All the locking related cmpxchg's in the following functions are
replaced with the _acquire variants:
- pv_queued_spin_steal_lock()
- trylock_clear_pending()
This change should help performance on architectures that use LL/SC.
The cmpxchg in pv_kick_node() is replaced with a relaxed version
with explicit memory barrier to make sure that it is fully ordered
in the writing of next->lock and the reading of pn->state whether
the cmpxchg is a success or failure without affecting performance in
non-LL/SC architectures.
On a 2-socket 12-core 96-thread Power8 system with pvqspinlock
explicitly enabled, the performance of a locking microbenchmark
with and without this patch on a 4.13-rc4 kernel with Xinhui's PPC
qspinlock patch were as follows:
# of thread w/o patch with patch % Change
----------- --------- ---------- --------
8 5054.8 Mop/s 5209.4 Mop/s +3.1%
16 3985.0 Mop/s 4015.0 Mop/s +0.8%
32 2378.2 Mop/s 2396.0 Mop/s +0.7%
Suggested-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrea Parri <parri.andrea@gmail.com>
Cc: Boqun Feng <boqun.feng@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Pan Xinhui <xinhui@linux.vnet.ibm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Will Deacon <will.deacon@arm.com>
Link: http://lkml.kernel.org/r/1502741222-24360-1-git-send-email-longman@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
struct call_single_data is used in IPIs to transfer information between
CPUs. Its size is bigger than sizeof(unsigned long) and less than
cache line size. Currently it is not allocated with any explicit alignment
requirements. This makes it possible for allocated call_single_data to
cross two cache lines, which results in double the number of the cache lines
that need to be transferred among CPUs.
This can be fixed by requiring call_single_data to be aligned with the
size of call_single_data. Currently the size of call_single_data is the
power of 2. If we add new fields to call_single_data, we may need to
add padding to make sure the size of new definition is the power of 2
as well.
Fortunately, this is enforced by GCC, which will report bad sizes.
To set alignment requirements of call_single_data to the size of
call_single_data, a struct definition and a typedef is used.
To test the effect of the patch, I used the vm-scalability multiple
thread swap test case (swap-w-seq-mt). The test will create multiple
threads and each thread will eat memory until all RAM and part of swap
is used, so that huge number of IPIs are triggered when unmapping
memory. In the test, the throughput of memory writing improves ~5%
compared with misaligned call_single_data, because of faster IPIs.
Suggested-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Huang, Ying <ying.huang@intel.com>
[ Add call_single_data_t and align with size of call_single_data. ]
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Aaron Lu <aaron.lu@intel.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Juergen Gross <jgross@suse.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/87bmnqd6lz.fsf@yhuang-mobile.sh.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
The new completion/crossrelease annotations interact unfavourable with
the extant flush_work()/flush_workqueue() annotations.
The problem is that when a single work class does:
wait_for_completion(&C)
and
complete(&C)
in different executions, we'll build dependencies like:
lock_map_acquire(W)
complete_acquire(C)
and
lock_map_acquire(W)
complete_release(C)
which results in the dependency chain: W->C->W, which lockdep thinks
spells deadlock, even though there is no deadlock potential since
works are ran concurrently.
One possibility would be to change the work 'lock' to recursive-read,
but that would mean hitting a lockdep limitation on recursive locks.
Also, unconditinoally switching to recursive-read here would fail to
detect the actual deadlock on single-threaded workqueues, which do
have a problem with this.
For now, forcefully disregard these locks for crossrelease.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Tejun Heo <tj@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: boqun.feng@gmail.com
Cc: byungchul.park@lge.com
Cc: david@fromorbit.com
Cc: johannes@sipsolutions.net
Cc: oleg@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
The flush_work() annotation as introduced by commit:
e159489baa ("workqueue: relax lockdep annotation on flush_work()")
hits on the lockdep problem with recursive read locks.
The situation as described is:
Work W1: Work W2: Task:
ARR(Q) ARR(Q) flush_workqueue(Q)
A(W1) A(W2) A(Q)
flush_work(W2) R(Q)
A(W2)
R(W2)
if (special)
A(Q)
else
ARR(Q)
R(Q)
where: A - acquire, ARR - acquire-read-recursive, R - release.
Where under 'special' conditions we want to trigger a lock recursion
deadlock, but otherwise allow the flush_work(). The allowing is done
by using recursive read locks (ARR), but lockdep is broken for
recursive stuff.
