A negative return value indicates an error; in fact the function at
present won't ever return zero.
Signed-off-by: Jan Beulich <[email protected]>
Reviewed-by: Juergen Gross <[email protected]>
Signed-off-by: Boris Ostrovsky <[email protected]>
Making use of "max_indirect_segments=" has issues:
- blkfront_setup_indirect() may end up with zero psegs when PAGE_SIZE
is sufficiently much larger than XEN_PAGE_SIZE
- the variable driven by the command line option
(xen_blkif_max_segments) has a somewhat different purpose, and hence
should namely never end up being zero
- as long as the specified value is lower than the legacy default,
we better don't use indirect segments at all (or we'd in fact lower
throughput)
Signed-off-by: Jan Beulich <[email protected]>
Signed-off-by: Konrad Rzeszutek Wilk <[email protected]>
Don't truncate the "feature-persistent" value read from xenstore: Any
non-zero value is supposed to enable the feature, just like is already
being done for feature_secdiscard.
Just like the other feature_* fields, feature_flush and feature_fua are
boolean flags, and hence fit well into a single bit.
Keep all bit fields together to limit gaps.
Signed-off-by: Jan Beulich <[email protected]>
Signed-off-by: Konrad Rzeszutek Wilk <[email protected]>
For single items being collected this should be preferred as being more
typesafe (as the compiler can check format string and to-be-written-to
variable match) and more efficient (requiring one less parameter to be
passed).
Signed-off-by: Jan Beulich <[email protected]>
Reviewed-by: Paul Durrant <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
If msi_setup_entry() fails to allocate an affinity mask, it logs a message
but continues on and allocates an MSI entry with entry->affinity == NULL.
Check for this case in pci_irq_get_affinity() so we don't try to
dereference a NULL pointer.
[bhelgaas: changelog]
Fixes: ee8d41e53e "pci/msi: Retrieve affinity for a vector"
Signed-off-by: Jan Beulich <[email protected]>
Signed-off-by: Bjorn Helgaas <[email protected]>
Reviewed-by: Christoph Hellwig <[email protected]>
CC: Thomas Gleixner <[email protected]>
Don't ignore errors here: Set backend state to unknown when
unsuccessful.
Signed-off-by: Jan Beulich <[email protected]>
Signed-off-by: David Vrabel <[email protected]>
This is more efficient than list_for_each_safe() when list modification
is accompanied by breaking out of the loop.
Signed-off-by: Jan Beulich <[email protected]>
Reviewed-by: Juergen Gross <[email protected]>
Signed-off-by: David Vrabel <[email protected]>
With the addition of uses of GCC's condition code outputs in commit:
35ccfb7114 ("x86, asm: Use CC_SET()/CC_OUT() in <asm/rwsem.h>")
... there's now an overlap of outputs and clobbers in __down_write_trylock().
Such overlaps are generally getting tagged with an error (occasionally
even with an ICE). I can't really tell why plain GCC 6.2 doesn't detect
this (judging by the code it is meant to), while the slightly modified
one I use does. Since condition code clobbers are never necessary on x86
(other than perhaps for documentation purposes, which doesn't really
get done consistently), remove it altogether rather than inventing
something like CC_CLOBBER (to accompany CC_SET/CC_OUT).
Signed-off-by: Jan Beulich <[email protected]>
Cc: Andrew Morton <[email protected]>
Cc: Andy Lutomirski <[email protected]>
Cc: Borislav Petkov <[email protected]>
Cc: Brian Gerst <[email protected]>
Cc: Denys Vlasenko <[email protected]>
Cc: H. Peter Anvin <[email protected]>
Cc: Josh Poimboeuf <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Paul E. McKenney <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
While commit 55f1ea1521 ("efi: Fix for_each_efi_memory_desc_in_map()
for empty memmaps") made an attempt to deal with empty memory maps, it
didn't address the case where the map field never gets set, as is
apparently the case when running under Xen.
Reported-by: <[email protected]>
Tested-by: <[email protected]>
Cc: Vitaly Kuznetsov <[email protected]>
Cc: Jiri Slaby <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: <[email protected]> # v4.7+
Signed-off-by: Jan Beulich <[email protected]>
[ Guard the loop with a NULL check instead of pointer underflow ]
Signed-off-by: Matt Fleming <[email protected]>
This should really only be done for XS_TRANSACTION_END messages, or
else at least some of the xenstore-* tools don't work anymore.
Fixes: 0beef634b8 ("xenbus: don't BUG() on user mode induced condition")
Reported-by: Richard Schütz <[email protected]>
Cc: <[email protected]>
Signed-off-by: Jan Beulich <[email protected]>
Tested-by: Richard Schütz <[email protected]>
Signed-off-by: David Vrabel <[email protected]>
The functions these get passed to have been taking pointers to const
since at least 2.6.16.
Acked-by: Jens Axboe <[email protected]>
Acked-by: Roger Pau Monné <[email protected]>
Signed-off-by: Jan Beulich <[email protected]>
Signed-off-by: Konrad Rzeszutek Wilk <[email protected]>
... for single items being collected: It is more typesafe (as the
compiler can check format string and to-be-written-to variable match)
and requires one less parameter to be passed.
