From dddd3379a619a4cb8247bfd3c94ca9ae3797aa2e Mon Sep 17 00:00:00 2001 From: Thomas Gleixner Date: Wed, 24 Nov 2010 10:05:55 +0100 Subject: perf: Fix inherit vs. context rotation bug It was found that sometimes children of tasks with inherited events had one extra event. Eventually it turned out to be due to the list rotation no being exclusive with the list iteration in the inheritance code. Cure this by temporarily disabling the rotation while we inherit the events. Signed-off-by: Thomas Gleixner Signed-off-by: Peter Zijlstra LKML-Reference: Cc: Signed-off-by: Ingo Molnar --- kernel/perf_event.c | 22 ++++++++++++++++++++-- 1 file changed, 20 insertions(+), 2 deletions(-) (limited to 'kernel') diff --git a/kernel/perf_event.c b/kernel/perf_event.c index 671f6c8c8a32..f365dd8ef8b0 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -1622,8 +1622,12 @@ static void rotate_ctx(struct perf_event_context *ctx) { raw_spin_lock(&ctx->lock); - /* Rotate the first entry last of non-pinned groups */ - list_rotate_left(&ctx->flexible_groups); + /* + * Rotate the first entry last of non-pinned groups. Rotation might be + * disabled by the inheritance code. + */ + if (!ctx->rotate_disable) + list_rotate_left(&ctx->flexible_groups); raw_spin_unlock(&ctx->lock); } @@ -6162,6 +6166,7 @@ int perf_event_init_context(struct task_struct *child, int ctxn) struct perf_event *event; struct task_struct *parent = current; int inherited_all = 1; + unsigned long flags; int ret = 0; child->perf_event_ctxp[ctxn] = NULL; @@ -6202,6 +6207,15 @@ int perf_event_init_context(struct task_struct *child, int ctxn) break; } + /* + * We can't hold ctx->lock when iterating the ->flexible_group list due + * to allocations, but we need to prevent rotation because + * rotate_ctx() will change the list from interrupt context. + */ + raw_spin_lock_irqsave(&parent_ctx->lock, flags); + parent_ctx->rotate_disable = 1; + raw_spin_unlock_irqrestore(&parent_ctx->lock, flags); + list_for_each_entry(event, &parent_ctx->flexible_groups, group_entry) { ret = inherit_task_group(event, parent, parent_ctx, child, ctxn, &inherited_all); @@ -6209,6 +6223,10 @@ int perf_event_init_context(struct task_struct *child, int ctxn) break; } + raw_spin_lock_irqsave(&parent_ctx->lock, flags); + parent_ctx->rotate_disable = 0; + raw_spin_unlock_irqrestore(&parent_ctx->lock, flags); + child_ctx = child->perf_event_ctxp[ctxn]; if (child_ctx && inherited_all) { -- cgit v1.2.3 From ee6dcfa40a50fe12a3ae0fb4d2653c66c3ed6556 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Fri, 26 Nov 2010 13:49:04 +0100 Subject: perf: Fix the software context switch counter 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 --- include/linux/perf_event.h | 29 +++++++++++++++-------------- kernel/perf_event.c | 2 -- 2 files changed, 15 insertions(+), 16 deletions(-) (limited to 'kernel') diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index 142e3d6042c7..de2c41758e29 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -909,20 +909,6 @@ extern int perf_num_counters(void); extern const char *perf_pmu_name(void); extern void __perf_event_task_sched_in(struct task_struct *task); extern void __perf_event_task_sched_out(struct task_struct *task, struct task_struct *next); - -extern atomic_t perf_task_events; - -static inline void perf_event_task_sched_in(struct task_struct *task) -{ - COND_STMT(&perf_task_events, __perf_event_task_sched_in(task)); -} - -static inline -void perf_event_task_sched_out(struct task_struct *task, struct task_struct *next) -{ - COND_STMT(&perf_task_events, __perf_event_task_sched_out(task, next)); -} - extern int perf_event_init_task(struct task_struct *child); extern void perf_event_exit_task(struct task_struct *child); extern void perf_event_free_task(struct task_struct *task); @@ -1031,6 +1017,21 @@ have_event: __perf_sw_event(event_id, nr, nmi, regs, addr); } +extern atomic_t perf_task_events; + +static inline void perf_event_task_sched_in(struct task_struct *task) +{ + COND_STMT(&perf_task_events, __perf_event_task_sched_in(task)); +} + +static inline +void perf_event_task_sched_out(struct task_struct *task, struct task_struct *next) +{ + perf_sw_event(PERF_COUNT_SW_CONTEXT_SWITCHES, 1, 1, NULL, 0); + + COND_STMT(&perf_task_events, __perf_event_task_sched_out(task, next)); +} + extern void perf_event_mmap(struct vm_area_struct *vma); extern struct perf_guest_info_callbacks *perf_guest_cbs; extern int perf_register_guest_info_callbacks(struct perf_guest_info_callbacks *callbacks); diff --git a/kernel/perf_event.c b/kernel/perf_event.c index f365dd8ef8b0..eac7e3364335 100644 --- a/kernel/perf_event.c +++ b/kernel/perf_event.c @@ -1287,8 +1287,6 @@ void __perf_event_task_sched_out(struct task_struct *task, { int ctxn; - perf_sw_event(PERF_COUNT_SW_CONTEXT_SWITCHES, 1, 1, NULL, 0); - for_each_task_context_nr(ctxn) perf_event_context_sched_out(task, ctxn, next); } -- cgit v1.2.3