linux-stable.git/include/linux/perf_event.h, branch linux-2.6.38.y

perf: Fix task context scheduling

2011-04-21T21:32:25+00:00

commit ab711fe08297de1485fff0a366e6db8828cafd6a upstream.

Jiri reported:

 |
 | - once an event is created by sys_perf_event_open, task context
 |   is created and it stays even if the event is closed, until the
 |   task is finished ... thats what I see in code and I assume it's
 |   correct
 |
 | - when the task opens event, perf_sched_events jump label is
 |   incremented and following callbacks are started from scheduler
 |
 |         __perf_event_task_sched_in
 |         __perf_event_task_sched_out
 |
 |   These callback *in/out set/unset cpuctx->task_ctx value to the
 |   task context.
 |
 | - close is called on event on CPU 0:
 |         - the task is scheduled on CPU 0
 |         - __perf_event_task_sched_in is called
 |         - cpuctx->task_ctx is set
 |         - perf_sched_events jump label is decremented and == 0
 |         - __perf_event_task_sched_out is not called
 |         - cpuctx->task_ctx on CPU 0 stays set
 |
 | - exit is called on CPU 1:
 |         - the task is scheduled on CPU 1
 |         - perf_event_exit_task is called
 |         - task_ctx_sched_out unsets cpuctx->task_ctx on CPU 1
 |         - put_ctx destroys the context
 |
 | - another call of perf_rotate_context on CPU 0 will use invalid
 |   task_ctx pointer, and eventualy panic.
 |

Cure this the simplest possibly way by partially reverting the
jump_label optimization for the sched_out case.

Reported-and-tested-by: Jiri Olsa 
Signed-off-by: Peter Zijlstra 
Cc: Oleg Nesterov 
LKML-Reference: <1301520405.4859.213.camel@twins>
Signed-off-by: Ingo Molnar 
Signed-off-by: Greg Kroah-Hartman

perf: Sysfs enumeration

2010-12-16T10:36:43+00:00

Simple sysfs emumeration of the PMUs.

Use a "event_source" bus, and add PMU devices using their name.

Each PMU device has a type attribute which contrains the value needed
for perf_event_attr::type to identify this PMU.

This is the minimal stub needed to start using this interface,
we'll consider extending the sysfs usage later.

Cc: Kay Sievers 
Cc: Greg KH 
Signed-off-by: Peter Zijlstra 
LKML-Reference: <20101117222056.316982569@chello.nl>
Signed-off-by: Ingo Molnar

perf: Dynamic pmu types

2010-12-16T10:36:43+00:00

Extend the perf_pmu_register() interface to allow for named and
dynamic pmu types.

Because we need to support the existing static types we cannot use
dynamic types for everything, hence provide a type argument.

If we want to enumerate the PMUs they need a name, provide one.

Signed-off-by: Peter Zijlstra 
LKML-Reference: <20101117222056.259707703@chello.nl>
Signed-off-by: Ingo Molnar

Merge branch 'perf/urgent' into perf/core

2010-12-16T10:22:27+00:00

Merge reason: We want to apply a dependent patch.

Signed-off-by: Ingo Molnar

perf: Fix duplicate events with multiple-pmu vs software events

2010-12-08T19:14:08+00:00

Because the multi-pmu bits can share contexts between struct pmu
instances we could get duplicate events by iterating the pmu list.

Signed-off-by: Peter Zijlstra 
Signed-off-by: Thomas Gleixner 
LKML-Reference: 
Signed-off-by: Ingo Molnar

perf events: Make sample_type identity fields available in all PERF_RECORD_ events

2010-12-05T01:02:20+00:00

If perf_event_attr.sample_id_all is set it will add the PERF_SAMPLE_ identity
info:

TID, TIME, ID, CPU, STREAM_ID

As a trailer, so that older perf tools can process new files, just ignoring the
extra payload.

With this its possible to do further analysis on problems in the event stream,
like detecting reordering of MMAP and FORK events, etc.

V2: Fixup header size in comm, mmap and task processing, as we have to take into
account different sample_types for each matching event, noticed by Thomas Gleixner.

Thomas also noticed a problem in v2 where if we didn't had space in the buffer we
wouldn't restore the header size.

Tested-by: Thomas Gleixner 
Reviewed-by: Thomas Gleixner 
Acked-by: Ian Munsie 
Acked-by: Peter Zijlstra 
Acked-by: Thomas Gleixner 
Cc: Frédéric Weisbecker 
Cc: Ian Munsie 
Cc: Mike Galbraith 
Cc: Peter Zijlstra 
Cc: Paul Mackerras 
Cc: Stephane Eranian 
Cc: Thomas Gleixner 
LKML-Reference: 
Signed-off-by: Arnaldo Carvalho de Melo

perf events: Separate the routines handling the PERF_SAMPLE_ identity fields

2010-12-05T00:56:48+00:00

Those will be made available in sample like events like MMAP, EXEC, etc in a
followup patch. So precalculate the extra id header space and have a separate
routine to fill them up.

V2: Thomas noticed that the id header needs to be precalculated at
inherit_events too:

LKML-Reference: 

Tested-by: Thomas Gleixner 
Reviewed-by: Thomas Gleixner 
Acked-by: Ian Munsie 
Acked-by: Peter Zijlstra 
Acked-by: Thomas Gleixner 
Cc: Frédéric Weisbecker 
Cc: Ian Munsie 
Cc: Mike Galbraith 
Cc: Peter Zijlstra 
Cc: Paul Mackerras 
Cc: Stephane Eranian 
Cc: Thomas Gleixner 
LKML-Reference: <1291318772-30880-2-git-send-email-acme@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo

perf events: Precalculate the header space for PERF_SAMPLE_ fields

2010-11-30T21:19:04+00:00

PERF_SAMPLE_{CALLCHAIN,RAW} have variable lenghts per sample, but the others
can be precalculated, reducing a bit the per sample cost.

Acked-by: Peter Zijlstra 
Cc: Frédéric Weisbecker 
Cc: Ian Munsie 
Cc: Mike Galbraith 
Cc: Peter Zijlstra 
Cc: Paul Mackerras 
Cc: Stephane Eranian 
LKML-Reference: 
Signed-off-by: Arnaldo Carvalho de Melo

perf: Introduce is_sampling_event()

2010-11-26T14:14:54+00:00

and use it when appropriate.

Signed-off-by: Franck Bui-Huu 
Signed-off-by: Peter Zijlstra 
LKML-Reference: <1290525705-6265-1-git-send-email-fbuihuu@gmail.com>
Signed-off-by: Ingo Molnar

perf: Fix the software context switch counter

2010-11-26T14:00:59+00:00

Stephane noticed that because the perf_sw_event() call is inside the
perf_event_task_sched_out() call it won't get called unless we
have a per-task counter.

Reported-by: Stephane Eranian 
Signed-off-by: Peter Zijlstra 
LKML-Reference: 
Signed-off-by: Ingo Molnar