Acked-by: Roger Pau Monné <[email protected]>
Acked-by: Jens Axboe <[email protected]>
Signed-off-by: Jan Beulich <[email protected]>
Signed-off-by: Konrad Rzeszutek Wilk <[email protected]>
... for single items being collected: It is more typesafe (as the
compiler can check format string and to-be-written-to variable match)
and requires one less parameter to be passed.
Signed-off-by: Jan Beulich <[email protected]>
Signed-off-by: Konrad Rzeszutek Wilk <[email protected]>
Acked-by: Roger Pau Monné <[email protected]>
Acked-by: Jens Axboe <[email protected]>
Commit 5dcef69486 ("intel_idle: add BXT support") added an 8-element
lookup array with just a 2-bit value used for lookups. As per the SDM
that bit field is really 3 bits wide. While this is supposedly benign
here, future re-use of the code for other CPUs might expose the issue.
Signed-off-by: Jan Beulich <[email protected]>
Signed-off-by: Rafael J. Wysocki <[email protected]>
Since irtl_ns_units[] has itself zero entries, make sure the caller
recognized those cases along with the MSR read returning zero, as zero
is not a valid value for exit_latency and target_residency.
Signed-off-by: Jan Beulich <[email protected]>
Signed-off-by: Rafael J. Wysocki <[email protected]>
As of Xen 4.7 PV CPUID doesn't expose either of CPUID[1].ECX[7] and
CPUID[0x80000007].EDX[7] anymore, causing the driver to fail to load on
both Intel and AMD systems. Doing any kind of hardware capability
checks in the driver as a prerequisite was wrong anyway: With the
hypervisor being in charge, all such checking should be done by it. If
ACPI data gets uploaded despite some missing capability, the hypervisor
is free to ignore part or all of that data.
Ditch the entire check_prereq() function, and do the only valid check
(xen_initial_domain()) in the caller in its place.
Signed-off-by: Jan Beulich <[email protected]>
Cc: <[email protected]>
Signed-off-by: David Vrabel <[email protected]>
No need to retain a local copy of the full request message, only the
type is really needed.
Signed-off-by: Jan Beulich <[email protected]>
Signed-off-by: David Vrabel <[email protected]>
xenbus_dev_request_and_reply() needs to track whether a transaction is
open. For XS_TRANSACTION_START messages it calls transaction_start()
and for XS_TRANSACTION_END messages it calls transaction_end().
If sending an XS_TRANSACTION_START message fails or responds with an
an error, the transaction is not open and transaction_end() must be
called.
If sending an XS_TRANSACTION_END message fails, the transaction is
still open, but if an error response is returned the transaction is
closed.
Commit 027bd7e899 ("xen/xenbus: Avoid synchronous wait on XenBus
stalling shutdown/restart") introduced a regression where failed
XS_TRANSACTION_START messages were leaving the transaction open. This
can cause problems with suspend (and migration) as all transactions
must be closed before suspending.
It appears that the problematic change was added accidentally, so just
remove it.
Signed-off-by: Jan Beulich <[email protected]>
Cc: Konrad Rzeszutek Wilk <[email protected]>
Cc: <[email protected]>
Signed-off-by: David Vrabel <[email protected]>
Inability to locate a user mode specified transaction ID should not
lead to a kernel crash. For other than XS_TRANSACTION_START also
don't issue anything to xenbus if the specified ID doesn't match that
of any active transaction.
Signed-off-by: Jan Beulich <[email protected]>
Cc: <[email protected]>
Signed-off-by: David Vrabel <[email protected]>
req_start is simply an alias of the "offset" function parameter, and
req_end is being used just once in each function. (And both variables
were loop invariant anyway, so should at least have got initialized
outside the loop.)
Signed-off-by: Jan Beulich <[email protected]>
Signed-off-by: David Vrabel <[email protected]>
There's no point calling xen_pcibk_config_read() here - all it'll do is
return whatever conf_space_read() returns for the field which was found
here (and which would be found there again). Also there's no point
clearing tmp_val before the call.
Signed-off-by: Jan Beulich <[email protected]>
Signed-off-by: David Vrabel <[email protected]>
Other than for raw BAR values, flags are properly separated in the
internal representation.
Signed-off-by: Jan Beulich <[email protected]>
Signed-off-by: David Vrabel <[email protected]>
Reads following writes with all address bits set to 1 should return all
changeable address bits as one, not the BAR size (nor, as was the case
for the upper half of 64-bit BARs, the high half of the region's end
address). Presumably this didn't cause any problems so far because
consumers use the value to calculate the size (usually via val & -val),
and do nothing else with it.
But also consider the exception here: Unimplemented BARs should always
return all zeroes.
And finally, the check for whether to return the sizing address on read
for the ROM BAR should ignore all non-address bits, not just the ROM
Enable one.
