We found a bug in 3.10.x.
The problem is that we accidentally have a hwpoisoned hugepage in free
hugepage list. It could happend in the the following scenario:
process A process B
migrate_huge_page
put_page (old hugepage)
linked to free hugepage list
hugetlb_fault
hugetlb_no_page
alloc_huge_page
dequeue_huge_page_vma
dequeue_huge_page_node
(steal hwpoisoned hugepage)
set_page_hwpoison_huge_page
dequeue_hwpoisoned_huge_page
(fail to dequeue)
I tested this bug, one process keeps allocating huge page, and I
use sysfs interface to soft offline a huge page, then received:
"MCE: Killing UCP:2717 due to hardware memory corruption fault at 8200034"
Upstream kernel is free from this bug because of these two commits:
The first one, although the problem is about memory leak, this patch
moves unset_migratetype_isolate(), which is important to avoid the race.
The latter is not a bug fix and it's too big, so I rewrite a small one.
The following patch can fix this bug.(please apply f15bdfa802bf first)
qib_user_sdma_queue_pkts() gets called with mmap_sem held for
writing. Except for get_user_pages() deep down in
qib_user_sdma_pin_pages() we don't seem to need mmap_sem at all. Even
more interestingly the function qib_user_sdma_queue_pkts() (and also
qib_user_sdma_coalesce() called somewhat later) call copy_from_user()
which can hit a page fault and we deadlock on trying to get mmap_sem
when handling that fault.
So just make qib_user_sdma_pin_pages() use get_user_pages_fast() and
leave mmap_sem locking for mm.
This deadlock has actually been observed in the wild when the node
is under memory pressure.
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Jan Kara <jack@suse.cz> Signed-off-by: Roland Dreier <roland@purestorage.com>
[Backported to 3.10: (Thanks to Ben Huthings)
- Adjust context
- Adjust indentation and nr_pages argument in qib_user_sdma_pin_pages()] Signed-off-by: Ben Hutchings <ben@decadent.org.uk> Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
After a successful page migration by soft offlining, the source page is
not properly freed and it's never reusable even if we unpoison it
afterward.
This is caused by the race between freeing page and setting PG_hwpoison.
In successful soft offlining, the source page is put (and the refcount
becomes 0) by putback_lru_page() in unmap_and_move(), where it's linked
to pagevec and actual freeing back to buddy is delayed. So if
PG_hwpoison is set for the page before freeing, the freeing does not
functions as expected (in such case freeing aborts in
free_pages_prepare() check.)
This patch tries to make sure to free the source page before setting
PG_hwpoison on it. To avoid reallocating, the page keeps
MIGRATE_ISOLATE until after setting PG_hwpoison.
This patch also removes obsolete comments about "keeping elevated
refcount" because what they say is not true. Unlike memory_failure(),
soft_offline_page() uses no special page isolation code, and the
soft-offlined pages have no elevated.
Due to an assumption in the VT8500 pinctrl driver, the value passed
from devicetree for 'wm,pull' was not explicitly translated before
being passed to pinconf.
Since v3.10, changes to 'enum pin_config_param', PIN_CONFIG_BIAS_PULL_(UP/DOWN)
no longer map 1-to-1 with the expected values in devicetree.
This patch adds a small translation between the devicetree values (0..2)
and the enum pin_config_param equivalent values.
Commit d61931d89b, "x86: Add optimized popcnt variants" introduced
compile flag -fcall-saved-rdi for lib/hweight.c. When combined with
options -fprofile-arcs and -O2, this flag causes gcc to generate
broken constructor code. As a result, a 64 bit x86 kernel compiled
with CONFIG_GCOV_PROFILE_ALL=y prints message "gcov: could not create
file" and runs into sproadic BUGs during boot.
The gcc people indicate that these kinds of problems are endemic when
using ad hoc calling conventions. It is therefore best to treat any
file compiled with ad hoc calling conventions as an isolated
environment and avoid things like profiling or coverage analysis,
since those subsystems assume a "normal" calling conventions.
This patch avoids the bug by excluding lib/hweight.o from coverage
profiling.
Reported-by: Meelis Roos <mroos@linux.ee> Cc: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Peter Oberparleiter <oberpar@linux.vnet.ibm.com> Link: http://lkml.kernel.org/r/52F3A30C.7050205@linux.vnet.ibm.com Signed-off-by: H. Peter Anvin <hpa@zytor.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
There was a large performance regression that was bisected to
commit 611ae8e3 ("x86/tlb: enable tlb flush range support for
x86"). This patch simply changes the default balance point
between a local and global flush for IvyBridge.
In the interest of allowing the tests to be reproduced, this
patch was tested using mmtests 0.15 with the following
configurations
Ebizzy was configured to run multiple iterations and threads.
Thread counts ranged from 1 to NR_CPUS*2. For each thread count,
it ran 100 iterations and each iteration lasted 10 seconds.
Ivybridge 4 threads
3.13.0-rc7 3.13.0-rc7
vanilla altshift-v3
Mean 1 6395.44 ( 0.00%) 6789.09 ( 6.16%)
Mean 2 7012.85 ( 0.00%) 8052.16 ( 14.82%)
Mean 3 6403.04 ( 0.00%) 6973.74 ( 8.91%)
Mean 4 6135.32 ( 0.00%) 6582.33 ( 7.29%)
Mean 5 6095.69 ( 0.00%) 6526.68 ( 7.07%)
Mean 6 6114.33 ( 0.00%) 6416.64 ( 4.94%)
Mean 7 6085.10 ( 0.00%) 6448.51 ( 5.97%)
Mean 8 6120.62 ( 0.00%) 6462.97 ( 5.59%)
Ivybridge 8 threads
3.13.0-rc7 3.13.0-rc7
vanilla altshift-v3
Mean 1 7336.65 ( 0.00%) 7787.02 ( 6.14%)
Mean 2 8218.41 ( 0.00%) 9484.13 ( 15.40%)
Mean 3 7973.62 ( 0.00%) 8922.01 ( 11.89%)
Mean 4 7798.33 ( 0.00%) 8567.03 ( 9.86%)
Mean 5 7158.72 ( 0.00%) 8214.23 ( 14.74%)
Mean 6 6852.27 ( 0.00%) 7952.45 ( 16.06%)
Mean 7 6774.65 ( 0.00%) 7536.35 ( 11.24%)
Mean 8 6510.50 ( 0.00%) 6894.05 ( 5.89%)
Mean 12 6182.90 ( 0.00%) 6661.29 ( 7.74%)
Mean 16 6100.09 ( 0.00%) 6608.69 ( 8.34%)
Ebizzy hits the worst case scenario for TLB range flushing every
time and it shows for these Ivybridge CPUs at least that the
default choice is a poor on. The patch addresses the problem.
