[Xenomai-help] deadlock for rtdm_mutex

[Xenomai-help] deadlock for rtdm_mutex

[Huan Fang]

Hi,

Sometimes I got below assertion failed:

xenomai/skins/rtdm/drvlib.c:1544 (mutex_owner != curr_thread)

Looking into the code,
/* Redefinition to clarify XENO_ASSERT output */
		#define mutex_owner xnsynch_owner(&mutex->synch_base)
		XENO_ASSERT(RTDM, mutex_owner != curr_thread,
			    err = -EDEADLK; goto unlock_out;);

Does it mean if the current owner of mutex is different from the requesting thread, then a -EDEADLK error is returned?(which is not documented in RTDM api doc)
In my case, if one thread tries to lock same mutex twice, the assertion is thrown out as well. Could you please explain a bit how does Xenomai detect deadlock?
Thanks.

Regards,
Huan

Re: [Xenomai-help] deadlock for rtdm_mutex

[Philippe Gerum]

On Thu, 2010-03-18 at 16:05 +0100, Huan Fang wrote:
> Hi,
> 
> Sometimes I got below assertion failed:
> 
> xenomai/skins/rtdm/drvlib.c:1544 (mutex_owner != curr_thread)
> 
> Looking into the code,
> /* Redefinition to clarify XENO_ASSERT output */
> 		#define mutex_owner xnsynch_owner(&mutex->synch_base)
> 		XENO_ASSERT(RTDM, mutex_owner != curr_thread,
> 			    err = -EDEADLK; goto unlock_out;);
> 
> Does it mean if the current owner of mutex is different from the requesting thread, then a -EDEADLK error is returned?(which is not documented in RTDM api doc)
> In my case, if one thread tries to lock same mutex twice, the assertion is thrown out as well. Could you please explain a bit how does Xenomai detect deadlock?
> Thanks.

#define XENO_ASSERT(subsystem,cond,action)  do { \
    if (unlikely(CONFIG_XENO_OPT_DEBUG_##subsystem > 0 && !(cond))) { \
        xnarch_trace_panic_freeze(); \
        xnlogerr("assertion failed at %s:%d (%s)\n", __FILE__, __LINE__,
(#cond)); \
        xnarch_trace_panic_dump(); \
        action; \
    } \
} while(0)

		XENO_ASSERT(RTDM, mutex_owner != curr_thread,
			    err = -EDEADLK; goto unlock_out;);

This means that RTDM mutexes do not allow recursion.

> 
> Regards,
> Huan
> _______________________________________________
> Xenomai-help mailing list
> Xenomai-help@domain.hid
> https://mail.gna.org/listinfo/xenomai-help


-- 
Philippe.

Re: [Xenomai-help] deadlock for rtdm_mutex

[Jan Kiszka]

Huan Fang wrote:
> Hi,
> 
> Sometimes I got below assertion failed:
> 
> xenomai/skins/rtdm/drvlib.c:1544 (mutex_owner != curr_thread)
> 
> Looking into the code,
> /* Redefinition to clarify XENO_ASSERT output */
> 		#define mutex_owner xnsynch_owner(&mutex->synch_base)
> 		XENO_ASSERT(RTDM, mutex_owner != curr_thread,
> 			    err = -EDEADLK; goto unlock_out;);
> 
> Does it mean if the current owner of mutex is different from the requesting thread, then a -EDEADLK error is returned?(which is not documented in RTDM api doc)

If the mutex is contended, the owner _must_ be different from the
current thread.

> In my case, if one thread tries to lock same mutex twice, the assertion is thrown out as well. Could you please explain a bit how does Xenomai detect deadlock?

Quite simple: RTDM mutexes do not support recursion, so you must not
acquire them twice in the same thread. That's a reasonable
simplification of the design you find in the mainline kernel as well e.g.

Jan

-- 
Siemens AG, Corporate Technology, CT T DE IT 1
Corporate Competence Center Embedded Linux

Re: [Xenomai-help] deadlock for rtdm_mutex

[Huan Fang]

Thanks for the explanation, now it's clear.

Regards,
Huan