrimeFDToHandle'
collect2: ld returned 1 exit status
From Adam Jackson's explaination:
most distros have changed it so ld defaults to --no-copy-dt-needed-entries,
so if you use something from libdrm you can't just assume libdrm_intel
will bring it in for you, you have to be explicit
Signed-off-by: Rob Clark <rob@ti.com>
This adds interfaces for the X driver to use to create a
prime handle from a buffer, and create a bo from a handle.
v2: use Chris's suggested naming (well from at least for consistency)
v3: git commit --amend fail
v4: fix as per Chris's suggestions, group assignments, add get tiling
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Dave Airlie <airlied@redhat.com>
These are just basic ioctl wrappers around the prime ioctls,
along with the capability reporting.
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
There can be scenarios, especially when re-importing an existing buffer,
where you end up with multiple 'struct omap_bo's wrapping a single GEM
object handle. Which causes badness when the first of the evil-clones
is omap_bo_del()'d.
To do this, introduce reference counting and a hashtable to track the
handles per fd.
First, to avoid bo's slipping through the crack if multiple 'struct
omap_device's are created for one drm fd, a hashtable mapping drm
fd to omap_device, and the omap_device itself is reference counted.
Per omap_device, we keep a handle_table mapping GEM handle to omap_bo.
When buffers are imported from flink name or dmabuf fd, the handle
table is consulted, and if an omap_bo already exists, it's refcnt is
incremented and it is returned. For good measure, to avoid the
handle_table being deleted before the omap_bo is freed, the omap_bo
holds a reference to the omap_device.
TODO: check the overhead of the hashtable. If too much we could maybe
get away with only tracking exported and imported bo's in the table.
TODO: all the import/export flink/dmabuf operations are generic DRM
ioctls. Really all this functionality could be handled by a generic
drm_bo and drm_device "base class" that could be extended by omap,
exynos, etc. That would also allow more common userspace code by
avoiding artificial libdrm_omap dependencies.
Signed-off-by: Rob Clark <rob@ti.com>
Since there is no getparam for hardware context support, Mesa always
tries to obtain a context by calling drm_intel_gem_context_create and
NULL-checking the result. On an older kernel without context support,
this caused libdrm to print an unwanted message to stderr:
DRM_IOCTL_I915_GEM_CONTEXT_CREATE failed: Invalid argument
In fact, this caused every Piglit test to fail with a "warn" status due
to the unrecognized error message.
Change the message to use DBG() rather than fprintf(), so people can
still get the debug message, but it won't spam normally.
Reviewed-by: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Hi list
The recently released libdrm 2.4.37 does not compile the Intel part:
test_decode.c: In function 'compare_batch':
test_decode.c:107: error: implicit declaration of function 'open_memstream'
PS: Please CC me.
Signed-off-by: Lauri Kasanen <cand@gmx.com>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Add relevant code to set up minimal state and call the appropriate
kernel IOCTLs.
This was missed in the previous cherry-picking for 2.3.36.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
I mistakenly "fixed" a bad decode with
commit 7d0a1d5ebb
Author: Ben Widawsky <ben@bwidawsk.net>
Date: Sun Jun 24 20:35:57 2012 -0700
intel/decode: VERTEX_ELEMENT_STATE, 1 means valid
However the actual fix is just to update the reference file, and
include GEN7 in the decode.
Props to Eric Anholt for putting the test in distcheck, or else I
wouldn't have caught this.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
To support this we extract the common execbuf2 functionality to be
called with, or without contexts.
The context'd execbuf does not support some of the dri1 stuff.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
int drm_intel_gem_bo_wait(drm_intel_bo *bo, uint64_t timeout_ns)
This should bump the libdrm version. We're waiting for context support
so we can do both features in one bump.
v2: don't return remaining timeout amount
use get param and fallback for older kernels
v3: only doing getparam at init
prototypes now have a signed input value
v4: update comments
fall back to correct polling behavior with new userspace and old kernel
v5: since the drmIoctl patch was not well received, return appropriate
values in this function instead. As Daniel pointed out, the polling
case (timeout == 0) should also return -ETIME.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
On r6xx or evergreen z/stencil surface don't support linear or
linear aligned surface, force 1D tiled mode for those.
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
Valgrind can't understand some of the fields passed to ioctls are overwritten
by kernel, so we need to initialize them. Almost all of our ioctl wrappers
already do it and the cost of remaining 3 is very small.
Signed-off-by: Marcin Slusarz <marcin.slusarz@gmail.com>
A bitmask property is similar to an enum. The enum value is a bit
position (0-63), and valid property values consist of a mask of
zero or more of (1 << enum_val[n]).
Signed-off-by: Rob Clark <rob@ti.com>
Reviewed-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Certain cards report the the wrong bank setup which causes
surface init to fail in the ddx and leads to no accel.
If we hit an invalid tiling parameter, just set a default
value and disable 2D tiling.
Should fix:
https://bugs.freedesktop.org/show_bug.cgi?id=43448
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
make headers_install in kernel. Copy to here.
v2: signed ns_timeout
Acked-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
this patch adds libdrm_exynos helper layer that inclues some intefaces
for exynos specific gem and virtual display driver and also adds exynos
module name to modtest and vbltest.
Changelog v2:
- fixed exynos broken ioctl.
the pointer of uint64_t *edid should be removed.
- removed unnecessary definitions.
- added drm prime interfaces.
this feature is used to share a buffer between drivers or memory managers
and for this, please, refer to below links:
http://www.mjmwired.net/kernel/Documentation/dma-buf-sharing.txthttp://lwn.net/Articles/488664/
this patch is based on a link below:
git://anongit.freedesktop.org/mesa/drm
commit id: d72a44c7c4
Reviewed-by: Rob Clark <rob@ti.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
Signed-off-by: Rob Clark <rob@ti.com>
This patch adds a new function,
drm_intel_bufmgr_gem_set_aub_annotations(), which can be used to
annotate the type and subtype of data stored in various sections of
each buffer. This data is used to populate type and subtype fields
when generating the .aub file, which improves the ability of later
debugging tools to analyze the contents of the .aub file.
If drm_intel_bufmgr_gem_set_aub_annotations() is not called, then we
fall back to the old set of annotations (annotate the portion of the
batchbuffer that is executed as AUB_TRACE_TYPE_BATCH, and everything
else as AUB_TRACE_TYPE_NOTYPE).
Reviewed-by: Eric Anholt <eric@anholt.net>
This is the same list of PCI ids added by Alex Deucher in xf86-video-ati commit
aacbd629b02cbee3f9e6a0ee452b4e3f21376bd3.
This is needed since the addition of the surface allocator helper in
commit c51f7f0e46 ; it needs to differentiate
pre and post-R600 GPUs.
Therefore we should maintain another PCI id list.
Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=48138
Signed-off-by: Anisse Astier <anisse@astier.eu>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
In the future we'll have more than just connector properties, so create
a dump_prop function that can handle any property (instead of the
current dump_props function that only handles connector properties).
Also, make this function print a lot more information about the existing
properties.
Also change the printed indentation of the modes to make the output more
readable.
The previous function dump_props also segfaulted when we didn't have
enought permissions. The new function does not segfault in this case (by
checking for the return value of drmModeGetProperty).
Reviewed-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
24 (16 direct, 8 indirect) bytes in 1 blocks are definitely lost in loss record 2 of 7
at 0x402994D: malloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
by 0x4A25950: drmMalloc (xf86drm.c:147)
by 0x4A2E26D: drmModeGetPlaneResources (xf86drmMode.c:951)
by 0x4025FF: dump_planes (modetest.c:276)
by 0x4052AF: main (modetest.c:1120)
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Don't "continue" without freeing the connector.
192 bytes in 6 blocks are indirectly lost in loss record 6 of 12
at 0x4C2779D: malloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
by 0x4E30DD8: drmMalloc (xf86drm.c:147)
by 0x4E35024: drmAllocCpy (xf86drmMode.c:73)
by 0x4E35D69: drmModeGetConnector (xf86drmMode.c:507)
by 0x402F22: dump_connectors (modetest.c:181)
by 0x40261B: main (modetest.c:801)
Reviewed-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Use unsigned int instead of int:
- modetest.c:90:1: warning: comparison between signed and unsigned integer expressions [-Wsign-compare]
- modetest.c:98:1: warning: comparison between signed and unsigned integer expressions [-Wsign-compare]
- modetest.c:118:1: warning: comparison between signed and unsigned integer expressions [-Wsign-compare]
- modetest.c:286:16: warning: comparison between signed and unsigned integer expressions [-Wsign-compare]
- modetest.c:303:17: warning: comparison between signed and unsigned integer expressions [-Wsign-compare]
- modetest.c:694:16: warning: comparison between signed and unsigned integer expressions [-Wsign-compare]
- modetest.c:1088:16: warning: comparison between signed and unsigned integer expressions [-Wsign-compare]
The 'fd' variable is global, we don't need to pass it as an argument:
- modetest.c:998:40: warning: unused parameter ‘fd’ [-Wunused-parameter]
We don't use the 'modeset' variable:
- modetest.c:1025:8: warning: variable ‘modeset’ set but not used [-Wunused-but-set-variable]
V2: rebase, clear some more warnings
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Redesigned primarily to allow us to better take advantage of BO's having
fixed GPU virtual addresses on GeForce 8 and up, and to reduce the overhead
of handling relocations on earlier chipsets.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Christoph Bumiller <e0425955@student.tuwien.ac.at>
Sync drm.h with from kernel headers for the new PRIME_HANDLE_TO_FD
and PRIME_FD_TO_HANDLE ioctls from Dave Airlie's "drm: base prime/
dma-buf support (v5)" kernel patch.
Signed-off-by: Rob Clark <rob@ti.com>
When compiling with linaro toolchain version 4.6.2 got this warning.
CC omap_drm.lo
omap_drm.c: In function 'omap_bo_new_impl':
omap_drm.c:139:6: warning: 'bo' may be used uninitialized in this function [-Wuninitialized]
This patch initialize bo to NULL avoiding the warning.
Signed-off-by: Víctor Manuel Jáquez Leal <vjaquez@igalia.com>
Signed-off-by: Rob Clark <rob@ti.com>
In syncing with the corresponding kernel header, the wrong license
header was inadvertantly copied over. The intention was for the
userspace headers to have a MIT license following the convention
of the rest of libdrm, xorg, etc.
Signed-off-by: Rob Clark <rob@ti.com>
This adds libdrm_omap helper layer (as used by xf86-video-omap,
omapdrmtest, etc).
Signed-off-by: Rob Clark <rob@ti.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
[danvet: pushed for Rob, he doesn't yet have commit access.]
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
This improves the performance of Mesa's GL_MAP_UNSYNCHRONIZED_BIT path
in GL_ARB_map_buffer_range. Improves Unigine Tropics performance at
1024x768 by 2.30482% +/- 0.0492146% (n=61)
v2: Fix comment grammar.
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
drmIoctl returns -1 on error with errno set to the error value. Other
users of it in this file just check for != 0, and only use errno when
they need to send an error value on to the caller of the API.
This will allow the driver to capture all of its execution state to a
file for later debugging. intel_gpu_dump is limited in that it only
captures batchbuffers, and Mesa's captures, while more complete, still
capture only a portion of the state involved in execution.
This is a squash commit of a long series of hacking as we tried to get
the resulting traces to work in the internal simulator. It contains
contributions by Yuanhan Liu and Kenneth Graunke.
v2: Drop the MI_FLUSH_ENABLE setup.
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
For example:
export INTEL_DEVID_OVERRIDE=0x162
If this variable is set, don't actually submit the batchbuffer to the
GPU, it probably contains commands for the wrong generation of hardware.
v2: Introduce a getter for the overridden devid, and avoid getenv per exec.
Reviewed-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Signed-off-by: Eric Anholt <eric@anholt.net>
xf86drmMode.h is missing a header protection. xf86drm.h has one so just
copy it and adjust the name.
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
Note that the regression test complains here: The batch that was
captured included a bug in its packet output, which was later fixed in
Mesa.
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
This requires pulling the gen6 3DSTATE_WM out to a function so it
doesn't override gen7's handler.
v2: Fix pasteo in interpreting ZW interpolation (thanks danvet!).
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Every access to either the GTT or CPU pointer is supposed to be
proceeded by a set_domain ioctl so that GEM is able to manage the cache
domains correctly and for the following access to be coherent. Of
course, some people explicitly want incoherent, non-blocking access
which is going to trigger warnings by this patch but are probably better
served by explicit suppression.
v2: Also mark the pointers as inaccessible following the explicit unmap
and implicit unmap upon return to the cache.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
The empty string used for the not case is replaced by the default
if-else clause and so causes the configure to fail in the absence of
valgrind. Which is not quite what was intended.
Instead use the common idiom of setting a variable depending on whether
the true or false branch is taken and emit the conditional code as a
second step.
Reported-by: Tobias Jakobi <liquid.acid@gmx.net>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
In particular, declare the hidden CPU mmaps to valgrind so that it knows
about those memory regions.
v2: Add an additional VG_CLEAR for the getparam
References: https://bugs.freedesktop.org/show_bug.cgi?id=35071
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Acked-by: Ben Widawsky <ben@bwidawsk.net>
[anholt: Ideally valgrind should just learn about the ioctls, and
removing the clear for the non-valgrindified code feels risky.]
Reviewed-by: Eric Anholt <eric@anholt.net>
The mipmap level computation was wrong, we need to know the block
width, height, depth of compressed texture to properly compute this.
Change API to provide block width, height, depth instead of nblk_x,
nblk_y, nblk_z.
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
The surface allocator is able to build complete miptree when allocating
surface for r600/r700/evergreen/northern islands GPU family. It also
compute bo size and alignment for render buffer, depth buffer and
scanout buffer.
v2 fix r6xx/r7xx 2D tiling width align computation
v3 add tile split support and fix 1d texture alignment
v4 rework to more properly support compressed format, split surface pixel
size and surface element size in separate fields
v5 support texture array (still issue on r6xx)
v6 split surface value computation and mipmap tree building, rework eg
and newer computation
v7 add a check for tile split and 2d tiled
v8 initialize mode value before testing it in all case, reenable
2D macro tile mode on r6xx for cubemap and array. Fix cubemap
to force array size to the number of face.
v9 fix handling of stencil buffer on evergreen
v10 on evergreen depth buffer need to have enough room for a stencil
buffer just after depth one
Signed-off-by: Jerome Glisse <jglisse@redhat.com>
This adds support for querying the kernel about the LLC support in the
hardware.
In case the ioctl fails, we assume that it is present on GEN6 and GEN7.
v2: fix the return code checking
Signed-off-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
Commit efd6e81e inadvertently broke the build by looking for "i?86" or
"x86_64" in $host_os. The correct variable to check is $host_cpu.
This was preventing libdrm_intel.so from being built.
Reviewed-by: Chad Versace <chad.versace@linux.intel.com>
This fixes a failure in 'make check' found by the tinderbox when trying to
build this code on Linux/ppc. This code is only designed to run on
Intel platforms, so don't even bother building it if we're not in that set.
Found-by: Tinderbox
Signed-off-by: Jeremy Huddleston <jeremyhu@apple.com>
If the pci_device's actual gen was > 4, then we stupidly set
bufmgr_gem->gen = 6. Luckily this caused no bugs, and this fix shouldn't
change any behavior, because all checks against the gen currently have one
of the forms below:
gen == 2
gen == 3
gen >= 4
Reviewed-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
Signed-off-by: Chad Versace <chad.versace@linux.intel.com>
This just gets packet name and length in place, with the remainder
unfinished. I've long since finished the work that got me started
fixing up the decode.
Since CC_STATE_POINTERS for gen6 and 7 are quite different but use the
same opcode, move gen6 out to a helper function too, so we can use a
helper function for gen7.
This puts the error message in a consistent location relative to the
packet, and while I'm here I made the error message a bit more
informative.
Now, most static length packets need to just declare their length in
the table and not worry.
The overflow checks were all thoroughly untested, and a bunch of the
ones I'm deleting were pretty broken. Now, in the case of overflow,
you just decode data of 0xd0d0d0d0, and instr_out prints the warning
message instead. Note that this still has the same issue of being
under-tested, but at least it's one place instead of per-packet.
A couple of BUFFER_FAIL uses are left where the length to be decoded
could be (significantly) larger than a page, and the decode didn't
just call instr_out (which doesn't dereference data itself unless it's
safe).
The .batch was generated using the dump-a-batch branch of
git://people.freedesktop.org/~anholt/mesa
using glxgears on gen7 hardware, using INTEL_DEVID_OVERRIDE for
non-gen7 (this means that offsets in the buffers for non-gen7 are 0!).
The .ref was generated by:
./test_decode tests/gen7-3d.batch -dump.
The .sh exists because you can't supply arguments to tests using the
simple automake tests driver. Something reasonable could be done
using automake's parallel-tests driver (in fact, a previous version of
the patch did that), but I was concerned that:
1) The parallel-tests driver is documented to be unstable -- they may
change interfaces on us later.
2) The parallel-tests driver hides the output of tests in .log files
scattered all over the tree, which was ugly and more painful to
work with.
v2: Actually add the batch files, add a .gitignore for the *-new.txt
files added after failures, and fix failure mode for undetected
chipset name.
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch> (v1)
Consumers often want to choose stdout vs stderr, and for testing I
want to output to an open_memstream file.
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
It was producing an unused code warning. I'm tempted to just remove
it, since it's unused, but I *might* use it soon.
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Eugeni Dodonov <eugeni@dodonov.net>
I'd rather be able to use c99 variable declarations (there's a lot of
awful code layout due to being c90ish), but I'll leave that for later.
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Eugeni Dodonov <eugeni@dodonov.net>
There was plenty of dropped useful data, and some horribly
mis-formatted data.
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Eugeni Dodonov <eugeni@dodonov.net>
We've got a different (better) set of warning flags in place in this
tree.
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Eugeni Dodonov <eugeni@dodonov.net>
My plan is to use this drm_intel_dump_batchbuffer() interface for the
current GPU tools, and the current Mesa batch dumping usage, while
eventually building more interesting interfaces for other uses.
Warnings are currently suppressed by using a helper lib with CFLAGS
set manually, because the code is totally not ready for libdrm's warnings
setup.
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Eugeni Dodonov <eugeni@dodonov.net>
Some comments weren't wrapped, and for some reason uint32_t *data got
an extra space (while other instances of "type *identifier" didn't),
and the indentation of the opcode-list structs got trashed.
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Eugeni Dodonov <eugeni@dodonov.net>
We generally go for kernel style in this tree, and this 4-space indent
stuff was bothering me. The new results have some ugly bits, but
they're in places where we desperately want to be using helper
functions anyway.
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Eugeni Dodonov <eugeni@dodonov.net>
These will be used by intel_decode.c, and were taken from intel-gpu-tools.
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Eugeni Dodonov <eugeni@dodonov.net>
This will make these macros reusable from intel_decode.c, which
doesn't have a bufmgr_gem context, without faking the struct. We
should generally only be using these macros from bufmgr_gem context
setup anyway.
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Eugeni Dodonov <eugeni@dodonov.net>
This is from commit dd9a5b4f7f.
We've been sharing this file between that repo and Mesa, and it's time
to build a real interface using it. I'm also hoping to apply some of
its packet-walking logic for AUB dumping and batch validation
purposes.
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Acked-by: Eugeni Dodonov <eugeni@dodonov.net>
gem_flink|gem_open are DRM_AUTH ioctl, and being lazy we do not
establish ourselves as authenticated before testing the ioctls. So
instead of aborting, skip the test unless we have root privileges (and
so DRM_MASTER and the DRM_AUTH restriction no longer applies).
A future test could assert that the flink fails without proper
authentication.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=43924
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
During free we unconditionally delete the bo from the vma cache. This
relies on the its list member being kept in a sane state. This fails
after the object is purged, as the purge operation performs a pure
deletion and doesn't reset the list member, leaving a pair of dangling
pointers.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Add structs and functions necessary for the new plane and fb handling code,
including a new header, drm_fourcc.h, that includes the surface formats
supported by various DRM drivers.
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Hopefully all the bugs in the callers have been found, so time to
handle the failures "gracefully" again.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
As the max number of VMA mappings is a hard per-process limit, we need
to include the number of currently active mappings when evicting in
order to make room for a new mmap.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
There is a per-process limit on the number of vma that the process can
keep open, so we cannot keep an unlimited cache of unused vma's (besides
keeping track of all those vma in the kernel adds considerable overhead).
However, in order to work around inefficiencies in the kernel it is
beneficial to reuse the vma, so keep a MRU cache of vma.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
As a precautionary measure munmap on buffer free so that we never leak
the vma. Also include a warning during debugging.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Valgrind throws warns about a user-after-free if you try to bind a
new subchannel after the old one in that slot was freed,
so remove it from the channel list.
Signed-off-by: Maarten Lankhorst <m.b.lankhorst@gmail.com>
Before this, consumers of the libdrm API that might map a buffer
either way had to track which way was chosen at map time to call the
appropriate unmap. This relaxes that requirement by making
drm_intel_bo_unmap() always appropriate.
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
This lets us replace the current inner drawing loop of mesa:
for each prim {
compute bo list
if (check_aperture_space(bo list)) {
batch_flush()
compute bo list
if (check_aperture_space(bo list)) {
whine_about_batch_size()
fall back;
}
}
upload state to BOs
}
with this inner loop:
for each prim {
retry:
upload state to BOs
if (check_aperture_space(batch)) {
if (!retried) {
reset_to_last_prim()
batch_flush()
} else {
if (batch_flush())
whine_about_batch_size()
goto retry;
}
}
}
This avoids having to implement code to walk over certain sets of GL
state twice (the "compute bo list" step). While it's not a
performance improvement, it's a significant win in code complexity:
about -200 lines, and one place to make mistakes related to aperture
space instead of N places to forget some BO we should have included.
Note how if we do a reset in the new loop , we immediately flush. We
don't need to check aperture space -- the kernel will tell us if we
actually ran out of aperture or not. And if we did run out of
aperture, it's because either the single prim was too big, or because
check_aperture was wrong at the point of setting up the last
primitive.
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
A few of the bitfield-based booleans are left in place. Changing them
to "bool" results in the same code size, so I'm erring on the side of
not changing things.
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
It's more compatible; at least the Intel driver now rejects 32 bit
depths since it generally can't support real 32 bit framebuffers
(supports 30, 36, and 64 bit, but not 32).
To enable usage of xf86drm.h from C++ programs/frameworks.
Signed-off-by: Tapani Pälli <tapani.palli@intel.com>
[ickle: also wrap xf86drmMode.h]
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Both drmIoctl and ioctl define second argument as unigned long.
