We were mapping an extra byte (and hence usually an extra page):
iommu_prepare_identity_map() expects to be given an 'end' argument which
is the last byte to be mapped; not the first byte *not* to be mapped.
Signed-off-by: David Woodhouse <[email protected]>
In commit 2a48fc0ab2 ('block: autoconvert
trivial BKL users to private mutex'), Arnd replaced the BKL usage with a
mutex. However, Maxim has already provided a better fix in commit
480792b7bf ('mtd: blktrans: kill BKL'),
which was simply to remove the BKL without replacing it — since he'd
already made it do all necessary locking for itself.
Signed-off-by: David Woodhouse <[email protected]>
Conflicts:
drivers/mtd/mtd_blkdevs.c
Merge Grant's device-tree bits so that we can apply the subsequent fixes.
Signed-off-by: David Woodhouse <[email protected]>
This reverts commit 432dc821c9.
The individual CFI geometry options were carefully set up to get sane
default values if the CFI_ADV_OPTIONS wasn't set, and it wasn't
appropriate to move them into an if/endif block.
Signed-off-by: David Woodhouse <[email protected]>
If i915 opregion is present, the acpi_video driver doesn't register
itself immediately; it defers that until the i915 opregion code is done.
But if that *fails*, the acpi_video driver was never getting registered.
And thus I have no backlight support on my Lenovo IdeaPad S10-3.
Call acpi_video_register() on the failure path, and it works again.
Signed-off-by: David Woodhouse <[email protected]>
Acked-by: Jesse Barnes <[email protected]>
Acked-by: Matthew Garrett <[email protected]>
drivers/pci/intel-iommu.c: In function 'dma_pte_addr':
drivers/pci/intel-iommu.c:239: warning: passing argument 1 of '__cmpxchg64' from incompatible pointer type
It seems that __cmpxchg64() now cares about the type of its pointer argument,
so give it a (uint64_t *) instead of a pointer to a structure which contains
only that.
Signed-off-by: David Woodhouse <[email protected]>
drivers/power/olpc_battery.c:387: error: unknown field 'owner' specified in initializer
drivers/power/olpc_battery.c:387: warning: excess elements in struct initializer
drivers/power/olpc_battery.c:387: warning: (near initialization for 'olpc_bat_eeprom.attr')
The .owner field has been dropped from the attr struct.
Signed-off-by: David Woodhouse <[email protected]>
We were seeing faults in the solos-pci receive tasklet when packets
arrived for a VCC which was currently being closed:
[18842.727906] EIP: [<e082f490>] br2684_push+0x19/0x234 [br2684] SS:ESP 0068:dfb89d14
[18845.090712] [<c13ecff3>] ? do_page_fault+0x0/0x2e1
[18845.120042] [<e082f490>] ? br2684_push+0x19/0x234 [br2684]
[18845.153530] [<e084fa13>] solos_bh+0x28b/0x7c8 [solos_pci]
[18845.186488] [<e084f711>] ? solos_irq+0x2d/0x51 [solos_pci]
[18845.219960] [<c100387b>] ? handle_irq+0x3b/0x48
[18845.247732] [<c10265cb>] ? irq_exit+0x34/0x57
[18845.274437] [<c1025720>] tasklet_action+0x42/0x69
[18845.303247] [<c102643f>] __do_softirq+0x8e/0x129
[18845.331540] [<c10264ff>] do_softirq+0x25/0x2a
[18845.358274] [<c102664c>] _local_bh_enable_ip+0x5e/0x6a
[18845.389677] [<c102666d>] local_bh_enable+0xb/0xe
[18845.417944] [<e08490a8>] ppp_unregister_channel+0x32/0xbb [ppp_generic]
[18845.458193] [<e08731ad>] pppox_unbind_sock+0x18/0x1f [pppox]
This patch uses an RCU-inspired approach to fix it. In the RX tasklet's
find_vcc() function we first refuse to use a VCC which already has the
ATM_VF_READY bit cleared. And in the VCC close function, we synchronise
with the tasklet to ensure that it can't still be using the VCC before
we continue and allow the VCC to be destroyed.
