math_state_restore() assumes it is called with irqs disabled,
but this is not true if the caller is __restore_xstate_sig().
This means that if ia32_fxstate == T and __copy_from_user()
fails, __restore_xstate_sig() returns with irqs disabled too.
This triggers:
BUG: sleeping function called from invalid context at kernel/locking/rwsem.c:41
dump_stack
___might_sleep
? _raw_spin_unlock_irqrestore
__might_sleep
down_read
? _raw_spin_unlock_irqrestore
print_vma_addr
signal_fault
sys32_rt_sigreturn
Change __restore_xstate_sig() to call set_used_math()
unconditionally. This avoids enabling and disabling interrupts
in math_state_restore(). If copy_from_user() fails, we can
simply do fpu_finit() by hand.
[ Note: this is only the first step. math_state_restore() should
not check used_math(), it should set this flag. While
init_fpu() should simply die. ]
Signed-off-by: Oleg Nesterov <oleg@redhat.com> Signed-off-by: Borislav Petkov <bp@suse.de> Cc: Andy Lutomirski <luto@amacapital.net> Cc: Borislav Petkov <bp@alien8.de> Cc: Dave Hansen <dave.hansen@intel.com> Cc: Fenghua Yu <fenghua.yu@intel.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Pekka Riikonen <priikone@iki.fi> Cc: Quentin Casasnovas <quentin.casasnovas@oracle.com> Cc: Rik van Riel <riel@redhat.com> Cc: Suresh Siddha <sbsiddha@gmail.com> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/20150307153844.GB25954@redhat.com Signed-off-by: Ingo Molnar <mingo@kernel.org> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The kernel crypto API logic requires the caller to provide the
length of (ciphertext || authentication tag) as cryptlen for the
AEAD decryption operation. Thus, the cipher implementation must
calculate the size of the plaintext output itself and cannot simply use
cryptlen.
The RFC4106 GCM decryption operation tries to overwrite cryptlen memory
in req->dst. As the destination buffer for decryption only needs to hold
the plaintext memory but cryptlen references the input buffer holding
(ciphertext || authentication tag), the assumption of the destination
buffer length in RFC4106 GCM operation leads to a too large size. This
patch simply uses the already calculated plaintext size.
In addition, this patch fixes the offset calculation of the AAD buffer
pointer: as mentioned before, cryptlen already includes the size of the
tag. Thus, the tag does not need to be added. With the addition, the AAD
will be written beyond the already allocated buffer.
Note, this fixes a kernel crash that can be triggered from user space
via AF_ALG(aead) -- simply use the libkcapi test application
from [1] and update it to use rfc4106-gcm-aes.
Using [1], the changes were tested using CAVS vectors to demonstrate
that the crypto operation still delivers the right results.
[1] http://www.chronox.de/libkcapi.html
CC: Tadeusz Struk <tadeusz.struk@intel.com> Signed-off-by: Stephan Mueller <smueller@chronox.de> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
As pointed by recent post[1] on exploiting DRAM physical imperfection,
/proc/PID/pagemap exposes sensitive information which can be used to do
attacks.
This disallows anybody without CAP_SYS_ADMIN to read the pagemap.
It's directly caused by 89d3cf6 [SCSI] libsas: add mutex for SMP task
execution, but shows a deeper cause: expander functions expect to be able to
cast to and treat domain devices as expanders. The correct fix is to only do
expander discover when we know we've got an expander device to avoid wrongly
casting a non-expander device.
Otherwise the guest can abuse that control to cause e.g. PCIe
Unsupported Request responses by disabling memory and/or I/O decoding
and subsequently causing (CPU side) accesses to the respective address
ranges, which (depending on system configuration) may be fatal to the
host.
Note that to alter any of the bits collected together as
PCI_COMMAND_GUEST permissive mode is now required to be enabled
globally or on the specific device.
This is CVE-2015-2150 / XSA-120.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Signed-off-by: David Vrabel <david.vrabel@citrix.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
To take down the MOB and GMR memory types, the driver may have to issue
fence objects and thus make sure that the fence manager is taken down
after those memory types.
Reorder device init accordingly.
According to a report from Yuxuan Shui, nilfs2 in kernel 3.19 got stuck
during recovery at mount time. The code path that caused the deadlock was
as follows:
nilfs_fill_super()
load_nilfs()
nilfs_salvage_orphan_logs()
* Do roll-forwarding, attach segment constructor for recovery,
and kick it.
nilfs_segctor_thread()
nilfs_segctor_thread_construct()
* A lock is held with nilfs_transaction_lock()
nilfs_segctor_do_construct()
nilfs_segctor_drop_written_files()
iput()
iput_final()
write_inode_now()
writeback_single_inode()
__writeback_single_inode()
do_writepages()
nilfs_writepage()
nilfs_construct_dsync_segment()
nilfs_transaction_lock() --> deadlock
This can happen if commit 7ef3ff2fea8b ("nilfs2: fix deadlock of segment
constructor over I_SYNC flag") is applied and roll-forward recovery was
performed at mount time. The roll-forward recovery can happen if datasync
write is done and the file system crashes immediately after that. For
instance, we can reproduce the issue with the following steps:
< nilfs2 is mounted on /nilfs (device: /dev/sdb1) >
# dd if=/dev/zero of=/nilfs/test bs=4k count=1 && sync
# dd if=/dev/zero of=/nilfs/test conv=notrunc oflag=dsync bs=4k
count=1 && reboot -nfh
< the system will immediately reboot >
# mount -t nilfs2 /dev/sdb1 /nilfs
The deadlock occurs because iput() can run segment constructor through
writeback_single_inode() if MS_ACTIVE flag is not set on sb->s_flags. The
above commit changed segment constructor so that it calls iput()
asynchronously for inodes with i_nlink == 0, but that change was
imperfect.
This fixes the another deadlock by deferring iput() in segment constructor
even for the case that mount is not finished, that is, for the case that
MS_ACTIVE flag is not set.
Normally _regulator_do_enable() isn't called on an already-enabled
rdev. That's because the main caller, _regulator_enable() always
calls _regulator_is_enabled() and only calls _regulator_do_enable() if
the rdev was not already enabled.
However, there is one caller of _regulator_do_enable() that doesn't
check: regulator_suspend_finish(). While we might want to make
regulator_suspend_finish() behave more like _regulator_enable(), it's
probably also a good idea to make _regulator_do_enable() robust if it
is called on an already enabled rdev.