Debugging/tracing tools (like strace or valgrind) on 64-bit machines see
different request value for ioctls with 32nd bit set, because casting
signed int to unsigned long extends 32nd bit to upper word, so 0x80000000
becomes 0xFFFFFFFF80000000)
Nobody noticed because higher 32 bits are chopped off on their way to kernel.
Hi Alex,
Enclosed is a revised version of the patch sent on Mar 18, against
the master branch of the drm userspace (i.e. libdrm). Details
summarised in this thread:
http://lists.freedesktop.org/archives/dri-devel/2011-March/009499.html
This patch reconciles libdrm with the the kernel change that Dave
pushed this morning. It *supersedes* the previously sent patch (i.e.
apply it to the master branch as it exists at the time of this writing,
not as an incremental patch to the one sent previously).
Regards,
Ilija
Signed-off-by: Ilija Hadzic <ihadzic@research.bell-labs.com>
Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
... and if asked to open a bo by the same global name, return a fresh
reference to the previously allocated buffer.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
The nvc0 gallium drivers passes NULL here to indicate to the memory manager
that a buffer is being used, but without creating an actual reloc.
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
For relaxed fencing the object may only consume the small set of active
pages, but still requires a fence region once bound into the aperture.
This is the size we need to use when computing the maximum possible
aperture space that could be used by a single batchbuffer and so avoid
hitting ENOSPC.
Reported-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
It makes sure that GPU object destruction is executed in order with
respect to the previous FIFO commands.
Signed-off-by: Francisco Jerez <currojerez@riseup.net>
Acked-by: Ben Skeggs <bskeggs@redhat.com>
Both the consumers of this API (sync objects and client throttling)
were expecting this behavior. The kernel used to actually behave the
desired (but incorrect) way for us anyway, but that got fixed a while
back.
If bufmgr.bo_mrb_exec is not set, drm_intel_bo_mrb_exec returns ENODEV
even though drm_intel_gem_bo_mrb_exec2 will work fine for the RENDER ring.
Fixes xf86-video-intel after commit 'add BLT ring support' (5bed685f76)
with kernels without BSD or BLT ring support (2.6.34 and before).
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=31443
Signed-off-by: Albert Damen <albrt@gmx.net>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
The intent of these was to catch mismatched map/unmap. What it
actually did was check whether there was ever a mapping of that type
(including in a previous life of the buffer through the userland BO
cache), not whether they were mismatched. We don't even actually want
to catch mismatched map/unmap, unless we also do refcounting, since at
one point Mesa would do map/map/use/unmap/unmap. Just remove this
code instead.
nouveau_bo_unmap called the CPU_FINI IOCTL even if it was a NOSYNC
mapping. It caused no harmful effects (actually CPU_FINI is a no-op on
recent enough kernels) besides the precious CPU cycles being wasted.
Signed-off-by: Francisco Jerez <currojerez@riseup.net>
The kernel has always allowed userspace to underallocate objects
supplied for fencing. However, the kernel only allocated the object size
for the fence in the GTT and so caused tiling corruption. More recently
the kernel does allocate the full fence region in the GTT for an
under-sized object and so advertises that clients may finally make use
of this feature. The biggest benefit is for texture-heavy GL games on
i945 such as World of Padman which go from needing over 1GiB of RAM to
play to fitting in the GTT!
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
As the higher layers check the error return from libdrm-intel and
are supposed to handle the error (and print their own warning in
extremis) the voluminous output on stderr is just noise and a hazard in
its own right.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
This works in conjunction with newer kernels. If we succeed in requesting
interface 1.4, the we know the kernel provides proper domain numbers. If
not, ignore the domain number as it's bogus (except on Alpha).
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Adam Jackson <ajax@redhat.com>
The high layers expect to receive a status code on error (on the
pessimistic assumption that the errno value will have been overwritten
by the time the failure is propagated all the way up), so convert
xf86drmMode.c to return -errno on an ioctl error and be consistent with
the rest of the libdrm API.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
If the mapping succeeds we have a valid pointer. If setting the domain
failures we may incur cache corruption. However the usual failure mode
is because of a hung GPU, in which case it is preferable to ignore the
minor error from setting the domain and continue on oblivious. If
these errors persist, we should rate limit the warning [or even just
remove it].
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Mesa uses the returned pitch from alloc_tiled, so make sure that we set
it correctly before modifying the stride used for the SET_TILING call.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Ensure that the user doesn't attempt to specify a stride to use with a
linear buffer by forcing such to be zero.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
execbuffer() returns ENOSPC if it cannot fit the batch buffer into the
aperture which is the error we want to diagnose here.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Rearrange the cache cleanup so that we always scan following a final
unreference, and guard against multiple scans in a single second.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
When allocating a tiled buffer, if we remove the desired tiling mode due
to it being beyond hardware limits, also remove the stride. This ensures
that we only ever use stride 0 with I915_TILING_NONE.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
As we now expose a method to allocate tiled buffers, it makes more sense
to defer the SET_TILING until required. Besides the slim chance that it
will be a no-op, by delaying the change we are less likely to stall on
waiting for a bound buffer to release a fence register.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
We need to inform the kernel if the tiling stride changes and not only
for changes of the tiling mode.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
We had two cases recently where the rounding to powers of two hurt
badly: 4:2:0 YUV HD video frames would round up from 2.2MB to 4MB, and
Urban Terror was hitting aperture size limitations. For UT, this is
because mipmap trees for power of two texture sizes will land right in
the middle between two cache buckets.
By giving a few more sizes between powers of two, Urban Terror on my
945 ends up consuming 207MB of GEM objects instead of 272MB, and HD
video decode on Ironlake goes from 99MB to 75MB.
cairo-perf-diff of the benchmarks for gl and xlib shows a 1.09x and
1.06x speedup and a 1.07x, 1.08x, and 1.11x slowdown. From this, I
think this patch was really a no-op in terms of performance for these
CPU-bound workloads.
If the pitch is too large for the hardware to tile, recompute the
required surface size based on the untiled pitch and alignments. For the
older hardware, which has smaller limits and greater restrictions, this
may be a considerable saving in allocation size.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
This introduces a new API to exec on BSD ring buffer, for H.264 VLD
decoding.
Signed-off-by: Xiang Hai hao <haihao.xiang@intel.com>
Signed-off-by: Zou Nan hai <nanhai.zou@intel.com>
Fixes:
Bug 26686 - Some textures are distorted with libdrm 2.4.18 in GTAVC>A3
http://bugs.freedesktop.org/show_bug.cgi?id=26686
This bug continues to haunt me. The kernel SET_TILING ioctl is
inconsistent in its return values when reporting an error. If one of its
sanity checks fail, then the input values are left unchanged. If the
kernel later fails to change the tiling mode, then the input values are
modified to match the current tiling on the object. In short, userspace
cannot trust the return values upon error and so we must assume that
upon error our current tiling mode matches reality and not update.
This reverts commit 7ca558494d.
This was pushed ahead of an essential review of bo level locking in
mesa, without which we cannot know whether removing this lock is safe.
Currently, all the tests for libdrm are built during 'make all', even
if you do not wish to run tests. Attached is a patch, based on
version 2.4.15, to make the tests build in 'make check'.
Thomas tracked down this error with kdm and commit b509640:
==4320== Invalid write of size 8
==4320== at 0x9A97998: do_bo_emit_reloc (in /usr/lib/libdrm_intel.so.1.0.0)
==4320== by 0x9A97B9C: drm_intel_gem_bo_emit_reloc (in /usr/lib/libdrm_intel.so.1.0.0)
==4320== by 0xAED3234: intel_batchbuffer_emit_reloc (in /usr/lib/xorg/modules/dri/i965_dri.so)
==4320== by 0xAF13827: brw_emit_vertices (in /usr/lib/xorg/modules/dri/i965_dri.so)
==4320== by 0xAF1F14D: brw_upload_state (in /usr/lib/xorg/modules/dri/i965_dri.so)
==4320== by 0xAF12122: brw_draw_prims (in /usr/lib/xorg/modules/dri/i965_dri.so)
==4320== by 0xB256824: vbo_exec_vtx_flush (in /usr/lib/xorg/modules/dri/libdricore.so)
==4320== by 0xB2523BB: vbo_exec_FlushVertices_internal (in /usr/lib/xorg/modules/dri/libdricore.so)
==4320== by 0xB252411: vbo_exec_FlushVertices (in /usr/lib/xorg/modules/dri/libdricore.so)
==4320== by 0xB195A3D: _mesa_PopAttrib (in /usr/lib/xorg/modules/dri/libdricore.so)
==4320== by 0x8DF0F02: __glXDisp_Render (in /usr/lib/xorg/modules/extensions/libglx.xorg)
==4320== by 0x8DF517F: __glXDispatch (in /usr/lib/xorg/modules/extensions/libglx.xorg)
==4320== Address 0x126a8b80 is 0 bytes after a block of size 16,368 alloc'd
==4320== at 0x4C23E03: malloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==4320== by 0x9A97A64: do_bo_emit_reloc (in /usr/lib/libdrm_intel.so.1.0.0)
==4320== by 0x9A97B9C: drm_intel_gem_bo_emit_reloc (in /usr/lib/libdrm_intel.so.1.0.0)
==4320== by 0xAED3234: intel_batchbuffer_emit_reloc (in /usr/lib/xorg/modules/dri/i965_dri.so)
==4320== by 0xAF191DB: upload_binding_table_pointers (in /usr/lib/xorg/modules/dri/i965_dri.so)
==4320== by 0xAF1F14D: brw_upload_state (in /usr/lib/xorg/modules/dri/i965_dri.so)
==4320== by 0xAF12122: brw_draw_prims (in /usr/lib/xorg/modules/dri/i965_dri.so)
==4320== by 0xB255EF6: vbo_exec_DrawArrays (in /usr/lib/xorg/modules/dri/libdricore.so)
==4320== by 0x8DF67A3: __glXDisp_DrawArrays (in /usr/lib/xorg/modules/extensions/libglx.xorg)
==4320== by 0x8DF0F02: __glXDisp_Render (in /usr/lib/xorg/modules/extensions/libglx.xorg)
==4320== by 0x8DF517F: __glXDispatch (in /usr/lib/xorg/modules/extensions/libglx.xorg)
==4320== by 0x446293: ??? (in /usr/bin/Xorg)
which is simply due to only allocating space for the pointers and not
the structs themselves. D'oh.
Reported-by: Thomas Bächler <thomas@archlinux.org>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
This reverts commit 245d695779.
libdrm now install headers in ${includedir}/libdrm and /usr/include/drm
is reserved for the kernel headers. We should only ever add one of these
to the CFLAGS.
The motivation behind this is that by shipping it here, it's essentially
an API which causes issues while bisecting across changes to the header
files.
Dump command stream + associated bo into a binary file
which follow a similar design as json file. It allows
to intercept a command stream and replay it in a standalone
program (see radeondb tools).
Fixes problem that libdrm_radeon was disabled in Makefile even when configure
claimed that radeon was enabled.
Signed-off-by: Pauli Nieminen <suokkos@gmail.com>
intel_bufmgr.h is installed in ${includedir} directly, and the other
headers are taken care of by libdrm.pc's Cflags.
Signed-off-by: Julien Cristau <jcristau@debian.org>
Signed-off-by: Eric Anholt <eric@anholt.net>
This is the largest untiled pitch requirement from gen2 through gen4.
It's only the case for gen3 rendering to color regions with depth, but
it's rare for this to be a significant factor in memory usage -- for
example, gen4 requires 1 or 2 times the element size, or up to 64
bytes depending on the size of the elements. This is easier than
encoding all the various little quirks for untiled pitch alignment,
since we rarely do untiled now.
bo->referenced_in_cs is checked if bo is already in cs. Adding and removing
reference in bo is done with atomic operations to allow parallel access to a
bo from multiple contexts.
cs->id generation code quarentees there is not duplicated ids which limits
number of cs->ids to 32. If there is more cs objects rest will get id 0.
V2:
- Fix configure to check for atomics operations if libdrm_radeon is only selected.
- Make atomic operations private to libdrm.
This optimization decreases cs_write_reloc share of torcs profiling from 4.3%
to 2.6%.
Tested-by: Michel Dänzer <michel@daenzer.net>
Signed-off-by: Pauli Nieminen <suokkos@gmail.com>
intel_atomic.h includes very usefull atomic operations for
lock free parrallel access of variables. Moving these to
core libdrm for code sharing with radeon.
Signed-off-by: Pauli Nieminen <suokkos@gmail.com>
Ensure that errors from the kernel are propagated back to the caller,
and not masked with return 0;
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
This allows Mesa to use drm_intel_bo_alloc_tiled() for its tiled
buffers, since it makes its decision about pitch before telling
libdrm. They happen to be the same choices for the tiled case.
- Currently reloc'ing a user bo to gart will first cause an allocation in vram,
which is then written to by cpu, then the bo gets moved to gart.
Acked-by: Francisco Jerez <currojerez@riseup.net>
Signed-off-by: Maarten Maathuis <madman2003@gmail.com>
This patch to libdrm adds support for the new execbuf2 ioctl. If
detected, it will be used instead of the old ioctl. By using the new
drm_intel_bufmgr_gem_enable_fenced_relocs(), you can indicate that any
time a fence register is actually required for a relocation target you
will call drm_intel_bo_emit_reloc_fence instead of
drm_intel_bo_emit_reloc, which will reduce fence register pressure.
Signed-off-by: Eric Anholt <eric@anholt.net>
This commit encompasses the changes necessary to run on top of the 0.0.16
nouveau interface, additional APIs to support the new features of the
interface, as well as code from Luca Barbieri to improve the pushbuf
interface, which just happens to break nouveau's libdrm ABI so was delayed
until now.
API changes as a result of 0.0.16 DRM interface:
1. No more bo_pin()/bo_unpin(), these were only there for UMS and we no
longer support it.
2. Any random nouveau_bo can be submitted to the GPU as a push buffer.
3. Relocations can be applied on any nouveau_bo
This patch changes the pushbuffer ABI to:
1. No longer use/expose nouveau_pushbuffer. Everything is directly
in nouveau_channel. This saves the extra "pushbuf" pointer dereference.
2. Use cur/end pointers instead of tracking the remaining size.
Pushing data now only needs to alter cur and not both cur and remaining.
The goal is to make the *_RING macros faster and make the interface simpler
and cleaner in the process.
The *_RING APIs are unchanged, but those are inlined and the ABI is changed.
Also, anything accessing pushbuf->remaining instead of using AVAIL_RING
will need to be fixed.
The SET_TILING is pernicious in that it overwrites the input arguments
following an error in order to report the current tiling state of the
buffer. This caught us by surprise as we then fed those arguments back
into to the ioctl unmodified following an EINTR and so the kernel then
reported success for the no-op. We interpreted this success as meaning
that the tiling on the buffer had changed so updated our state and
started using the buffer incorrectly in the new tiled/untiled manner.
This lead to all sorts of random corruption and GPU hangs, even though
the batch buffers would look sane (when the GPU had not wandered off
into forbidden territory).
References:
Bug 25475 - [i915] Xorg crash / Execbuf while wedged
http://bugs.freedesktop.org/show_bug.cgi?id=25475
Bug 25554 - i830_uxa_prepare_access: gtt bo map failed: Input/output error
http://bugs.freedesktop.org/show_bug.cgi?id=25554
(And probably every other weird bug in the last few months.)
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
As the kernel reports the total number of fences, we must guess how many
fences are likely to be pinned. In the typical system these will be only
used by the scanout buffers, of which there may be one per pipe, and any
number of manually pinned fenced buffers. So take a conservative guess
and reserve two fences for use by the system.
Note this reduces the number of fences to 3 for i915 and prior.
Reference:
http://bugs.freedesktop.org/show_bug.cgi?id=25911
The latest intel driver 2.10.0 causes kernel oops and system hangs
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
If there is section size mismatch reusing the section object
makes section start fail.
Reseting the object before doing error checking prevents the
possible flood of errors.
List of changes:
Fixes the cursor size to 64x64, you still need ti supply width and height
Explicitly make the cursor format A8R8G8B8
Explicitly make the scanout format X8R8G8B8
We don't intend libdrm-radeon to become clever enough to
decode cs for all GPU we support. Better to let an external
tool do the job. This will print raw cs in an easy to parse
way.
If the available modes changes between the two GETCONNECTOR ioctls, that
caused the kernel to skip filling one array and led to a crash (as the size
of the allocated and initialised block of memory differed from the reported
size, and might be NULL if no modes were present at first).
This bug manifest its self on my machine due to spurious false positive
detections of a connected TV-out.
Fixes: http://bugs.freedesktop.org/show_bug.cgi?id=25912
Crash whilst probing modes
Based upon the similar fixes for the GETRESOURCES ioctls by Chris Wilson,
in the following commits:
commit e6c136ca7a
commit 85fb3e55fd
commit d1308f4fe7
Signed-off-by: Peter Clifton <pcjc2@cam.ac.uk>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
If the count is 0, then the malloc is permitted to return NULL, so don't
throw an error in that case.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Peter Clifton hit an issue whereby he had a spurious TV hotplug event
that occurred between the two GETRESOURCES ioctls that caused the kernel
to skip filling one array and led to a crash (as the size of the
allocated and initialised block of memory differed from the reported
size).
Fixes: http://bugs.freedesktop.org/show_bug.cgi?id=25912
Crash whilst probing modes
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reported-by: Peter Clifton <pcjc2@cam.ac.uk>
as Michel pointed out we are exposing too much info for these object
for this to be maintainable going forward.
This patch set minimises the exposed parts of the radeon_bo and
radeon_cs objects to the piece necessary for ddx/mesa to operate
at a decent speed.
The major problem is mesa contains a legacy BO/CS managers which we still
need to expose functionality to, and we really cannot change the API
until we can drop the non-KMS codepaths.
Signed-off-by: Dave Airlie <airlied@redhat.com>
Don't store the error return in bo_gem->gtt_virtual or else we will
attempt to use that as a valid pointer in future mappings.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
This removes the foremost prolific user of mutexes in libdrm_intel.so.
The other uses of the bufmgr_gem->mutex to serial access to individual
bos are currently required by Mesa, and are far less frequent.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
[anholt: This chunk looks good...]
Acked-by: Eric Anholt <eric@anholt.net>
This has the unfortunate behaviour of releasing our malloc cache, but
the alternative is for X to consume a couple of gigabytes of ram and
die during testing. Fortunately the extra mallocs have little impact on
performance whereas avoiding swap and death, lots.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Instead of forcing the caller to check after every emit_reloc(), we can
flag the object as being in error, propagating that error upwards through
the relocation tree, and failing the eventual batch buffer execution.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
EAGAIN cannot be raised by the current code, but the system call maybe
interrupted and so return EINTR.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Hitting this error lead to a segfault:
intel_bufmgr_gem.c:919: Error mapping buffer 48607 (pixmap):
Cannot allocate memory.
because the errno was reused as the function return value after being
reset by the fprintf(), so caller thought the mapping had succeeded. The
convention established by libdrm is that the return value is the
negative errno and that uses of libdrm cannot trust the value of errno
afterwards, but must use the return code.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Buffers on the relocation tree are guarded by the reference to the batch
object and so do not need an extra reference whilst constructing the
list of execution buffer objects.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Having been bitten by a missing EINTR check during mmap_gtt(), I thought
it prudent to add some more protection around the ioctls.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
This allows us to keep the assert added in the previous commit that we do
not modify the tree_reloc_size after inserting the buffer into a relocation
tree, which was being hit here:
#0 0xb78c2424 in __kernel_vsyscall ()
#1 0xb74f6401 in raise () from /lib/libc.so.6
#2 0xb74f7b42 in abort () from /lib/libc.so.6
#3 0xb74ef5a8 in __assert_fail () from /lib/libc.so.6
#4 0xb737e78b in drm_intel_bo_gem_set_in_aperture_size (bufmgr_gem=<value optimized out>, bo_gem=0x6) at intel_bufmgr_gem.c:373
#5 0xb737f519 in drm_intel_gem_bo_set_tiling (bo=0xa1030a0, tiling_mode=0xbff6c85c, stride=0) at intel_bufmgr_gem.c:1386
#6 0xb737f67f in drm_intel_gem_bo_unreference_final (bo=0xa1030a0, time=<value optimized out>) at intel_bufmgr_gem.c:768
#7 0xb737f5e3 in drm_intel_gem_bo_unreference_locked_timed (bo=0xa1e50d0, time=<value optimized out>) at intel_bufmgr_gem.c:805
#8 drm_intel_gem_bo_unreference_final (bo=0xa1e50d0, time=<value optimized out>) at intel_bufmgr_gem.c:756
#9 0xb737fcbb in drm_intel_gem_bo_unreference (bo=0xa1e50d0) at intel_bufmgr_gem.c:821
#10 0xb737b4e6 in drm_intel_bo_unreference (bo=0x0) at intel_bufmgr.c:80
#11 0xb7325625 in intel_batch_flush (scrn=0x9d91f78, flush=1) at i830_batchbuffer.c:200
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
For the older chipsets, i.e. pre-i965, which have severe alignment
restrictions for tiled buffers we need to pessimistically assume that we
will waste the size of buffer to meet those alignment constraints.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
If the kernel immediately frees the backing store for a buffer when
marking it purgeable, then there is not point adding to the cache. Free
it immediately, instead.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
The nouveau_drmif.h is not the Nouveau DRM kernel ABI file, but purely
user space stuff. Remove it, it does not belong in include/drm/.
Copy the right header from Nouveau kernel v2.6.31-rc9-757-gaca551c.
Signed-off-by: Pekka Paalanen <pq@iki.fi>
The old code increments the command stream size by another kbyte, but does
not make sure that the requested packet size fits into the stream. The patch
ensures that the whole next packet fits there and rounds the allocated size to
a power of two.
Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
The kernel will now write data to the DRM fd for various event types if
requested. Currently, the only supported event is a vblank event: it contains
the vblank count for the event as well as a timestamp from when the event
ocurred. Since the DRM fd is now pollable, it's easy to integrate into
existing event loops.
Notably when freeing a batchbuffer, we often end up freeing many of the
buffers it points at as well. Avoiding repeated calls brings us a 9% CPU
win for cairo-gl.
[ # ] backend test min(s) median(s) stddev. count
before:
[ 0] gl firefox-talos-gfx 58.941 58.966 0.75% 3/3
after:
[ 0] gl firefox-talos-gfx 54.186 54.195 0.49% 3/3
If the target we're asking about hasn't ever been used as a relocation
target, then it obviously hasn't been used as a target by the batch's reloc
tree. This is the common case for good GL programming where you only map
fresh buffers, and gives us a 5% win in cairo-gl.
