On umount path, jbd2_journal_destroy() writes latest transaction ID
(->j_tail_sequence) to be used at next mount.
The bug is that ->j_tail_sequence is not holding latest transaction ID
in some cases. So, at next mount, there is chance to conflict with
remaining (not overwritten yet) transactions.
mount (id=10)
write transaction (id=11)
write transaction (id=12)
umount (id=10) <= the bug doesn't write latest ID
mount (id=10)
write transaction (id=11)
crash
mount
[recovery process]
transaction (id=11)
transaction (id=12) <= valid transaction ID, but old commit
must not replay
Like above, this bug become the cause of recovery failure, or FS
corruption.
So why ->j_tail_sequence doesn't point latest ID?
Because if checkpoint transactions was reclaimed by memory pressure
(i.e. bdev_try_to_free_page()), then ->j_tail_sequence is not updated.
(And another case is, __jbd2_journal_clean_checkpoint_list() is called
with empty transaction.)
So in above cases, ->j_tail_sequence is not pointing latest
transaction ID at umount path. Plus, REQ_FLUSH for checkpoint is not
done too.
So, to fix this problem with minimum changes, this patch updates
->j_tail_sequence, and issue REQ_FLUSH. (With more complex changes,
some optimizations would be possible to avoid unnecessary REQ_FLUSH
for example though.)
One of the strange things that the original sg driver did was let the
user provide both a data-out buffer (it followed the sg_header+cdb)
_and_ specify a reply length greater than zero. What happened was that
the user data-out buffer was copied into some kernel buffers and then
the mid level was told a read type operation would take place with the
data from the device overwriting the same kernel buffers. The user would
then read those kernel buffers back into the user space.
From what I can tell, the above action was broken by commit fad7f01e61bf
("sg: set dxferp to NULL for READ with the older SG interface") in 2008
and syzkaller found that out recently.
Make sure that a user space pointer is passed through when data follows
the sg_header structure and command. Fix the abnormal case when a
non-zero reply_len is also given.
Fixes: fad7f01e61bf737fe8a3740d803f000db57ecac6 Cc: <stable@vger.kernel.org> #v2.6.28+ Signed-off-by: Douglas Gilbert <dgilbert@interlog.com> Reviewed-by: Ewan Milne <emilne@redhat.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 685a50b681ddd07ff2b7714797b5793adcc691e7)
break_stripe_batch_list breaks up a batch and copies some flags from
the batch head to the members, preserving others.
It doesn't preserve or copy STRIPE_PREREAD_ACTIVE. This is not
normally a problem as STRIPE_PREREAD_ACTIVE is cleared when a
stripe_head is added to a batch, and is not set on stripe_heads
already in a batch.
However there is no locking to ensure one thread doesn't set the flag
after it has just been cleared in another. This does occasionally happen.
md/raid5 maintains a count of the number of stripe_heads with
STRIPE_PREREAD_ACTIVE set: conf->preread_active_stripes. When
break_stripe_batch_list clears STRIPE_PREREAD_ACTIVE inadvertently
this could becomes incorrect and will never again return to zero.
md/raid5 delays the handling of some stripe_heads until
preread_active_stripes becomes zero. So when the above mention race
happens, those stripe_heads become blocked and never progress,
resulting is write to the array handing.
So: change break_stripe_batch_list to preserve STRIPE_PREREAD_ACTIVE
in the members of a batch.
URL: https://bugzilla.kernel.org/show_bug.cgi?id=108741
URL: https://bugzilla.redhat.com/show_bug.cgi?id=1258153
URL: http://thread.gmane.org/5649C0E9.2030204@zoner.cz Reported-by: Martin Svec <martin.svec@zoner.cz> (and others) Tested-by: Tom Weber <linux@junkyard.4t2.com> Fixes: 1b956f7a8f9a ("md/raid5: be more selective about distributing flags across batch.") Cc: stable@vger.kernel.org (v4.1 and later) Signed-off-by: NeilBrown <neilb@suse.com> Signed-off-by: Shaohua Li <shli@fb.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 88e2df1b99cfc0087ccd2faa8ca61d7263fc007a)
The Home Agent and PCU PCI devices in Broadwell-EP have a non-BAR register
where a BAR should be. We don't know what the side effects of sizing the
"BAR" would be, and we don't know what address space the "BAR" might appear
to describe.
Mark these devices as having non-compliant BARs so the PCI core doesn't
touch them.
When bch_cache_set_alloc() fails to kzalloc the cache_set, the
asyncronous closure handling tries to dereference a cache_set that
hadn't yet been allocated inside of cache_set_flush() which is called
by __cache_set_unregister() during cleanup. This appears to happen only
during an OOM condition on bcache_register.
Signed-off-by: Eric Wheeler <bcache@linux.ewheeler.net> Cc: stable@vger.kernel.org Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit d09a05998d79dcfaa25de84624dce9f806fe4e7c)
Fix null pointer dereference by changing register_cache() to return an int
instead of being void. This allows it to return -ENOMEM or -ENODEV and
enables upper layers to handle the OOM case without NULL pointer issues.
See this thread:
http://thread.gmane.org/gmane.linux.kernel.bcache.devel/3521
Fixes this error:
gargamel:/sys/block/md5/bcache# echo /dev/sdh2 > /sys/fs/bcache/register
The bch_writeback_thread might BUG_ON in read_dirty() if
dc->sb==BDEV_STATE_DIRTY and bch_sectors_dirty_init has not yet completed
its related initialization. This patch downs the dc->writeback_lock until
after initialization is complete, thus preventing bch_writeback_thread
from proceeding prematurely.
See this thread:
http://thread.gmane.org/gmane.linux.kernel.bcache.devel/3453
Signed-off-by: Eric Wheeler <bcache@linux.ewheeler.net> Tested-by: Marc MERLIN <marc@merlins.org> Cc: <stable@vger.kernel.org> Signed-off-by: Jens Axboe <axboe@fb.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 7269554a57352f66aefb3e85cb7e11c4b63bba59)
The callers of steal_account_process_tick() expect it to return
whether a jiffy should be considered stolen or not.
Currently the return value of steal_account_process_tick() is in
units of cputime, which vary between either jiffies or nsecs
depending on CONFIG_VIRT_CPU_ACCOUNTING_GEN.
If cputime has nsecs granularity and there is a tiny amount of
stolen time (a few nsecs, say) then we will consider the entire
tick stolen and will not account the tick on user/system/idle,
causing /proc/stats to show invalid data.
The fix is to change steal_account_process_tick() to accumulate
the stolen time and only account it once it's worth a jiffy.
(Thanks to Frederic Weisbecker for suggestions to fix a bug in my
first version of the patch.)
Signed-off-by: Chris Friesen <chris.friesen@windriver.com> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Reviewed-by: Thomas Gleixner <tglx@linutronix.de> Cc: <stable@vger.kernel.org> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/56DBBDB8.40305@mail.usask.ca Signed-off-by: Ingo Molnar <mingo@kernel.org> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 23745ba7ffac8cffbe648812fe7dc485d6df9404)
early_init_dt_alloc_reserved_memory_arch passes end as 0 to
__memblock_alloc_base, when limits are not specified. But
__memblock_alloc_base takes end value of 0 as MEMBLOCK_ALLOC_ACCESSIBLE
and limits the end to memblock.current_limit. This results in regions
never being placed in HIGHMEM area, for e.g. CMA.
Let __memblock_alloc_base allocate from anywhere in memory if limits are
not specified.
Acked-by: Marek Szyprowski <m.szyprowski@samsung.com> Signed-off-by: Vinayak Menon <vinmenon@codeaurora.org> Cc: stable@vger.kernel.org Signed-off-by: Rob Herring <robh@kernel.org> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 444cf5487d5f51a3ecce2a0dfe237156290dfc7f)
Allow device initialization to finish gracefully when it is in
FTL rebuild failure state. Also, recover device out of this state
after successfully secure erasing it.
Signed-off-by: Selvan Mani <smani@micron.com> Signed-off-by: Vignesh Gunasekaran <vgunasekaran@micron.com> Signed-off-by: Asai Thambi S P <asamymuthupa@micron.com> Cc: stable@vger.kernel.org Signed-off-by: Jens Axboe <axboe@fb.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 65963ead8aefa685ec2e22d403461101c243683e)
pci and block layers have changed a lot compared to when SRSI support was added.
