linux_dsm_epyc7002/kernel
Peter Zijlstra 10c6db110d perf: Fix loss of notification with multi-event
When you do:
        $ perf record -e cycles,cycles,cycles noploop 10

You expect about 10,000 samples for each event, i.e., 10s at
1000samples/sec. However, this is not what's happening. You
get much fewer samples, maybe 3700 samples/event:

$ perf report -D | tail -15
Aggregated stats:
           TOTAL events:      10998
            MMAP events:         66
            COMM events:          2
          SAMPLE events:      10930
cycles stats:
           TOTAL events:       3644
          SAMPLE events:       3644
cycles stats:
           TOTAL events:       3642
          SAMPLE events:       3642
cycles stats:
           TOTAL events:       3644
          SAMPLE events:       3644

On a Intel Nehalem or even AMD64, there are 4 counters capable
of measuring cycles, so there is plenty of space to measure those
events without multiplexing (even with the NMI watchdog active).
And even with multiplexing, we'd expect roughly the same number
of samples per event.

The root of the problem was that when the event that caused the buffer
to become full was not the first event passed on the cmdline, the user
notification would get lost. The notification was sent to the file
descriptor of the overflowed event but the perf tool was not polling
on it.  The perf tool aggregates all samples into a single buffer,
i.e., the buffer of the first event. Consequently, it assumes
notifications for any event will come via that descriptor.

The seemingly straight forward solution of moving the waitq into the
ringbuffer object doesn't work because of life-time issues. One could
perf_event_set_output() on a fd that you're also blocking on and cause
the old rb object to be freed while its waitq would still be
referenced by the blocked thread -> FAIL.

Therefore link all events to the ringbuffer and broadcast the wakeup
from the ringbuffer object to all possible events that could be waited
upon. This is rather ugly, and we're open to better solutions but it
works for now.

