net: qualcomm: rmnet: Fix warning seen with fill_info
When the last rmnet device attached to a real device is removed, the
real device is unregistered from rmnet. As a result, the real device
lookup fails resulting in a warning when the fill_info handler is
called as part of the rmnet device unregistration.
Fix this by returning the rmnet flags as 0 when no real device is
present.
David Sterba [Mon, 16 Apr 2018 19:10:14 +0000 (21:10 +0200)]
btrfs: fix unaligned access in readdir
The last update to readdir introduced a temporary buffer to store the
emitted readdir data, but as there are file names of variable length,
there's a lot of unaligned access.
This was observed on a sparc64 machine:
Kernel unaligned access at TPC[102f3080] btrfs_real_readdir+0x51c/0x718 [btrfs]
Matt Redfearn [Tue, 17 Apr 2018 15:40:00 +0000 (16:40 +0100)]
MIPS: memset.S: Fix clobber of v1 in last_fixup
The label .Llast_fixup\@ is jumped to on page fault within the final
byte set loop of memset (on < MIPSR6 architectures). For some reason, in
this fault handler, the v1 register is randomly set to a2 & STORMASK.
This clobbers v1 for the calling function. This can be observed with the
following test code:
static int __init __attribute__((optimize("O0"))) test_clear_user(void)
{
register int t asm("v1");
char *test;
int j, k;
pr_info("\n\n\nTesting clear_user\n");
test = vmalloc(PAGE_SIZE);
for (j = 256; j < 512; j++) {
t = 0xa5a5a5a5;
if ((k = clear_user(test + PAGE_SIZE - 256, j)) != j - 256) {
pr_err("clear_user (%px %d) returned %d\n", test + PAGE_SIZE - 256, j, k);
}
if (t != 0xa5a5a5a5) {
pr_err("v1 was clobbered to 0x%x!\n", t);
}
}
return 0;
}
late_initcall(test_clear_user);
Which demonstrates that v1 is indeed clobbered (MIPS64):
Testing clear_user
v1 was clobbered to 0x1!
v1 was clobbered to 0x2!
v1 was clobbered to 0x3!
v1 was clobbered to 0x4!
v1 was clobbered to 0x5!
v1 was clobbered to 0x6!
v1 was clobbered to 0x7!
Since the number of bytes that could not be set is already contained in
a2, the andi placing a value in v1 is not necessary and actively
harmful in clobbering v1.
Merge tag 'ceph-for-4.17-rc2' of git://github.com/ceph/ceph-client
Pull ceph fixes from Ilya Dryomov:
"A couple of follow-up patches for -rc1 changes in rbd, support for a
timeout on waiting for the acquisition of exclusive lock and a fix for
uninitialized memory access in CephFS, marked for stable"
* tag 'ceph-for-4.17-rc2' of git://github.com/ceph/ceph-client:
rbd: notrim map option
rbd: adjust queue limits for "fancy" striping
rbd: avoid Wreturn-type warnings
ceph: always update atime/mtime/ctime for new inode
rbd: support timeout in rbd_wait_state_locked()
rbd: refactor rbd_wait_state_locked()
perf test BPF: Fixup BPF test using epoll_pwait syscall function probe
Since e145242ea0df ("syscalls/core, syscalls/x86: Clean up syscall stub
naming convention") changed the main syscall function for 'epoll_pwait'
to something other than the expected 'SyS_epoll_pwait the' 'perf test
BPF' entries started failing, fix it by using something called from the
main syscall function instead, 'epoll_wait', which should keep this test
working in older kernels too.
In the 'perf test "mmap interface"' we try creating events for several
tracepoints, but when perf_evsel__new() fails we're not showing which
one is failing, fix that to help diagnosing problems, such as the
syscall tracepoints ones being found and fixes in this merge window.
Now the failing tests shows:
# perf test -v "mmap interface"
4: Read samples using the mmap interface :
--- start ---
test child forked, pid 14311
<SNIP>
perf_evsel__new(sys_enter_getppid)
test child finished with -1
---- end ----
Read samples using the mmap interface: FAILED!
#
Now to check why the syscalls:sys_enter_getppid is failing...
# ls -la /sys/kernel/debug/tracing/events/syscalls/sys_enter_getppid
ls: cannot access '/sys/kernel/debug/tracing/events/syscalls/sys_enter_getppid': No such file or directory
#
Introduced in a4ff8e8620d3 ("mm: introduce MAP_FIXED_NOREPLACE"), and
now that we have that define in the just syncronized
tools/arch/*/include/uapi/asm/mman.h files, add support for it.
This should really transition to autogeneration of string tables as
done for various other things:
Also it has to be ifdef'ed, because arches like mips still don't define
it. The proper solution will be to have per-arch tables for these
values to support cross-analysis.
Bogus trimming in tun_net_xmit() causes truncated vlan packets.
skb->len is correct whether or not skb_vlan_tag_present() is true. There
is no more reason to adjust the skb length on xmit in this driver than
any other driver. tun_put_user() adds 4 bytes to the total for tagged
packets because it transmits the tag inline to userspace. This is
similar to a nic transmitting the tag inline on the wire.
Reproducing the bug by sending any tagged packet through back-to-back
connected tap interfaces:
socat TUN,tun-type=tap,iff-up,tun-name=in TUN,tun-type=tap,iff-up,tun-name=out &
ip link add link in name in.20 type vlan id 20
ip addr add 10.9.9.9/24 dev in.20
ip link set in.20 up
tshark -nxxi in -f arp -c1 2>/dev/null &
tshark -nxxi out -f arp -c1 2>/dev/null &
ping -c 1 10.9.9.5 >/dev/null 2>&1
The output from the 'in' and 'out' interfaces are different when the
bug is present:
tipc: fix infinite loop when dumping link monitor summary
When configuring the number of used bearers to MAX_BEARER and issuing
command "tipc link monitor summary", the command enters infinite loop
in user space.
This issue happens because function tipc_nl_node_dump_monitor() returns
the wrong 'prev_bearer' value when all potential monitors have been
scanned.
The correct behavior is to always try to scan all monitors until either
the netlink message is full, in which case we return the bearer identity
of the affected monitor, or we continue through the whole bearer array
until we can return MAX_BEARERS. This solution also caters for the case
where there may be gaps in the bearer array.
