On Fri, Oct 02, 2020 at 09:16:11AM -0400, Liang, Kan wrote: > Tested-by: Kan Liang <kan.li...@linux.intel.com>
--- Subject: perf/x86: Fix n_metric for cancelled txn From: Peter Zijlstra <pet...@infradead.org> Date: Mon Oct 5 10:10:24 CEST 2020 When a group that has TopDown members is failed to be scheduled, any later TopDown groups will not return valid values. Here is an example. A background perf that occupies all the GP counters and the fixed counter 1. $perf stat -e "{cycles,cycles,cycles,cycles,cycles,cycles,cycles, cycles,cycles}:D" -a A user monitors a TopDown group. It works well, because the fixed counter 3 and the PERF_METRICS are available. $perf stat -x, --topdown -- ./workload retiring,bad speculation,frontend bound,backend bound, 18.0,16.1,40.4,25.5, Then the user tries to monitor a group that has TopDown members. Because of the cycles event, the group is failed to be scheduled. $perf stat -x, -e '{slots,topdown-retiring,topdown-be-bound, topdown-fe-bound,topdown-bad-spec,cycles}' -- ./workload <not counted>,,slots,0,0.00,, <not counted>,,topdown-retiring,0,0.00,, <not counted>,,topdown-be-bound,0,0.00,, <not counted>,,topdown-fe-bound,0,0.00,, <not counted>,,topdown-bad-spec,0,0.00,, <not counted>,,cycles,0,0.00,, The user tries to monitor a TopDown group again. It doesn't work anymore. $perf stat -x, --topdown -- ./workload ,,,,, In a txn, cancel_txn() is to truncate the event_list for a canceled group and update the number of events added in this transaction. However, the number of TopDown events added in this transaction is not updated. The kernel will probably fail to add new Topdown events. Fixes: 7b2c05a15d29 ("perf/x86/intel: Generic support for hardware TopDown metrics") Reported-by: Andi Kleen <a...@linux.intel.com> Reported-by: Kan Liang <kan.li...@linux.intel.com> Signed-off-by: Peter Zijlstra (Intel) <pet...@infradead.org> Tested-by: Kan Liang <kan.li...@linux.intel.com> --- arch/x86/events/core.c | 3 +++ arch/x86/events/perf_event.h | 1 + 2 files changed, 4 insertions(+) --- a/arch/x86/events/core.c +++ b/arch/x86/events/core.c @@ -1066,6 +1066,7 @@ static int add_nr_metric_event(struct cp if (cpuc->n_metric == INTEL_TD_METRIC_NUM) return -EINVAL; cpuc->n_metric++; + cpuc->n_txn_metric++; } return 0; @@ -2065,6 +2066,7 @@ static void x86_pmu_start_txn(struct pmu perf_pmu_disable(pmu); __this_cpu_write(cpu_hw_events.n_txn, 0); __this_cpu_write(cpu_hw_events.n_txn_pair, 0); + __this_cpu_write(cpu_hw_events.n_txn_metric, 0); } /* @@ -2091,6 +2093,7 @@ static void x86_pmu_cancel_txn(struct pm __this_cpu_sub(cpu_hw_events.n_added, __this_cpu_read(cpu_hw_events.n_txn)); __this_cpu_sub(cpu_hw_events.n_events, __this_cpu_read(cpu_hw_events.n_txn)); __this_cpu_sub(cpu_hw_events.n_pair, __this_cpu_read(cpu_hw_events.n_txn_pair)); + __this_cpu_sub(cpu_hw_events.n_metric, __this_cpu_read(cpu_hw_events.n_txn_metric)); perf_pmu_enable(pmu); } --- a/arch/x86/events/perf_event.h +++ b/arch/x86/events/perf_event.h @@ -236,6 +236,7 @@ struct cpu_hw_events { int n_txn; /* the # last events in the below arrays; added in the current transaction */ int n_txn_pair; + int n_txn_metric; int assign[X86_PMC_IDX_MAX]; /* event to counter assignment */ u64 tags[X86_PMC_IDX_MAX];