From: Namhyung Kim Date: Fri, 5 Nov 2021 20:58:47 +0000 (-0700) Subject: perf evsel: Fix missing exclude_{host,guest} setting X-Git-Tag: v6.1-rc5~2702^2~12 X-Git-Url: http://review.tizen.org/git/?a=commitdiff_plain;h=3500eeebeda842e8499617b8983a4c55fd6bdfe3;p=platform%2Fkernel%2Flinux-starfive.git perf evsel: Fix missing exclude_{host,guest} setting The current logic for the perf missing feature has a bug that it can wrongly clear some modifiers like G or H. Actually some PMUs don't support any filtering or exclusion while others do. But we check it as a global feature. For example, the cycles event can have 'G' modifier to enable it only in the guest mode on x86. When you don't run any VMs it'll return 0. # perf stat -a -e cycles:G sleep 1 Performance counter stats for 'system wide': 0 cycles:G 1.000721670 seconds time elapsed But when it's used with other pmu events that don't support G modifier, it'll be reset and return non-zero values. # perf stat -a -e cycles:G,msr/tsc/ sleep 1 Performance counter stats for 'system wide': 538,029,960 cycles:G 16,924,010,738 msr/tsc/ 1.001815327 seconds time elapsed This is because of the missing feature detection logic being global. Add a hashmap to set pmu-specific exclude_host/guest features. Committer notes: Fix 'perf test python' by adding a stub for evsel__find_pmu() in tools/perf/util/python.c, document that it is used so far only for the above reasons so that if anybody needs this in the python binding usecases, we can revisit this. Reported-by: Stephane Eranian Signed-off-by: Namhyung Kim Cc: Andi Kleen Cc: Ian Rogers Cc: Jiri Olsa Cc: Peter Zijlstra Cc: Ravi Bangoria Link: http://lore.kernel.org/lkml/20211105205847.120950-1-namhyung@kernel.org Signed-off-by: Arnaldo Carvalho de Melo --- diff --git a/tools/perf/util/evsel.c b/tools/perf/util/evsel.c index 2cfc293..3cc1f8f 100644 --- a/tools/perf/util/evsel.c +++ b/tools/perf/util/evsel.c @@ -1824,7 +1824,7 @@ static void evsel__disable_missing_features(struct evsel *evsel) evsel->open_flags &= ~(unsigned long)PERF_FLAG_FD_CLOEXEC; if (perf_missing_features.mmap2) evsel->core.attr.mmap2 = 0; - if (perf_missing_features.exclude_guest) + if (evsel->pmu && evsel->pmu->missing_features.exclude_guest) evsel->core.attr.exclude_guest = evsel->core.attr.exclude_host = 0; if (perf_missing_features.lbr_flags) evsel->core.attr.branch_sample_type &= ~(PERF_SAMPLE_BRANCH_NO_FLAGS | @@ -1917,10 +1917,27 @@ bool evsel__detect_missing_features(struct evsel *evsel) perf_missing_features.mmap2 = true; pr_debug2_peo("switching off mmap2\n"); return true; - } else if (!perf_missing_features.exclude_guest && - (evsel->core.attr.exclude_guest || evsel->core.attr.exclude_host)) { - perf_missing_features.exclude_guest = true; - pr_debug2_peo("switching off exclude_guest, exclude_host\n"); + } else if ((evsel->core.attr.exclude_guest || evsel->core.attr.exclude_host) && + (evsel->pmu == NULL || evsel->pmu->missing_features.exclude_guest)) { + if (evsel->pmu == NULL) { + evsel->pmu = evsel__find_pmu(evsel); + if (evsel->pmu) + evsel->pmu->missing_features.exclude_guest = true; + else { + /* we cannot find PMU, disable attrs now */ + evsel->core.attr.exclude_host = false; + evsel->core.attr.exclude_guest = false; + } + } + + if (evsel->exclude_GH) { + pr_debug2_peo("PMU has no exclude_host/guest support, bailing out\n"); + return false; + } + if (!perf_missing_features.exclude_guest) { + perf_missing_features.exclude_guest = true; + pr_debug2_peo("switching off exclude_guest, exclude_host\n"); + } return true; } else if (!perf_missing_features.sample_id_all) { perf_missing_features.sample_id_all = true; diff --git a/tools/perf/util/evsel.h b/tools/perf/util/evsel.h index 846c827..dcc87c2 100644 --- a/tools/perf/util/evsel.h +++ b/tools/perf/util/evsel.h @@ -22,6 +22,7 @@ struct target; struct hashmap; struct bperf_leader_bpf; struct bperf_follower_bpf; +struct perf_pmu; typedef int (evsel__sb_cb_t)(union perf_event *event, void *data); @@ -153,6 +154,9 @@ struct evsel { }; unsigned long open_flags; int precise_ip_original; + + /* for missing_features */ + struct perf_pmu *pmu; }; struct perf_missing_features { diff --git a/tools/perf/util/pmu.h b/tools/perf/util/pmu.h index 15bbec3..541889f 100644 --- a/tools/perf/util/pmu.h +++ b/tools/perf/util/pmu.h @@ -49,6 +49,10 @@ struct perf_pmu { struct list_head caps; /* HEAD struct perf_pmu_caps -> list */ struct list_head list; /* ELEM */ struct list_head hybrid_list; + + struct { + bool exclude_guest; + } missing_features; }; extern struct perf_pmu perf_pmu__fake; diff --git a/tools/perf/util/python.c b/tools/perf/util/python.c index 8feef3a..563a9ba 100644 --- a/tools/perf/util/python.c +++ b/tools/perf/util/python.c @@ -70,6 +70,18 @@ void perf_stat__collect_metric_expr(struct evlist *evsel_list) } /* + * This one is needed not to drag the PMU bandwagon, jevents generated + * pmu_sys_event_tables, etc and evsel__find_pmu() is used so far just for + * doing per PMU perf_event_attr.exclude_guest handling, not really needed, so + * far, for the perf python binding known usecases, revisit if this become + * necessary. + */ +struct perf_pmu *evsel__find_pmu(struct evsel *evsel __maybe_unused) +{ + return NULL; +} + +/* * Add this one here not to drag util/metricgroup.c */ int metricgroup__copy_metric_events(struct evlist *evlist, struct cgroup *cgrp,