At the moment, _regulator_do_enable() is _not_ robust for already
enabled rdevs if we're using an ena_pin. Each time
_regulator_do_enable() is called for an rdev using an ena_pin the
reference count of the ena_pin is incremented even if the rdev was
already enabled. This is not as intended because the ena_pin is for
something else: for keeping track of how many active rdevs there are
sharing the same ena_pin.
Here's how the reference counting works here:
* Each time _regulator_enable() is called we increment
rdev->use_count, so _regulator_enable() calls need to be balanced
with _regulator_disable() calls.
* There is no explicit reference counting in _regulator_do_enable()
which is normally just a warapper around rdev->desc->ops->enable()
with code for supporting delays. It's not expected that the
"ops->enable()" call do reference counting.
* Since regulator_ena_gpio_ctrl() does have reference counting
(handling the sharing of the pin amongst multiple rdevs), we
shouldn't call it if the current rdev is already enabled.
Note that as part of this we cleanup (remove) the initting of
ena_gpio_state in regulator_register(). In _regulator_do_enable(),
_regulator_do_disable() and _regulator_is_enabled() is is clear that
ena_gpio_state should be the state of whether this particular rdev has
requested the GPIO be enabled. regulator_register() was initting it
as the actual state of the pin.
Fixes: 967cfb18c0e3 ("regulator: core: manage enable GPIO list") Signed-off-by: Doug Anderson <dianders@chromium.org> Signed-off-by: Mark Brown <broonie@kernel.org> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The _regulator_do_enable() call ought to be a no-op when called on an
already-enabled regulator. However, as an optimization
_regulator_enable() doesn't call _regulator_do_enable() on an already
enabled regulator. That means we never test the case of calling
_regulator_do_enable() during normal usage and there may be hidden
bugs or warnings. We have seen warnings issued by the tps65090 driver
and bugs when using the GPIO enable pin.
Let's match the same optimization that _regulator_enable() in
regulator_suspend_finish(). That may speed up suspend/resume and also
avoids exposing hidden bugs.
[Use much clearer commit message from Doug Anderson]
Signed-off-by: Javier Martinez Canillas <javier.martinez@collabora.co.uk> Signed-off-by: Mark Brown <broonie@kernel.org> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
EEH recovery for bnx2x based adapters is not reliable on all Power
systems using the default hot reset, which can result in an
unrecoverable EEH error. Forcing the use of fundamental reset
during EEH recovery fixes this.
Signed-off-by: Brian King <brking@linux.vnet.ibm.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The commit [ef403edb7558: ALSA: hda - Don't access stereo amps for
mono channel widgets] fixed the handling of mono widgets in general,
but it still misses an exceptional case: namely, a mono mixer widget
taking a single stereo input. In this case, it has stereo volumes
although it's a mono widget, and thus we have to take care of both
left and right input channels, as stated in HD-audio spec ("7.1.3
Widget Interconnection Rules").
This patch covers this missing piece by adding proper checks of stereo
amps in both the generic parser and the proc output codes.
MacBook Air 5,2 has the same problem as MacBook Pro 8,1 where the
built-in mic records only the right channel. Apply the same
workaround as MBP8,1 to spread the mono channel via a Cirrus codec
vendor-specific COEF setup.
CS420x codecs seem to deal only the single amps of ADC nodes even
though the nodes receive multiple inputs. This leads to the
inconsistent amp value after S3/S4 resume, for example.
The fix is just to set codec->single_adc_amp flag. Then the driver
handles these ADC amps as if single connections.
The current HDA generic parser initializes / modifies the amp values
always in stereo, but this seems causing the problem on ALC3229 codec
that has a few mono channel widgets: namely, these mono widgets react
to actions for both channels equally.
In the driver code, we do care the mono channel and create a control
only for the left channel (as defined in HD-audio spec) for such a
node. When the control is updated, only the left channel value is
changed. However, in the resume, the right channel value is also
restored from the initial value we took as stereo, and this overwrites
the left channel value. This ends up being the silent output as the
right channel has been never touched and remains muted.
This patch covers the places where unconditional stereo amp accesses
are done and converts to the conditional accesses.
Compaq Presario CQ60 laptop with CX20561 gives a wrong pin for the
built-in mic NID 0x17 instead of NID 0x1d, and it results in the
non-working mic. This patch just remaps the pin correctly via fixup.
There was no check about the id string of user control elements, so we
accepted even a control element with an empty string, which is
obviously bogus. This patch adds more sanity checks of id strings.
The device complies to the UAC1 standard but hides that fact with
proprietary descriptors. The autodetect quirk for Roland devices
catches the audio interface but misses the MIDI part, so a specific
quirk is needed.
Commit fd316941c ("spi/pl022: disable port when unused") introduced a race,
which leads to possible driver lock up (easily reproducible on SMP).
The problem happens in giveback() function where the completion of the transfer
is signalled to SPI subsystem and then the HW SPI controller is disabled. Another
transfer might be setup in between, which brings driver in locked-up state.
Lockup! SPI controller is disabled and the data will never be received. Whole
SPI subsystem is waiting for transfer ACK and blocked.
So, only signal transfer completion after disabling the controller.
Fixes: fd316941c (spi/pl022: disable port when unused) Signed-off-by: Alexander Sverdlin <alexander.sverdlin@nokia.com> Signed-off-by: Mark Brown <broonie@kernel.org> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Problem: When IMA and VTPM are both enabled in kernel config,
kernel hangs during bootup on LE OS.
Why?: IMA calls tpm_pcr_read() which results in tpm_ibmvtpm_send
and tpm_ibmtpm_recv getting called. A trace showed that
tpm_ibmtpm_recv was hanging.
Resolution: tpm_ibmtpm_recv was hanging because tpm_ibmvtpm_send
was sending CRQ message that probably did not make much sense
to phype because of Endianness. The fix below sends correctly
converted CRQ for LE. This was not caught before because it
seems IMA is not enabled by default in kernel config and
IMA exercises this particular code path in vtpm.
Tested with IMA and VTPM enabled in kernel config and VTPM
enabled on both a BE OS and a LE OS ppc64 lpar. This exercised
CRQ and TPM command code paths in vtpm.
Patch is against Peter's tpmdd tree on github which included
Vicky's previous vtpm le patches.
Signed-off-by: Joy Latten <jmlatten@linux.vnet.ibm.com> Reviewed-by: Ashley Lai <ashley@ahsleylai.com> Signed-off-by: Peter Huewe <peterhuewe@gmx.de> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The cpuset.sched_relax_domain_level can control how far we do
immediate load balancing on a system. However, it was found on recent
kernels that echo'ing a value into cpuset.sched_relax_domain_level
did not reduce any immediate load balancing.
