Alexander noticed that we use RCU iteration on rb->event_list but do
not use list_{add,del}_rcu() to add,remove entries to that list, nor
do we observe proper grace periods when re-using the entries.
Merge ring_buffer_detach() into ring_buffer_attach() such that
attaching to the NULL buffer is detaching.
Furthermore, ensure that between any 'detach' and 'attach' of the same
event we observe the required grace period, but only when strictly
required. In effect this means that only ioctl(.request =
PERF_EVENT_IOC_SET_OUTPUT) will wait for a grace period, while the
normal initial attach and final detach will not be delayed.
This patch should, I think, do the right thing under all
circumstances, the 'normal' cases all should never see the extra grace
period, but the two cases:
1) PERF_EVENT_IOC_SET_OUTPUT on an event which already has a
ring_buffer set, will now observe the required grace period between
removing itself from the old and attaching itself to the new buffer.
This case is 'simple' in that both buffers are present in
perf_event_set_output() one could think an unconditional
synchronize_rcu() would be sufficient; however...
2) an event that has a buffer attached, the buffer is destroyed
(munmap) and then the event is attached to a new/different buffer
using PERF_EVENT_IOC_SET_OUTPUT.
This case is more complex because the buffer destruction does:
ring_buffer_attach(.rb = NULL)
followed by the ioctl() doing:
ring_buffer_attach(.rb = foo);
and we still need to observe the grace period between these two
calls due to us reusing the event->rb_entry list_head.
In order to make 2 happen we use Paul's latest cond_synchronize_rcu()
call.
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Reported-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20140507123526.GD13658@twins.programming.kicks-ass.net
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
The perf cpu offline callback takes down all cpu context
events and releases swhash->swevent_hlist.
This could race with task context software event being just
scheduled on this cpu via perf_swevent_add while cpu hotplug
code already cleaned up event's data.
The race happens in the gap between the cpu notifier code
and the cpu being actually taken down. Note that only cpu
ctx events are terminated in the perf cpu hotplug code.
It's easily reproduced with:
$ perf record -e faults perf bench sched pipe
while putting one of the cpus offline:
# echo 0 > /sys/devices/system/cpu/cpu1/online
Console emits following warning:
WARNING: CPU: 1 PID: 2845 at kernel/events/core.c:5672 perf_swevent_add+0x18d/0x1a0()
Modules linked in:
CPU: 1 PID: 2845 Comm: sched-pipe Tainted: G W 3.14.0+ #256
Hardware name: Intel Corporation Montevina platform/To be filled by O.E.M., BIOS AMVACRB1.86C.0066.B00.0805070703 05/07/2008
0000000000000009 ffff880077233ab8 ffffffff81665a23 0000000000200005
0000000000000000 ffff880077233af8 ffffffff8104732c 0000000000000046
ffff88007467c800 0000000000000002 ffff88007a9cf2a0 0000000000000001
Call Trace:
[<ffffffff81665a23>] dump_stack+0x4f/0x7c
[<ffffffff8104732c>] warn_slowpath_common+0x8c/0xc0
[<ffffffff8104737a>] warn_slowpath_null+0x1a/0x20
[<ffffffff8110fb3d>] perf_swevent_add+0x18d/0x1a0
[<ffffffff811162ae>] event_sched_in.isra.75+0x9e/0x1f0
[<ffffffff8111646a>] group_sched_in+0x6a/0x1f0
[<ffffffff81083dd5>] ? sched_clock_local+0x25/0xa0
[<ffffffff811167e6>] ctx_sched_in+0x1f6/0x450
[<ffffffff8111757b>] perf_event_sched_in+0x6b/0xa0
[<ffffffff81117a4b>] perf_event_context_sched_in+0x7b/0xc0
[<ffffffff81117ece>] __perf_event_task_sched_in+0x43e/0x460
[<ffffffff81096f1e>] ? put_lock_stats.isra.18+0xe/0x30
[<ffffffff8107b3c8>] finish_task_switch+0xb8/0x100
[<ffffffff8166a7de>] __schedule+0x30e/0xad0
[<ffffffff81172dd2>] ? pipe_read+0x3e2/0x560
[<ffffffff8166b45e>] ? preempt_schedule_irq+0x3e/0x70
[<ffffffff8166b45e>] ? preempt_schedule_irq+0x3e/0x70
[<ffffffff8166b464>] preempt_schedule_irq+0x44/0x70
[<ffffffff816707f0>] retint_kernel+0x20/0x30
[<ffffffff8109e60a>] ? lockdep_sys_exit+0x1a/0x90
[<ffffffff812a4234>] lockdep_sys_exit_thunk+0x35/0x67
[<ffffffff81679321>] ? sysret_check+0x5/0x56
Fixing this by tracking the cpu hotplug state and displaying
the WARN only if current cpu is initialized properly.
