perf: Add sched_task callback during ctx reschedule
Subsystems (2)
PATCH v2 1/2 perf: Add sched_task callback during ctx reschedule
From: Kan Liang <[email protected]>
Several similar kernel warnings can be triggered,
[56605.607840] CPU0 PEBS record size 0, expected 32, config 0 cpuc->record_size=208
when the below commands are running in parallel for a while on SPR.
while true; do perf record --no-buildid -a --intr-regs=AX -e cpu/event=0xd0,umask=0x81/pp -c 10003 -o /dev/null ./triad; done &
while true; do perf record -o /tmp/out -W -d -e '{ld_blocks.store_forward:period=1000000, MEM_TRANS_RETIRED.LOAD_LATENCY:u:precise=2:ldlat=4}' -c 1037 ./triad; done
*The triad program is just the generation of loads/stores.
The warnings are triggered when an unexpected PEBS record (with a different config and size) is found.
In a context switch, different events may be applied to the old task and the new task. The sched_task callback is used to switch the PMU-specific data. For the PEBS, the callback flushes the DS area and parses the PEBS records from the old task when schedule out. The new task never sees the stale PEBS records.
However, the exec() doesn't have similar support. The new command may see the PEBS records from the old command. In the perf_event_exec(), the ctx_resched() is invoked to reschedule the context. It may updates the active events, which may change the global PEBS configuration. The PEBS record from the old command may have a different config and size from the new command. The warning is triggered.
The sched_task callback should be invoked in all the places where the context can be changed. Add the sched_task callback in the ctx_resched() as well.
--- kernel/events/core.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-)
kernel/events/core.c +6 -1
@@ -2642,6 +2642,8 @@ static void perf_event_sched_in(struct perf_cpu_context *cpuctx,ctx_sched_in (ctx,EVENT_FLEXIBLE ); } +static void perf_ctx_sched_task_cb (struct perf_event_context *ctx,bool sched_in); +/* * We want to maintain the following priority of scheduling: * - CPU pinned (EVENT_CPU | EVENT_PINNED)
@@ -2680,6 +2682,7 @@ static void ctx_resched(struct perf_cpu_context *cpuctx,perf_ctx_disable (&cpuctx->ctx);if (task_ctx) {perf_ctx_disable (task_ctx); +perf_ctx_sched_task_cb (task_ctx, false);task_ctx_sched_out (task_ctx, event_type); }
@@ -2698,8 +2701,10 @@ static void ctx_resched(struct perf_cpu_context *cpuctx,perf_event_sched_in (cpuctx, task_ctx);perf_ctx_enable (&cpuctx->ctx); -if (task_ctx) +if (task_ctx) { +perf_ctx_sched_task_cb (task_ctx, true);perf_ctx_enable (task_ctx); + } }void perf_pmu_resched (struct pmu *pmu)
Signature · 1 line
2.35.1
PATCH v2 2/2 perf/x86/intel/ds: Use the size from each PEBS record
From: Kan Liang <[email protected]>
The kernel warning for the unexpected PEBS record can also be observed during a context switch, when the below commands are running in parallel for a while on SPR.
while true; do perf record --no-buildid -a --intr-regs=AX -e cpu/event=0xd0,umask=0x81/pp -c 10003 -o /dev/null ./triad; done &
while true; do perf record -o /tmp/out -W -d -e '{ld_blocks.store_forward:period=1000000, MEM_TRANS_RETIRED.LOAD_LATENCY:u:precise=2:ldlat=4}' -c 1037 ./triad; done
*The triad program is just the generation of loads/stores.
The current PEBS code assumes that all the PEBS records in the DS buffer have the same size, aka cpuc->pebs_record_size. It's true for the most cases, since the DS buffer is always flushed in every context switch.
However, there is a corner case that breaks the assumption. A system-wide PEBS event with the large PEBS config may be enabled during a context switch. Some PEBS records for the system-wide PEBS may be generated while the old task is sched out but the new one hasn't been sched in yet. When the new task is sched in, the cpuc->pebs_record_size may be updated for the per-task PEBS events. So the existing system-wide PEBS records have a different size from the later PEBS records.
Two methods were considered to fix the issue.
One is to flush the DS buffer for the system-wide PEBS right before the new task sched in. It has to be done in the generic code via the sched_task() call back. However, the sched_task() is shared among different ARCHs. The movement may impact other ARCHs, e.g., AMD BRS requires the sched_task() is called after the PMU has started on a ctxswin. The method is dropped.