Given the current state of pci and block layers, this driver do not have to do
any specific handling.
Signed-off-by: Asai Thambi S P <asamymuthupa@micron.com> Signed-off-by: Selvan Mani <smani@micron.com> Signed-off-by: Jens Axboe <axboe@fb.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 13af0df20f8e78dc1e7239a29b2862addde3953e)
When dqget() in __dquot_initialize() fails e.g. due to IO error,
__dquot_initialize() will pass an array of uninitialized pointers to
dqput_all() and thus can lead to deference of random data. Fix the
problem by properly initializing the array.
CC: stable@vger.kernel.org Signed-off-by: Nikolay Borisov <kernel@kyup.com> Signed-off-by: Jan Kara <jack@suse.cz> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit ab1cc52b3f62f2445c60cbe390d26c50ebc0f3bd)
A number of spots in the xdr decoding follow a pattern like
n = be32_to_cpup(p++);
READ_BUF(n + 4);
where n is a u32. The only bounds checking is done in READ_BUF itself,
but since it's checking (n + 4), it won't catch cases where n is very
large, (u32)(-4) or higher. I'm not sure exactly what the consequences
are, but we've seen crashes soon after.
Instead, just break these up into two READ_BUF()s.
Cc: stable@vger.kernel.org Signed-off-by: J. Bruce Fields <bfields@redhat.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit d876f71611ad9b720cc890075b3c4bec25bd54b5)
Let the target core check task existence instead of the SRP target
driver. Additionally, let the target core check the validity of the
task management request instead of the ib_srpt driver.
There are still a couple of minor issues in the X.509 leap year handling:
(1) To avoid doing a modulus-by-400 in addition to a modulus-by-100 when
determining whether the year is a leap year or not, I divided the year
by 100 after doing the modulus-by-100, thereby letting the compiler do
one instruction for both, and then did a modulus-by-4.
Unfortunately, I then passed the now-modified year value to mktime64()
to construct a time value.
Since this isn't a fast path and since mktime64() does a bunch of
divisions, just condense down to "% 400". It's also easier to read.
(2) The default month length for any February where the year doesn't
divide by four exactly is obtained from the month_length[] array where
the value is 29, not 28.
This is fixed by altering the table.
Reported-by: Rudolf Polzer <rpolzer@google.com> Signed-off-by: David Howells <dhowells@redhat.com> Acked-by: David Woodhouse <David.Woodhouse@intel.com> Acked-by: Arnd Bergmann <arnd@arndb.de>
cc: stable@vger.kernel.org Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit e62c5259a62f3da2a911f8fe6275dbf43d3b624f)
Make the X.509 ASN.1 time object decoder fill in a time64_t rather than a
struct tm to make comparison easier (unfortunately, this makes readable
display less easy) and export it so that it can be used by the PKCS#7 code
too.
Further, tighten up its parsing to reject invalid dates (eg. weird
characters, non-existent hour numbers) and unsupported dates (eg. timezones
other than 'Z' or dates earlier than 1970).
Signed-off-by: David Howells <dhowells@redhat.com> Reviewed-by: David Woodhouse <David.Woodhouse@intel.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit f85d91f88486b34679c532a5687466eaf335258f)
Extract both parts of the AuthorityKeyIdentifier, not just the keyIdentifier,
as the second part can be used to match X.509 certificates by issuer and
serialNumber.
Signed-off-by: David Howells <dhowells@redhat.com> Tested-by: Vivek Goyal <vgoyal@redhat.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 3ccbbbf7b5ceb75bef47692c39f443a3efb38437)
Since a crypto_ahash_import() can be called against a request context
that has not had a crypto_ahash_init() performed, the request context
needs to be cleared to insure there is no random data present. If not,
the random data can result in a kernel oops during crypto_ahash_update().
Cc: <stable@vger.kernel.org> # 3.14.x- Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit dad41d54081e1bd2ef601c702ff4ea0f7428a965)
check_reshape() is called from raid5d thread. raid5d thread shouldn't
call mddev_suspend(), because mddev_suspend() waits for all IO finish
but IO is handled in raid5d thread, we could easily deadlock here.
This issue is introduced by 738a273 ("md/raid5: fix allocation of 'scribble' array.")
Cc: stable@vger.kernel.org (v4.1+) Reported-and-tested-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com> Reviewed-by: NeilBrown <neilb@suse.com> Signed-off-by: Shaohua Li <shli@fb.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 503f8305ab1b82d8788a2f161e7c52c0c0f6aeac)
'max_discard_sectors' is in sectors, while 'stripe' is in bytes.
This fixes the problem where DISCARD would get disabled on some larger
RAID5 configurations (6 or more drives in my testing), while it worked
as expected with smaller configurations.
Fixes: 620125f2bf8 ("MD: raid5 trim support") Cc: stable@vger.kernel.org v3.7+ Signed-off-by: Jes Sorensen <Jes.Sorensen@redhat.com> Signed-off-by: Shaohua Li <shli@fb.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 5255a738ee6ecc0e479728efe5668efd64901197)
Commit bb44fa317be6c1dd0650feaae3326ac11f2d37a4 ("PCI: Add
dev->has_secondary_link to track downstream PCIe links") and commit 0af1534fb7b3dba6e11d6c2670912725a2087217 ("PCI: Disable IO/MEM decoding
for devices with non-compliant BARs") added one-bit flags to the
pci_dev structure. Technically, this broke kABI (according to the
checker, anyway). In reality, these bits were added after a bunch
of other one-bit fields, the result being that their addition didn't
extend the size of the structure, nor did it change the offsets of
any existing fields of the structure.
This commit simply wraps these two fields in "#ifndef __GENKSYMS__"
to hide them from the checker.
The PCI config header (first 64 bytes of each device's config space) is
defined by the PCI spec so generic software can identify the device and
manage its usage of I/O, memory, and IRQ resources.
Some non-spec-compliant devices put registers other than BARs where the
BARs should be. When the PCI core sizes these "BARs", the reads and writes
it does may have unwanted side effects, and the "BAR" may appear to
describe non-sensical address space.
Add a flag bit to mark non-compliant devices so we don't touch their BARs.
Turn off IO/MEM decoding to prevent the devices from consuming address
space, since we can't read the BARs to find out what that address space
would be.
A PCIe Port is an interface to a Link. A Root Port is a PCI-PCI bridge in
a Root Complex and has a Link on its secondary (downstream) side. For
other Ports, the Link may be on either the upstream (closer to the Root
Complex) or downstream side of the Port.
The usual topology has a Root Port connected to an Upstream Port. We
previously assumed this was the only possible topology, and that a
Downstream Port's Link was always on its downstream side, like this:
+---------------------+
+------+ | Downstream |
| Root | | Upstream Port +--Link--
| Port +--Link--+ Port |
+------+ | Downstream |
| Port +--Link--
+---------------------+
But systems do exist (see URL below) where the Root Port is connected to a
Downstream Port. In this case, a Downstream Port's Link may be on either
the upstream or downstream side:
+---------------------+
+------+ | Upstream |
| Root | | Downstream Port +--Link--
| Port +--Link--+ Port |
+------+ | Downstream |
| Port +--Link--
+---------------------+
We can't use the Port type to determine which side the Link is on, so add a
bit in struct pci_dev to keep track.
A Root Port's Link is always on the Port's secondary side. A component
(Endpoint or Port) on the other end of the Link obviously has the Link on
its upstream side. If that component is a Port, it is part of a Switch or
a Bridge. A Bridge has a PCI or PCI-X bus on its secondary side, not a
Link. The internal bus of a Switch connects the Port to another Port whose
Link is on the downstream side.
Commit 5942ddbc500d ("mtd: introduce mtd_block_markbad interface")
incorrectly changed onenand_block_markbad() to call mtd_block_markbad
instead of onenand_chip's block_markbad function. As a result the function
will now recurse and deadlock. Fix by reverting the change.
Fixes: 5942ddbc500d ("mtd: introduce mtd_block_markbad interface") Signed-off-by: Aaro Koskinen <aaro.koskinen@iki.fi> Acked-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Cc: <stable@vger.kernel.org> Signed-off-by: Brian Norris <computersforpeace@gmail.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 084b44e9cbc6eec0b0b13138918fb5935208f3b4)
We were setting the queue depth correctly, then setting it back to
two. If you hit this as a bisection point then please send me an email
as it would imply we've been hiding other bugs with this one.