[ # ] backend test min(s) median(s) stddev. count
before:
[ 0] gl firefox-talos-gfx 64.680 64.756 0.06% 3/3
after:
[ 0] gl firefox-talos-gfx 60.816 60.970 0.29% 3/3
Only build libdrm_intel automatically if we have support for atomic
operations. To force configure to build drm pass --enable-intel, which
will cause the configure to error if no support is found. Or pass
--disable-intel to explicitly prevent libdrm_intel from being built.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
I slipped it in with the alloc_tiled changes, since we were explicitly
throwing the parameter away. It caught some bogus released code, which
we've now fixed, so remove the asserts to keep old drivers working.
This simplifies driver code in handling object allocation, and also gives us
an opportunity to possibly cache tiled buffers if it turns out to be a win.
[anholt: This is chopped out of the execbuf2 patch, as it seems to be useful
separately and cleans up the execbuf2 changes to be more obvious]
In conjunction with the atomic operation patch, it may be more
convenient for some people to disable building libdrm-intel and its
dependencies upon the atomic intrinsics then it is for them to use a
supported compiler.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
As the target architecture for Intel GPUs is the x86, we can presume to
have reasonable compiler support for Intel atomic intrinsics, i.e. gcc,
and so use those in preference to pulling in a complicated mess of
fragile assembly.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
[anholt: hand-resolved against my previous commit. This brings cairo-gl
firefox-talos-gfx time from 65 seconds back down to 62 seconds.]
Signed-off-by: Eric Anholt <eric@anholt.net>
Set the DONTNEED flag on cached buffers so that the kernel is free to
discard those when under memory pressure.
[anholt: This takes firefox-talos-gfx time from ~62 seconds to ~65 seconds
on my GM965, but it seems like a hit worth taking for the improved
functionality from saving memory]
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Eric Anholt <eric@anholt.net>
There are a bunch of places in GL where if we can't do this we have to
flush the batchbuffer, and the cost of lookups here is outweighed by flush
savings.
I thought I was going to do all sorts of crazy experiments with it. I never
did, and it turned out the free-after-a-few-seconds plan is working out fine.
New ids for G41, Clarkdale and Arrandale.
Make sure we don't need to count fence also on new chips.
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
This reverts commit 0a732983f0.
Paul Nieminen and Dave Airlie pointed out on IRC that this shouldn't be
necessary. I was seeing visual corruption in X before I made this change, but
I can't reproduce that anymore so it was probably an unrelated issue.
airlied: modified the interface to drop busy return value, just return
it normally, also fixed int->uint32_t for domain
Signed-off-by: Pauli Nieminen <suokkos@gmail.com>
Using this call in OUT_BATCH_TABLE reduces radeonEmitState cpu usage from
9% to 5% and emit_vpu goes from 7% to 1.5%. I did use calgrind to profile
gears for cpu hotspots with r500 card.
Signed-off-by: Pauli Nieminen <suokkos@gmail.com>
- The bo was mapped with sysmem == NULL, so this means cpu prep is called.
- The bo was unmapped with sysmem != NULL, so this means cpu finish is not called.
- This can lead to a non-zero "cpu writers" count in ttm_bo.
The goal of the BO cache is to keep buffers on hand for fast continuous use,
as in every frame of a game or every batchbuffer of the X Server. Keeping
older buffers on hand not only doesn't serve this purpose, it may hurt
performance by resulting in disk cache getting kicked out, or even driving
the system to swap.
Bug #20766.
Several POSIX extensions are used in the libdrm code (e.g., mknod and ffs).
Set _XOPEN_SOURCE and _GNU_SOURCE to something reasonable to ensure that
prototypes for these functions are available. This is done in configure.ac
using AC_USE_SYSTEM_MACROS. This requires autoconf 2.60 or later. Eventually
the code should check for the existance of these defines and do something
reasonable if they are not available.
Inspired by a patch by Pauli Nieminen and suggestions from Julien Cristau.
Thanks.
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
The logbase2 would overflow and wrap the size around to 0, making the code
allocate a 4kb object instead. By simplifying the code to just walk the
14-entry bucket array comparing sizes instead of indexing on
ffs(1 << logbase2(size)), we avoid silly math errors and have code of
approximately the same speed.
Many thanks to Simon Farnsworth for debugging and providing a working patch.
Bug #27365.
Normal map() should operate as before, and map_range()/map_flush() should
give correct results but lacking any performance difference from map().
Nothing exiting being done here yet, but the interface is a good start.
The commit 651e3dc6dd, "drm: Fix
compilation on 2.6.30" broke all builds on kernels before 2.6.26.
Variadic functions cannot be inlined.
The variadic inline function is replaced with a variadic macro.
Also, the function dev_name() added by that same commit is never used,
so it is removed.
Signed-off-by: Pekka Paalanen <pq@iki.fi>
No idea what this reg does.. NVIDIA put 0x802 into it on some chips, the
criteria aren't too clear on when they do that however. Hopefully 0x800
will work everywhere...
Patch allows the backlight to be manipulated under gnome on apple powerpc
based NV30 machines. It works fine on my powerbook, and should also work
for older NV17/NV18 machines.
Note that older powerpc specific tools (pbbuttonsd) have some problems with
this implementation (because the device is not yet there at the start time
of the daemon, and the code makes incorrect assumptions about the max
brightness values). However, IMHO these things can and should be addressed
in the daemon.
Some style/warning fixes applied by Stuart Bennett <stuart@freedesktop.org>
This avoids making objects significantly bigger than they would be
otherwise, which would result in some failing at binding to the GTT.
Found from firefox hanging on:
http://upload.wikimedia.org/wikipedia/commons/b/b7/Singapore_port_panorama.jpg
due to a software fallback trying to do a GTT-mapped copy between two 73MB
BOs that were instead each 128MB, and failing because both couldn't fit
simultaneously.
The cost here is that we get no opportunity to cache these objects and
avoid the mapping. But since the objects are a significant percentage
of the aperture size, each mapped access is likely having to fault and rebind
the object most of the time anyway.
Bug #20152 (2/3)
The convention is that all APIs are per-bufmgr, so make this one the same.
Then, have it return -1 on failure so that the application can know what's
going on and do something sensible.
Signed-off-by: Keith Packard <keithp@keithp.com>
This wraps the new DRM_IOCTL_I915_GET_PIPE_FROM_CRTC_ID ioctl,
allowing applications to discover the pipe number corresponding
to a given CRTC ID. This is necessary for doing pipe-specific
operations such as waiting for vblank on a given CRTC.
Scanout buffers need to be freed through the kernel as it holds a reference
to them; exposing this API allows applications allocating scanout buffers to
flag them as not reusable.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
Add assertions to drm_intel_gem_bo_reference,
drm_intel_gem_bo_reference_locked and drm_intel_gem_bo_unreference_locked
that the object has not been freed (refcount > 0). Mistakes in refcounting
lead to attempts to insert a bo into a free list more than once which causes
application failure as empty free lists are dereferenced as buffer objects.
Signed-off-by: Keith Packard <keithp@keithp.com>
This patch speeds up drmModeGetConnector by pre-allocating mode &
property info space before calling into the kernel. In many cases this
pre-allocation will be sufficient to hold the returned values (it's easy
enough to tweak if the common case becomes larger), which means we don't
have to make the second call, which saves a lot of time.
Acked-by: Jakob Bornecrantz <wallbraker@gmail.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
libdrm has some support for GTT mapping already, but there are bugs
with it (no surprise since it hasn't been used much).
In fixing 20803, I found that sharing bo_gem->virtual was a bad idea,
since a previously mapped object might not end up getting GTT mapped,
leading to corruption. So this patch splits the fields according to
use, taking care to unmap both at free time (but preserving the map
caching).
There's still a risk we might run out of mappings (there's a sysctl
tunable for max number of mappings per process, defaulted to 64k or so
it looks like) but at least GTT maps will work with these changes (and
some others for fixing PAT breakage in the kernel).
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
drm_handle_t is defined to be a u32 on linux and a u64 on everything
else. This addresses an issue on FreeBSD amd64 where the map offsets
may be greater than 32bits. When the handle is cast to 32bit, mmap
cannot match the requested map and causes X to crash.
This should be a NOOP on linux since drm_handle_t is always 32bit.
Signed-off-by: Robert Noland <rnoland@2hip.net>
NV04 was completely busted. Push buffers were getting allocated at the
end of VRAM, overwriting PRAMIN. So, it turns out PRAMIN is in VRAM on
all chips. Question answered!
Intel 855 chips present the same pci id for both heads. This prevents
us from attaching to the dummy second head. All other chips that I
am aware of either only present a single pci id, or different ids
for each head so that we only match on the correct head.
The commit 957b10695b, "Move vblank_init
to driver load time." forgot to add the function declaration in
linux-core/drmP.h.
Signed-off-by: Pekka Paalanen <pq@iki.fi>
This helps avoid the n^2 performance cost of counting tree size when we
get a lot of relocations into our batch buffer. rgb10text on keithp's laptop
went from 136k glyphs/sec to 234k glyphs/sec.
This avoids using the oldest BO in the BO cache and waiting for it to be
idle before we turn around and render to it with the GPU. Thanks to
Chris Wilson for pointing out how silly we were being.
Loading nouveau.ko would fail with unknown symbols, if the backlight
class device support is not provided in the kernel. Let's make the
backlight support dependant on the kernel configuration.
This is a bit ugly, the proper way would be to check for the config in
Makefile.kernel whether to build nouveau_backlight.o at all, and if not,
nouveau_drv.h should provide the stubs.
Signed-off-by: Pekka Paalanen <pq@iki.fi>
Several nvidia-based systems don't support backlight control via the
standard ACPI control mechanisms. Instead, it's necessary for the driver
to modify the backlight control registers directly. This patch adds
support for determining whether the registers appear to be in use, and
if so registers a kernel backlight device to control them. The backlight
can then be controlled via existing userspace tools.
Signed-off-by: Matthew Garrett <mjg@redhat.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
This function is unused, and yet creates build problems: the symbol
init_mm is not exported by the latest -rc kernels and I don't believe it
ever will be. Even CONFIG_UNUSED_SYMBOLS does not provide it anymore.
If this function is needed in the future, it needs to be reinvented in
any case. So remove it.
Signed-off-by: Pekka Paalanen <pq@iki.fi>
Intel developers have stated, that their DRM development continues
elsewhere in some Linux kernel trees. This makes the code in drm.git
just dead weight. This removal allows further cleanup of compatibility
code.
shared-core and bsd-core are left untouched this time.
Signed-off-by: Pekka Paalanen <pq@iki.fi>
Acked-by: Eric Anholt <eric@anholt.net>
nouveau_notifier.c had two places where void* was used in arithmetic,
fixed by using char*.
nouveau_dma_wait(), nouveau_notifier_wait_status() and
nouveau_resource_alloc() had signed/unsigned comparison warnings, fixed
by changing the function parameter into an unsigned type.
Signed-off-by: Pekka Paalanen <pq@iki.fi>
NVIDIA do this fun little sequence after updating the PRAMIN page tables.
On 9xxx chips, none of the PRAMIN BAR bindings (except the initial one)
worked, hence the majority of the setup needed to create a channel
ended up in the wrong place, causing all sorts of fun.
This is done by NVIDIA on nv8x chips also, so we'll do it for them too,
even though they appear to work without it.
I'm not 100% sure that the nv94 one we were using won't work. The context
layouts are identical (well.. same ctxprog, so of course!), only a couple
of registers differ. But, be safe until we actually get some 9xxx chips
working.
Our PFIFO/PGRAPH context save/load functions don't really work well
(at all?) on nv5x yet. Depending on what random state the card is
in before the drm loads, fbcon probably won't work correctly.
Luckily we've setup the GPU in such a way that it'll actually do a
hw context switch for the first context. Not sure of how successful
this'd be currently on the older chips (actually, pretty sure it won't
work), so NV50 only for now.
This patch tries to use the available fence count to figure out whether a
given batch can succeed or not (just like the aperture check).
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Eric Anholt <eric@anholt.net>
drm_fops.c reads the current process' EUID directly from task_struct.
Apparently starting in 2.6.28-rc4 this fails to build.
In Linus' tree, commit b6dff3ec5e116e3af6f537d4caedcad6b9e5082a
"CRED: Separate task security context from task_struct"
moves the euid field from task_struct to another struct.
Earlier commit 9e2b2dc4133f65272a6d3c5dcb2ce63f8a87cae9
"CRED: Introduce credential access wrappers" implements the wrapper
macros to access e.g. euid. This is in 2.6.27-rc4, and this contains the
definition of current_euid() that will be used in the DRM compatibility header
for kernels before 2.6.27. That commit also creates <linux/cred.h>, which
contains the upstream definition of current_euid().
drm_fops.c is fixed to use current_euid(), and drm_compat.h will offer
the compatibility definition for kernels <2.6.27.
Signed-off-by: Pekka Paalanen <pq@iki.fi>
It is impossible to replace the original semantics of this call purely
in userland, since the fb_id would change.
after discussion with Dr_Jakob
Signed-Off-By: Owain Ainsworth <oga@openbsd.org>
Acked-By: Jakob Bornecrantz <jakob@vmware.com>
Michel caught a case where we might overwrite a success or other return
value with EBUSY, so check the return value before checking for the
timeout condition.
In some cases, vblank interrupts may be disabled or otherwise broken.
The kernel has a 3s timeout builtin to handle these cases, but the X
server's SIGALM for cursor handling may interrupt vblank wait ioctls,
causing libdrm to restart the ioctl, making the kernel's timeout
useless.
This change tracks time across ioctl restarts and returns EBUSY to the
caller if the expected vblank sequence doesn't occur within 1s of the
first call.
Fixes fdo bz #18041, which is caused by a drmWaitVBlank hanging due to
the corresponding pipe getting disabled (thus preventing further events
from coming in).
The values are really going to continue meaning pipe, not plane, and that's
what they're called in the kernel copy of the header. Userland hasn't ever
made the switch to pipe!=plane, since userland checks are based on DRM
version, which is still stuck at 1.6. However, Mesa did start using
plane[AB] names, so provide a compat define.
Remember tiling mode values provided by appplications, and
record tiling mode when creating a buffer from another application. This
eliminates any need to ask the kernel for tiling values and also makes
reused buffers get the right tiling.
Signed-off-by: Keith Packard <keithp@keithp.com>
Applications may actually care if the mapping operation failed, so when
it happens, return an error indication. errno is probably trashed by
fprintf though.
Signed-off-by: Keith Packard <keithp@keithp.com>
The execbuffer ioctl returns ENOMEM when it fails to pin all of the buffers
in the GTT. This is usually caused by the DRM client attempting to use too
much memory in a single request. Dumping out the requested and available
memory values should help point out failures in the DRM code to catch over
commitments of this form.
Signed-off-by: Keith Packard <keithp@keithp.com>
Add mode setting files to libdrm, including xf86drmMode.* and the new
drm_mode.h header. Also add a couple of tests to sanity check the
kernel interfaces and update code to support them.
I either messed up when I pulled these from a mmio-trace last time, or
the previous values didn't work on my card. Hopefully it's the former!
In any case, at least one of the original NV50 chipsets work now.
A small subset of CS need to be atomic (relocation+IB commit to
ring) right now, because of the way relocation are handled, we
need to protect the whole ioctl.
BO are referenced once by reloc to make sure that they not destroyed
before we get a chance to flush the cmd stream, so we need to unreference
them once in cs submit or cs erase if cs i never submitted so bo can
be destructed.
Relocation now consist of the following informations (in this order) :
handle buffer object handle identifier
start_offset start offset of first data of the buffer object used by the cs
end_offset end offset of last data of the buffer object used by the cs
read_domain read domain (either VRAM, or GTT as GPU is invalid for CS)
write_domain write domain (either VRAM, or GTT as GPU is invalid for CS)
flags flags used for further optimization (like discard previous
buffer content or forget buffer content after cs which can
help in avoiding moving content in or out)
The set_page_locked() function has changed its name again.
2.6.28 offers __set_page_locked() instead, which uses non-atomic
__set_bit() to do the work. In this case, offer our own
set_page_locked() using the atomic set_bit(), because I do not know
if atomic access is really necessary. Atomic behaviour is the one
previously expected.
Signed-off-by: Pekka Paalanen <pq@iki.fi>
I wanted to avoid doing this, as it's a bunch of churn, but there was a
conflict between the dri_ symbols in libdrm and the symbols that were in
Mesa in 7.2, which broke Mesa 7.2 AIGLX when the 2D driver had loaded new
libdrm symbols. The new naming was recommended by cworth for giving the
code a unique prefix identifying where the code lives.
Additionally, take the opportunity to fix up two API mistakes: emit_reloc's
arguments were in a nonsensical order, and set_tiling lacked the stride
argument that the kernel will want to use soon. API compatibility with
released code is maintained using #defines.
This relies on a new kernel ioctl to get the available aperture size.
In order to provide reasonable performance from dri_bufmgr_check_aperture, we
now require that once a buffer has been used as the target of a relocation,
it gets no further relocations added to it. This cuts the cost of
check_aperture from 10% to 1% in the 3D driver with no code changes, but
slightly complicates our plans for the 2D driver.
This turns the various nvXX_graph_init_ctxvals() methods into tables,
and speeds up compliation of nv50_graph.c quite a bit. This has bothered
me for a while, but others are complaining now so it's time to fix it :)
The drm_compat.c version of kmap_atomic_prot_pfn() uses the macro
pgd_offset_k(), which references the symbol init_mm.
Starting in 2.6.25, init_mm is no longer exported by default.
The only user of kmap_atomic_prot_pfn() is i915, so this should
not hurt anyone, and it allows people to load drm.ko.
Signed-off-by: Pekka Paalanen <pq@iki.fi>
Olaf Kirch noticed that the i915_set_status_page() function of the i915
kernel driver calls ioremap with an address offset that is supplied by
userspace via ioctl. The function zeroes the mapped memory via memset
and tells the hardware about the address. Turns out that access to that
ioctl is not restricted to root so users could probably exploit that to
do nasty things. We haven't tried to write actual exploit code though.
It only affects the Intel G33 series and newer.
Don't count on ioctl returning -errno; use errno directly.
Signed-off-by: Keith Packard <keithp@keithp.com>
Signed-off-by: Eric Anholt <eric@anholt.net>
We want to be able to use the bufmgr from multiple threads for GL, and thus
we need to protect the internal structures.
The pthread-stubs package is used so that programs not linked against
pthreads get weak symbols to stubs and don't eat most of the cost.
Allocate memory from different pools. This allows the OS to track memory
allocations for us, much like the linux memory debugging. This will ease
tracking down memory leaks since the OS can track the number of allocations
from each pool and help to point us in the right direction. Also replace
drm_alloc and friends with static __inline__ versions while we are here.
When using bufmgr_fake without DRM, the X server idles the ring whenever it
wants to wait for something to complete (brutal, but effective). In this
case, bufmgr_fake must treat the pending fence as having passed. However, it
wasn't recording the fences as it emitted them, nor cleaning buffers as they
passed.
Signed-off-by: Keith Packard <keithp@keithp.com>
We want to be able to use the bufmgr from multiple threads for GL, and thus
we need to protect the internal structures.
The pthread-stubs package is used so that programs not linked against
pthreads get weak symbols to stubs and don't eat most of the cost.
We don't explicitly check for error here and M_WAITOK will just put the
process to sleep waiting on resources to become available.
Suggested by John Baldwin
d_mmap gets called twice and we are only able to associate the file_priv
during the first call. The second call will return EBADF and we need to
assume that the call was succesful. d_mmap will not tolerate having an
error returned for the second call.
We need a way of getting at the underlying handle for use with mode
setting. We can either export it in the dri_bo object or provide a new
callback to get it.
I'd swapped the operands, so if we weren't in lockstep with the hardware we
said the sequence was always passed. Additionally, a race was available that
we might have failed at recovering from. Instead, I've replaced the logic
with new stuff that should be more robust and not rely on all the parties in
userland following the same IRQ_EMIT() == 1 protocol. Also, in a radical
departure from past efforts, include a long comment describing the failure
modes and how we're working around them.
Thanks to haihao for catching the original issue.
dri_bufmgr.h is replaced by intel_bufmgr.h, and several functions are renamed,
though the structures and many functions remain dri_bufmgr_* and dri_bo_*
The driver is in control of the show, so when you try and unload a module
the driver detach routine is called first. It is what drives the whole
unload process and so lots of panics occur if dev->driver is already
free.
Airlied inadvertently discovered that the IGP gart needs to be un-cached
for radeon rs485 and rs690 to work. Initial tests by placing a wbinvd()
after allocating the gart were successful. This is an attempt at a more
appropriate method of achieving success.
When the mapping is clean this flag will be set. This can be used
by a driver to save migrating and allocating pages for an object
that will first be used in VRAM.
We shuffled all the links around to disconnect the entry, but
never free it. We would incorrectly free the last entry in the
hash chain if nothing matched.
sarea_priv needs to be NULL before i915_initialized is called to
properly reset it. The stale value produces a panic any time something
opens/closes drm without calling initialize. i.e. version checking
This patch allows you to --enable-udev, and will avoid having libdrm
make device nodes. If you are using udev, you should really --enable-udev
your libdrm.
This resolves and issue on amd64 FreeBSD and it looks like the
linux ioctl syscall should be unsigned long as well.
Signed-off-by: Robert Noland <rnoland@2hip.net>
The user visible ioctl does this, but since we call into GEM internals
directly, we have to flush things ourselves. Fixes initial fb console
corruption.
When i915_gem_retire_request has a flush which matches an object write
domain, clear the write domain. This will move the object to the inactive
list rather than the flushing list, avoiding trouble with objects left stuck
on the flushing list.
In i915_gem_object_wait_rendering, if the object write domain is being
written by the GPU, the appropriate flushing commands are written to the
device and an additional request queued to mark that flush. Finally, the
function blocks on that new request.
The bug was that the write_domain in the object was cleared before the
function blocked.
If the wait is interrupted by a signal, the flushing commands may still be
pending. With the current write_domain information lost, the restarted
syscall will drop right through the write_domain test as that value was
lost, and so the function will not block at all. Oops.
Fixed by simply moving the write_domain clear until after the wait_request
succeeds. Note that the restarted system call will generate an additional
flush sequence and request, but that should be 'harmless', aside from a
slight performance impact.
Someday we'll track flushing more accurately and clear write_domains more
efficiently, but for now, this should suffice.
This bug was discovered in the 2d gem development by running x11perf
-copypixwin500 and noticing that the window got cleared accidentally.
This makes our handling of cliprects sane. drm_clip_rect always has exclusive
bottom-right corners, but the hardware expects inclusive bottom-right corners,
so we adjust this here.