Signed-off-by: David Woodhouse <[email protected]>
Tested-by: Nathan Williams <[email protected]>
Cc: [email protected]
Signed-off-by: David S. Miller <[email protected]>
Certain revisions of this chipset appear to be broken. There is a shadow
GTT which mirrors the real GTT but contains pre-translated physical
addresses, for performance reasons. When a GTT update happens, the
translations are done once and the resulting physical addresses written
back to the shadow GTT.
Except sometimes, the physical address is actually written back to the
_real_ GTT, not the shadow GTT. Thus we start to see faults when that
physical address is fed through translation again.
Signed-off-by: David Woodhouse <[email protected]>
Only build the parts which are actually needed, and turn the DMA support
(which is currently hard-coded off) into a config option (still forced off).
Signed-off-by: David Woodhouse <[email protected]>
... by killing the gratuitous 'res_nand' early_param() and turning it
into a module_param() called 'reserved_mb' instead.
Signed-off-by: David Woodhouse <[email protected]>
New inodes need to be locked as we're creating them, so they don't get used
by other things (like NFSd) before they're ready.
Pointed out by Al Viro.
Signed-off-by: David Woodhouse <[email protected]>
fix memory leak introduced by the patch 6e03a201bb:
firmware: speed up request_firmware()
1. vfree won't release pages there were allocated explicitly and mapped
using vmap. The memory has to be vunmap-ed and the pages needs
to be freed explicitly
2. page array is moved into the 'struct
firmware' so that we can free it from release_firmware()
and not only in fw_dev_release()
The fix doesn't break the firmware load speed.
Cc: Johannes Berg <[email protected]>
Cc: Ming Lei <[email protected]>
Cc: Catalin Marinas <[email protected]>
Singed-off-by: Kay Sievers <[email protected]>
Signed-off-by: David Woodhouse <[email protected]>
Signed-off-by: Tomas Winkler <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>
This reverts commit 66803762 ("mtd: mxc_nand: add RESET command support").
Support for NAND_CMD_RESET was added separately in commit d4840180
("mtd: mxc_nand: set NFC registers after reset"), causing a build error:
drivers/mtd/nand/mxc_nand.c: In function 'mxc_nand_command':
drivers/mtd/nand/mxc_nand.c:689: error: duplicate case value
drivers/mtd/nand/mxc_nand.c:606: error: previously used here
Signed-off-by: David Woodhouse <[email protected]>
We're about to call this from a bunch of places which already hold
c->erase_completion_lock, so add an assertion and change its existing
callers to do the same.
Signed-off-by: David Woodhouse <[email protected]>
jffs2_garbage_collect_pass() would previously return -EAGAIN if it
couldn't find anything to garbage collect from, and there were blocks on
the erase_pending_list. If the blocks were actually in the process of
being erased, though, then they wouldn't be on that list. Check for
nr_erasing_blocks being non-zero instead.
Fix jffs2_reserve_space() to wait for the in-progress erases to
complete, when jffs2_garbage_collect_pass() returns -EAGAIN.
And fix jffs2_erase_succeeded() to actually wake up the erase_wait wq
that jffs2_reserve_space() is now using.
Signed-off-by: David Woodhouse <[email protected]>
Doh. Pointed out by Guillaume LECERF <[email protected]> since I managed
to miss it in my test builds. S'what I get for hacking at 2am, I suppose.
Signed-off-by: David Woodhouse <[email protected]>
Accept values of 2-5 for TopBottom, where the newly-added 4 and 5 values
mean a uniform layout. It does indicate WP layout but we don't handle that.
Also don't say "broken" when swapping erase regions in a top-boot chip.
That got retrospectively documented in the spec.
Signed-off-by: David Woodhouse <[email protected]>
drivers/mtd/nand/denali.c:1427: error: conflicting types for ‘enable_dma’
arch/powerpc/include/asm/dma.h:189: note: previous definition of ‘enable_dma’ was here
Signed-off-by: David Woodhouse <[email protected]>
... instead of comparing with DMA_ERROR_CODE, which will only work on
powerpc/sparc/x86.