Cc: <stable@vger.kernel.org> Signed-off-by: Alan Cox <alan@linux.intel.com> Reviewed-by: Hannes Reinicke <hare@suse.de> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit cf438ddac48b27e7de0514f96d912e132c908df4)
aac_fib_map_free() calls pci_free_consistent() without checking that
dev->hw_fib_va is not NULL and dev->max_fib_size is not zero.If they are
indeed NULL/0, this will result in a hang as pci_free_consistent() will
attempt to invalidate cache for the entire 64-bit address space
(which would take a very long time).
Fixed by adding a check to make sure that dev->hw_fib_va and
dev->max_fib_size are not NULL and 0 respectively.
Fixes: 9ad5204d6 - "[SCSI]aacraid: incorrect dma mapping mask during blinked recover or user initiated reset" Cc: stable@vger.kernel.org Signed-off-by: Raghava Aditya Renukunta <raghavaaditya.renukunta@pmcs.com> Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de> Reviewed-by: Tomas Henzl <thenzl@redhat.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 6d2cd58f288e320c5a023219ac8726b30657ddbb)
The driver utilizes an array of atomic variables to keep track of IO
submissions to each vector. To submit an IO multiple threads iterate
through the array to find a vector which has empty slots to send an
IO. The reading and updating of the variable is not atomic, causing race
conditions when a thread uses a full vector to submit an IO.
Fixed by mapping each FIB to a vector, the submission path then uses
said vector to submit IO thereby removing the possibly of a race
condition.The vector assignment is started from 1 since vector 0 is
reserved for the use of AIF management FIBS.If the number of MSIx
vectors is 1 (MSI or INTx mode) then all the fibs are allocated to
vector 0.
Fixes: 495c0217 "aacraid: MSI-x support" Cc: stable@vger.kernel.org # v4.1 Signed-off-by: Raghava Aditya Renukunta <raghavaaditya.renukunta@pmcs.com> Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de> Reviewed-by: Tomas Henzl <thenzl@redhat.com> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 34bb7098221c2a37f5d9ef3591971a25719ae21a)
Request-based DM's blk-mq support (dm-mq) was reported to be 50% slower
than if an underlying null_blk device were used directly. One of the
reasons for this drop in performance is that blk_insert_clone_request()
was calling blk_mq_insert_request() with @async=true. This forced the
use of kblockd_schedule_delayed_work_on() to run the blk-mq hw queues
which ushered in ping-ponging between process context (fio in this case)
and kblockd's kworker to submit the cloned request. The ftrace
function_graph tracer showed:
Fixing blk_insert_clone_request()'s blk_mq_insert_request() call to
_not_ use kblockd to submit the cloned requests isn't enough to
eliminate the observed context switches.
In addition to this dm-mq specific blk-core fix, there are 2 DM core
fixes to dm-mq that (when paired with the blk-core fix) completely
eliminate the observed context switching:
1) don't blk_mq_run_hw_queues in blk-mq request completion
Motivated by desire to reduce overhead of dm-mq, punting to kblockd
just increases context switches.
In my testing against a really fast null_blk device there was no benefit
to running blk_mq_run_hw_queues() on completion (and no other blk-mq
driver does this). So hopefully this change doesn't induce the need for
yet another revert like commit 621739b00e16ca2d !
2) use blk_mq_complete_request() in dm_complete_request()
blk_complete_request() doesn't offer the traditional q->mq_ops vs
.request_fn branching pattern that other historic block interfaces
do (e.g. blk_get_request). Using blk_mq_complete_request() for
blk-mq requests is important for performance. It should be noted
that, like blk_complete_request(), blk_mq_complete_request() doesn't
natively handle partial completions -- but the request-based
DM-multipath target does provide the required partial completion
support by dm.c:end_clone_bio() triggering requeueing of the request
via dm-mpath.c:multipath_end_io()'s return of DM_ENDIO_REQUEUE.
dm-mq fix #2 is _much_ more important than #1 for eliminating the
context switches.
Before: cpu : usr=15.10%, sys=59.39%, ctx=7905181, majf=0, minf=475
After: cpu : usr=20.60%, sys=79.35%, ctx=2008, majf=0, minf=472
With these changes multithreaded async read IOPs improved from ~950K
to ~1350K for this dm-mq stacked on null_blk test-case. The raw read
IOPs of the underlying null_blk device for the same workload is ~1950K.
Fixes: 7fb4898e0 ("block: add blk-mq support to blk_insert_cloned_request()") Fixes: bfebd1cdb ("dm: add full blk-mq support to request-based DM") Cc: stable@vger.kernel.org # 4.1+ Reported-by: Sagi Grimberg <sagig@dev.mellanox.co.il> Signed-off-by: Mike Snitzer <snitzer@redhat.com> Acked-by: Jens Axboe <axboe@kernel.dk> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit fb1840e257d26b6aad69b864cb7ed4a67ab986b6)
In commit bcff24887d00 ("ext4: don't read blocks from disk after extents
being swapped") bh is not updated correctly in the for loop and wrong
data has been written to disk. generic/324 catches this on sub-page
block size ext4.
Fixes: bcff24887d00 ("ext4: don't read blocks from disk after extentsbeing swapped") Signed-off-by: Eryu Guan <guaneryu@gmail.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu> Cc: stable@vger.kernel.org Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 31a37d7c7ef7b4f90b600fcddd1c385a39f9d34c)
The N_IRDA line discipline may access the previous line discipline's closed
and already-fre private data on open [1].
The tty->disc_data field _never_ refers to valid data on entry to the
line discipline's open() method. Rather, the ldisc is expected to
initialize that field for its own use for the lifetime of the instance
(ie. from open() to close() only).
[1]
==================================================================
BUG: KASAN: use-after-free in irtty_open+0x422/0x550 at addr ffff8800331dd068
Read of size 4 by task a.out/13960
=============================================================================
BUG kmalloc-512 (Tainted: G B ): kasan: bad access detected
-----------------------------------------------------------------------------
...
Call Trace:
[<ffffffff815fa2ae>] __asan_report_load4_noabort+0x3e/0x40 mm/kasan/report.c:279
[<ffffffff836938a2>] irtty_open+0x422/0x550 drivers/net/irda/irtty-sir.c:436
[<ffffffff829f1b80>] tty_ldisc_open.isra.2+0x60/0xa0 drivers/tty/tty_ldisc.c:447
[<ffffffff829f21c0>] tty_set_ldisc+0x1a0/0x940 drivers/tty/tty_ldisc.c:567
[< inline >] tiocsetd drivers/tty/tty_io.c:2650
[<ffffffff829da49e>] tty_ioctl+0xace/0x1fd0 drivers/tty/tty_io.c:2883
[< inline >] vfs_ioctl fs/ioctl.c:43
[<ffffffff816708ac>] do_vfs_ioctl+0x57c/0xe60 fs/ioctl.c:607
[< inline >] SYSC_ioctl fs/ioctl.c:622
[<ffffffff81671204>] SyS_ioctl+0x74/0x80 fs/ioctl.c:613
[<ffffffff852a7876>] entry_SYSCALL_64_fastpath+0x16/0x7a
Use a local variable for the exported and imported state so that
alignment is not an issue. On export, set a local variable from the
request context and then memcpy the contents of the local variable to
the export memory area. On import, memcpy the import memory area into
a local variable and then use the local variable to set the request
context.
Cc: <stable@vger.kernel.org> # 3.14.x- Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit b053d66b66e702f74ddc986863b30eaac41e7f4c)
Since the exported information can be exposed to user-space, instead of
exporting the entire request context only export the minimum information
needed.
Cc: <stable@vger.kernel.org> # 3.14.x- Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 5badf7e00f0968d820bf7bba9081339bfca3489c)
commit 9ce119f318ba ("tty: Fix GPF in flush_to_ldisc()") fixed a
GPF caused by a line discipline which does not define a receive_buf()
method.
However, the vt driver (and speakup driver also) pushes selection
data directly to the line discipline receive_buf() method via
tty_ldisc_receive_buf(). Fix the same problem in tty_ldisc_receive_buf().