This complements Michel Daenzer's commit 57aea290e1e0a26d1e74df6cff777eb9f038f1f8
to Mesa. See also http://bugs.freedesktop.org/show_bug.cgi?id=16123 .
Main fix is an oops that was triggered by the gtt pwrite path when we don't
have the gtt initialized. Also, settle on -EBADF for "bad object handle",
and -EINVAL for "reading/writing beyond object boundary".
This is around 3x or so speedup, since we would read wide rows at a time, and
clflush each tile 8 times as a result. We'll want code related to this anyway
when we do fault-based per-page clflushing for sw fallbacks.
DRAW_INDEX writes a vertex count to VAP_VF_CNTL. Docs say that behaviour
is undefined (i.e. lockups happen) when this write is not followed by the
right number of vertex indices.
Thus we used to do the wrong thing when drawing across many cliprects was
necessary, because we emitted a sequence
DRAW_INDEX, DRAW_INDEX, INDX_BUFFER, INDX_BUFFER
instead of
DRAW_INDEX, INDX_BUFFER, DRAW_INDEX, INDX_BUFFER
The latter is what we're doing now and which ought to be correct.
This is an initial import of the atom bios parser with modesetting support
for r500 hw using atombios. It also includes a simple memory manager
layer that translates a radeon GEM style interface onto TTM internally.
So far this memory manager has only been used for pinned object allocation
for the DDX to test modesetting.
this lets us debug the X server through xkb startup.
Not sure what the correct answer is, probably X needs to drop
the lock when execing stuff, with input hotplug it can get
xkb stuff at any time I believe.
this lets us debug the X server through xkb startup.
Not sure what the correct answer is, probably X needs to drop
the lock when execing stuff, with input hotplug it can get
xkb stuff at any time I believe.
This resolves a panic on FreeBSD which was caused by trying
to re-initialize the swap lock. It's just much easier to
initialize all of the locks at load time. It should also
ensure that the vblank structures are available earlier.
This increases overhead for the large-readpixels case due to the repeated
page cache accessing, but greatly reduces overhead for the small-readpixels
case.
Thanks to the reworked vblank-rework, we can just use the hardware frame
counter directly, and make the RADEON_PARAM_VBLANK_CRTC getparam just return
what was set by the corresponding setparam.
Caused drm_update_vblank_count() not to do its thing when called from
drm_modeset_ctl() -> drm_vblank_get().
The vblank functionality no longer needs to be suspended during a modeset, so
rename the field to vblank_inmodeset.
In my last push I forgot to convert users of drm_update_vblank_count
over to drm_vblank_get/put, since that's where any interrupt off->on
update accounting is done now. Since the modeset ioctl did something
similar (an open coded update of the counter) convert it over to using
get/put too, which saves us from having to deal with every combination
of interrupt off & on between calls.
We depend on the VM fully now for memory protection, separate DMA objects
for VRAM and GART are unneccesary. However, until the next interface break
(soon) a client can't depend on the objects being the same and must still
call NV_OBJ_SET_DMA_* methods appropriately.
Remove the unused (and broken) "in vblank" code now that the core has
been fixed to use a counter while interrupts are enabled. Also make the
vblank pipe get/set ioctls into dumb stub functions, since with the new
code we can no longer let userspace control whether vblank interrupts
are enabled, or the core code will misbehave.
The current code uses the hw vblank counter exclusively, which can lead
to wakeups during the active period rather than during the vblank period
if the hw counter counts displayed frames rather than vblank periods.
This change coverts the code over to using the counter while interrupts
are enabled, fixing that issue. It also includes a couple of related
changes: one to not enable the new enable/disable behavior until the
modeset ioctl is called (to preserve old client behavior) and another to
account for lost events due to mode setting with the new counter scheme.
BSD will require similar changes to its drm_irq.c code, but they should
be straightforward.
The current code can sleep in an interrupt handler, that is bad. So
instead if we can't grab the lock, flag it and run the tasklet on
unlock.
Signed-off-by: Robert Noland <rnoland@2hip.net>
modifications to make it work correctly on my test hardware (altered the
backlight write function, made it enable the legacy backlight controller
interrupts on mobile hardware, sorted the interrupt function so we don't
get an excessive number of vblank interrupts). This lets the backlight
keys on my T61 work properly, though there's a 750msec or so delay
between the request and the brightness actually changing - this sounds
awfully like the hardware spinning waiting for a status flag to become
ready, but as far as I can tell they're all set correctly. If anyone can
figure out what's wrong here, it'd be nice to know.
Some of the functions are still stubs and just tell the hardware that
the request was successful. These can be filled in as kernel modesetting
gets integrated. I think it's worth getting this in anyway, since it's
required for backlight control to work properly on some new platforms.
Signed-off-by: Matthew Garrett <mjg@redhat.com>
One of our systems has been returning 0xffffffff from all MCHBAR reads, which
means we'll need to figure out why, or add an alternate detection method.
Various chips have exciting interactions between the CPU and the GPU's
different ways of accessing interleaved memory, so we need some kernel
assistance in determining how it works.
Only fully tested on GM965 so far.
- There is one fb, used for as many outputs as possible.
- Eventually smaller screens will be scaled to see the full console, but for the moment this'll do.
When scheduled swaps occur, we need to blit between front & back buffers. I
the buffers are tiled, we need to set the appropriate XY_SRC_COPY tile bit,
only on 965 chips, since it will cause corruption on pre-965 (e.g. 945).
Bug reported by and fix tested by Tomas Janousek <tomi@nomi.cz>.
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
When scheduled swaps occur, we need to blit between front & back buffers. If
the buffers are tiled, we need to set the appropriate XY_SRC_COPY tile bit, but
only on 965 chips, since it will cause corruption on pre-965 (e.g. 945).
Bug reported by and fix tested by Tomas Janousek <tomi@nomi.cz>.
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
On 9xx chips, bus mastering needs to be enabled at resume time for much of the
chip to function. With this patch, vblank interrupts will work as expected
on resume, along with other chip functions. Fixes kernel bugzilla #10844.
Signed-off-by: Jie Luo <clotho67@gmail.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
We won't get a PFIFO context switch when the same channel ID is recreated if
the hw still thinks the channel is already active, which causes fun issues.
Should allow X to be stopped and started without tearing down the entire
card state in lastclose().
This avoids seeing garbage from engine setup etc before X gets around
to pointing the CRTCs at a new scanout buffer. Not actually a noticable
problem before G80 as PRAMIN is forced to the end of VRAM by the hardware
already.
We won't get a PFIFO context switch when the same channel ID is recreated if
the hw still thinks the channel is already active, which causes fun issues.
Should allow X to be stopped and started without tearing down the entire
card state in lastclose().
With the interrupt enable/disable using only the mask register, it was wrong
to use the enable register to detect which pipes had vblank detection
turned on. Also, as we keep a local copy of the mask register around, and
MSI machines smack the hardware during the interrupt handler, it is more
efficient and more correct to use the local copy.
It would be nice if one day the DRM driver was the canonical source for
register definitions and core macros. To that end, this patch cleans things up
quite a bit, removing redundant definitions (some with different names
referring to the same register) and generally tidying up the header file.
Clean up queues, free objects. On the next entervt, unmark the hardware to
let the user try again (presumably after resetting the chip). Someday we'll
automatically recover...
Clean up queues, free objects. On the next entervt, unmark the hardware to
let the user try again (presumably after resetting the chip). Someday we'll
automatically recover...
We depend on the VM fully now for memory protection, separate DMA objects
for VRAM and GART are unneccesary. However, until the next interface break
(soon) a client can't depend on the objects being the same and must still
call NV_OBJ_SET_DMA_* methods appropriately.
This avoids seeing garbage from engine setup etc before X gets around
to pointing the CRTCs at a new scanout buffer. Not actually a noticable
problem before G80 as PRAMIN is forced to the end of VRAM by the hardware
already.
Noting that the interrupt mask register was more reliable than the interrupt
enable register for managing interrupts in user_irq_on/user_irq_off, this
patch replaces the remaining IER frobbing with IMR instead.
The test which exposes IER related failures is:
$ glxgears & glxgears & glxgears
(reposition the glxgears windows away from the upper left corner)
$ while :; do x11perf -rect100 -reps 800 -repeat 1; sleep 1; done &
$ while :; do runoa; runet; done &
While waiting for the hardware to idle on leavevt or lastclose, poll
for the sync sequence number instead of waiting for an interrupt. This
allows the code to bail if the hardware hangs for some reason. Also, this
avoids issues with signals as the exisiting wait function is interruptible.
find_or_create_page doesn't quite set up pages correctly; any newly created
pages aren't hooked into the shmem object quite right; user space mmaps of
those pages end up mapping pages full of zeros which then get written to the
real pages inappropriately. This patch requires that the kernel export
shmem_getpage.
When a software fallback has completed, usermode must notify the kernel so
that any scanout buffers can be synchronized. This ioctl should be called
whenever a fallback completes to flush CPU and chipset caches.
Lots of conflicts, seems to load ok, but I'm sure some bugs snuck in.
Conflicts:
linux-core/drmP.h
linux-core/drm_lock.c
linux-core/i915_gem.c
shared-core/drm.h
shared-core/i915_dma.c
shared-core/i915_drv.h
shared-core/i915_irq.c
Normally when X is running, panic messages will be invisible and the machine
will just appear to hard hang. This patch adds support for switching back to
the fbcon framebuffer on panic (through the use of a panic notifier
registration) so we can see what happened.
Note that in order to be really useful, X will have to run its VT in something
other than KD_GRAPHICS mode. Also, not all kernel errors result in panics,
some go through BUG() which may trigger another type of event, not resulting in
a switch.
This fixes registration when MSI is set up after the stub function fills in
dev->irq. Otherwise /proc/interrupts would report attachment to the fasteoi
interrupt. dev->irq is still exposed (and updated at IRQ setup)
for the drivers that use it for whatever reason.
In leavevt_ioctl, queue an MI_FLUSH and then block waiting for it to
complete. This will empty the active and flushing lists. That leaves only
the inactive list to evict.
Pin/unpin need to know whether to remove/add objects from the inactive list,
inactive objects cannot be in any GPU write domain as those would be on the
flushing list instead. However, inactive objects may be in the CPU write
domain.
Now that gem_object_unbind waits for rendering to complete, objects should
not be active when they are being pulled from the GTT. BUG_ON if this is
broken.
Record the last execbuffer sequence for each client.
Record that sequence in the throttle ioctl as the 'throttle sequence'.
Wait for the last throttle sequence in the throttle ioctl.
When i915_wait_request clears object from the active list, it may end up
freeing them and not moving them to the inactive list. This ends up
unbinding objects from the GTT without there ever being new objects visible
to i915_gem_evict_something on the inactive list. As the only success
condition required the presence of objects on the inactive list, this would
falsely assume that no GTT space had been made available, and end up
returning -ENOMEM to the application.
The interrupt identity register must be writen before any work occurs lest
we drop an interrupt on the floor. This patch just shuffles code around to
make sure that IIR is written as early as possible.
We want request retirement to occur about once a second when the request
queue is non-empty. This was done with a timer that queued a work_struct,
using a delayed_work instead makes a lot more sense.
In the short-circuit code for the breadcrumb already being new enough, we
need to update the sarea_priv copy of the breadcrumb just as if we had
waited. Otherwise userland error checking will notice that we returned
too early based on its wrong information, and call wait_irq again (leading
to spinning until someone else comes along and updates the sarea_priv).
This bug was hidden when we had interrupt masking disabled, such as in
master, since the interrupt handler would update sarea_priv.
Thanks to Thomas Hellstrom for catching the issue, no thanks to the kernel
developer who authoritatively told me that they would get restarted on their
own.
This was insufficient once we started masking interrupts to only when someone
was waiting for them (and would thus retire requests themselves). It was
replaced by the retire_timer.
This patch should fixe hard lockup and convert them in
softlockup (ie you can ssh the box but the gpu is busted
and we are waiting in loop for it to come back to reason).
We need a version to depend on from the other components that require GEM and
the bufmgr code. Some interfaces will be removed before the 2.4.0 release.
Use GEM for ring buffer setup and framebuffer allocation. This means reworking
the hardware status page stuff a bit (just use the basic range allocator for
vram for now) and #ifdef'ing out the TTM & DRI2 code. Works well enough to
load/unload several times and display fbcon on my T61 (though there's still
some unexplained console corruption).
This is the create (may want location flags), pread/pwrite/mmap
(performance tuning hints), and set_domain (will 32 bits be enough for
everyone?) ioctls. Left in the generic set are just flink/open/close.
The 2D driver must be updated for this change, and API but not ABI is broken
for 3D. The driver version is bumped to mark this.
Use new GEM based ring buffer initialization. Still need to init GEM & use it
for framebuffer allocation etc.
Conflicts:
shared-core/i915_dma.c
shared-core/i915_drv.h
This requires that the X Server use the execbuf interface for buffer
submission, as it no longer has direct access to the ring. This is
therefore a flag day for the gem interface.
This also adds enter/leavevt ioctls for use by the X Server. These would
get stubbed out in a modesetting implementation, but are required while
in an environment where the device's state is only managed by the DRM while
X has the VT.
Port over EDID quirks from X.Org so we can handle more monitors. This meant
adding size info to the drm_display_mode struct, but other than that the
changes were isolated to the DRM EDID handling code (as they should be).
This seems to be the key to getting at least some radeon
cards working. Most, if not all drivers need it enabled,
so just request it once the driver has attached.
Without the user IRQ running constantly, there's no wakeup when the ring
empties to go retire requests and free buffers. Use a 1 second timer to make
that happen more often.
Instead of throttling and execbuffer time, have the application ask to
throttle explicitly. This allows the throttle to happen less often, and
without holding the DRM lock.
The code was discarding the dri_bo_gem structure and saving only the kernel
handle. This lost the mmap address, causing pain when the next buffer user
wanted to map the buffer.
A check in drm_sysfs_connector_remove was supposed to allow it to be called
even with unregistered objects, to make cleanup paths a little simpler.
However, device_is_regsitered didn't always seem to return what we thought it
would, so we'd sometimes end up leaving objects lying around rather than
unregistering them.
Fix this situation up by requiring devices to be registered before being
removed. Any problems resulting from this change should be easier to track
down than the alternative (which is leaving kobjects registered after unload).
We need to initialize the edid_blob_ptr to NULL when we init a connector,
otherwise drm_mode_connector_update_edid_property may think there's a valid
EDID lying around and try to destroy it, causing a crash.
Idea being if you want to add new crtc/output/encoder dynamically later,
you just increase the generation counter and userspace should re-read
all the resources
Without kernel modesetting, this requires cooperation of the userspace
modesetting driver. We may have to leave the vblank interrupt enabled otherwise
to avoid problems.
Only compensate when the driver counter actually appears to have moved
backwards.
The compensation deltas need to be incremental instead of absolute; drop the
vblank_offset field and just use atomic_sub().
Turns out the radeon driver is affected by the same problem that prompted i915
to revert to less useful counter flipping at the end of the vblank interval. In
the long term, we can hopefully implement more reliable methods to achieve
counter flipping at the beginning of vblank, but otherwise this should be an
acceptable workaround.
This reverts commit 6671ad1917.
The vblank ioctl needs to update the userspace parameters when interrupted by
a signal, which was prevented by this. Let's see if this breaks other ioctls...
set_domain can block waiting for rendering to complete. If that process is
interrupted by a signal, it can return -EINTR. Catch this error in all
callers and correctly deal with the result.
This creates a default group attached to the legacy drm minor nodes.
It covers all the objects in the set. make set resources only return
objects for this set. Need to fix up other functions to only work on
objects in their allowed set.
Okay we have crtc, encoder and connectors.
No more outputs exposed beyond driver internals
I've broken intel tv connector stuff.
Really for TV we should have one TV connector, with a sub property for the
type of signal been driven over it
Use subclassing from the drivers to allocate the objects. This saves
two objects being allocated for each crtc/output and generally makes
exit paths cleaner.
This splits a lot of the core modesetting code out into a file of
helper functions, that are only called from themselves and/or the driver.
The driver gets called into more often or can call these functions from itself
if it is a helper using driver.
I've broken framebuffer resize doing this but I didn't like the API for that
in any case.
Object domain transfer can involve adding flush ops to the request queue,
and so the DRM lock must be held to avoid having the X server smash pointers
badly.
The interrupt enable register cannot be used to temporarily disable
interrupts, instead use the interrupt mask register.
Note that this change means that a pile of buffers will be left stuck on the
chip as the final interrupts will not be recognized to come and drain things.
Add code to get panel modes from the VBIOS if present and check whether certain
outputs exist. Should make our display detection code a little more robust.
When reading from multiple domains, allow each cache to continue
to hold data until writes occur somewhere. This is done by
first leaving the read_domains alone at bind time (presumably the CPU read
cache contains valid data still) and then in set_domain, if no write_domain
is specified, the new read domains are simply merged into the existing read
domains.
A huge comment was added above set_domain to explain how things are
expected to work.
Newly allocated objects need to be in the CPU domain as they've just been
cleared by the CPU. Also, unmapping objects from the GTT needs to put them
into the CPU domain, both to flush rendering as well as to ensure that any
paging action gets flushed before we remap to the GTT.
Commands in the ring are parsed and started when the head pointer passes by
them, but they are not necessarily finished until a MI_FLUSH happens. This
patch inserts a flush after the execbuffer (the only place a flush wasn't
already happening).
Recording the tail pointer in a local variable improves performance, but if
someone messes up and fails to reload at the right time, the driver will
write commands to the wrong part of the ring and scramble execution badly.
This change (available by setting I915_RING_VALIDATE to 1) checks to make
sure the cached tail pointer matches the hardware tail pointer at each ring
buffer addition, calling BUG_ON when that's not true.
There are now 3 lists. Active is buffers currently in the ringbuffer.
Flushing is not in the ringbuffer, but needs a flush before unbinding.
Inactive is as before. This prevents object_free → unbind →
wait_rendering → object_reference and a kernel oops about weird refcounting.
This also avoids an synchronous extra flush and wait when freeing a buffer
which had a write_domain set (such as a temporary rendered to and then from
using the 2d engine). It will sit around on the flushing list until the
appropriate flush gets emitted, or we need the GTT space for another
operation.
The dummy read page will point to NULL if drm_bo_driver_init failed at
firstopen (modeset is not enabled), and will cause kernel oops at
subsequent drm_lastclose call, so be sure to check it.
This lets us get some qualities we desire, such as using the full 32-bit
range (except zero), avoiding DRM_WAIT_ON, and a 1:1 mapping of active
sequence numbers to request structs, which will be used soon for throttling
and interrupt-driven list cleanup.
Additionally, a boolean active field is added to indicate which list an
object is on, rather than smashing last_rendering_cookie to 0 to show
inactive. This will help with flush-reduction later on, and makes the code
clearer.
It would be nice if one day the DRM driver was the canonical source for
register definitions and core macros. To that end, this patch cleans
things up quite a bit, removing redundant definitions (some with
different names referring to the same register) and generally tidying up
the header file.
In order to avoid recursive ->detect->interrupt->detect->interrupt->...
we need to disable TV hotplug interrupts in
intel_tv.c:intel_tv_detect_type. We also need to enable the TV interrupt
detection and hotplug sequence properly in i915_irq.c.
No need to fill the ring that much; wait for it to become nearly empty
before adding the execbuffer request. A better fix will involve scheduling
ring insertion in the irq handler.
drm_crtc->fb may point to NULL, f.e X server will allocate a new fb
and assign it to the CRTC at startup, when X server exits, it will destroy
the allocated fb, making drm_crtc->fb points to NULL.
pread and pwrite must update the memory domains to ensure consistency with
the GPU. At some point, it should be possible to avoid clflush through this
path, but that isn't working for me.
The exec list contains all objects, in order of use. The lru list contains
only unpinned objects ready to be evicted. This required two changes -- the
first was to not migrate pinned objects from exec to lru, the second was to
search for the first unpinned object in the exec list when doing eviction.
Now, the LRU list has objects that are completely done rendering and ready
to kick out, while the execution list has things with active rendering,
which have associated cookies and reference counts on them.
Even if the TV encoder hasn't been fused off, we may not have a TV connector on
the platform. The BDB in the BIOS should give us this info in some cases.
Domain information is about buffer relationships, not buffer contents. That
means a relocation contains the domain information as it knows how the
source buffer references the target buffer.
This also adds the set_domain ioctl so that user space can move buffers to
the cpu domain.
This should already have been generally safe since we don't change contents
and put in new relocations between execbufs, so if we were writing in a new
relocation then we'd already waited rendering to complete when we moved
the target of the relocation. However, doing the right thing will be required
if we do buffer reuse.
The kernel has removed nopage so move the old nopage codepaths into a compat vm file and switch to using the fault paths.
nopfn is on its way out in the future also, so we should switch to using fault
for that path as well soon
If objects on the lru aren't ref counted, they'll get pulled from the gtt as
soon as they are freed. This change does cause objects to get stuck in the
gtt until they're forced out by new requests. The lru should get cleaned
when the irq occurs.
When batch buffers are executing, the ring may be stuck for a long time.
Monitor the ACTHD pointer which will show if the execution engine is
actually hung.
pages come back from find_or_create_page locked, but must not stay locked
for long. Unlock them immediately instead of waiting until we're done with
them to avoid deadlock when applications try to touch them.
Track named objects in /proc/dri/0/gem_names.
Track total object count in /proc/dri/0/gem_objects.
Initialize device gem data.
return -ENODEV for gem ioctls if the driver doesn't support gem.
Call unlock_page when unbinding from gtt.
Add numerous misssing calls to drm_gem_object_unreference.
Names are just another unique integer set (from another idr object).
Names are removed when the user refernces (handles) are all destroyed --
this required that handles for objects be counted separately from
internal kernel references (so that we can tell when the handles are all
gone).
mixed 32/64 bit systems need 'special' help for ioctl where the user-space
and kernel-space datatypes differ. Fixing the datatypes to be the same size,
and align the same way for both 32 and 64-bit ppc and x86 environments will
elimiante the need to have magic 32/64-bit ioctl translation code.
This is possibly temporary. I can trigger an unending IRQ storm on G8x
in some circumstances, and have no idea how to handle that particular PFIFO
exception correctly yet.
I swore I'd actually do this properly and not go the horrible route
we did with nv4x, but I won't get around to it just yet with so many
*actually* interesting things to do first.. One day.
Since someone already added nv86, why not!
It's not really a graphics memory allocator, just something to track ranges
of address space. It doesn't involve actual allocation, and was consuming
some desired namespace.
Enum can be of pretty much any size since C leaves the choice of size up to the implementation. So avoid using it in new interfaces like the vblank pre- & post-modeset ioctl. Thanks to hch for spotting this.