The other method is implemented here. It doesn't assume that all the PEBS records have the same size any more. The size from each PEBS record is used to parse the record. For the previous platform (PEBS format < 4), which doesn't support adaptive PEBS, there is nothing changed.
---
Changes since V1:
- Del the unused variable 'cpuc'
arch/x86/events/intel/ds.c | 32 +++++++++++++++++++++++++------ arch/x86/include/asm/perf_event.h | 6 ++++++ 2 files changed, 32 insertions(+), 6 deletions(-)
arch/x86/events/intel/ds.c +26 -6
@@ -1546,6 +1546,15 @@ static inline u64 get_pebs_status(void *n)return ((struct pebs_basic *)n)->applicable_counters; } +static inline u64 get_pebs_size (void *n) +{ +struct cpu_hw_events *cpuc =this_cpu_ptr (&cpu_hw_events); + +if (x86_pmu.intel_cap.pebs_format <4 ) +return cpuc->pebs_record_size; +return intel_adaptive_pebs_size (((struct pebs_basic *)n)->format_size); +} +#define PERF_X86_EVENT_PEBS_HSW_PREC \ (PERF_X86_EVENT_PEBS_ST_HSW | \ PERF_X86_EVENT_PEBS_LD_HSW | \
@@ -1903,9 +1912,9 @@ static void setup_pebs_adaptive_sample_data(struct perf_event *event, } } -WARN_ONCE (next_record != __pebs + (format_size >>48 ), +WARN_ONCE (next_record != __pebs +intel_adaptive_pebs_size (format_size),"PEBS record size %llu, expected %llu, config %llx\n" , - format_size >>48 , +intel_adaptive_pebs_size (format_size), (u64)(next_record - __pebs), basic->format_size); }
@@ -1927,7 +1936,7 @@ get_next_pebs_record_by_bit(void *base, void *top, int bit)if (base ==NULL )return NULL ; -for (at = base; at < top; at += cpuc->pebs_record_size) { +for (at = base; at < top; at +=get_pebs_size (at)) {unsigned long status =get_pebs_status (at);if (test_bit (bit, (unsigned long *)&status)) {
@@ -2030,7 +2039,6 @@ __intel_pmu_pebs_event(struct perf_event *event, struct perf_sample_data *, struct pt_regs *)) { -struct cpu_hw_events *cpuc =this_cpu_ptr (&cpu_hw_events);struct hw_perf_event *hwc = &event->hw;struct x86_perf_regs perf_regs;struct pt_regs *regs = &perf_regs.regs;
@@ -2054,7 +2062,7 @@ __intel_pmu_pebs_event(struct perf_event *event,while (count >1 ) {setup_sample (event, iregs, at, data, regs);perf_event_output (event, data, regs); - at += cpuc->pebs_record_size; + at +=get_pebs_size (at); at =get_next_pebs_record_by_bit (at, top, bit); count--; }
@@ -2278,7 +2286,19 @@ static void intel_pmu_drain_pebs_icl(struct pt_regs *iregs, struct perf_sample_dreturn ; } -for (at = base; at < top; at += cpuc->pebs_record_size) { +/* +* The cpuc->pebs_record_size may be different from the +* size of each PEBS record. For example, a system-wide +* PEBS event with the large PEBS config may be enabled +* during a context switch. Some PEBS records for the +* system-wide PEBS may be generated while the old task +* is sched out but the new one isn't sched in. When the +* new task is sched in, the cpuc->pebs_record_size may +* be updated for the per-task PEBS events. So the +* existing system-wide PEBS records have a different +* size from the later PEBS records. +*/ +for (at = base; at < top; at +=get_pebs_size (at)) {u64 pebs_status; pebs_status =get_pebs_status (at) & cpuc->pebs_enabled;
arch/x86/include/asm/perf_event.h +6 -0
@@ -386,6 +386,12 @@ static inline bool is_topdown_idx(int idx)/* * Adaptive PEBS v4 */ +#define INTEL_ADAPTIVE_PEBS_SIZE_OFF 48 + +static inline u64 intel_adaptive_pebs_size (u64 format_size) +{ +return (format_size >>INTEL_ADAPTIVE_PEBS_SIZE_OFF ); +}struct pebs_basic {u64 format_size;
Signature · 1 line
2.35.1