linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Namhyung Kim <namhyung@kernel.org>
To: Jiri Olsa <jolsa@kernel.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>,
	lkml <linux-kernel@vger.kernel.org>,
	Peter Zijlstra <a.p.zijlstra@chello.nl>,
	Ingo Molnar <mingo@kernel.org>,
	Mark Rutland <mark.rutland@arm.com>,
	Alexander Shishkin <alexander.shishkin@linux.intel.com>,
	Michael Petlan <mpetlan@redhat.com>,
	Ian Rogers <irogers@google.com>,
	Stephane Eranian <eranian@google.com>,
	Alexei Budankov <abudankov@huawei.com>
Subject: Re: [PATCH 07/22] perf daemon: Add daemon command
Date: Tue, 19 Jan 2021 13:08:17 +0900	[thread overview]
Message-ID: <CAM9d7ciSkns0=0LqWK2BrTFOON2z4wx5vjb=zhP5=iUoXXjeDQ@mail.gmail.com> (raw)
In-Reply-To: <20210102220441.794923-8-jolsa@kernel.org>

Hi Jiri,

On Sun, Jan 3, 2021 at 7:05 AM Jiri Olsa <jolsa@kernel.org> wrote:
>
> Adding daemon command that allows to run record sessions
> on background. Each session represents one perf record
> process and is configured in config file.
>
> Example:
>
>   # cat ~/.perfconfig
>   [daemon]
>   base=/opt/perfdata
>
>   [session-cycles]
>   run = -m 10M -e cycles --overwrite --switch-output -a
>
>   [session-sched]
>   run = -m 20M -e sched:* --overwrite --switch-output -a
>
> Starting the daemon:
>
>   # perf daemon start
>
> Check sessions:
>
>   # perf daemon
>   [771394:daemon] base: /opt/perfdata
>   [771395:cycles] perf record -m 10M -e cycles --overwrite --switch-output -a
>   [771396:sched] perf record -m 20M -e sched:* --overwrite --switch-output -a
>
> Check sessions with more info:
>
>   # perf daemon -v
>   [771394:daemon] base: /opt/perfdata
>     output:  /opt/perfdata/output
>   [771395:cycles] perf record -m 10M -e cycles --overwrite --switch-output -a
>     base:    /opt/perfdata/session-cycles
>     output:  /opt/perfdata/session-cycles/output
>   [771396:sched] perf record -m 20M -e sched:* --overwrite --switch-output -a
>     base:    /opt/perfdata/session-sched
>     output:  /opt/perfdata/session-sched/output
>
> The 'output' file is perf record output for specific session.
>
> Note you have to stop all running perf processes manually at
> this point, stop command is coming in following patches.
>
> Adding empty perf-daemon.txt to skip compile warning,
> the man page is populated in following patch.
>
> Signed-off-by: Jiri Olsa <jolsa@kernel.org>
> ---
[SNIP]
> +static int session_config(struct daemon *daemon, const char *var, const char *value)
> +{
> +       struct session *session;
> +       char name[100];
> +
> +       if (get_session_name(var, name, sizeof(name)))
> +               return -EINVAL;
> +
> +       var = strchr(var, '.');
> +       if (!var)
> +               return -EINVAL;
> +
> +       var++;
> +
> +       session = daemon__find_session(daemon, name);
> +       if (!session) {
> +               session = daemon__add_session(daemon, name);
> +               if (!session)
> +                       return -ENOMEM;
> +
> +               pr_debug("reconfig: found new session %s\n", name);
> +               /* This is new session, trigger reconfig to start it. */
> +               session->state = SESSION_STATE__RECONFIG;
> +       } else if (session->state == SESSION_STATE__KILL) {
> +               /*
> +                * The session was marked to kill and we still
> +                * found it in config file.
> +                */
> +               pr_debug("reconfig: found current session %s\n", name);
> +               session->state = SESSION_STATE__OK;
> +       }
> +
> +       if (!strcmp(var, "run")) {
> +               if (session->run && strcmp(session->run, value)) {
> +                       free(session->run);
> +                       pr_debug("reconfig: session %s is changed\n", name);
> +                       session->state = SESSION_STATE__RECONFIG;
> +               }
> +               session->run = strdup(value);

Please check the result.

> +       }
> +
> +       return 0;
> +}
> +
> +static int server_config(const char *var, const char *value, void *cb)
> +{
> +       struct daemon *daemon = cb;
> +
> +       if (strstarts(var, "session-"))
> +               return session_config(daemon, var, value);
> +       else if (!strcmp(var, "daemon.base"))
> +               daemon->base = strdup(value);

It seems these config items are mandatory.  Is there a check
for their presence?