On my 865G machine, it seems the CPU will receive interrupt before
irq_postinstall is called. This will cause kernel oops because vblank is not
inited at that time. Clear interrupt status before install seems fixing this
problem.
Signed-off-by: Hong Liu <hong.liu@intel.com>
This tries to automatically fetch a git revision string and if succeeds,
it #defines GIT_REVISION string macro. Packagers can override it by
'make GIT_REVISION=foo'.
Update Nouveau to use GIT_REVISION, if defined, instead of DRIVER_DATE
in struct drm_driver.
Signed-off-by: Pekka Paalanen <pq@iki.fi>
The batchbuffer submission paths were fixed to use the 965-specific command,
but the vblank tasklet was not. When the older version is sent, the 965 will
lock up.
TV out needs to do load detection, which means we have to find an
available pipe to use for the detection. Port over the pipe reservation
code for this purpose.
Put off registering new outputs with sysfs until they're properly configured,
or we may get duplicates if the type hasn't been set yet (as is the case with
SDVO initialization). This also means moving de-registration into the cleanup
function instead of output destroy, since the latter occurs during the normal
course of setup when an output isn't found (and therefore not registered with
sysfs yet.
If the driver is 'modeset' enabled, it'll register it's interrupt
handler at load time. Set the devname in this case so that
/proc/interrupts makes sense.
This patch ties outputs, output properties and hotplug events into the
DRM core. Each output has a corresponding directory under the primary
DRM device (usually card0) containing dpms, edid, modes, and connection
status files.
New hotplug change events occur when outputs are added or hotplug events
are detected.
This interface was defined completely wrong, however userspace has only
ever used 4 values from it (0x1, 0x2, 0x3 and 0x6), so fix the interface to do what userspace actually expected but define new defines for new users to use
it properly.
Previously, the R300_CMD_WAIT command would write the passed directly to the
hardware. However this is incorrect because the R300_WAIT_* values used are
internal interface values that do not map directly to the hardware.
The new function I have added translates the R300_WAIT_* values into appropriate
values for the hardware before writing the register.
Thanks to John Bridgman for pointing this out. :-)
More or less a workaround for issues on some chipsets where a context
switch results in critical data in PRAMIN being overwritten by the GPU.
The correct fix is known, but may take some time before it's a feasible
option.
This is the correct fix for the RS690 and hopefully the dma coherent work.
For now we limit everybody to a 32-bit DMA mask but it is possible for
RS690 to use a 40-bit DMA mask for the GART table itself,
and the PCIE cards can use 40-bits for the table entries.
Signed-off-by: Dave Airlie <airlied@redhat.com>
DMA command submission. It's worth remembering that all new bright ideas on how
to make this command reader work properly and according to docs
will probably fail :( Bring in some old code.
The i915_vblank_swap() function schedules an automatic buffer swap
upon receipt of the vertical sync interrupt. Such an operation is
lengthy so it can't be allowed to happen in normal interrupt context,
thus the DRM implements this by scheduling the work in a kernel
softirq-scheduled tasklet. In order for the buffer swap to work
safely, the DRM's central lock must be taken, via a call to
drm_lock_take() located in drivers/char/drm/drm_irq.c within the
function drm_locked_tasklet_func(). The lock-taking logic uses a
non-interrupt-blocking spinlock to implement the manipulations needed
to take the lock. This semantic would be safe if all attempts to use
the spinlock only happen from process context. However this buffer
swap happens from softirq context which is really a form of interrupt
context. Thus we have an unsafe situation, in that
drm_locked_tasklet_func() can block on a spinlock already taken by a
thread in process context which will never get scheduled again because
of the blocked softirq tasklet. This wedges the kernel hard.
To trigger this bug, run a dual-head cloned mode configuration which
uses the i915 drm, then execute an opengl application which
synchronizes buffer swaps against the vertical sync interrupt. In my
testing, a lockup always results after running anywhere from 5 minutes
to an hour and a half. I believe dual-head is needed to really
trigger the problem because then the vertical sync interrupt handling
is no longer predictable (due to being interrupt-sourced from two
different heads running at different speeds). This raises the
probability of the tasklet trying to run while the userspace DRI is
doing things to the GPU (and manipulating the DRM lock).
The fix is to change the relevant spinlock semantics to be the
interrupt-blocking form. After this change I am no longer able to
trigger the lockup; the longest test run so far was 20 hours (test
stopped after that point).
Note: I have examined the places where this spinlock is being
employed; all are reasonably short bounded sequences and should be
suitable for interrupts being blocked without impacting overall kernel
interrupt response latency.
Signed-off-by: Mike Isely <isely@pobox.com>
If we ever want to be able to use the 3D engine we have no choice. It
appears that the tiling setup (required for 3D on G8x) is in the page tables.
The immediate benefit of this change however is that it's now not possible
for a client to use the GPU to render over the top of important engine setup
tables, which also live in VRAM.
G8x VRAM size is limited to 512MiB at the moment, as we use a 1-1 mapping
of real vram pages to their offset within the start of a channel's VRAM
DMA object and only populate a single PDE for VRAM use.
Conflicts:
linux-core/drm_compat.c
linux-core/drm_compat.h
linux-core/drm_ttm.c
shared-core/i915_dma.c
Bump driver minor to 13 due to introduction of new
relocation type.
My 965GM gets interrupts stuck when using the old PIPE_VBLANK interrupt.
Switch to the PIPE_EVENT interrupt mechanism, and set the PIPE*STAT
registers to use START_VBLANK on 965 and VBLANK on previous chips.
Will hopefully work a bit better than previous code, which depended on
knowing the channel's most recent PUT value. Some chips always return
0 on reading these regs, and currently userspace is the only other entity
which knows the value.
This patch fixes bits of the DRM so to make the radeon DRI work on
non-cache coherent PCI DMA variants of the PowerPC processors.
It moves the few places that needs change to wrappers to that
other architectures with similar issues can easily add their
own changes to those wrappers, at least until we have more useful
generic kernel API.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
PCI- or high memory.
This is substantially more efficient than drm_bo_kmap,
since the mapping only lives on a single processor.
Unmapping is done use kunmap_atomic(). Flushes only a single tlb() entry.
Add a support utility int drm_bo_pfn_prot() that returns the
pfn and desired page protection for a given bo offset.
This is all intended for relocations in bound TTMS or vram.
Mapping-accessing-unmapping must be atomic, either using preempt_xx() macros
or a spinlock.
PCI- or high memory.
This is substantially more efficient than drm_bo_kmap,
since the mapping only lives on a single processor.
Unmapping is done use kunmap_atomic(). Flushes only a single tlb() entry.
Add a support utility int drm_bo_pfn_prot() that returns the
pfn and desired page protection for a given bo offset.
This is all intended for relocations in bound TTMS or vram.
Mapping-accessing-unmapping must be atomic, either using preempt_xx() macros
or a spinlock.
This change adds a driver feature that for i915 is controlled by a module
parameter. You now need to do insmod i915.ko modeset=1 to enable it the
modesetting paths.
It also fixes up lots of X paths. I can run my new DDX driver on this code
with and without modesetting enabled
Allow mode to be set with fb_id set to -1, meaning set
the mode with the current fb (if we have one bound).
Allow intelfb to hook back up it's fb if modesetting
clears it (maybe temporary).
Move any crtc->fb related register changes to set_base
in intel_fb.
General intelfb cleanups.
Texture uploads could hit the blitter coordinate limit, adjust the texture
offset when uploading the pieces. Make sure to check the end address of the
upload too.
This fixes at least two problems:
* The vblank_disable_fn timer callback could get called after the DRM was
de-initialized, e.g. on X server shutdown.
* Leak of vblank related resources when disabling and re-enabling the IRQ, e.g.
on an X server reset.
On many chipsets, the checks for DPLL enable or VGA mode will prevent the
pipeconf regs from being restored, which could result in a blank display or X
failing to come back after resume. So restore them unconditionally along with
actually restoring pipe B's palette correctly.
Make sure we have enough room for all the GR registers or we'll end up
clobbering the AR index register (which should actually be harmless
unless the BIOS is making an assumption about it).
On resume, if the interrupt state isn't restored correctly, we may end
up with a flood of unexpected or ill-timed interrupts, which could cause
the kernel to disable the interrupt or vblank events to happen at the
wrong time. So save/restore them properly.
There were two problems with the existing callback code: the vblank
enable callback happened multiple times per disable, making drivers more
complex than they had to be, and there was a race between the final
decrement of the vblank usage counter and the next enable call, which
could have resulted in a put->schedule disable->get->enable->disable
sequence, which would be bad.
So add a new vblank_enabled array to track vblank enable on per-pipe
basis, and add a lock to protect it along with the refcount +
enable/disable calls to fix the race.
sequence number may actually turn up before the corresponding fence
object has been queued on the ring.
Fence drivers can use this member to determine whether a
sequence number must be re-reported.
In hibernate, we may end up calling the VGA save regs function twice, so we
need to make sure it's idempotent. That means leaving ARX in index mode after
the first save operation. Fixes hibernate on 965.
We need to return an accurate vblank count to the callers of
->get_vblank_counter, and in the Intel case the actual frame count
register isn't udpated until the next active line is displayed, so we
need to return one more than the frame count register if we're currently
in a vblank period.
However, none of the various ways of doing this is working yet, so
disable the logic for now. This may result in a few missed events, but
should fix the hangs some people have seen due to the current code
tripping the wraparound logic in drm_update_vblank_count.
The frame count registers don't increment until the start of the next
frame, so make sure we return an incremented count if called during the
actual vblank period.
Ack the IRQs correctly (PIPExSTAT first followed by IIR). Don't read
vblank counter registers on disabled pipes (might hang otherwise). And
deal with flipped pipe/plane mappings if present.
i915_flush_ttm was unconditionally executing a clflush instruction
to (obviously) flush the cache. Instead, check if the cpu supports
clflush, and if not, fall back to calling wbinvd to flush the entire
cache.
Signed-off-by: Kyle McMartin <kmcmartin@redhat.com>
(1 << bits) is an undefined value when bits == 32.
gcc may generate 1 with this expression
which will lead to an infinite retry loop in
drm_ht_just_insert_please.
Because of the different implement of hash_long,
this issue is more frequenly see on 64 bit system
This is necessary for AGP to work after running bios init scripts on nv3x, and
is seen in mmio traces of all cards (nv04-nv4x)
I'm not making the equivalent change to nv40_mc.c, as early cards (6200, 6800gt)
use the 0x000018XX PBUS and later cards use the 0x000880XX PBUS and I don't know
the effects of using the wrong one
As DRM_DEBUG macro already prints out the __FUNCTION__ string (see
drivers/char/drm/drmP.h), it is not worth doing this again. At some
other places the ending "\n" was added.
airlied:- I cleaned up a few that this patch missed also
I couldn't figure out what drm_bo_type_dc was for; Dave Airlie finally clued
me in that it was the 'normal' buffer objects with kernel allocated pages
that could be mmapped from the drm device file.
I thought that 'drm_bo_type_device' was a more descriptive name.
I also added a bunch of comments describing the use of the type enum values and
the functions that use them.
Flags pending validation were stored in a misleadingly named field, 'mask'.
As 'mask' is already used to indicate pieces of a flags field which are
changing, it seems better to use a name reflecting the actual purpose of
this field. I chose 'proposed_flags' as they may not actually end up in
'flags', and in an case will be modified when they are moved over.
This affects the API, but not ABI of the user-mode interface.
Previously, dummy_read_page was used only for read-only user allocations; it
filled in pages that were not present in the user address map (presumably,
these were allocated but never written to pages).
This patch allows them to be used for read-only ttms allocated from the
kernel, so that applications can over-allocate buffers without forcing every
page to be allocated.
I'm hoping to use the dummy_read_page for kernel allocated buffers to avoid
allocating extra pages for read-only buffers (like vertex and batch buffers).
This also eliminates the 'write' parameter to drm_ttm_set_user and just
has DRM_TTM_PAGE_WRITE passed into drm_ttm_create.
Aside from changing drm_bind_ttm to drm_ttm_bind, this patch
adds only documentation and fixes the functions inside drm_ttm.c
to all be prefixed with drm_ttm_.
drivers/char/drm/mga_dma.c::mga_do_cleanup_dma() and I think there's a small
problem.
The variable is only used inside #if __OS_HAS_AGP which is fine, but all
that
ever happens is an assignment to the variable - it is never actually used
for
anything. The variable is nicely initialized to zero which is also what the
return statement at the end of function returns (always at the moment).
It looks to me like that function should be returning 'err' instead of
always
just returning 0. Here's a patch to do that.
Signed-off-by: Jesper Juhl <jesper.juhl@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
drmBOSetStatus does not bother to set the fence_class parameter.
Fortunately, drm_bo_setstatus_ioctl doesn't end up using it as it
calls drm_bo_handle_validate with use_old_fence_class = 1.
Document parameters and usage for drm_bo_handle_validate. Change parameter
order to match drm_bo_do_validate (fence_class has been moved to after
flags, hint and mask values). Existing users of this function have been
changed, but out-of-tree users must be modified separately.
Add comments about the parameters to drm_bo_do_validate, along
with comments for the DRM_BO_HINT options. Remove the 'do_wait'
parameter as it is duplicated by DRM_BO_HINT_DONT_BLOCK.
Creating a ttm was done with drm_ttm_init while destruction was done with
drm_destroy_ttm. Renaming these to drm_ttm_create and drm_ttm_destroy makes
their use clearer. Passing page_flags to the create function will allow that
to know whether user or kernel pages are needed, with the goal of allowing
kernel ttms to be saved for later reuse.
One of the costs of superioctl has been the need to perform relocations
inside the kernel. The cost of mapping the buffers to the CPU and writing
data is fairly high, especially if those buffers have been mapped and read
by the GPU.
If we assume that buffers don't move around very often, we can have the
client compute the relocations itself using the previous GPU address. When
that object doesn't move, the kernel can skip computing and writing the
updated data.
Here's a patch which adds a new field to struct drm_bo_info_req called
'presumed_offset', and a new DRM_BO_HINT_PRESUMED_OFFSET that is set when
this field has been filled in by the client.
There are two separate optimizations performed when the presumed_offset is
correct:
1. i915_exec_reloc checks to see if all previous buffer offsets were guessed
correctly. If so, there's no need for it to look at *any* of the
relocations for a buffer. When this happens, it skips the whole
relocation process, simply returning success.
2. i915_apply_reloc checks to see if the target buffer offset was guessed
correctly. If so, it skips mapping the relocatee, computing the
relocation and writing the value. If no relocations are needed, the
relocatee should never be mapped to the CPU, and so the kernel shouldn't
need to wait for any fences to pass.
If drmMinor >= 6, the intel DDX driver will enable vblank events on both
pipes. If drmMinor >= 10 on pre-965 chipsets, the intel DDX driver will
swap the pipe<->plane mapping to allow for framebuffer compression on
laptop screens. This means the secondary vblank counter (corresponding
to pipe B) will be incremented when vblank interrupts occur.
Now Mesa waits for vblank events on whichever plane has a greater
portion of the displayed window. So it will happly ask to wait for the
primary counter even though that one won't increment.
So we can fix this in either the DDX driver, Mesa or the kernel (though
I thought we already had several times).
Since current (and previous) userspace assumes it's talking about a pipe
== plane situation and now uses planes when talking to the kernel, we
should probably just hide the mapping details there (indeed they already
are hidden there for vblank swaps), which this patch does.
So as far as userland is concerned, whether we call things planes or
pipes is irrelevant, as long as kernel developers understand that
userland hands them planes and they have to figure out which pipe that
corresponds to (which will typically be the same on 965+ hardware and
reversed on pre-965 mobile chips).
This fix is actually a bit of a cleanup too--it moves lock freeing to
drm_rmmap_locked and out of drm_lastclose. This makes it symmetrical with
addmap and also prevents the lock from being incorrectly freed from driver
mappings.
This should work on all radeon but there is still many things todo:
- add crtc2
- tmds
- lvds
- add bios data table so we don't need to hardcode dac/crtc infos
- separate clock control to make power saving easier & cleaner
- tiling (warning tiling shouldn't be enable in double scan or interlace)
- surface reg manager (this goes along with tiling)
- suspend/resume hook
- avivo & r500 family support
- atom bios support (for posting card mostly)
- finish superioctl skeleton
- what else ? :)
so really want to get a list of modes per output not the global hammer list.
also we remove the mode ids and let the user pass back the full mode description
need to fix up add/remove mode for user modes now
We needed to specifically check for driver support and test the correct
vbl_received value. Also pulled over support for _DRM_VBLANK_NEXTONMISS
from the linux code.
This allow the user to retrieve a list of properties for an output.
Properties can either be 32-bit values or an enum with an associated name.
Range properties are to be supported.
This API is probably not all correct, I may make properties part of the general
resource get when I think about it some more.
So basically you can create properties and attached them to whatever outputs you want,
so it should be possible to create some generics and just attach them to every output.
Since the drm_mode_card_res structure contains user pointers, we have to use
put_user and copy_to_user to write stuff out. The DRM ioctl wrapper will only
take care of copying the base drm_mode_card_res struct, not the included
arrays.
This header file is shared across linux and bsd, but is not installed
for user space to access. It's the place to put prototypes and data
types that aren't platform or chipset specific, but still internal to
the drm.
This way driver can get_edid in output status detection
(using all workaround which are in get_edid) and then provide
this edid data in get_mode callback of output.
This patch is originally from malc0_, but since it used some NV40_*
regs, I edited them into hex values with a comment.
This seems to correspond quite well with my own mmio-trace,
for the parts I cared to check.
This code relied on the CPU and GPU address for the aperture being the same,
On some r5xx hardware I was playing with I noticed that this isn't always true.
I wonder if this will fix some of those r4xx DRI issues we've seen in the past.
Conflicts:
linux-core/Makefile.kernel
linux-core/drm_stub.c
linux-core/i915_drv.c
shared-core/i915_dma.c
shared-core/i915_drv.h
Fixup suspend/resume conflicts (basically use what's in DRM master for now).
Also fix up a few other conflicts that snuck in (i915_dma changes etc.).
The vblank_init function wanted a couple of cleanups.
Also, drm_irq_install wasn't checking the new return value of irq_postinstall.
If it returns a failure, assume IRQs didn't get set up and take appropriate
action.
This mapping allows cached objects to be mapped in/out of the TT space
with the appropriate flushing calls.
It should put back the old CACHED functionality for snooped mappings
Conflicts:
linux-core/drmP.h
linux-core/drm_drv.c
linux-core/drm_irq.c
shared-core/i915_drv.h
shared-core/i915_irq.c
shared-core/mga_drv.h
shared-core/mga_irq.c
shared-core/radeon_drv.h
shared-core/radeon_irq.c
Merge in the latest master bits and update the remaining drivers (except
mach64 which math_b is working on). Also remove the 9xx hack from the i915
driver; it seems to be correct.
Add suspend/resume support to the i915 driver. Moves some of the
initialization into the driver load routine, and fixes up places where we
assumed no dev_private existed in some of the cleanup paths. This allows
us to suspend/resume properly even if X isn't running.
Make DRM devices use real Linux devices instead of class devices, which are
going away. While we're at it, clean up some of the interfaces to take
struct drm_device * or struct device * and use the global drm_class where
needed instead of passing it around.
Implement a version check IOCTL for drivers that don't use
drmMMInit from user-space.
Remove the minor check from the kernel code. That's really up
to the driver.
Bump major.
Previously, the lock would get released on the first close by the X Server
(during AIGLX setup), and the Radeon driver would then hang in initialization
due to unexpected failure in DRM calls that required the lock to be held.
Based on a patch by Kostik Belousov.
Remove need for lock for now.
May create races when we clean memory areas or on takedown.
Needs to be fixed.
Really do a validate on buffer creation in order to avoid problems with
fixed memory buffers.
We now always create a drm_ref_object for user objects and this is then the only
things that holds a reference to the user object. This way unreference on will
destroy the user object when the last drm_ref_object goes way.
The buffer object type is still tracked internally, but it is no longer
part of the user space visible ioctl interface. If the bo create ioctl
specifies a non-NULL buffer address we assume drm_bo_type_user,
otherwise drm_bo_type_dc. Kernel side allocations call
drm_buffer_object_create() directly and can still specify drm_bo_type_kernel.
Not 100% this makes sense either, but with this patch, the buffer type
is no longer exported and we can clean up the internals later on.
This adds the initial i915 superioctl interface. The interface should be
sufficent even if the implementation may needs fixes/optimisations internally
in the drm wrt caching etc.
This isn't 100% as command submission via PCI-e GART buffers doesn't work.
I've hacked around that for the time being. This is essentially the code
that was used at the POWER.org event to show Bimini.
All nv30 functions in nv30_graph.c that can be used on nv20 are renamed
as accordingly. nv20 specific parts from nv20_graph.c are moved into
nv30_graph.c.
I should not have renamed this field.
I should not have renamed this field.
I should not have renamed this field.
On the plus side, it was at least binary compatible.
Conflicts:
linux-core/drmP.h
linux-core/drm_bo.c
linux-core/drm_drv.c
linux-core/drm_objects.h
shared-core/drm.h
shared-core/i915_dma.c
shared-core/i915_drv.h
shared-core/i915_irq.c
Mostly removing typedefs that snuck into the modesetting code and
updating to the latest TTM APIs. As of today, the i915 driver builds,
but there are likely to be problems, so debugging and bugfixes will
come next.
Modify the TTM backend bind arguments.
Export a number of functions needed for driver-specific super-ioctls.
Add a function to map buffer objects from the kernel, regardless of where they're
currently placed.
A number of error fixes.
This branch replaces the NO_MOVE/NO_EVICT flags to buffer validation with a
separate privileged ioctl to pin buffers like NO_EVICT meant before. The
functionality that was supposed to be covered by NO_MOVE may be reintroduced
later, possibly in a different way, after the superioctl branch is merged.
Previously any ioctls that weren't explicitly listed in the compat ioctl
table would fail with ENOTTY. If the incoming ioctl number is outside the
range of the table, assume that it Just Works, and pass it off to drm_ioctl.
This make the fence related ioctls work on 64-bit PowerPC.
We can figure out which pipe a given plane is mapped to by looking at the
display control registers instead of tracking it in a new SAREA private field.
If this becomes a performance problem, we could move to an ioctl based solution
by adding a new parameter for the DDX to set (defaulting to the old behavior if
the param was never set of course).
This mod makes the SAREA track plane to pipe mappings and corrects the name of
the plane info variables (they were mislabeled as pipe info since until now all
code assumed a direct mapping between planes and pipes).
It also updates the flip ioctl argument to take a set of planes rather than
pipes, since planes are flipped while pipes generate vblank events.