However, there appears to be no need to acquire the lock if we're not
'special', so if we remove the 'else' clause things become much
simpler and no longer need the recursion thing at all.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Tejun Heo <tj@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: boqun.feng@gmail.com
Cc: byungchul.park@lge.com
Cc: david@fromorbit.com
Cc: johannes@sipsolutions.net
Cc: oleg@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Pull tracing fixes from Steven Rostedt:
"Various bug fixes:
- Two small memory leaks in error paths.
- A missed return error code on an error path.
- A fix to check the tracing ring buffer CPU when it doesn't exist
(caused by setting maxcpus on the command line that is less than
the actual number of CPUs, and then onlining them manually).
- A fix to have the reset of boot tracers called by lateinit_sync()
instead of just lateinit(). As some of the tracers register via
lateinit(), and if the clear happens before the tracer is
registered, it will never start even though it was told to via the
kernel command line"
* tag 'trace-v4.13-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
tracing: Fix freeing of filter in create_filter() when set_str is false
tracing: Fix kmemleak in tracing_map_array_free()
ftrace: Check for null ret_stack on profile function graph entry function
ring-buffer: Have ring_buffer_alloc_read_page() return error on offline CPU
tracing: Missing error code in tracer_alloc_buffers()
tracing: Call clear_boot_tracer() at lateinit_sync
Performing the following task with kmemleak enabled:
# cd /sys/kernel/tracing/events/irq/irq_handler_entry/
# echo 'enable_event:kmem:kmalloc:3 if irq >' > trigger
# echo 'enable_event:kmem:kmalloc:3 if irq > 31' > trigger
# echo scan > /sys/kernel/debug/kmemleak
# cat /sys/kernel/debug/kmemleak
unreferenced object 0xffff8800b9290308 (size 32):
comm "bash", pid 1114, jiffies 4294848451 (age 141.139s)
hex dump (first 32 bytes):
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
backtrace:
[<ffffffff81cef5aa>] kmemleak_alloc+0x4a/0xa0
[<ffffffff81357938>] kmem_cache_alloc_trace+0x158/0x290
[<ffffffff81261c09>] create_filter_start.constprop.28+0x99/0x940
[<ffffffff812639c9>] create_filter+0xa9/0x160
[<ffffffff81263bdc>] create_event_filter+0xc/0x10
[<ffffffff812655e5>] set_trigger_filter+0xe5/0x210
[<ffffffff812660c4>] event_enable_trigger_func+0x324/0x490
[<ffffffff812652e2>] event_trigger_write+0x1a2/0x260
[<ffffffff8138cf87>] __vfs_write+0xd7/0x380
[<ffffffff8138f421>] vfs_write+0x101/0x260
[<ffffffff8139187b>] SyS_write+0xab/0x130
[<ffffffff81cfd501>] entry_SYSCALL_64_fastpath+0x1f/0xbe
[<ffffffffffffffff>] 0xffffffffffffffff
The function create_filter() is passed a 'filterp' pointer that gets
allocated, and if "set_str" is true, it is up to the caller to free it, even
on error. The problem is that the pointer is not freed by create_filter()
when set_str is false. This is a bug, and it is not up to the caller to free
the filter on error if it doesn't care about the string.
Link: http://lkml.kernel.org/r/1502705898-27571-2-git-send-email-chuhu@redhat.com
Cc: stable@vger.kernel.org
Fixes: 38b78eb85 ("tracing: Factorize filter creation")
Reported-by: Chunyu Hu <chuhu@redhat.com>
Tested-by: Chunyu Hu <chuhu@redhat.com>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
There's a small race when function graph shutsdown and the calling of the
registered function graph entry callback. The callback must not reference
the task's ret_stack without first checking that it is not NULL. Note, when
a ret_stack is allocated for a task, it stays allocated until the task exits.
The problem here, is that function_graph is shutdown, and a new task was
created, which doesn't have its ret_stack allocated. But since some of the
functions are still being traced, the callbacks can still be called.
The normal function_graph code handles this, but starting with commit
8861dd303c ("ftrace: Access ret_stack->subtime only in the function
profiler") the profiler code references the ret_stack on function entry, but
doesn't check if it is NULL first.
Link: https://bugzilla.kernel.org/show_bug.cgi?id=196611
Cc: stable@vger.kernel.org
Fixes: 8861dd303c ("ftrace: Access ret_stack->subtime only in the function profiler")
Reported-by: lilydjwg@gmail.com
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
This was reported many times, and this was even mentioned in commit
52ee2dfdd4 ("pids: refactor vnr/nr_ns helpers to make them safe") but
somehow nobody bothered to fix the obvious problem: task_tgid_nr_ns() is
not safe because task->group_leader points to nowhere after the exiting
task passes exit_notify(), rcu_read_lock() can not help.