> +
> +       return 0;
> +}
> +
> +static int client_config(const char *var, const char *value, void *cb)
> +{
> +       struct daemon *daemon = cb;
> +
> +       if (!strcmp(var, "daemon.base"))
> +               daemon->base = strdup(value);
> +
> +       return 0;
> +}
> +
> +static int setup_client_config(struct daemon *daemon)
> +{
> +       struct perf_config_set *set;
> +       int err = -ENOMEM;
> +
> +       set = perf_config_set__load_file(daemon->config_real);
> +       if (set) {
> +               err = perf_config_set(set, client_config, daemon);
> +               perf_config_set__delete(set);
> +       }
> +
> +       return err;
> +}
> +
> +static int setup_server_config(struct daemon *daemon)
> +{
> +       struct perf_config_set *set;
> +       struct session *session;
> +       int err = -ENOMEM;
> +
> +       pr_debug("reconfig: started\n");
> +
> +       /*
> +        * Mark all session for kill, the server config will
> +        * set proper state for found sessions.
> +        */
> +       list_for_each_entry(session, &daemon->sessions, list)
> +               session->state = SESSION_STATE__KILL;

Probably we can put them in a different state like INIT or READY?

> +
> +       set = perf_config_set__load_file(daemon->config_real);
> +       if (set) {
> +               err = perf_config_set(set, server_config, daemon);
> +               perf_config_set__delete(set);
> +       }
> +
> +       return err;
> +}
> +
> +static int session__signal(struct session *session, int sig)
> +{
> +       if (session->pid < 0)
> +               return -1;
> +       return kill(session->pid, sig);
> +}
> +
> +static int session__run(struct session *session, struct daemon *daemon)
> +{
> +       char buf[PATH_MAX];
> +       char **argv;
> +       int argc, fd;
> +
> +       if (asprintf(&session->base, "%s/session-%s",
> +                    daemon->base, session->name) < 0) {
> +               perror("asprintf failed");
> +               return -1;
> +       }
> +
> +       if (mkdir(session->base, 0755) && errno != EEXIST) {
> +               perror("mkdir failed");
> +               return -1;
> +       }
> +
> +       session->pid = fork();
> +       if (session->pid < 0)
> +               return -1;
> +       if (session->pid > 0) {
> +               pr_info("reconfig: ruining session [%s:%d]: %s\n",
> +                       session->name, session->pid, session->run);
> +               return 0;
> +       }
> +
> +       if (chdir(session->base)) {
> +               perror("chdir failed");
> +               return -1;
> +       }
> +
> +       fd = open("/dev/null", O_RDONLY);
> +       if (fd < 0) {
> +               perror("failed to open /dev/null");
> +               return -1;
> +       }
> +
> +       close(0);
> +       dup2(fd, 0);

The man page says dup2() will close the second file descriptor.

> +       close(fd);
> +
> +       fd = open(SESSION_OUTPUT, O_RDWR|O_CREAT|O_TRUNC, 0644);
> +       if (fd < 0) {
> +               perror("failed to open session output");
> +               return -1;
> +       }
> +
> +       close(1);
> +       close(2);
> +       dup2(fd, 1);
> +       dup2(fd, 2);

Ditto.

> +       close(fd);
> +
> +       scnprintf(buf, sizeof(buf), "%s record %s", daemon->perf, session->run);
> +
> +       argv = argv_split(buf, &argc);
> +       if (!argv)
> +               exit(-1);
> +
> +       exit(execve(daemon->perf, argv, NULL));
> +       return -1;
> +}

[SNIP]
> +static int daemon__reconfig(struct daemon *daemon)
> +{
> +       struct session *session, *n;
> +
> +       list_for_each_entry_safe(session, n, &daemon->sessions, list) {
> +               /* No change. */
> +               if (session->state == SESSION_STATE__OK)
> +                       continue;
> +
> +               /* Remove session. */
> +               if (session->state == SESSION_STATE__KILL) {
> +                       if (session->pid > 0) {
> +                               session__kill(session);
> +                               pr_info("reconfig: session '%s' killed\n", session->name);
> +                       }
> +                       session__remove(session);
> +                       continue;
> +               }
> +
> +               /* Reconfig session. */
> +               pr_debug2("reconfig: session '%s' start\n", session->name);
> +               if (session->pid > 0) {
> +                       session__kill(session);
> +                       pr_info("reconfig: session '%s' killed\n", session->name);
> +               }
> +               if (session__run(session, daemon))

Does it call a config function?  Or is it called already?

> +                       return -1;
> +               pr_debug2("reconfig: session '%s' done\n", session->name);
> +               session->state = SESSION_STATE__OK;

I think RUNNING is a better name.