The reason this occurred was because the update_domain_attr_tree() traversal
did not update for the "top_cpuset". This resulted in nothing being changed
when modifying the sched_relax_domain_level parameter.
This patch is able to address that problem by having update_domain_attr_tree()
allow updates for the root in the cpuset traversal.
Fixes: fc560a26acce ("cpuset: replace cpuset->stack_list with cpuset_for_each_descendant_pre()") Signed-off-by: Jason Low <jason.low2@hp.com> Signed-off-by: Zefan Li <lizefan@huawei.com> Signed-off-by: Tejun Heo <tj@kernel.org> Tested-by: Serge Hallyn <serge.hallyn@canonical.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
cancel[_delayed]_work_sync() are implemented using
__cancel_work_timer() which grabs the PENDING bit using
try_to_grab_pending() and then flushes the work item with PENDING set
to prevent the on-going execution of the work item from requeueing
itself.
try_to_grab_pending() can always grab PENDING bit without blocking
except when someone else is doing the above flushing during
cancelation. In that case, try_to_grab_pending() returns -ENOENT. In
this case, __cancel_work_timer() currently invokes flush_work(). The
assumption is that the completion of the work item is what the other
canceling task would be waiting for too and thus waiting for the same
condition and retrying should allow forward progress without excessive
busy looping
Unfortunately, this doesn't work if preemption is disabled or the
latter task has real time priority. Let's say task A just got woken
up from flush_work() by the completion of the target work item. If,
before task A starts executing, task B gets scheduled and invokes
__cancel_work_timer() on the same work item, its try_to_grab_pending()
will return -ENOENT as the work item is still being canceled by task A
and flush_work() will also immediately return false as the work item
is no longer executing. This puts task B in a busy loop possibly
preventing task A from executing and clearing the canceling state on
the work item leading to a hang.
task A task B worker
executing work
__cancel_work_timer()
try_to_grab_pending()
set work CANCELING
flush_work()
block for work completion
completion, wakes up A
__cancel_work_timer()
while (forever) {
try_to_grab_pending()
-ENOENT as work is being canceled
flush_work()
false as work is no longer executing
}
This patch removes the possible hang by updating __cancel_work_timer()
to explicitly wait for clearing of CANCELING rather than invoking
flush_work() after try_to_grab_pending() fails with -ENOENT.
Link: http://lkml.kernel.org/g/20150206171156.GA8942@axis.com
v3: bit_waitqueue() can't be used for work items defined in vmalloc
area. Switched to custom wake function which matches the target
work item and exclusive wait and wakeup.
v2: v1 used wake_up() on bit_waitqueue() which leads to NULL deref if
the target bit waitqueue has wait_bit_queue's on it. Use
DEFINE_WAIT_BIT() and __wake_up_bit() instead. Reported by Tomeu
Vizoso.
when multiport is off, virtio console invokes config access from irq
context, config access is blocking on s390.
Fix this up by scheduling work from config irq - similar to what we do
for multiport configs.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com> Reviewed-by: Amit Shah <amit.shah@redhat.com> Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
when multiport is off, we don't initialize config work,
but we then cancel uninitialized control_work on freeze.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com> Reviewed-by: Amit Shah <amit.shah@redhat.com> Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
commit 6ae9200f2cab7 ("enlarge console.name") increased the storage
for the console name to 16 bytes, but not the corresponding
struct console_cmdline::name storage. Console names longer than
8 bytes cause read beyond end-of-string and failure to match
console; I'm not sure if there are other unexpected consequences.
Signed-off-by: Peter Hurley <peter@hurleysoftware.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
If the part of the compression data are corrupted, or the compression
data is totally fake, the memory access over the limit is possible.
This is the log from my system usning lz4 decompression.
[6502]data abort, halting
[6503]r0 0x00000000 r1 0x00000000 r2 0xdcea0ffc r3 0xdcea0ffc
[6509]r4 0xb9ab0bfd r5 0xdcea0ffc r6 0xdcea0ff8 r7 0xdce80000
[6515]r8 0x00000000 r9 0x00000000 r10 0x00000000 r11 0xb9a98000
[6522]r12 0xdcea1000 usp 0x00000000 ulr 0x00000000 pc 0x820149bc
[6528]spsr 0x400001f3
and the memory addresses of some variables at the moment are
ref:0xdcea0ffc, op:0xdcea0ffc, oend:0xdcea1000
As you can see, COPYLENGH is 8bytes, so @ref and @op can access the momory
over @oend.
and a non-hugetlbfs page has no page_hstate(). This works 99% of the
time because page_hstate() determines the hstate from the page order
alone. Since the page order of a THP page matches the default hugetlbfs
page order, it works.
But, if you change the default huge page size on the boot command-line
(say default_hugepagesz=1G), then we might not even *have* a 2MB hstate
so page_hstate() returns null and copy_huge_page() oopses pretty fast
since copy_huge_page() dereferences the hstate:
Mel noticed that the migration code is really the only user of these
functions. This moves all the copy code over to migrate.c and makes
copy_huge_page() work for THP by checking for it explicitly.
I believe the bug was introduced in commit b32967ff101a ("mm: numa: Add
THP migration for the NUMA working set scanning fault case")
[akpm@linux-foundation.org: fix coding-style and comment text, per Naoya Horiguchi] Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com> Acked-by: Mel Gorman <mgorman@suse.de> Reviewed-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com> Cc: Hillf Danton <dhillf@gmail.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Tested-by: Dave Jiang <dave.jiang@intel.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
This is a single hunk introduced later in the upstream commit cb900f41215447433cbc456d1c4294e858a84d7c (mm, hugetlb: convert
hugetlbfs to use split pmd lock). We need page_hstate even for
!HUGETLB_PAGE case for the next patch (mm: thp: give transparent
hugepage code a separate copy_page).
mm/migrate.c: In function `do_move_page_to_node_array':
include/linux/hugetlb.h:140:33: warning: statement with no effect [-Wunused-value]
#define isolate_huge_page(p, l) false
^
mm/migrate.c:1170:4: note: in expansion of macro `isolate_huge_page'
isolate_huge_page(page, &pagelist);
Commit 746c9e9f92dd "of/base: Fix PowerPC address parsing hack" limited
the applicability of the workaround whereby a missing ranges is treated
as an empty ranges. This workaround was hiding a bug in the etsec2
device tree nodes, which have children with reg, but did not have
ranges.