Signed-off-by: Jan Beulich <[email protected]>
Reviewed-by: Boris Ostrovsky <[email protected]>
Signed-off-by: David Vrabel <[email protected]>
The copying of ring data was wrong for two cases: For a full ring
nothing got copied at all (as in that case the canonicalized producer
and consumer indexes are identical). And in case one or both of the
canonicalized (after the resize) indexes would point into the second
half of the buffer, the copied data ended up in the wrong (free) part
of the new buffer. In both cases uninitialized data would get passed
back to the caller.
Fix this by simply copying the old ring contents twice: Once to the
low half of the new buffer, and a second time to the high half.
This addresses the inability to boot a HVM guest with 64 or more
vCPUs. This regression was caused by 8620015499 (xen/evtchn:
dynamically grow pending event channel ring).
Reported-by: Konrad Rzeszutek Wilk <[email protected]>
Signed-off-by: Jan Beulich <[email protected]>
Cc: <[email protected]> # 4.4+
Signed-off-by: David Vrabel <[email protected]>
Huge pages are not normally available to PV guests. Not suppressing
hugetlbfs use results in an endless loop of page faults when user mode
code tries to access a hugetlbfs mapped area (since the hypervisor
denies such PTEs to be created, but error indications can't be
propagated out of xen_set_pte_at(), just like for various of its
siblings), and - once killed in an oops like this:
kernel BUG at .../fs/hugetlbfs/inode.c:428!
invalid opcode: 0000 [#1] SMP
...
RIP: e030:[<ffffffff811c333b>] [<ffffffff811c333b>] remove_inode_hugepages+0x25b/0x320
...
Call Trace:
[<ffffffff811c3415>] hugetlbfs_evict_inode+0x15/0x40
[<ffffffff81167b3d>] evict+0xbd/0x1b0
[<ffffffff8116514a>] __dentry_kill+0x19a/0x1f0
[<ffffffff81165b0e>] dput+0x1fe/0x220
[<ffffffff81150535>] __fput+0x155/0x200
[<ffffffff81079fc0>] task_work_run+0x60/0xa0
[<ffffffff81063510>] do_exit+0x160/0x400
[<ffffffff810637eb>] do_group_exit+0x3b/0xa0
[<ffffffff8106e8bd>] get_signal+0x1ed/0x470
[<ffffffff8100f854>] do_signal+0x14/0x110
[<ffffffff810030e9>] prepare_exit_to_usermode+0xe9/0xf0
[<ffffffff814178a5>] retint_user+0x8/0x13
This is CVE-2016-3961 / XSA-174.
Reported-by: Vitaly Kuznetsov <[email protected]>
Signed-off-by: Jan Beulich <[email protected]>
Cc: Andrew Morton <[email protected]>
Cc: Andy Lutomirski <[email protected]>
Cc: Boris Ostrovsky <[email protected]>
Cc: Borislav Petkov <[email protected]>
Cc: Brian Gerst <[email protected]>
Cc: David Vrabel <[email protected]>
Cc: Denys Vlasenko <[email protected]>
Cc: H. Peter Anvin <[email protected]>
Cc: Juergen Gross <[email protected]>
Cc: Linus Torvalds <[email protected]>
Cc: Luis R. Rodriguez <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Toshi Kani <[email protected]>
Cc: [email protected]
Cc: xen-devel <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
There's no reason to defer this until the connect phase, and in fact
there are frontend implementations expecting this to be available
earlier. Move it into the probe function.
Acked-by: Roger Pau Monné <[email protected]>
Signed-off-by: Jan Beulich <[email protected]>
Cc: Bob Liu <[email protected]>
Signed-off-by: Konrad Rzeszutek Wilk <[email protected]>
"max" is rather ambiguous and carries pretty little meaning, the more
that there are also "max_queues" and "max_ring_page_order". Make this
"max_indirect_segments" instead, and at once change the type from int
to uint (to match the respective variable's type).
Acked-by: Roger Pau Monné <[email protected]>
Signed-off-by: Jan Beulich <[email protected]>
Signed-off-by: Konrad Rzeszutek Wilk <[email protected]>
set_memory_nx() (and set_memory_x()) currently differ in behavior from
all other set_memory_*() functions when encountering a virtual address
space hole within the kernel address range: They stop processing at the
hole, but nevertheless report success (making the caller believe the
operation was carried out on the entire range). While observed to be a
problem - triggering the CONFIG_DEBUG_WX warning - only with out of
tree code, I suspect (but didn't check) that on x86-64 the
CONFIG_DEBUG_PAGEALLOC logic in free_init_pages() would, when called
from free_initmem(), have the same effect on the set_memory_nx() called
from mark_rodata_ro().