Next was a tlbflush microbenchmark written by Alex Shi at
http://marc.info/?l=linux-kernel&m=133727348217113 . It
measures access costs while the TLB is being flushed. The
expectation is that if there are always full TLB flushes that
the benchmark would suffer and it benefits from range flushing
There are 320 iterations of the test per thread count. The
number of entries is randomly selected with a min of 1 and max
of 512. To ensure a reasonably even spread of entries, the full
range is broken up into 8 sections and a random number selected
within that section.
iteration 1, random number between 0-64
iteration 2, random number between 64-128 etc
This is still a very weak methodology. When you do not know
what are typical ranges, random is a reasonable choice but it
can be easily argued that the opimisation was for smaller ranges
and an even spread is not representative of any workload that
matters. To improve this, we'd need to know the probability
distribution of TLB flush range sizes for a set of workloads
that are considered "common", build a synthetic trace and feed
that into this benchmark. Even that is not perfect because it
would not account for the time between flushes but there are
limits of what can be reasonably done and still be doing
something useful. If a representative synthetic trace is
provided then this benchmark could be revisited and the shift values retuned.
Ivybridge 4 threads
3.13.0-rc7 3.13.0-rc7
vanilla altshift-v3
Mean 1 10.50 ( 0.00%) 10.50 ( 0.03%)
Mean 2 17.59 ( 0.00%) 17.18 ( 2.34%)
Mean 3 22.98 ( 0.00%) 21.74 ( 5.41%)
Mean 5 47.13 ( 0.00%) 46.23 ( 1.92%)
Mean 8 43.30 ( 0.00%) 42.56 ( 1.72%)
Ivybridge 8 threads
3.13.0-rc7 3.13.0-rc7
vanilla altshift-v3
Mean 1 9.45 ( 0.00%) 9.36 ( 0.93%)
Mean 2 9.37 ( 0.00%) 9.70 ( -3.54%)
Mean 3 9.36 ( 0.00%) 9.29 ( 0.70%)
Mean 5 14.49 ( 0.00%) 15.04 ( -3.75%)
Mean 8 41.08 ( 0.00%) 38.73 ( 5.71%)
Mean 13 32.04 ( 0.00%) 31.24 ( 2.49%)
Mean 16 40.05 ( 0.00%) 39.04 ( 2.51%)
For both CPUs, average access time is reduced which is good as
this is the benchmark that was used to tune the shift values in
the first place albeit it is now known *how* the benchmark was
used.
The scheduler benchmarks were somewhat inconclusive. They
showed gains and losses and makes me reconsider how stable those
benchmarks really are or if something else might be interfering
with the test results recently.
Network benchmarks were inconclusive. Almost all results were
flat except for netperf-udp tests on the 4 thread machine.
These results were unstable and showed large variations between
reboots. It is unknown if this is a recent problems but I've
noticed before that netperf-udp results tend to vary.
Based on these results, changing the default for Ivybridge seems
like a logical choice.
Signed-off-by: Mel Gorman <mgorman@suse.de> Tested-by: Davidlohr Bueso <davidlohr@hp.com> Reviewed-by: Alex Shi <alex.shi@linaro.org> Reviewed-by: Rik van Riel <riel@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/n/tip-cqnadffh1tiqrshthRj3Esge@git.kernel.org Signed-off-by: Ingo Molnar <mingo@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
To use spin_{un}lock_irq is dangerous if caller disabled interrupt.
During aio buffer migration, we have a possibility to see the following
call stack.
Mac Pro 1,1 with ALC889A codec needs the VREF setup on NID 0x18 to
VREF50, in order to make the speaker working. The same fixup was
already needed for MacBook Air 1,1, so we can reuse it.
The commit 44dcbbb1cd61 introduced the usage of bitreverse helpers but
forgot to add the dependency. This patch adds the selection for
CONFIG_BITREVERSE.
Add DSB after icache flush to complete the cache maintenance operation.
The function __flush_icache_all() is used only for user space mappings
and an ISB is not required because of an exception return before executing
user instructions. An exception return would behave like an ISB.
Signed-off-by: Vinayak Kale <vkale@apm.com> Acked-by: Will Deacon <will.deacon@arm.com> Signed-off-by: Catalin Marinas <catalin.marinas@arm.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When __kernel_clock_gettime is called with a CLOCK_MONOTONIC_COARSE or
CLOCK_REALTIME_COARSE clock id, it returns incorrectly to whatever the
caller has placed in x2 ("ret x2" to return from the fast path). Fix
this by saving x30/LR to x2 only in code that will call
__do_get_tspec, restoring x30 afterward, and using a plain "ret" to
return from the routine.
Also: while the resulting tv_nsec value for CLOCK_REALTIME and
CLOCK_MONOTONIC must be computed using intermediate values that are
left-shifted by cs_shift (x12, set by __do_get_tspec), the results for
coarse clocks should be calculated using unshifted values
(xtime_coarse_nsec is in units of actual nanoseconds). The current
code shifts intermediate values by x12 unconditionally, but x12 is
uninitialized when servicing a coarse clock. Fix this by setting x12
to 0 once we know we are dealing with a coarse clock id.
With the 64K page size configuration, __create_page_tables in head.S
maps enough memory to get started but using 64K pages rather than 512M
sections with a single pgd/pud/pmd entry pointing to a pte table.
create_mapping() may override the pgd/pud/pmd table entry with a block
(section) one if the RAM size is more than 512MB and aligned correctly.
For the end of this block to be accessible, the old TLB entry must be
invalidated.
Reported-by: Mark Salter <msalter@redhat.com> Tested-by: Mark Salter <msalter@redhat.com> Signed-off-by: Catalin Marinas <catalin.marinas@arm.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Whilst the text segment for our VDSO is marked as PT_LOAD in the ELF
headers, it is mapped by the kernel and not actually subject to
demand-paging. ld doesn't realise this, and emits a p_align field of 64k
(the maximum supported page size), which conflicts with the load address
picked by the kernel on 4k systems, which will be 4k aligned. This
causes GDB to fail with "Failed to read a valid object file image from
memory" when attempting to load the VDSO.
This patch passes the -n option to ld, which prevents it from aligning
PT_LOAD segments to the maximum page size.
Update wall-to-monotonic fields in the VDSO data page
unconditionally. These are used to service CLOCK_MONOTONIC_COARSE,
which is not guarded by use_syscall.