Thanks,
Namhyung


> +       }
> +
> +       return 0;
> +}
> +

  reply	other threads:[~2021-01-19  4:10 UTC|newest]

Thread overview: 44+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-01-02 22:04 [PATCHv2 00/22] perf tools: Add daemon command Jiri Olsa
2021-01-02 22:04 ` [PATCH 01/22] perf tools: Make perf_config_from_file static Jiri Olsa
2021-01-18 15:51   ` Arnaldo Carvalho de Melo
2021-01-02 22:04 ` [PATCH 02/22] perf tools: Add config set interface Jiri Olsa
2021-01-18 15:53   ` Arnaldo Carvalho de Melo
2021-01-02 22:04 ` [PATCH 03/22] perf tools: Add debug_set_display_time function Jiri Olsa
2021-01-18 16:02   ` Arnaldo Carvalho de Melo
2021-01-19 14:59   ` Arnaldo Carvalho de Melo
2021-01-19 17:39     ` Jiri Olsa
2021-01-19 19:42       ` Arnaldo Carvalho de Melo
2021-01-02 22:04 ` [PATCH 04/22] perf tools: Add perf_home_perfconfig function Jiri Olsa
2021-01-18 16:05   ` Arnaldo Carvalho de Melo
2021-01-02 22:04 ` [PATCH 05/22] perf tools: Make perf_config_system global Jiri Olsa
2021-01-18 16:03   ` Arnaldo Carvalho de Melo
2021-01-02 22:04 ` [PATCH 06/22] perf tools: Make perf_config_global gobal Jiri Olsa
2021-01-18 16:05   ` Arnaldo Carvalho de Melo
2021-01-02 22:04 ` [PATCH 07/22] perf daemon: Add daemon command Jiri Olsa
2021-01-19  4:08   ` Namhyung Kim [this message]
2021-01-19 18:31     ` Jiri Olsa
2021-01-21  4:53       ` Namhyung Kim
2021-01-27  7:09   ` Namhyung Kim
2021-01-27 22:01     ` Jiri Olsa
2021-01-02 22:04 ` [PATCH 08/22] perf daemon: Add config file change check Jiri Olsa
2021-01-19  5:31   ` Namhyung Kim
2021-01-19 17:49     ` Jiri Olsa
2021-01-21  4:54       ` Namhyung Kim
2021-01-02 22:04 ` [PATCH 09/22] perf daemon: Add signalfd support Jiri Olsa
2021-01-02 22:04 ` [PATCH 10/22] perf daemon: Add signal command Jiri Olsa
2021-01-02 22:04 ` [PATCH 11/22] perf daemon: Add stop command Jiri Olsa
2021-01-19  5:35   ` Namhyung Kim
2021-01-02 22:04 ` [PATCH 12/22] perf daemon: Allow only one daemon over base directory Jiri Olsa
2021-01-19  5:37   ` Namhyung Kim
2021-01-19 17:44     ` Jiri Olsa
2021-01-02 22:04 ` [PATCH 13/22] perf daemon: Set control fifo for session Jiri Olsa
2021-01-02 22:04 ` [PATCH 14/22] perf daemon: Add ping command Jiri Olsa
2021-01-02 22:04 ` [PATCH 15/22] perf daemon: Use control to stop session Jiri Olsa
2021-01-02 22:04 ` [PATCH 16/22] perf daemon: Add up time for daemon/session list Jiri Olsa
2021-01-02 22:04 ` [PATCH 17/22] perf daemon: Add man page for perf-daemon Jiri Olsa
2021-01-02 22:04 ` [PATCH 18/22] perf test: Add daemon list command test Jiri Olsa
2021-01-02 22:04 ` [PATCH 19/22] perf test: Add daemon reconfig test Jiri Olsa
2021-01-02 22:04 ` [PATCH 20/22] perf test: Add daemon stop command test Jiri Olsa
2021-01-02 22:04 ` [PATCH 21/22] perf test: Add daemon signal " Jiri Olsa
2021-01-02 22:04 ` [PATCH 22/22] perf test: Add daemon ping " Jiri Olsa
2021-01-18 12:55 ` [PATCHv2 00/22] perf tools: Add daemon command Jiri Olsa

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='CAM9d7ciSkns0=0LqWK2BrTFOON2z4wx5vjb=zhP5=iUoXXjeDQ@mail.gmail.com' \
    --to=namhyung@kernel.org \
    --cc=a.p.zijlstra@chello.nl \
    --cc=abudankov@huawei.com \
    --cc=acme@kernel.org \
    --cc=alexander.shishkin@linux.intel.com \
    --cc=eranian@google.com \
    --cc=irogers@google.com \
    --cc=jolsa@kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=mark.rutland@arm.com \
    --cc=mingo@kernel.org \
    --cc=mpetlan@redhat.com \
    /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).