Jon Maloy [Tue, 17 Apr 2018 19:25:42 +0000 (21:25 +0200)]
tipc: fix use-after-free in tipc_nametbl_stop
When we delete a service item in tipc_nametbl_stop() we loop over
all service ranges in the service's RB tree, and for each service
range we loop over its pertaining publications while calling
tipc_service_remove_publ() for each of them.
However, tipc_service_remove_publ() has the side effect that it also
removes the comprising service range item when there are no publications
left. This leads to a "use-after-free" access when the inner loop
continues to the next iteration, since the range item holding the list
we are looping no longer exists.
We fix this by moving the delete of the service range item outside
the said function. Instead, we now let the two functions calling it
test if the list is empty and perform the removal when that is the
case.
powerpc/xive: Fix trying to "push" an already active pool VP
When setting up a CPU, we "push" (activate) a pool VP for it.
However it's an error to do so if it already has an active
pool VP.
This happens when doing soft CPU hotplug on powernv since we
don't tear down the CPU on unplug. The HW flags the error which
gets captured by the diagnostics.
Fix this by making sure to "pull" out any already active pool
first.
Fixes: 243e25112d06 ("powerpc/xive: Native exploitation of the XIVE interrupt controller") Cc: [email protected] # v4.12+ Signed-off-by: Benjamin Herrenschmidt <[email protected]> Signed-off-by: Michael Ellerman <[email protected]>
Commit 43b18595d660 ("btrfs: qgroup: Use separate meta reservation type
for delalloc") merged into mainline is not the latest version submitted
to mail list in Dec 2017.
It has a fatal wrong @qgroup_free parameter, which results increasing
qgroup metadata pertrans reserved space, and causing a lot of early EDQUOT.
Fix it by applying the correct diff on top of current branch.
Fixes: 43b18595d660 ("btrfs: qgroup: Use separate meta reservation type for delalloc") Signed-off-by: Qu Wenruo <[email protected]> Signed-off-by: David Sterba <[email protected]>
btrfs: delayed-inode: Remove wrong qgroup meta reservation calls
Commit 4f5427ccce5d ("btrfs: delayed-inode: Use new qgroup meta rsv for
delayed inode and item") merged into mainline was not latest version
submitted to the mail list in Dec 2017.
Which lacks the following fixes:
1) Remove btrfs_qgroup_convert_reserved_meta() call in
btrfs_delayed_item_release_metadata()
2) Remove btrfs_qgroup_reserve_meta_prealloc() call in
btrfs_delayed_inode_reserve_metadata()
Those fixes will resolve unexpected EDQUOT problems.
Fixes: 4f5427ccce5d ("btrfs: delayed-inode: Use new qgroup meta rsv for delayed inode and item") Signed-off-by: Qu Wenruo <[email protected]> Signed-off-by: David Sterba <[email protected]>
Qu Wenruo [Thu, 21 Dec 2017 05:42:04 +0000 (13:42 +0800)]
btrfs: qgroup: Use independent and accurate per inode qgroup rsv
Unlike reservation calculation used in inode rsv for metadata, qgroup
doesn't really need to care about things like csum size or extent usage
for the whole tree COW.
Qgroups care more about net change of the extent usage.
That's to say, if we're going to insert one file extent, it will mostly
find its place in COWed tree block, leaving no change in extent usage.
Or causing a leaf split, resulting in one new net extent and increasing
qgroup number by nodesize.
Or in an even more rare case, increase the tree level, increasing qgroup
number by 2 * nodesize.
So here instead of using the complicated calculation for extent
allocator, which cares more about accuracy and no error, qgroup doesn't
need that over-estimated reservation.
This patch will maintain 2 new members in btrfs_block_rsv structure for
qgroup, using much smaller calculation for qgroup rsv, reducing false
EDQUOT.
Qu Wenruo [Fri, 22 Dec 2017 08:06:39 +0000 (16:06 +0800)]
btrfs: qgroup: Commit transaction in advance to reduce early EDQUOT
Unlike previous method that tries to commit transaction inside
qgroup_reserve(), this time we will try to commit transaction using
fs_info->transaction_kthread to avoid nested transaction and no need to
worry about locking context.
Since it's an asynchronous function call and we won't wait for
transaction commit, unlike previous method, we must call it before we
hit the qgroup limit.
So this patch will use the ratio and size of qgroup meta_pertrans
reservation as indicator to check if we should trigger a transaction
commit. (meta_prealloc won't be cleaned in transaction committ, it's
useless anyway)
Jan Kara [Thu, 12 Apr 2018 15:22:23 +0000 (17:22 +0200)]
udf: Fix leak of UTF-16 surrogates into encoded strings
OSTA UDF specification does not mention whether the CS0 charset in case
of two bytes per character encoding should be treated in UTF-16 or
UCS-2. The sample code in the standard does not treat UTF-16 surrogates
in any special way but on systems such as Windows which work in UTF-16
internally, filenames would be treated as being in UTF-16 effectively.
In Linux it is more difficult to handle characters outside of Base
Multilingual plane (beyond 0xffff) as NLS framework works with 2-byte
characters only. Just make sure we don't leak UTF-16 surrogates into the
resulting string when loading names from the filesystem for now.
Mark Rutland [Mon, 16 Apr 2018 15:45:01 +0000 (16:45 +0100)]
arm64: signal: don't force known signals to SIGKILL
Since commit:
a7e6f1ca90354a31 ("arm64: signal: Force SIGKILL for unknown signals in force_signal_inject")
... any signal which is not SIGKILL will be upgraded to a SIGKILL be
force_signal_inject(). This includes signals we do expect, such as
SIGILL triggered by do_undefinstr().
Fix the check to use a logical AND rather than a logical OR, permitting
signals whose layout is SIL_FAULT.
Fixes: a7e6f1ca90354a31 ("arm64: signal: Force SIGKILL for unknown signals in force_signal_inject") Cc: Will Deacon <[email protected]> Reviewed-by: Dave Martin <[email protected]> Signed-off-by: Mark Rutland <[email protected]> Signed-off-by: Catalin Marinas <[email protected]>
The "cooling-min-level" and "cooling-max-level" properties are not
parsed by any part of kernel currently and the max cooling state of a
CPU cooling device is found by referring to the cpufreq table instead.
Imre Deak [Mon, 16 Apr 2018 15:53:09 +0000 (18:53 +0300)]
drm/i915: Fix LSPCON TMDS output buffer enabling from low-power state
LSPCON adapters in low-power state may ignore the first I2C write during
TMDS output buffer enabling, resulting in a blank screen even with an
otherwise enabled pipe. Fix this by reading back and validating the
written value a few times.