This unexpected behavior is a result of change_page_attr_set_clr()
special casing changes to only the NX bit, in that it passes "false" as
the "checkalias" argument to __change_page_attr_set_clr(). Since this
flag becomes the "primary" argument of both __change_page_attr() and
__cpa_process_fault(), the latter would so far return success without
adjusting cpa->numpages. Success to the higher level callers, however,
means that whatever cpa->numpages currently holds is the count of
successfully processed pages. The cases when __change_page_attr() calls
__cpa_process_fault(), otoh, don't generally mean the entire range got
processed (as can be seen from one of the two success return paths in
__cpa_process_fault() already adjusting ->numpages).
Signed-off-by: Jan Beulich <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Thomas Gleixner <[email protected]>
Considering that pci_pericom_setup(()'s only difference to
pci_default_setup() is the setting of the uartclk field, and taking
into account that this field already gets taken care of by having the
base_baud field filled in the pci_boards[] entries, there's no need
for both the function and the quirks table entry.
Signed-off-by: Jan Beulich <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>
Documentation/Changes still lists this as the minimal required version,
so it ought to remain usable for the time being.
Fixes: d2036f30cf ("scripts/kconfig/Makefile: Allow KBUILD_DEFCONFIG to be a target")
Signed-off-by: Jan Beulich <[email protected]>
Cc: Michael Ellerman <[email protected]>
Signed-off-by: Michal Marek <[email protected]>
For PAE kernels "unsigned long" is not suitable to hold page protection
flags, since _PAGE_NX doesn't fit there. This is the reason for quite a
few W+X pages getting reported as insecure during boot (observed namely
for the entire initrd range).
Fixes: 281d4078be ("x86: Make page cache mode a real type")
Signed-off-by: Jan Beulich <[email protected]>
Reviewed-by: Juergen Gross <[email protected]>
Cc: [email protected]
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Thomas Gleixner <[email protected]>
The combined effect of commits 6423fc3416 ("igb: do not re-init SR-IOV
during probe") and ceee3450b3 ("igb: make sure SR-IOV init uses the
right number of queues") causes VFs no longer getting set up, leading
to NULL pointer dereferences due to the adapter's ->vf_data being NULL
while ->vfs_allocated_count is non-zero. The first commit not only
neglected the side effect of igb_sriov_reinit() that the second commit
tried to account for, but also that of setting IGB_FLAG_HAS_MSIX,
without which igb_enable_sriov() is effectively a no-op. Calling
igb_{,re}set_interrupt_capability() as done here seems to address this,
but I'm not sure whether this is better than sinply reverting the other
two commits.
Signed-off-by: Jan Beulich <[email protected]>
Tested-by: Aaron Brown <[email protected]>
Signed-off-by: Jeff Kirsher <[email protected]>
While in most cases commit b1d9e6b064 ("LSM: Switch to lists of hooks")
retained previous error returns, in three cases it altered them without
any explanation in the commit message. Restore all of them - in the
security_old_inode_init_security() case this led to reiserfs using
uninitialized data, sooner or later crashing the system (the only other
user of this function - ocfs2 - was unaffected afaict, since it passes
pre-initialized structures).
Signed-off-by: Jan Beulich <[email protected]>
Signed-off-by: Casey Schaufler <[email protected]>
Signed-off-by: James Morris <[email protected]>
Over the years I found it desirable to be able to use all sorts of
relations, not just (in)equality. And apparently I'm not the only one,
as there's at least one example in the tree where the programmer
assumed this would work (see DEBUG_UART_8250_WORD in
arch/arm/Kconfig.debug). Another possible use would e.g. be to fold the
two SMP/NR_CPUS prompts into one: SMP could be promptless, simply
depending on NR_CPUS > 1.
A (desirable) side effect of this change - resulting from numeric
values now necessarily being compared as numbers rather than as
strings - is that comparing hex values now works as expected: Other
than int ones (which aren't allowed to have leading zeroes), zeroes
following the 0x prefix made them compare unequal even if their values
were equal.
Signed-off-by: Jan Beulich <[email protected]>
Signed-off-by: Michal Marek <[email protected]>
At the very least we should tell people that what they wrote is not
what the utility understands.
Signed-off-by: Jan Beulich <[email protected]>
Acked-by: Paul Bolle <[email protected]>
Signed-off-by: Michal Marek <[email protected]>
Older compilers don't recognize that "v" can't be used uninitialized;
other code using hvm_get_parameter() zeros the value too, so follow
suit here.
Signed-off-by: Jan Beulich <[email protected]>
Signed-off-by: David Vrabel <[email protected]>
Older compilers don't recognize that "v" can't be used uninitialized;
other code using hvm_get_parameter() zeros the value too, so follow
suit here.
Signed-off-by: Jan Beulich <[email protected]>
Signed-off-by: David Vrabel <[email protected]>
It's not clear to me why only the enabling operation got handled so
far.
Signed-off-by: Jan Beulich <[email protected]>
Reviewed-by: Konrad Rzeszutek Wilk <[email protected]>
Signed-off-by: David Vrabel <[email protected]>
Otherwise the guest can abuse that control to cause e.g. PCIe
Unsupported Request responses by disabling memory and/or I/O decoding
and subsequently causing (CPU side) accesses to the respective address
ranges, which (depending on system configuration) may be fatal to the
host.
Note that to alter any of the bits collected together as
PCI_COMMAND_GUEST permissive mode is now required to be enabled
globally or on the specific device.