We really need to change __unhash_process() to nullify group_leader,
parent, and real_parent, but this needs some cleanups. Until then we
can turn task_tgid_nr_ns() into another user of __task_pid_nr_ns() and
fix the problem.
Reported-by: Troy Kensinger <tkensinger@google.com>
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pull perf fixes from Thomas Gleixner:
"Two fixes for the perf subsystem:
- Fix an inconsistency of RDPMC mm struct tagging across exec() which
causes RDPMC to fault.
- Correct the timestamp mechanics across IOC_DISABLE/ENABLE which
causes incorrect timestamps and total time calculations"
* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
perf/core: Fix time on IOC_ENABLE
perf/x86: Fix RDPMC vs. mm_struct tracking
Pull irq fixes from Thomas Gleixner:
"A pile of smallish changes all over the place:
- Add a missing ISB in the GIC V1 driver
- Remove an ACPI version check in the GIC V3 ITS driver
- Add the missing irq_pm_shutdown function for BRCMSTB-L2 to avoid
spurious wakeups
- Remove the artifical limitation of ITS instances to the number of
NUMA nodes which prevents utilizing the ITS hardware correctly
- Prevent a infinite parsing loop in the GIC-V3 ITS/MSI code
- Honour the force affinity argument in the GIC-V3 driver which is
required to make perf work correctly
- Correctly report allocation failures in GIC-V2/V3 to avoid using
half allocated and initialized interrupts.
- Fixup checks against nr_cpu_ids in the generic IPI code"
* 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
genirq/ipi: Fixup checks against nr_cpu_ids
genirq: Restore trigger settings in irq_modify_status()
MAINTAINERS: Remove Jason Cooper's irqchip git tree
irqchip/gic-v3-its-platform-msi: Fix msi-parent parsing loop
irqchip/gic-v3-its: Allow GIC ITS number more than MAX_NUMNODES
irqchip: brcmstb-l2: Define an irq_pm_shutdown function
irqchip/gic: Ensure we have an ISB between ack and ->handle_irq
irqchip/gic-v3-its: Remove ACPICA version check for ACPI NUMA
irqchip/gic-v3: Honor forced affinity setting
irqchip/gic-v3: Report failures in gic_irq_domain_alloc
irqchip/gic-v2: Report failures in gic_irq_domain_alloc
irqchip/atmel-aic: Remove root argument from ->fixup() prototype
irqchip/atmel-aic: Fix unbalanced refcount in aic_common_rtc_irq_fixup()
irqchip/atmel-aic: Fix unbalanced of_node_put() in aic_common_irq_fixup()
Pull watchdog fix from Thomas Gleixner:
"A fix for the hardlockup watchdog to prevent false positives with
extreme Turbo-Modes which make the perf/NMI watchdog fire faster than
the hrtimer which is used to verify.
Slightly larger than the minimal fix, which just would increase the
hrtimer frequency, but comes with extra overhead of more watchdog
timer interrupts and thread wakeups for all users.
With this change we restrict the overhead to the extreme Turbo-Mode
systems"
* 'core-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
kernel/watchdog: Prevent false positives with turbo modes
When forcing a signal, SIGNAL_UNKILLABLE is removed to prevent recursive
faults, but this is undesirable when tracing. For example, debugging an
init process (whether global or namespace), hitting a breakpoint and
SIGTRAP will force SIGTRAP and then remove SIGNAL_UNKILLABLE.
Everything continues fine, but then once debugging has finished, the
init process is left killable which is unlikely what the user expects,
resulting in either an accidentally killed init or an init that stops
reaping zombies.
Link: http://lkml.kernel.org/r/20170815112806.10728-1-jamie.iles@oracle.com
Signed-off-by: Jamie Iles <jamie.iles@oracle.com>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Recursive loops with module loading were previously handled in kmod by
restricting the number of modprobe calls to 50 and if that limit was
breached request_module() would return an error and a user would see the
following on their kernel dmesg:
request_module: runaway loop modprobe binfmt-464c
Starting init:/sbin/init exists but couldn't execute it (error -8)
This issue could happen for instance when a 64-bit kernel boots a 32-bit
userspace on some architectures and has no 32-bit binary format
hanlders. This is visible, for instance, when a CONFIG_MODULES enabled
64-bit MIPS kernel boots a into o32 root filesystem and the binfmt
handler for o32 binaries is not built-in.
After commit 6d7964a722 ("kmod: throttle kmod thread limit") we now
don't have any visible signs of an error and the kernel just waits for
the loop to end somehow.