The problem was noticed on GLK machines with an onboard LSPCON adapter
after entering/exiting DC5 power state. Doing an I2C read of the adapter
ID as the first transaction - instead of the I2C write to enable the
TMDS buffers - returns the correct value. Based on this we assume that
the transaction itself is sent properly, it's only the adapter that is
not ready for some reason to accept this first write after waking from
low-power state. In my case the second I2C write attempt always
succeeded.
On Geminilake, sometimes audio card is not getting
detected after reboot. This is a spurious issue happening on
Geminilake. HW codec and HD audio controller link was going
out of sync for which there was a fix in i915 driver but
was not getting invoked for GLK. Extending this fix to GLK as well.
Tested by Du,Wenkai on GLK board.
Bspec: 21829
v2: Instead of checking GEN9_BC, BXT and GLK macros, use IS_GEN9 macro (Jani N)
Chris Wilson [Sat, 14 Apr 2018 09:12:33 +0000 (10:12 +0100)]
drm/i915: Call i915_perf_fini() on init_hw error unwind
We have to cleanup after i915_perf_init(), even on the error path, as it
passes a pointer into the module to the sysfs core. If we fail to
unregister the sysctl table, we leave a dangling pointer which then may
explode anytime later.
Jani Nikula [Wed, 11 Apr 2018 13:15:18 +0000 (16:15 +0300)]
drm/i915/bios: filter out invalid DDC pins from VBT child devices
The VBT contains the DDC pin to use for specific ports. Alas, sometimes
the field appears to contain bogus data, and while we check for it later
on in intel_gmbus_get_adapter() we fail to check the returned NULL on
errors. Oops results.
The simplest approach seems to be to catch and ignore the bogus DDC pins
already at the VBT parsing phase, reverting to fixed per port default
pins. This doesn't guarantee display working, but at least it prevents
the oops. And we continue to be fuzzed by VBT.
One affected machine is Dell Latitude 5590 where a BIOS upgrade added
invalid DDC pins.
drm/i915/pmu: Inspect runtime PM state more carefully while estimating RC6
While thinking about sporadic failures of perf_pmu/rc6-runtime-pm* tests
on some CI machines I have concluded that: a) the PMU readout of RC6 can
race against runtime PM transitions, and b) there are other reasons than
being runtime suspended which can cause intel_runtime_pm_get_if_in_use to
fail.
Therefore when estimating RC6 the code needs to assert we are indeed in
suspended state, and if not, the best we can do is return the last known
RC6 value.
Without this check we can calculate the estimated value based on un-
initialized or inappropriate internal state, which can result in over-
estimation, or in any case incorrect value being returned.
v2:
* Re-arrange the code a bit to avoid second unlock and return branch.
(Chris Wilson)
v3:
* Insert some strategic blank lines and improve commit msg.
(Chris Wilson)
Xidong Wang [Wed, 4 Apr 2018 09:38:24 +0000 (10:38 +0100)]
drm/i915: Do no use kfree() to free a kmem_cache_alloc() return value
Along the eb_lookup_vmas() error path, the return value from
kmem_cache_alloc() was freed using kfree(). Fix it to use the proper
kmem_cache_free() instead.
Michael Ellerman [Tue, 10 Apr 2018 06:20:53 +0000 (16:20 +1000)]
selftests/filesystems: Don't run dnotify_test by default
In commit ce290a19609d ("selftests: add devpts selftests"), the
filesystems directory was added to the top-level selftests Makefile.
That had the effect of causing the existing dnotify_test in the
filesystems directory to now be run as part of the default selftests
test-run. Unfortunately dnotify_test is actually an infinite loop.
Fix it by moving dnotify_test to TEST_GEN_PROGS_EXTENDED, which says
that it's a generated file (ie. built) but should not be run as part
of the default test suite run (it's an "extended" test).
While we're here cleanup a few other things, devpts_pts should be in
TEST_GEN_PROGS to indicate that it's built, and with the above two
changes we no longer need a custom all or clean rule.
cifs: smb2ops: Fix NULL check in smb2_query_symlink
The current code null checks variable err_buf, which is always null
when it is checked, hence utf16_path is free'd and the function
returns -ENOENT everytime it is called, making it impossible for the
execution path to reach the following code:
err_buf = err_iov.iov_base;
Fix this by null checking err_iov.iov_base instead of err_buf. Also,
notice that err_buf no longer needs to be initialized to NULL.
Addresses-Coverity-ID: 1467876 ("Logically dead code") Fixes: 2d636199e400 ("cifs: Change SMB2_open to return an iov for the error parameter") Signed-off-by: Gustavo A. R. Silva <[email protected]> Signed-off-by: Steve French <[email protected]> Reviewed-by: Pavel Shilovsky <[email protected]>
Eric Biggers [Tue, 17 Apr 2018 19:07:06 +0000 (12:07 -0700)]
KEYS: DNS: limit the length of option strings
Adding a dns_resolver key whose payload contains a very long option name
resulted in that string being printed in full. This hit the WARN_ONCE()
in set_precision() during the printk(), because printk() only supports a
precision of up to 32767 bytes:
precision 1000000 too large
WARNING: CPU: 0 PID: 752 at lib/vsprintf.c:2189 vsnprintf+0x4bc/0x5b0
Fix it by limiting option strings (combined name + value) to a much more
reasonable 128 bytes. The exact limit is arbitrary, but currently the
only recognized option is formatted as "dnserror=%lu" which fits well
within this limit.
Also ratelimit the printks.
Reproducer:
perl -e 'print "#", "A" x 1000000, "\x00"' | keyctl padd dns_resolver desc @s
This bug was found using syzkaller.
Reported-by: Mark Rutland <[email protected]> Fixes: 4a2d789267e0 ("DNS: If the DNS server returns an error, allow that to be cached [ver #2]") Signed-off-by: Eric Biggers <[email protected]> Signed-off-by: David S. Miller <[email protected]>
For some firmware variants - specifically 'capture packed stream' - RSS
filters are not valid. We must check if RSS is actually active rather
than merely enabled.
Fixes: 42356d9a137b ("sfc: support RSS spreading of ethtool ntuple filters") Signed-off-by: Bert Kenward <[email protected]> Signed-off-by: David S. Miller <[email protected]>
vlan: Fix reading memory beyond skb->tail in skb_vlan_tagged_multi
Syzkaller spotted an old bug which leads to reading skb beyond tail by 4
bytes on vlan tagged packets.