Reported-by: Geert Uytterhoeven <[email protected]>
Signed-off-by: David Woodhouse <[email protected]>
Randy Dunlap observed a build problem with the following config:
CONFIG_SM_FTL=y
CONFIG_SM_FTL_MUSEUM=y
CONFIG_MTD_NAND=m
The ECC calculation routines are then built as a module, but referenced
by the sm_ftl code in the kernel, resulting in a build failure. The
simple fix is to make CONFIG_SM_FTL depend on MTD_NAND unconditionally
-- it's pointless without hardware support anyway.
Fix some typos which Randy pointed out, too.
Reported-By: Randy Dunlap <[email protected]>
Signed-off-by: David Woodhouse <[email protected]>
(and remove the CONFIG_MTD_NAND_SMARTMEDIA option which isn't going to be
used now that we're doing it this way)
Signed-off-by: David Woodhouse <[email protected]>
Ever since jffs2_garbage_collect_metadata() was first half-written in
February 2001, it's been broken on architectures where 'char' is signed.
When garbage collecting a symlink with target length above 127, the payload
length would end up negative, causing interesting and bad things to happen.
Cc: [email protected]
Signed-off-by: David Woodhouse <[email protected]>
The hotplug notifier will call find_domain() to see if the device in
question has been assigned an IOMMU domain. However, this should never
be called for devices with a "dummy" domain, such as graphics devices
when intel_iommu=igfx_off is set and the corresponding IOMMU isn't even
initialised. If you do that, it'll oops as it dereferences the (-1)
pointer.
The notifier function should check iommu_no_mapping() for the
device before doing anything else.
Cc: [email protected]
Signed-off-by: David Woodhouse <[email protected]>
Some HP BIOSes report an RMRR region (a region which needs a 1:1 mapping
in the IOMMU for a given device) which has an end address lower than its
start address. Detect that and warn, rather than triggering the
BUG() in dma_pte_clear_range().
Cc: [email protected]
Signed-off-by: David Woodhouse <[email protected]>
The BIOS errors where an IOMMU is reported either at zero or a bogus
address are causing problems even when the IOMMU is disabled -- because
interrupt remapping uses the same hardware. Ensure that the checks get
applied for the interrupt remapping initialisation too.
Cc: [email protected]
Signed-off-by: David Woodhouse <[email protected]>
Chris Wright has some patches which let us fall back to swiotlb nicely
if IOMMU initialisation fails. But those are a bit much for 2.6.32.
Instead, let's shift the check for the biggest problem, the HP and Acer
BIOS bug which reports a DMAR at physical address zero. That one can
actually be checked much earlier -- before we even admit to having
detected an IOMMU in the first place. So the swiotlb init goes ahead as
we want.
Signed-off-by: David Woodhouse <[email protected]>
This reverts commit 71b7d0d90d.
The problem which that commit attempted to fix was a bootloader issue,
which had been misunderstood. The 'fix' causes lots of false bad blocks
for existing users with sane firmware.
Thanks to Mathieu Berland for diagnosing the problem coherently.
Signed-off-by: David Woodhouse <[email protected]>
We want this to happen after the PCI quirks, which are now running at
the very end of the fs_initcalls.
This works around the BIOS problems which were originally addressed by
commit db8be50c43 ('USB: Work around BIOS
bugs by quiescing USB controllers earlier'), which was reverted in
commit d93a8f829f.
Signed-off-by: David Woodhouse <[email protected]>
Having this as a device_initcall() means that some real device drivers
can actually initialise _before_ the quirks are run, which is wrong.
We want it to run _before_ device_initcall(), but _after_ fs_initcall(),
since some arch-specific PCI initialisation like pcibios_assign_resources()
is done at fs_initcall().
We could use rootfs_initcall() but I actually want to use that for the
IOMMU initialisation, which has to come after the quirks, but still
before the real devices. So use fs_initcall_sync() instead -- since this
is entirely synchronous, it doesn't hurt that it'll escape the
synchronisation.
Signed-off-by: David Woodhouse <[email protected]>
This function may have done more in the past, but all it does now is
apply the PCI_FIXUP_FINAL quirks. So name it sensibly and put it where
it belongs.
