Re: [microblaze-uclinux] pthread_mutex_unlock() on unlocked mutexlock system hard

[John Williams <jwilliams@xxxxxxxxxxxxxx>]

Hi Yashi,

Yasushi SHOJI wrote:

> attached code locks system hard.  I'm not sure where to look
> for. armnommu works just fine with same code, so this problem should
> be microblaze specific. (or SUZAKU specific, I don't know)

pthreads support on microblaze is experimental - I worked on it for as 
long as it took to get the thread_demos to ocmpile, link and run without 
  total failure!

That said, your demo seems to run fine on my system (pretty standard 
mbvanilla_net). Output:

# /mnt/yashitest
process wait
process run 0

As for source of problem - by locks hard I assume you mean totally dead 
system?  no ping reply, no interrupts, nothing?

If you use the mdm based debugging (see website for uClinux-specific 
details), you can fire up xmd, connect, and basically trap the 
microblaze "live".  Then load mb-gdb, you can do a stack unwind, see 
where the execution context is, and see how it got there.  This only 
works in kernel space though, so if you're locking in a library or app, 
you're out of luck at this stage.

The arch-specific pthreads support lives in 
uClibc/libpthread/linuxthreads/sysdeps/microblaze

Just browsing there now, I see some horrible code of mine - a user/lib 
space cli() critical section for the semaphore implementation.  yuck. 
But, probably unlikely to be the cause of your problem.

If I think of anythign else I'll let you know.

Cheers,

John

___________________________
microblaze-uclinux mailing list
microblaze-uclinux@itee.uq.edu.au
Project Home Page : http://www.itee.uq.edu.au/~jwilliams/mblaze-uclinux
Mailing List Archive : http://www.itee.uq.edu.au/~listarch/microblaze-uclinux/