Re: [Starpu-devel] Example sched_ctx segfaults

[Andra Hugo <andra.hugo@inria.fr>]

Hi,

I'll check this out.

Andra

----- Mail original -----
> De: "Nathalie Furmento" <nathalie.furmento@labri.fr>
> À: starpu-devel@lists.gforge.inria.fr
> Envoyé: Mardi 19 Février 2013 13:47:07
> Objet: [Starpu-devel] Example sched_ctx segfaults
> 
> 
> All,
> 
> I am getting a segfault when running the example sched_ctx/sched_ctx
> with the defaut configuration on the hannibal machine.
> 
> It does not  occur all the time, but here a backtrace
> 
> $ libtool --mode=execute gdb ./examples/sched_ctx/sched_ctx
> ...
> tasks executed 1000 out of 1000
> 
> Program received signal SIGSEGV, Segmentation fault.
> [Switching to Thread 0x7fffeba35700 (LWP 19025)]
> 0x00007ffff7ef6164 in starpu_sched_ctx_remove_workers
> (workers_to_remove=0x7fffe5bba2c0, nworkers_to_remove=3,
> sched_ctx_id=2)
> at core/sched_ctx.c:626
> 626        _starpu_check_workers(workers_to_remove,
> nworkers_to_remove);
> (gdb) bt
> #0  0x00007ffff7ef6164 in starpu_sched_ctx_remove_workers
> (workers_to_remove=0x7fffe5bba2c0, nworkers_to_remove=3,
> sched_ctx_id=2)
> at core/sched_ctx.c:626
> #1  0x00007ffff7ef6762 in
> _starpu_decrement_nsubmitted_tasks_of_sched_ctx (sched_ctx_id=2) at
> core/sched_ctx.c:737
> #2  0x00007ffff7ed2b54 in _starpu_handle_job_termination
> (j=0x7fffdc017550) at core/jobs.c:270
> #3  0x00007ffff7f27456 in _starpu_cuda_driver_run_once
> (d=0x7fffeba34ea0) at drivers/cuda/driver_cuda.c:497
> #4  0x00007ffff7f2752d in _starpu_cuda_worker (arg=0x7ffff7f8bc40) at
> drivers/cuda/driver_cuda.c:537
> #5  0x00007ffff0fa6b50 in start_thread (arg=<optimized out>) at
> pthread_create.c:304
> #6  0x00007ffff0cf170d in clone () at
> ../sysdeps/unix/sysv/linux/x86_64/clone.S:112
> #7  0x0000000000000000 in ?? ()
> (gdb) p workers_to_remove
> $1 = (int *) 0x7fffe5bba2c0
> (gdb) p nworkers_to_remove
> $2 = 3
> (gdb) p *workers_to_remove
> $3 = 0
> 
> (gdb) f 1
> #1  0x00007ffff7ef6762 in
> _starpu_decrement_nsubmitted_tasks_of_sched_ctx (sched_ctx_id=2) at
> core/sched_ctx.c:737
> 737                starpu_sched_ctx_remove_workers(workerids,
> nworkers,
> sched_ctx_id);
> (gdb) p sched_ctx->workers
> $4 = (struct starpu_sched_ctx_worker_collection *) 0x3c059b0
> (gdb) p *sched_ctx->workers
> $5 = {workerids = 0x3c05200, nworkers = 3537031890, type =
> -757935406,
> has_next = 0xd2d2d2d2d2d2d2d2, get_next = 0xd2d2d2d2d2d2d2d2, add =
> 0xd2d2d2d2d2d2d2d2, remove = 0xd2d2d2d2d2d2d2d2, init =
> 0xd2d2d2d2d2d2d2d2,
>    deinit = 0xd2d2d2d2d2d2d2d2, init_iterator = 0x7ffff7f257d2
> <list_init_iterator>}
> 
> The object sched_ctx->worker seems quite messy, which would explain
> why
> the list returned by _get_workers_list is NULL.
> 
> Cheers,
> 
> Nathalie
> 
> 
> _______________________________________________
> Starpu-devel mailing list
> Starpu-devel@lists.gforge.inria.fr
> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/starpu-devel
>