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: stable@vger.kernel.org
Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Jiri Olsa <jolsa@redhat.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1396861448-10097-1-git-send-email-jolsa@redhat.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
When removing a (sibling) event we do:
raw_spin_lock_irq(&ctx->lock);
perf_group_detach(event);
raw_spin_unlock_irq(&ctx->lock);
<hole>
perf_remove_from_context(event);
raw_spin_lock_irq(&ctx->lock);
...
raw_spin_unlock_irq(&ctx->lock);
Now, assuming the event is a sibling, it will be 'unreachable' for
things like ctx_sched_out() because that iterates the
groups->siblings, and we just unhooked the sibling.
So, if during <hole> we get ctx_sched_out(), it will miss the event
and not call event_sched_out() on it, leaving it programmed on the
PMU.
The subsequent perf_remove_from_context() call will find the ctx is
inactive and only call list_del_event() to remove the event from all
other lists.
Hereafter we can proceed to free the event; while still programmed!
Close this hole by moving perf_group_detach() inside the same
ctx->lock region(s) perf_remove_from_context() has.
The condition on inherited events only in __perf_event_exit_task() is
likely complete crap because non-inherited events are part of groups
too and we're tearing down just the same. But leave that for another
patch.
Most-likely-Fixes: e03a9a55b4 ("perf: Change close() semantics for group events")
Reported-by: Vince Weaver <vincent.weaver@maine.edu>
Tested-by: Vince Weaver <vincent.weaver@maine.edu>
Much-staring-at-traces-by: Vince Weaver <vincent.weaver@maine.edu>
Much-staring-at-traces-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20140505093124.GN17778@laptop.programming.kicks-ass.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Pull cgroup updates from Tejun Heo:
"A lot updates for cgroup:
- The biggest one is cgroup's conversion to kernfs. cgroup took
after the long abandoned vfs-entangled sysfs implementation and
made it even more convoluted over time. cgroup's internal objects
were fused with vfs objects which also brought in vfs locking and
object lifetime rules. Naturally, there are places where vfs rules
don't fit and nasty hacks, such as credential switching or lock
dance interleaving inode mutex and cgroup_mutex with object serial
number comparison thrown in to decide whether the operation is
actually necessary, needed to be employed.
After conversion to kernfs, internal object lifetime and locking
rules are mostly isolated from vfs interactions allowing shedding
of several nasty hacks and overall simplification. This will also
allow implmentation of operations which may affect multiple cgroups
which weren't possible before as it would have required nesting
i_mutexes.
- Various simplifications including dropping of module support,
easier cgroup name/path handling, simplified cgroup file type
handling and task_cg_lists optimization.
- Prepatory changes for the planned unified hierarchy, which is still
a patchset away from being actually operational. The dummy
hierarchy is updated to serve as the default unified hierarchy.
Controllers which aren't claimed by other hierarchies are
associated with it, which BTW was what the dummy hierarchy was for
anyway.
- Various fixes from Li and others. This pull request includes some
patches to add missing slab.h to various subsystems. This was
triggered xattr.h include removal from cgroup.h. cgroup.h
indirectly got included a lot of files which brought in xattr.h
which brought in slab.h.