Reported-by: Stephane Eranian <eranian@google.com>
Finished-by: Stephane Eranian <eranian@google.com>
Reviewed-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/20111126014731.GA7030@quad
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2011-12-05 09:33:03 +01:00
..
debug Merge branch 'modsplit-Oct31_2011' of git://git.kernel.org/pub/scm/linux/kernel/git/paulg/linux 2011-11-06 19:44:47 -08:00
events perf: Fix loss of notification with multi-event 2011-12-05 09:33:03 +01:00
gcov gcov: disable CONSTRUCTORS for UML 2011-07-26 16:49:45 -07:00
irq Merge branch 'modsplit-Oct31_2011' of git://git.kernel.org/pub/scm/linux/kernel/git/paulg/linux 2011-11-06 19:44:47 -08:00
power PM / QoS: Set cpu_dma_pm_qos->name 2011-11-07 23:02:24 +01:00
time kernel: Fix files explicitly needing EXPORT_SYMBOL infrastructure 2011-10-31 19:30:05 -04:00
trace kernel: Fix files explicitly needing EXPORT_SYMBOL infrastructure 2011-10-31 19:30:05 -04:00
.gitignore
acct.c
async.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
audit_tree.c audit_tree,rcu: Convert call_rcu(__put_tree) to kfree_rcu() 2011-07-20 14:10:11 -07:00
audit_watch.c kill path_lookup() 2011-03-14 09:15:23 -04:00
audit.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
audit.h
auditfilter.c netlink: kill loginuid/sessionid/sid members from struct netlink_skb_parms 2011-03-03 10:55:40 -08:00
auditsc.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
backtracetest.c
bounds.c memcg: remove direct page_cgroup-to-page pointer 2011-03-23 19:46:28 -07:00
capability.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
cgroup_freezer.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
cgroup.c memcg: replace ss->id_lock with a rwlock 2011-11-02 16:07:03 -07:00
compat.c kernel: Fix files explicitly needing EXPORT_SYMBOL infrastructure 2011-10-31 19:30:05 -04:00
configs.c kernel/configs.c: include MODULE_*() when CONFIG_IKCONFIG_PROC=n 2011-07-25 20:57:15 -07:00
cpu_pm.c cpu_pm: call notifiers during suspend 2011-09-23 12:05:29 +05:30
cpu.c Merge branch 'modsplit-Oct31_2011' of git://git.kernel.org/pub/scm/linux/kernel/git/paulg/linux 2011-11-06 19:44:47 -08:00
cpuset.c Merge branch 'modsplit-Oct31_2011' of git://git.kernel.org/pub/scm/linux/kernel/git/paulg/linux 2011-11-06 19:44:47 -08:00
crash_dump.c Merge branch 'modsplit-Oct31_2011' of git://git.kernel.org/pub/scm/linux/kernel/git/paulg/linux 2011-11-06 19:44:47 -08:00
cred.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
delayacct.c KVM: Steal time implementation 2011-07-14 12:59:14 +03:00
dma.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
elfcore.c
exec_domain.c
exit.c oom: remove oom_disable_count 2011-10-31 17:30:45 -07:00
extable.c extable, core_kernel_data(): Make sure all archs define _sdata 2011-05-20 08:56:56 +02:00
fork.c Merge branch 'writeback-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/wfg/linux 2011-11-06 19:02:23 -08:00
freezer.c Merge branch 'modsplit-Oct31_2011' of git://git.kernel.org/pub/scm/linux/kernel/git/paulg/linux 2011-11-06 19:44:47 -08:00
futex_compat.c userns: user namespaces: convert several capable() calls 2011-03-23 19:47:08 -07:00
futex.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
groups.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
hrtimer.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
hung_task.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
irq_work.c kernel: fix two implicit header assumptions in irq_work.c 2011-10-31 09:20:12 -04:00
itimer.c
jump_label.c jump-label: initialize jump-label subsystem much earlier 2011-10-25 11:55:15 -07:00
kallsyms.c Merge branch 'core-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip 2011-03-25 17:52:22 -07:00
Kconfig.freezer
Kconfig.hz
Kconfig.locks arch:Kconfig.locks Remove unused config option. 2011-04-10 17:01:05 +02:00
Kconfig.preempt sched: Isolate preempt counting in its own config option 2011-06-10 15:15:40 +02:00
kexec.c [S390] kdump: Add infrastructure for unmapping crashkernel memory 2011-10-30 15:16:42 +01:00
kfifo.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
kmod.c kmod: prevent kmod_loop_msg overflow in __request_module() 2011-10-26 13:10:39 +10:30
kprobes.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
ksysfs.c kernel: ksysfs.c is implicitly using stat.h 2011-10-31 09:20:13 -04:00
kthread.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
latencytop.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
lockdep_internals.h
lockdep_proc.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
lockdep_states.h
lockdep.c Merge branch 'core-rcu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip 2011-10-26 16:26:53 +02:00
Makefile Merge branch 'devel-stable' of http://ftp.arm.linux.org.uk/pub/linux/arm/kernel/git-cur/linux-2.6-arm 2011-10-28 12:02:27 -07:00
module.c Merge branch 'modsplit-Oct31_2011' of git://git.kernel.org/pub/scm/linux/kernel/git/paulg/linux 2011-11-06 19:44:47 -08:00
mutex-debug.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
mutex-debug.h mutex: Use p->on_cpu for the adaptive spin 2011-04-14 08:52:33 +02:00
mutex.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
mutex.h mutex: Use p->on_cpu for the adaptive spin 2011-04-14 08:52:33 +02:00
notifier.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
nsproxy.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
padata.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
panic.c module,bug: Add TAINT_OOT_MODULE flag for modules not built in-tree 2011-11-07 07:54:42 +10:30
