From: Tvrtko Ursulin <tursulin@ursulin.net>
To: Lucas De Marchi <lucas.demarchi@intel.com>,
igt-dev@lists.freedesktop.org
Cc: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
Subject: Re: [PATCH i-g-t 8/8] lib/igt_drm_clients: lazy stat process
Date: Wed, 3 Apr 2024 17:24:52 +0100 [thread overview]
Message-ID: <aaab8699-399e-4d1d-8922-1e303eef103e@ursulin.net> (raw)
In-Reply-To: <20240402221716.1840148-9-lucas.demarchi@intel.com>
On 02/04/2024 23:17, Lucas De Marchi wrote:
> Only try to get task data like pid and name when that is required.
> Typically most of the processes in a system don't have a drm fd open.
> So only open and use /proc/<pid>/stat when it's needed.
This sounds like a very good optimisation to me too and I also don't
understand how it could be making things less efficient (as you say in
the cover letter). Unfortunately I will not have time to cross check in
the next week or so.
Regards,
Tvrtko
> Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
> ---
> lib/igt_drm_clients.c | 50 +++++++++++++++++++++----------------------
> 1 file changed, 25 insertions(+), 25 deletions(-)
>
> diff --git a/lib/igt_drm_clients.c b/lib/igt_drm_clients.c
> index c174c96ab..1e9781c68 100644
> --- a/lib/igt_drm_clients.c
> +++ b/lib/igt_drm_clients.c
> @@ -367,26 +367,33 @@ static size_t readat2buf(int at, const char *name, char *buf, const size_t sz)
> }
> }
>
> -static bool get_task_name(const char *buffer, char *out, unsigned long sz)
> +static void get_task_data(int pid_dir, unsigned int *pid, char *task, size_t tasksz)
> {
> - char *s = index(buffer, '(');
> - char *e = rindex(buffer, ')');
> - unsigned int len;
> + char buf[4096];
> + char *s, *e;
> + size_t len;
>
> + if (!readat2buf(pid_dir, "stat", buf, sizeof(buf)))
> + return;
> +
> + s = strchr(buf, '(');
> + e = strchr(s, ')');
> if (!s || !e)
> - return false;
> - assert(e >= s);
> + return;
>
> len = e - ++s;
> - if(!len || (len + 1) >= sz)
> - return false;
> + if (!len)
> + return;
>
> - strncpy(out, s, len);
> - out[len] = 0;
> + if (len + 1 > tasksz)
> + len = tasksz - 1;
>
> - return true;
> + strncpy(task, s, len);
> + task[len] = 0;
> + *pid = atoi(buf);
> }
>
> +
> static bool is_drm_fd(int fd_dir, const char *name, unsigned int *minor)
> {
> struct stat stat;
> @@ -480,13 +487,11 @@ igt_drm_clients_scan(struct igt_drm_clients *clients,
> return clients;
>
> while ((proc_dent = readdir(proc_dir)) != NULL) {
> - unsigned int client_pid, minor = 0;
> + unsigned int client_pid = 0, minor = 0;
> int pid_dir = -1, fd_dir = -1;
> struct dirent *fdinfo_dent;
> char client_name[64] = { };
> DIR *fdinfo_dir = NULL;
> - char buf[4096];
> - size_t count;
>
> if (proc_dent->d_type != DT_DIR)
> continue;
> @@ -498,17 +503,6 @@ igt_drm_clients_scan(struct igt_drm_clients *clients,
> if (pid_dir < 0)
> continue;
>
> - count = readat2buf(pid_dir, "stat", buf, sizeof(buf));
> - if (!count)
> - goto next;
> -
> - client_pid = atoi(buf);
> - if (!client_pid)
> - goto next;
> -
> - if (!get_task_name(buf, client_name, sizeof(client_name)))
> - goto next;
> -
> fd_dir = openat(pid_dir, "fd", O_DIRECTORY | O_RDONLY);
> if (fd_dir < 0)
> goto next;
> @@ -541,6 +535,12 @@ igt_drm_clients_scan(struct igt_drm_clients *clients,
> minor, info.id))
> continue; /* Skip duplicate fds. */
>
> + if (!client_pid) {
> + get_task_data(pid_dir, &client_pid, client_name,
> + sizeof(client_name));
> + assert(client_pid > 0);
> + }
> +
> c = igt_drm_clients_find(clients, IGT_DRM_CLIENT_PROBE,
> minor, info.id);
> if (!c)
next prev parent reply other threads:[~2024-04-03 16:24 UTC|newest]
Thread overview: 25+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-04-02 22:17 [PATCH i-g-t 0/8] Refactors and fixes for drm_clients Lucas De Marchi
2024-04-02 22:17 ` [PATCH i-g-t 1/8] lib/igt_drm_clients: Use calloc Lucas De Marchi
2024-04-03 15:27 ` Umesh Nerlige Ramappa
2024-04-03 17:23 ` Kamil Konieczny
2024-04-02 22:17 ` [PATCH i-g-t 2/8] lib/igt_drm_clients: Fix sizeof calculation Lucas De Marchi
2024-04-03 15:28 ` Umesh Nerlige Ramappa
2024-04-03 17:25 ` Kamil Konieczny
2024-04-02 22:17 ` [PATCH i-g-t 3/8] lib/igt_drm_clients: Fix leaks Lucas De Marchi
2024-04-03 15:39 ` Umesh Nerlige Ramappa
2024-04-03 16:09 ` Tvrtko Ursulin
2024-04-03 17:26 ` Kamil Konieczny
2024-04-02 22:17 ` [PATCH i-g-t 4/8] gputop: Free clients on exit Lucas De Marchi
2024-04-03 16:03 ` Umesh Nerlige Ramappa
2024-04-03 16:13 ` Tvrtko Ursulin
2024-04-03 17:28 ` Kamil Konieczny
2024-04-02 22:17 ` [PATCH i-g-t 5/8] lib/igt_drm_fdinfo: Simplify find_kv() Lucas De Marchi
2024-04-03 17:48 ` Kamil Konieczny
2024-04-02 22:17 ` [PATCH i-g-t 6/8] lib/igt_drm_fdinfo: Stop passing key twice Lucas De Marchi
2024-04-03 17:56 ` Kamil Konieczny
2024-04-02 22:17 ` [PATCH i-g-t 7/8] lib/igt_drm_fdinfo: Remove prefix arg from parse functions Lucas De Marchi
2024-04-03 23:53 ` Umesh Nerlige Ramappa
2024-04-02 22:17 ` [PATCH i-g-t 8/8] lib/igt_drm_clients: lazy stat process Lucas De Marchi
2024-04-03 16:24 ` Tvrtko Ursulin [this message]
2024-04-02 23:59 ` ✗ Fi.CI.BAT: failure for Refactors and fixes for drm_clients Patchwork
2024-04-03 0:31 ` ✓ CI.xeBAT: success " Patchwork
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=aaab8699-399e-4d1d-8922-1e303eef103e@ursulin.net \
--to=tursulin@ursulin.net \
--cc=igt-dev@lists.freedesktop.org \
--cc=lucas.demarchi@intel.com \
--cc=umesh.nerlige.ramappa@intel.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