Jan Kiszka wrote:
> M. Koehrer wrote:
>> Hi all,
>>
>> some more interesing measurements:
>> 1) Even when I remove the printf() completely, the system freezes.
>> My realtimetask is then
>> static void realtimetask(void *arg)
>> {
>>     system("ls -l");
>>     rt_task_sleep(1000000000ULL);
>>
>> }
>>
>> 2) When I replace the printf() by a gepid() the behaviour is the same.
> 
> Hah, it finally oopses! Hurray!! Now the funny part begins...
> 

It must be a tight race window. My oops disappears once logging goes to
serial console, so I only caught the following from the stalled console:

BUG: sleeping function called from invalid context at
include/asm/semaphore.h
in_atomic():0, irqs_disabled():1
 [...] lock_kernel+0x28
       default_llseek+0x15
       sys_llseek+0x4e
       sysenter_past_esp+0x65

This all happens shortly after the test case terminated. I'm able scroll
for a short while up in the log, but then the box goes South. With the
serial console attached I only get that trap-#14 reports by Xenomai that
the test program got switched to secondary mode.

I'm not able to continue my test on this particular system for the rest
of the year. I will try to reproduce it again elsewhere (the SPINLOCK
debugging switches seem to reveal something - sometime), but there is no
guarantee that I can do anything about it.

Jan