[Xenomai-help] Problem with absolute timeout

[Xenomai-help] Problem with absolute timeout

[Kolja Waschk]

Hi,

Sorry for bothering the list with this probably silly problem, but I just don't have the right idea.

The POSIX skin sem_timedwait() call doesn't wait!? I expected the following test code to take 2 seconds until it finishes (and that holds true on a PC), but instead, it returns immediately. What am I doing wrong here?


#include <errno.h>
#include <sys/time.h>
#include <semaphore.h>

int main()
{
     int r;
     sem_t s;
     struct timeval tv;
     struct timespec ts;

     r = sem_init(&s, 0, 0);
     if (r != 0) { perror("sem_init"); return errno; }

     r = gettimeofday(&tv, 0);
     if (r != 0) { perror("gettimeofday"); return errno; }

     // Same result with TIMEVAL_TO_TIMESPEC
     ts.tv_sec = tv.tv_sec;
     ts.tv_nsec = tv.tv_usec * 1000;

     ts.tv_sec += 2;

     r = sem_timedwait(&s, &ts);
     if (r != 0) { perror("sem_timedwait"); return errno; }
}

The result is always !0, errno = 110(ETIMEDOUT), immediately after starting.

/tmp # time ./totest 
sem_timedwait: Connection timed out
Command exited with non-zero status 110
real	0m 0.00s
user	0m 0.00s
sys	0m 0.00s


Test environment: Blackfin uClinux-dist 2009R1.1-RC4 (Xenomai 2.4.7, 2.6.28.10/I-Pipe 1.10.0)
but also tried and failed with Xenomai 2.5-rc4 in Kernel 2.6.31.6.

Thanks for any hints!
Kolja

Re: [Xenomai-help] Problem with absolute timeout

[Philippe Gerum]

On Thu, 2010-03-18 at 14:29 +0100, Kolja Waschk wrote:
> Hi,
> 
> Sorry for bothering the list with this probably silly problem, but I just don't have the right idea.
> 
> The POSIX skin sem_timedwait() call doesn't wait!? I expected the following test code to take 2 seconds until it finishes (and that holds true on a PC), but instead, it returns immediately. What am I doing wrong here?
> 
> 
> #include <errno.h>
> #include <sys/time.h>
> #include <semaphore.h>
> 
> int main()
> {
>      int r;
>      sem_t s;
>      struct timeval tv;
>      struct timespec ts;
> 
>      r = sem_init(&s, 0, 0);
>      if (r != 0) { perror("sem_init"); return errno; }
> 
>      r = gettimeofday(&tv, 0);
>      if (r != 0) { perror("gettimeofday"); return errno; }

gettimeofday() is not the right time base for a Xenomai call. Try
clock_gettime(CLOCK_MONOTONIC/CLOCK_REALTIME...). CLOCK_MONOTONIC is
faster if you don't require the returned time to be adjusted on the
linux time.

> 
>      // Same result with TIMEVAL_TO_TIMESPEC
>      ts.tv_sec = tv.tv_sec;
>      ts.tv_nsec = tv.tv_usec * 1000;
> 
>      ts.tv_sec += 2;
> 
>      r = sem_timedwait(&s, &ts);
>      if (r != 0) { perror("sem_timedwait"); return errno; }
> }
> 
> The result is always !0, errno = 110(ETIMEDOUT), immediately after starting.
> 
> /tmp # time ./totest 
> sem_timedwait: Connection timed out
> Command exited with non-zero status 110
> real	0m 0.00s
> user	0m 0.00s
> sys	0m 0.00s
> 
> 
> Test environment: Blackfin uClinux-dist 2009R1.1-RC4 (Xenomai 2.4.7, 2.6.28.10/I-Pipe 1.10.0)
> but also tried and failed with Xenomai 2.5-rc4 in Kernel 2.6.31.6.
> 
> Thanks for any hints!
> Kolja
> 
> 
> 
> _______________________________________________
> Xenomai-help mailing list
> Xenomai-help@domain.hid
> https://mail.gna.org/listinfo/xenomai-help


-- 
Philippe.

Re: [Xenomai-help] Problem with absolute timeout

[Gilles Chanteperdrix]

Philippe Gerum wrote:
> On Thu, 2010-03-18 at 14:29 +0100, Kolja Waschk wrote:
>> Hi,
>>
>> Sorry for bothering the list with this probably silly problem, but I just don't have the right idea.
>>
>> The POSIX skin sem_timedwait() call doesn't wait!? I expected the following test code to take 2 seconds until it finishes (and that holds true on a PC), but instead, it returns immediately. What am I doing wrong here?
>>
>>
>> #include <errno.h>
>> #include <sys/time.h>
>> #include <semaphore.h>
>>
>> int main()
>> {
>>      int r;
>>      sem_t s;
>>      struct timeval tv;
>>      struct timespec ts;
>>
>>      r = sem_init(&s, 0, 0);
>>      if (r != 0) { perror("sem_init"); return errno; }
>>
>>      r = gettimeofday(&tv, 0);
>>      if (r != 0) { perror("gettimeofday"); return errno; }
> 
> gettimeofday() is not the right time base for a Xenomai call. Try
> clock_gettime(CLOCK_MONOTONIC/CLOCK_REALTIME...). CLOCK_MONOTONIC is
> faster if you don't require the returned time to be adjusted on the
> linux time.

There is no single rule for the POSIX API, some calls allow to use
CLOCK_REALTIME, some CLOCK_MONOTONIC, other to choose the clock, yet
other to use relative timings. For the case of sem_timedwait, you have
to use CLOCK_REALTIME, as documented here:

http://www.xenomai.org/documentation/xenomai-2.5/html/api/group__posix__sem.html#g8678e7f5136343ef8d8123a185878f78

> 
>>      // Same result with TIMEVAL_TO_TIMESPEC
>>      ts.tv_sec = tv.tv_sec;
>>      ts.tv_nsec = tv.tv_usec * 1000;
>>
>>      ts.tv_sec += 2;
>>
>>      r = sem_timedwait(&s, &ts);
>>      if (r != 0) { perror("sem_timedwait"); return errno; }
>> }
>>
>> The result is always !0, errno = 110(ETIMEDOUT), immediately after starting.
>>
>> /tmp # time ./totest 
>> sem_timedwait: Connection timed out
>> Command exited with non-zero status 110
>> real	0m 0.00s
>> user	0m 0.00s
>> sys	0m 0.00s
>>
>>
>> Test environment: Blackfin uClinux-dist 2009R1.1-RC4 (Xenomai 2.4.7, 2.6.28.10/I-Pipe 1.10.0)
>> but also tried and failed with Xenomai 2.5-rc4 in Kernel 2.6.31.6.
>>
>> Thanks for any hints!
>> Kolja
>>
>>
>>
>> _______________________________________________
>> Xenomai-help mailing list
>> Xenomai-help@domain.hid
>> https://mail.gna.org/listinfo/xenomai-help
> 
> 


-- 
Gilles Chanteperdrix, Free Electrons
Kernel, drivers, real-time and embedded Linux
development, consulting, training and support.
http://free-electrons.com