Although this *particular* recursive loop could also be addressed by
doing a sanity check on search_binary_handler() and disallowing a
modular binfmt to be required for modprobe, a generic solution for any
recursive kernel kmod issues is still needed.
This should catch these loops. We can investigate each loop and address
each one separately as they come in, this however puts a stop gap for
them as before.
Link: http://lkml.kernel.org/r/20170809234635.13443-3-mcgrof@kernel.org
Fixes: 6d7964a722 ("kmod: throttle kmod thread limit")
Signed-off-by: Luis R. Rodriguez <mcgrof@kernel.org>
Reported-by: Matt Redfearn <matt.redfearn@imgtec.com>
Tested-by: Matt Redfearn <matt.redfearn@imgetc.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Colin Ian King <colin.king@canonical.com>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Cc: Daniel Mentz <danielmentz@google.com>
Cc: David Binderman <dcb314@hotmail.com>
Cc: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jessica Yu <jeyu@redhat.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Michal Marek <mmarek@suse.com>
Cc: Miroslav Benes <mbenes@suse.cz>
Cc: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Petr Mladek <pmladek@suse.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The hardlockup detector on x86 uses a performance counter based on unhalted
CPU cycles and a periodic hrtimer. The hrtimer period is about 2/5 of the
performance counter period, so the hrtimer should fire 2-3 times before the
performance counter NMI fires. The NMI code checks whether the hrtimer
fired since the last invocation. If not, it assumess a hard lockup.
The calculation of those periods is based on the nominal CPU
frequency. Turbo modes increase the CPU clock frequency and therefore
shorten the period of the perf/NMI watchdog. With extreme Turbo-modes (3x
nominal frequency) the perf/NMI period is shorter than the hrtimer period
which leads to false positives.
A simple fix would be to shorten the hrtimer period, but that comes with
the side effect of more frequent hrtimer and softlockup thread wakeups,
which is not desired.
Implement a low pass filter, which checks the perf/NMI period against
kernel time. If the perf/NMI fires before 4/5 of the watchdog period has
elapsed then the event is ignored and postponed to the next perf/NMI.
That solves the problem and avoids the overhead of shorter hrtimer periods
and more frequent softlockup thread wakeups.
Fixes: 58687acba5 ("lockup_detector: Combine nmi_watchdog and softlockup detector")
Reported-and-tested-by: Kan Liang <Kan.liang@intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: dzickus@redhat.com
Cc: prarit@redhat.com
Cc: ak@linux.intel.com
Cc: babu.moger@oracle.com
Cc: peterz@infradead.org
Cc: eranian@google.com
Cc: acme@redhat.com
Cc: stable@vger.kernel.org
Cc: atomlin@redhat.com
Cc: akpm@linux-foundation.org
Cc: torvalds@linux-foundation.org
Link: http://lkml.kernel.org/r/alpine.DEB.2.20.1708150931310.1886@nanos
irq_modify_status starts by clearing the trigger settings from
irq_data before applying the new settings, but doesn't restore them,
leaving them to IRQ_TYPE_NONE.
That's pretty confusing to the potential request_irq() that could
follow. Instead, snapshot the settings before clearing them, and restore
them if the irq_modify_status() invocation was not changing the trigger.
Fixes: 1e2a7d7849 ("irqdomain: Don't set type when mapping an IRQ")
Reported-and-tested-by: jeffy <jeffy.chen@rock-chips.com>
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Jon Hunter <jonathanh@nvidia.com>
Cc: stable@vger.kernel.org
Link: http://lkml.kernel.org/r/20170818095345.12378-1-marc.zyngier@arm.com
With the new lockdep crossrelease feature, which checks completions usage,
a false positive is reported in the workqueue code:
> Worker A : acquired of wfc.work -> wait for cpu_hotplug_lock to be released
> Task B : acquired of cpu_hotplug_lock -> wait for lock#3 to be released
> Task C : acquired of lock#3 -> wait for completion of barr->done
> (Task C is in lru_add_drain_all_cpuslocked())
> Worker D : wait for wfc.work to be released -> will complete barr->done
Such a dead lock can not happen because Task C's barr->done and Worker D's
barr->done can not be the same instance.
The reason of this false positive is we initialize all wq_barrier::done
at insert_wq_barrier() via init_completion(), which makes them belong to
the same lock class, therefore, impossible circles are reported.
To fix this, explicitly initialize the lockdep map for wq_barrier::done
in insert_wq_barrier(), so that the lock class key of wq_barrier::done
is a subkey of the corresponding work_struct, as a result we won't build
a dependency between a wq_barrier with a unrelated work, and we can
differ wq barriers based on the related works, so the false positive
above is avoided.
