From 3812d2987733c5a00e103be4e23d63ec9342043a Mon Sep 17 00:00:00 2001 From: Adrian Hunter Date: Fri, 10 Jun 2022 14:33:15 +0300 Subject: perf record: Add finished init event In preparation for recording sideband events in a virtual machine guest so that they can be injected into a host perf.data file. This is needed to enable injecting events after the initial synthesized user events (that have an all zero id sample) but before regular events. Committer notes: Add entry about PERF_RECORD_FINISHED_INIT to tools/perf/Documentation/perf.data-file-format.txt. Committer testing: Before: # perf report -D | grep FINISHED 0 0x5910 [0x8]: PERF_RECORD_FINISHED_ROUND FINISHED_ROUND events: 1 ( 0.5%) # After: # perf record -- sleep 1 [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0.020 MB perf.data (7 samples) ] # perf report -D | grep FINISHED 0 0x5068 [0x8]: PERF_RECORD_FINISHED_INIT: unhandled! 0 0x5390 [0x8]: PERF_RECORD_FINISHED_ROUND FINISHED_ROUND events: 1 ( 0.5%) FINISHED_INIT events: 1 ( 0.5%) # Signed-off-by: Adrian Hunter Acked-by: Ian Rogers Tested-by: Arnaldo Carvalho de Melo Cc: Jiri Olsa Cc: Namhyung Kim Link: https://lore.kernel.org/r/20220610113316.6682-5-adrian.hunter@intel.com Signed-off-by: Arnaldo Carvalho de Melo --- tools/lib/perf/include/perf/event.h | 1 + 1 file changed, 1 insertion(+) (limited to 'tools/lib') diff --git a/tools/lib/perf/include/perf/event.h b/tools/lib/perf/include/perf/event.h index e7758707cadd..9f7ca070da87 100644 --- a/tools/lib/perf/include/perf/event.h +++ b/tools/lib/perf/include/perf/event.h @@ -389,6 +389,7 @@ enum perf_user_event_type { /* above any possible kernel type */ PERF_RECORD_TIME_CONV = 79, PERF_RECORD_HEADER_FEATURE = 80, PERF_RECORD_COMPRESSED = 81, + PERF_RECORD_FINISHED_INIT = 82, PERF_RECORD_HEADER_MAX }; -- cgit v1.2.3 From b47bb18661eaed30790d70b7563a5220b3c59594 Mon Sep 17 00:00:00 2001 From: Adrian Hunter Date: Mon, 11 Jul 2022 12:31:53 +0300 Subject: perf tools: Add machine_pid and vcpu to id_index When injecting events from a guest perf.data file, the events will have separate sample ID numbers. These ID numbers can then be used to determine which machine an event belongs to. To facilitate that, add machine_pid and vcpu to id_index records. For backward compatibility, these are added at the end of the record, and the length of the record is used to determine if they are present or not. Note, this is needed because the events from a guest perf.data file contain the pid/tid of the process running at that time inside the VM not the pid/tid of the (QEMU) hypervisor thread. So a way is needed to relate guest events back to the guest machine and VCPU, and using sample ID numbers for that is relatively simple and convenient. Signed-off-by: Adrian Hunter Cc: Andi Kleen Cc: Ian Rogers Cc: Jiri Olsa Cc: Namhyung Kim Cc: kvm@vger.kernel.org Link: https://lore.kernel.org/r/20220711093218.10967-11-adrian.hunter@intel.com Signed-off-by: Arnaldo Carvalho de Melo --- tools/lib/perf/include/internal/evsel.h | 4 +++ tools/lib/perf/include/perf/event.h | 5 ++++ tools/perf/util/session.c | 40 ++++++++++++++++++++++---- tools/perf/util/synthetic-events.c | 51 +++++++++++++++++++++++++-------- tools/perf/util/synthetic-events.h | 1 + 5 files changed, 84 insertions(+), 17 deletions(-) (limited to 'tools/lib') diff --git a/tools/lib/perf/include/internal/evsel.h b/tools/lib/perf/include/internal/evsel.h index 2a912a1f1989..a99a75d9e78f 100644 --- a/tools/lib/perf/include/internal/evsel.h +++ b/tools/lib/perf/include/internal/evsel.h @@ -30,6 +30,10 @@ struct perf_sample_id { struct perf_cpu cpu; pid_t tid; + /* Guest machine pid and VCPU, valid only if machine_pid is non-zero */ + pid_t machine_pid; + struct perf_cpu vcpu; + /* Holds total ID period value for PERF_SAMPLE_READ processing. */ u64 period; }; diff --git a/tools/lib/perf/include/perf/event.h b/tools/lib/perf/include/perf/event.h index 9f7ca070da87..c2dbd3e88885 100644 --- a/tools/lib/perf/include/perf/event.h +++ b/tools/lib/perf/include/perf/event.h @@ -237,6 +237,11 @@ struct id_index_entry { __u64 tid; }; +struct id_index_entry_2 { + __u64 machine_pid; + __u64 vcpu; +}; + struct perf_record_id_index { struct perf_event_header header; __u64 nr; diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c index 4c9513bc6d89..5141fe164e97 100644 --- a/tools/perf/util/session.c +++ b/tools/perf/util/session.c @@ -2756,18 +2756,35 @@ int perf_event__process_id_index(struct perf_session *session, { struct evlist *evlist = session->evlist; struct perf_record_id_index *ie = &event->id_index; + size_t sz = ie->header.size - sizeof(*ie); size_t i, nr, max_nr; + size_t e1_sz = sizeof(struct id_index_entry); + size_t e2_sz = sizeof(struct id_index_entry_2); + size_t etot_sz = e1_sz + e2_sz; + struct id_index_entry_2 *e2; - max_nr = (ie->header.size - sizeof(struct perf_record_id_index)) / - sizeof(struct id_index_entry); + max_nr = sz / e1_sz; nr = ie->nr; - if (nr > max_nr) + if (nr > max_nr) { + printf("Too big: nr %zu max_nr %zu\n", nr, max_nr); return -EINVAL; + } + + if (sz >= nr * etot_sz) { + max_nr = sz / etot_sz; + if (nr > max_nr) { + printf("Too big2: nr %zu max_nr %zu\n", nr, max_nr); + return -EINVAL; + } + e2 = (void *)ie + sizeof(*ie) + nr * e1_sz; + } else { + e2 = NULL; + } if (dump_trace) fprintf(stdout, " nr: %zu\n", nr); - for (i = 0; i < nr; i++) { + for (i = 0; i < nr; i++, (e2 ? e2++ : 0)) { struct id_index_entry *e = &ie->entries[i]; struct perf_sample_id *sid; @@ -2775,15 +2792,28 @@ int perf_event__process_id_index(struct perf_session *session, fprintf(stdout, " ... id: %"PRI_lu64, e->id); fprintf(stdout, " idx: %"PRI_lu64, e->idx); fprintf(stdout, " cpu: %"PRI_ld64, e->cpu); - fprintf(stdout, " tid: %"PRI_ld64"\n", e->tid); + fprintf(stdout, " tid: %"PRI_ld64, e->tid); + if (e2) { + fprintf(stdout, " machine_pid: %"PRI_ld64, e2->machine_pid); + fprintf(stdout, " vcpu: %"PRI_lu64"\n", e2->vcpu); + } else { + fprintf(stdout, "\n"); + } } sid = evlist__id2sid(evlist, e->id); if (!sid) return -ENOENT; + sid->idx = e->idx; sid->cpu.cpu = e->cpu; sid->tid = e->tid; + + if (!e2) + continue; + + sid->machine_pid = e2->machine_pid; + sid->vcpu.cpu = e2->vcpu; } return 0; } diff --git a/tools/perf/util/synthetic-events.c b/tools/perf/util/synthetic-events.c index ed9623702f34..2ae59c03ae77 100644 --- a/tools/perf/util/synthetic-events.c +++ b/tools/perf/util/synthetic-events.c @@ -1759,19 +1759,26 @@ int perf_event__synthesize_id_sample(__u64 *array, u64 type, const struct perf_s return (void *)array - (void *)start; } -int perf_event__synthesize_id_index(struct perf_tool *tool, perf_event__handler_t