From: Ian Rogers <irogers@google.com>
To: Adrian Hunter <adrian.hunter@intel.com>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>,
Jiri Olsa <jolsa@redhat.com>, Namhyung Kim <namhyung@kernel.org>,
Andi Kleen <ak@linux.intel.com>,
linux-kernel@vger.kernel.org, kvm@vger.kernel.org
Subject: Re: [PATCH 10/35] perf tools: Add machine_pid and vcpu to id_index
Date: Tue, 19 Jul 2022 10:48:19 -0700 [thread overview]
Message-ID: <CAP-5=fUOHiOyKi0_Mp9EMD5Jz-K0+6R8Vz6=+rpAbhL1neqxEQ@mail.gmail.com> (raw)
In-Reply-To: <20220711093218.10967-11-adrian.hunter@intel.com>
On Mon, Jul 11, 2022 at 2:33 AM Adrian Hunter <adrian.hunter@intel.com> wrote:
>
> 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 <adrian.hunter@intel.com>
> ---
> 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(-)
>
> 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 @@ int perf_event__synthesize_id_index(struct perf_tool *tool, perf_event__handler_
> 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);
Given there is only 1 use in the file defining the function, should
this just be static with no header file declaration?
Thanks,
Ian
> 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);
> --
> 2.25.1
>
next prev parent reply other threads:[~2022-07-19 17:48 UTC|newest]
Thread overview: 83+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-07-11 9:31 [PATCH 00/35] perf intel-pt: Add support for tracing virtual machine user space on the host Adrian Hunter
2022-07-11 9:31 ` [PATCH 01/35] perf tools: Fix dso_id inode generation comparison Adrian Hunter
2022-07-18 14:57 ` Arnaldo Carvalho de Melo
2022-07-19 10:18 ` Adrian Hunter
2022-07-19 15:13 ` Ian Rogers
2022-07-19 19:16 ` Arnaldo Carvalho de Melo
2022-07-11 9:31 ` [PATCH 02/35] perf tools: Export dsos__for_each_with_build_id() Adrian Hunter
2022-07-19 16:55 ` Ian Rogers
2022-07-11 9:31 ` [PATCH 03/35] perf ordered_events: Add ordered_events__last_flush_time() Adrian Hunter
2022-07-19 16:56 ` Ian Rogers
2022-07-11 9:31 ` [PATCH 04/35] perf tools: Export perf_event__process_finished_round() Adrian Hunter
2022-07-19 17:04 ` Ian Rogers
2022-08-09 11:37 ` Adrian Hunter
2022-07-11 9:31 ` [PATCH 05/35] perf tools: Factor out evsel__id_hdr_size() Adrian Hunter
2022-07-19 17:09 ` Ian Rogers
2022-08-09 11:49 ` Adrian Hunter
2022-08-09 17:07 ` Ian Rogers
2022-07-11 9:31 ` [PATCH 06/35] perf tools: Add perf_event__synthesize_id_sample() Adrian Hunter
2022-07-19 17:10 ` Ian Rogers
2022-07-11 9:31 ` [PATCH 07/35] perf script: Add --dump-unsorted-raw-trace option Adrian Hunter
2022-07-19 17:11 ` Ian Rogers
2022-07-11 9:31 ` [PATCH 08/35] perf buildid-cache: Add guestmount'd files to the build ID cache Adrian Hunter
2022-07-19 17:41 ` Ian Rogers
2022-08-09 12:21 ` Adrian Hunter
2022-07-11 9:31 ` [PATCH 09/35] perf buildid-cache: Do not require purge files to also be in the file system Adrian Hunter
2022-07-19 17:44 ` Ian Rogers
2022-07-11 9:31 ` [PATCH 10/35] perf tools: Add machine_pid and vcpu to id_index Adrian Hunter
2022-07-19 17:48 ` Ian Rogers [this message]
2022-08-09 12:19 ` Adrian Hunter
2022-07-11 9:31 ` [PATCH 11/35] perf session: Create guest machines from id_index Adrian Hunter
2022-07-19 17:51 ` Ian Rogers
2022-07-11 9:31 ` [PATCH 12/35] perf tools: Add guest_cpu to hypervisor threads Adrian Hunter
2022-07-20 0:23 ` Ian Rogers
2022-07-11 9:31 ` [PATCH 13/35] perf tools: Add machine_pid and vcpu to perf_sample Adrian Hunter