- add forgotten init value
- use the same PGRAPH_DEBUG than the blob
- remove init of ddx reg : it should be done with object
- better handle of channel destruction
hope I didn't break anything ;)
The i830 and newer intel 2D code adds the AGP base to map offsets already,
because it wasn't doing the AGP enable which used to set dev->agp->base.
Credit goes to Zhenyu for finding the issue.
The original XGI kernel driver strobed 0xB03F each time a page was
allocated to back a GART page. When the driver was converted to use
the DRM SG interface, this code was lost. Returning it fixes a long
standing issue where the X-server would work fine the first time, but
acceleration commands would be ignored on the second X-server
invocation.
Since the heaps weren't marked as uninitialized, SG memory was never
re-allocated. This prevented the X-server from being able to restart
without re-loading the kernel module.
The DRM_XGI_PCIE_ALLOC and DRM_XGI_FB_ALLOC ioctls (and the matching
free ioctls) are unified to DRM_XGI_ALLOC. The desired memory region
is selected by xgi_mem_alloc::location. The region is magically
encoded in xgi_mem_alloc::index, which is used to release the memory.
Bump to version 0.11.0. This update requires a new DDX.
Pass the master's file pointer, as supplied to xgi_bootstrap, to
xgi_cmdlist_initialize. Associate that pointer with the memory
allocated for the command list buffer. By doing this the memory will
be automatically cleaned up when the master closes the device. This
allows the removal of some clean up code.
Need to find another way of doing this, ideally someone'd hunt down which
object/method controls it! The Xv blit adaptor is likely now broken on
cards that have pNv->WaitVSyncPossible enabled.
1. DRM_NOUVEAU_GPUOBJ_FREE
Used to free GPU objects. The obvious usage case is for Gr objects,
but notifiers can also be destroyed in the same way.
GPU objects gain a destructor method and private data fields with
this change, so other specialised cases (like notifiers) can be
implemented on top of gpuobjs.
2. DRM_NOUVEAU_CHANNEL_FREE
3. DRM_NOUVEAU_CARD_INIT
Ideally we'd do init during module load, but this isn't currently
possible. Doing init during firstopen() is bad as X has a love of
opening/closing the DRM many times during startup. Once the
modesetting-101 branch is merged this can go away.
IRQs are enabled in nouveau_card_init() now, rather than having the
X server call drmCtlInstHandler(). We'll need this for when we give
the kernel module its own channel.
4. DRM_NOUVEAU_GETPARAM
Add CHIPSET_ID value, which will return the chipset id derived
from NV_PMC_BOOT_0.
4. Use list_* in a few places, rather than home-brewed stuff.
When the GE is shut down, an empty command packet without a begin-link
must be sent. After this command is sent, wait for the hardware to go
idle. Finally, turn off the GE and disable MMIO.
This should let us allocate buffers without holding the hardware lock.
While here, add DRM_DEBUG info for the drm_bo ioctls, so you can see something
more specific than just the cmd value per ioctl.
The core DRM lastclose routine automatically destroys all mappings and
releases SG memory. XP10 DRM and DDX assumed this data stayed around
until module unload. xgi_bootstrap was reworked to recreate all these
mappings. In addition, the drm_addmap for the GART backing store was
moved into the kernel. This causes a change to the ioctl protocol and
a version bump.
Based on review comments from airlied, XGI_CHECK_PCI_CONFIG is
removed. He believes (and I tend to agree) that this is a largely
unnecessary workaround for a bug elsewhere.
There were numerous unnecessary fields in xgi_cmd_info. The remaining
fields had pretty crummy names. Cut out the cruft, and rename the
rest. As a result, the unused parameter "triggerCounter" to
triggerHWCommandList can be removed.
Debug print fix in drm_release().
Forgotten local variable init in drm_setversion().
Unnecessary put_user() in drm_addmap_ioctl().
ioctl->cmd check broken in drm_ioctl(); workaround.
The data is now in kernel space, copied in/out as appropriate according to the
This results in DRM_COPY_{TO,FROM}_USER going away, and error paths to deal
with those failures. This also means that XFree86 4.2.0 support for i810 DRM
is lost.
As a fallout, replace filp storage with file_priv storage for "unique
identifier of a client" all over the DRM. There is a 1:1 mapping, so this
should be a noop. This could be a minor performance improvement, as everything
on Linux dereferenced filp to get file_priv anyway, while only the mmap ioctls
went the other direction.
This was used to make all ioctl handlers return -errno on linux and errno on
*BSD. Instead, just return -errno in shared code, and flip sign on return from
shared code to *BSD code.
Generate the begin command once in a temporary buffer. Then,
depending on whether the command is to be written directly to the
hardware or to a secondary buffer, copy to command to the correct place.
Moved the getCurBatchBeginPort before its only caller. Modified
function to return the command ID instead of the port offset.
Function also now assumes input begin type is value.
Added code to ioctl handler to validate begin type.
xgi_cmdlist_initialize wasn't correctly checking for errors from
xgi_pcie_alloc. Furthermore, xgi_bootstrap, the one caller of
xgi_cmdlist_initialize, wasn't check its return value.
For reasons that I don't understand, the drm_addmap call would succeed
in xgi_driver_load, but writes to the map later would oops. Moving it
to xgi_bootstrap fixes this problem.
Following my nv28 kmmio dumps, nouveau_wait_for_idle() is modified to
read PTIMER and NV03_PMC_ENABLE. Also a timeout based on PTIMER value is
added, so wait_for_idle() cannot stall indefinitely (unless PTIMER is
halted). The timeout was selected as 1 giga-ticks, which for me is 1s.
NV03_PGRAPH_NSTATUS and NV03_PGRAPH_NSOURCE.
The prefix NV03 is chosen because nv10reg.h had no versioned prefix,
and the code using these registers does not check card_type.
The ioctlss XGI_ESC_DEVICE_INFO, XGI_ESC_MEM_COLLECT,
XGI_ESC_PCIE_CHECK, XGI_ESC_GET_SCREEN_INFO, XGI_ESC_PUT_SCREEN_INFO,
XGI_ESC_MMIO_INFO, and XGI_ESC_SAREA_INFO, are completely unnecessary.
The will be doubly useless when the driver is converted to the DRM
infrastructure.
Most occurances of U32 were converted to u32. These are cases where
the data represents something that will be written to the hardware.
Other cases were converted to 'unsigned int'.
U32 was the last type in xgi_types.h, so that file is removed.
These two structures were used as the request and reply for certain
ioctls. Having a different type for an ioctl's input and output is
just wierd. In addition, each structure contained fields (e.g., pid)
that had no business being there.
This change requires updates to user-space.
Allows multiple references to a single object, needed to support PCI(E)GART
scatter-gather DMA objects which would quickly fill PRAMIN if each channel
had its own.
Handle per-channel private instmem areas. This is needed to support NV50,
but might be something we want to do on earlier chipsets at some point?
Everything that touches PRAMIN is a GPU object.
Two large blocks of code were moved out of this function into separate
functions. This brought some much needed sanity to the indentation.
Some dead varaibles were removed.
Dave Airlie pointed out on IRC that idr_replace lets us know if the ID hasn't
been allocated, so we don't need a special pointer value for allocated IDs that
don't have valid information yet.
There's a difference between a drawable ID not having valid drawable
information and not being allocated at all. Not making the distinction would
break i915 DRM swap scheduling with older X servers that don't push drawable
cliprect information to the DRM.
The whole purpose of xgi_pcie_heap_check is to log information about
entries on the used_list. If XGI_DEBUG is not set, it doesn't print
anything. Therefore we can #ifdef the whole function body.
Convert open-code list iteration to use list_for_each_entry.
Comment in the code explains it. Basically, I put an if-statement
around a block of code to prevent a NULL pointer dereference that
should never happen in the first place. Eventually, this will need to
come out.
This function used to return 'void *', which was then cast to
'xgi_pcie_block_t *' at the only caller. I changed the return type to
'struct xgi_pcie_block_s *' and removed the explicit cast.
For various reasons, this ioctl was a bad idea.
At channel creation we now automatically create DMA objects covering
available VRAM and GART memory, where the client used to do this themselves.
However, there is still a need to be able to create DMA objects pointing at
specific areas of memory (ie. notifiers). Each channel is now allocated a
small amount of memory from which a client can suballocate things (such as
notifiers), and have a DMA object created which covers the suballocated area.
The NOTIFIER_ALLOC ioctl exposes this functionality.
NV04/NV10 load_context()/save_context() are stubs. I don't know enough about
how they work to implement them sanely. The "old" context_switch() code
remains hooked up, so it shouldn't break anything.
NV20 will probably break if load_context() works. No inital context values
are filled in, so when the first channel is created PGRAPH will probably end
up having its state zeroed. Some setup from nv20_graph_init() will probably
need to be moved to the per-channel context setup.
- use a timer for disabling vblank events to avoid enable/disable calls too
often
- make i915 work with pre-965 chips again (would like to structure this
better, but this hack works on my test system)
s/u64/drm_u64_t/ to allow userspace code using drm.h to compile.
Move 64 bit arg member to the beginning to avoid alignment issues with 32
bit userspace on 64 bit kernels.
It's possible that we disable vblank interrupts and clear the
corresponding flag in irq_enable_reg, but receive an interrupt at just
the wrong time, causing us to not ack it properly, nor report to the
core kernel that it was handled. Fix that case by always handling
vblank interrupts, even if the irq_enable_reg field is clear.
Fix range of frame counter registers.
Use DRM_ERR() instead of Linux specific error codes in shared code.
Remove duplicate register definitions and superfluous local variables.
Commit 9b01bd5b284bbf519b726b39f1352023cb5e9e69 introduced a
compat_ioctl handler for RADEON_SETPARAM, the sole purpose of which was
to handle the fact that on i386, alignof(uint64_t)==4.
Unfortunately, this handler was installed for _all_ 64-bit
architectures, instead of only x86_64 and ia64. And thus it breaks
32-bit compatibility on every other arch, where 64-bit integers are
aligned to 8 bytes in 32-bit mode just the same as in 64-bit mode.
Arnd has a cunning plan to use 'compat_u64' with appropriate alignment
attributes according to the 32-bit ABI, but for now let's just make the
compat_radeon_cp_setparam routine entirely disappear on 64-bit machines
whose 32-bit compat support isn't for i386. It would be a no-op with
compat_u64 anyway.
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
If the driver doesn't support vertical blank interrupts, it won't call
drm_vblank_init(), and dev->num_crtcs will be 0.
Also fix an off-by-one test against dev->num_crtcs.
- use correct refcount variable in get/put routines
- extract counter update from drm_vblank_get
- make signal handling callback per-crtc
- update interrupt handling logic, drivers should use drm_handle_vblank
- move wakeup and counter update logic to new drm_handle_vblank routine
- fixup usage of get/put in light of counter update extraction
- fix longstanding bug in signal code, update pending counter only
*after* we're sure we'll setup signal handling
Introduce tile members for future tiled buffer support.
Allow user-space to explicitly define a fence-class.
Remove the implicit fence-class mechanism.
64-bit wide buffer object flag member.
These require that the status page be referenced by a pointer in GTT, rather
than phsyical memory. So, we have the X Server allocate that memory and tell
us the address, instead.
This reverts commit 6e860d08d0.
As I said not a good plan - this macro will have to stay for now,
trying to do the vbl code with the inline was a bit messy - may need specialised
drm wait on functions
held across any operations that modify mode lists, crtc config, output
config, etc. It should be taken at high level entry points (currently just
initial config and user IOCTL).
Seems to work ok on my system, but needs more testing (with lockdep) and
review from some fresh eyes.
Conflicts:
linux-core/drm_crtc.c
linux-core/drm_fb.c
Lots of changes to merge with alanh's latest stuff:
o fix use of fb->pitch now that it has the right value
o add new helper for finding the CRTC given an FB
o fix new fb_probe/fb_remove functions to take a CRTC
o fixup callers of new FB routines
o port drm_fb changes to intel_fb
o check for errors after creating fb buffer object
o go back to using cfb_imageblit since the accel stubs aren't ready
places).
Add new FB hooks to the drm driver structure and make i915 use them for an
Intel specific FB driver. This will allow acceleration and better handling
of the command stream.
Change from DIRECTCOLOR to TRUECOLOR, and enable
support for PSEUDOCOLOR. DIRECTCOLOR support needs more work.
Add the ability to change the mode on the fbdev device.
Support depth 8, 15, 16 and 24 (and 32).
Add a /dev/fbX device per CRTC, but there's some code which
doesn't allocate the fbX device unless the output is actually
enabled. Read the code on this as it impacts the fbcon map flags.
Pick CRTC's based on the available outputs. More work could
be done here to match modes, so cloning could be achieved on
outputs. This fits more inline with what the X code does.
i915_driver_irq_postinstall was forcing vblank interrupts to pipe A when
called with vblank interrupts disabled. This caused vblank interrupts to be
accidentally re-enabled when VT switching the X server. Instead, start the
driver with vblank interrupts enabled on pipe A to support older X servers,
but then leave control over the state to the X server if it is able to do so.
This just cleans up the xf86drm.c to what I want and drm.h,
I need to fix up the kernel internals to suit these changes now.
I've moved to using struct instead of typedefs for the bo and it doesn't look
that bad so I'll do the same thing for mm and fence..
There can still be other contexts that may use page flipping later on, so don't
just unilaterally 'clean it up', which could lead to the wrong page being
displayed, e.g. when running 3D apps with a GLX compositing manager such as
compiz using page flipping.
The mode list sets all the output modes to UNVERIFIED, then probes a new list,
If a mode is on the new list and not on the old, it adds it to the old,
if a mode is on the new list and old, it just updates the status to the new
mode status.
If a mode is on the old list and not on the new, prune invalid modes should
remove all UNVERIFIED modes
VGA outs.
The drm_initial_config path has specific handling code to name
match for VGA, TMDS or LVDS. This is pretty restrictive and should
probably be dealt with to be more generic.
Check for the PENDING message when reading the attached
displays. Ensures the command has completed before continuing.
(probably need to check PENDING in other SDVO calls too)
When the kernel driver is loaded it sets up a lot of stuff..
it tears down the same stuff on unload.
This add a new map type called DRM_DRIVER which means the driver will clean the mapping up
and fix up the map cleaner
- allow drm_buffer_object_create to be called w/o dev_mapping
- fixup i915 init code to allocate memory, fb and set modes right
- pass fb to drm_initial_config for setup
- change some debug output to make it easier to spot
- fixup lvds code to use DDC probing correctly
- move EDID structures to drm_edid.h
- add EDID info structure to drm_output
- add a few routines to intel_display for getting current mode info
- add some prototypes to intel_drv.h and drm_crtc.h
The IGPGART setup code was traced using mmio-trace on fglrx by myself
and Phillip Ezolt <phillipezolt@gmail.com> on dri-devel.
This code doesn't let the 3D driver work properly as the card has no
vertex shader support.
Thanks to Matthew Garrett + Ubuntu for providing me some hardware to do this
work on.
- pull in FB DDC code (we'll have to rewrite it anyway it appears)
- add comments
- note a few FIXMEs
- make it less quiet, and more informative when it actually does print
This adds the user interfaces from Jakob and hooks them up for 3 ioctls
GetResources, GetCrtc and GetOutput.
I've made the ids for everything fbs, crtcs, outputs and modes go via idr as
per krh's suggestion on irc as it make the code nice and consistent.
It's a good idea to keep these synchronized; even though the DRM doesn't use all
the defines, maintaining two different copies is prone to errors when the diff
gets bigger.
The PGRAPH init for the various cards will need cleaning up at some point,
a lot of the values written there are per-context state left over from the
all the hardcoding done in the ddx.
It's possible some cards get broken by this commit, let me know.
Tested on: NV5, NV18, NV28, NV35, NV40, NV4E
Note that just like the values written to R300_RB3D_DSTCACHE_CTLSTAT these
values are really unknown; ideally more reverse engineering should be done to
determine what these values mean and when they should be set.
graphics objects:
- No longer takes flags/dmaobj parameters, requires some major changes
to the ddx to setup the object through the FIFO. This change is
likely to cause breakages on some cards (tested on NV05,NV28,NV35,
NV40 and NV4E).
dma objects:
- now takes a "class" parameter, not really used yet but we may need
it at some point.
- parameters are checked, so clients can't randomly create DMA objects
pointing at whatever they feel like.
misc:
- Added FB_SIZE/AGP_SIZE getparams
- Read PFIFO_INTR in PFIFO irq handler, not PMC_INTR
- Dump PGRAPH trap info on PGRAPH_INTR_NOTIFY if NSOURCE isn't
NOTIFICATION_PENDING.
Add refcounting of user waiters to the DRM hardware lock, so that we can use the
DRM_LOCK_CONT flag more conservatively.
Also add a kernel waiter refcount that if nonzero transfers the lock for the kernel context,
when it is released. This is useful when waiting for idle and can be used
for very simple fence object driver implementations for the new memory manager.
It also resolves the AIGLX startup deadlock for the sis and the via drivers.
i810, i830 still require that the hardware lock is really taken so the deadlock remains
for those two. I'm not sure about ffb. Anyone familiar with that code?
The MI_WAIT_FOR_EVENT instruction does not support waiting for several events
at once, so this should fix the lockups with page flipping when both pipes are
enabled.
Always use dev_priv->sarea_priv->pf_current_page directly. This allows clients
to modify it as well while they hold the HW lock, e.g. in order to sync pages
between pipes.
The assumption is that synchronous flips are not isolated usually, and waiting
for all of them could result in stalling the pipeline for long periods of time.
Also use i915_emit_mi_flush() instead of an old-fashioned way to achieve the
same effect.
Unfortunately, emitting asynchronous flips during vertical blank results in
tearing. So we have to wait for the previous vertical blank and emit a
synchronous flip.
Leave it to the client to wait for the flip to complete when necessary,
but wait for a previous flip to complete before emitting another one. This
should help avoid unnecessary stalling of the ring due to pending flips.
Call i915_do_cleanup_pageflip() unconditionally in preclose.
Memory types are either fixed (on-card or pre-bound AGP) or not fixed
(dynamically bound) to an aperture. They also carry information about:
1) Whether they can be mapped cached.
2) Whether they are at all mappable.
3) Whether they need an ioremap to be accessible from kernel space.
In this way VRAM memory and, for example, pre-bound AGP appear
identical to the memory manager.
This also makes support for unmappable VRAM simple to implement.
Hook into nv20 pgraph switching functions (they're identical for nv3x).
Actually call nv30_pgraph_context_init so the ctx_table is allocated.
Thanks to Carlos Martin for the help.
This means the loop will wait up to ~10ms for ring buffer space to become
available, rather than just however long it takes to check the space 10000
times. This matches other drivers' behavior when waiting for ring buffer/fifo
space.
* Pulled in some registers from nv10reg.h. Needed for context switching.
* Filled in nv30 graphics context (based on nv40_graph.c).
* Figure out nv30 context table, set up on context creation. Allows the cards automatic switching to work.
The overflows could lead to the AGP aperture overlapping the framebuffer area
in the card's address space when the latter is located at the very end of the
32 bit address space, which would result in a freeze on X server startup,
probably because the card read commands from the framebuffer instead of from
AGP.
See http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=392915 .
This comes from the Re: [patch] paravirt: isolate module ops on lkml
It needs some testing, please report any regressions caused.
Signed-off-by: Dave Airlie <airlied@linux.ie>
Now I can get 3D + working grctx switching on my NV40 without
the binary driver initialising the card first. However, this
change also breaks 3D on my C51 even *with* the binary driver's
help. So, it's likely that the weird voodoo is card-specific.
This is enough to get grctx switching going on my NV40 and C51 after
the binary driver has initialised the card first.
Bumping the drm patchlevel because the ddx needs some modifications to
have NV4x work at all with these changes.
With this change the GPU is responsible for doing the channel switch
itself. This is needed for the upcoming NV4x PGRAPH context work as
we don't yet know enough to manually swap PGRAPH contexts.
The vmalloc_32 function together with the memset to clear
the new pages are replaced with a vmalloc_user.
A pre-2.6.18 compat vmalloc_user is added.
Please replace any breakage on machines with > 1GB of memory.
ioremap must be balanced by an iounmap and failing to do so can result
in a memory leak.
Tested (compilation only) to make sure the files are compiling without
any warning/error due to new changes
Signed-off-by: Amol Lad <amol@verismonetworks.com>
Signed-off-by: Dave Airlie <airlied@linux.ie>
I need the following patch to fix compilation of
latest drm/linux-core on my ppc64 machine.
/home/mb/develop/git/drm/linux-core/savage_bci.c: In function ‘savage_driver_firstopen’:
/home/mb/develop/git/drm/linux-core/savage_bci.c:587: error: ‘DRM_MTRR_WC’ undeclared (first use in this function)
/home/mb/develop/git/drm/linux-core/savage_bci.c:587: error: (Each undeclared identifier is reported only once
/home/mb/develop/git/drm/linux-core/savage_bci.c:587: error: for each function it appears in.)
/home/mb/develop/git/drm/linux-core/savage_bci.c: In function ‘savage_driver_lastclose’:
/home/mb/develop/git/drm/linux-core/savage_bci.c:664: error: ‘DRM_MTRR_WC’ undeclared (first use in this function)
I looked at in-kernel drmP.h and it actually
has the same fix in it.
Signed-off-by: Michael Buesch <mb@bu3sch.de>
Replace r300_check_offset() with generic radeon_check_offset(), which doesn't
reject valid offsets when the framebuffer area is at the very end of the card's
32 bit address space. Make radeon_check_and_fixup_offset() use
radeon_check_offset() as well.
This fixes https://bugs.freedesktop.org/show_bug.cgi?id=7697 .
The current version didn't build on BSD, where the new functionality isn't used
yet anyway. Whoever changes that will hopefully be able to make the OSes share
this file as well.
Previously, if there were several buffer swaps scheduled for the same vertical
blank, all but the first blit emitted stood a chance of exhibiting tearing. In
order to avoid this, split the blits along slices of each output top to bottom.
- Do important card init in firstopen
- Give each channel it's own cmdbuf dma object
- Move RAMHT config state to the same place as RAMRO/RAMFC
- Make sure instance mem for objects is *after* RAM{FC,HT,RO}
On X init, PFIFO and PGRAPH are reset to defaults. This causes the GPU to
loose the configuration done by the drm. Perhaps a CARD_INIT ioctl a proper
solution to having this problem again in the future..
This will come in very handy for tiled buffers on intel hardware.
Also add some padding to interface structures to allow future binary backwards
compatible changes.
This makes an allocated block actually align itself and returns any
wasted space to the manager.
Also add some functions to grow and shrink the managed area.