Signed-off-by: Scott Wood <scottwood@freescale.com> Reported-by: Alexander Graf <agraf@suse.de> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Firstly, handle zero length calls properly. Believe it or not there
are a few of these happening during early boot.
Next, we can't just drop to a memcpy() call in the forward copy case
where dst <= src. The reason is that the cache initializing stores
used in the Niagara memcpy() implementations can end up clearing out
cache lines before we've sourced their original contents completely.
For example, considering NG4memcpy, the main unrolled loop begins like
this:
Assume dst is 64 byte aligned and let's say that dst is src - 8 for
this memcpy() call. That store at the end there is the one to the
first line in the cache line, thus clearing the whole line, which thus
clobbers "src + 0x28" before it even gets loaded.
To avoid this, just fall through to a simple copy only mildly
optimized for the case where src and dst are 8 byte aligned and the
length is a multiple of 8 as well. We could get fancy and call
GENmemcpy() but this is good enough for how this thing is actually
used.
Reported-by: David Ahern <david.ahern@oracle.com> Reported-by: Bob Picco <bpicco@meloft.net> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
With the increase in number of CPUs calls to functions that dump
output to console (e.g., arch_trigger_all_cpu_backtrace) can take
a long time to complete. If IRQs are disabled eventually the NMI
watchdog kicks in and creates more havoc. Avoid by telling the NMI
watchdog everything is ok.
Signed-off-by: David Ahern <david.ahern@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The reason is that state is never reset (stays with PERF_HES_UPTODATE set).
Add a call to sparc_pmu_enable_event during the added_event handling.
Clean up the encoding since pmu_start calls sparc_pmu_enable_event which
does the same. Passing PERF_EF_RELOAD to sparc_pmu_start means the call
to sparc_perf_event_set_period can be removed as well.
With this patch:
$ perf stat ls
...
Performance counter stats for 'ls':
Signed-off-by: David Ahern <david.ahern@oracle.com> Acked-by: Bob Picco <bob.picco@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
perf_pmu_disable is called by core perf code before pmu->del and the
enable function is called by core perf code afterwards. No need to
call again within sparc_pmu_del.
Ditto for pmu->add and sparc_pmu_add.
Signed-off-by: David Ahern <david.ahern@oracle.com> Acked-by: Bob Picco <bob.picco@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
A bug was reported that the semtimedop() system call was always
failing eith ENOSYS.
Since SEMCTL is defined as 3, and SEMTIMEDOP is defined as 4,
the comparison "call <= SEMCTL" will always prevent SEMTIMEDOP
from getting through to the semaphore ops switch statement.
This is corrected by changing the comparison to "call <= SEMTIMEDOP".
Load balancing can be triggered in the critical sections protected by
srmmu_context_spinlock in destroy_context() and switch_mm() and can hang
the cpu waiting for the rq lock of another cpu that in turn has called
switch_mm hangning on srmmu_context_spinlock leading to deadlock.
So, disable interrupt while taking srmmu_context_spinlock in
destroy_context() and switch_mm() so we don't deadlock.
See also commit 77b838fa1ef0 ("[SPARC64]: destroy_context() needs to disable
interrupts.")
Signed-off-by: Andreas Larsson <andreas@gaisler.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
While working on sk_forward_alloc problems reported by Denys
Fedoryshchenko, we found that tcp connect() (and fastopen) do not call
sk_wmem_schedule() for SYN packet (and/or SYN/DATA packet), so
sk_forward_alloc is negative while connect is in progress.
We can fix this by calling regular sk_stream_alloc_skb() both for the
SYN packet (in tcp_connect()) and the syn_data packet in
tcp_send_syn_data()
Then, tcp_send_syn_data() can avoid copying syn_data as we simply
can manipulate syn_data->cb[] to remove SYN flag (and increment seq)
Instead of open coding memcpy_fromiovecend(), simply use this helper.
This leaves in socket write queue clean fast clone skbs.
This was tested against our fastopen packetdrill tests.
Reported-by: Denys Fedoryshchenko <nuclearcat@nuclearcat.com> Signed-off-by: Eric Dumazet <edumazet@google.com> Acked-by: Yuchung Cheng <ycheng@google.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Commit db31c55a6fb2 (net: clamp ->msg_namelen instead of returning an
error) introduced the clamping of msg_namelen when the unsigned value
was larger than sizeof(struct sockaddr_storage). This caused a
msg_namelen of -1 to be valid. The native code was subsequently fixed by
commit dbb490b96584 (net: socket: error on a negative msg_namelen).
In addition, the native code sets msg_namelen to 0 when msg_name is
NULL. This was done in commit (6a2a2b3ae075 net:socket: set msg_namelen
to 0 if msg_name is passed as NULL in msghdr struct from userland) and
subsequently updated by 08adb7dabd48 (fold verify_iovec() into
copy_msghdr_from_user()).
This patch brings the get_compat_msghdr() in line with
copy_msghdr_from_user().
Fixes: db31c55a6fb2 (net: clamp ->msg_namelen instead of returning an error) Cc: David S. Miller <davem@davemloft.net> Cc: Dan Carpenter <dan.carpenter@oracle.com> Signed-off-by: Catalin Marinas <catalin.marinas@arm.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
tcp_send_fin() does not account for the memory it allocates properly, so
sk_forward_alloc can be negative in cases where we've sent a FIN:
ss example output (ss -amn | grep -B1 f4294):
tcp FIN-WAIT-1 0 1 192.168.0.1:45520 192.0.2.1:8080
skmem:(r0,rb87380,t0,tb87380,f4294966016,w1280,o0,bl0) Acked-by: Eric Dumazet <edumazet@google.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
for throw routes to trigger evaluation of other policy rules
EAGAIN needs to be propagated up to fib_rules_lookup
similar to how its done for IPv4
A simple testcase for verification is:
ip -6 rule add lookup 33333 priority 33333
ip -6 route add throw 2001:db8::1
ip -6 route add 2001:db8::1 via fe80::1 dev wlan0 table 33333
ip route get 2001:db8::1
Signed-off-by: Steven Barth <cyrus@openwrt.org> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The custom USB_DEVICE_CLASS macro matches
bDeviceClass, bDeviceSubClass and bDeviceProtocol
but the common USB_DEVICE_AND_INTERFACE_INFO matches
bInterfaceClass, bInterfaceSubClass and bInterfaceProtocol instead, which are
not specified.
