Re: [PATCH 1/7] perf bench: Add build-id injection benchmark

[Namhyung Kim]

Hi Ian,

On Thu, Sep 24, 2020 at 7:13 AM Ian Rogers <irogers@xxxxxxxxxx> wrote:
>
> On Wed, Sep 23, 2020 at 1:05 AM Namhyung Kim <namhyung@xxxxxxxxxx> wrote:
> >
> > Sometimes I can see perf record piped with perf inject take long time
> > processing build-id.  So add inject-build-id benchmark to the
> > internals benchmark suite to measure its overhead regularly.
> >
> > It runs perf inject command internally and feeds the given number of
> > synthesized events (MMAP2 + SAMPLE basically).
> >
> >   Usage: perf bench internals inject-build-id <options>
> >
> >     -i, --iterations <n>  Number of iterations used to compute average (default: 100)
> >     -m, --nr-mmaps <n>    Number of mmap events for each iteration (default: 100)
> >     -n, --nr-samples <n>  Number of sample events per mmap event (default: 100)
> >     -v, --verbose         be more verbose (show iteration count, DSO name, etc)
> >
> > By default, it measures average processing time of 100 MMAP2 events
> > and 10000 SAMPLE events.  Below is a result on my laptop.
> >
> >   $ perf bench internals inject-build-id
> >   # Running 'internals/inject-build-id' benchmark:
> >     Average build-id injection took: 22.997 msec (+- 0.067 msec)
> >     Average time per event: 2.255 usec (+- 0.007 usec)
>
> This is great! Some suggestions below.

Thanks!

>
> > Signed-off-by: Namhyung Kim <namhyung@xxxxxxxxxx>
> > ---
[SNIP]
> > +
> > +static const char *const bench_usage[] = {
> > +       "perf bench internals inject-build-id <options>",
> > +       NULL
> > +};
> > +
>
> Perhaps a comment:
> /* Helper for collect_dso that adds the given file as a dso to
> dso_list if it contains a buildid. Stops after 4 such dsos.*/

Will add.. please see below.

>
> > +static int add_dso(const char *fpath, const struct stat *sb __maybe_unused,
> > +                  int typeflag, struct FTW *ftwbuf __maybe_unused)
> > +{
> > +       struct bench_dso *dso;
> > +       unsigned char build_id[BUILD_ID_SIZE];
> > +
> > +       if (typeflag == FTW_D || typeflag == FTW_SL) {
> > +               return 0;
> > +       }
> > +
> > +       if (filename__read_build_id(fpath, build_id, BUILD_ID_SIZE) < 0)
> > +               return 0;
> > +
> > +       dso = malloc(sizeof(*dso));
> > +       if (dso == NULL)
> > +               return -1;
> > +
> > +       dso->name = realpath(fpath, NULL);
> > +       if (dso->name == NULL) {
> > +               free(dso);
> > +               return -1;
> > +       }
> > +
> > +       dso->ino = nr_dsos++;
> > +       list_add(&dso->list, &dso_list);
> > +       pr_debug2("  Adding DSO: %s\n", fpath);
> > +
> > +       /* stop if we collected 4x DSOs than needed */
> > +       if ((unsigned)nr_dsos > 4 * nr_mmaps)
> > +               return 1;
> > +
> > +       return 0;
> > +}
> > +
> > +static void collect_dso(void)
> > +{
> > +       if (nftw("/usr/lib/", add_dso, 10, FTW_PHYS) < 0)
> > +               return;
> > +
> > +       pr_debug("  Collected %d DSOs\n", nr_dsos);
>
> Should this fail if the count isn't 4?

The add_dso would stop if it collected enough DSOs.
I chose it as 4 x nr_mmaps (default: 100).

It's gonna pick a DSO in the list randomly during benchmark
and I want to reduce the chance it selects the same one in the
same iteration. So instead of having nr_mmaps DSOs, it keeps
4 times more DSOs than needed.

>
> > +}
> > +
> > +static void release_dso(void)
> > +{
> > +       struct bench_dso *dso;
> > +
> > +       while (!list_empty(&dso_list)) {
> > +               dso = list_first_entry(&dso_list, struct bench_dso, list);
> > +               list_del(&dso->list);
> > +               free(dso->name);
> > +               free(dso);
> > +       }
> > +}
> > +
>
> Perhaps a comment and move next to synthesize_mmap.
> /* Fake address used by mmap events. */

OK, will do.  (and it's used by sample events too)

>
> > +static u64 dso_map_addr(struct bench_dso *dso)
> > +{
> > +       return 0x400000ULL + dso->ino * 8192ULL;
> > +}
[SNIP]