params.c kernel: params.c needs module.h not moduleparam.h 2011-10-31 09:20:13 -04:00
pid_namespace.c pidns: call pid_ns_prepare_proc() from create_pid_namespace() 2011-03-23 19:46:58 -07:00
pid.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
posix-cpu-timers.c Merge branch 'core-locking-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip 2011-10-26 16:17:32 +02:00
posix-timers.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
printk.c printk: remove bounds checking for log_prefix 2011-10-31 17:30:53 -07:00
profile.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
ptrace.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
range.c range: fix bogus misuse of module.h to get printk() 2011-10-31 09:20:11 -04:00
rcu.h rcu: Add grace-period, quiescent-state, and call_rcu trace events 2011-09-28 21:38:21 -07:00
rcupdate.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
rcutiny_plugin.h kernel: fix up module header handling in rcutiny files 2011-10-31 09:20:13 -04:00
rcutiny.c kernel: fix up module header handling in rcutiny files 2011-10-31 09:20:13 -04:00
rcutorture.c rcu: Make rcu_torture_boost() exit loops at end of test 2011-09-28 21:38:46 -07:00
rcutree_plugin.h rcu: Remove rcu_needs_cpu_flush() to avoid false quiescent states 2011-09-28 21:38:48 -07:00
rcutree_trace.c rcu: Simplify quiescent-state accounting 2011-09-28 21:38:22 -07:00
rcutree.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
rcutree.h rcu: Remove rcu_needs_cpu_flush() to avoid false quiescent states 2011-09-28 21:38:48 -07:00
relay.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
res_counter.c memcg: res_counter_read_u64(): fix potential races on 32-bit machines 2011-03-23 19:46:22 -07:00
resource.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
rtmutex_common.h rtmutex: Simplify PI algorithm and make highest prio task get lock 2011-01-27 21:13:51 -05:00
rtmutex-debug.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
rtmutex-debug.h
rtmutex-tester.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
rtmutex.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
rtmutex.h
rwsem.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
sched_autogroup.c Fix common misspellings 2011-03-31 11:26:23 -03:00
sched_autogroup.h sched: Skip autogroup when looking for all rt sched groups 2011-07-01 10:39:08 +02:00
sched_clock.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
sched_cpupri.c sched/cpupri: Remove cpupri->pri_active 2011-08-14 12:01:11 +02:00
sched_cpupri.h sched/cpupri: Remove cpupri->pri_active 2011-08-14 12:01:11 +02:00
sched_debug.c sched: Get rid of lock_depth 2011-04-24 13:18:38 +02:00
sched_fair.c sched: Wrap scheduler p->cpus_allowed access 2011-10-06 12:46:56 +02:00
sched_features.h sched: Kill WAKEUP_PREEMPT 2011-08-14 12:00:41 +02:00
sched_idletask.c sched: Drop the rq argument to sched_class::select_task_rq() 2011-04-14 08:52:36 +02:00
sched_rt.c sched: Warn on rt throttling 2011-10-06 12:47:04 +02:00
sched_stats.h locking, sched: Annotate thread_group_cputimer as raw 2011-09-13 11:11:55 +02:00
sched_stoptask.c sched: Implement hierarchical task accounting for SCHED_OTHER 2011-08-14 12:01:13 +02:00
sched.c Merge branch 'next' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc 2011-11-06 17:12:03 -08:00
seccomp.c
semaphore.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
signal.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
smp.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
softirq.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
spinlock.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
srcu.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
stacktrace.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
stop_machine.c Merge branch 'modsplit-Oct31_2011' of git://git.kernel.org/pub/scm/linux/kernel/git/paulg/linux 2011-11-06 19:44:47 -08:00
sys_ni.c Cross Memory Attach 2011-10-31 17:30:44 -07:00
sys.c Merge branch 'modsplit-Oct31_2011' of git://git.kernel.org/pub/scm/linux/kernel/git/paulg/linux 2011-11-06 19:44:47 -08:00
sysctl_binary.c ipv4: NET_IPV4_ROUTE_GC_INTERVAL removal 2011-10-03 14:13:01 -04:00
sysctl_check.c xfs: remove subdirectories 2011-08-12 16:21:35 -05:00
sysctl.c Merge branch 'akpm' (Andrew's incoming) 2011-10-31 17:46:07 -07:00
taskstats.c Make TASKSTATS require root access 2011-09-19 17:04:37 -07:00
test_kprobes.c
time.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
timeconst.pl
timer.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
tracepoint.c Tracepoint: Dissociate from module mutex 2011-08-10 20:38:14 -04:00
tsacct.c Make taskstats round statistics down to nearest 1k bytes/events 2011-09-19 17:10:57 -07:00
uid16.c userns: user namespaces: convert several capable() calls 2011-03-23 19:47:08 -07:00
up.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
user_namespace.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
user-return-notifier.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
user.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
utsname_sysctl.c Merge branch 'modsplit-Oct31_2011' of git://git.kernel.org/pub/scm/linux/kernel/git/paulg/linux 2011-11-06 19:44:47 -08:00
utsname.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
wait.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00
watchdog.c watchdog: move watchdog_*_all_cpus under CONFIG_SYSCTL 2011-10-31 17:30:53 -07:00
workqueue_sched.h
workqueue.c kernel: Map most files to use export.h instead of module.h 2011-10-31 09:20:12 -04:00