In the Armada 370/XP driver, when we receive an IRQ 0, we read the
list of doorbells that caused the interrupt from register
ARMADA_370_XP_IN_DRBEL_CAUSE_OFFS. This gives the list of IPIs that
were generated. However, instead of acknowledging only the IPIs that
were generated, we acknowledge *all* the IPIs, by writing
~IPI_DOORBELL_MASK in the ARMADA_370_XP_IN_DRBEL_CAUSE_OFFS register.
This creates a race condition: if a new IPI that isn't part of the
ones read into the temporary "ipimask" variable is fired before we
acknowledge all IPIs, then we will simply loose it. This is causing
scheduling hangs on SMP intensive workloads.
It is important to mention that this ARMADA_370_XP_IN_DRBEL_CAUSE_OFFS
register has the following behavior: "A CPU write of 0 clears the bits
in this field. A CPU write of 1 has no effect". This is what allows us
to simply write ~ipimask to acknoledge the handled IPIs.
Notice that the same problem is present in the MSI implementation, but
it will be fixed as a separate patch, so that this IPI fix can be
pushed to older stable versions as appropriate (all the way to 3.8),
while the MSI code only appeared in 3.13.
Signed-off-by: Lior Amsalem <alior@marvell.com> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Fixes: 344e873e5657e8dc0 'arm: mvebu: Add IPI support via doorbells' Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Jason Cooper <jason@lakedaemon.net> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In s390 des and 3des ctr mode there is one preallocated page
used to speed up the en/decryption. This page is not protected
against concurrent usage and thus there is a potential of data
corruption with multiple threads.
The fix introduces locking/unlocking the ctr page and a slower
fallback solution at concurrency situations.
Signed-off-by: Harald Freudenberger <freude@linux.vnet.ibm.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In s390 des and des3_ede cbc mode the iv value is not protected
against concurrency access and modifications from another running
en/decrypt operation which is using the very same tfm struct
instance. This fix copies the iv to the local stack before
the crypto operation and stores the value back when done.
Signed-off-by: Harald Freudenberger <freude@linux.vnet.ibm.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The aes-ctr mode uses one preallocated page without any concurrency
protection. When multiple threads run aes-ctr encryption or decryption
this can lead to data corruption.
The patch introduces locking for the page and a fallback solution with
slower en/decryption performance in concurrency situations.
Signed-off-by: Harald Freudenberger <freude@linux.vnet.ibm.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Setting an empty security context (length=0) on a file will
lead to incorrectly dereferencing the type and other fields
of the security context structure, yielding a kernel BUG.
As a zero-length security context is never valid, just reject
all such security contexts whether coming from userspace
via setxattr or coming from the filesystem upon a getxattr
request by SELinux.
Setting a security context value (empty or otherwise) unknown to
SELinux in the first place is only possible for a root process
(CAP_MAC_ADMIN), and, if running SELinux in enforcing mode, only
if the corresponding SELinux mac_admin permission is also granted
to the domain by policy. In Fedora policies, this is only allowed for
specific domains such as livecd for setting down security contexts
that are not defined in the build host policy.
Reproducer:
su
setenforce 0
touch foo
setfattr -n security.selinux foo
Caveat:
Relabeling or removing foo after doing the above may not be possible
without booting with SELinux disabled. Any subsequent access to foo
after doing the above will also trigger the BUG.
Reported-by: Matthew Thode <mthode@mthode.org> Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov> Signed-off-by: Paul Moore <pmoore@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The minimum pstate is supposed to be a percentage of the maximum P
state available. Calculate min using max pstate and not the
current max which may have been limited by the user
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch addresses Bug 60727
(https://bugzilla.kernel.org/show_bug.cgi?id=60727)
which was due to the truncation of intermediate values in the
calculations, which causes the code to consistently underestimate the
current cpu frequency, specifically 100% cpu utilization was truncated
down to the setpoint of 97%. This patch fixes the problem by keeping
the results of all intermediate calculations as fixed point numbers
rather scaling them back and forth between integers and fixed point.
References: https://bugzilla.kernel.org/show_bug.cgi?id=60727 Signed-off-by: Brennan Shacklett <bpshacklett@gmail.com> Acked-by: Dirk Brandewie <dirk.j.brandewie@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When sysfs for no_turbo is set, then also some p states in turbo regions
are observed. This patch will set IDA Engage bit when no_turbo is set to
explicitly disengage turbo.
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com> Acked-by: Dirk Brandewie <dirk.j.brandewie@intel.com> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Enable the intel_pstate driver for Haswell CPUs. One missing Ivy Bridge
model (0x3E) is also included. Models referenced from
tools/power/x86/turbostat/turbostat.c:has_nehalem_turbo_ratio_limit
As part of normal operaions, the hrtimer subsystem frequently calls
into the timekeeping code, creating a locking order of
hrtimer locks -> timekeeping locks
clock_was_set_delayed() was suppoed to allow us to avoid deadlocks
between the timekeeping the hrtimer subsystem, so that we could
notify the hrtimer subsytem the time had changed while holding
the timekeeping locks. This was done by scheduling delayed work
that would run later once we were out of the timekeeing code.
But unfortunately the lock chains are complex enoguh that in
scheduling delayed work, we end up eventually trying to grab
an hrtimer lock.
Sasha Levin noticed this in testing when the new seqlock lockdep
enablement triggered the following (somewhat abrieviated) message:
[ 251.100221] ======================================================
[ 251.100221] [ INFO: possible circular locking dependency detected ]
[ 251.100221] 3.13.0-rc2-next-20131206-sasha-00005-g8be2375-dirty #4053 Not tainted
[ 251.101967] -------------------------------------------------------
[ 251.101967] kworker/10:1/4506 is trying to acquire lock:
[ 251.101967] (timekeeper_seq){----..}, at: [<ffffffff81160e96>] retrigger_next_event+0x56/0x70
[ 251.101967]
[ 251.101967] but task is already holding lock:
[ 251.101967] (hrtimer_bases.lock#11){-.-...}, at: [<ffffffff81160e7c>] retrigger_next_event+0x3c/0x70
[ 251.101967]
[ 251.101967] which lock already depends on the new lock.