> > +static int setup_injection(struct bench_data *data)
> > +{
> > +       int ready_pipe[2];
> > +       int dev_null_fd;
> > +       char buf;
> > +
> > +       if (pipe(ready_pipe) < 0)
> > +               return -1;
> > +
> > +       if (pipe(data->input_pipe) < 0)
> > +               return -1;
> > +
> > +       if (pipe(data->output_pipe) < 0)
> > +               return -1;
> > +
> > +       data->pid = fork();
> > +       if (data->pid < 0)
> > +               return -1;
> > +
> > +       if (data->pid == 0) {
> > +               const char **inject_argv;
> > +
> > +               close(data->input_pipe[1]);
> > +               close(data->output_pipe[0]);
> > +               close(ready_pipe[0]);
> > +
> > +               dup2(data->input_pipe[0], STDIN_FILENO);
> > +               close(data->input_pipe[0]);
> > +               dup2(data->output_pipe[1], STDOUT_FILENO);
> > +               close(data->output_pipe[1]);
> > +
> > +               dev_null_fd = open("/dev/null", O_WRONLY);
> > +               if (dev_null_fd < 0)
> > +                       exit(1);
> > +
> > +               dup2(dev_null_fd, STDERR_FILENO);
> > +
> > +               inject_argv = calloc(3, sizeof(*inject_argv));
> > +               if (inject_argv == NULL)
> > +                       exit(1);
> > +
> > +               inject_argv[0] = strdup("inject");
> > +               inject_argv[1] = strdup("-b");
> > +
> > +               /* signal that we're ready to go */
> > +               close(ready_pipe[1]);
> > +
> > +               cmd_inject(2, inject_argv);
> > +
> > +               exit(0);
> > +       }
> > +
> > +       signal(SIGPIPE, sigpipe_handler);
> > +
> > +       close(ready_pipe[1]);
> > +       close(data->input_pipe[0]);
> > +       close(data->output_pipe[1]);
> > +
> > +       /* wait for child ready */
> > +       if (read(ready_pipe[0], &buf, 1) < 0)
> > +               return -1;
> > +       close(ready_pipe[0]);
> > +
> > +       return 0;
> > +}
>
> This feels like generic scaffolding that could be shared by other perf
> command benchmarks.

Maybe.. the thing is perf inject usually works on pipes so it needed
a new process to run the test.  Probably we can simply run others
in the same process.

>
> > +
> > +static int inject_build_id(struct bench_data *data)
> > +{
> > +       int flag, status;
> > +       unsigned int i, k;
> > +       char buf[8192];
> > +       u64 nread = 0;
> > +       u64 len = nr_mmaps / 2 * sizeof(struct perf_record_header_build_id);
> > +
> > +       flag = fcntl(data->output_pipe[0], F_GETFL, 0);
> > +       if (fcntl(data->output_pipe[0], F_SETFL, flag | O_NONBLOCK) < 0)
> > +               return -1;
> > +
> > +       /* this makes the child to run */
> > +       if (perf_header__write_pipe(data->input_pipe[1]) < 0)
> > +               return -1;
> > +
> > +       len += synthesize_attr(data);
> > +       len += synthesize_fork(data);
> > +
> > +       for (i = 0; i < nr_mmaps; i++) {
> > +               struct bench_dso *dso;
> > +               int idx = rand() % (nr_dsos - 1);
> > +
> > +               dso = list_first_entry(&dso_list, struct bench_dso, list);
> > +               while (idx--)
> > +                       dso = list_next_entry(dso, list);
> > +
> > +               pr_debug("   [%2d] injecting: %s\n", i+1, dso->name);
> > +               len += synthesize_mmap(data, dso);
> > +
> > +               for (k = 0; k < nr_samples; k++)
> > +                       len += synthesize_sample(data, dso);
> > +
> > +               /* read out data from child */
> > +               while (true) {
> > +                       int n;
> > +
> > +                       n = read(data->output_pipe[0], buf, sizeof(buf));
> > +                       if (n <= 0)
> > +                               break;
> > +                       nread += n;
> > +               }
> > +       }
> > +
> > +       /* wait to read data at least as we wrote + some build-ids */
> > +       while (nread < len) {
> > +               int n;
> > +
> > +               n = read(data->output_pipe[0], buf, sizeof(buf));
> > +               if (n < 0)
> > +                       break;
> > +               nread += n;
> > +       }
> > +       close(data->input_pipe[1]);
> > +       close(data->output_pipe[0]);
> > +
> > +       wait(&status);
> > +       pr_debug("   Child %d exited with %d\n", data->pid, status);
> > +
> > +       return 0;
> > +}
> >
>
> Perhaps we can read the highwater mark (VmHWM) from /proc/[pid]/status
> as this would capture cases like buildid injection doing unnecessary
> symbol generation.

Good idea!  I'll add it and check we can see the difference.

Thanks
Namhyung