This is caused because skb_vlan_tagged_multi() did not check
skb_headlen.
Fixes: 58e998c6d239 ("offloading: Force software GSO for multiple vlan tags.") Reported-and-tested-by: [email protected] Signed-off-by: Toshiaki Makita <[email protected]> Signed-off-by: David S. Miller <[email protected]>
Baolin Wang [Fri, 13 Apr 2018 05:27:58 +0000 (13:27 +0800)]
timekeeping: Remove __current_kernel_time()
The __current_kernel_time() function based on 'struct timespec' is no
longer recommended for new code, and the only user of this function has
been replaced by commit 6909e29fdefb ("kdb: use __ktime_get_real_seconds
instead of __current_kernel_time").
Matt Redfearn [Tue, 17 Apr 2018 14:52:21 +0000 (15:52 +0100)]
MIPS: memset.S: Fix return of __clear_user from Lpartial_fixup
The __clear_user function is defined to return the number of bytes that
could not be cleared. From the underlying memset / bzero implementation
this means setting register a2 to that number on return. Currently if a
page fault is triggered within the memset_partial block, the value
loaded into a2 on return is meaningless.
The label .Lpartial_fixup\@ is jumped to on page fault. In order to work
out how many bytes failed to copy, the exception handler should find how
many bytes left in the partial block (andi a2, STORMASK), add that to
the partial block end address (a2), and subtract the faulting address to
get the remainder. Currently it incorrectly subtracts the partial block
start address (t1), which has additionally been clobbered to generate a
jump target in memset_partial. Fix this by adding the block end address
instead.
This issue was found with the following test code:
int j, k;
for (j = 0; j < 512; j++) {
if ((k = clear_user(NULL, j)) != j) {
pr_err("clear_user (NULL %d) returned %d\n", j, k);
}
}
Which now passes on Creator Ci40 (MIPS32) and Cavium Octeon II (MIPS64).
Mark Rutland [Mon, 16 Apr 2018 13:44:41 +0000 (14:44 +0100)]
arm64: kasan: avoid pfn_to_nid() before page array is initialized
In arm64's kasan_init(), we use pfn_to_nid() to find the NUMA node a
span of memory is in, hoping to allocate shadow from the same NUMA node.
However, at this point, the page array has not been initialized, and
thus this is bogus.
Let's fix this by using early_pfn_to_nid(), as other architectures do in
their kasan init code. Note that early_pfn_to_nid acquires the nid from
the memblock array, which we iterate over in kasan_init(), so this
should be fine.
Stefan Hajnoczi [Tue, 17 Apr 2018 06:25:58 +0000 (14:25 +0800)]
VSOCK: make af_vsock.ko removable again
Commit c1eef220c1760762753b602c382127bfccee226d ("vsock: always call
vsock_init_tables()") introduced a module_init() function without a
corresponding module_exit() function.
Modules with an init function can only be removed if they also have an
exit function. Therefore the vsock module was considered "permanent"
and could not be removed.
This patch adds an empty module_exit() function so that "rmmod vsock"
works. No explicit cleanup is required because:
1. Transports call vsock_core_exit() upon exit and cannot be removed
while sockets are still alive.
2. vsock_diag.ko does not perform any action that requires cleanup by
vsock.ko.
x86/mm: Prevent kernel Oops in PTDUMP code with HIGHPTE=y
The walk_pte_level() function just uses __va to get the virtual address of
the PTE page, but that breaks when the PTE page is not in the direct
mapping with HIGHPTE=y.
The result is an unhandled kernel paging request at some random address
when accessing the current_kernel or current_user file.
x86,sched: Allow topologies where NUMA nodes share an LLC
Intel's Skylake Server CPUs have a different LLC topology than previous
generations. When in Sub-NUMA-Clustering (SNC) mode, the package is divided
into two "slices", each containing half the cores, half the LLC, and one
memory controller and each slice is enumerated to Linux as a NUMA
node. This is similar to how the cores and LLC were arranged for the
Cluster-On-Die (CoD) feature.
CoD allowed the same cache line to be present in each half of the LLC.
But, with SNC, each line is only ever present in *one* slice. This means
that the portion of the LLC *available* to a CPU depends on the data being
accessed:
Remote socket: entire package LLC is shared
Local socket->local slice: data goes into local slice LLC
Local socket->remote slice: data goes into remote-slice LLC. Slightly
higher latency than local slice LLC.
The biggest implication from this is that a process accessing all
NUMA-local memory only sees half the LLC capacity.
The CPU describes its cache hierarchy with the CPUID instruction. One of
the CPUID leaves enumerates the "logical processors sharing this
cache". This information is used for scheduling decisions so that tasks
move more freely between CPUs sharing the cache.
But, the CPUID for the SNC configuration discussed above enumerates the LLC
as being shared by the entire package. This is not 100% precise because the
entire cache is not usable by all accesses. But, it *is* the way the
hardware enumerates itself, and this is not likely to change.
The userspace visible impact of all the above is that the sysfs info
reports the entire LLC as being available to the entire package. As noted
above, this is not true for local socket accesses. This patch does not
correct the sysfs info. It is the same, pre and post patch.
The current code emits the following warning:
sched: CPU #3's llc-sibling CPU #0 is not on the same node! [node: 1 != 0]. Ignoring dependency.
The warning is coming from the topology_sane() check in smpboot.c because
the topology is not matching the expectations of the model for obvious
reasons.
To fix this, add a vendor and model specific check to never call
topology_sane() for these systems. Also, just like "Cluster-on-Die" disable
the "coregroup" sched_domain_topology_level and use NUMA information from
the SRAT alone.
This is OK at least on the hardware we are immediately concerned about
because the LLC sharing happens at both the slice and at the package level,
which are also NUMA boundaries.
Jiri Olsa [Sun, 15 Apr 2018 09:23:51 +0000 (11:23 +0200)]
perf: Fix sample_max_stack maximum check
The syzbot hit KASAN bug in perf_callchain_store having the entry stored
behind the allocated bounds [1].
We miss the sample_max_stack check for the initial event that allocates
callchain buffers. This missing check allows to create an event with
sample_max_stack value bigger than the global sysctl maximum:
# sysctl -a | grep perf_event_max_stack
kernel.perf_event_max_stack = 127
Note the '-C 1', which forces perf record to create just single event.