[ 251.101967]
[ 251.101967]
[ 251.101967] the existing dependency chain (in reverse order) is:
[ 251.101967]
-> #5 (hrtimer_bases.lock#11){-.-...}:
[snipped]
-> #4 (&rt_b->rt_runtime_lock){-.-...}:
[snipped]
-> #3 (&rq->lock){-.-.-.}:
[snipped]
-> #2 (&p->pi_lock){-.-.-.}:
[snipped]
-> #1 (&(&pool->lock)->rlock){-.-...}:
[ 251.101967] [<ffffffff81194803>] validate_chain+0x6c3/0x7b0
[ 251.101967] [<ffffffff81194d9d>] __lock_acquire+0x4ad/0x580
[ 251.101967] [<ffffffff81194ff2>] lock_acquire+0x182/0x1d0
[ 251.101967] [<ffffffff84398500>] _raw_spin_lock+0x40/0x80
[ 251.101967] [<ffffffff81153e69>] __queue_work+0x1a9/0x3f0
[ 251.101967] [<ffffffff81154168>] queue_work_on+0x98/0x120
[ 251.101967] [<ffffffff81161351>] clock_was_set_delayed+0x21/0x30
[ 251.101967] [<ffffffff811c4bd1>] do_adjtimex+0x111/0x160
[ 251.101967] [<ffffffff811e2711>] compat_sys_adjtimex+0x41/0x70
[ 251.101967] [<ffffffff843a4b49>] ia32_sysret+0x0/0x5
[ 251.101967]
-> #0 (timekeeper_seq){----..}:
[snipped]
[ 251.101967] other info that might help us debug this:
[ 251.101967]
[ 251.101967] Chain exists of:
timekeeper_seq --> &rt_b->rt_runtime_lock --> hrtimer_bases.lock#11
So the best solution is to avoid calling clock_was_set_delayed() while
holding the timekeeping lock, and instead using a flag variable to
decide if we should call clock_was_set() once we've released the locks.
This works for the case here, where the do_adjtimex() was the deadlock
trigger point. Unfortuantely, in update_wall_time() we still hold
the jiffies lock, which would deadlock with the ipi triggered by
clock_was_set(), preventing us from calling it even after we drop the
timekeeping lock. So instead call clock_was_set_delayed() at that point.
Since 48cdc135d4840 (Implement a shadow timekeeper), we have to
call timekeeping_update() after any adjustment to the timekeeping
structure in order to make sure that any adjustments to the structure
persist.
In the timekeeping suspend path, we udpate the timekeeper
structure, so we should be sure to update the shadow-timekeeper
before releasing the timekeeping locks. Currently this isn't done.
In most cases, the next time related code to run would be
timekeeping_resume, which does update the shadow-timekeeper, but
in an abundence of caution, this patch adds the call to
timekeeping_update() in the suspend path.
Cc: Sasha Levin <sasha.levin@oracle.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Prarit Bhargava <prarit@redhat.com> Cc: Richard Cochran <richardcochran@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Signed-off-by: John Stultz <john.stultz@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
A think-o in the calculation of the monotonic -> tai time offset
results in CLOCK_TAI timers and nanosleeps to expire late (the
latency is ~2x the tai offset).
Fix this by adding the tai offset from the realtime offset instead
of subtracting.
Cc: Sasha Levin <sasha.levin@oracle.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Prarit Bhargava <prarit@redhat.com> Cc: Richard Cochran <richardcochran@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Signed-off-by: John Stultz <john.stultz@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Since 48cdc135d4840 (Implement a shadow timekeeper), we have to
call timekeeping_update() after any adjustment to the timekeeping
structure in order to make sure that any adjustments to the structure
persist.
Unfortunately, the updates to the tai offset via adjtimex do not
trigger this update, causing adjustments to the tai offset to be
made and then over-written by the previous value at the next
update_wall_time() call.
This patch resovles the issue by calling timekeeping_update()
right after setting the tai offset.
Cc: Sasha Levin <sasha.levin@oracle.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Prarit Bhargava <prarit@redhat.com> Cc: Richard Cochran <richardcochran@gmail.com> Cc: Ingo Molnar <mingo@kernel.org> Signed-off-by: John Stultz <john.stultz@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Doing some different tests, I discovered that function graph tracing, when
filtered via the set_ftrace_filter and set_ftrace_notrace files, does
not always keep with them if another function ftrace_ops is registered
to trace functions.
The reason is that function graph just happens to trace all functions
that the function tracer enables. When there was only one user of
function tracing, the function graph tracer did not need to worry about
being called by functions that it did not want to trace. But now that there
are other users, this becomes a problem.
# echo 1 > /proc/sys/kernel/stack_tracer_enabled
# cat trace
[..]
1) + 20.825 us | }
1) + 21.651 us | }
1) + 30.924 us | } /* SyS_ioctl */
1) | do_page_fault() {
1) | __do_page_fault() {
1) 0.274 us | down_read_trylock();
1) 0.098 us | find_vma();
1) | handle_mm_fault() {
1) | _raw_spin_lock() {
1) 0.102 us | preempt_count_add();
1) 0.097 us | do_raw_spin_lock();
1) 2.173 us | }
1) | do_wp_page() {
1) 0.079 us | vm_normal_page();
1) 0.086 us | reuse_swap_page();
1) 0.076 us | page_move_anon_rmap();
1) | unlock_page() {
1) 0.082 us | page_waitqueue();
1) 0.086 us | __wake_up_bit();
1) 1.801 us | }
1) 0.075 us | ptep_set_access_flags();
1) | _raw_spin_unlock() {
1) 0.098 us | do_raw_spin_unlock();
1) 0.105 us | preempt_count_sub();
1) 1.884 us | }
1) 9.149 us | }
1) + 13.083 us | }
1) 0.146 us | up_read();
When the stack tracer was enabled, it enabled all functions to be traced, which
now the function graph tracer also traces. This is a side effect that should
not occur.
To fix this a test is added when the function tracing is changed, as well as when
the graph tracer is enabled, to see if anything other than the ftrace global_ops
function tracer is enabled. If so, then the graph tracer calls a test trampoline
that will look at the function that is being traced and compare it with the
filters defined by the global_ops.
As an optimization, if there's no other function tracers registered, or if
the only registered function tracers also use the global ops, the function
graph infrastructure will call the registered function graph callback directly
and not go through the test trampoline.
Fixes: d2d45c7a03a2 "tracing: Have stack_tracer use a separate list of functions" Signed-off-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The synchronization needed after ftrace_ops are unregistered must happen
after the callback is disabled from becing called by functions.
The current location happens after the function is being removed from the
internal lists, but not after the function callbacks were disabled, leaving
the functions susceptible of being called after their callbacks are freed.
This affects perf and any externel users of function tracing (LTTng and
SystemTap).