process, - struct evlist *evlist, struct machine *machine) +int __perf_event__synthesize_id_index(struct perf_tool *tool, perf_event__handler_t process, + struct evlist *evlist, struct machine *machine, size_t from) { union perf_event *ev; struct evsel *evsel; - size_t nr = 0, i = 0, sz, max_nr, n; + size_t nr = 0, i = 0, sz, max_nr, n, pos; + size_t e1_sz = sizeof(struct id_index_entry); + size_t e2_sz = sizeof(struct id_index_entry_2); + size_t etot_sz = e1_sz + e2_sz; + bool e2_needed = false; int err; - max_nr = (UINT16_MAX - sizeof(struct perf_record_id_index)) / - sizeof(struct id_index_entry); + max_nr = (UINT16_MAX - sizeof(struct perf_record_id_index)) / etot_sz; - evlist__for_each_entry(evlist, evsel) + pos = 0; + evlist__for_each_entry(evlist, evsel) { + if (pos++ < from) + continue; nr += evsel->core.ids; + } if (!nr) return 0; @@ -1779,31 +1786,38 @@ int perf_event__synthesize_id_index(struct perf_tool *tool, perf_event__handler_ pr_debug2("Synthesizing id index\n"); n = nr > max_nr ? max_nr : nr; - sz = sizeof(struct perf_record_id_index) + n * sizeof(struct id_index_entry); + sz = sizeof(struct perf_record_id_index) + n * etot_sz; ev = zalloc(sz); if (!ev) return -ENOMEM; + sz = sizeof(struct perf_record_id_index) + n * e1_sz; + ev->id_index.header.type = PERF_RECORD_ID_INDEX; - ev->id_index.header.size = sz; ev->id_index.nr = n; + pos = 0; evlist__for_each_entry(evlist, evsel) { u32 j; - for (j = 0; j < evsel->core.ids; j++) { + if (pos++ < from) + continue; + for (j = 0; j < evsel->core.ids; j++, i++) { struct id_index_entry *e; + struct id_index_entry_2 *e2; struct perf_sample_id *sid; if (i >= n) { + ev->id_index.header.size = sz + (e2_needed ? n * e2_sz : 0); err = process(tool, ev, NULL, machine); if (err) goto out_err; nr -= n; i = 0; + e2_needed = false; } - e = &ev->id_index.entries[i++]; + e = &ev->id_index.entries[i]; e->id = evsel->core.id[j]; @@ -1816,11 +1830,18 @@ int perf_event__synthesize_id_index(struct perf_tool *tool, perf_event__handler_ e->idx = sid->idx; e->cpu = sid->cpu.cpu; e->tid = sid->tid; + + if (sid->machine_pid) + e2_needed = true; + + e2 = (void *)ev + sz; + e2[i].machine_pid = sid->machine_pid; + e2[i].vcpu = sid->vcpu.cpu; } } - sz = sizeof(struct perf_record_id_index) + nr * sizeof(struct id_index_entry); - ev->id_index.header.size = sz; + sz = sizeof(struct perf_record_id_index) + nr * e1_sz; + ev->id_index.header.size = sz + (e2_needed ? nr * e2_sz : 0); ev->id_index.nr = nr; err = process(tool, ev, NULL, machine); @@ -1830,6 +1851,12 @@ out_err: return err; } +int perf_event__synthesize_id_index(struct perf_tool *tool, perf_event__handler_t process, + struct evlist *evlist, struct machine *machine) +{ + return __perf_event__synthesize_id_index(tool, process, evlist, machine, 0); +} + int __machine__synthesize_threads(struct machine *machine, struct perf_tool *tool, struct target *target, struct perf_thread_map *threads, perf_event__handler_t process, bool needs_mmap, diff --git a/tools/perf/util/synthetic-events.h b/tools/perf/util/synthetic-events.h index b136ec3ec95d..81cb3d6af0b9 100644 --- a/tools/perf/util/synthetic-events.h +++ b/tools/perf/util/synthetic-events.h @@ -55,6 +55,7 @@ int perf_event__synthesize_extra_attr(struct perf_tool *tool, struct evlist *evs int perf_event__synthesize_extra_kmaps(struct