Re: [PATCH 2/2] perf timechart: implement IO mode
From: Namhyung Kim
Date: Tue Jun 10 2014 - 03:50:44 EST
On Mon, 9 Jun 2014 13:21:12 +0400, Stanislav Fomichev wrote:
> TIMECHART OPTIONS
> -----------------
> @@ -54,6 +64,15 @@ TIMECHART OPTIONS
> duration or tasks with given name. If number is given it's interpreted
> as number of nanoseconds. If non-numeric string is given it's
> interpreted as task name.
> +--io-skip-eagain::
> + Don't draw EAGAIN IO events.
> +--io-min-time::
> + Draw small events as if they lasted min-time. Useful when you need
> + to see very small and fast IO. Default value is 1ms.
> +--io-merge-dist::
> + Merge events that are merge-dist nanoseconds apart.
> + Reduces number of figures on the SVG and makes it more render-friendly.
> + Default value is 1us.
Looks like these are advanced options. If so, I think it's better to
split them to a different changeset so that the main change remains
simpler, smaller and easier to review.
>
> EXAMPLES
> --------
> @@ -75,6 +94,14 @@ Record system-wide timechart:
>
> $ perf timechart --highlight gcc
>
> +Record system-wide IO events:
> +
> + $ perf timechart record -I
I think it's not working currently.. Even with massaging of the decl
shadow errors, I still wasn't able to run it due to some errors. Please
see below..
> +
> + then generate timechart:
> +
> + $ perf timechart
[SNIP]
> +static int timechart__io_record(int argc, const char **argv)
> +{
> + unsigned int rec_argc, i;
> + const char **rec_argv;
> + const char **p;
> + char *filter = NULL;
> +
> + const char * const common_args[] = {
> + "record", "-a", "-R", "-c", "1",
> + };
> + unsigned int common_args_nr = ARRAY_SIZE(common_args);
> +
> + const char * const disk_events[] = {
> + "-e", "syscalls:sys_enter_read",
I guess you need to get rid of all "-e" part in the arrays as you add it
automatically below, otherwise I saw following errors:
$ perf timechart record -I
invalid or unsupported event: '-e'
Run 'perf list' for a list of valid events
usage: perf record [<options>] [<command>]
or: perf record [<options>] -- <command> [<options>]
-e, --event <event> event selector. use 'perf list' to list available events
> + "-e", "syscalls:sys_enter_pread64",
And even with above change, it still failed to run since it seems my
system doesn't have the pread64 syscall:
$ perf timechart record -I
invalid or unsupported event: 'syscalls:sys_enter_pread64'
Run 'perf list' for a list of valid events
usage: perf record [<options>] [<command>]
or: perf record [<options>] -- <command> [<options>]
-e, --event <event> event selector. use 'perf list' to list available events
Then I stopped to test/review. I'll review if you send v2 with fixes.
Thanks,
Namhyung
> + "-e", "syscalls:sys_enter_readv",
> + "-e", "syscalls:sys_enter_preadv",
> + "-e", "syscalls:sys_enter_write",
> + "-e", "syscalls:sys_enter_pwrite64",
> + "-e", "syscalls:sys_enter_writev",
> + "-e", "syscalls:sys_enter_pwritev",
> + "-e", "syscalls:sys_enter_sync",
> + "-e", "syscalls:sys_enter_sync_file_range",
> + "-e", "syscalls:sys_enter_fsync",
> + "-e", "syscalls:sys_enter_msync",
> +
> + "-e", "syscalls:sys_exit_read",
> + "-e", "syscalls:sys_exit_pread64",
> + "-e", "syscalls:sys_exit_readv",
> + "-e", "syscalls:sys_exit_preadv",
> + "-e", "syscalls:sys_exit_write",
> + "-e", "syscalls:sys_exit_pwrite64",
> + "-e", "syscalls:sys_exit_writev",
> + "-e", "syscalls:sys_exit_pwritev",
> + "-e", "syscalls:sys_exit_sync",
> + "-e", "syscalls:sys_exit_sync_file_range",
> + "-e", "syscalls:sys_exit_fsync",
> + "-e", "syscalls:sys_exit_msync",
> + };
> + unsigned int disk_events_nr = ARRAY_SIZE(disk_events);
> +
> + const char * const net_events[] = {
> + "-e", "syscalls:sys_enter_recvfrom",
> + "-e", "syscalls:sys_enter_recvmmsg",
> + "-e", "syscalls:sys_enter_recvmsg",
> + "-e", "syscalls:sys_enter_sendto",
> + "-e", "syscalls:sys_enter_sendmsg",
> + "-e", "syscalls:sys_enter_sendmmsg",
> +
> + "-e", "syscalls:sys_exit_recvfrom",
> + "-e", "syscalls:sys_exit_recvmmsg",
> + "-e", "syscalls:sys_exit_recvmsg",
> + "-e", "syscalls:sys_exit_sendto",
> + "-e", "syscalls:sys_exit_sendmsg",
> + "-e", "syscalls:sys_exit_sendmmsg",
> + };
> + unsigned int net_events_nr = ARRAY_SIZE(net_events);
> +
> + const char * const poll_events[] = {
> + "-e", "syscalls:sys_enter_epoll_pwait",
> + "-e", "syscalls:sys_enter_epoll_wait",
> + "-e", "syscalls:sys_enter_poll",
> + "-e", "syscalls:sys_enter_ppoll",
> + "-e", "syscalls:sys_enter_pselect6",
> + "-e", "syscalls:sys_enter_select",
> +
> + "-e", "syscalls:sys_exit_epoll_pwait",
> + "-e", "syscalls:sys_exit_epoll_wait",
> + "-e", "syscalls:sys_exit_poll",
> + "-e", "syscalls:sys_exit_ppoll",
> + "-e", "syscalls:sys_exit_pselect6",
> + "-e", "syscalls:sys_exit_select",
> + };
> + unsigned int poll_events_nr = ARRAY_SIZE(poll_events);
> +
> + rec_argc = common_args_nr +
> + disk_events_nr * 4 +
> + net_events_nr * 4 +
> + poll_events_nr * 4 +
> + argc + 1;
> + rec_argv = calloc(rec_argc + 1, sizeof(char *));
> +
> + if (rec_argv == NULL)
> + return -ENOMEM;
> +
> + if (asprintf(&filter, "common_pid != %d", getpid()) < 0)
> + return -ENOMEM;
> +
> + p = rec_argv;
> + for (i = 0; i < common_args_nr; i++)
> + *p++ = strdup(common_args[i]);
> + for (i = 0; i < disk_events_nr; i++) {
> + *p++ = "-e";
> + *p++ = strdup(disk_events[i]);
> + *p++ = "--filter";
> + *p++ = filter;
> + }
> + for (i = 0; i < net_events_nr; i++) {
> + *p++ = "-e";
> + *p++ = strdup(net_events[i]);
> + *p++ = "--filter";
> + *p++ = filter;
> + }
> + for (i = 0; i < poll_events_nr; i++) {
> + *p++ = "-e";
> + *p++ = strdup(poll_events[i]);
> + *p++ = "--filter";
> + *p++ = filter;
> + }
> +
> + for (i = 0; i < (unsigned int)argc; i++)
> + *p++ = argv[i];
> +
> + return cmd_record(rec_argc, rec_argv, NULL);
> +}
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/