Fixes: cdbe61bfe704 "ftrace: Allow dynamically allocated function tracers" Signed-off-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
ftrace_trace_function is a variable that holds what function will be called
directly by the assembly code (mcount). If just a single function is
registered and it handles recursion itself, then the assembly will call that
function directly without any helper function. It also passes in the
ftrace_op that was registered with the callback. The ftrace_op to send is
stored in the function_trace_op variable.
The ftrace_trace_function and function_trace_op needs to be coordinated such
that the called callback wont be called with the wrong ftrace_op, otherwise
bad things can happen if it expected a different op. Luckily, there's no
callback that doesn't use the helper functions that requires this. But
there soon will be and this needs to be fixed.
Use a set_function_trace_op to store the ftrace_op to set the
function_trace_op to when it is safe to do so (during the update function
within the breakpoint or stop machine calls). Or if dynamic ftrace is not
being used (static tracing) then we have to do a bit more synchronization
when the ftrace_trace_function is set as that takes affect immediately
(as oppose to dynamic ftrace doing it with the modification of the trampoline).
Signed-off-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When the mode is set with 16bpp on QEMU, the output gets totally broken.
The culprit is the bogus register values set for 16bpp, which was likely
copied from from a wrong place.
Add new definitions for hotplug live status bits for VLV2 since they're
in reverse order from the gen4x ones.
Changelog:
- Restored gen4 bit definitions
- Added new definitions for VLV2
- Added platform check for IS_VALLEYVIEW() in dp_detect to use the correct
bit defintions
- Replaced a lost trailing brace for the added switch()
Signed-off-by: Todd Previte <tprevite@gmail.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=73951
[danvet: Switch to _VLV postfix instead of prefix and regroupg
comments again so that the g4x warning is right next to those defines.
Also add a _G4X suffix for those special ones. Also cc stable.] Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The 'offset' field of the 'scatterlist' structure was wrongly
programmed with the offset value from the base of stolen area,
whereas this field indicates the offset from where the interested
data starts within the first PAGE pointed to by 'scattterlist'
structure. As a result when a new GEM object allocated from stolen
area is mapped to GTT, it could lead to an overwrite of GTT entries
as the page count calculation will go wrong, refer the function
'sg_page_count'.
In very rare cases (such as a memory failure stress test) it is possible
to fill the entire ring without emitting a request. Under this
circumstance, the outstanding request is flushed and waited upon. After
space on the ring is cleared, we return to emitting the new command -
except that we just cleared the seqno allocated for this operation and
trigger the sanity check that a request is only ever emitted with a
valid seqno. The fix is to rearrange the code to make sure the
allocation of the seqno for this operation is after any required flushes
of outstanding operations.
The bug exists since the preallocation was introduced in
commit 9d7730914f4cd496e356acfab95b41075aa8eae8
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date: Tue Nov 27 16:22:52 2012 +0000
drm/i915: Preallocate next seqno before touching the ring
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Mika Kuoppala <mika.kuoppala@intel.com> Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Reviewed-by: Jani Nikula <jani.nikula@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Commit de7b7d59d54852c introduced tiled GART, but a linear copy is
still performed. This may result in errors on eviction, fix it by
checking tiling from memtype.
This reverts commit be35f48610 ("dm: wait until embedded kobject is
released before destroying a device") and provides an improved fix.
The kobject release code that calls the completion must be placed in a
non-module file, otherwise there is a module unload race (if the process
calling dm_kobject_release is preempted and the DM module unloaded after
the completion is triggered, but before dm_kobject_release returns).
To fix this race, this patch moves the completion code to dm-builtin.c
which is always compiled directly into the kernel if BLK_DEV_DM is
selected.
The patch introduces a new dm_kobject_holder structure, its purpose is
to keep the completion and kobject in one place, so that it can be
accessed from non-module code without the need to export the layout of
struct mapped_device to that code.
s5p_mfc_get_node_type() relies on get_index() helper function, which in
turn relies on video_device index numbers assigned on driver
registration. All this code is not really needed, because there is
already access to respective video_device structures via common
s5p_mfc_dev structure. This fixes the issues introduced by patch 1056e4388b0454917a512618c8416a98628fc9ce ("v4l2-dev: Fix race condition
on __video_register_device"), which has been merged in v3.12-rc1.
As the dvb-frontend kthread can be called anytime, it can race
with some get status ioctl. So, it seems better to avoid one to
race with the other while reading a 32 bits register.
I can't see any other reason for having a mutex there at I2C, except
to provide such kind of protection, as the I2C core already has a
mutex to protect I2C transfers.
Note: instead of this approach, it could eventually remove the dib8000
specific mutex for it, and either group the 4 ops into one xfer or
to manually control the I2C mutex. The main advantage of the current
approach is that the changes are smaller and more puntual.
A 3% of system memory bonus is sometimes too excessive in comparison to
other processes.
With commit a63d83f427fb ("oom: badness heuristic rewrite"), the OOM
killer tries to avoid killing privileged tasks by subtracting 3% of
overall memory (system or cgroup) from their per-task consumption. But
as a result, all root tasks that consume less than 3% of overall memory
are considered equal, and so it only takes 33+ privileged tasks pushing
the system out of memory for the OOM killer to do something stupid and
kill dhclient or other root-owned processes. For example, on a 32G
machine it can't tell the difference between the 1M agetty and the 10G
fork bomb member.
The changelog describes this 3% boost as the equivalent to the global
overcommit limit being 3% higher for privileged tasks, but this is not
the same as discounting 3% of overall memory from _every privileged task
individually_ during OOM selection.
Replace the 3% of system memory bonus with a 3% of current memory usage
bonus.
By giving root tasks a bonus that is proportional to their actual size,
they remain comparable even when relatively small. In the example
above, the OOM killer will discount the 1M agetty's 256 badness points
down to 179, and the 10G fork bomb's 262144 points down to 183500 points
and make the right choice, instead of discounting both to 0 and killing
agetty because it's first in the task list.
Signed-off-by: David Rientjes <rientjes@google.com> Reported-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Cc: Michal Hocko <mhocko@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The hw i2c engines are disabled by default as the
current implementation is still experimental. Print
a warning when users enable it so that it's obvious
when the option is enabled.
v2: check for non-0 rather than 1
Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Reviewed-by: Christian König <christian.koenig@amd.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This bug was introduced in commit 7e664b3dec431e ("dm space map metadata:
fix extending the space map").
When extending a dm-thin metadata volume we:
- Switch the space map into a simple bootstrap mode, which allocates
all space linearly from the newly added space.
- Add new bitmap entries for the new space
- Increment the reference counts for those newly allocated bitmap
entries
- Commit changes to disk
- Switch back out of bootstrap mode.