Otherwise it opens event for every cpu, then the sample_max_stack check
fails on the second event and all's fine.
The fix is to run the sample_max_stack check also for the first event
with callchains.
Thomas Richter [Mon, 16 Apr 2018 13:23:14 +0000 (15:23 +0200)]
perf list: Add s390 support for detailed/verbose PMU event description
'perf list' with flags -d and -v print a description (-d) or a very
verbose explanation (-v) of CPU specific counter events. These
descriptions are provided with the json files in directory
pmu-events/arch/s390/*.json.
Display of these descriptions on s390 requires the corresponding json
files.
On s390 this does not work because function is_pmu_core() does not
detect the s390 directory name where the CPU specific events are listed.
On x86 it is:
perf script: Extend misc field decoding with switch out event type
Append 'p' sign to 'S' tag designating the type of context switch out event so
'Sp' means preemption context switch. Documentation is extended to cover
new presentation changes.
perf report: Extend raw dump (-D) out with switch out event type
Print additional 'preempt' tag for PERF_RECORD_SWITCH[_CPU_WIDE] OUT records when
event header misc field contains PERF_RECORD_MISC_SWITCH_OUT_PREEMPT bit set
designating preemption context switch out event:
perf/core: Store context switch out type in PERF_RECORD_SWITCH[_CPU_WIDE]
Store preempting context switch out event into Perf trace as a part of
PERF_RECORD_SWITCH[_CPU_WIDE] record.
Percentage of preempting and non-preempting context switches help
understanding the nature of workloads (CPU or IO bound) that are running
on a machine;
The event is treated as preemption one when task->state value of the
thread being switched out is TASK_RUNNING. Event type encoding is
implemented using PERF_RECORD_MISC_SWITCH_OUT_PREEMPT bit;
tools/headers: Synchronize kernel ABI headers, v4.17-rc1
Sync the following tooling headers with the latest kernel version:
tools/arch/arm/include/uapi/asm/kvm.h
- New ABI: KVM_REG_ARM_*
tools/arch/x86/include/asm/required-features.h
- Removal of NEED_LA57 dependency
tools/arch/x86/include/uapi/asm/kvm.h
- New KVM ABI: KVM_SYNC_X86_*
tools/include/uapi/asm-generic/mman-common.h
- New ABI: MAP_FIXED_NOREPLACE flag
tools/include/uapi/linux/bpf.h
- New ABI: BPF_F_SEQ_NUMBER functions
tools/include/uapi/linux/if_link.h
- New ABI: IFLA tun and rmnet support
tools/include/uapi/linux/kvm.h
- New ABI: hyperv eventfd and CONN_ID_MASK support plus header cleanups
tools/include/uapi/sound/asound.h
- New ABI: SNDRV_PCM_FORMAT_FIRST PCM format specifier
tools/perf/arch/x86/entry/syscalls/syscall_64.tbl
- The x86 system call table description changed due to the ptregs changes and the renames, in:
d5a00528b58c: syscalls/core, syscalls/x86: Rename struct pt_regs-based sys_*() to __x64_sys_*() 5ac9efa3c50d: syscalls/core, syscalls/x86: Clean up compat syscall stub naming convention ebeb8c82ffaf: syscalls/x86: Use 'struct pt_regs' based syscall calling for IA32_EMULATION and x32
Also fix the x86 syscall table warning:
-Warning: Kernel ABI header at 'tools/arch/x86/entry/syscalls/syscall_64.tbl' differs from latest version at 'arch/x86/entry/syscalls/syscall_64.tbl'
+Warning: Kernel ABI header at 'tools/perf/arch/x86/entry/syscalls/syscall_64.tbl' differs from latest version at 'arch/x86/entry/syscalls/syscall_64.tbl'
None of these changes impact existing tooling code, so we only have to copy the kernel version.
xen/sndif: Sync up with the canonical definition in Xen
This is the sync up with the canonical definition of the sound
protocol in Xen:
1. Protocol version was referenced in the protocol description,
but missed its definition. Fixed by adding a constant
for current protocol version.
2. Some of the request descriptions have "reserved" fields
missed: fixed by adding corresponding entries.
3. Extend the size of the requests and responses to 64 octets.
Bump protocol version to 2.
4. Add explicit back and front synchronization
In order to provide explicit synchronization between backend and
frontend the following changes are introduced in the protocol:
- add new ring buffer for sending asynchronous events from
backend to frontend to report number of bytes played by the
frontend (XENSND_EVT_CUR_POS)
- introduce trigger events for playback control: start/stop/pause/resume
- add "req-" prefix to event-channel and ring-ref to unify naming
of the Xen event channels for requests and events
5. Add explicit back and front parameter negotiation
In order to provide explicit stream parameter negotiation between
backend and frontend the following changes are introduced in the protocol:
add XENSND_OP_HW_PARAM_QUERY request to read/update
configuration space for the parameters given: request passes
desired parameter's intervals/masks and the response to this request
returns allowed min/max intervals/masks to be used.
Petr Mladek [Mon, 16 Apr 2018 11:36:47 +0000 (13:36 +0200)]
livepatch: Allow to call a custom callback when freeing shadow variables
We might need to do some actions before the shadow variable is freed.
For example, we might need to remove it from a list or free some data
that it points to.
This is already possible now. The user can get the shadow variable
by klp_shadow_get(), do the necessary actions, and then call
klp_shadow_free().
This patch allows to do it a more elegant way. The user could implement
the needed actions in a callback that is passed to klp_shadow_free()
as a parameter. The callback usually does reverse operations to
the constructor callback that can be called by klp_shadow_*alloc().
It is especially useful for klp_shadow_free_all(). There we need to do
these extra actions for each found shadow variable with the given ID.
Note that the memory used by the shadow variable itself is still released
later by rcu callback. It is needed to protect internal structures that
keep all shadow variables. But the destructor is called immediately.
The shadow variable must not be access anyway after klp_shadow_free()
is called. The user is responsible to protect this any suitable way.
Be aware that the destructor is called under klp_shadow_lock. It is
the same as for the contructor in klp_shadow_alloc().
Petr Mladek [Mon, 16 Apr 2018 11:36:46 +0000 (13:36 +0200)]
livepatch: Initialize shadow variables safely by a custom callback
The existing API allows to pass a sample data to initialize the shadow
data. It works well when the data are position independent. But it fails
miserably when we need to set a pointer to the shadow structure itself.