2022-07-20 0:36 ` Ian Rogers
2022-07-11 9:31 ` [PATCH 14/35] perf tools: Use sample->machine_pid to find guest machine Adrian Hunter
2022-07-20 0:37 ` Ian Rogers
2022-07-11 9:31 ` [PATCH 15/35] perf script: Add machine_pid and vcpu Adrian Hunter
2022-07-20 0:39 ` Ian Rogers
2022-07-11 9:31 ` [PATCH 16/35] perf dlfilter: " Adrian Hunter
2022-07-20 0:42 ` Ian Rogers
2022-07-11 9:32 ` [PATCH 17/35] perf auxtrace: Add machine_pid and vcpu to auxtrace_error Adrian Hunter
2022-07-20 0:43 ` Ian Rogers
2022-07-11 9:32 ` [PATCH 18/35] perf script python: Add machine_pid and vcpu Adrian Hunter
2022-07-20 0:43 ` Ian Rogers
2022-07-11 9:32 ` [PATCH 19/35] perf script python: intel-pt-events: " Adrian Hunter
2022-07-20 0:44 ` Ian Rogers
2022-07-11 9:32 ` [PATCH 20/35] perf tools: Remove also guest kcore_dir with host kcore_dir Adrian Hunter
2022-07-20 0:45 ` Ian Rogers
2022-07-11 9:32 ` [PATCH 21/35] perf tools: Make has_kcore_dir() work also for guest kcore_dir Adrian Hunter
2022-07-20 0:49 ` Ian Rogers
2022-07-11 9:32 ` [PATCH 22/35] perf tools: Automatically use guest kcore_dir if present Adrian Hunter
2022-07-20 0:51 ` Ian Rogers
2022-07-11 9:32 ` [PATCH 23/35] perf tools: Add reallocarray_as_needed() Adrian Hunter
2022-07-20 0:55 ` Ian Rogers
2022-08-09 16:48 ` Adrian Hunter
2022-07-11 9:32 ` [PATCH 24/35] perf inject: Add support for injecting guest sideband events Adrian Hunter
2022-07-20 1:06 ` Ian Rogers
2022-08-11 17:19 ` Adrian Hunter
2022-07-11 9:32 ` [PATCH 25/35] perf machine: Use realloc_array_as_needed() in machine__set_current_tid() Adrian Hunter
2022-07-11 9:32 ` [PATCH 26/35] perf tools: Handle injected guest kernel mmap event Adrian Hunter
2022-07-20 1:09 ` Ian Rogers
2022-07-11 9:32 ` [PATCH 27/35] perf tools: Add perf_event__is_guest() Adrian Hunter
2022-07-20 1:11 ` Ian Rogers
2022-07-20 14:06 ` Arnaldo Carvalho de Melo
2022-07-20 14:56 ` Ian Rogers
2022-07-11 9:32 ` [PATCH 28/35] perf intel-pt: Remove guest_machine_pid Adrian Hunter
2022-07-20 1:12 ` Ian Rogers
2022-07-11 9:32 ` [PATCH 29/35] perf intel-pt: Add some more logging to intel_pt_walk_next_insn() Adrian Hunter
2022-07-20 1:13 ` Ian Rogers
2022-07-11 9:32 ` [PATCH 30/35] perf intel-pt: Track guest context switches Adrian Hunter
2022-07-20 1:13 ` Ian Rogers
2022-07-11 9:32 ` [PATCH 31/35] perf intel-pt: Disable sync switch with guest sideband Adrian Hunter
2022-07-20 1:14 ` Ian Rogers
2022-07-11 9:32 ` [PATCH 32/35] perf intel-pt: Determine guest thread from " Adrian Hunter
2022-07-20 1:15 ` Ian Rogers
2022-07-11 9:32 ` [PATCH 33/35] perf intel-pt: Add machine_pid and vcpu to auxtrace_error Adrian Hunter
2022-07-20 5:27 ` Ian Rogers
2022-07-11 9:32 ` [PATCH 34/35] perf intel-pt: Use guest pid/tid etc in guest samples Adrian Hunter
2022-07-20 5:28 ` Ian Rogers
2022-07-11 9:32 ` [PATCH 35/35] perf intel-pt: Add documentation for tracing guest machine user space Adrian Hunter
2022-07-20 5:29 ` Ian Rogers
2022-07-18 15:28 ` [PATCH 00/35] perf intel-pt: Add support for tracing virtual machine user space on the host Arnaldo Carvalho de Melo
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to='CAP-5=fUOHiOyKi0_Mp9EMD5Jz-K0+6R8Vz6=+rpAbhL1neqxEQ@mail.gmail.com' \
--to=irogers@google.com \
--cc=acme@kernel.org \
--cc=adrian.hunter@intel.com \
--cc=ak@linux.intel.com \
--cc=jolsa@redhat.com \
--cc=kvm@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=namhyung@kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).