Also define the empty lockdep_init_map_crosslock() for !CROSSRELEASE
to make the code simple and away from unnecessary #ifdefs.
Reported-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Boqun Feng <boqun.feng@gmail.com>
Cc: Byungchul Park <byungchul.park@lge.com>
Cc: Lai Jiangshan <jiangshanlai@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Tejun Heo <tj@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20170817094622.12915-1-boqun.feng@gmail.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
This implements refcount_t overflow protection on x86 without a noticeable
performance impact, though without the fuller checking of REFCOUNT_FULL.
This is done by duplicating the existing atomic_t refcount implementation
but with normally a single instruction added to detect if the refcount
has gone negative (e.g. wrapped past INT_MAX or below zero). When detected,
the handler saturates the refcount_t to INT_MIN / 2. With this overflow
protection, the erroneous reference release that would follow a wrap back
to zero is blocked from happening, avoiding the class of refcount-overflow
use-after-free vulnerabilities entirely.
Only the overflow case of refcounting can be perfectly protected, since
it can be detected and stopped before the reference is freed and left to
be abused by an attacker. There isn't a way to block early decrements,
and while REFCOUNT_FULL stops increment-from-zero cases (which would
be the state _after_ an early decrement and stops potential double-free
conditions), this fast implementation does not, since it would require
the more expensive cmpxchg loops. Since the overflow case is much more
common (e.g. missing a "put" during an error path), this protection
provides real-world protection. For example, the two public refcount
overflow use-after-free exploits published in 2016 would have been
rendered unexploitable:
http://perception-point.io/2016/01/14/analysis-and-exploitation-of-a-linux-kernel-vulnerability-cve-2016-0728/http://cyseclabs.com/page?n=02012016
This implementation does, however, notice an unchecked decrement to zero
(i.e. caller used refcount_dec() instead of refcount_dec_and_test() and it
resulted in a zero). Decrements under zero are noticed (since they will
have resulted in a negative value), though this only indicates that a
use-after-free may have already happened. Such notifications are likely
avoidable by an attacker that has already exploited a use-after-free
vulnerability, but it's better to have them reported than allow such
conditions to remain universally silent.
On first overflow detection, the refcount value is reset to INT_MIN / 2
(which serves as a saturation value) and a report and stack trace are
produced. When operations detect only negative value results (such as
changing an already saturated value), saturation still happens but no
notification is performed (since the value was already saturated).
On the matter of races, since the entire range beyond INT_MAX but before
0 is negative, every operation at INT_MIN / 2 will trap, leaving no
overflow-only race condition.
As for performance, this implementation adds a single "js" instruction
to the regular execution flow of a copy of the standard atomic_t refcount
operations. (The non-"and_test" refcount_dec() function, which is uncommon
in regular refcount design patterns, has an additional "jz" instruction
to detect reaching exactly zero.) Since this is a forward jump, it is by
default the non-predicted path, which will be reinforced by dynamic branch
prediction. The result is this protection having virtually no measurable
change in performance over standard atomic_t operations. The error path,
located in .text.unlikely, saves the refcount location and then uses UD0
to fire a refcount exception handler, which resets the refcount, handles
reporting, and returns to regular execution. This keeps the changes to
.text size minimal, avoiding return jumps and open-coded calls to the
error reporting routine.
Example assembly comparison:
refcount_inc() before:
.text:
ffffffff81546149: f0 ff 45 f4 lock incl -0xc(%rbp)
refcount_inc() after:
.text:
ffffffff81546149: f0 ff 45 f4 lock incl -0xc(%rbp)
ffffffff8154614d: 0f 88 80 d5 17 00 js ffffffff816c36d3
...
.text.unlikely:
ffffffff816c36d3: 48 8d 4d f4 lea -0xc(%rbp),%rcx
ffffffff816c36d7: 0f ff (bad)
These are the cycle counts comparing a loop of refcount_inc() from 1
to INT_MAX and back down to 0 (via refcount_dec_and_test()), between
unprotected refcount_t (atomic_t), fully protected REFCOUNT_FULL
(refcount_t-full), and this overflow-protected refcount (refcount_t-fast):
2147483646 refcount_inc()s and 2147483647 refcount_dec_and_test()s:
cycles protections
atomic_t 82249267387 none
refcount_t-fast 82211446892 overflow, untested dec-to-zero
refcount_t-full 144814735193 overflow, untested dec-to-zero, inc-from-zero
This code is a modified version of the x86 PAX_REFCOUNT atomic_t
overflow defense from the last public patch of PaX/grsecurity, based
on my understanding of the code. Changes or omissions from the original
code are mine and don't reflect the original grsecurity/PaX code. Thanks
to PaX Team for various suggestions for improvement for repurposing this
code to be a refcount-only protection.