Unfortunately, we might need to initialize the pointer surprisingly
often because of struct list_head. It is even worse because the list
might be hidden in other common structures, for example, struct mutex,
struct wait_queue_head.
For example, this was needed to fix races in ALSA sequencer. It required
to add mutex into struct snd_seq_client. See commit b3defb791b26ea06
("ALSA: seq: Make ioctls race-free") and commit d15d662e89fc667b9
("ALSA: seq: Fix racy pool initializations")
This patch makes the API more safe. A custom constructor function and data
are passed to klp_shadow_*alloc() functions instead of the sample data.
Note that ctor_data are no longer a template for shadow->data. It might
point to any data that might be necessary when the constructor is called.
Also note that the constructor is called under klp_shadow_lock. It is
an internal spin_lock that synchronizes alloc() vs. get() operations,
see klp_shadow_get_or_alloc(). On one hand, this adds a risk of ABBA
deadlocks. On the other hand, it allows to do some operations safely.
For example, we could add the new structure into an existing list.
This must be done only once when the structure is allocated.
x86/acpi: Prevent X2APIC id 0xffffffff from being accounted
RongQing reported that there are some X2APIC id 0xffffffff in his machine's
ACPI MADT table, which makes the number of possible CPU inaccurate.
The reason is that the ACPI X2APIC parser has no sanity check for APIC ID
0xffffffff, which is an invalid id in all APIC types. See "Intel® 64
Architecture x2APIC Specification", Chapter 2.4.1.
Add a sanity check to acpi_parse_x2apic() which ignores the invalid id.
and then divides the normalized delta of the corresponding TSC timestamps
by the result to calulate the TSC frequency.
tscfreq = ((tstsc1 - tstsc2 ) * 1e6) / hpetref
This uses do_div() which takes an u32 as the divisor, which worked so far
because the HPET frequency was low enough that 'hpetref' never exceeded
32bit.
On Skylake machines the HPET frequency increased so 'hpetref' can exceed
32bit. do_div() truncates the divisor, which causes the calibration to
fail.
Use div64_u64() to avoid the problem.
[ tglx: Fixes whitespace mangled patch and rewrote changelog ]
powerpc/64s: Default l1d_size to 64K in RFI fallback flush
If there is no d-cache-size property in the device tree, l1d_size could
be zero. We don't actually expect that to happen, it's only been seen
on mambo (simulator) in some configurations.
A zero-size l1d_size leads to the loop in the asm wrapping around to
2^64-1, and then walking off the end of the fallback area and
eventually causing a page fault which is fatal.
Just default to 64K which is correct on some CPUs, and sane enough to
not cause a crash on others.
Fixes: aa8a5e0062ac9 ('powerpc/64s: Add support for RFI flush of L1-D cache') Signed-off-by: Madhavan Srinivasan <[email protected]>
[mpe: Rewrite comment and change log] Signed-off-by: Michael Ellerman <[email protected]>
The struct sigaction for user space in arch/s390/include/uapi/asm/signal.h
is ill defined. The kernel uses two structures 'struct sigaction' and
'struct old_sigaction', the correlation in the kernel for both 31 and
64 bit is as follows
The correlation of the (single) uapi definition for 'struct sigaction'
under '#ifndef __KERNEL__':
31-bit: sys_sigaction -> uapi struct sigaction
31-bit: sys_rt_sigaction -> no structure available
64-bit: sys_sigaction -> no structure available
64-bit: sys_rt_sigaction -> uapi struct sigaction
This is quite confusing. To make it a bit less confusing make the
uapi definition of 'struct sigaction' usable for sys_rt_sigaction for
both 31-bit and 64-bit.
Daniel Stone [Fri, 30 Mar 2018 14:11:30 +0000 (15:11 +0100)]
drm/exynos: Move GEM BOs to drm_framebuffer
Since drm_framebuffer can now store GEM objects directly, place them
there rather than in our own subclass. As this makes the framebuffer
create_handle and destroy functions the same as the GEM framebuffer
helper, we can reuse those.
Song Liu [Fri, 13 Apr 2018 18:55:13 +0000 (11:55 -0700)]
trace_kprobe: Remove warning message "Could not insert probe at..."
This warning message is not very helpful, as the return value should
already show information about the error. Also, this message will
spam dmesg if the user space does testing in a loop, like:
for x in {0..5}
do
echo p:xx xx+$x >> /sys/kernel/debug/tracing/kprobe_events
done
Randy Dunlap [Mon, 16 Apr 2018 19:32:55 +0000 (12:32 -0700)]
textsearch: fix kernel-doc warnings and add kernel-api section
Make lib/textsearch.c usable as kernel-doc.
Add textsearch() function family to kernel-api documentation.
Fix kernel-doc warnings in <linux/textsearch.h>:
../include/linux/textsearch.h:65: warning: Incorrect use of kernel-doc format:
* get_next_block - fetch next block of data
../include/linux/textsearch.h:82: warning: Incorrect use of kernel-doc format:
* finish - finalize/clean a series of get_next_block() calls
Matt Redfearn [Thu, 29 Mar 2018 09:28:23 +0000 (10:28 +0100)]
MIPS: memset.S: EVA & fault support for small_memset
The MIPS kernel memset / bzero implementation includes a small_memset
branch which is used when the region to be set is smaller than a long (4
bytes on 32bit, 8 bytes on 64bit). The current small_memset
implementation uses a simple store byte loop to write the destination.
There are 2 issues with this implementation:
1. When EVA mode is active, user and kernel address spaces may overlap.
Currently the use of the sb instruction means kernel mode addressing is
always used and an intended write to userspace may actually overwrite
some critical kernel data.
2. If the write triggers a page fault, for example by calling
__clear_user(NULL, 2), instead of gracefully handling the fault, an OOPS
is triggered.
Fix these issues by replacing the sb instruction with the EX() macro,
which will emit EVA compatible instuctions as required. Additionally
implement a fault fixup for small_memset which sets a2 to the number of
bytes that could not be cleared (as defined by __clear_user).
Merge branch 'timers-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull missed timer updates from Thomas Gleixner:
"This is a branch which got forgotten during the merge window, but it
contains only fixes and hardware enablement. No fundamental changes.