This will be used in the future to manage the buffer object swap cache.
otherwise data will be missing, which becomes apparent when the kernel evicts
batch buffers which are likely to be written into in the evicted state,
and then rebound to the AGP aperture.
This means we cannot rely on the AGP module to flush the
cache for us.
driver we're on.
Avoid global cache flushes before inserting pages.
In general, they are never mapped, and not accessed through the kernel map, so
a cache flush should not be necessary. The exception is pages that are bound
cached. We might need a cache flush for those.
The mm_lock function is used when leaving vt. It evicts _all_ buffers.
Buffers with the DRM_BO_NO_MOVE attribute set will be guaranteed to
get the same offset when / if they are rebound.
Ok, I lied before.. it was a fluke it worked and required magic to repeat it..
It actually helps to fill in RAMFC entries in the correct place.
The code also clears RAMIN entirely instead of just the hash-table.
Fix buffer bound caching policy changing, Allow
on-the-fly changing of caching policy on bound buffers if the hardware
supports it.
Allow drivers to use driver-specific AGP memory types for TTM AGP pages.
Will make AGP drivers much easier to migrate.
Adapt for new functions in the 2.6.19 kernel.
Remove the ability to have multiple regions in one TTM.
This simplifies a lot of code.
Remove the ability to access TTMs from user space.
We don't need it anymore without ttm regions.
Don't change caching policy for evicted buffers. Instead change it only
when the buffer is accessed by the CPU (on the first page fault).
This tremendously speeds up eviction rates.
Current code is safe for kernels <= 2.6.14.
Should also be OK with 2.6.19 and above.
Added preliminary support for context switches (triggers the interrupts, but hangs after the switch ; something's not quite right yet).
Removed the PFIFO_REINIT ioctl. I hope it's that a good idea...
Requires the upcoming commit to the DDX.
mach64_state.c: convert the DRM_MACH64_BLIT ioctl to submit a pointer to
user-space memory rather than a DMA buffer index, similar to DRM_MACH64_VERTEX.
This change allows the DDX to map the DMA buffers read-only and eliminate a
security problem where a client can alter the contents of the DMA buffer after
submission to the DRM.
This change also affects the DRI/DRM interface. Performace-wise, it basically
affects PCI mode where I get a ~12% speedup for some Mesa demos I tested.
This is mainly due to eliminating an ioctl for allocating the DMA buffer.
mach64_dma.c: move the responsibility for allocating memory for the DMA ring
in PCI mode to the DDX.
This change affects the DDX/DRM interface and unifies a couple of PCI/AGP code
paths for ring memory in the DRM.
Bump the mach64 DRM version major and date.
Map the DMA buffers from the same linear address as the vertex bufs. If
dev->agp_buffer_token is not set, mach64 drm maps the DMA buffers from
linear address 0x0.
Initialize the spinlock unconditionally when struct drm_device is filled in,
and return early in drm_locked_tasklet() if the driver doesn't support IRQs.
This fixes issues on X server startup with versions of xf86-video-intel that
enable the IRQ before they have a context ID.
(cherry picked from 7af93dd984 commit)
It looks like 'after a while', I915REG_INT_IDENTITY_R for some reason always has
VSYNC_PIPEB_FLAG set in the interrupt handler, even though pipe B is disabled.
So we only increase dev->vbl_received if the corresponding bit is also set in
dev->vblank_pipe.
(cherry picked from 881ba56992 commit)
It looks like this would have caused signals to always get sent on the next
vertical blank, regardless of the sequence number.
(cherry picked from cf6b2c5299 commit)
When this flag is set and the target sequence is missed, wait for the next
vertical blank instead of returning immediately.
(cherry picked from 89e323e490 commit)
Initialize it to default value if it hasn't been set by the X server yet.
In i915_vblank_pipe_set(), only update dev_priv->vblank_pipe and call
i915_enable_interrupt() if the argument passed from userspace is valid to avoid
corrupting dev_priv->vblank_pipe on invalid arguments.
(cherry picked from 87c57cba1a commit)
Handle relative as well as absolute target sequence numbers.
Return error if target sequence has already passed, so userspace can deal with
this situation as it sees fit.
On success, return the sequence number of the vertical blank when the buffer
swap is expected to take place.
Also add DRM_IOCTL_I915_VBLANK_SWAP definition for userspace code that may want
to use ioctl() instead of drmCommandWriteRead().
(cherry picked from d5a0f10751 commit)
This makes it easier for userspace to know when it needs to allocate an ID.
Also free drawable information memory when it's no longer needed.
(cherry picked from df7551ef73 commit)
This uses the core facility to schedule a driver callback that will be called
ASAP after the given vertical blank interrupt with the HW lock held.
(cherry picked from 257771fa29 commit)
Actually make the existing ioctls for adding and removing drawables do
something useful, and add another ioctl for the X server to update drawable
information. The only kind of drawable information tracked so far is cliprects.
(cherry picked from 29598e5253 commit)
When the vertical blank interrupt is enabled for both pipes, pipe A is
considered primary and pipe B secondary. When it's only enabled for one pipe,
it's always considered primary for backwards compatibility.
(cherry picked from 0c7d7f4361 commit)
It looks like 'after a while', I915REG_INT_IDENTITY_R for some reason always has
VSYNC_PIPEB_FLAG set in the interrupt handler, even though pipe B is disabled.
So we only increase dev->vbl_received if the corresponding bit is also set in
dev->vblank_pipe.
Initialize it to default value if it hasn't been set by the X server yet.
In i915_vblank_pipe_set(), only update dev_priv->vblank_pipe and call
i915_enable_interrupt() if the argument passed from userspace is valid to avoid
corrupting dev_priv->vblank_pipe on invalid arguments.
Handle relative as well as absolute target sequence numbers.
Return error if target sequence has already passed, so userspace can deal with
this situation as it sees fit.
On success, return the sequence number of the vertical blank when the buffer
swap is expected to take place.
Also add DRM_IOCTL_I915_VBLANK_SWAP definition for userspace code that may want
to use ioctl() instead of drmCommandWriteRead().
Actually make the existing ioctls for adding and removing drawables do
something useful, and add another ioctl for the X server to update drawable
information. The only kind of drawable information tracked so far is cliprects.
When the vertical blank interrupt is enabled for both pipes, pipe A is
considered primary and pipe B secondary. When it's only enabled for one pipe,
it's always considered primary for backwards compatibility.
This takes up two more ring buffer entries per rectangle blitted but makes sure
the blit is performed top to bottom, reducing the likelyhood of tearing.
Change the fence object interface somewhat to allow some more flexibility.
Make list IOCTLS really restartable.
Try to avoid busy-waits in the kernel using immediate return to user-space with an -EAGAIN.
- Mark the correct RAMIN slots as free (oops)
- Remove a VRAM alloc that shouldn't have been there (oops)
- Move HT init out of firstopen() and into dma_init()
- Setup PFIFO_RAM{HT,FC,RO} in pfifo_init()
Add a function to free buffers on hold for destruction if their
fence object has expired.
Add a timer to periodically call that function when there are
buffers pending deletion.
First warning result from open-coded PTR_ERR,
the rest is caused by code like this:
*(u32 *) ((u32) buf_priv->kernel_virtual + used)
I've also fixed a missing PTR_ERR in i830_dma.c
From: Denis Vlasenko <vda.linux@googlemail.com>
Signed-off-by: Dave Airlie <airlied@linux.ie>
The overflows could cause valid offsets to get rejected under some
circumstances, e.g. when the framebuffer resides at the very end of the card's
address space.
0x00 EXE fence. Signals when command stream interpreter has reached the point
where the fence was emitted.
0x01 FLUSH fence. Signals when command stream interpreter has reached the point
where the fence was emitted, and all previous drawing operations have been
completed and flushed.
Implements busy wait (for fastest response time / high CPU) and
lazy wait (User interrupt or timer driven).
First warning result from open-coded PTR_ERR,
the rest is caused by code like this:
*(u32 *) ((u32) buf_priv->kernel_virtual + used)
I've also fixed a missing PTR_ERR in i830_dma.c
From: Denis Vlasenko <vda.linux@googlemail.com>
Signed-off-by: Dave Airlie <airlied@linux.ie>
Commit 2a47f6bfec caused the CHIP_IS_AGP flag to
get 'restored' with PCI(e) cards. I can't think of a way to fix this without
introducing a (otherwise redundant) CHIP_IS_PCI flag.
Not doing this might waste bus bandwidth or even cause memory corruption or
system crashes on systems that check bus transfers. No such incident has been
reported though.
32-bit physical device addresses are mapped directly to user-tokens. No
duplicate maps are allowed, and the addresses are assumed to be outside
of the range 0x10000000 through 0x30000000. The user-token is identical
to the 32-bit physical start-address of the map.
64-bit physical device addressed are mapped to user-tokens in the range
0x10000000 to 0x30000000 with page-size increments. The user_token should
not be interpreted as an address.
Other map types, like upcoming TTM maps are mapped to user-tokens in the
range
0x10000000 to 0x30000000 with page-size increments. The user_token should
not be interpreted as an address.
This keeps compatibility with buggy drivers, while still implementing a
hashed map lookup. The SiS and via device driver major bumps are
reverted.
0x10000000 to 0x90000000 in PAGE_SIZE increments.
Implement hashed map lookups.
This potentially breaks both 2D and 3D drivers. If so, the corresponding
2D and 3D driver should be fixed, and it's corresponding drm device driver
should have its major bumped as soon as possible.
Bump sis and via drm device driver majors.
The SiS and Unichrome 3D drivers are fixed in Mesa CVS HEAD and
mesa_6_4_branch.
i915 vblanks can be generated from either pipe a or b, however a disabled
pipe generates no interrupts. This change allows the X server to select
which pipe generates vblank interrupts.
more problems with flickering (bug #6637). drm may not be appropriate
place for this, since doing that flush there might both be overkill and
insufficient in some cases. However, it's hard to figure out when that
flush is needed, so this has to suffice. There does not seem to be a
performance penalty associated with it.
Linux, which broke on FreeBSD. DRM_COPY_*_IOCTL checks for the size
parameter matching the ioctl's command size there, since the copin/out
happened earlier.
chips may be problematic). Leave the existing entries for new chips in
though. Remove ids not known by ddx (secondary ids, non-existant,...).
Correct some entries (name/family). Make the radeon family enum look
more alike the ddx/dri versions. See #5413
that particular file. Its contents have changed a good bit since the
original sis code, and the original sis code didn't care much about
attribution since it routinely disclaims Precision Insight/VA Linux
from responsibility. Also, adjust formatting around license headers
(have a comment open immediately before the "Copyright" line, not as a
runon of any previous comments) for automatic processing into FreeBSD,
where /*- is used to signal the beginning of license headers for
automatic compilation of license lists.
single DRM_COPY_FROM_USER rather than DRM_VERIFYAREA_READ followed by
tons of DRM_COPY_FROM_USER_UNCHECKED. I don't like the look of the
temporary space allocation, but I like the simplification in the rest
of the file. Tested with glxgears, tuxracer, and q3 on a savage4.
drm_mtrr_{add,del} for handling the MTRR setup. Still has a LOR issue
with DRM_VERIFYAREA_READ/DRM_COPY_FROM_USER_UNCHECKED in savage_bci.c
-- this won't work with the fine-grained locking in use, and just doing
a single copyin to a temporary will probably work fine. Also note that
the module leaks approximately 4 kb on unload.
up a good bit, I think. Also, remove the agp_uninit() function which
has lain around as a noop for years now. The FreeBSD DRM is now all
compiling, with the exception of via. One known sleeping-with-lock-held
issue remains.
radeon_cp.c to use a drm_local_map_t-type mapping (drm_core_ioremap
rather than drm_ioremap), which contains private device mapping
information on BSD. I also changed the ati_pcigart interface to use
"void *" for pointers to kva rather than "unsigned long". While PCIGART
support appears to be broken on FreeBSD currently, I think this is not
new, and BusType PCI remains working on my r100 in Linux.
firstopen, by making drm_addmap require the drm device lock to be held.
Also, make matching of kernel maps match linux by requiring shm matches
to have the contains_lock flag set if the offset doesn't match.
drm_agp_foo_ioctl pair. Modifies the MGA DRM to use the drm_agp_foo
functions instead of the drm_foo_agp functions. The drm_foo_agp
functions are no longer exported by drm.ko.
Ensures that dma->seg_count and dma->page_count are properly set in
drm_addbufs_{agp,sg,fb}. drm_addbufs_pci was already correct.
Ensures that mga_do_agp_dma_bootstrap correctly sets agp_buffer_token.
At this point PCI DMA is still broken.
Xorg bug: #4797 Reviewed by: Dave Airlie, Eric Anholt Signed-off-by: Ian
Romanick <idr@us.ibm.com>
mga_do_agp_dma_bootstrap fails causes problems if
mga_do_pci_dma_bootstrap succeeds. This commit makes it possible to do
a "minimal" clean up instead. I'm still trying to figure out what is
causing the failures in mga_do_agp_dma_bootstrap...
Signed-off-by: Ian Romanick <idr@us.ibm.com>
R200_EMIT_PP_TXCTLALL_0-5 (replaces R200_EMIT_PP_TXFILTER_0-5, 2 more
regs) and R200_EMIT_ATF_TFACTOR (replaces R200_EMIT_TFACTOR_0 (8 consts
instead of 6)
- Comment out the "is this mapping/bufs in allocated AGP" bits in BSD
because they break mga (which uses AGP allocation that doesn't track
entries). It's not a security issue when we still have the related
ioctls marked root-only.
- Apply some power-of-two alignment restrictions to hopefully avoid some
panicing in bad cases of drm_pci_alloc() on FreeBSD.
- Add verbosity to some error handling that I found useful while debugging.
allocate the resource RF_ACTIVE, pull out the appropriate value, and
return it. However, allocating large framebuffers RF_ACTIVE would run
the system out of KVA, and this also left open the possibility of the
resource getting moved after getting the offset. Instead, when either
of these are called, allocate the resource if it isn't allocated
already (non-RF_ACTIVE) and store it in the DRM device, to be cleaned
up on lastclose.
me to match other drivers and avoid ifdeffing. The linux via_drv.c will
be moved from shared-core to linux-core soon by repocopy.
Submitted by: Jake Burkholder <jake@FreeBSD.org> Tested by: unichrome
caller on fb / agp memory alloc and free. Otherwise malicious clients
can register allocations on other clients or free memory used by other
clients which will lead to severe memory manager inconsistensies.
understandable: preinit -> load postinit -> (removed) presetup ->
firstopen postsetup -> (removed) open_helper -> open prerelease ->
preclose free_filp_priv -> postclose pretakedown -> lastclose
postcleanup -> unload release -> reclaim_buffers_locked version ->
(removed)
postinit and version were replaced with generic code in the Linux DRM
(drivers now set their version numbers and description in the driver
structure, like on BSD). postsetup wasn't used at all. Fixes the savage
hooks for initializing and tearing down mappings at the right times.
Testing involved at least starting X, running glxgears, killing
glxgears, exiting X, and repeating.
Tested on: FreeBSD (g200, g400, r200, r128) Linux (r200, savage4)
buffers that have been allocated from AGP. This includes some new
capable(CAP_SYS_ADMIN) checks, these functions are also protected by
the root requirement on the IOCTL macros.
driver's preinit routine, and by using DRM_COPY_TO_USER_IOCTL when
copying out to an ioctl's data pointer. Pulled from the latest version
of my drm-hook-rename.diff and only compile-tested after that.
code duplication, and it also hands you the map pointer so you don't
need to re-find it.
- Remove the permanent maps flag. Instead, for register and framebuffer
maps, we always check whether there's already a map of that type and
offset around. Move the Radeon map initialization into presetup (first
open) so it happens again after every takedown.
- Remove the split cleanup of maps between driver takedown (last close) and
cleanup (module unload). Instead, always tear down maps on takedown,
and drivers can recreate them on first open.
- Make MGA always use addmap, instead of allocating consistent memory in
the PCI case and then faking up a map for it, which accomplished nearly
the same thing, in a different order. Note that the maps are exposed to
the user again: we may want to expose a flag to avoid this, but it's
not a security concern, and saves us a lot of code.
- Remove rmmaps in the MGA driver. Since the function is only called during
takedown anyway, we can let them die a natural death.
- Make removal of maps happen in one function, which is called by both
drm_takedown and drm_rmmap_ioctl.
Reviewed by: idr (previous revision) Tested on: mga (old/new/pci dma),
radeon, savage
zero does not mean no privs, instead it grants write access
_DRM_READ_ONLY only applies to non-root users. Problem is only in CVS,
initmaps are not in the kernel yet.
zero does not mean no privs, instead it grants write access
_DRM_READ_ONLY only applies to non-root users. Problem is only in CVS,
initmaps are not in the kernel yet.
This patch adds serveral new ioctls and a new query to get_param query to
support PCI MGA cards.
Two ioctls were added to implement interrupt based waiting. With this
change, the client-side driver no longer needs to map the primary DMA
region or the MMIO region. Previously, end-of-frame waiting was done by
busy waiting in the client-side driver until one of the MMIO registers
(the current DMA pointer) matched a pointer to the end of primary DMA
space. By using interrupts, the busy waiting and the extra mappings are
removed.
A third ioctl was added to bootstrap DMA. This ioctl, which is used by the
X-server, moves a *LOT* of code from the X-server into the kernel. This
allows the kernel to do whatever needs to be done to setup DMA buffers.
The entire process and the locations of the buffers are hidden from
user-mode.
Additionally, a get_param query was added to differentiate between G4x0
cards and G550 cards. A gap was left in the numbering sequence so that,
if needed, G450 cards could be distinguished from G400 cards. According
to Ville Syrjälä, the G4x0 cards and the G550 cards handle
anisotropic filtering differently. This seems the most compatible way
to let the client-side driver know which card it's own. Doing this very
small change now eliminates the need to bump the DRM minor version
twice.
http://marc.theaimsgroup.com/?l=dri-devel&m=106625815319773&w=2
A number of ioctl handlers in linux-core were also modified so that they
could be called in-kernel. In these cases, the in-kernel callable
version kept the existing name (e.g., drm_agp_acquire) and the ioctl
handler added _ioctl to the name (e.g., drm_agp_acquire_ioctl).
This patch also replaces the drm_agp_do_release function with
drm_agp_release. drm_agp_release (drm_core_agp_release in the previous
patch) is very similar to drm_agp_do_release, and I saw no reason to
have both.
This commit *breaks the build* on BSD. Eric said that he would make the
required updates to the BSD side soon.
Xorg bug: 3259 Reviewed by: Eric Anholt
There were two problems. First, the 'warp' and 'primary' pointers weren't
cleared, so mga_do_cleanup_dma, which gets called multiple times, would
try to ioremapfree them multiple times. This resulted in the new error
messages to syslog. The second problem was the, since the dev_private
structure isn't reallocated and cleaned out in mga_do_init_dma, when
the server is reloaded idle-waits would wait for impossible values.
I have given this patch some more riggorous testing. This includes:
- Load module, start server, run GL app, stop server, unload module.
- Load module, start server, run GL app, stop server, unload module, reload
module, restart server, run GL app.
- Load module, start server, run GL app, stop server, restart server, run
GL app, stop server, unload module.
In all three cases, everything worked as expected. Please let me know if
there are any further regressions with this patch.
Xorg bug: 3408 Reported by: Chris Rankin
the driver is loaded and is always expected to be around, it should
only be deallocated when the driver is unloaded.
Xorg bug: #3408 Reported by: Chris Rankin
linux-core to free pci memory without freeing the structure. Linux-core
internals often create pci dma handle structures on the stack due to
the lack of a drm_local_map_t to store them in properly. Fix the
original drm_pci_free to actually free the dma handle structure instead
of leaking it.
Signed-off-by: Sergey Vlasov <vsu@altlinux.ru>
There's two choices when fb is or isn't loaded as we treat ourselves as a
PCI driver in the latter case.
If we are a PCI driver, then register the suspend/resume functions
directly. If not, then we register as a sysdev and pick up the
suspend/resume actions and pump them down into a generic *power
function.
It'll be nice when this little mess is sorted out with regard to being a
real PCI driver ;-/
that a device absolutely is, absolutely is not, or may or may not be
AGP. Modify the i915 DRM to use this to force all i9x5 devices to be
"AGP" (even the PCI-e devices).
Reported by: Lukas Hejtmanek
two routines (one for G200 and one for G400) were replaced with static
const variables and a single function that returns the correct size.
The code to generate an error if the allocated WARP region is too small
was refactored from mga_warp_install_{g200,g400}_microcode to
mga_wrap_install_microcode.
mga_warp_microcode_size is global to the MGA DRM because it will soon be
used by code in another file.
is now allocated (and partially filled in) by the new
mga_driver_preinit function.
This allows the driver to detect the type of card (i.e., G200 class vs.
G400 class) on its own. The chipset value passed to mga_dma_init is now
ignored. This same technique is used by the radeon DRM.
As a result of this, mga_driver_pretakedown was converted to
mga_driver_postcleanup. This routine gets called in some other places
than might be expected, and it sets the dev_private pointer to NULL.
That little gem took over an hour to track down. :(
platform-specific drm_device_is_agp function. Added implementation of
this function the the Linux-specific portion of the MGA driver to
detect PCI G450 cards. Added code to the Linux-specific portion of the
generic DRM layer to not initialize AGP infrastructure if the card is
not AGP (this matches what already existed in BSD).
Bumped the driver date and the driver patch-level for MGA.
This mostly fixes bugzilla #3248. The BSD side still needs an
implementation of mga_driver_device_is_agp.
with IOMMUs and such. There is one usage of the forbidden vtophys()
left in drm_scatter.c which will be fixed up soon. This required a KPI
change for drm_pci_alloc/free() to return/use a drm_dma_handle_t that
keeps track of os-specific bits, rather than just passing around the
vaddr/busaddr/size.
Submitted by: Tonnerre Lombard (partially) Tested on: FreeBSD: Rage128
AGP/PCI Linux: Savage4 AGP/PCI
flag DMA_QUIESCENT (typically the X server), but gets interrupted by a
signal. The locking IOCTL should then return an error, but if
DMA_QUIESCENT succeeds it returns 0, and the client falsely thinks it
has the lock. In addition The client waits for DMA_QUISCENT and
possibly DMA_READY without having the lock.
1) Security fix: Stopped possible MMIO access to PCI DMA area for the
unichrome Pro.
2) Fixed an odd cpu usage problem by padding small AGP DMA submissions.
Bumped patchlevel.
New PCI command parser. Moved from via_dma.c to via_verifier.c so functions
with similar functionality are close to eachother.
Moved video related functions to via_video.c, which might be extended in
the future, as new video functionality is added.