Cc: <stable@vger.kernel.org> Signed-off-by: Peter Hurley <peter@hurleysoftware.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 3beac2b0bffd8e977ef1d464dd424f22af965a96)
Commit 8996eafdcbad ("crypto: ahash - ensure statesize is non-zero")
added a check to prevent ahash algorithms from successfully registering
if the import and export functions were not implemented. This prevents
an oops in the hash_accept function of algif_hash. This commit causes
the ccp-crypto module SHA support and AES CMAC support from successfully
registering and causing the ccp-crypto module load to fail because the
ahash import and export functions are not implemented.
Update the CCP Crypto API support to provide import and export support
for ahash algorithms.
Cc: <stable@vger.kernel.org> # 3.14.x- Signed-off-by: Tom Lendacky <thomas.lendacky@amd.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 1ad241d40ef8d9a50ced5c35bf55e9a21a997516)
The usbvision driver crashes when a specially crafted usb device with invalid
number of interfaces or endpoints is detected. This fix adds checks that the
device has proper configuration expected by the driver.
Poison pointer values should be small enough to find a room in
non-mmap'able/hardly-mmap'able space. E.g. on x86 "poison pointer space"
is located starting from 0x0. Given unprivileged users cannot mmap
anything below mmap_min_addr, it should be safe to use poison pointers
lower than mmap_min_addr.
The current poison pointer values of LIST_POISON{1,2} might be too big for
mmap_min_addr values equal or less than 1 MB (common case, e.g. Ubuntu
uses only 0x10000). There is little point to use such a big value given
the "poison pointer space" below 1 MB is not yet exhausted. Changing it
to a smaller value solves the problem for small mmap_min_addr setups.
The values are suggested by Solar Designer:
http://www.openwall.com/lists/oss-security/2015/05/02/6
Signed-off-by: Vasily Kulikov <segoon@openwall.com> Cc: Solar Designer <solar@openwall.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 46460a03f44f1915ded434057fa46332438b3a6e)
If a user key gets negatively instantiated, an error code is cached in the
payload area. A negatively instantiated key may be then be positively
instantiated by updating it with valid data. However, the ->update key
type method must be aware that the error code may be there.
The following may be used to trigger the bug in the user key type:
keyctl request2 user user "" @u
keyctl add user user "a" @u
keyctl request2 trusted user "" @u
keyctl add trusted user "a" @u
This should also affect encrypted keys - but that has to be correctly
parameterised or it will fail with EINVAL before getting to the bit that
will crashes.
Reported-by: Dmitry Vyukov <dvyukov@google.com> Signed-off-by: David Howells <dhowells@redhat.com> Acked-by: Mimi Zohar <zohar@linux.vnet.ibm.com> Signed-off-by: James Morris <james.l.morris@oracle.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit d979e967f848caf908a1401b7ad67cf13f06ef9f)
Currently if userspace restores the pit counters with a count of 0
on channels 1 or 2 and the guest attempts to read the count on those
channels, then KVM will perform a mod of 0 and crash. This will ensure
that 0 values are converted to 65536 as per the spec.
This is CVE-2015-7513.
Signed-off-by: Andy Honig <ahonig@google.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 90352f3f473a29db1289ec31facc1ac18cc66e9e)
[Al Viro] The bug is in being too enthusiastic about optimizing ->setattr()
away - instead of "copy verbatim with metadata" + "chmod/chown/utimes"
(with the former being always safe and the latter failing in case of
insufficient permissions) it tries to combine these two. Note that copyup
itself will have to do ->setattr() anyway; _that_ is where the elevated
capabilities are right. Having these two ->setattr() (one to set verbatim
copy of metadata, another to do what overlayfs ->setattr() had been asked
to do in the first place) combined is where it breaks.
Signed-off-by: Miklos Szeredi <miklos@szeredi.hu> Cc: <stable@vger.kernel.org> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 2cadb57dff500076a87b934cac64bb5a2293b644)
Parameter of trace_btrfs_work_queued() can be freed in its workqueue.
So no one use use that pointer after queue_work().
Fix the user-after-free bug by move the trace line before queue_work().
Reported-by: Dave Jones <davej@codemonkey.org.uk> Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: Chris Mason <clm@fb.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit b9a54ed91c7bbd5c18a4170be078d9f7e28560ed)
I see no_space in v4.4-rc1 again in xfstests generic/102.
It happened randomly in some node only.
(one of 4 phy-node, and a kvm with non-virtio block driver)
By bisect, we can found the first-bad is:
commit bdced438acd8 ("block: setup bi_phys_segments after splitting")'
But above patch only triggered the bug by making bio operation
faster(or slower).
Main reason is in our space_allocating code, we need to commit
page writeback before wait it complish, this patch fixed above
bug.
BTW, there is another reason for generic/102 fail, caused by
disable default mixed-blockgroup, I'll fix it in xfstests.
Signed-off-by: Zhao Lei <zhaolei@cn.fujitsu.com> Signed-off-by: Chris Mason <clm@fb.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit d5b55a7aae08c0e0785430126bcc4a9ae7f5c737)
btrfs will report no_space when we run following write and delete
file loop:
# FILE_SIZE_M=[ 75% of fs space ]
# DEV=[ some dev ]
# MNT=[ some dir ]
#
# mkfs.btrfs -f "$DEV"
# mount -o nodatacow "$DEV" "$MNT"
# for ((i = 0; i < 100; i++)); do dd if=/dev/zero of="$MNT"/file0 bs=1M count="$FILE_SIZE_M"; rm -f "$MNT"/file0; done
#
Reason:
iput() and evict() is run after write pages to block device, if
write pages work is not finished before next write, the "rm"ed space
is not freed, and caused above bug.
Fix:
We can add "-o flushoncommit" mount option to avoid above bug, but
it have performance problem. Actually, we can to wait for on-the-fly
writes only when no-space happened, it is which this patch do.
Signed-off-by: Zhao Lei <zhaolei@cn.fujitsu.com> Signed-off-by: Chris Mason <clm@fb.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 42bd8f4fda813558c3045c60ad6436b1c7430ec7)
It looks like smack and yama weren't aware that the ptrace mode
can have flags ORed into it - PTRACE_MODE_NOAUDIT until now, but
only for /proc/$pid/stat, and with the PTRACE_MODE_*CREDS patch,
all modes have flags ORed into them.
Signed-off-by: Jann Horn <jann@thejh.net> Acked-by: Kees Cook <keescook@chromium.org> Acked-by: Casey Schaufler <casey@schaufler-ca.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: James Morris <james.l.morris@oracle.com> Cc: "Serge E. Hallyn" <serge.hallyn@ubuntu.com> Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Cc: Andy Lutomirski <luto@kernel.org> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: "Eric W. Biederman" <ebiederm@xmission.com> Cc: Willy Tarreau <w@1wt.eu> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit ee6ad435c9872610c5a52cad02e331951cf2fb25)
This doesn't seem to fix a regression -- I don't think the CLAC was
ever there.
I double-checked in a debugger: entries through the int80 gate do
not automatically clear AC.
Stable maintainers: I can provide a backport to 4.3 and earlier if
needed. This needs to be backported all the way to 3.10.
Reported-by: Brian Gerst <brgerst@gmail.com> Signed-off-by: Andy Lutomirski <luto@kernel.org> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Borislav Petkov <bp@alien8.de> Cc: Denys Vlasenko <dvlasenk@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: <stable@vger.kernel.org> # v3.10 and later Fixes: 63bcff2a307b ("x86, smap: Add STAC and CLAC instructions to control user space access") Link: http://lkml.kernel.org/r/b02b7e71ae54074be01fc171cbd4b72517055c0e.1456345086.git.luto@kernel.org Signed-off-by: Ingo Molnar <mingo@kernel.org> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 1f9780e372264c0cd71571c94da08cc49ae327e3)
In __request_region, if a conflict with a BUSY and MUXED resource is
detected, then the caller goes to sleep and waits for the resource to be
released. A pointer on the conflicting resource is kept. At wake-up
this pointer is used as a parent to retry to request the region.
A first problem is that this pointer might well be invalid (if for
example the conflicting resource have already been freed). Another
problem is that the next call to __request_region() fails to detect a
remaining conflict. The previously conflicting resource is passed as a
parameter and __request_region() will look for a conflict among the
children of this resource and not at the resource itself. It is likely
to succeed anyway, even if there is still a conflict.