Signed-off-by: David Woodhouse <[email protected]>
Asus decided to ship a BIOS which configures sound DMA to go via the
dedicated IOMMU unit, but assigns precisely zero TLB entries to that
unit. Which causes the whole thing to deadlock, including the DMA
traffic on the _other_ IOMMU units. Nice one.
Signed-off-by: David Woodhouse <[email protected]>
We may end up doing DMA to/from these. Until the new MTD API fixes the
issues, this should stop things from falling over.
Original idea from Gilles Casse <[email protected]>
Signed-off-by: David Woodhouse <[email protected]>
As with orion_nand in commit f33dabbe79
("register orion_nand using platform_driver_probe()"), avoid .init.text
problems by using platform_device_probe(). This isn't going to be
hotplugged anyway.
Reported-by: Uwe Kleine-König <[email protected]>
Signed-off-by: David Woodhouse <[email protected]>
Just make it depend on BROKEN for now, in case people scream really loud
about it (and because we might want to keep some of this logic for an
upcoming BIOS workaround, so I don't just want to rip it out entirely
just yet). But for graphics devices, it really ought to be unnecessary.
Signed-off-by: David Woodhouse <[email protected]>
If end_pfn is equal to (unsigned long)-1, then the loop will never end.
Seen on 32-bit kernel, but could have happened on 64-bit too once we get
hardware that supports 64-bit guest addresses.
Change both functions to a 'do {} while' loop with the test at the end,
and check for the PFN having wrapper round to zero.
Reported-by: Benjamin LaHaise <[email protected]>
Tested-by: Benjamin LaHaise <[email protected]>
Signed-off-by: David Woodhouse <[email protected]>
This means we're limited to 44-bit addresses on 32-bit kernels, and
makes it sane for us to use 'unsigned long' for PFNs throughout.
Which is just as well, really, since we already do that.
Reported-by: Benjamin LaHaise <[email protected]>
Tested-by: Benjamin LaHaise <[email protected]>
Signed-off-by: David Woodhouse <[email protected]>
Since commit 19943b0e30 ('intel-iommu:
Unify hardware and software passthrough support'), hardware passthrough
mode will do the same as software passthrough mode was doing -- it'll
still use the IOMMU normally for devices which can't address all of
memory. This means that we don't need to bother with swiotlb.
Signed-off-by: David Woodhouse <[email protected]>
All callers of the former were also calling the latter, in one order or
the other, and failing to correctly clean up if the second returned
failure.
Signed-off-by: David Woodhouse <[email protected]>
If DMAR is configured in but absent, we really do want to make sure that
the dma mask is set appropriately. Otherwise we get mapping failures on
highmem. Spotted by Zhenyu Wang.
Signed-off-by: David Woodhouse <[email protected]>
This makes the hardware passthrough mode work a lot more like the
software version, so that the behaviour of a kernel with 'iommu=pt'
is the same whether the hardware supports passthrough or not.
In particular:
- We use a single si_domain for the pass-through devices.
- 32-bit devices can be taken out of the pass-through domain so that
they don't have to use swiotlb.
- Devices will work again after being removed from a KVM guest.
- A potential oops on OOM (in init_context_pass_through()) is fixed.
Signed-off-by: David Woodhouse <[email protected]>
Yet another reason why trusting this stuff to the BIOS was a bad idea.
The HP DC7900 BIOS reports an iommu at an address which just returns all
ones, when VT-d is disabled in the BIOS.
Fix up the missing iounmap in the error paths while we're at it.
Signed-off-by: David Woodhouse <[email protected]>
In commit 07613ba2 ("agp: switch AGP to use page array instead of
unsigned long array") we switched the mask_memory() method to take a
'struct page *' instead of an address. This is painful, because in some
cases it has to be an IOMMU-mapped virtual bus address (in fact,
shouldn't it _always_ be a dma_addr_t returned from pci_map_xxx(), and
we just happen to get lucky most of the time?)
Signed-off-by: David Woodhouse <[email protected]>
We only ever obtain this lock immediately before the iova_rbtree_lock,
and release it immediately after the iova_rbtree_lock. So ditch it and
just use iova_rbtree_lock.
[v2: Remove the lockdep bits this time too]
Signed-off-by: David Woodhouse <[email protected]>
We did before, in the end -- but it was at the bottom of a long stack of
functions. Add an inline wrapper get_valid_domain_for_dev() which will
use the cached one _first_ and only make the out-of-line call if it's
not already set.