Signed-off-by: Ondrej Zary <linux@rainbow-software.org> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
[I would really like an ACK on that one from dhowells; it appears to be
quite straightforward, but...]
MSG_PEEK isn't passed to ->recvmsg() via msg->msg_flags; as the matter of
fact, neither the kernel users of rxrpc, nor the syscalls ever set that bit
in there. It gets passed via flags; in fact, another such check in the same
function is done correctly - as flags & MSG_PEEK.
It had been that way (effectively disabled) for 8 years, though, so the patch
needs beating up - that case had never been tested. If it is correct, it's
-stable fodder.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
It should be checking flags, not msg->msg_flags. It's ->sendmsg()
instances that need to look for that in ->msg_flags, ->recvmsg() ones
(including the other ->recvmsg() instance in that file, as well as
unix_dgram_recvmsg() this one claims to be imitating) check in flags.
Braino had been introduced in commit dcda13 ("caif: Bugfix - use MSG_TRUNC
in receive") back in 2010, so it goes quite a while back.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The rds_iw_update_cm_id function stores a large 'struct rds_sock' object
on the stack in order to pass a pair of addresses. This happens to just
fit withint the 1024 byte stack size warning limit on x86, but just
exceed that limit on ARM, which gives us this warning:
net/rds/iw_rdma.c:200:1: warning: the frame size of 1056 bytes is larger than 1024 bytes [-Wframe-larger-than=]
As the use of this large variable is basically bogus, we can rearrange
the code to not do that. Instead of passing an rds socket into
rds_iw_get_device, we now just pass the two addresses that we have
available in rds_iw_update_cm_id, and we change rds_iw_get_mr accordingly,
to create two address structures on the stack there.
Signed-off-by: Arnd Bergmann <arnd@arndb.de> Acked-by: Sowmini Varadhan <sowmini.varadhan@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
sysctl has sysctl.net.core.rmem_*/wmem_* parameters which can be
set to incorrect values. Given that 'struct sk_buff' allocates from
rcvbuf, incorrectly set buffer length could result to memory
allocation failures. For example, set them as follows:
Moreover, the possible minimum is 1, so we can get another kernel panic:
...
BUG: unable to handle kernel paging request at ffff88013caee5c0
IP: [<ffffffff815604cf>] __alloc_skb+0x12f/0x1f0
...
Signed-off-by: Alexey Kodanev <alexey.kodanev@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The commit [63e51fd708f5: ALSA: hda - Don't take unresponsive D3
transition too serious] introduced a conditional fallback behavior to
the HD-audio controller depending on the flag set. However, it
introduced a silly bug, too, that the flag was evaluated in a reverse
way. This resulted in a regression of HD-audio controller driver
where it can't go to the fallback mode at communication errors.
Unfortunately (or fortunately?) this didn't come up until recently
because the affected code path is an error handling that happens only
on an unstable hardware chip. Most of recent chips work stably, thus
they didn't hit this problem. Now, we've got a regression report with
a VIA chip, and this seems indeed requiring the fallback to the
polling mode, and finally the bug was revealed.
The fix is a oneliner to remove the wrong logical NOT in the check.
(Lesson learned - be careful about double negation.)
The bug should be backported to stable, but the patch won't be
applicable to 3.13 or earlier because of the code splits. The stable
fix patches for earlier kernels will be posted later manually.
When running the test which causes the race as shown in the previous patch,
we can hit the BUG "get_page() on refcount 0 page" in hugetlb_fault().
This race happens when pte turns into migration entry just after the first
check of is_hugetlb_entry_migration() in hugetlb_fault() passed with false.
To fix this, we need to check pte_present() again after huge_ptep_get().
This patch also reorders taking ptl and doing pte_page(), because
pte_page() should be done in ptl. Due to this reordering, we need use
trylock_page() in page != pagecache_page case to respect locking order.
Fixes: 66aebce747ea ("hugetlb: fix race condition in hugetlb_fault()") Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com> Cc: Hugh Dickins <hughd@google.com> Cc: James Hogan <james.hogan@imgtec.com> Cc: David Rientjes <rientjes@google.com> Cc: Mel Gorman <mel@csn.ul.ie> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Michal Hocko <mhocko@suse.cz> Cc: Rik van Riel <riel@redhat.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Luiz Capitulino <lcapitulino@redhat.com> Cc: Nishanth Aravamudan <nacc@linux.vnet.ibm.com> Cc: Lee Schermerhorn <lee.schermerhorn@hp.com> Cc: Steve Capper <steve.capper@linaro.org> Cc: <stable@vger.kernel.org> [3.2+] Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Jiri Slaby <jslaby@suse.cz> [backport to 3.12]
In case CLK_GATE_HIWORD_MASK flag is passed to clk_register_gate(), the bit #
should be no higher than 15, however the corresponding check is obviously off-
by-one.
Fixes: 045779942c04 ("clk: gate: add CLK_GATE_HIWORD_MASK") Signed-off-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com> Signed-off-by: Michael Turquette <mturquette@linaro.org> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Sometimes while CPU have some load and ath5k doing the wireless
interface reset the whole WiSoC completely freezes. Set of tests shows
that using atomic delay function while we wait interface reset helps to
avoid such freezes.
The easiest way to reproduce this issue: create a station interface,
start continous scan with wpa_supplicant and load CPU by something. Or
just create multiple station interfaces and put them all in continous
scan.
This patch partially reverts the commit 1846ac3dbec0 ("ath5k: Use
usleep_range where possible"), which replaces initial udelay()
by usleep_range().
I do not know actual source of this issue, but all looks like that HW
freeze is caused by transaction on internal SoC bus, while wireless
block is in reset state.
Also I should note that I do not know how many chips are affected, but I
did not see this issue with chips, other than AR5312.
CC: Jiri Slaby <jirislaby@gmail.com> CC: Nick Kossifidis <mickflemm@gmail.com> CC: Luis R. Rodriguez <mcgrof@do-not-panic.com> Fixes: 1846ac3dbec0 ("ath5k: Use usleep_range where possible") Reported-by: Christophe Prevotaux <c.prevotaux@rural-networks.com> Tested-by: Christophe Prevotaux <c.prevotaux@rural-networks.com> Tested-by: Eric Bree <ebree@nltinc.com> Signed-off-by: Sergey Ryazanov <ryazanov.s.a@gmail.com> Signed-off-by: Kalle Valo <kvalo@codeaurora.org> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
i915.ko depends upon the acpi/video.ko module and so refuses to load if
ACPI is disabled at runtime if for example the BIOS is broken beyond
repair. acpi/video provides an optional service for i915.ko and so we
should just allow the modules to load, but do no nothing in order to let
the machines boot correctly.