Instead, the parent of the conflicting resource should be passed to
__request_region().
As a fix, this patch doesn't update the parent resource pointer in the
case we have to wait for a muxed region right after.
Reported-and-tested-by: Vincent Pelletier <plr.vincent@gmail.com> Signed-off-by: Simon Guinot <simon.guinot@sequanux.org> Tested-by: Vincent Donnefort <vdonnefort@gmail.com> Cc: stable@kernel.org Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit aa1311b426d5cc249887c8cbfa21a6dda2c5a201)
Mount filesystem
Create subvol with ID 257
Unmount filesystem
Mount filesystem
Delete subvol with ID 257
btrfs_drop_snapshot()
Add root corresponding to subvol 257 into
btrfs_transaction->dropped_roots list
Create new subvol (i.e. create_subvol())
257 is returned as the next free objectid
btrfs_read_fs_root_no_name()
Finds the btrfs_root instance corresponding to the old subvol with ID 257
in btrfs_fs_info->fs_roots_radix.
Returns error since btrfs_root_item->refs has the value of 0.
To fix the issue the commit initializes tree root's and subvolume root's
highest_objectid when loading the roots from disk.
Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit c19cd7e350c5ad9f3f4b3e486dbb4dab737a22a4)
If we failed to create a hard link we were not always releasing the
the transaction handle we got before, resulting in a memory leak and
preventing any other tasks from being able to commit the current
transaction.
Fix this by always releasing our transaction handle.
Signed-off-by: Filipe Manana <fdmanana@suse.com> Reviewed-by: Liu Bo <bo.li.liu@oracle.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 9bf972e8aa6110d750cf1ddab68511f478a6a751)
We weren't accounting for the insertion of an inline extent item for the
symlink inode nor that we need to update the parent inode item (through
the call to btrfs_add_nondir()). So fix this by including two more
transaction units.
Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit a1f535acffbd95ae6ae81656e8bba39af094c3f0)
When a symlink is successfully created it always has an inline extent
containing the source path. However if an error happens when creating
the symlink, we can leave in the subvolume's tree a symlink inode without
any such inline extent item - this happens if after btrfs_symlink() calls
btrfs_end_transaction() and before it calls the inode eviction handler
(through the final iput() call), the transaction gets committed and a
crash happens before the eviction handler gets called, or if a snapshot
of the subvolume is made before the eviction handler gets called. Sadly
we can't just avoid this by making btrfs_symlink() call
btrfs_end_transaction() after it calls the eviction handler, because the
later can commit the current transaction before it removes any items from
the subvolume tree (if it encounters ENOSPC errors while reserving space
for removing all the items).
So make send fail more gracefully, with an -EIO error, and print a
message to dmesg/syslog informing that there's an empty symlink inode,
so that the user can delete the empty symlink or do something else
about it.
Reported-by: Stephen R. van den Berg <srb@cuci.nl> Signed-off-by: Filipe Manana <fdmanana@suse.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit e92c51b734d57e04f0c9b43106b5294834339be6)
There is one ENOSPC case that's very confusing. There's Available
greater than zero but no file operation succeds (besides removing
files). This happens when the metadata are exhausted and there's no
possibility to allocate another chunk.
In this scenario it's normal that there's still some space in the data
chunk and the calculation in df reflects that in the Avail value.
To at least give some clue about the ENOSPC situation, let statfs report
zero value in Avail, even if there's still data space available.
Current:
/dev/sdb1 4.0G 3.3G 719M 83% /mnt/test
New:
/dev/sdb1 4.0G 3.3G 0 100% /mnt/test
We calculate the remaining metadata space minus global reserve. If this
is (supposedly) smaller than zero, there's no space. But this does not
hold in practice, the exhausted state happens where's still some
positive delta. So we apply some guesswork and compare the delta to a 4M
threshold. (Practically observed delta was 2M.)
We probably cannot calculate the exact threshold value because this
depends on the internal reservations requested by various operations, so
some operations that consume a few metadata will succeed even if the
Avail is zero. But this is better than the other way around.
Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 4e3fa12f124507ad17f999c28ef35803596ff2c6)
We hit this panic on a few of our boxes this week where we have an
ordered_extent with an NULL inode. We do an igrab() of the inode in writepages,
but weren't doing it in writepage which can be called directly from the VM on
dirty pages. If the inode has been unlinked then we could have I_FREEING set
which means igrab() would return NULL and we get this panic. Fix this by trying
to igrab in btrfs_writepage, and if it returns NULL then just redirty the page
and return AOP_WRITEPAGE_ACTIVATE; so the VM knows it wasn't successful. Thanks,
Signed-off-by: Josef Bacik <jbacik@fb.com> Reviewed-by: Liu Bo <bo.li.liu@oracle.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit bb055e837f904fdc80d4d82819b0a9aaf35dce4a)
Looks like oversight, call brelse() when checksum fails. Further down the
code, in the non error path, we do call brelse() and so we don't see
brelse() in the goto error paths.
Signed-off-by: Anand Jain <anand.jain@oracle.com> Reviewed-by: David Sterba <dsterba@suse.com> Signed-off-by: David Sterba <dsterba@suse.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit c0109d289de5a48e54a2d070981a629fc241f112)
ext4 can update bh->b_state non-atomically in _ext4_get_block() and
ext4_da_get_block_prep(). Usually this is fine since bh is just a
temporary storage for mapping information on stack but in some cases it
can be fully living bh attached to a page. In such case non-atomic
update of bh->b_state can race with an atomic update which then gets
lost. Usually when we are mapping bh and thus updating bh->b_state
non-atomically, nobody else touches the bh and so things work out fine
but there is one case to especially worry about: ext4_finish_bio() uses
BH_Uptodate_Lock on the first bh in the page to synchronize handling of
PageWriteback state. So when blocksize < pagesize, we can be atomically
modifying bh->b_state of a buffer that actually isn't under IO and thus
can race e.g. with delalloc trying to map that buffer. The result is
that we can mistakenly set / clear BH_Uptodate_Lock bit resulting in the
corruption of PageWriteback state or missed unlock of BH_Uptodate_Lock.
Fix the problem by always updating bh->b_state bits atomically.
CC: stable@vger.kernel.org Reported-by: Nikolay Borisov <kernel@kyup.com> Signed-off-by: Jan Kara <jack@suse.cz> Signed-off-by: Theodore Ts'o <tytso@mit.edu> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit a7635d6a0849007d2192bc02c038cc1b9d91b274)
dax_fault() currently relies on the get_block callback to attach an
io completion callback to the mapping buffer head so that it can
run unwritten extent conversion after zeroing allocated blocks.
Instead of this hack, pass the conversion callback directly into
dax_fault() similar to the get_block callback. When the filesystem
allocates unwritten extents, it will set the buffer_unwritten()
flag, and hence the dax_fault code can call the completion function
in the contexts where it is necessary without overloading the
mapping buffer head.
Note: The changes to ext4 to use this interface are suspect at best.
In fact, the way ext4 did this end_io assignment in the first place
looks suspect because it only set a completion callback when there
wasn't already some other write() call taking place on the same
inode. The ext4 end_io code looks rather intricate and fragile with
all it's reference counting and passing to different contexts for
modification via inode private pointers that aren't protected by
locks...
Signed-off-by: Dave Chinner <dchinner@redhat.com> Acked-by: Jan Kara <jack@suse.cz> Signed-off-by: Dave Chinner <david@fromorbit.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 0e3029cfab4a7884b32e1b2d3c19c12eded9804a)
The SPI tx and rx buffers are both supposed to be scan_bytes amount of
bytes large and a common allocation is used to allocate both buffers. This
puts the beginning of the tx buffer scan_bytes bytes after the rx buffer.
The initialization of the tx buffer pointer is done adding scan_bytes to
the beginning of the rx buffer, but since the rx buffer is of type __be16
this will actually add two times as much and the tx buffer ends up pointing
after the allocated buffer.
Fix this by using scan_count, which is scan_bytes / 2, instead of
scan_bytes when initializing the tx buffer pointer.
