Re: [PATCH] tst: Provide test for getrusage

[Adhemerval Zanella <adhemerval.zanella@linaro.org>]

On 15/03/2021 07:42, Lukasz Majewski wrote:
> This commit adds test to check if getrusage provides correct time
> accounting for some busy wait loop.
> 
> To be more specific - the time is measured in a loop by wrappers
> from glibc's libsupport and checked if timeout has passed.
> 
> Afterwards, the data from struct rusage, namely ru_stime and ru_utime
> is summed up and compared with time from busy waiting.
> ---
>   sysdeps/unix/sysv/linux/Makefile        |  2 +-
>   sysdeps/unix/sysv/linux/tst-getrusage.c | 55 +++++++++++++++++++++++++
>   2 files changed, 56 insertions(+), 1 deletion(-)
>   create mode 100644 sysdeps/unix/sysv/linux/tst-getrusage.c
> 
> diff --git a/sysdeps/unix/sysv/linux/Makefile b/sysdeps/unix/sysv/linux/Makefile
> index 6a45ac820a..a93fa4ca50 100644
> --- a/sysdeps/unix/sysv/linux/Makefile
> +++ b/sysdeps/unix/sysv/linux/Makefile
> @@ -113,7 +113,7 @@ tests += tst-clone tst-clone2 tst-clone3 tst-fanotify tst-personality \
>   	 tst-tgkill tst-sysvsem-linux tst-sysvmsg-linux tst-sysvshm-linux \
>   	 tst-timerfd tst-ppoll tst-futimens tst-utime tst-utimes \
>   	 tst-clock_adjtime tst-adjtimex tst-ntp_adjtime tst-futimes \
> -	 tst-lutimes tst-ntp_gettime tst-ntp_gettimex
> +	 tst-lutimes tst-ntp_gettime tst-ntp_gettimex tst-getrusage
>   tests-internal += tst-ofdlocks-compat tst-sigcontext-get_pc
>   tests-time64 += \
>     tst-futimes-time64\
> diff --git a/sysdeps/unix/sysv/linux/tst-getrusage.c b/sysdeps/unix/sysv/linux/tst-getrusage.c
> new file mode 100644
> index 0000000000..068becdadb
> --- /dev/null
> +++ b/sysdeps/unix/sysv/linux/tst-getrusage.c
> @@ -0,0 +1,55 @@
> +/* Test for getrusage
> +   Copyright (C) 2021 Free Software Foundation, Inc.
> +   This file is part of the GNU C Library.
> +
> +   The GNU C Library is free software; you can redistribute it and/or
> +   modify it under the terms of the GNU Lesser General Public
> +   License as published by the Free Software Foundation; either
> +   version 2.1 of the License, or (at your option) any later version.
> +
> +   The GNU C Library is distributed in the hope that it will be useful,
> +   but WITHOUT ANY WARRANTY; without even the implied warranty of
> +   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
> +   Lesser General Public License for more details.
> +
> +   You should have received a copy of the GNU Lesser General Public
> +   License along with the GNU C Library; if not, see
> +   <https://www.gnu.org/licenses/>.  */
> +
> +#include <time.h>
> +#include <sys/resource.h>
> +#include <support/check.h>
> +#include <support/timespec.h>
> +
> +#define TST_GETRUSAGE_TIMEOUT 1
> +
> +static int
> +do_test (void)
> +{
> +  struct timespec tv_then, tv_now, gru_u, gru_s, s, r;
> +  struct rusage ru;
> +  int ret;
> +
> +  tv_then = xclock_now (CLOCK_REALTIME);
> +  /* Busy wait for 1 second.  */
> +  do
> +  {
> +    tv_now = xclock_now (CLOCK_REALTIME);
> +    r = timespec_sub (tv_now, tv_then);
> +  } while ( r.tv_sec != TST_GETRUSAGE_TIMEOUT );
> +
> +  ret = getrusage (RUSAGE_SELF, &ru);
> +  if (ret == -1)
> +    FAIL_EXIT1 ("getrusage failed: %m\n");
> +
> +  gru_u = make_timespec (ru.ru_utime.tv_sec, ru.ru_utime.tv_usec * 1000);
> +  gru_s = make_timespec (ru.ru_stime.tv_sec, ru.ru_stime.tv_usec * 1000);
> +  s = timespec_add (gru_u, gru_s);
> +
> +  TEST_COMPARE (support_timespec_check_in_range
> +                ((struct timespec) { 1, 0 }, s, 0.8, 1.2), 1);
> +
> +  return 0;
> +}
> +
> +#include <support/test-driver.c>
> 

Timer accounting tests are tricky to get right and prone to show a
lot of false positive failures and has bitten us in the past (check
04deeaa9ea74 and 9a29f1a2ae3d4b).

So I think we should be as simple as possible and take in consideration
that tests will run a timer shared machine so even with a busy wait the
result timer might not take in consideration the desirable timeout.

So for geturage (which afaik is also deprecated on future POSIX) I
more simpler tests would be to:

   1. Check for invalid 'who' (the EFAULT result from man-pages only
      trigger where syscall is issued directly, which is not always
      the case for glibc).

   2. Issue getusage and check if the ru_utime.tv_nsec and
      ru_stime.tv_nsec are within the microseconds range.

   3. Do some very short busy wait (10ms) and check if the
      ru_utime.tv_sec and ru_stime.tv_sec are equal or larger
      than the previous one.

Also, I think this tests should generic, so move to resource
folder.