- Various fixes for the imx-tpm clocksource driver
- A new timer driver for the NCPM7xx SoC family"
* 'timers-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
clocksource/drivers/imx-tpm: Add different counter width support
clocksource/drivers/imx-tpm: Correct some registers operation flow
clocksource/drivers/imx-tpm: Fix typo of clock name
dt-bindings: timer: tpm: fix typo of clock name
clocksource/drivers/npcm: Add NPCM7xx timer driver
dt-binding: timer: document NPCM7xx timer DT bindings
Tyler Hicks [Wed, 28 Mar 2018 23:41:52 +0000 (23:41 +0000)]
eCryptfs: don't pass up plaintext names when using filename encryption
Both ecryptfs_filldir() and ecryptfs_readlink_lower() use
ecryptfs_decode_and_decrypt_filename() to translate lower filenames to
upper filenames. The function correctly passes up lower filenames,
unchanged, when filename encryption isn't in use. However, it was also
passing up lower filenames when the filename wasn't encrypted or
when decryption failed. Since 88ae4ab9802e, eCryptfs refuses to lookup
lower plaintext names when filename encryption is enabled so this
resulted in a situation where userspace would see lower plaintext
filenames in calls to getdents(2) but then not be able to lookup those
filenames.
An example of this can be seen when enabling filename encryption on an
eCryptfs mount at the root directory of an Ext4 filesystem:
$ ls -1i /lower
12 ECRYPTFS_FNEK_ENCRYPTED.FWYZD8TcW.5FV-TKTEYOHsheiHX9a-w.NURCCYIMjI8pn5BDB9-h3fXwrE--
11 lost+found
$ ls -1i /upper
ls: cannot access '/upper/lost+found': No such file or directory
? lost+found
12 test
With this change, the lower lost+found dentry is ignored:
$ ls -1i /lower
12 ECRYPTFS_FNEK_ENCRYPTED.FWYZD8TcW.5FV-TKTEYOHsheiHX9a-w.NURCCYIMjI8pn5BDB9-h3fXwrE--
11 lost+found
$ ls -1i /upper
12 test
Additionally, some potentially noisy error/info messages in the related
code paths are turned into debug messages so that the logs can't be
easily filled.
Fixes: 88ae4ab9802e ("ecryptfs_lookup(): try either only encrypted or plaintext name") Reported-by: Guenter Roeck <[email protected]> Cc: Al Viro <[email protected]> Signed-off-by: Tyler Hicks <[email protected]>
Merge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Pull kvm fixes from Paolo Bonzini:
"Bug fixes, plus a new test case and the associated infrastructure for
writing nested virtualization tests"
* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
kvm: selftests: add vmx_tsc_adjust_test
kvm: x86: move MSR_IA32_TSC handling to x86.c
X86/KVM: Properly update 'tsc_offset' to represent the running guest
kvm: selftests: add -std=gnu99 cflags
x86: Add check for APIC access address for vmentry of L2 guests
KVM: X86: fix incorrect reference of trace_kvm_pi_irte_update
X86/KVM: Do not allow DISABLE_EXITS_MWAIT when LAPIC ARAT is not available
kvm: selftests: fix spelling mistake: "divisable" and "divisible"
X86/VMX: Disable VMX preemption timer if MWAIT is not intercepted
It would allocate memory in this function when the cork->opt is NULL. But
the memory isn't freed if failed in the latter rt check, and return error
directly. It causes the memleak if its caller is ip_make_skb which also
doesn't free the cork->opt when meet a error.
Paolo Bonzini [Tue, 27 Mar 2018 20:46:11 +0000 (22:46 +0200)]
kvm: selftests: add vmx_tsc_adjust_test
The test checks the behavior of setting MSR_IA32_TSC in a nested guest,
and the TSC_OFFSET VMCS field in general. It also introduces the testing
infrastructure for Intel nested virtualization.
KarimAllah Ahmed [Sat, 14 Apr 2018 03:10:52 +0000 (05:10 +0200)]
X86/KVM: Properly update 'tsc_offset' to represent the running guest
Update 'tsc_offset' on vmentry/vmexit of L2 guests to ensure that it always
captures the TSC_OFFSET of the running guest whether it is the L1 or L2
guest.
Local variable description: ----req_u@packet_setsockopt
Variable was created at:
packet_setsockopt+0x13f/0x5a90 net/packet/af_packet.c:3612
SYSC_setsockopt+0x4b8/0x570 net/socket.c:1849
Fixes: f6fb8f100b80 ("af-packet: TPACKET_V3 flexible buffer implementation.") Signed-off-by: Eric Dumazet <[email protected]> Reported-by: syzbot <[email protected]> Signed-off-by: David S. Miller <[email protected]>
Thomas Falcon [Sun, 15 Apr 2018 23:53:36 +0000 (18:53 -0500)]
ibmvnic: Clear pending interrupt after device reset
Due to a firmware bug, the hypervisor can send an interrupt to a
transmit or receive queue just prior to a partition migration, not
allowing the device enough time to handle it and send an EOI. When
the partition migrates, the interrupt is lost but an "EOI-pending"
flag for the interrupt line is still set in firmware. No further
interrupts will be sent until that flag is cleared, effectively
freezing that queue. To workaround this, the driver will disable the
hardware interrupt and send an H_EOI signal prior to re-enabling it.
This will flush the pending EOI and allow the driver to continue
operation.
tcp: clear tp->packets_out when purging write queue
Clear tp->packets_out when purging the write queue, otherwise
tcp_rearm_rto() mistakenly assumes TCP write queue is not empty.
This results in NULL pointer dereference.
Also, remove the redundant `tp->packets_out = 0` from
tcp_disconnect(), since tcp_disconnect() calls
tcp_write_queue_purge().
Dan Williams [Mon, 9 Apr 2018 23:38:01 +0000 (16:38 -0700)]
tools/testing/nvdimm: support nfit_test_dimm attributes under nfit_test.1
The nfit_test.1 bus provides a pmem topology without blk-aperture
enabling, so it presents different failure modes for label space
handling. Allow custom DSM command error injection.
Dan Williams [Mon, 9 Apr 2018 19:34:24 +0000 (12:34 -0700)]
libnvdimm, dimm: handle EACCES failures from label reads
The new support for the standard _LSR and _LSW methods neglected to also
update the nvdimm_init_config_data() and nvdimm_set_config_data() to
return the translated error code from failed commands. This precision is
necessary because the locked status that was previously returned on
ND_CMD_GET_CONFIG_SIZE commands is now returned on
ND_CMD_{GET,SET}_CONFIG_DATA commands.
If the kernel misses this indication it can inadvertently fall back to
label-less mode when it should otherwise avoid all access to locked
regions.