Fixes: aacff892cbd5 ("staging:iio:adis: Preallocate transfer message") Signed-off-by: Lars-Peter Clausen <lars@metafoo.de> Cc: <Stable@vger.kernel.org> Signed-off-by: Jonathan Cameron <jic23@kernel.org> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 3e77cb858ce67861e142186e3b131c04ddbc6edd)
By checking the effective credentials instead of the real UID / permitted
capabilities, ensure that the calling process actually intended to use its
credentials.
To ensure that all ptrace checks use the correct caller credentials (e.g.
in case out-of-tree code or newly added code omits the PTRACE_MODE_*CREDS
flag), use two new flags and require one of them to be set.
The problem was that when a privileged task had temporarily dropped its
privileges, e.g. by calling setreuid(0, user_uid), with the intent to
perform following syscalls with the credentials of a user, it still passed
ptrace access checks that the user would not be able to pass.
While an attacker should not be able to convince the privileged task to
perform a ptrace() syscall, this is a problem because the ptrace access
check is reused for things in procfs.
In particular, the following somewhat interesting procfs entries only rely
on ptrace access checks:
/proc/$pid/stat - uses the check for determining whether pointers
should be visible, useful for bypassing ASLR
/proc/$pid/maps - also useful for bypassing ASLR
/proc/$pid/cwd - useful for gaining access to restricted
directories that contain files with lax permissions, e.g. in
this scenario:
lrwxrwxrwx root root /proc/13020/cwd -> /root/foobar
drwx------ root root /root
drwxr-xr-x root root /root/foobar
-rw-r--r-- root root /root/foobar/secret
Therefore, on a system where a root-owned mode 6755 binary changes its
effective credentials as described and then dumps a user-specified file,
this could be used by an attacker to reveal the memory layout of root's
processes or reveal the contents of files he is not allowed to access
(through /proc/$pid/cwd).
[akpm@linux-foundation.org: fix warning] Signed-off-by: Jann Horn <jann@thejh.net> Acked-by: Kees Cook <keescook@chromium.org> Cc: Casey Schaufler <casey@schaufler-ca.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: James Morris <james.l.morris@oracle.com> Cc: "Serge E. Hallyn" <serge.hallyn@ubuntu.com> Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com> Cc: Andy Lutomirski <luto@kernel.org> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: "Eric W. Biederman" <ebiederm@xmission.com> Cc: Willy Tarreau <w@1wt.eu> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit ab88ce5feca4204ecf4e7ef6c6693ff67edc2169)
c118baf80256 ("arch/powerpc/mm/numa.c: do not allocate bootmem memory for non existing nodes")
avoids allocating bootmem memory for non existent nodes.
But when DEBUG_PER_CPU_MAPS=y is enabled, my powerNV system failed to boot
because in sched_init_numa(), cpumask_or() operation was done on
unallocated nodes.
Fix that by making cpumask_or() operation only on existing nodes.
[ Tested with and w/o DEBUG_PER_CPU_MAPS=y on x86 and PowerPC. ]
Reported-by: Jan Stancek <jstancek@redhat.com> Tested-by: Jan Stancek <jstancek@redhat.com> Signed-off-by: Raghavendra K T <raghavendra.kt@linux.vnet.ibm.com> Cc: <gkurz@linux.vnet.ibm.com> Cc: <grant.likely@linaro.org> Cc: <nikunj@linux.vnet.ibm.com> Cc: <vdavydov@parallels.com> Cc: <linuxppc-dev@lists.ozlabs.org> Cc: <linux-mm@kvack.org> Cc: <peterz@infradead.org> Cc: <benh@kernel.crashing.org> Cc: <paulus@samba.org> Cc: <mpe@ellerman.id.au> Cc: <anton@samba.org> Link: http://lkml.kernel.org/r/1452884483-11676-1-git-send-email-raghavendra.kt@linux.vnet.ibm.com Signed-off-by: Ingo Molnar <mingo@kernel.org> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 8cf0abcfb3b1ce60a9bd866db451a093dc015233)
The critical section protected by usbhid->lock in hid_ctrl() is too
big and because of this it causes a recursive deadlock. "Too big" means
the case statement and the call to hid_input_report() do not need to be
protected by the spinlock (no URB operations are done inside them).
The deadlock happens because in certain rare cases drivers try to grab
the lock while handling the ctrl irq which grabs the lock before them
as described above. For example newer wacom tablets like 056a:033c try
to reschedule proximity reads from wacom_intuos_schedule_prox_event()
calling hid_hw_request() -> usbhid_request() -> usbhid_submit_report()
which tries to grab the usbhid lock already held by hid_ctrl().
There are two ways to get out of this deadlock:
1. Make the drivers work "around" the ctrl critical region, in the
wacom case for ex. by delaying the scheduling of the proximity read
request itself to a workqueue.
2. Shrink the critical region so the usbhid lock protects only the
instructions which modify usbhid state, calling hid_input_report()
with the spinlock unlocked, allowing the device driver to grab the
lock first, finish and then grab the lock afterwards in hid_ctrl().
The error path in perf_event_open() is such that asking for a sampling
event on a PMU that doesn't generate interrupts will end up in dropping
the perf_sched_count even though it hasn't been incremented for this
event yet.
Given a sufficient amount of these calls, we'll end up disabling
scheduler's jump label even though we'd still have active events in the
system, thereby facilitating the arrival of the infernal regions upon us.
I'm fixing this by moving account_event() inside perf_event_alloc().
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: <stable@vger.kernel.org> Cc: Arnaldo Carvalho de Melo <acme@infradead.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Stephane Eranian <eranian@google.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Vince Weaver <vincent.weaver@maine.edu> Cc: vince@deater.net Link: http://lkml.kernel.org/r/1456917854-29427-1-git-send-email-alexander.shishkin@linux.intel.com Signed-off-by: Ingo Molnar <mingo@kernel.org> Signed-off-by: He Kuang <hekuang@huawei.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 919e67a63aa967566a909f4f6e1c13f8e88cf76e)
Because event_sched_out() checks event->pending_disable _before_
actually disabling the event, it can happen that the event fires after
it checks but before it gets disabled.
This would leave event->pending_disable set and the queued irq_work
will try and process it.
However, if the event trigger was during schedule(), the event might
have been de-scheduled by the time the irq_work runs, and
perf_event_disable_local() will fail.
Fix this by checking event->pending_disable _after_ we call
event->pmu->del(). This depends on the latter being a compiler
barrier, such that the compiler does not lift the load and re-creates
the problem.
Tested-by: Alexander Shishkin <alexander.shishkin@linux.intel.com> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Reviewed-by: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: dvyukov@google.com Cc: eranian@google.com Cc: oleg@redhat.com Cc: panand@redhat.com Cc: sasha.levin@oracle.com Cc: vince@deater.net Link: http://lkml.kernel.org/r/20160224174948.040469884@infradead.org Signed-off-by: Ingo Molnar <mingo@kernel.org> Signed-off-by: He Kuang <hekuang@huawei.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 882f862db7f3509f055208b2e3e5bd263265a03b)
When doing delayed allocation, update of on-disk inode size is postponed
until IO submission time. However hole punch or zero range fallocate
calls can end up discarding the tail page cache page and thus on-disk
inode size would never be properly updated.
Make sure the on-disk inode size is updated before truncating page
cache.
Signed-off-by: Jan Kara <jack@suse.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu> Reviewed-by: Mingming Cao <mingming.cao@oracle.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit f2b132595b89d9236b386e1d6ed3fcf5e9edf4cb)
Current code implementing FALLOC_FL_COLLAPSE_RANGE and
FALLOC_FL_INSERT_RANGE is prone to races with buffered writes and page
faults. If buffered write or write via mmap manages to squeeze between
filemap_write_and_wait_range() and truncate_pagecache() in the fallocate
implementations, the written data is simply discarded by
truncate_pagecache() although it should have been shifted.
Fix the problem by moving filemap_write_and_wait_range() call inside
i_mutex and i_mmap_sem. That way we are protected against races with
both buffered writes and page faults.