But, the disk commit may allocate space itself, if so this fact will be
lost when switching out of bootstrap mode.
The bug exhibited itself as an error when the bitmap_root, with an
erroneous ref count of 0, was subsequently decremented as part of a
later disk commit. This would cause the disk commit to fail, and thinp
to enter read_only mode. The metadata was not damaged (thin_check
passed).
The fix is to put the increments + commit into a loop, running until
the commit has not allocated extra space. In practise this loop only
runs twice.
With this fix the following device mapper testsuite test passes:
dmtest run --suite thin-provisioning -n thin_remove_works_after_resize
Signed-off-by: Joe Thornber <ejt@redhat.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When extending a metadata space map we should do the first commit whilst
still in bootstrap mode -- a mode where all blocks get allocated in the
new area.
That way the commit overhead is allocated from the newly added space.
Otherwise we risk running out of space.
With this fix, and the previous commit "dm space map common: make sure
new space is used during extend", the following device mapper testsuite
test passes:
dmtest run --suite thin-provisioning -n /resize_metadata_no_io/
Signed-off-by: Joe Thornber <ejt@redhat.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
There may be other parts of the kernel holding a reference on the dm
kobject. We must wait until all references are dropped before
deallocating the mapped_device structure.
The dm_kobject_release method signals that all references are dropped
via completion. But dm_kobject_release doesn't free the kobject (which
is embedded in the mapped_device structure).
This is the sequence of operations:
* when destroying a DM device, call kobject_put from dm_sysfs_exit
* wait until all users stop using the kobject, when it happens the
release method is called
* the release method signals the completion and should return without
delay
* the dm device removal code that waits on the completion continues
* the dm device removal code drops the dm_mod reference the device had
* the dm device removal code frees the mapped_device structure that
contains the kobject
Using kobject this way should avoid the module unload race that was
mentioned at the beginning of this thread:
https://lkml.org/lkml/2014/1/4/83
If a snapshot is created and later deleted the origin dm_thin_device's
snapshotted_time will have been updated to reflect the snapshot's
creation time. The 'shared' flag in the dm_thin_lookup_result struct
returned from dm_thin_find_block() is an approximation based on
snapshotted_time -- this is done to avoid 0(n), or worse, time
complexity. In this case, the shared flag would be true.
But because the 'shared' flag reflects an approximation a block can be
incorrectly assumed to be shared (e.g. false positive for 'shared'
because the snapshot no longer exists). This could result in discards
issued to a thin device not being passed down to the pool's underlying
data device.
To fix this we double check that a thin block is really still in-use
after a mapping is removed using dm_pool_block_is_used(). If the
reference count for a block is now zero the discard is allowed to be
passed down.
Also add a 'definitely_not_shared' member to the dm_thin_new_mapping
structure -- reflects that the 'shared' flag in the response from
dm_thin_find_block() can only be held as definitive if false is
returned.
When a task enters call_refreshresult with status 0 from call_refresh and
!rpcauth_uptodatecred(task) it enters call_refresh again with no rate-limiting
or max number of retries.
Instead of trying forever, make use of the retry path that other errors use.
This only seems to be possible when the crrefresh callback is gss_refresh_null,
which only happens when destroying the context.
To reproduce:
1) mount with sec=krb5 (or sec=sys with krb5 negotiated for non FSID specific
operations).
2) reboot - the client will be stuck and will need to be hard rebooted
An NFS4ERR_RECALLCONFLICT is returned by server from a GET_LAYOUT
only when a Server Sent a RECALL do to that GET_LAYOUT, or
the RECALL and GET_LAYOUT crossed on the wire.
In any way this means we want to wait at most until in-flight IO
is finished and the RECALL can be satisfied.
So a proper wait here is more like 1/10 of a second, not 15 seconds
like we have now. In case of a server bug we delay exponentially
longer on each retry.
Current code totally craps out performance of very large files on
most pnfs-objects layouts, because of how the map changes when the
file has grown into the next raid group.
[Stable: This will patch back to 3.9. If there are earlier still
maintained trees, please tell me I'll send a patch]
If clp is new (cl_count = 1) and it matches another client in
nfs4_discover_server_trunking, the nfs_put_client will free clp before
->cl_preserve_clid is set.
Both nfs41_walk_client_list and nfs40_walk_client_list expect the
'status' variable to be set to the value -NFS4ERR_STALE_CLIENTID
if the loop fails to find a match.
The problem is that the 'pos->cl_cons_state > NFS_CS_READY' changes
the value of 'status', and sets it either to the value '0' (which
indicates success), or to the value EINTR.
decode_op_hdr() cannot distinguish between an XDR decoding error and
the perfectly valid errorcode NFS4ERR_IO. This is normally not a
problem, but for the particular case of OPEN, we need to be able
to increment the NFSv4 open sequence id when the server returns
a valid response.
This corrects a problem in spi_pump_messages() that leads to an spi
message hanging forever when a call to transfer_one_message() fails.
This failure occurs in my MCP2210 driver when the cs_change bit is set
on the last transfer in a message, an operation which the hardware does
not support.
Rationale
Since the transfer_one_message() returns an int, we must presume that it
may fail. If transfer_one_message() should never fail, it should return
void. Thus, calls to transfer_one_message() should properly manage a
failure.
Fixes: ffbbdd21329f3 (spi: create a message queueing infrastructure) Signed-off-by: Daniel Santos <daniel.santos@pobox.com> Signed-off-by: Mark Brown <broonie@linaro.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
At IO preparation we calculate the max pages at each device and
allocate a BIO per device of that size. The calculation was wrong
on some unaligned corner cases offset/length combination and would
make prepare return with -ENOMEM. This would be bad for pnfs-objects
that would in that case IO through MDS. And fatal for exofs were it
would fail writes with EIO.
Fix it by doing the proper math, that will work in all cases. (I
ran a test with all possible offset/length combinations this time
round).
Also when reading we do not need to allocate for the parity units
since we jump over them.
Also lower the max_io_length to take into account the parity pages
so not to allocate BIOs bigger than PAGE_SIZE
With commit d8d14bd09cdd ("fs/compat: fix lookup_dcookie() parameter
handling") I changed the type of the len parameter of the
lookup_dcookie() syscall.
However I missed that there was still a stale declaration in
arch/tile/.. which now causes a compile error on tile:
In file included from fs/dcookies.c:28:0:
include/linux/compat.h:425:17: error: conflicting types for 'compat_sys_lookup_dcookie'
fs/dcookies.c:207:1: error: conflicting types for 'compat_sys_lookup_dcookie'
Simply remove the declaration in the tile architecture, which is only a
leftover from before the different compat lookup_dcookie() versions have
been merged. The correct declaration is now in include/linux/compat.h
The build error was reported by Fenguang's build bot.
