From: Thomas Gleixner <tglx@kernel.org>
To: Wake Liu <wakel@google.com>, Shuah Khan <shuah@kernel.org>
Cc: linux-kselftest@vger.kernel.org, linux-kernel@vger.kernel.org,
Wake Liu <wakel@google.com>
Subject: Re: [PATCH] selftests: timers: Use CPU time for CPU timers in posix_timers test
Date: Mon, 20 Jul 2026 16:14:52 +0200 [thread overview]
Message-ID: <87fr1dn3xf.ffs@fw13> (raw)
In-Reply-To: <20260708121544.583552-1-wakel@google.com>
On Wed, Jul 08 2026 at 12:15, Wake Liu wrote:
> The posix_timers test compares the elapsed real time (GTOD) with the
> expected timer delay to verify if the timer expired. This works fine for
> real-time timers (ITIMER_REAL), but is flaky for CPU-time timers
> (ITIMER_PROF, ITIMER_VIRTUAL, and CLOCK_*_CPUTIME_ID) on loaded systems
> (like shared test runners) because the test process might be scheduled
> out, resulting in elapsed real time being larger than the consumed CPU
> time.
>
> Fix this by using the appropriate measurement method for each timer:
> - Use CPU process/thread time (via clock_gettime or getrusage) for CPU
> timers.
> - Keep using monotonic clock (real time) for real-time timers.
>
> This makes the test robust against scheduling delays.
Lacks Signed-off-by ....
> +#include <sys/resource.h>
> +
> +static long long get_clock_time_us(clockid_t clock_id)
> +{
> + struct timespec ts;
Newline between declaration and code.
> + if (clock_gettime(clock_id, &ts) < 0)
> + return -1;
> + return (long long)ts.tv_sec * USECS_PER_SEC + ts.tv_nsec / 1000;
> +}
> +
> +static long long get_itimer_time_us(int which)
> +{
> + struct rusage usage;
> +
> + switch (which) {
> + case ITIMER_REAL:
> + return get_clock_time_us(CLOCK_MONOTONIC);
> + case ITIMER_PROF:
> + return get_clock_time_us(CLOCK_PROCESS_CPUTIME_ID);
> + case ITIMER_VIRTUAL:
> + if (getrusage(RUSAGE_SELF, &usage) < 0)
> + return -1;
> + return (long long)usage.ru_utime.tv_sec * USECS_PER_SEC + usage.ru_utime.tv_usec;
> + default:
> + return -1;
> + }
> +}
> +
> +static int check_diff_us(long long start, long long end)
> {
> long long diff;
>
> - diff = end.tv_usec - start.tv_usec;
> - diff += (end.tv_sec - start.tv_sec) * USECS_PER_SEC;
> + diff = end - start;
Move that to the declaration line.
> if (llabs(diff - DELAY * USECS_PER_SEC) > USECS_PER_SEC / 2) {
> printf("Diff too high: %lld..", diff);
> @@ -100,7 +123,7 @@ static int check_diff(struct timeval start, struct timeval end)
>
> - if (gettimeofday(&end, NULL) < 0)
> - fatal_error(name, "gettimeofday()");
> + end = get_clock_time_us(which);
> + if (end < 0)
> + fatal_error(name, "get_clock_time_us()");
>
> - ksft_test_result(check_diff(start, end) == 0,
> + ksft_test_result(check_diff_us(start, end) == 0,
> "timer_create() per %s\n", name);
Please get rid of the line break.
next prev parent reply other threads:[~2026-07-20 14:14 UTC|newest]
Thread overview: 4+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-07-08 12:15 [PATCH] selftests: timers: Use CPU time for CPU timers in posix_timers test Wake Liu
2026-07-20 14:14 ` Thomas Gleixner [this message]
2026-07-21 10:37 ` [PATCH v2] " Wake Liu
2026-07-21 10:38 ` [PATCH] " Wake Liu
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=87fr1dn3xf.ffs@fw13 \
--to=tglx@kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-kselftest@vger.kernel.org \
--cc=shuah@kernel.org \
--cc=wakel@google.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