Reported-by: Bill Augur <bill-auger@programmer.net> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Daniel Vetter <daniel.vetter@ffwll.ch> Cc: Jani Nikula <jani.nikula@intel.com> Acked-by: Aaron Lu <aaron.lu@intel.com>
[ rjw: Fixed up the new comment in acpi_video_init() ] Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The ignore check that got added in 6ce901eb61 ("HID: input: fix confusion
on conflicting mappings") needs to properly check for VARIABLE reports
as well (ARRAY reports should be ignored), otherwise legitimate keyboards
might break.
Fixes: 6ce901eb61 ("HID: input: fix confusion on conflicting mappings") Reported-by: Fredrik Hallenberg <megahallon@gmail.com> Reported-by: David Herrmann <dh.herrmann@gmail.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
(Note that the number of keys is the same, but key '5' is moved down and
the shape of key '4' is changed. Keys '1' to '3' are exactly the same.)
The keys 1-4 report the same scan-code in HID in both layouts, even though
the keysym they produce is usually different depending on the XKB-keymap
used by user-space.
However, key '5' (US 'backslash'/'pipe') reports 0x31 for the upper layout
and 0x32 for the lower layout, as defined by the HID spec. This is highly
confusing as the linux-input API uses a single keycode for both.
So far, this was never a problem as there never has been a keyboard with
both of those keys present at the same time. It would have to look
something like this:
HID can represent such a keyboard, but the linux-input API cannot.
Furthermore, any user-space mapping would be confused by this and,
luckily, no-one ever produced such hardware.
Now, the HID input layer fixed this mess by mapping both 0x31 and 0x32 to
the same keycode (KEY_BACKSLASH==0x2b). As only one of both physical keys
is present on a hardware, this works just fine.
Lets introduce hardware-vendors into this:
------------------------------------------
Unfortunately, it seems way to expensive to produce a different device for
American and European layouts. Therefore, hardware-vendors put both keys,
(0x31 and 0x32) on the same keyboard, but only one of them is hooked up
to the physical button, the other one is 'dead'.
This means, they can use the same hardware, with a different button-layout
and automatically produce the correct HID events for American *and*
European layouts. This is unproblematic for normal keyboards, as the
'dead' key will never report any KEY-DOWN events. But RollOver keyboards
send the whole matrix on each key-event, allowing n-key roll-over mode.
This means, we get a 0x31 and 0x32 event on each key-press. One of them
will always be 0, the other reports the real state. As we map both to the
same keycode, we will get spurious key-events, even though the real
key-state never changed.
The easiest way would be to blacklist 'dead' keys and never handle those.
We could simply read the 'country' tag of USB devices and blacklist either
key according to the layout. But... hardware vendors... want the same
device for all countries and thus many of them set 'country' to 0 for all
devices. Meh..
So we have to deal with this properly. As we cannot know which of the keys
is 'dead', we either need a heuristic and track those keys, or we simply
make use of our value-tracking for HID fields. We simply ignore HID events
for absolute data if the data didn't change. As HID tracks events on the
HID level, we haven't done the keycode translation, yet. Therefore, the
'dead' key is tracked independently of the real key, therefore, any events
on it will be ignored.
This patch simply discards any HID events for absolute data if it didn't
change compared to the last report. We need to ignore relative and
buffered-byte reports for obvious reasons. But those cannot be affected by
this bug, so we're fine.
Preferably, we'd do this filtering on the HID-core level. But this might
break a lot of custom drivers, if they do not follow the HID specs.
Therefore, we do this late in hid-input just before we inject it into the
input layer (which does the exact same filtering, but on the keycode
level).
If this turns out to break some devices, we might have to limit filtering
to EV_KEY events. But lets try to do the Right Thing first, and properly
filter any absolute data that didn't change.
This patch is tagged for 'stable' as it fixes a lot of n-key RollOver
hardware. We might wanna wait with backporting for a while, before we know
it doesn't break anything else, though.
Reported-by: Adam Goode <adam@spicenitz.org> Reported-by: Fredrik Hallenberg <megahallon@gmail.com> Tested-by: Fredrik Hallenberg <megahallon@gmail.com> Signed-off-by: David Herrmann <dh.herrmann@gmail.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
When the snapshot target is unloaded, snapshot_dtr() waits until
pending_exceptions_count drops to zero. Then, it destroys the snapshot.
Therefore, the function that decrements pending_exceptions_count
should not touch the snapshot structure after the decrement.
pending_complete() calls free_pending_exception(), which decrements
pending_exceptions_count, and then it performs up_write(&s->lock) and it
calls retry_origin_bios() which dereferences s->origin. These two
memory accesses to the fields of the snapshot may touch the dm_snapshot
struture after it is freed.
This patch moves the call to free_pending_exception() to the end of
pending_complete(), so that the snapshot will not be destroyed while
pending_complete() is in progress.
The function dm_get_md finds a device mapper device with a given dev_t,
increases the reference count and returns the pointer.
dm_get_md calls dm_find_md, dm_find_md takes _minor_lock, finds the
device, tests that the device doesn't have DMF_DELETING or DMF_FREEING
flag, drops _minor_lock and returns pointer to the device. dm_get_md then
calls dm_get. dm_get calls BUG if the device has the DMF_FREEING flag,
otherwise it increments the reference count.
There is a possible race condition - after dm_find_md exits and before
dm_get is called, there are no locks held, so the device may disappear or
DMF_FREEING flag may be set, which results in BUG.
To fix this bug, we need to call dm_get while we hold _minor_lock. This
patch renames dm_find_md to dm_get_md and changes it so that it calls
dm_get while holding the lock.
I created a dm-raid1 device backed by a device that supports DISCARD
and another device that does NOT support DISCARD with the following
dm configuration:
Notice that the mirror device /dev/mapper/moo advertises DISCARD
support even though one of the mirror halves doesn't.
If I issue a DISCARD request (via fstrim, mount -o discard, or ioctl
BLKDISCARD) through the mirror, kmirrord gets stuck in an infinite
loop in do_region() when it tries to issue a DISCARD request to sdb.
The problem is that when we call do_region() against sdb, num_sectors
is set to zero because q->limits.max_discard_sectors is zero.
Therefore, "remaining" never decreases and the loop never terminates.