Commit d5dc77bfeeab ("consolidate compat lookup_dcookie()") coverted all
architectures to the new compat_sys_lookup_dcookie() syscall.
The "len" paramater of the new compat syscall must have the type
compat_size_t in order to enforce zero extension for architectures where
the ABI requires that the caller of a function performed zero and/or
sign extension to 64 bit of all parameters.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Ingo Molnar <mingo@kernel.org> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Hendrik Brueckner <brueckner@linux.vnet.ibm.com> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
We got a report that the pwritev syscall does not work correctly in
compat mode on s390.
It turned out that with commit 72ec35163f9f ("switch compat readv/writev
variants to COMPAT_SYSCALL_DEFINE") we lost the zero extension of a
couple of syscall parameters because the some parameter types haven't
been converted from unsigned long to compat_ulong_t.
This is needed for architectures where the ABI requires that the caller
of a function performed zero and/or sign extension to 64 bit of all
parameters.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Ingo Molnar <mingo@kernel.org> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Hendrik Brueckner <brueckner@linux.vnet.ibm.com> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Commit 91c2e0bcae72 ("unify compat fanotify_mark(2), switch to
COMPAT_SYSCALL_DEFINE") added a new unified compat fanotify_mark syscall
to be used by all architectures.
Unfortunately the unified version merges the split mask parameter in a
wrong way: the lower and higher word got swapped.
This was discovered with glibc's tst-fanotify test case.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com> Reported-by: Andreas Krebbel <krebbel@linux.vnet.ibm.com> Cc: "James E.J. Bottomley" <jejb@parisc-linux.org> Acked-by: "David S. Miller" <davem@davemloft.net> Acked-by: Al Viro <viro@ZenIV.linux.org.uk> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Ralf Baechle <ralf@linux-mips.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
There is currently no facility in ACPI to express the hookup of voltage
regulators, the expectation is that the regulators that exist in the
system will be handled transparently by firmware if they need software
control at all. This means that if for some reason the regulator API is
enabled on such a system it should assume that any supplies that devices
need are provided by the system at all relevant times without any software
intervention.
Tell the regulator core to make this assumption by calling
regulator_has_full_constraints(). Do this as soon as we know we are using
ACPI so that the information is available to the regulator core as early
as possible. This will cause the regulator core to pretend that there is
an always on regulator supplying any supply that is requested but that has
not otherwise been mapped which is the behaviour expected on a system with
ACPI.
Should the ability to specify regulators be added in future revisions of
ACPI then once we have support for ACPI mappings in the kernel the same
assumptions will apply. It is also likely that systems will default to a
mode of operation which does not require any interpretation of these
mappings in order to be compatible with existing operating system releases
so it should remain safe to make these assumptions even if the mappings
exist but are not supported by the kernel.
Signed-off-by: Mark Brown <broonie@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
turbostat uses inline assembly to call cpuid. On 32-bit x86, on systems
that have certain security features enabled by default that make -fPIC
the default, this causes a build error:
turbostat.c: In function ‘check_cpuid’:
turbostat.c:1906:2: error: PIC register clobbered by ‘ebx’ in ‘asm’
asm("cpuid" : "=a" (fms), "=c" (ecx), "=d" (edx) : "a" (1) : "ebx");
^
GCC provides a header cpuid.h, containing a __get_cpuid function that
works with both PIC and non-PIC. (On PIC, it saves and restores ebx
around the cpuid instruction.) Use that instead.
Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
turbostat's Makefile puts arch/x86/include/uapi/ in the include path, so
that it can include <asm/msr.h> from it. It isn't in general safe to
include even uapi headers directly from the kernel tree without
processing them through scripts/headers_install.sh, but asm/msr.h
happens to work.
However, that include path can break with some versions of system
headers, by overriding some system headers with the unprocessed versions
directly from the kernel source. For instance:
In file included from /build/x86-generic/usr/include/bits/sigcontext.h:28:0,
from /build/x86-generic/usr/include/signal.h:339,
from /build/x86-generic/usr/include/sys/wait.h:31,
from turbostat.c:27:
../../../../arch/x86/include/uapi/asm/sigcontext.h:4:28: fatal error: linux/compiler.h: No such file or directory
This occurs because the system bits/sigcontext.h on that build system
includes <asm/sigcontext.h>, and asm/sigcontext.h in the kernel source
includes <linux/compiler.h>, which scripts/headers_install.sh would have
filtered out.
Since turbostat really only wants a single header, just include that one
header rather than putting an entire directory of kernel headers on the
include path.
In the process, switch from msr.h to msr-index.h, since turbostat just
wants the MSR numbers.
Signed-off-by: Josh Triplett <josh@joshtriplett.org> Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
We should use page->pages instead of page->pobjects when calculating
the number of cpu partial slabs. This also fixes the mapping of slabs
and nodes.
As there's no variable storing the number of total/active objects in
cpu partial slabs, and we don't have user interfaces requiring those
statistics, I just add WARN_ON for those cases.
Acked-by: Christoph Lameter <cl@linux.com> Reviewed-by: Wanpeng Li <liwanp@linux.vnet.ibm.com> Signed-off-by: Li Zefan <lizefan@huawei.com> Signed-off-by: Pekka Enberg <penberg@kernel.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
With some SDIO devices, timeout errors can happen when reading data.
To solve this issue, the DMA transfer has to be activated before sending
the command to the device. This order is incorrect in PDC mode. So we
have to take care if we are using DMA or PDC to know when to send the
MMC command.
Signed-off-by: Ludovic Desroches <ludovic.desroches@atmel.com> Acked-by: Nicolas Ferre <nicolas.ferre@atmel.com> Signed-off-by: Chris Ball <cjb@laptop.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Under function mmc_blk_issue_rq, after an MMC discard operation,
the MMC request data structure may be freed in memory. Later in
the same function, the check of req->cmd_flags & MMC_REQ_SPECIAL_MASK
is dangerous and invalid. It causes the MMC host not to be released
when it should.
This patch fixes the issue by marking the special request down before
the discard/flush operation.
Reported by: Harold (SoonYeal) Yang <haroldsy@broadcom.com> Signed-off-by: Ray Jui <rjui@broadcom.com> Reviewed-by: Seungwon Jeon <tgih.jun@samsung.com> Acked-by: Seungwon Jeon <tgih.jun@samsung.com> Signed-off-by: Chris Ball <cjb@laptop.org> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The VM is currently heavily tuned to avoid swapping. Whether that is
good or bad is a separate discussion, but as long as the VM won't swap
to make room for dirty cache, we can not consider anonymous pages when
calculating the amount of dirtyable memory, the baseline to which
dirty_background_ratio and dirty_ratio are applied.