Cc: <[email protected]> Fixes: 4b27db7e26cd ("acpi, nfit: add support for the _LSI, _LSR, and...") Signed-off-by: Dan Williams <[email protected]>
Paolo Abeni [Fri, 13 Apr 2018 11:59:25 +0000 (13:59 +0200)]
team: avoid adding twice the same option to the event list
When parsing the options provided by the user space,
team_nl_cmd_options_set() insert them in a temporary list to send
multiple events with a single message.
While each option's attribute is correctly validated, the code does
not check for duplicate entries before inserting into the event
list.
Exploiting the above, the syzbot was able to trigger the following
splat:
This changeset addresses the avoiding list_add() if the current
option is already present in the event list.
Reported-and-tested-by: [email protected] Signed-off-by: Paolo Abeni <[email protected]> Fixes: 2fcdb2c9e659 ("team: allow to send multiple set events in one message") Signed-off-by: David S. Miller <[email protected]>
Michael Ellerman [Mon, 16 Apr 2018 13:25:19 +0000 (23:25 +1000)]
powerpc/lib: Fix off-by-one in alternate feature patching
When we patch an alternate feature section, we have to adjust any
relative branches that branch out of the alternate section.
But currently we have a bug if we have a branch that points to past
the last instruction of the alternate section, eg:
FTR_SECTION_ELSE
1: b 2f
or 6,6,6
2:
ALT_FTR_SECTION_END(...)
nop
This will result in a relative branch at 1 with a target that equals
the end of the alternate section.
That branch does not need adjusting when it's moved to the non-else
location. Currently we do adjust it, resulting in a branch that goes
off into the link-time location of the else section, which is junk.
The fix is to not patch branches that have a target == end of the
alternate section.
Fixes: d20fe50a7b3c ("KVM: PPC: Book3S HV: Branch inside feature section") Fixes: 9b1a735de64c ("powerpc: Add logic to patch alternative feature sections") Cc: [email protected] # v2.6.27+ Signed-off-by: Michael Ellerman <[email protected]>
Jia-Ju Bai [Wed, 11 Apr 2018 01:15:31 +0000 (09:15 +0800)]
xen: xen-pciback: Replace GFP_ATOMIC with GFP_KERNEL in pcistub_reg_add
pcistub_reg_add() is never called in atomic context.
pcistub_reg_add() is only called by pcistub_quirk_add, which is
only set in DRIVER_ATTR().
Despite never getting called from atomic context,
pcistub_reg_add() calls kzalloc() with GFP_ATOMIC,
which does not sleep for allocation.
GFP_ATOMIC is not necessary and can be replaced with GFP_KERNEL,
which can sleep and improve the possibility of sucessful allocation.
This is found by a static analysis tool named DCNS written by myself.
And I also manually check it.
Jia-Ju Bai [Mon, 9 Apr 2018 15:04:25 +0000 (23:04 +0800)]
xen: xen-pciback: Replace GFP_ATOMIC with GFP_KERNEL in xen_pcibk_config_quirks_init
xen_pcibk_config_quirks_init() is never called in atomic context.
The call chains ending up at xen_pcibk_config_quirks_init() are:
[1] xen_pcibk_config_quirks_init() <- xen_pcibk_config_init_dev() <-
pcistub_init_device() <- pcistub_seize() <- pcistub_probe()
[2] xen_pcibk_config_quirks_init() <- xen_pcibk_config_init_dev() <-
pcistub_init_device() <- pcistub_init_devices_late() <-
xen_pcibk_init()
pcistub_probe() is only set as ".probe" in struct pci_driver.
xen_pcibk_init() is is only set as a parameter of module_init().
These functions are not called in atomic context.
Despite never getting called from atomic context,
xen_pcibk_config_quirks_init() calls kzalloc() with GFP_ATOMIC,
which does not sleep for allocation.
GFP_ATOMIC is not necessary and can be replaced with GFP_KERNEL,
which can sleep and improve the possibility of sucessful allocation.
Jia-Ju Bai [Mon, 9 Apr 2018 15:04:12 +0000 (23:04 +0800)]
xen: xen-pciback: Replace GFP_ATOMIC with GFP_KERNEL in pcistub_device_alloc
pcistub_device_alloc() is never called in atomic context.
The call chain ending up at pcistub_device_alloc() is:
[1] pcistub_device_alloc() <- pcistub_seize() <- pcistub_probe()
pcistub_probe() is only set as ".probe" in struct pci_driver.
This function is not called in atomic context.
Despite never getting called from atomic context,
pcistub_device_alloc() calls kzalloc() with GFP_ATOMIC,
which does not sleep for allocation.
GFP_ATOMIC is not necessary and can be replaced with GFP_KERNEL,
which can sleep and improve the possibility of sucessful allocation.
Jia-Ju Bai [Mon, 9 Apr 2018 15:03:53 +0000 (23:03 +0800)]
xen: xen-pciback: Replace GFP_ATOMIC with GFP_KERNEL in pcistub_init_device
pcistub_init_device() is never called in atomic context.
The call chain ending up at pcistub_init_device() is:
[1] pcistub_init_device() <- pcistub_seize() <- pcistub_probe()
[2] pcistub_init_device() <- pcistub_init_devices_late() <-
xen_pcibk_init()
pcistub_probe() is only set as ".probe" in struct pci_driver.
xen_pcibk_init() is is only set as a parameter of module_init().
These functions are not called in atomic context.
Despite never getting called from atomic context,
pcistub_init_device() calls kzalloc() with GFP_ATOMIC,
which does not sleep for allocation.
GFP_ATOMIC is not necessary and can be replaced with GFP_KERNEL,
which can sleep and improve the possibility of sucessful allocation.
This is found by a static analysis tool named DCNS written by myself.
And I also manually check it.
Jia-Ju Bai [Mon, 9 Apr 2018 15:03:36 +0000 (23:03 +0800)]
xen: xen-pciback: Replace GFP_ATOMIC with GFP_KERNEL in pcistub_probe
pcistub_probe() is never called in atomic context.
This function is only set as ".probe" in struct pci_driver.
Despite never getting called from atomic context,
pcistub_probe() calls kmalloc() with GFP_ATOMIC,
which does not sleep for allocation.
GFP_ATOMIC is not necessary and can be replaced with GFP_KERNEL,
which can sleep and improve the possibility of sucessful allocation.
This is found by a static analysis tool named DCNS written by myself.
And I also manually check it.
projects / linux.git / log