This is CVE-2015-2150 / XSA-120.
Signed-off-by: Jan Beulich <[email protected]>
Reviewed-by: Konrad Rzeszutek Wilk <[email protected]>
Cc: <[email protected]>
Signed-off-by: David Vrabel <[email protected]>
The conditional in local_irq_restore() otherwise can cause code
bloat (the if and else blocks may get translated into separate
code paths despite the generated code being identical, dependent
on compiler internal heuristics). Note that this adjustment gets
the code in sync with the comment preceding it (which was
slightly wrong from at least from 2.6.37 onwards).
The code bloat was observed in reality with an experimental x86
patch.
Signed-off-by: Jan Beulich <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Steven Rostedt <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
We don't really need a helper symbol for that. For one, it's
pointlessly getting set to Y for all configurations (even 64-bit
ones). And then the purpose can be fulfilled by suitably
adjusting X86_UP_APIC: Hide its prompt when PCI_MSI, and default
it to PCI_MSI.
Tested-by: Bryan O'Donoghue <[email protected]>
Signed-off-by: Jan Beulich <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
Since dependencies are transitive, we don't really need to
repeat those of X86_UP_IOAPIC.
Furthermore avoid the symbol getting entered into .config when
it is off by having the default simply Y and the dependencies
solely handled via the intended for that purpose "depends on".
Signed-off-by: Jan Beulich <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
Settings without prompts shouldn't normally have defaults other
than Y, as otherwise they (a) needlessly enlarge the resulting
.config and (b) if they ever get a prompt added later, the
tracked setting of off will prevent the devloper from then being
prompted for his/her choice when doing an incremental update of
the configuration (make oldconfig).
Signed-off-by: Jan Beulich <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
Commit 81907478c4 ("sched/fair: Avoid using uninitialized variable
in preferred_group_nid()") unconditionally initializes max_group with
NODE_MASK_NONE, this means that when !max_faults (max_group didn't get
set), we'll now continue the iteration with an empty mask.
Which in turn makes the actual body of the loop go away, so we'll just
iterate until completion; short circuit this by breaking out of the
loop as soon as this would happen.
Signed-off-by: Jan Beulich <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Cc: Linus Torvalds <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
Just like for AVX2 (which simply needs an #if -> #ifdef conversion),
SSSE3 assembler support should be checked for before using it.
Signed-off-by: Jan Beulich <[email protected]>
Cc: Jim Kukunas <[email protected]>
Acked-by: Thomas Gleixner <[email protected]>
Signed-off-by: NeilBrown <[email protected]>
At least some gcc versions - validly afaict - warn about potentially
using max_group uninitialized: There's no way the compiler can prove
that the body of the conditional where it and max_faults get set/
updated gets executed; in fact, without knowing all the details of
other scheduler code, I can't prove this either.
Generally the necessary change would appear to be to clear max_group
prior to entering the inner loop, and break out of the outer loop when
it ends up being all clear after the inner one. This, however, seems
inefficient, and afaict the same effect can be achieved by exiting the
outer loop when max_faults is still zero after the inner loop.
[ mingo: changed the solution to zero initialization: uninitialized_var()
needs to die, as it's an actively dangerous construct: if in the future
a known-proven-good piece of code is changed to have a true, buggy
uninitialized variable, the compiler warning is then supressed...
The better long term solution is to clean up the code flow, so that
even simple minded compilers (and humans!) are able to read it without
getting a headache. ]
Signed-off-by: Jan Beulich <[email protected]>
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Cc: Rik van Riel <[email protected]>
Cc: Linus Torvalds <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
This makes a difference if the compiler decides not to inline the
function, as then the function's reference to acpisleep_dmi_table[]
yields a section mismatch warning.
Signed-off-by: Jan Beulich <[email protected]>
Signed-off-by: Rafael J. Wysocki <[email protected]>
As a module_init() function, this should have been this way from the
beginning.
Signed-off-by: Jan Beulich <[email protected]>
Signed-off-by: David Vrabel <[email protected]>
Using the native code here can't work properly, as the hypervisor would
normally have cleared the two reason bits by the time Dom0 gets to see
the NMI (if passed to it at all). There's a shared info field for this,
and there's an existing hook to use - just fit the two together. This
is particularly relevant so that NMIs intended to be handled by APEI /
GHES actually make it to the respective handler.
Note that the hook can (and should) be used irrespective of whether
being in Dom0, as accessing port 0x61 in a DomU would be even worse,
while the shared info field would just hold zero all the time. Note
further that hardware NMI handling for PVH doesn't currently work
anyway due to missing code in the hypervisor (but it is expected to
work the native rather than the PV way).
Signed-off-by: Jan Beulich <[email protected]>
Reviewed-by: Boris Ostrovsky <[email protected]>
Signed-off-by: David Vrabel <[email protected]>
The old scheme can lead to failure in certain cases - the
problem is that after bumping step_size the next (non-final)
iteration is only guaranteed to make available a memory block
the size of what step_size was before. E.g. for a memory block
[0,3004600000) we'd have:
iter start end step amount
1 3004400000 30045fffff 2M 2M
2 3004000000 30043fffff 64M 4M
3 3000000000 3003ffffff 2G 64M
4 2000000000 2fffffffff 64G 64G
Yet to map 64G with 4k pages (as happens e.g. under PV Xen) we
need slightly over 128M, but the first three iterations made
only about 70M available.