To fix this: before entering the loop, check for the combination of
REQ_DISCARD and no discard and return -EOPNOTSUPP to avoid hanging up
the mirror device.
This bug was found by the unfortunate coincidence of pvmove and a
discard operation in the RHEL 6.5 kernel; upstream is also affected.
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Acked-by: "Martin K. Petersen" <martin.petersen@oracle.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
It may be possible that a device claims discard support but it rejects
discards with -EOPNOTSUPP. It happens when using loopback on ext2/ext3
filesystem driven by the ext4 driver. It may also happen if the
underlying devices are moved from one disk on another.
If discard error happens, we reject the bio with -EOPNOTSUPP, but we do
not degrade the array.
This patch fixes failed test shell/lvconvert-repair-transient.sh in the
lvm2 testsuite if the testsuite is extracted on an ext2 or ext3
filesystem and it is being driven by the ext4 driver.
`do_cmd_ioctl()` in "comedi_fops.c" handles the `COMEDI_CMD` ioctl.
This returns `-EAGAIN` if it has copied a modified `struct comedi_cmd`
back to user-space. (This occurs when the low-level Comedi driver's
`do_cmdtest()` handler returns non-zero to indicate a problem with the
contents of the `struct comedi_cmd`, or when the `struct comedi_cmd` has
the `CMDF_BOGUS` flag set.)
`compat_cmd()` in "comedi_compat32.c" handles the 32-bit compatible
version of the `COMEDI_CMD` ioctl. Currently, it never copies a 32-bit
compatible version of `struct comedi_cmd` back to user-space, which is
at odds with the way the regular `COMEDI_CMD` ioctl is handled. To fix
it, change `compat_cmd()` to copy a 32-bit compatible version of the
`struct comedi_cmd` back to user-space when the main ioctl handler
returns `-EAGAIN`.
Signed-off-by: Ian Abbott <abbotti@mev.co.uk> Reviewed-by: H Hartley Sweeten <hsweeten@visionengravers.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
For all pll-s on sun6i n == 0 means use a multiplier of 1, rather then 0 as
it means on sun4i / sun5i / sun7i. n_start = 1 is already correctly set
for sun6i pll6, but was missing for pll1, this commit fixes this.
Cc: Chen-Yu Tsai <wens@csie.org> Signed-off-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The CPU_2X clock does not have a classical in-kernel user, but is,
amongst other things, required for OCM and debug access. Make sure this
clock is not mistakenly disabled during boot up by enabling it in the
platform's clock driver.
DMA_BIT_MASK of 64 is not valid dma address mask for OMAPs, it should
be set to 32.
The 64 was introduced by commit (in 2009): a152ff24b978 ASoC: OMAP: Make DMA 64 aligned
But the dma_mask and coherent_dma_mask can not be used to specify
alignment.
Fixes: a152ff24b978 (ASoC: OMAP: Make DMA 64 aligned) Reported-by: Grygorii Strashko <Grygorii.Strashko@linaro.org> Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com> Signed-off-by: Mark Brown <broonie@kernel.org> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Each inode of nilfs2 stores a root node of a b-tree, and it turned out to
have a memory overrun issue:
Each b-tree node of nilfs2 stores a set of key-value pairs and the number
of them (in "bn_nchildren" member of nilfs_btree_node struct), as well as
a few other "bn_*" members.
Since the value of "bn_nchildren" is used for operations on the key-values
within the b-tree node, it can cause memory access overrun if a large
number is incorrectly set to "bn_nchildren".
For instance, nilfs_btree_node_lookup() function determines the range of
binary search with it, and too large "bn_nchildren" leads
nilfs_btree_node_get_key() in that function to overrun.
As for intermediate b-tree nodes, this is prevented by a sanity check
performed when each node is read from a drive, however, no sanity check
has been done for root nodes stored in inodes.
This patch fixes the issue by adding missing sanity check against b-tree
root nodes so that it's called when on-memory inodes are read from ifile,
inode metadata file.
We have a race condition between move_pages() and freeing hugepages, where
move_pages() calls follow_page(FOLL_GET) for hugepages internally and
tries to get its refcount without preventing concurrent freeing. This
race crashes the kernel, so this patch fixes it by moving FOLL_GET code
for hugepages into follow_huge_pmd() with taking the page table lock.
This patch intentionally removes page==NULL check after pte_page.
This is justified because pte_page() never returns NULL for any
architectures or configurations.
This patch changes the behavior of follow_huge_pmd() for tail pages and
then tail pages can be pinned/returned. So the caller must be changed to
properly handle the returned tail pages.
We could have a choice to add the similar locking to
follow_huge_(addr|pud) for consistency, but it's not necessary because
currently these functions don't support FOLL_GET flag, so let's leave it
for future development.
int main(int argc, char *argv[]) {
int i;
int nr_hp = strtol(argv[1], NULL, 0);
int nr_p = nr_hp * HPS / PS;
int ret;
void **addrs;
int *status;
int *nodes;
pid_t pid;
Currently we have many duplicates in definitions around
follow_huge_addr(), follow_huge_pmd(), and follow_huge_pud(), so this
patch tries to remove the m. The basic idea is to put the default
implementation for these functions in mm/hugetlb.c as weak symbols
(regardless of CONFIG_ARCH_WANT_GENERAL_HUGETL B), and to implement
arch-specific code only when the arch needs it.
For follow_huge_addr(), only powerpc and ia64 have their own
implementation, and in all other architectures this function just returns
ERR_PTR(-EINVAL). So this patch sets returning ERR_PTR(-EINVAL) as
default.
As for follow_huge_(pmd|pud)(), if (pmd|pud)_huge() is implemented to
always return 0 in your architecture (like in ia64 or sparc,) it's never
called (the callsite is optimized away) no matter how implemented it is.
So in such architectures, we don't need arch-specific implementation.
In some architecture (like mips, s390 and tile,) their current
arch-specific follow_huge_(pmd|pud)() are effectively identical with the
common code, so this patch lets these architecture use the common code.
One exception is metag, where pmd_huge() could return non-zero but it
expects follow_huge_pmd() to always return NULL. This means that we need
arch-specific implementation which returns NULL. This behavior looks
strange to me (because non-zero pmd_huge() implies that the architecture
supports PMD-based hugepage, so follow_huge_pmd() can/should return some
relevant value,) but that's beyond this cleanup patch, so let's keep it.