perf_tool *tool, perf_event__handler_t process, struct machine *machine); int perf_event__synthesize_features(struct perf_tool *tool, struct perf_session *session, struct evlist *evlist, perf_event__handler_t process); int perf_event__synthesize_id_index(struct perf_tool *tool, perf_event__handler_t process, struct evlist *evlist, struct machine *machine); +int __perf_event__synthesize_id_index(struct perf_tool *tool, perf_event__handler_t process, struct evlist *evlist, struct machine *machine, size_t from); int perf_event__synthesize_id_sample(__u64 *array, u64 type, const struct perf_sample *sample); int perf_event__synthesize_kernel_mmap(struct perf_tool *tool, perf_event__handler_t process, struct machine *machine); int perf_event__synthesize_mmap_events(struct perf_tool *tool, union perf_event *event, pid_t pid, pid_t tgid, perf_event__handler_t process, struct machine *machine, bool mmap_data); -- cgit v1.2.3 From 7151c1d17820c0cfcda0c890f55a868cb4336afc Mon Sep 17 00:00:00 2001 From: Adrian Hunter Date: Mon, 11 Jul 2022 12:32:00 +0300 Subject: perf auxtrace: Add machine_pid and vcpu to auxtrace_error Add machine_pid and vcpu to struct perf_record_auxtrace_error. The existing fmt member is used to identify the new format. The new members make it possible to easily differentiate errors from guest machines. Signed-off-by: Adrian Hunter Acked-by: Ian Rogers Cc: Andi Kleen Cc: Jiri Olsa Cc: Namhyung Kim Cc: kvm@vger.kernel.org Link: https://lore.kernel.org/r/20220711093218.10967-18-adrian.hunter@intel.com Signed-off-by: Arnaldo Carvalho de Melo --- tools/lib/perf/include/perf/event.h | 2 ++ tools/perf/util/auxtrace.c | 30 +++++++++++++++++----- tools/perf/util/auxtrace.h | 4 +++ .../util/scripting-engines/trace-event-python.c | 4 ++- tools/perf/util/session.c | 4 +++ 5 files changed, 37 insertions(+), 7 deletions(-) (limited to 'tools/lib') diff --git a/tools/lib/perf/include/perf/event.h b/tools/lib/perf/include/perf/event.h index c2dbd3e88885..556bb06798f2 100644 --- a/tools/lib/perf/include/perf/event.h +++ b/tools/lib/perf/include/perf/event.h @@ -279,6 +279,8 @@ struct perf_record_auxtrace_error { __u64 ip; __u64 time; char msg[MAX_AUXTRACE_ERROR_MSG]; + __u32 machine_pid; + __u32 vcpu; }; struct perf_record_aux { diff --git a/tools/perf/util/auxtrace.c b/tools/perf/util/auxtrace.c index 511dd3caa1bc..6edab8a16de6 100644 --- a/tools/perf/util/auxtrace.c +++ b/tools/perf/util/auxtrace.c @@ -1189,9 +1189,10 @@ void auxtrace_buffer__free(struct auxtrace_buffer *buffer) free(buffer); } -void auxtrace_synth_error(struct perf_record_auxtrace_error *auxtrace_error, int type, - int code, int cpu, pid_t pid, pid_t tid, u64 ip, - const char *msg, u64 timestamp) +void auxtrace_synth_guest_error(struct perf_record_auxtrace_error *auxtrace_error, int type, + int code, int cpu, pid_t pid, pid_t tid, u64 ip, + const char *msg, u64 timestamp, + pid_t machine_pid, int vcpu) { size_t size; @@ -1207,12 +1208,26 @@ void auxtrace_synth_error(struct perf_record_auxtrace_error *auxtrace_error, int auxtrace_error->ip = ip; auxtrace_error->time = timestamp; strlcpy(auxtrace_error->msg, msg, MAX_AUXTRACE_ERROR_MSG); - - size = (void *)auxtrace_error->msg - (void *)auxtrace_error + - strlen(auxtrace_error->msg) + 1; + if (machine_pid) { + auxtrace_error->fmt = 2; + auxtrace_error->machine_pid = machine_pid; + auxtrace_error->vcpu = vcpu; + size = sizeof(*auxtrace_error); + } else { + size = (void *)auxtrace_error->msg - (void *)auxtrace_error + + strlen(auxtrace_error->msg) + 1; + } auxtrace_error->header.size = PERF_ALIGN(size, sizeof(u64)); } +void auxtrace_synth_error(struct perf_record_auxtrace_error *auxtrace_error, int type, + int code, int cpu, pid_t pid, pid_t tid, u64 ip, + const char *msg, u64 timestamp) +{ + auxtrace_synth_guest_error(auxtrace_error, type, code, cpu, pid, tid, + ip, msg, timestamp, 0, -1); +} + int perf_event__synthesize_auxtrace_info(struct auxtrace_record *itr, struct perf_tool *tool, struct perf_session *session, @@ -1662,6 +1677,9 @@ size_t perf_event__fprintf_auxtrace_error(union perf_event *event, FILE *fp) if (!e->fmt) msg = (const char *)&e->time; + if (e->fmt >= 2 && e->machine_pid) + ret += fprintf(fp, " machine_pid %d vcpu %d", e->machine_pid, e->vcpu); + ret += fprintf(fp, " cpu %d pid %d tid %d ip %#"PRI_lx64" code %u: %s\n", e->cpu, e->pid, e->tid, e->ip, e->code, msg); return ret; diff --git a/tools/perf/util/auxtrace.h b/tools/perf/util/auxtrace.h index cd0d25c2751c..6a4fbfd34c6b 100644 --- a/tools/perf/util/auxtrace.h +++ b/tools/perf/util/auxtrace.h @@ -595,6 +595,10 @@ int auxtrace_index__process(int fd, u64 size, struct perf_session *session, bool needs_swap); void auxtrace_index__free(struct list_head *head); +void auxtrace_synth_guest_error(struct perf_record_auxtrace_error *auxtrace_error, int type, + int code, int cpu, pid_t pid, pid_t tid, u64 ip, + const char *msg, u64 timestamp, + pid_t machine_pid, int vcpu); void auxtrace_synth_error(struct perf_record_auxtrace_error *auxtrace_error, int type, int code, int cpu, pid_t pid, pid_t tid, u64 ip, const char *msg, u64 timestamp); diff --git a/tools/perf/util/scripting-engines/trace-event-python.c b/tools/perf/util/scripting-engines/trace-event-python.c index adba01b7d9dd..3367c5479199 100644 --- a/tools/perf/util/scripting-engines/trace-event-python.c +++ b/tools/perf/util/scripting-engines/trace-event-python.c @@ -1559,7 +1559,7 @@ static void python_process_auxtrace_error(struct perf_session *session __maybe_u msg = (const char *)&e->time; } - t = tuple_new(9); + t = tuple_new(11); tuple_set_u32(t, 0, e->type); tuple_set_u32(t, 1, e->code); @@ -1570,6 +1570,8 @@ static void python_process_auxtrace_error(struct perf_session *session __maybe_u tuple_set_u64(t, 6, tm); tuple_set_string(t, 7, msg); tuple_set_u32(t, 8, cpumode); + tuple_set_s32(t, 9, e->machine_pid); + tuple_set_s32(t, 10, e->vcpu); call_object(handler, t, handler_name); diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c index f3e9fa557bc9..7ea0b91013ea 100644 --- a/tools/perf/util/session.c +++ b/tools/perf/util/session.c @@ -895,6 +895,10 @@ static void perf_event__auxtrace_error_swap(union perf_event *event, event->auxtrace_error.ip = bswap_64(event->auxtrace_error.ip); if (event->auxtrace_error.fmt) event->auxtrace_error.time = bswap_64(event->auxtrace_error.time); + if (event->auxtrace_error.fmt >= 2) { + event->auxtrace_error.machine_pid = bswap_32(event->auxtrace_error.machine_pid); + event->auxtrace_error.vcpu = bswap_32(event->auxtrace_error.vcpu); + } } static void perf_event__thread_map_swap(union perf_event *event, -- cgit v1.2.3