Linus,
please pull the latest perf-urgent-for-linus git tree from:
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git
perf-urgent-for-linus
Two fixes for the perf subsystem:
- Fix an inconsistency of RDPMC mm struct tagging across exec() which
causes RDPMC to fault.
- Correct the timestamp mechanics across IOC_DISABLE/ENABLE which causes
incorrect timestamps and total time calculations.
Thanks,
tglx
------------------>
Peter Zijlstra (2):
perf/x86: Fix RDPMC vs. mm_struct tracking
perf/core: Fix time on IOC_ENABLE
arch/x86/events/core.c | 16 +++++++---------
include/linux/perf_event.h | 4 ++--
kernel/events/core.c | 47 ++++++++++++++++++++++++++++++++++++++--------
3 files changed, 48 insertions(+), 19 deletions(-)
diff --git a/arch/x86/events/core.c b/arch/x86/events/core.c
index 8e3db8f642a7..af12e294caed 100644
--- a/arch/x86/events/core.c
+++ b/arch/x86/events/core.c
@@ -2114,7 +2114,7 @@ static void refresh_pce(void *ignored)
load_mm_cr4(this_cpu_read(cpu_tlbstate.loaded_mm));
}
-static void x86_pmu_event_mapped(struct perf_event *event)
+static void x86_pmu_event_mapped(struct perf_event *event, struct mm_struct
*mm)
{
if (!(event->hw.flags & PERF_X86_EVENT_RDPMC_ALLOWED))
return;
@@ -2129,22 +2129,20 @@ static void x86_pmu_event_mapped(struct perf_event
*event)
* For now, this can't happen because all callers hold mmap_sem
* for write. If this changes, we'll need a different solution.
*/
- lockdep_assert_held_exclusive(¤t->mm->mmap_sem);
+ lockdep_assert_held_exclusive(&mm->mmap_sem);
- if (atomic_inc_return(¤t->mm->context.perf_rdpmc_allowed) == 1)
- on_each_cpu_mask(mm_cpumask(current->mm), refresh_pce, NULL, 1);
+ if (atomic_inc_return(&mm->context.perf_rdpmc_allowed) == 1)
+ on_each_cpu_mask(mm_cpumask(mm), refresh_pce, NULL, 1);
}
-static void x86_pmu_event_unmapped(struct perf_event *event)
+static void x86_pmu_event_unmapped(struct perf_event *event, struct mm_struct
*mm)
{
- if (!current->mm)
- return;
if (!(event->hw.flags & PERF_X86_EVENT_RDPMC_ALLOWED))
return;
- if (atomic_dec_and_test(¤t->mm->context.perf_rdpmc_allowed))
- on_each_cpu_mask(mm_cpumask(current->mm), refresh_pce, NULL, 1);
+ if (atomic_dec_and_test(&mm->context.perf_rdpmc_allowed))
+ on_each_cpu_mask(mm_cpumask(mm), refresh_pce, NULL, 1);
}
static int x86_pmu_event_idx(struct perf_event *event)
diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index a3b873fc59e4..b14095bcf4bb 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -310,8 +310,8 @@ struct pmu {
* Notification that the event was mapped or unmapped. Called
* in the context of the mapping task.
*/
- void (*event_mapped) (struct perf_event *event); /*optional*/
- void (*event_unmapped) (struct perf_event *event); /*optional*/
+ void (*event_mapped) (struct perf_event *event, struct
mm_struct *mm); /* optional */
+ void (*event_unmapped) (struct perf_event *event, struct
mm_struct *mm); /* optional */
/*
* Flags for ->add()/->del()/ ->start()/->stop(). There are
diff --git a/kernel/events/core.c b/kernel/events/core.c
index 426c2ffba16d..ee20d4c546b5 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -2217,6 +2217,33 @@ static int group_can_go_on(struct perf_event *event,
return can_add_hw;
}
+/*
+ * Complement to update_event_times(). This computes the tstamp_* values to
+ * continue 'enabled' state from @now, and effectively discards the time
+ * between the prior tstamp_stopped and now (as we were in the OFF state, or
+ * just switched (context) time base).
+ *
+ * This further assumes '@event->state == INACTIVE' (we just came from OFF) and
+ * cannot have been scheduled in yet. And going into INACTIVE state means
+ * '@event->tstamp_stopped = @now'.
+ *
+ * Thus given the rules of update_event_times():
+ *
+ * total_time_enabled = tstamp_stopped - tstamp_enabled
+ * total_time_running = tstamp_stopped - tstamp_running
+ *
+ * We can insert 'tstamp_stopped == now' and reverse them to compute new
+ * tstamp_* values.
+ */
+static void __perf_event_enable_time(struct perf_event *event, u64 now)
+{
+ WARN_ON_ONCE(event->state != PERF_EVENT_STATE_INACTIVE);
+
+ event->tstamp_stopped = now;
+ event->tstamp_enabled = now - event->total_time_enabled;
+ event->tstamp_running = now - event->total_time_running;
+}
+
static void add_event_to_ctx(struct perf_event *event,
struct perf_event_context *ctx)
{
@@ -2224,9 +2251,12 @@ static void add_event_to_ctx(struct perf_event *event,
list_add_event(event, ctx);
perf_group_attach(event);
- event->tstamp_enabled = tstamp;
- event->tstamp_running = tstamp;
- event->tstamp_stopped = tstamp;
+ /*
+ * We can be called with event->state == STATE_OFF when we create with
+ * .disabled = 1. In that case the IOC_ENABLE will call this function.
+ */
+ if (event->state == PERF_EVENT_STATE_INACTIVE)
+ __perf_event_enable_time(event, tstamp);
}
static void ctx_sched_out(struct perf_event_context *ctx,
@@ -2471,10 +2501,11 @@ static void __perf_event_mark_enabled(struct perf_event
*event)
u64 tstamp = perf_event_time(event);
event->state = PERF_EVENT_STATE_INACTIVE;
- event->tstamp_enabled = tstamp - event->total_time_enabled;
+ __perf_event_enable_time(event, tstamp);
list_for_each_entry(sub, &event->sibling_list, group_entry) {
+ /* XXX should not be > INACTIVE if event isn't */
if (sub->state >= PERF_EVENT_STATE_INACTIVE)
- sub->tstamp_enabled = tstamp - sub->total_time_enabled;
+ __perf_event_enable_time(sub, tstamp);
}
}
@@ -5090,7 +5121,7 @@ static void perf_mmap_open(struct vm_area_struct *vma)
atomic_inc(&event->rb->aux_mmap_count);
if (event->pmu->event_mapped)
- event->pmu->event_mapped(event);
+ event->pmu->event_mapped(event, vma->vm_mm);
}
static void perf_pmu_output_stop(struct perf_event *event);
@@ -5113,7 +5144,7 @@ static void perf_mmap_close(struct vm_area_struct *vma)
unsigned long size = perf_data_size(rb);
if (event->pmu->event_unmapped)
- event->pmu->event_unmapped(event);
+ event->pmu->event_unmapped(event, vma->vm_mm);
/*
* rb->aux_mmap_count will always drop before rb->mmap_count and
@@ -5411,7 +5442,7 @@ static int perf_mmap(struct file *file, struct
vm_area_struct *vma)
vma->vm_ops = &perf_mmap_vmops;
if (event->pmu->event_mapped)
- event->pmu->event_mapped(event);
+ event->pmu->event_mapped(event, vma->vm_mm);
return ret;
}