There are several merge commits - one to pull in kernfs updates
necessary for converting cgroup (already in upstream through
driver-core), others for interfering changes in the fixes branch"
* 'for-3.15' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup: (74 commits)
cgroup: remove useless argument from cgroup_exit()
cgroup: fix spurious lockdep warning in cgroup_exit()
cgroup: Use RCU_INIT_POINTER(x, NULL) in cgroup.c
cgroup: break kernfs active_ref protection in cgroup directory operations
cgroup: fix cgroup_taskset walking order
cgroup: implement CFTYPE_ONLY_ON_DFL
cgroup: make cgrp_dfl_root mountable
cgroup: drop const from @buffer of cftype->write_string()
cgroup: rename cgroup_dummy_root and related names
cgroup: move ->subsys_mask from cgroupfs_root to cgroup
cgroup: treat cgroup_dummy_root as an equivalent hierarchy during rebinding
cgroup: remove NULL checks from [pr_cont_]cgroup_{name|path}()
cgroup: use cgroup_setup_root() to initialize cgroup_dummy_root
cgroup: reorganize cgroup bootstrapping
cgroup: relocate setting of CGRP_DEAD
cpuset: use rcu_read_lock() to protect task_cs()
cgroup_freezer: document freezer_fork() subtleties
cgroup: update cgroup_transfer_tasks() to either succeed or fail
cgroup: drop task_lock() protection around task->cgroups
cgroup: update how a newly forked task gets associated with css_set
...
Use the ctx pmu instead of the event pmu.
When a group leader is a software event but the group contains
hardware events, the entire group is on the hardware PMU.
Using the hardware PMU for the transaction makes most sense since
that's the most expensive one to programm (and software PMUs generally
don't have TXN support anyway).
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-sctoo9t2f3nn2c9g568928q3@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Currently perf_branch_stack_sched_in iterates over the set of pmus,
checks that each pmu has a flush_branch_stack callback, then overwrites
the pmu before calling the callback. This is either redundant or broken.
In systems with a single hw pmu, pmu == cpuctx->ctx.pmu, and thus the
assignment is redundant.
In systems with multiple hw pmus (i.e. multiple pmus with task_ctx_nr ==
perf_hw_context) the pmus share the same perf_cpu_context. Thus the
assignment can cause one of the pmus to flush its branch stack
repeatedly rather than causing each of the pmus to flush their branch
stacks. Worse still, if only some pmus have the callback the assignment
can result in a branch to NULL.
This patch removes the redundant assignment.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1392054264-23570-3-git-send-email-mark.rutland@arm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Drew Richardson reported that he could make the kernel go *boom* when hotplugging
while having perf events active.
It turned out that when you have a group event, the code in
__perf_event_exit_context() fails to remove the group siblings from
the context.
We then proceed with destroying and freeing the event, and when you
re-plug the CPU and try and add another event to that CPU, things go
*boom* because you've still got dead entries there.
Reported-by: Drew Richardson <drew.richardson@arm.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Cc: Will Deacon <will.deacon@arm.com>
Cc: <stable@vger.kernel.org>
Link: http://lkml.kernel.org/n/tip-k6v5wundvusvcseqj1si0oz0@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
On Mon, Feb 10, 2014 at 08:45:16AM -0800, Dave Hansen wrote:
> The reason I coded this up was that NMIs were firing off so fast that
> nothing else was getting a chance to run. With this patch, at least the
> printk() would come out and I'd have some idea what was going on.
It will start spewing to early_printk() (which is a lot nicer to use
from NMI context too) when it fails to queue the IRQ-work because its
already enqueued.
It does have the false-positive for when two CPUs trigger the warn
concurrently, but that should be rare and some extra clutter on the
early printk shouldn't be a problem.
Cc: hpa@zytor.com
Cc: tglx@linutronix.de
Cc: dzickus@redhat.com
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: mingo@kernel.org
Fixes: 6a02ad66b2 ("perf/x86: Push the duration-logging printk() to IRQ context")
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20140211150116.GO27965@twins.programming.kicks-ass.net
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
If !NULL, @skip_css makes cgroup_taskset_for_each() skip the matching
css. The intention of the interface is to make it easy to skip css's
(cgroup_subsys_states) which already match the migration target;
however, this is entirely unnecessary as migration taskset doesn't
include tasks which are already in the target cgroup. Drop @skip_css
from cgroup_taskset_for_each().
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Li Zefan <lizefan@huawei.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Cc: Daniel Borkmann <dborkman@redhat.com>
Ingo Molnar
Peter Zijlstra
Vince Weaver
Borislav Petkov
Jiri Olsa
Yan, Zheng
Linus Torvalds
Mark Rutland
Tejun Heo