The condition (new_mapped_ram_size > mapped_ram_size) for
bumping step_size is just not suitable. Instead we want to bump
it when we know we have enough memory available to cover a block
of the new step_size. And rather than making that condition more
complicated than needed, simply adjust step_size by the largest
possible factor we know we can cover at that point - which is
shifting it left by one less than the difference between page
table level shifts. (Interestingly the original STEP_SIZE_SHIFT
definition had a comment hinting at that having been the
intention, just that it should have been PUD_SHIFT-PMD_SHIFT-1
instead of (PUD_SHIFT-PMD_SHIFT)/2, and of course for non-PAE
32-bit we can't really use these two constants as they're equal
there.)
Furthermore the comment in get_new_step_size() didn't get
updated when the bottom-down mapping logic got added. Yet while
an overflow (flushing step_size to zero) of the shift doesn't
matter for the top-down method, it does for bottom-up because
round_up(x, 0) = 0, and an upper range boundary of zero can't
really work well.
Signed-off-by: Jan Beulich <[email protected]>
Acked-by: Yinghai Lu <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
When X86_LOCAL_APIC (i.e. unconditionally on x86-64),
first_system_vector will never end up being higher than
LOCAL_TIMER_VECTOR (0xef), and hence building stubs for vectors
0xef...0xff is pointlessly reducing code density. Deal with this at
build time already.
Taking into consideration that X86_64 implies X86_LOCAL_APIC, also
simplify (and hence make easier to read and more consistent with the
change done here) some #if-s in arch/x86/kernel/irqinit.c.
While we could further improve the packing of the IRQ entry stubs (the
four ones now left in the last set could be fit into the four padding
bytes each of the final four sets have) this doesn't seem to provide
any real benefit: Both irq_entries_start and common_interrupt getting
cache line aligned, eliminating the 30th set would just produce 32
bytes of padding between the 29th and common_interrupt.
[ tglx: Folded lguest fix from Dan Carpenter ]
Signed-off-by: Jan Beulich <[email protected]>
Cc: Dan Carpenter <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: [email protected]
Cc: Rusty Russell <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Link: http://lkml.kernel.org/r/20141115185718.GB6530@mwanda
Signed-off-by: Thomas Gleixner <[email protected]>
While f3761db164 ("x86, irq: Fix build error caused by
9eabc99a635a77cbf09") addressed the original build problem,
declaration, inline stub, and definition still seem misplaced: It isn't
really IO-APIC related, and it's being used solely in arch/x86/pci/.
This also means stubbing it out when !CONFIG_X86_IO_APIC was at least
questionable.
Signed-off-by: Jan Beulich <[email protected]>
Cc: Jiang Liu <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Thomas Gleixner <[email protected]>
When xenvif_alloc() fails, it returns a non-NULL error indicator. To
avoid eventual races, we shouldn't store that into struct backend_info
as readers of it only check for NULL.
Signed-off-by: Jan Beulich <[email protected]>
Acked-by: Ian Campbell <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
When a PF driver unloads, it may find it necessary to leave the VFs
around simply because of pciback having marked them as assigned to a
guest. Utilize a suitable notification to let go of the VFs, thus
allowing the PF to go back into the state it was before its driver
loaded (which in particular allows the driver to be loaded again with
it being able to create the VFs anew, but which also allows to then
pass through the PF instead of the VFs).
Don't do this however for any VFs currently in active use by a guest.
Signed-off-by: Jan Beulich <[email protected]>
[v2: Removed the switch statement, moved it about]
[v3: Redid it a bit differently]
Signed-off-by: Konrad Rzeszutek Wilk <[email protected]>
Signed-off-by: David Vrabel <[email protected]>
Observing that per-CPU data (in the SMP case) is reachable by
exploiting 64-bit address wraparound (building on the default kernel
load address being at 16Mb), the one byte shorter RIP-relative
addressing form can be used for most per-CPU accesses. The one
exception are the "stable" reads, where the use of the "P" operand
modifier prevents the compiler from using RIP-relative addressing, but
is unavoidable due to the use of the "p" constraint (side note: with
gcc 4.9.x the intended effect of this isn't being achieved anymore,
see gcc bug 63637).
With the dependency on the minimum kernel load address, arbitrarily
low values for CONFIG_PHYSICAL_START are now no longer possible. A
link time assertion is being added, directing to the need to increase
that value when it triggers.
Signed-off-by: Jan Beulich <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Thomas Gleixner <[email protected]>
Both this_cpu_off and cpu_info aren't getting modified post boot, yet
are being accessed on enough code paths that grouping them with other
frequently read items seems desirable. For cpu_info this at the same
time implies removing the cache line alignment (which afaict became
pointless when it got converted to per-CPU data years ago).
