[PATCH v4 3/3] kselftest: Extend vDSO selftest to clock_getres

[Christophe Leroy]

Le 04/06/2019 à 15:43, Vincenzo Frascino a écrit :
> On 04/06/2019 14:39, Christophe Leroy wrote:
>>
>>
>> Le 04/06/2019 à 15:32, Vincenzo Frascino a écrit :
>>> Hi Christophe,
>>>
>>> On 04/06/2019 14:16, Christophe Leroy wrote:
>>>> Hi Vincenzo
>>>>
>>>> Le 28/05/2019 à 13:57, Vincenzo Frascino a écrit :
>>>>> Hi Michael,
>>>>>
>>>>> thank you for your reply.
>>>>>
>>>>> On 28/05/2019 07:19, Michael Ellerman wrote:
>>>>>> Vincenzo Frascino <vincenzo.frascino at arm.com> writes:
>>>>>>
>>>>>>> The current version of the multiarch vDSO selftest verifies only
>>>>>>> gettimeofday.
>>>>>>>
>>>>>>> Extend the vDSO selftest to clock_getres, to verify that the
>>>>>>> syscall and the vDSO library function return the same information.
>>>>>>>
>>>>>>> The extension has been used to verify the hrtimer_resoltion fix.
>>>>>>
>>>>>> This is passing for me even without patch 1 applied, shouldn't it fail
>>>>>> without the fix? What am I missing?
>>>>>>
>>>>>
>>>>> This is correct, because during the refactoring process I missed an "n" :)
>>>>>
>>>>> if·((x.tv_sec·!=·y.tv_sec)·||·(x.tv_sec·!=·y.tv_sec))
>>>>>
>>>>> Should be:
>>>>>
>>>>> if·((x.tv_sec·!=·y.tv_sec)·||·(x.tv_nsec·!=·y.tv_nsec))
>>>>>
>>>>> My mistake, I am going to fix the test and re-post v5 of this set.
>>>>>
>>>>> Without my patch if you pass "highres=off" to the kernel (as a command line
>>>>> parameter) it leads to a broken implementation of clock_getres since the value
>>>>> of CLOCK_REALTIME_RES does not change at runtime.
>>>>>
>>>>> Expected result (with highres=off):
>>>>>
>>>>> # uname -r
>>>>> 5.2.0-rc2
>>>>> # ./vdso_clock_getres
>>>>> clock_id: CLOCK_REALTIME [FAIL]
>>>>> clock_id: CLOCK_BOOTTIME [PASS]
>>>>> clock_id: CLOCK_TAI [PASS]
>>>>> clock_id: CLOCK_REALTIME_COARSE [PASS]
>>>>> clock_id: CLOCK_MONOTONIC [FAIL]
>>>>> clock_id: CLOCK_MONOTONIC_RAW [PASS]
>>>>> clock_id: CLOCK_MONOTONIC_COARSE [PASS]
>>>>>
>>>>> The reason of this behavior is that the only clocks supported by getres on
>>>>> powerpc are CLOCK_REALTIME and CLOCK_MONOTONIC, the rest on the clocks use
>>>>> always syscalls.
>>>>
>>>> vdso64 is supposed to implement CLOCK_{REALTIME/MONOTONIC}_COARSE, so I
>>>> guess it should fail for them too ?
>>>>
>>>> Or is your test done on vdso32 ?
>>>>
>>>
>>> Based on what I can see in kernel/vdso64 in 5.2-rc3:
>>>
>>> /*
>>>    * Exact prototype of clock_getres()
>>>    *
>>>    * int __kernel_clock_getres(clockid_t clock_id, struct timespec *res);
>>>    *
>>>    */
>>> V_FUNCTION_BEGIN(__kernel_clock_getres)
>>>     .cfi_startproc
>>> 	/* Check for supported clock IDs */
>>> 	cmpwi	cr0,r3,CLOCK_REALTIME
>>> 	cmpwi	cr1,r3,CLOCK_MONOTONIC
>>> 	cror	cr0*4+eq,cr0*4+eq,cr1*4+eq
>>> 	bne	cr0,99f
>>>
>>> 	li	r3,0
>>> 	cmpldi	cr0,r4,0
>>> 	crclr	cr0*4+so
>>> 	beqlr
>>> 	lis	r5,CLOCK_REALTIME_RES at h
>>> 	ori	r5,r5,CLOCK_REALTIME_RES at l
>>> 	std	r3,TSPC64_TV_SEC(r4)
>>> 	std	r5,TSPC64_TV_NSEC(r4)
>>> 	blr
>>>
>>> 	/*
>>> 	 * syscall fallback
>>> 	 */
>>> 99:
>>> 	li	r0,__NR_clock_getres
>>> 	sc
>>> 	blr
>>>     .cfi_endproc
>>> V_FUNCTION_END(__kernel_clock_getres)
>>>
>>> it does not seem so for what concerns vdso64. I did run again the test both on
>>> ppc and ppc64 qemu instances and the result is the same to what I reported in
>>> this thread.
>>>
>>> Am I missing something?
>>
>> I was thinking about
>> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=5c929885f1bb
>> but apparently clock_getres() was left aside. Should we do something
>> about it ?
>>
> 
> Sure, but I would like this series to be merged first (since the topic is
> different). I am happy, after that, to push a separate one on top that addresses
> the problem.
> 
> Please let me know if it works for you and Michael.

No problem for myself.

By the way, next time (or next spin ?) I recommend you to handle your 
patches independently and not as a series since they are all 
independant. It would have avoided confusion and the need for you to 
resend all 3 patches everytime you did a change in one of them.

Christophe

> 
>> Christophe
>>
>>>
>>>> Christophe
>>>>
>>>>>
>>>>>> # uname -r
>>>>>> 5.2.0-rc2-gcc-8.2.0
>>>>>>
>>>>>> # ./vdso_clock_getres
>>>>>> clock_id: CLOCK_REALTIME [PASS]
>>>>>> clock_id: CLOCK_BOOTTIME [PASS]
>>>>>> clock_id: CLOCK_TAI [PASS]
>>>>>> clock_id: CLOCK_REALTIME_COARSE [PASS]
>>>>>> clock_id: CLOCK_MONOTONIC [PASS]
>>>>>> clock_id: CLOCK_MONOTONIC_RAW [PASS]
>>>>>> clock_id: CLOCK_MONOTONIC_COARSE [PASS]
>>>>>>
>>>>>> cheers
>>>>>>
>>>>>>> Cc: Shuah Khan <shuah at kernel.org>
>>>>>>> Signed-off-by: Vincenzo Frascino <vincenzo.frascino at arm.com>
>>>>>>> ---
>>>>>>>
>>>>>>> Note: This patch is independent from the others in this series, hence it
>>>>>>> can be merged singularly by the kselftest maintainers.
>>>>>>>
>>>>>>>     tools/testing/selftests/vDSO/Makefile         |   2 +
>>>>>>>     .../selftests/vDSO/vdso_clock_getres.c        | 124 ++++++++++++++++++
>>>>>>>     2 files changed, 126 insertions(+)
>>>>>>>     create mode 100644 tools/testing/selftests/vDSO/vdso_clock_getres.c
>>>>>
>>>
>