New device-specific generic IRQ IOCTL, similar to the general VBLANK IOCTL,
but with support for multiple device IRQ sources and functionality.
Support for Unichrome Pro PM800/CN400 video DMA commands in verifier and
PCI parser.
Support for Unichrome Pro PM800/CN400 HQV IRQs in the new generic IRQ
IOCTL.
Bumped minor. New version 2.6.0.
what wait_event_interruptible_timeout() does, use the function and just
change the return values appropriately.
Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
so we need to consult the EFI memory map before we try to set the write
combine attribute of a page. This patch will try to map a page write
combined if it's not an AGP page and the EFI memory map says it's ok,
otherwise it falls back to a regular, uncached mapping. Can someone
please apply this to the drm tree?
From: Jesse Barnes
1. Initialize futex locks to zero on device init.
2. Remove some stray defines from via_drm.h
3. Prepare via_drm.h for drm client inclusion. The goal is to share a
common file with common definitions.
4. Sync shared / shared-core via_drm.h
5. Bump minor, because of the futex lock initialization.
after ~5min: buffer aging was subtly broken. Part of this may have also
affected vertex DMA buffer aging and client-side texture heap aging,
though with less fatal consequences. Bumped minor version and driver
date.
command and vertex DMA don't work at the same time. Command DMA
performance is superior and works with all vertex formats. Bumped minor
version and driver date.
event counters in the SAVAGE_EVENT_EMIT/WAIT ioctls. This is needed for
reliable client-side texture heap aging. Slightly simplified DMA buffer
aging while at it. Bumped minor version and driver date.
privileges on Radeon hardware. Essentially, a malicious program could
submit a packet containing an offset (possibly in main memory) to be
rendered from/to, while a separate thread switched that offset in
userspace rapidly between a valid value and an invalid one.
radeon_check_and_fixup_offset() would pull the offset in from user
space, check it, and spit it back out to user space to be copied in
later by the emit code. It would sometimes catch the bad value, but
sometimes the malicious program could modify it after the check and get
an invalid offset rendered from/to.
Fix this by allocating a temporary buffer and copying the data in at once.
While here, make the cliprects stuff not do the VERIFYAREA_READ and
COPY_FROM_USER_UNCHECKED gymnastics, avoiding a lock order reversal on
FreeBSD. Performance impact is negligible -- no difference on r200 to
~1% improvement on rv200 in quake3 tests (P4 1Ghz, demofour at
1024x768, n=4 or 5).
FreeBSD. Add drm_get_resource_{start|len} so linux-specific stuff
doesn't need to be in shared code.
- Fix mach64 build by using __DECONST to work around passing a const
pointer to useracc, which is unfortunately not marked const.
- Get rid of a lot of maplist code by not having dev->maplist be a pointer,
and by sticking the link entries directly in drm_local_map_t rather
than having a separate structure for the linked list.
- Factor out map uninit and removal into its own routine, rather than
duplicating in both drm_takedown() and drm_rmmap().
- Hook up more driver functions, and correct FreeBSD-specific bits of
radeon_cp.c, making radeon work.
- Baby steps towards using bus_space as we should.
- Add $(DESTDIR) for distributors doing package creation
- Remove OS-specific include path from build
- Add /usr/include/drm for driver-kernel API
- Install all of shared-core/*.h in /usr/include/drm
- Rename xf86drm.h to libdrm.h since we're not X biased anymore
- Include backwards compat for xf86drm.h name, with a warning
- Fix libdrm source to account for drm.h living in /usr/include/drm
Implemented SAVAGE_CMD_DMA_IDX and SAVAGE_CMD_VB_IDX for ELTs support in
the _savage_render_stage of the 3D driver
Bumped minor version and driver date
- create libdrm.so target
- build it by default
- drop xf86drmCompat.c from the build
- make 'clean' target never fail
- use pattern rules for .c -> .o for parallelism
- add 'install' target
The attached patch adds a new buffer type DRM_FB_BUFFER. It works like AGP
memory but uses video memory.
From: austinyuan@viatech.com.cn (fd.o bug 1668) Signed-off-by: Dave Airlie
<airlied@linux.ie>
passed to the cmdbuf ioctl (try xeyes on top of glxgears ;-)
Tightened the texture state check
Bumped Savage DRM version to 2.1.0 so that DRI driver can (theoretically)
depend on the DRM to manage the scissor registers
Release video futexes when context is destroyed (This was previously done
by the X server).
Added New Unichrome Pro VIDEO DMA commands to the verifier.
Added Quiescent heavyweight lock mode.
implementation errors). Direct hardware (MMIO, BCI) access is no longer
needed in the Mesa driver. Bumped version to 2.0.0. Corresponding
changes to the DDX and Mesa drivers are being committed.
bug in the cmdbuf_size ioctl. Modified ring-buffer jump code AGAIN, due
to new oddities discovered on the Unichrome Pro with faster processors.
Bumped patchlevel and driver date.
1. Improved security check of AGP texture adresses.
2. Hopefully last fix of ring-buffer jump oddities.
3. Added ioctl to check available space and command regulator lag in
ring-buffer. This is needed for 3D application responsiveness.
1. The command verifier was never initialized in the non-core source tree.
2. Check added that the AGP ring buffer has been initialized before
accepting command buffer.
3. Free space check in the AGP buffer is moved to after command
verification, which is more optimal in most cases.
drivers/char/drm/sis_ds.h:88:12: warning: dubious one-bit signed
bitfield drivers/char/drm/sis_ds.h:89:16: warning: dubious one-bit
signed bitfield
Signed-off-by: Randy Dunlap <rddunlap@osdl.org> Signed-off-by: Dave Airlie
<airlied@linux.ie>
driver. It can now handle the 3D OpenGL commands from the Mesa
unichrome driver.
Added vsync frequency detection support. This will be used in the future
for XvMC and better frame timing.
Bumped minor version number and driver date.
ioctls with dev_lock, which is a major step toward being able to remove
Giant. Covers some new pieces (dev->unique*) in the core, and avoids
one call down into system internals with the drm lock held, which is
usually bad (FreeBSD LOR #23, #27).
code. Remove the "drv" from sisdrv, as it's unnecessary. Use the
drm_pci functions in i915 instead of per-os implementations of the
same. Avoid whitespace within fields in drm_pciids.txt (one of the r300
definitions), since it breaks the bsd pciids script. Tested on sis,
mga, r128. i915 needs more work.
done by Martin Lexa (martin at martinlexa dot cz). Now that we've got
porting for all three major BSDs (and the fourth being very similar to
FreeBSD), move the mostly-duplication drm_os_* files into drmP.h.
Remove some cruft from linux heritage and from pieces of the DRM that
have since been removed.
Note that things are still not quite working for even FreeBSD, but these
are first steps at cleanup, and just a WIP checkpoint.
martinlexa dot cz). Now that we've got porting for all three major BSDs
(and the fourth being very similar to FreeBSD), move the
mostly-duplication drm_os_* files into drmP.h. Remove some cruft from
linux heritage and from pieces of the DRM that have since been removed.
Note that things are still not quite working for even FreeBSD, but these
are first steps at cleanup, and just a WIP checkpoint.
DRM_IOCTL_VIA_DMA_INIT DRM_IOCTL_VIA_CMDBUFFER DRM_IOCTL_VIA_FLUSH
The first ioctl sets up an area in AGP memory that will be used as the ring
buffer. The second ioctl copies a command buffer from user space memory
to the ring buffer. The third ioctl waits for engine idle until it
returns.
The motivation for this patch is to avoid the wait for engine idle call
before each buffer flush in the current DRI driver. With this patch,
the DRI driver can continue to flush its buffer as long as there is
free space in the ring buffer.
This patch adds an additional copy operation on the command buffer. This
buffer copying is necessary to support multiple DRI clients rendering
simultaneously. Otherwise, more CPU time will be spent in the busy loop
waiting for engine idle between DRI context switch. Even in the single
client case, the tradeoff is reasonable in comparision to the kernel
call to check for free buffer space for the client to render directly
to the ring buffer.
hotplug event plus the addition of one requesting RESET. Put your
scripts in /etc/hotplug.d/drm to run. kernel class_simple generates the
ADD/REMOVE events. No cards currently request RESET, the flag is there
to stop you from resetting your boot display.
kernel allocated and then mmap-installed to userspace, but instead
makes it use the kernel virtual address directly instead.
From: Arjan van de Ven <arjanv@redhat.com>
xc/xc/programs/Xserver/hw/xfree86/os-support/linux/drm/xf86drm.c so
that it matches the prototypes in xf86drm.h XFree86 bug: Reported by:
Submitted by: Reviewed by: Obtained from:
and/or d->send_sizes. When these functions are called from gamma_dma,
these pointers are user pointers and are thus not safe to deref. This
patch copies over the pointers inside gamma_dma_priority and
gamma_dma_send_buffers.
Submitted-by: Robert T. Johnson <rtjohnso@eecs.berkeley.edu> Signed-off-by:
Dave Airlie <airlied@linux.ie>
old DRM interface, the devname was set in DRM(setunique), but with the
current DRM interface >=1.1 the devname is not being set in
DRM(set_busid).
From: Alan Swanson Approved-by: Dave Airlie <airlied@linux.ie>
2. Big change to the XvMC part of the SAREA. OpenGL clients will not suffer
from this, and via XvMC is still alpha. Needed to make future additions
to XvMC (More decoders and overlays) possible.
3. Bumped version number to 1.3.0.
allows the mesa drivers to use a single definition of the DRM
sarea/IOCTLS located in the drm driver directory. Adjustments were made
to the 2D drivers to not include these changes. Changes to the mesa
copy of DRM were copied to the DRI copy. XFree86 bug: Reported by:
Submitted by: Reviewed by: Obtained from:
dev->lock.hw_lock is already set. This fixes the case of two X Servers
running on the same head on different VTs with interface 1.1, by making
the 2nd head fail to inizialize like before.
irq handler, fixes kernel ooops. comment out some setting of flags that
are done in DRM(setup) (not sure why both of the above is done in the
irq handler)
a busid that doesn't correspond to the device the DRM is attached to.
This is a breaking of backwards-compatibility only for the
multiple-DRI-head case with X Servers that don't use interface 1.1.
- Move irq_busid to drm_irq.h and make it only return the IRQ for the
current device. Retains compatibility with previous X Servers, cleans
up unnecessary code. This means no irq_busid on !__HAVE_IRQ, but can be
changed if necessary.
- Bump interface version to 1.2. This version when set signifies that the
control ioctl should ignore the irq number passed in and enable the
interrupt handler for the attached device. Otherwise it errors out when
the passed-in irq is not equal to the device's.
- Store the highest version the interface has been set to in the device.
- Fix a recursion on DRM_LOCK in irq_uninstall on FreeBSD. This leaves
irq_uninstall being done without the lock in some cases, but it was
racey anyways.
the 2D driver initializes MC_FB_LOCATION and related registers sanely
the DRM deduces the layout from these registers
clients use the new SETPARAM ioctl to tell the DRM where they think the
framebuffer is located in the card's address space
the DRM uses all this information to check client state and fix it up if
necessary
This is a prerequisite for things like direct rendering with IGP chips and
video capturing.
in bufs_info sysctl handler. dev->dma and dev->dma_lock existence are
protected by DRM_LOCK(). Fixes panic on sysctl hw.dri when the device
is uninitialied (when you aren't in X).
server or client to notify the DRM that it expects a certain version of
the device dependent or device independent interface. If the major
doesn't match or minor is too large, EINVAL is returned. A major of -1
means that the requestor doesn't care about that portion of the
interface. The ioctl returns the actual versions in the same struct.
- Introduce DRM DI interface version 1.1. If the server requests version
1.1, then the DRM sets the unique itself according to the busid of the
device it probed, which may then be accessed as normal using getunique.
- Request version 1.1 in libdrm's drmOpenByBusID, allowing the X Server to
request based on a BusID. Introduce a wrapper for DRM_IOCTL_SET_VERSION
and bump libdrm minor version.
- Pass the busid in DRIScreenInit if libdrm can handle both a busid and
name. This allows drmOpenByBusID to be used to find the DRM instead of
just the driver name, which allows us in the future to tie a DRM more
strongly to the device it probed to. Introduce a function
DRICreatePCIBusID which creates a busid in the form pci:oooo:bb:dd.f
similar to linux's pci_name() function. This matches the format used by
the DRM in version 1.1. libdrm knows how to match both this format and
the old PCI:b:d:f format.
- Use the new DRICreatePCIBusID function in the *_dri.c to request the new,
more exact busid format.
get the drm_file_t * based on the filp passed in ioctl handlers.
- Use this macro on BSD for simplification and improve its error reporting.
Make failure to find the drm_file_t * print as an error, not debug.
This failure may be part of the problem with KDE.
- Make debug and error print macros include the pid on BSD.
data into temporary variables with the lock held then outputting to
sysctls with the lock released. Rearranged a little extra code to aid
this. Note that drm_memory_debug.h hasn't had this fix applied, but I
consider that code to be just about dead anyway.
used by root (the X Server) which are not locked. However, it should
deal with lost-IRQ issues on -current which I think people have been
experiencing but I am unable to reproduce (though I understand why they
would occur, because of a bug of mine). Note that most of the locking
(DRM_LOCK()/UNLOCK()) is all covered by Giant still, so it doesn't
matter yet.
- Remove locking on FreeBSD-stable and NetBSD. These are covered by the
fact that there is no reentrancy of the kernel except by interrupts,
which are locked using spldrm()/splx() instead.
from __HAVE_DMA. This will be useful for adding vblank sync support to
sis and tdfx. Rename dma_service to irq_handler, which is more
accurately what it is.
- Fix the #if _HAVE_DMA_IRQ in radeon, r128, mga, i810, i830, gamma to have
the right number of underscores. This may have been a problem in the
case that the server died without doing its DRM_IOCTL_CONTROL to
uninit.
just a single instance. Moved the PCI ID lists from <card>_drv.c in BSD
to <card>.h. The PCI ID lists include a driver private field, which may
be used by drivers for chip family or other information. Based on work
by jonsmirl.
- Make tdfx_drv.c and tdfx.h match other drivers.
- Fixed up linking of sis shared files.
Tested with Radeon and SiS on Linux and FreeBSD, including a Linux setup
with
2 SiS cards in a machine, but only one head being used (with DRI)
Change some nearby memset()s to bzero()s or to calloc allocations to
take advantage of M_ZERO). Reverse some error tests to reduce high
levels of indentation. Move the sg_cleanup() call out of the maplist
loop in DRM(takedown)-- I can't see any need for it to be inside.
allocate framebuffer memory without sisfb, and a new ioctl to be used
by the X Server which tells the DRM what region of framebuffer memory
to allocate from. Also fixes a possibility to panic the kernel I
believe. Tested on linux with sisfb and FreeBSD (without sisfb) with
new DRI only.
2.5.0. It still has some issues, including a flicker in the fps meter
in tuxracer and I've seen garbage left behind after moving/closing
windows. However, it's usable. Add the Option "EnablePageFlip" "YES" to
use it.
respectively. Splited the work out of the ioctls and renamed (with the
_ioctl prefix). Added some more documentation. Did the same for
drm_sgpsupport.h.
DRM_*MEMORYBARRIER we had were related to an MMIO space. This means
arch-specific code on the BSDs, unfortunately. Also add
DRM_MEMORYBARRIER() and change the DRM_READMEMORYBARRIER()s that used
to be read/write barriers to it.
it. To do this we need to save the bus address along with the virtual
address in the seglist. Also fix some error handling and a few bits of
whitespace.
irq handler in DRM(irq_install). Modify all drivers to ensure irq
handler is removed before cleanup and cleanup is called at takedown.
Remove unused buffer private struct fields in i810, i830. Check for
lock on init/cleanup in all drivers except i810/i830. The current DDX
for i810 and i830 doesn't hold the lock on kernel init (FIXME?).
DRM(write_string). This is the first part of removing much of the
support code for gamma from the BSD DRM, since it appears that no new
drivers are using it and nobody has ever shown interest in gamma on
BSD.
use correct address for ring read pointer writeback (yes, we seem to have
been running with bogus values for the ring read pointer, which
'worked' because the return value of radeon_wait_ring() is never
checked and the ring usually never fills up)
kernel build system. This is based on suggestions and examples from
David Woodhouse. This approach has the advantage that the build
requirements of a wider range of standard kernels are now supported
transparently, but the disadvantage of some extra complexity to handle
building against clean vendor-distributed kernel source trees. This has
been tested with some recent Red Hat and SuSE distributions.
- Reset 'bound' flag for an agp entry after undbind succeeded in
drm_agpsupport.h (Egbert Eich).
- Ignore hw_lock for drm device if lock was set by a different instance (ie
Xserver) to prevent second server from spinning in driver release
function (currently only relevant for i8xx drm drivers) (David Dawes).
- Use the agpgart "key" for the unique handle for bindings rather than the
memory address (the key is guaranteed to be unique) (David Dawes).
post-drm-filp-0-1-branch world. The filp is a void * cast from the
current pid. This is a temporary solution which maintains the status
quo until a proper solution is implemented.
What is really needed is a unique pointer per open, hopefully with a device
private area. This can be done in FreeBSD for all entry points except
mmap, but is difficult (sys/dev/streams/streams.c is an example). I
have partially completed code for this but have not had time to debug,
so this is a temporary fix.
pageflipping after a mode switch
take current page into account in AdjustFrame(); writing the CRTC offset
via the CP was probably a bad idea as this can happen asynchronously,
reverted
take frame offset into account when flipping pages
handle CRTC2 as well for pageflipping (untested)
preserve GEN_INT_CNTL on mode switches to prevent interrupts from getting
disabled
Whitwell).
- Major rework of the 2D i830/i845G support, including:
- Improve VESA mode selection, and fix refresh rate selection.
- Don't duplicate functions provided in the vbe modules.
- Don't duplicate functions provided in the vgahw module.
- Rewrite memory allocation.
- Rewrite initialisation and save/restore state handling.
- Decouple the i810 support from i830 and later.
- Remove various unnecessary hacks and workarounds.
- Fix an 845G problem with the ring buffer not in pre-allocated memory.
- Fix screen blanking.
- Fix some HW cursor glitches, and turn HW cursor off at VT switch and
exit.
- Don't attempt to use the i830's function 1 entity.
- Fix problems with option handling. (David Dawes).
- Add mode VBE-aware mode handling functions to the vbe module (David
Dawes).
place they're used). Use fd locking on -current. Actually copy in data
from userspace to kernel in the linux-compat ioctl path. Make sure
ioctl sizes are as expected in the ioctl handler functions.
1) Remove redundant header inclusion
2) Silence bitop on non-long argument warnings (change int to long)
3) Move to ISO C (gcc 2.6) initializers (accepted by older gccs also) All
of these are syntax changes that should not impact functionality.
ioctl. The DRM reads them from memory addresses the chip writes to on
updates. Fall back to reading the registers directly with an old DRM.
(Tim Smith, cleanups by myself)
agp chipsets to drm_agpsupport.h, redid the card detection common code
to use a structure (avoids endian porting issues), changed the tdfx
driver to use the kernel pci id '#defines'
be ported, rolled back r128 and i810 version bumps so 4.1.0 works with
cvs kernel modules, added Config.in and updated Makefile.kernel,
incorporated lots of drm fixes inspired by patches sent by Redhat, made
DRM(realloc) usage check for NULL allocations, restructure driver init
routines to export dev_priv only when initialized and to check for all
error conditions.
glide header files.
The changes include:
- Brian Paul's changes to the tdfx client-side 3D driver to make it
dlopen() the correct glide library (Voodoo3 or Voodoo5). This allows
both types of the glide library to co-exist, and allows Voodoo3/Voodoo5
cards to be mixed in multi-head configs.
- DRM kernel driver changes to allow a driver to set up multiple instances
(minor numbers), one for each card present that the driver supports.
This is currently implemented and tested only for the tdfx DRM driver.
- Add some missing missing <stdarg.h> includes.
- Some log message cleanups.
- Change the 2D tdfx driver to access VGA legacy registers via their PCI
I/O space access points rather than their legacy addresses, and fix
some problems with the way the VGA-related bits are initialised.
Status:
- With these changes, multi-head direct rendering works with multiple
Voodoo3 and/or Voodoo5 cards. This has been tested with two PCI Voodoo3
cards and an AGP Voodoo5 card, and all permutations of those.
Caveats:
- Xinerama is not supported. If Xinerama is enabled, then direct rendering
gets disabled.
- The text mode on secondary screens will show junk after the X server
exits.
- On some hardware, starting the X server on multiple 3dfx cards will
result in a hard lockup. One workaround is to enable APIC support in a
uni-processor kernel, or use an SMP kernel.
more flexible and allow older module versions to still work. Might not
be final code if we move over to DRI kernel module version, but will be
if we use XFree86 version. + Only build tdfx.o instead of tdfx-4_1_0.o
since tdfx.o has never changed its API.
- Don't include xf86Priv.h in the drm module.
- Fix 'struct stat' in loader libc wrapping to avoid unnecessary
XFree86LOADER #ifdefs in code that uses it.
- Make drmOpenMinor() static.
dev->agp, when agp_acquire called, and..... Uncomment MUST_HAVE_AGP
around agp_* calls, so that support for both agp/pci can happen in a
single kernel driver.
- Fix old-style DMA for gamma driver (please test).
- Pull out IRQ handling into drm_dma.h (please test on i810, gamma).
- Lots of general cleanups, remove compiler warnings etc.
against 2.2.18pre21
As usual, since all 2.4.0-test* kernels set LINUX_VERSION_CODE to the same
value, if you are running a 2.4.0-test kernel, you MUST be running
2.4.0-test11-pre4 or later (although anything after 2.4.0-test11-pre1
should work fine -- I tested with pre4/pre5). I expect 2.2.x support to
continue to work for all recent kernels, but I tested with 2.2.18pre21
-- we use the old intermodule symbol communication for 2.2.x kernels,
so they should all continue to work.
several fixes including: (Jeff) Really disable multitexture (Broken
since Mesa 3.4 integration.)
(Jeff) Various changes in mga_state.c in the kernel module, which includes
a fix to the bug where the first OGL application hangs the machine
w/out rendering anything. This also includes defines for the warp
registers so they are easily human readable.
(Rik and Jeff) Fixed all the schedule loops in the kernel to look like they
are supposed too.
(Jeff) Configurable agp modes: Add the option "AGPMode2x" or "AGPMode4x" to
your XF86Config file.
(Rik) Various cleanups to the mga kernel driver to make it easier to read
and debug.
(Rik) Removed alot of DRM_DEBUG statements from the kernel driver.
- New security model for i810
- Enable i810 dri by default
- New indexed vertex path for mga
- Mga kernel driver rework
- Removed dead files in i810 driver