Re: [LTP] [PATCH] timers: support clockid of CLOCK_REALTIME_COARSE/CLOCK_MONOTONIC_COARSE

[Garrett C. <yan...@gm...>]

On Sun, Feb 21, 2010 at 11:55 PM, Shi Weihua <sh...@cn...> wrote:
> at 2010-2-18 1:14, Garrett Cooper wrote:
>> On Tue, Feb 16, 2010 at 11:46 PM, Rishikesh <ris...@li...> wrote:
>>> On 02/10/2010 03:16 PM, Shi Weihua wrote:
>>>> mainline commit da15cfdae03351c689736f8d142618592e3cebc3
>>>>   (subject: time: Introduce CLOCK_REALTIME_COARSE)
>>>> added two clocks (CLOCK_REALTIME_COARSE/CLOCK_MONOTONIC_COARSE) in
>>>> 2.6.32-rc1, but it has not been responded in ltp. then, the following
>>>> error message occured.
>>>> -----
>>>> clock_gettime03    3  TFAIL  :  failed to produce expected error [expected errno = 22 (Invalid argument), TEST_RETURN = 0]: TEST_ERRNO=???(0): Success
>>>> clock_gettime03    4  TFAIL  :  failed to produce expected error [expected errno = 22 (Invalid argument), TEST_RETURN = 0]: TEST_ERRNO=???(0): Success
>>>> timer_create04    1  TFAIL  :  didn't fail as expected [expected errno = 22 (Invalid argument)]: TEST_ERRNO=???(95): Operation not supported
>>>> timer_create04    2  TFAIL  :  didn't fail as expected [expected errno = 22 (Invalid argument)]: TEST_ERRNO=???(95): Operation not supported
>>>> -----
>>>>
>>>> Following patch fixed this bug. and, passed on 2.6.32-rc1, 2.6.33-rc6, 2.6.23.1-42.fc8.
>>>>
>>>
>>> Can anyone provide their review & Acked ?
>>>
>>> Thanks "Shi Weihua" for patch.
>>
>> 1. Please provide before and after (once the items were fixed) output.
>
> before:
> ------
> clock_gettime03    1  TPASS  :  got expected failure: TEST_ERRNO=EFAULT(14): Bad address
> clock_gettime03    2  TPASS  :  got expected failure: TEST_ERRNO=EFAULT(14): Bad address
> clock_gettime03    3  TFAIL  :  failed to produce expected error [expected errno = 22 (Invalid argument), TEST_RETURN = 0]: TEST_ERRNO=???(0): Success
> clock_gettime03    4  TFAIL  :  failed to produce expected error [expected errno = 22 (Invalid argument), TEST_RETURN = 0]: TEST_ERRNO=???(0): Success
> clock_gettime03    5  TPASS  :  got expected failure: TEST_ERRNO=EFAULT(14): Bad address
> clock_gettime03    6  TPASS  :  got expected failure: TEST_ERRNO=EFAULT(14): Bad address
> <snip>
> timer_create04    1  TFAIL  :  didn't fail as expected [expected errno = 22 (Invalid argument)]: TEST_ERRNO=???(95): Operation not supported
> timer_create04    2  TFAIL  :  didn't fail as expected [expected errno = 22 (Invalid argument)]: TEST_ERRNO=???(95): Operation not supported
> timer_create04    3  TPASS  :  failed as expected: TEST_ERRNO=EFAULT(14): Bad address
> timer_create04    4  TPASS  :  failed as expected: TEST_ERRNO=EFAULT(14): Bad address
> timer_create04    5  TPASS  :  failed as expected: TEST_ERRNO=EFAULT(14): Bad address
> timer_create04    6  TPASS  :  failed as expected: TEST_ERRNO=EFAULT(14): Bad address
> ------
>
> after ( patch merged )
> ------
> clock_gettime03    1  TPASS  :  got expected failure: TEST_ERRNO=EFAULT(14): Bad address
> clock_gettime03    2  TPASS  :  got expected failure: TEST_ERRNO=EFAULT(14): Bad address
> clock_gettime03    3  TPASS  :  got expected failure: TEST_ERRNO=EINVAL(22): Invalid argument
> clock_gettime03    4  TPASS  :  got expected failure: TEST_ERRNO=EINVAL(22): Invalid argument
> clock_gettime03    5  TPASS  :  got expected failure: TEST_ERRNO=EFAULT(14): Bad address
> clock_gettime03    6  TPASS  :  got expected failure: TEST_ERRNO=EFAULT(14): Bad address
> <snip>
> timer_create04    1  TPASS  :  failed as expected: TEST_ERRNO=EINVAL(22): Invalid argument
> timer_create04    2  TPASS  :  failed as expected: TEST_ERRNO=EINVAL(22): Invalid argument
> timer_create04    3  TPASS  :  failed as expected: TEST_ERRNO=EFAULT(14): Bad address
> timer_create04    4  TPASS  :  failed as expected: TEST_ERRNO=EFAULT(14): Bad address
> timer_create04    5  TPASS  :  failed as expected: TEST_ERRNO=EFAULT(14): Bad address
> timer_create04    6  TPASS  :  failed as expected: TEST_ERRNO=EFAULT(14): Bad address
> ------
>
>> 2. *sigh* These constants should be autoconf conditionalized --
>> otherwise when Linux devs decide to change their minds about the
>> values of the constants or the values are different on different
>> architectures you'll run into red-herring errors. The test should be
>> resistant to this behavior, but currently isn't. This would also
>> prevent `ENOSYS'-like errors from showing up on kernels < 2.6.32-rc1.
>
> i checked the frequency of adding clock in include/linux/time.h.
>  2005-04-16    CLOCK_REALTIME/CLOCK_MONOTONIC/CLOCK_PROCESS_CPUTIME_ID/CLOCK_THREAD_CPUTIME_ID
>                (when git initialized)
>  2008-08-21    CLOCK_MONOTONIC_RAW added
>  2009-08-21    CLOCK_REALTIME_COARSE/CLOCK_MONOTONIC_COARSE added
>  (from http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=history;f=include/linux/time.h;h=6e026e45a179e2c66b347eb3834501a13be7ee2e;hb=aea187c46f7d03ce985e55eb1398d0776a15b928)
> so, the frequency is slow, i think.
> And, should to accept my patch at first. ;-)

    I've added the autoconf-conditionalized variables along with your
changes, and some other minor cleanups to git just moments ago.
Cheers,
-Garrett