Skip to content

L
linux
Switch branch/tag
  1. 15 May, 2009 9 commits
  3. 13 May, 2009 1 commit
  5. 12 May, 2009 1 commit
    • Paul Mackerras's avatar
      perf_counter: call hw_perf_save_disable/restore around group_sched_in · e758a33d
      Paul Mackerras authored 
      I noticed that when enabling a group via the PERF_COUNTER_IOC_ENABLE
ioctl on the group leader, the counters weren't enabled and counting
immediately on return from the ioctl, but did start counting a little
while later (presumably after a context switch).

The reason was that __perf_counter_enable calls group_sched_in which
calls hw_perf_group_sched_in, which on powerpc assumes that the caller
has called hw_perf_save_disable already.  Until commit 46d686c6
("perf_counter: put whole group on when enabling group leader") it was
true that all callers of group_sched_in had called
hw_perf_save_disable first, and the powerpc hw_perf_group_sched_in
relies on that (there isn't an x86 version).

This fixes the problem by putting calls to hw_perf_save_disable /
hw_perf_restore around the calls to group_sched_in and
counter_sched_in in __perf_counter_enable.  Having the calls to
hw_perf_save_disable/restore around the counter_sched_in call is
harmless and makes this call consistent with the other call sites
of counter_sched_in, which have all called hw_perf_save_disable first.

[ Impact: more precise counter group disable/enable functionality ]
Signed-off-by: default avatarPaul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
LKML-Reference: <18953.25733.53359.147452@cargo.ozlabs.ibm.com>
Signed-off-by: default avatarIngo Molnar <mingo@elte.hu>
      e758a33d
  7. 11 May, 2009 4 commits
    • Paul Mackerras's avatar
      perf_counter: call atomic64_set for counter->count · 615a3f1e
      Paul Mackerras authored 
      A compile warning triggered because we are calling
atomic_set(&counter->count). But since counter->count
is an atomic64_t, we have to use atomic64_set.

So the count can be set short, resulting in the reset ioctl
only resetting the low word.

[ Impact: clear counter properly during the reset ioctl ]
Signed-off-by: default avatarPaul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
LKML-Reference: <18951.48285.270311.981806@drongo.ozlabs.ibm.com>
Signed-off-by: default avatarIngo Molnar <mingo@elte.hu>
      615a3f1e
    • Paul Mackerras's avatar
      perf_counter: don't count scheduler ticks as context switches · a08b159f
      Paul Mackerras authored 
      The context-switch software counter gives inflated values at present
because each scheduler tick and each process-wide counter
enable/disable prctl gets counted as a context switch.

This happens because perf_counter_task_tick, perf_counter_task_disable
and perf_counter_task_enable all call perf_counter_task_sched_out,
which calls perf_swcounter_event to record a context switch event.

This fixes it by introducing a variant of perf_counter_task_sched_out
with two underscores in front for internal use within the perf_counter
code, and makes perf_counter_task_{tick,disable,enable} call it.  This
variant doesn't record a context switch event, and takes a struct
perf_counter_context *.  This adds the new variant rather than
changing the behaviour or interface of perf_counter_task_sched_out
because that is called from other code.

[ Impact: fix inflated context-switch event counts ]
Signed-off-by: default avatarPaul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
LKML-Reference: <18951.48034.485580.498953@drongo.ozlabs.ibm.com>
Signed-off-by: default avatarIngo Molnar <mingo@elte.hu>
      a08b159f
    • Paul Mackerras's avatar
      perf_counter: Put whole group on when enabling group leader · 6751b71e
      Paul Mackerras authored 
      Currently, if you have a group where the leader is disabled and there
are siblings that are enabled, and then you enable the leader, we only
put the leader on the PMU, and not its enabled siblings.  This is
incorrect, since the enabled group members should be all on or all off
at any given point.

This fixes it by adding a call to group_sched_in in
__perf_counter_enable in the case where we're enabling a group leader.

To avoid the need for a forward declaration this also moves
group_sched_in up before __perf_counter_enable.  The actual content of
group_sched_in is unchanged by this patch.

[ Impact: fix bug in counter enable code ]
Signed-off-by: default avatarPaul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
LKML-Reference: <18951.34946.451546.691693@drongo.ozlabs.ibm.com>
Signed-off-by: default avatarIngo Molnar <mingo@elte.hu>
      6751b71e
    • Mike Galbraith's avatar
      perf_counter, x86: clean up throttling printk · 88233923
      Mike Galbraith authored 
      s/PERFMON/perfcounters for perfcounter interrupt throttling warning.

'perfmon' is the CPU feature name that is Intel-only, while we do
throttling in a generic way.

[ Impact: cleanup ]
Signed-off-by: default avatarMike Galbraith <efault@gmx.de>
Cc: Robert Richter <robert.richter@amd.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <new-submission>
Signed-off-by: default avatarIngo Molnar <mingo@elte.hu>
      88233923
  9. 10 May, 2009 1 commit
  11. 09 May, 2009 1 commit
  13. 08 May, 2009 4 commits
    • Peter Zijlstra's avatar
      perf_counter: add PERF_RECORD_CPU · f370e1e2
      Peter Zijlstra authored 
      Allow recording the CPU number the event was generated on.

RFC: this leaves a u32 as reserved, should we fill in the
     node_id() there, or leave this open for future extention,
     as userspace can already easily do the cpu->node mapping
     if needed.