Signed-off-by: Jan Beulich <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Thomas Gleixner <[email protected]>
- don't include unneeded headers
- drop redundant entry point label
- complete unwind annotations
- use .L prefix on local labels to not clutter the symbol table
Signed-off-by: Jan Beulich <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Thomas Gleixner <[email protected]>
- don't include unneeded headers
- don't open-code PER_CPU_VAR()
- drop redundant entry point label
- complete unwind annotations
- use .L prefix on local label to not clutter the symbol table
Signed-off-by: Jan Beulich <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Thomas Gleixner <[email protected]>
In particular seeing zero in eft->month is problematic, as it results in
-1 (converted to unsigned int, i.e. yielding 0xffffffff) getting passed
to rtc_year_days(), where the value gets used as an array index
(normally resulting in a crash). This was observed with the driver
enabled on x86 on some Fujitsu system (with possibly not up to date
firmware, but anyway).
Perhaps efi_read_alarm() should not fail if neither enabled nor pending
are set, but the returned time is invalid?
Signed-off-by: Jan Beulich <[email protected]>
Reported-by: Raymund Will <[email protected]>
Cc: Alessandro Zummo <[email protected]>
Cc: Jingoo Han <[email protected]>
Acked-by: Lee, Chun-Yi <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
With the conversion of the register saving code from macros to
functions, and with those functions not clobbering most of the
registers they spill, there's no need to annotate most of the
spill operations; the only exceptions being %rbx (always
modified) and %rcx (modified on the error_kernelspace: path).
Also remove a bogus commented out annotation - there's no
register %orig_rax after all.
Signed-off-by: Jan Beulich <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
Relying on static functions used just once to get inlined (and
subsequently have dead code paths eliminated) is wrong: Compilers are
free to decide whether they do this, regardless of optimization level.
With this not happening for vdso_addr() (observed with gcc 4.1.x), an
unresolved reference to align_vdso_addr() causes the build to fail.
[ hpa: vdso_addr() is never actually used on x86-32, as calculate_addr
in map_vdso() is always false. It ought to be possible to clean
this up further, but this fixes the immediate problem. ]
Signed-off-by: Jan Beulich <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Acked-by: Andy Lutomirski <[email protected]>
Tested-by: Boris Ostrovsky <[email protected]>
Tested-by: Andrew Morton <[email protected]>
Signed-off-by: H. Peter Anvin <[email protected]>
Certain ld versions (observed with 2.20.0) put an empty .rela.dyn
section into shared object files, breaking the assumption on the number
of sections to be copied to the final output. Simply discard any empty
SHT_REL and SHT_RELA sections to address this.
Signed-off-by: Jan Beulich <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Acked-by: Andy Lutomirski <[email protected]>
Tested-by: Boris Ostrovsky <[email protected]>
Tested-by: Andrew Morton <[email protected]>
Signed-off-by: H. Peter Anvin <[email protected]>
In 2.6.29 io_tlb_orig_addr[] got converted from storing virtual addresses
to storing physical ones. While checking virtual addresses against NULL
is a legitimate thing to catch invalid entries, checking physical ones
against zero isn't: There's no guarantee that PFN 0 is reserved on a
particular platform.
Since it is unclear whether the check in swiotlb_tbl_unmap_single() is
actually needed, retain it but check against a guaranteed invalid physical
address. This requires setting up the array in a suitable fashion. And
since the original code failed to invalidate array entries when regions
get unmapped, this is being fixed at once along with adding a similar
check to swiotlb_tbl_sync_single().
Obviously the less intrusive change would be to simply drop the check in
swiotlb_tbl_unmap_single().
Signed-off-by: Jan Beulich <[email protected]>
Signed-off-by: Konrad Rzeszutek Wilk <[email protected]>
Simplify populate_msi_sysfs() by
- Swapping the order of the two allocations and storing the
msi_dev_attr-derived pointer right after allocation, allowing the
cleanup code to pick things up without extra effort.
- Using kasprintf() instead of the kmalloc()/sprintf() pair.
Signed-off-by: Jan Beulich <[email protected]>
Signed-off-by: Bjorn Helgaas <[email protected]>
Acked-by: Greg Kroah-Hartman <[email protected]>
Recent increased use of typeof() throughout the tree resulted in a
number of symbols (25 in a typical distro config of ours) not getting a
proper CRC calculated for them anymore, due to the parser in genksyms
not coping with several of these uses (interestingly in the majority of
[if not all] cases the problem is due to the use of typeof() in code
preceding a certain export, not in the declaration/definition of the
exported function/object itself; I wasn't able to find a way to address
this more general parser shortcoming).
The use of parameter_declaration is a little more relaxed than would be
ideal (permitting not just a bare type specification, but also one with
identifier), but since the same code is being passed through an actual
compiler, there's no apparent risk of allowing through any broken code.
Otoh using parameter_declaration instead of the ad hoc
"decl_specifier_seq '*'" / "decl_specifier_seq" pair allows all types to
be handled rather than just plain ones and pointers to plain ones.