Justification of non-trivial changes:
- in s390, follow_huge_pmd() checks !MACHINE_HAS_HPAGE at first, and this
patch removes the check. This is OK because we can assume MACHINE_HAS_HPAGE
is true when follow_huge_pmd() can be called (note that pmd_huge() has
the same check and always returns 0 for !MACHINE_HAS_HPAGE.)
- in s390 and mips, we use HPAGE_MASK instead of PMD_MASK as done in common
code. This patch forces these archs use PMD_MASK, but it's OK because
they are identical in both archs.
In s390, both of HPAGE_SHIFT and PMD_SHIFT are 20.
In mips, HPAGE_SHIFT is defined as (PAGE_SHIFT + PAGE_SHIFT - 3) and
PMD_SHIFT is define as (PAGE_SHIFT + PAGE_SHIFT + PTE_ORDER - 3), but
PTE_ORDER is always 0, so these are identical.
Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com> Acked-by: Hugh Dickins <hughd@google.com> Cc: James Hogan <james.hogan@imgtec.com> Cc: David Rientjes <rientjes@google.com> Cc: Mel Gorman <mel@csn.ul.ie> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Michal Hocko <mhocko@suse.cz> Cc: Rik van Riel <riel@redhat.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Luiz Capitulino <lcapitulino@redhat.com> Cc: Nishanth Aravamudan <nacc@linux.vnet.ibm.com> Cc: Lee Schermerhorn <lee.schermerhorn@hp.com> Cc: Steve Capper <steve.capper@linaro.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
A race condition starts to be visible in recent mmotm, where a PG_hwpoison
flag is set on a migration source page *before* it's back in buddy page
poo= l.
This is problematic because no page flag is supposed to be set when
freeing (see __free_one_page().) So the user-visible effect of this race
is that it could trigger the BUG_ON() when soft-offlining is called.
The root cause is that we call lru_add_drain_all() to make sure that the
page is in buddy, but that doesn't work because this function just
schedule= s a work item and doesn't wait its completion.
drain_all_pages() does drainin= g directly, so simply dropping
lru_add_drain_all() solves this problem.
Microsoft Natural Wireless Ergonomic Keyboard 7000 has special My
Favorites 1..5 keys which are handled through a vendor-defined usage
page (0xff05).
Apply MS_ERGONOMY quirks handling to USB PID 0x071d (Microsoft Microsoft
2.4GHz Transceiver V1.0) so that the My Favorites 1..5 keys are reported
as KEY_F14..18 events.
The switch back is limited to ULT even on HP. The contrary
finding arose by bad luck in BIOS versions for testing.
This fixes spontaneous resume from S3 on some HP laptops.
This changeset removes all the code that allows the driver to write to
the EEPROM and update the recorded error counters and power on hours.
These two stats are unused and writing them exposes a timing risk
which could leave the EEPROM in a bad state preventing further normal
operation of the HCA.
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Mitko Haralanov <mitko.haralanov@intel.com> Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Roland Dreier <roland@purestorage.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
When a PCM draining is performed to an empty stream that has been
already in PREPARED state, the current code just ignores and leaves as
it is, although the drain is supposed to set all such streams to SETUP
state. This patch covers that overlooked case.
Ben Hutchings [Thu, 29 Jan 2015 02:50:33 +0000 (02:50 +0000)]
splice: Apply generic position and size checks to each write
We need to check the position and size of file writes against various
limits, using generic_write_check(). This was not being done for
the splice write path. It was fixed upstream by commit 8d0207652cbe
("->splice_write() via ->write_iter()") but we can't apply that.
CVE-2014-7822
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
[ kamal: port to 3.13-stable: context ] Signed-off-by: Kamal Mostafa <kamal@canonical.com> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The stack guard page error case has long incorrectly caused a SIGBUS
rather than a SIGSEGV, but nobody actually noticed until commit fee7e49d4514 ("mm: propagate error from stack expansion even for guard
page") because that error case was never actually triggered in any
normal situations.
Now that we actually report the error, people noticed the wrong signal
that resulted. So far, only the test suite of libsigsegv seems to have
actually cared, but there are real applications that use libsigsegv, so
let's not wait for any of those to break.
Reported-and-tested-by: Takashi Iwai <tiwai@suse.de> Tested-by: Jan Engelhardt <jengelh@inai.de> Acked-by: Heiko Carstens <heiko.carstens@de.ibm.com> # "s390 still compiles and boots" Cc: linux-arch@vger.kernel.org Cc: stable@vger.kernel.org Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
The core VM already knows about VM_FAULT_SIGBUS, but cannot return a
"you should SIGSEGV" error, because the SIGSEGV case was generally
handled by the caller - usually the architecture fault handler.
That results in lots of duplication - all the architecture fault
handlers end up doing very similar "look up vma, check permissions, do
retries etc" - but it generally works. However, there are cases where
the VM actually wants to SIGSEGV, and applications _expect_ SIGSEGV.
In particular, when accessing the stack guard page, libsigsegv expects a
SIGSEGV. And it usually got one, because the stack growth is handled by
that duplicated architecture fault handler.
However, when the generic VM layer started propagating the error return
from the stack expansion in commit fee7e49d4514 ("mm: propagate error
from stack expansion even for guard page"), that now exposed the
existing VM_FAULT_SIGBUS result to user space. And user space really
expected SIGSEGV, not SIGBUS.
To fix that case, we need to add a VM_FAULT_SIGSEGV, and teach all those
duplicate architecture fault handlers about it. They all already have
the code to handle SIGSEGV, so it's about just tying that new return
value to the existing code, but it's all a bit annoying.
This is the mindless minimal patch to do this. A more extensive patch
would be to try to gather up the mostly shared fault handling logic into
one generic helper routine, and long-term we really should do that
cleanup.
Just from this patch, you can generally see that most architectures just
copied (directly or indirectly) the old x86 way of doing things, but in
the meantime that original x86 model has been improved to hold the VM
semaphore for shorter times etc and to handle VM_FAULT_RETRY and other
"newer" things, so it would be a good idea to bring all those
improvements to the generic case and teach other architectures about
them too.
Reported-and-tested-by: Takashi Iwai <tiwai@suse.de> Tested-by: Jan Engelhardt <jengelh@inai.de> Acked-by: Heiko Carstens <heiko.carstens@de.ibm.com> # "s390 still compiles and boots" Cc: linux-arch@vger.kernel.org Cc: stable@vger.kernel.org Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
This replaces four copies in various stages of mm_fault_error() handling
with just a single one. It will also allow for more natural placement
of the unlocking after some further cleanup.
projects / linux / log