Signed-off-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: David S. Miller <davem@davemloft.net>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Elena Reshetova <elena.reshetova@intel.com>
Cc: Eric Biggers <ebiggers3@gmail.com>
Cc: Eric W. Biederman <ebiederm@xmission.com>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Hans Liljestrand <ishkamiel@gmail.com>
Cc: James Bottomley <James.Bottomley@hansenpartnership.com>
Cc: Jann Horn <jannh@google.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Manfred Spraul <manfred@colorfullife.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Serge E. Hallyn <serge@hallyn.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: arozansk@redhat.com
Cc: axboe@kernel.dk
Cc: kernel-hardening@lists.openwall.com
Cc: linux-arch <linux-arch@vger.kernel.org>
Link: http://lkml.kernel.org/r/20170815161924.GA133115@beast
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Pull audit fixes from Paul Moore:
"Two small fixes to the audit code, both explained well in the
respective patch descriptions, but the quick summary is one
use-after-free fix, and one silly fanotify notification flag fix"
* tag 'audit-pr-20170816' of git://git.kernel.org/pub/scm/linux/kernel/git/pcmoore/audit:
audit: Receive unmount event
audit: Fix use after free in audit_remove_watch_rule()
Pull networking fixes from David Miller:
1) Fix TCP checksum offload handling in iwlwifi driver, from Emmanuel
Grumbach.
2) In ksz DSA tagging code, free SKB if skb_put_padto() fails. From
Vivien Didelot.
3) Fix two regressions with bonding on wireless, from Andreas Born.
4) Fix build when busypoll is disabled, from Daniel Borkmann.
5) Fix copy_linear_skb() wrt. SO_PEEK_OFF, from Eric Dumazet.
6) Set SKB cached route properly in inet_rtm_getroute(), from Florian
Westphal.
7) Fix PCI-E relaxed ordering handling in cxgb4 driver, from Ding
Tianhong.
8) Fix module refcnt leak in ULP code, from Sabrina Dubroca.
9) Fix use of GFP_KERNEL in atomic contexts in AF_KEY code, from Eric
Dumazet.
10) Need to purge socket write queue in dccp_destroy_sock(), also from
Eric Dumazet.
11) Make bpf_trace_printk() work properly on 32-bit architectures, from
Daniel Borkmann.
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (47 commits)
bpf: fix bpf_trace_printk on 32 bit archs
PCI: fix oops when try to find Root Port for a PCI device
sfc: don't try and read ef10 data on non-ef10 NIC
net_sched: remove warning from qdisc_hash_add
net_sched/sfq: update hierarchical backlog when drop packet
net_sched: reset pointers to tcf blocks in classful qdiscs' destructors
ipv4: fix NULL dereference in free_fib_info_rcu()
net: Fix a typo in comment about sock flags.
ipv6: fix NULL dereference in ip6_route_dev_notify()
tcp: fix possible deadlock in TCP stack vs BPF filter
dccp: purge write queue in dccp_destroy_sock()
udp: fix linear skb reception with PEEK_OFF
ipv6: release rt6->rt6i_idev properly during ifdown
af_key: do not use GFP_KERNEL in atomic contexts
tcp: ulp: avoid module refcnt leak in tcp_set_ulp
net/cxgb4vf: Use new PCI_DEV_FLAGS_NO_RELAXED_ORDERING flag
net/cxgb4: Use new PCI_DEV_FLAGS_NO_RELAXED_ORDERING flag
PCI: Disable Relaxed Ordering Attributes for AMD A1100
PCI: Disable Relaxed Ordering for some Intel processors
PCI: Disable PCIe Relaxed Ordering if unsupported
...
James reported that on MIPS32 bpf_trace_printk() is currently
broken while MIPS64 works fine:
bpf_trace_printk() uses conditional operators to attempt to
pass different types to __trace_printk() depending on the
format operators. This doesn't work as intended on 32-bit
architectures where u32 and long are passed differently to
u64, since the result of C conditional operators follows the
"usual arithmetic conversions" rules, such that the values
passed to __trace_printk() will always be u64 [causing issues
later in the va_list handling for vscnprintf()].
For example the samples/bpf/tracex5 test printed lines like
below on MIPS32, where the fd and buf have come from the u64
fd argument, and the size from the buf argument:
[...] 1180.941542: 0x00000001: write(fd=1, buf= (null), size=6258688)
Instead of this:
[...] 1625.616026: 0x00000001: write(fd=1, buf=009e4000, size=512)
One way to get it working is to expand various combinations
of argument types into 8 different combinations for 32 bit
and 64 bit kernels. Fix tested by James on MIPS32 and MIPS64
as well that it resolves the issue.
