Skip to contentSkip to conversation

perf, powerpc: Fix power_pmu_event_init to not use event->ctx

PATCH In review Paul Mackerras 2 participants 5 messages Oct 19, 2010

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 perf, powerpc: Fix power_pmu_event_init to not use event->ctx

Commit c3f00c70 ("perf: Separate find_get_context() from event initialization") changed the generic perf_event code to call perf_event_alloc, which calls the arch-specific event_init code, before looking up the context for the new event. Unfortunately, power_pmu_event_init uses event->ctx->task to see whether the new event is a per-task event or a system-wide event, and thus crashes since event->ctx is NULL at the point where power_pmu_event_init gets called.

(The reason it needs to know whether it is a per-task event is because there are some hardware events on Power systems which only count when the processor is not idle, and there are some fixed-function counters which count such events. For example, the "run cycles" event counts cycles when the processor is not idle. If the user asks to count cycles, we can use "run cycles" if this is a per-task event, since the processor is running when the task is running, by definition. We can't use "run cycles" if the user asks for "cycles" on a system-wide counter.)

Fortunately the information we need is in the event->attach_state field, so we just use that instead.

---
 arch/powerpc/kernel/perf_event.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/powerpc/kernel/perf_event.c +1 -1
index 9cb4924..3129c85 100644
@@ -1092,7 +1092,7 @@ static int power_pmu_event_init(struct perf_event *event)
 	 * XXX we should check if the task is an idle task.
 	 */
 	flags = 0;
-	if (event->ctx->task)
+	if (event->attach_state & PERF_ATTACH_TASK)
 		flags |= PPMU_ONLY_COUNT_RUN;
 
 	/*
maintainer Acked-by on v1 1/1
#1 lore raw
Quoted 19 lines · Paul Mackerras

Commit c3f00c70 ("perf: Separate find_get_context() from event initialization") changed the generic perf_event code to call perf_event_alloc, which calls the arch-specific event_init code, before looking up the context for the new event. Unfortunately, power_pmu_event_init uses event->ctx->task to see whether the new event is a per-task event or a system-wide event, and thus crashes since event->ctx is NULL at the point where power_pmu_event_init gets called.

(The reason it needs to know whether it is a per-task event is because there are some hardware events on Power systems which only count when the processor is not idle, and there are some fixed-function counters which count such events. For example, the "run cycles" event counts cycles when the processor is not idle. If the user asks to count cycles, we can use "run cycles" if this is a per-task event, since the processor is running when the task is running, by definition. We can't use "run cycles" if the user asks for "cycles" on a system-wide counter.)

Quoted 2 lines

Fortunately the information we need is in the event->attach_state field, so we just use that instead.

Ah, sorry for the breakage.

Quoted 2 lines
Quoted 17 lines
---
 arch/powerpc/kernel/perf_event.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/powerpc/kernel/perf_event.c +1 -1
index 9cb4924..3129c85 100644
@@ -1092,7 +1092,7 @@ static int power_pmu_event_init(struct perf_event *event)
 	 * XXX we should check if the task is an idle task.
 	 */
 	flags = 0;
-	if (event->ctx->task)
+	if (event->attach_state & PERF_ATTACH_TASK)
 		flags |= PPMU_ONLY_COUNT_RUN;
 
 	/*
#4 lore raw
In reply to Paul Mackerras
Commit-ID:  57fa7214330be2e292ddb1402834ff0b221ef29a
Gitweb:     http://git.kernel.org/tip/57fa7214330be2e292ddb1402834ff0b221ef29a
Author:     Paul Mackerras <[email protected]>
AuthorDate: Tue, 19 Oct 2010 16:55:35 +1100
Committer:  Ingo Molnar <[email protected]>
CommitDate: Tue, 19 Oct 2010 09:18:34 +0200

perf, powerpc: Fix power_pmu_event_init to not use event->ctx

Commit c3f00c70 ("perf: Separate find_get_context() from event initialization") changed the generic perf_event code to call perf_event_alloc, which calls the arch-specific event_init code, before looking up the context for the new event. Unfortunately, power_pmu_event_init uses event->ctx->task to see whether the new event is a per-task event or a system-wide event, and thus crashes since event->ctx is NULL at the point where power_pmu_event_init gets called.

(The reason it needs to know whether it is a per-task event is because there are some hardware events on Power systems which only count when the processor is not idle, and there are some fixed-function counters which count such events. For example, the "run cycles" event counts cycles when the processor is not idle. If the user asks to count cycles, we can use "run cycles" if this is a per-task event, since the processor is running when the task is running, by definition. We can't use "run cycles" if the user asks for "cycles" on a system-wide counter.)

Fortunately the information we need is in the event->attach_state field, so we just use that instead.

Cc Benjamin Herrenschmidt <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Frederic Weisbecker <[email protected]>

LKML-Reference: <20101019055535.GA10398@drongo>

---
 arch/powerpc/kernel/perf_event.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
arch/powerpc/kernel/perf_event.c +1 -1
index 9cb4924..3129c85 100644
@@ -1092,7 +1092,7 @@ static int power_pmu_event_init(struct perf_event *event)
 	 * XXX we should check if the task is an idle task.
 	 */
 	flags = 0;
-	if (event->ctx->task)
+	if (event->attach_state & PERF_ATTACH_TASK)
 		flags |= PPMU_ONLY_COUNT_RUN;
 
 	/*
maintainer
#2 lore raw
In reply to Peter Zijlstra
Quoted 9 lines · Peter Zijlstra
Quoted 9 lines

(The reason it needs to know whether it is a per-task event is because there are some hardware events on Power systems which only count when the processor is not idle, and there are some fixed-function counters which count such events. For example, the "run cycles" event counts cycles when the processor is not idle. If the user asks to count cycles, we can use "run cycles" if this is a per-task event, since the processor is running when the task is running, by definition. We can't use "run cycles" if the user asks for "cycles" on a system-wide counter.)

Right, so the problem comes from you using run-cycles as cycles? How does that interact with things like cpu_relax() will that generate a difference in run-cycles vs cycles?

#3 lore raw
Quoted 12 lines · Peter Zijlstra
Quoted 9 lines · Peter Zijlstra
Quoted 9 lines

(The reason it needs to know whether it is a per-task event is because there are some hardware events on Power systems which only count when the processor is not idle, and there are some fixed-function counters which count such events. For example, the "run cycles" event counts cycles when the processor is not idle. If the user asks to count cycles, we can use "run cycles" if this is a per-task event, since the processor is running when the task is running, by definition. We can't use "run cycles" if the user asks for "cycles" on a system-wide counter.)

Quoted 3 lines

Right, so the problem comes from you using run-cycles as cycles? How does that interact with things like cpu_relax() will that generate a difference in run-cycles vs cycles?

Cycles in cpu_relax() are still counted as run cycles. Whether cycles are considered "run" cycles or "idle" cycles is controlled by a bit in an SPR that we set and clear in the idle loop and in the interrupt entry code. Only cycles when we're actually in the idle loop count as "idle" cycles.

Paul.