This takes the average time taken for a 1-page intel_map_sg() from 5961
cycles to 4812 cycles on my Lenovo x200s test box -- a modest 20%.
Signed-off-by: David Woodhouse <[email protected]>
Our current strategy for pass-through mode is to put all devices into
the 1:1 domain at startup (which is before we know what their dma_mask
will be), and only _later_ take them out of that domain, if it turns out
that they really can't address all of memory.
However, when there are a bunch of PCI devices behind a bridge, they all
end up with the same source-id on their DMA transactions, and hence in
the same IOMMU domain. This means that we _can't_ easily move them from
the 1:1 domain into their own domain at runtime, because there might be DMA
in-flight from their siblings.
So we have to adjust our pass-through strategy: For PCI devices not on
the root bus, and for the bridges which will take responsibility for
their transactions, we have to start up _out_ of the 1:1 domain, just in
case.
This fixes the BUG() we see when we have 32-bit-capable devices behind a
PCI-PCI bridge, and use the software identity mapping.
It does mean that we might end up using 'normal' mapping mode for some
devices which could actually live with the faster 1:1 mapping -- but
this is only for PCI devices behind bridges, which presumably aren't the
devices for which people are most concerned about performance.
Signed-off-by: David Woodhouse <[email protected]>
At boot time, the dma_mask won't have been set on any devices, so we
assume that all devices will be 64-bit capable (and thus get a 1:1 map).
Signed-off-by: David Woodhouse <[email protected]>
This should fix kernel.org bug #11821, where the dcdbas driver makes up
a platform device and then uses dma_alloc_coherent() on it, in an
attempt to get memory < 4GiB.
Signed-off-by: David Woodhouse <[email protected]>
We need to give people a little more time to fix the broken drivers.
Re-introduce this, but tied in properly with the 'iommu=pt' support this
time. Change the config option name and make it default to 'no' too.
Signed-off-by: David Woodhouse <[email protected]>
We do this twice, and it's about to get more complicated. This makes the
code slightly clearer about what it's doing, too.
Signed-off-by: David Woodhouse <[email protected]>
When we reattach a device to the si_domain (because it's been removed
from a VM), we weren't calling domain_context_mapping() to actually tell
the hardware about that.
We should really put the call to domain_context_mapping() into
domain_add_dev_info() -- we never call the latter without also doing the
former, and we can keep the error paths simple that way. But that's a
cleanup which can wait for 2.6.32 now.
Signed-off-by: David Woodhouse <[email protected]>
We should check iommu_dummy() _first_, because that means it's attached
to an iommu that we've just disabled completely. At the moment, we might
try to put the device into the identity mapping domain.
Signed-off-by: David Woodhouse <[email protected]>
The aligned_nrpages() function rounds up to the next VM page, but
returns its result as a number of DMA pages.
Purely theoretical except on IA64, which doesn't boot with VT-d right
now anyway.
Signed-off-by: David Woodhouse <[email protected]>
Check dma_pte_present() and only free the page if there _is_ one.
Kind of surprising that there was no warning about this.
Signed-off-by: David Woodhouse <[email protected]>
On Wed, 2009-07-01 at 16:59 -0700, Linus Torvalds wrote:
> I also _really_ hate how you do
>
> (unsigned long)pte >> VTD_PAGE_SHIFT ==
> (unsigned long)first_pte >> VTD_PAGE_SHIFT
Kill this, in favour of just looking to see if the incremented pte
pointer has 'wrapped' onto the next page. Which means we have to check
it _after_ incrementing it, not before.
Signed-off-by: David Woodhouse <[email protected]>
This would have found the bug in i386 pci_unmap_addr() a long time ago.
We shouldn't just silently return without doing anything.
Signed-off-by: David Woodhouse <[email protected]>
Since we're using cmpxchg64() anyway (because that's the only way to do
an atomic 64-bit store on i386), we might as well ditch the extra
locking and just use cmpxchg64() to ensure that we don't add the page
twice.
Signed-off-by: David Woodhouse <[email protected]>
David Woodhouse