A simple workload that occupies a significant size (40+%, depending on
memory layout, storage speeds etc.) of memory with anon/tmpfs pages and
uses the remainder for a streaming writer demonstrates this problem. In
that case, the actual cache pages are a small fraction of what is
considered dirtyable overall, which results in an relatively large
portion of the cache pages to be dirtied. As kswapd starts rotating
these, random tasks enter direct reclaim and stall on IO.
Only consider free pages and file pages dirtyable.
Tejun reported stuttering and latency spikes on a system where random
tasks would enter direct reclaim and get stuck on dirty pages. Around
50% of memory was occupied by tmpfs backed by an SSD, and another disk
(rotating) was reading and writing at max speed to shrink a partition.
: The problem was pretty ridiculous. It's a 8gig machine w/ one ssd and 10k
: rpm harddrive and I could reliably reproduce constant stuttering every
: several seconds for as long as buffered IO was going on on the hard drive
: either with tmpfs occupying somewhere above 4gig or a test program which
: allocates about the same amount of anon memory. Although swap usage was
: zero, turning off swap also made the problem go away too.
:
: The trigger conditions seem quite plausible - high anon memory usage w/
: heavy buffered IO and swap configured - and it's highly likely that this
: is happening in the wild too. (this can happen with copying large files
: to usb sticks too, right?)
This patch (of 2):
The dirty_balance_reserve is an approximation of the fraction of free
pages that the page allocator does not make available for page cache
allocations. As a result, it has to be taken into account when
calculating the amount of "dirtyable memory", the baseline to which
dirty_background_ratio and dirty_ratio are applied.
However, currently the reserve is subtracted from the sum of free and
reclaimable pages, which is non-sensical and leads to erroneous results
when the system is dominated by unreclaimable pages and the
dirty_balance_reserve is bigger than free+reclaimable. In that case, at
least the already allocated cache should be considered dirtyable.
Fix the calculation by subtracting the reserve from the amount of free
pages, then adding the reclaimable pages on top.
After thp split in hwpoison_user_mappings(), we hold page lock on the
raw error page only between try_to_unmap, hence we are in danger of race
condition.
I found in the RHEL7 MCE-relay testing that we have "bad page" error
when a memory error happens on a thp tail page used by qemu-kvm:
Triggering MCE exception on CPU 10
mce: [Hardware Error]: Machine check events logged
MCE exception done on CPU 10
MCE 0x38c535: Killing qemu-kvm:8418 due to hardware memory corruption
MCE 0x38c535: dirty LRU page recovery: Recovered
qemu-kvm[8418]: segfault at 20 ip 00007ffb0f0f229a sp 00007fffd6bc5240 error 4 in qemu-kvm[7ffb0ef14000+420000]
BUG: Bad page state in process qemu-kvm pfn:38c400
page:ffffea000e310000 count:0 mapcount:0 mapping: (null) index:0x7ffae3c00
page flags: 0x2fffff0008001d(locked|referenced|uptodate|dirty|swapbacked)
Modules linked in: hwpoison_inject mce_inject vhost_net macvtap macvlan ...
CPU: 0 PID: 8418 Comm: qemu-kvm Tainted: G M -------------- 3.10.0-54.0.1.el7.mce_test_fixed.x86_64 #1
Hardware name: NEC NEC Express5800/R120b-1 [N8100-1719F]/MS-91E7-001, BIOS 4.6.3C19 02/10/2011
Call Trace:
dump_stack+0x19/0x1b
bad_page.part.59+0xcf/0xe8
free_pages_prepare+0x148/0x160
free_hot_cold_page+0x31/0x140
free_hot_cold_page_list+0x46/0xa0
release_pages+0x1c1/0x200
free_pages_and_swap_cache+0xad/0xd0
tlb_flush_mmu.part.46+0x4c/0x90
tlb_finish_mmu+0x55/0x60
exit_mmap+0xcb/0x170
mmput+0x67/0xf0
vhost_dev_cleanup+0x231/0x260 [vhost_net]
vhost_net_release+0x3f/0x90 [vhost_net]
__fput+0xe9/0x270
____fput+0xe/0x10
task_work_run+0xc4/0xe0
do_exit+0x2bb/0xa40
do_group_exit+0x3f/0xa0
get_signal_to_deliver+0x1d0/0x6e0
do_signal+0x48/0x5e0
do_notify_resume+0x71/0xc0
retint_signal+0x48/0x8c
The reason of this bug is that a page fault happens before unlocking the
head page at the end of memory_failure(). This strange page fault is
trying to access to address 0x20 and I'm not sure why qemu-kvm does
this, but anyway as a result the SIGSEGV makes qemu-kvm exit and on the
way we catch the bad page bug/warning because we try to free a locked
page (which was the former head page.)
To fix this, this patch suggests to shift page lock from head page to
tail page just after thp split. SIGSEGV still happens, but it affects
only error affected VMs, not a whole system.
audit_syscall_exit() saves a result of regs_return_value() in intermediate
"int" variable and passes it to __audit_syscall_exit(), which expects its
second argument as a "long" value. This will result in truncating the
value returned by a system call and making a wrong audit record.
I don't know why gcc compiler doesn't complain about this, but anyway it
causes a problem at runtime on arm64 (and probably most 64-bit archs).
Signed-off-by: AKASHI Takahiro <takahiro.akashi@linaro.org> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Eric Paris <eparis@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Eric Paris <eparis@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When the audit queue overflows and times out (audit_backlog_wait_time), the
audit queue overflow timeout is set to zero. Once the audit queue overflow
timeout condition recovers, the timeout should be reset to the original value.
See also:
https://lkml.org/lkml/2013/9/2/473
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com> Signed-off-by: Dan Duval <dan.duval@oracle.com> Signed-off-by: Chuck Anderson <chuck.anderson@oracle.com> Signed-off-by: Richard Guy Briggs <rgb@redhat.com> Signed-off-by: Eric Paris <eparis@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Having this struct in module memory could Oops when if the module is
unloaded while the buffer still persists in a pipe.
Since sock_pipe_buf_ops is essentially the same as fuse_dev_pipe_buf_steal
merge them into nosteal_pipe_buf_ops (this is the same as
default_pipe_buf_ops except stealing the page from the buffer is not
allowed).