Fixes: 9c959c863f ("tracing: Allow BPF programs to call bpf_trace_printk()")
Reported-by: James Hogan <james.hogan@imgtec.com>
Tested-by: James Hogan <james.hogan@imgtec.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Although audit_watch_handle_event() can handle FS_UNMOUNT event, it is
not part of AUDIT_FS_WATCH mask and thus such event never gets to
audit_watch_handle_event(). Thus fsnotify marks are deleted by fsnotify
subsystem on unmount without audit being notified about that which leads
to a strange state of existing audit rules with dead fsnotify marks.
Add FS_UNMOUNT to the mask of events to be received so that audit can
clean up its state accordingly.
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Paul Moore <paul@paul-moore.com>
audit_remove_watch_rule() drops watch's reference to parent but then
continues to work with it. That is not safe as parent can get freed once
we drop our reference. The following is a trivial reproducer:
mount -o loop image /mnt
touch /mnt/file
auditctl -w /mnt/file -p wax
umount /mnt
auditctl -D
<crash in fsnotify_destroy_mark()>
Grab our own reference in audit_remove_watch_rule() earlier to make sure
mark does not get freed under us.
CC: stable@vger.kernel.org
Reported-by: Tony Jones <tonyj@suse.de>
Signed-off-by: Jan Kara <jack@suse.cz>
Tested-by: Tony Jones <tonyj@suse.de>
Signed-off-by: Paul Moore <paul@paul-moore.com>
Conflicts:
include/linux/mm_types.h
mm/huge_memory.c
I removed the smp_mb__before_spinlock() like the following commit does:
8b1b436dd1 ("mm, locking: Rework {set,clear,mm}_tlb_flush_pending()")
and fixed up the affected commits.
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Patch series "fixes of TLB batching races", v6.
It turns out that Linux TLB batching mechanism suffers from various
races. Races that are caused due to batching during reclamation were
recently handled by Mel and this patch-set deals with others. The more
fundamental issue is that concurrent updates of the page-tables allow
for TLB flushes to be batched on one core, while another core changes
the page-tables. This other core may assume a PTE change does not
require a flush based on the updated PTE value, while it is unaware that
TLB flushes are still pending.
This behavior affects KSM (which may result in memory corruption) and
MADV_FREE and MADV_DONTNEED (which may result in incorrect behavior). A
proof-of-concept can easily produce the wrong behavior of MADV_DONTNEED.
Memory corruption in KSM is harder to produce in practice, but was
observed by hacking the kernel and adding a delay before flushing and
replacing the KSM page.
Finally, there is also one memory barrier missing, which may affect
architectures with weak memory model.
This patch (of 7):
Setting and clearing mm->tlb_flush_pending can be performed by multiple
threads, since mmap_sem may only be acquired for read in
task_numa_work(). If this happens, tlb_flush_pending might be cleared
while one of the threads still changes PTEs and batches TLB flushes.
This can lead to the same race between migration and
change_protection_range() that led to the introduction of
tlb_flush_pending. The result of this race was data corruption, which
means that this patch also addresses a theoretically possible data
corruption.
An actual data corruption was not observed, yet the race was was
confirmed by adding assertion to check tlb_flush_pending is not set by
two threads, adding artificial latency in change_protection_range() and
using sysctl to reduce kernel.numa_balancing_scan_delay_ms.
Link: http://lkml.kernel.org/r/20170802000818.4760-2-namit@vmware.com
Fixes: 2084140594 ("mm: fix TLB flush race between migration, and
change_protection_range")
Signed-off-by: Nadav Amit <namit@vmware.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Minchan Kim <minchan@kernel.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jeff Dike <jdike@addtoit.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
As Tetsuo points out:
"Commit 385386cff4 ("mm: vmstat: move slab statistics from zone to
node counters") broke "Slab:" field of /proc/meminfo . It shows nearly
0kB"
In addition to /proc/meminfo, this problem also affects the slab
counters OOM/allocation failure info dumps, can cause early -ENOMEM from
overcommit protection, and miscalculate image size requirements during
suspend-to-disk.
This is because the patch in question switched the slab counters from
the zone level to the node level, but forgot to update the global
accessor functions to read the aggregate node data instead of the
aggregate zone data.
Use global_node_page_state() to access the global slab counters.