Signed-off-by: Jan Kara <jack@suse.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu> Reviewed-by: Mingming Cao <mingming.cao@oracle.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 181aaebde9360b8235df647ee36dafdc041d4964)
Currently ext4_alloc_file_blocks() was handling protection against
unlocked DIO. However we now need to sometimes call it under i_mmap_sem
and sometimes not and DIO protection ranks above it (although strictly
speaking this cannot currently create any deadlocks). Also
ext4_zero_range() was actually getting & releasing unlocked DIO
protection twice in some cases. Luckily it didn't introduce any real bug
but it was a land mine waiting to be stepped on. So move DIO protection
out from ext4_alloc_file_blocks() into the two callsites.
Signed-off-by: Jan Kara <jack@suse.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu> Reviewed-by: Mingming Cao <mingming.cao@oracle.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 9621787d69783fc23d14e1332377d7170d6928ed)
Currently, page faults and hole punching are completely unsynchronized.
This can result in page fault faulting in a page into a range that we
are punching after truncate_pagecache_range() has been called and thus
we can end up with a page mapped to disk blocks that will be shortly
freed. Filesystem corruption will shortly follow. Note that the same
race is avoided for truncate by checking page fault offset against
i_size but there isn't similar mechanism available for punching holes.
Fix the problem by creating new rw semaphore i_mmap_sem in inode and
grab it for writing over truncate, hole punching, and other functions
removing blocks from extent tree and for read over page faults. We
cannot easily use i_data_sem for this since that ranks below transaction
start and we need something ranking above it so that it can be held over
the whole truncate / hole punching operation. Also remove various
workarounds we had in the code to reduce race window when page fault
could have created pages with stale mapping information.
Signed-off-by: Jan Kara <jack@suse.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu> Reviewed-by: Mingming Cao <mingming.cao@oracle.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 248766f068fd1d3d95479f470bc926d1136141d6)
Yes, all of these are needed. :) This is admittedly a bit odd, but
kvm-unit-tests access.flat tests this if you run it with "-cpu host"
and of course ept=0.
KVM runs the guest with CR0.WP=1, so it must handle supervisor writes
specially when pte.u=1/pte.w=0/CR0.WP=0. Such writes cause a fault
when U=1 and W=0 in the SPTE, but they must succeed because CR0.WP=0.
When KVM gets the fault, it sets U=0 and W=1 in the shadow PTE and
restarts execution. This will still cause a user write to fault, while
supervisor writes will succeed. User reads will fault spuriously now,
and KVM will then flip U and W again in the SPTE (U=1, W=0). User reads
will be enabled and supervisor writes disabled, going back to the
originary situation where supervisor writes fault spuriously.
When SMEP is in effect, however, U=0 will enable kernel execution of
this page. To avoid this, KVM also sets NX=1 in the shadow PTE together
with U=0. If the guest has not enabled NX, the result is a continuous
stream of page faults due to the NX bit being reserved.
The fix is to force EFER.NX=1 even if the CPU is taking care of the EFER
switch. (All machines with SMEP have the CPU_LOAD_IA32_EFER vm-entry
control, so they do not use user-return notifiers for EFER---if they did,
EFER.NX would be forced to the same value as the host).
There is another bug in the reserved bit check, which I've split to a
separate patch for easier application to stable kernels.
Cc: stable@vger.kernel.org Cc: Andy Lutomirski <luto@amacapital.net> Reviewed-by: Xiao Guangrong <guangrong.xiao@linux.intel.com> Fixes: f6577a5fa15d82217ca73c74cd2dcbc0f6c781dd Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit eac525506a083a389ba173880979a6291401af2d)
Linux guests on Haswell (and also SandyBridge and Broadwell, at least)
would crash if you decided to run a host command that uses PEBS, like
perf record -e 'cpu/mem-stores/pp' -a
This happens because KVM is using VMX MSR switching to disable PEBS, but
SDM [2015-12] 18.4.4.4 Re-configuring PEBS Facilities explains why it
isn't safe:
When software needs to reconfigure PEBS facilities, it should allow a
quiescent period between stopping the prior event counting and setting
up a new PEBS event. The quiescent period is to allow any latent
residual PEBS records to complete its capture at their previously
specified buffer address (provided by IA32_DS_AREA).
There might not be a quiescent period after the MSR switch, so a CPU
ends up using host's MSR_IA32_DS_AREA to access an area in guest's
memory. (Or MSR switching is just buggy on some models.)
The guest can learn something about the host this way:
If the guest doesn't map address pointed by MSR_IA32_DS_AREA, it results
in #PF where we leak host's MSR_IA32_DS_AREA through CR2.
After that, a malicious guest can map and configure memory where
MSR_IA32_DS_AREA is pointing and can therefore get an output from
host's tracing.
This is not a critical leak as the host must initiate with PEBS tracing
and I have not been able to get a record from more than one instruction
before vmentry in vmx_vcpu_run() (that place has most registers already
overwritten with guest's).
We could disable PEBS just few instructions before vmentry, but
disabling it earlier shouldn't affect host tracing too much.
We also don't need to switch MSR_IA32_PEBS_ENABLE on VMENTRY, but that
optimization isn't worth its code, IMO.
(If you are implementing PEBS for guests, be sure to handle the case
where both host and guest enable PEBS, because this patch doesn't.)
Fixes: 26a4f3c08de4 ("perf/x86: disable PEBS on a guest entry.") Cc: <stable@vger.kernel.org> Reported-by: Jiří Olša <jolsa@redhat.com> Signed-off-by: Radim Krčmář <rkrcmar@redhat.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit eb34a645aee3906baac9cad7defdabf61ac40bfd)
d_instantiate(new_dentry, old_inode) is absolutely wrong thing to
do - it will oops if new_dentry used to be positive, for starters.
What we need is d_invalidate() the target and be done with that.
Cc: stable@vger.kernel.org # v3.18+ Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit c62aadae234ffad0901c20ac1a1aa4e13cce1c20)
Failing to allocate an inode for child means that cache for *parent* is
incompletely populated. So it's parent directory inode ('dir') that
needs NCPI_DIR_CACHE flag removed, *not* the child inode ('inode', which
is what we'd failed to allocate in the first place).
Fucked-up-in: commit 5e993e25 ("ncpfs: get rid of d_validate() nonsense") Fucked-up-by: Al Viro <viro@zeniv.linux.org.uk> Cc: stable@vger.kernel.org # v3.19 Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 906e5a6e6e73316fa4741ca53be014c9477a100c)
Overlayfs must update uid/gid after chown, otherwise functions
like inode_owner_or_capable() will check user against stale uid.
Catched by xfstests generic/087, it chowns file and calls utimes.
Signed-off-by: Konstantin Khlebnikov <koct9i@gmail.com> Signed-off-by: Miklos Szeredi <miklos@szeredi.hu> Cc: <stable@vger.kernel.org> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 546a8b3c4059af5fd8466f3d1848321e7613904c)
After rename file dentry still holds reference to lower dentry from
previous location. This doesn't matter for data access because data comes
from upper dentry. But this stale lower dentry taints dentry at new
location and turns it into non-pure upper. Such file leaves visible
whiteout entry after remove in directory which shouldn't have whiteouts at
all.
Overlayfs already tracks pureness of file location in oe->opaque. This
patch just uses that for detecting actual path type.
Comment from Vivek Goyal's patch:
Here are the details of the problem. Do following.
$ mkdir upper lower work merged upper/dir/
$ touch lower/test
$ sudo mount -t overlay overlay -olowerdir=lower,upperdir=upper,workdir=
work merged
$ mv merged/test merged/dir/
$ rm merged/dir/test
$ ls -l merged/dir/
/usr/bin/ls: cannot access merged/dir/test: No such file or directory
total 0
c????????? ? ? ? ? ? test
Basic problem seems to be that once a file has been unlinked, a whiteout
has been left behind which was not needed and hence it becomes visible.
Whiteout is visible because parent dir is of not type MERGE, hence
od->is_real is set during ovl_dir_open(). And that means ovl_iterate()
passes on iterate handling directly to underlying fs. Underlying fs does
not know/filter whiteouts so it becomes visible to user.
Why did we leave a whiteout to begin with when we should not have.
ovl_do_remove() checks for OVL_TYPE_PURE_UPPER() and does not leave
whiteout if file is pure upper. In this case file is not found to be pure
upper hence whiteout is left.