[ Impact: extend perfcounter output record format ]
Signed-off-by: default avatarPeter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
LKML-Reference: <20090508170029.008627711@chello.nl>
Signed-off-by: default avatarIngo Molnar <mingo@elte.hu>
      f370e1e2
    • Peter Zijlstra's avatar
      perf_counter: add PERF_RECORD_CONFIG · a85f61ab
      Peter Zijlstra authored 
      Much like CONFIG_RECORD_GROUP records the hw_event.config to
identify the values, allow to record this for all counters.

[ Impact: extend perfcounter output record format ]
Signed-off-by: default avatarPeter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
LKML-Reference: <20090508170028.923228280@chello.nl>
Signed-off-by: default avatarIngo Molnar <mingo@elte.hu>
      a85f61ab
    • Peter Zijlstra's avatar
      perf_counter: rework ioctl()s · 3df5edad
      Peter Zijlstra authored 
      Corey noticed that ioctl()s on grouped counters didn't work on
the whole group. This extends the ioctl() interface to take a
second argument that is interpreted as a flags field. We then
provide PERF_IOC_FLAG_GROUP to toggle the behaviour.

Having this flag gives the greatest flexibility, allowing you
to individually enable/disable/reset counters in a group, or
all together.

[ Impact: fix group counter enable/disable semantics ]
Reported-by: default avatarCorey Ashford <cjashfor@linux.vnet.ibm.com>
Signed-off-by: default avatarPeter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <20090508170028.837558214@chello.nl>
Signed-off-by: default avatarIngo Molnar <mingo@elte.hu>
      3df5edad
    • Peter Zijlstra's avatar
      perf_counter: optimize perf_counter_task_tick() · 7fc23a53
      Peter Zijlstra authored 
      perf_counter_task_tick() does way too much work to find out
there's nothing to do. Provide an easy short-circuit for the
normal case where there are no counters on the system.

[ Impact: micro-optimization ]
Signed-off-by: default avatarPeter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
LKML-Reference: <20090508170028.750619201@chello.nl>
Signed-off-by: default avatarIngo Molnar <mingo@elte.hu>
      7fc23a53
  15. 06 May, 2009 1 commit
  17. 05 May, 2009 6 commits
  19. 04 May, 2009 5 commits
    • Ingo Molnar's avatar
      perf_counter: fix fixed-purpose counter support on v2 Intel-PERFMON · 066d7dea
      Ingo Molnar authored 
      Fixed-purpose counters stopped working in a simple 'perf stat ls' run:

   <not counted>  cache references
   <not counted>  cache misses

Due to:

  ef7b3e09: perf_counter, x86: remove vendor check in fixed_mode_idx()

Which made x86_pmu.num_counters_fixed matter: if it's nonzero, the
fixed-purpose counters are utilized.

But on v2 perfmon this field is not set (despite there being
fixed-purpose PMCs). So add a quirk to set the number of fixed-purpose
counters to at least three.

[ Impact: add quirk for three fixed-purpose counters on certain Intel CPUs ]

Cc: Robert Richter <robert.richter@amd.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1241002046-8832-28-git-send-email-robert.richter@amd.com>
Signed-off-by: default avatarIngo Molnar <mingo@elte.hu>
      066d7dea
    • Ingo Molnar's avatar
      perf_counter: convert perf_resource_mutex to a spinlock · 1dce8d99
      Ingo Molnar authored 
      Now percpu counters can be initialized very early. But the init
sequence uses mutex_lock(). Fortunately, perf_resource_mutex should
be a spinlock anyway, so convert it.

[ Impact: fix crash due to early init mutex use ]

LKML-Reference: <new-submission>
Signed-off-by: default avatarIngo Molnar <mingo@elte.hu>
      1dce8d99
    • Ingo Molnar's avatar
      perf_counter: initialize the per-cpu context earlier · 0d905bca
      Ingo Molnar authored 
      percpu scheduling for perfcounters wants to take the context lock,
but that lock first needs to be initialized. Currently it is an
early_initcall() - but that is too late, the task tick runs much
sooner than that.

Call it explicitly from the scheduler init sequence instead.

[ Impact: fix access-before-init crash ]

LKML-Reference: <new-submission>
Signed-off-by: default avatarIngo Molnar <mingo@elte.hu>
      0d905bca
    • Peter Zijlstra's avatar
      perf_counter: x86: fixup nmi_watchdog vs perf_counter boo-boo · ba77813a
      Peter Zijlstra authored 
      Invert the atomic_inc_not_zero() test so that we will indeed detect the
first activation.

Also rename the global num_counters, since its easy to confuse with
x86_pmu.num_counters.

[ Impact: fix non-working perfcounters on AMD CPUs, cleanup ]
Signed-off-by: default avatarPeter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1241455664.7620.4938.camel@twins>
Signed-off-by: default avatarIngo Molnar <mingo@elte.hu>
      ba77813a
    • Ingo Molnar's avatar
      perf_counter: round-robin per-CPU counters too · b82914ce
      Ingo Molnar authored 
      This used to be unstable when we had the rq->lock dependencies,
but now that they are that of the past we can turn on percpu
counter RR too.

[ Impact: handle counter over-commit for per-CPU counters too ]

LKML-Reference: <new-submission>
Signed-off-by: default avatarIngo Molnar <mingo@elte.hu>
      b82914ce
  21. 03 May, 2009 1 commit
  23. 01 May, 2009 6 commits
GitLab Nexedi Edition | About GitLab | About Nexedi | 沪ICP备2021021310号-2 | 沪ICP备2021021310号-7