Fixes: 385386cff4 ("mm: vmstat: move slab statistics from zone to node counters")
Link: http://lkml.kernel.org/r/20170801134256.5400-1-hannes@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Reported-by: Tetsuo Handa <penguin-kernel@i-love.sakura.ne.jp>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Josef Bacik <josef@toxicpanda.com>
Cc: Vladimir Davydov <vdavydov.dev@gmail.com>
Cc: Stefan Agner <stefan@agner.ch>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
No acquisition might be in progress on commit of a crosslock. Completion
operations enabling crossrelease are the case like:
CONTEXT X CONTEXT Y
--------- ---------
trigger completion context
complete AX
commit AX
wait_for_complete AX
acquire AX
wait
where AX is a crosslock.
When no acquisition is in progress, we should not perform commit because
the lock does not exist, which might cause incorrect memory access. So
we have to track the number of acquisitions of a crosslock to handle it.
Moreover, in case that more than one acquisition of a crosslock are
overlapped like:
CONTEXT W CONTEXT X CONTEXT Y CONTEXT Z
--------- --------- --------- ---------
acquire AX (gen_id: 1)
acquire A
acquire AX (gen_id: 10)
acquire B
commit AX
acquire C
commit AX
where A, B and C are typical locks and AX is a crosslock.
Current crossrelease code performs commits in Y and Z with gen_id = 10.
However, we can use gen_id = 1 to do it, since not only 'acquire AX in X'
but 'acquire AX in W' also depends on each acquisition in Y and Z until
their commits. So make it use gen_id = 1 instead of 10 on their commits,
which adds an additional dependency 'AX -> A' in the example above.
Signed-off-by: Byungchul Park <byungchul.park@lge.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: akpm@linux-foundation.org
Cc: boqun.feng@gmail.com
Cc: kernel-team@lge.com
Cc: kirill@shutemov.name
Cc: npiggin@gmail.com
Cc: walken@google.com
Cc: willy@infradead.org
Link: http://lkml.kernel.org/r/1502089981-21272-8-git-send-email-byungchul.park@lge.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Two boots + a make defconfig, the first didn't have the redundant bit
in, the second did:
lock-classes: 1168 1169 [max: 8191]
direct dependencies: 7688 5812 [max: 32768]
indirect dependencies: 25492 25937
all direct dependencies: 220113 217512
dependency chains: 9005 9008 [max: 65536]
dependency chain hlocks: 34450 34366 [max: 327680]
in-hardirq chains: 55 51
in-softirq chains: 371 378
in-process chains: 8579 8579
stack-trace entries: 108073 88474 [max: 524288]
combined max dependencies: 178738560 169094640
max locking depth: 15 15
max bfs queue depth: 320 329
cyclic checks: 9123 9190
redundant checks: 5046
redundant links: 1828
find-mask forwards checks: 2564 2599
find-mask backwards checks: 39521 39789
So it saves nearly 2k links and a fair chunk of stack-trace entries, but
as expected, makes no real difference on the indirect dependencies.
At the same time, you see the max BFS depth increase, which is also
expected, although it could easily be boot variance -- these numbers are
not entirely stable between boots.
The down side is that the cycles in the graph become larger and thus
the reports harder to read.
XXX: do we want this as a CONFIG variable, implied by LOCKDEP_SMALL?
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Byungchul Park <byungchul.park@lge.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Nikolay Borisov <nborisov@suse.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: akpm@linux-foundation.org
Cc: boqun.feng@gmail.com
Cc: iamjoonsoo.kim@lge.com
Cc: kernel-team@lge.com
Cc: kirill@shutemov.name
Cc: npiggin@gmail.com
Cc: walken@google.com
Link: http://lkml.kernel.org/r/20170303091338.GH6536@twins.programming.kicks-ass.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Since its inception, our understanding of ACQUIRE, esp. as applied to
spinlocks, has changed somewhat. Also, I wonder if, with a simple
change, we cannot make it provide more.
The problem with the comment is that the STORE done by spin_lock isn't
itself ordered by the ACQUIRE, and therefore a later LOAD can pass over
it and cross with any prior STORE, rendering the default WMB
insufficient (pointed out by Alan).
Now, this is only really a problem on PowerPC and ARM64, both of
which already defined smp_mb__before_spinlock() as a smp_mb().
At the same time, we can get a much stronger construct if we place
that same barrier _inside_ the spin_lock(). In that case we upgrade
the RCpc spinlock to an RCsc. That would make all schedule() calls
fully transitive against one another.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Will Deacon <will.deacon@arm.com>
Cc: Alan Stern <stern@rowland.harvard.edu>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Nicholas Piggin <npiggin@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Paul McKenney <paulmck@linux.vnet.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