So why file was not PURE_UPPER in this case? I think because dentry is
still carrying some leftover state which was valid before rename. For
example, od->numlower was set to 1 as it was a lower file. After rename,
this state is not valid anymore as there is no such file in lower.
ovl_remove_upper() should do d_drop() only after it successfully
removes the dir, otherwise a subsequent getcwd() system call will
fail, breaking userspace programs.
This is to fix: https://bugzilla.kernel.org/show_bug.cgi?id=110491
Signed-off-by: Rui Wang <rui.y.wang@intel.com> Reviewed-by: Konstantin Khlebnikov <koct9i@gmail.com> Signed-off-by: Miklos Szeredi <miklos@szeredi.hu> Cc: <stable@vger.kernel.org> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit e786702fff38e2b5142029d6de615abf1c8e436f)
Just like for CCMP we need to check that for GCMP the fragments
have PNs that increment by one; the spec was updated to fix this
security issue and now has the following text:
The receiver shall discard MSDUs and MMPDUs whose constituent
MPDU PN values are not incrementing in steps of 1.
Adapt the code for CCMP to work for GCMP as well, luckily the
relevant fields already alias each other so no code duplication
is needed (just check the aliasing with BUILD_BUG_ON.)
Cc: stable@vger.kernel.org Signed-off-by: Johannes Berg <johannes.berg@intel.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 87e0016ccb1f9cbe377d4af19cb840acbbdff206)
gs_destroy_candev() erroneously calls kfree() on a struct gs_can *, which is
allocated through alloc_candev() and should instead be freed using
free_candev() alone.
The inappropriate use of kfree() causes the kernel to hang when
gs_destroy_candev() is called.
Only the struct gs_usb * which is allocated through kzalloc() should be freed
using kfree() when the device is disconnected.
Signed-off-by: Maximilian Schneider <max@schneidersoft.net> Cc: linux-stable <stable@vger.kernel.org> Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 01ff3a0a01366a231593476cfe775596ebdba30f)
This is a clone of commit 2ab957492d13b ("ip_forward: Drop frames with
attached skb->sk") for ipv6.
This commit has exactly the same reasons as the above mentioned commit,
namely to prevent panics during netfilter reload or a misconfigured stack.
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit b014bae072a1dad1767c5c6e7e7b480165685c3e)
Pause/unpause graph tracing around do_suspend_lowlevel as it has
inconsistent call/return info after it jumps to the wakeup vector.
The graph trace buffer will otherwise become misaligned and
may eventually crash and hang on suspend.
To reproduce the issue and test the fix:
Run a function_graph trace over suspend/resume and set the graph
function to suspend_devices_and_enter. This consistently hangs the
system without this fix.
Signed-off-by: Todd Brandt <todd.e.brandt@linux.intel.com> Cc: All applicable <stable@vger.kernel.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 8ef267aabd98f9df0279b9bb4245a3b985ead692)
Games with ordering and barriers are way too brittle. Just
bump ->d_seq before and after updating ->d_inode and ->d_flags
type bits, so that verifying ->d_seq would guarantee they are
coherent.
Cc: stable@vger.kernel.org # v3.13+ Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Sasha Levin <sasha.levin@oracle.com> Signed-off-by: Dan Duval <dan.duval@oracle.com>
(cherry picked from commit c8ce76e3c6cd937e8b3fd8ae3573f23767b70eca)
HDSPM driver contains a code issuing zero-division potentially in
system sample rate ctl code. This patch fixes it by not processing
a zero or invalid rate value as a divisor, as well as excluding the
invalid value to be passed via the given ctl element.
For interim responses we only need to parse a header and update
a number credits. Now it is done for all SMB2+ command except
SMB2_READ which is wrong. Fix this by adding such processing.
Signed-off-by: Pavel Shilovsky <pshilovsky@samba.org> Tested-by: Shirish Pargaonkar <shirishpargaonkar@gmail.com> CC: Stable <stable@vger.kernel.org> Signed-off-by: Steve French <smfrench@gmail.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 78b821d76e779822877604052f06219294e9e038)
When opening a file, SMB2_open() attempts to parse the lease state from the
SMB2 CREATE Response. However, the parsing code was not careful to ensure
that the create contexts are not empty or invalid, which can lead to out-
of-bounds memory access. This can be seen easily by trying
to read a file from a OSX 10.11 SMB3 server. Here is sample crash output:
Calling return copy_to_user(...) in an ioctl will not
do the right thing if there's a pagefault:
copy_to_user returns the number of bytes not copied
in this case.
Fix up vfio to do
return copy_to_user(...)) ?
-EFAULT : 0;
everywhere.
Cc: stable@vger.kernel.org Signed-off-by: Michael S. Tsirkin <mst@redhat.com> Signed-off-by: Alex Williamson <alex.williamson@redhat.com> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 1590808b43559d8330599158453f28f1b16ffd54)
both do_last() and walk_component() risk picking a NULL inode out
of dentry about to become positive, *then* checking its flags and
seeing that it's not negative anymore and using (already stale by
then) value they'd fetched earlier. Usually ends up oopsing soon
after that...
Cc: stable@vger.kernel.org # v3.13+ Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 9b77cd137fd841d7a14e1c9428cfc49f4df0306e)
... into returning a positive to path_openat(), which would interpret that
as "symlink had been encountered" and proceed to corrupt memory, etc.
It can only happen due to a bug in some ->open() instance or in some LSM
hook, etc., so we report any such event *and* make sure it doesn't trick
us into further unpleasantness.
Cc: stable@vger.kernel.org # v3.6+, at least Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 3960cde3e356057bd60adce1b625a7d178b9581c)
Commit 4167e9b2cf10 ("mm: remove GFP_THISNODE") removed the GFP_THISNODE
flag combination due to confusing semantics. It noted that
alloc_misplaced_dst_page() was one such user after changes made by
commit e97ca8e5b864 ("mm: fix GFP_THISNODE callers and clarify").
Unfortunately when GFP_THISNODE was removed, users of
alloc_misplaced_dst_page() started waking kswapd and entering direct
reclaim because the wrong GFP flags are cleared. The consequence is
that workloads that used to fit into memory now get reclaimed which is
addressed by this patch.
The problem can be demonstrated with "mutilate" that exercises memcached
which is software dedicated to memory object caching. The configuration
uses 80% of memory and is run 3 times for varying numbers of clients.
The results on a 4-socket NUMA box are
The metric is queries/second with the more the better. The results are
way outside of the noise and the reason for the improvement is obvious
from some of the vmstats
The vanilla kernel is swapping like crazy with large amounts of direct
reclaim and kswapd activity. The figures are aggregate but it's known
that the bad activity is throughout the entire test.
Note that simple streaming anon/file memory consumers also see this
problem but it's not as obvious. In those cases, kswapd is awake when
it should not be.
As there are at least two reclaim-related bugs out there, it's worth
spelling out the user-visible impact. This patch only addresses bugs
related to excessive reclaim on NUMA hardware when the working set is
larger than a NUMA node. There is a bug related to high kswapd CPU
usage but the reports are against laptops and other UMA hardware and is
not addressed by this patch.
Signed-off-by: Mel Gorman <mgorman@techsingularity.net> Cc: Vlastimil Babka <vbabka@suse.cz> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: David Rientjes <rientjes@google.com> Cc: <stable@vger.kernel.org> [4.1+] Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit 419ddc3099727a291a67d41498c3d1caddb75392)
pmd_trans_unstable()/pmd_none_or_trans_huge_or_clear_bad() were
introduced to locklessy (but atomically) detect when a pmd is a regular
(stable) pmd or when the pmd is unstable and can infinitely transition
from pmd_none() and pmd_trans_huge() from under us, while only holding
the mmap_sem for reading (for writing not).
While holding the mmap_sem only for reading, MADV_DONTNEED can run from
under us and so before we can assume the pmd to be a regular stable pmd
we need to compare it against pmd_none() and pmd_trans_huge() in an
atomic way, with pmd_trans_unstable(). The old pmd_trans_huge() left a
tiny window for a race.
Useful applications are unlikely to notice the difference as doing
MADV_DONTNEED concurrently with a page fault would lead to undefined
behavior.
[akpm@linux-foundation.org: tidy up comment grammar/layout] Signed-off-by: Andrea Arcangeli <aarcange@redhat.com> Reported-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
(cherry picked from commit d347d0e9ae617bd44ca7679786ebf11a06d50372)
projects / users / jedix / linux-maple.git / log