Signed-off-by: Jan Beulich <[email protected]>
Cc: Michal Marek <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
... to match the function's parameters. While reportedly commutative,
using the proper order allows for leveraging the instruction permitting
the source operand to be in memory.
[ hpa: This code originated in the dpdk toolkit. This was a bug in dpdk
which has recently been fixed in part due to an earlier version of
this patch. ]
Signed-off-by: Jan Beulich <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Acked-by: Daniel Borkmann <[email protected]>
Cc: Francesco Fusco <[email protected]>
Cc: Thomas Graf <[email protected]>
Cc: David S. Miller <[email protected]>
Signed-off-by: H. Peter Anvin <[email protected]>
consolidate duplicate code is skb_checksum_setup() helpers
Realizing that the skb_maybe_pull_tail() calls in the IP-protocol
specific portions of both helpers are terminal ones (i.e. no further
pulls are expected), their maximum size to be pulled can be made match
their minimal size needed, thus making the code identical and hence
possible to be moved into another helper.
Signed-off-by: Jan Beulich <[email protected]>
Cc: Paul Durrant <[email protected]>
Cc: David Miller <[email protected]>
Cc: Eric Dumazet <[email protected]>
Reviewed-by: Paul Durrant <[email protected]>
Signed-off-by: David S. Miller <[email protected]>
According to Documentation/Changes, make 3.80 is still being supported
for building the kernel, hence make files must not make (unconditional)
use of features introduced only in newer versions. Commit 8779657d29
("stackprotector: Introduce CONFIG_CC_STACKPROTECTOR_STRONG") however
introduced an "else ifdef" construct which make 3.80 doesn't understand.
Also correct a warning message still referencing the old config option
name.
Apart from that I question the use of "ifdef" here (but it was used that
way already prior to said commit): ifeq (,y) would seem more to the
point.
Signed-off-by: Jan Beulich <[email protected]>
Acked-by: Kees Cook <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Michal Marek <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
"ret", being set to -1 early on, gets cleared by the first invocation of
lz4_decompress()/lz4_decompress_unknownoutputsize(), and hence subsequent
failures wouldn't be noticed by the caller without setting it back to -1
right after those calls.
Reported-by: Matthew Daley <[email protected]>
Signed-off-by: Jan Beulich <[email protected]>
Cc: Kyungsik Lee <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Commit 7ea6c6c1 ("Move cper.c from drivers/acpi/apei to
drivers/firmware/efi") results in CONFIG_EFI being enabled even
when the user doesn't want this. Since ACPI APEI used to build
fine without UEFI (and as far as I know also has no functional
depency on it), at least in that case using a reverse dependency
is wrong (and a straight one isn't needed).
Whether the same is true for ACPI_EXTLOG I don't know - if there
is a functional dependency, it should depend on EFI rather than
selecting it. It certainly has (currently) no build dependency.
Adjust Kconfig and build logic so that the bad dependency gets
avoided.
Signed-off-by: Jan Beulich <[email protected]>
Acked-by: Tony Luck <[email protected]>
Cc: Matt Fleming <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
According to Documentation/Changes, make 3.80 is still being supported
for building the kernel, hence make files must not make (unconditional)
use of features introduced only in newer versions.
Commit 1bf49dd4be ("./Makefile: export initial ramdisk compression
config option") however introduced "else ifeq" constructs which make
3.80 doesn't understand. Replace the logic there with more conventional
(in the kernel build infrastructure) list constructs (except that the
list here is intentionally limited to exactly one element).
Signed-off-by: Jan Beulich <[email protected]>
Cc: P J P <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Commit fad1a86e25 ("procfs: call default get_unmapped_area on
MMU-present architectures"), as its title says, took care of only the
MMU case, leaving the !MMU side still in the regressed state (returning
-EIO in all cases where pde->proc_fops->get_unmapped_area is NULL).
From the fad1a86e25 changelog:
"Commit c4fe244857 ("sparc: fix PCI device proc file mmap(2)") added
proc_reg_get_unmapped_area in proc_reg_file_ops and
proc_reg_file_ops_no_compat, by which now mmap always returns EIO if
get_unmapped_area method is not defined for the target procfs file, which
causes regression of mmap on /proc/vmcore.
To address this issue, like get_unmapped_area(), call default
current->mm->get_unmapped_area on MMU-present architectures if
pde->proc_fops->get_unmapped_area, i.e. the one in actual file operation
in the procfs file, is not defined"
Signed-off-by: Jan Beulich <[email protected]>
Cc: HATAYAMA Daisuke <[email protected]>
Cc: Alexey Dobriyan <[email protected]>
Cc: David S. Miller <[email protected]>
Cc: <[email protected]> [3.12.x]
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Even though the omission was found only during code review
(originally in the Xen hypervisor, looking through ACPI v5 flags
and their meanings and uses), we shouldn't be creating a
corresponding platform device in that case.
Signed-off-by: Jan Beulich <[email protected]>
Cc: John Stultz <[email protected]>
Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: Ingo Molnar <[email protected]>
Jan Beulich