[sheepdog] [PATCH 0/7] fix local driver to pass test scripts

MORITA Kazutaka morita.kazutaka at lab.ntt.co.jp
Wed Sep 5 07:03:20 CEST 2012


At Tue, 04 Sep 2012 19:21:51 +0800,
Liu Yuan wrote:
> 
> On 09/04/2012 04:05 PM, MORITA Kazutaka wrote:
> > This depends on the series I sent two hours ago.  I've pushed all of
> > patches to the devel branch for easier review.
> > 
> > Basically, this series does the same changes as we have done to the
> > corosync driver.  With this series, local driver can pass all
> > testcases.
> 
> I hit a crash on 026.
> 
> (gdb) bt
> #0  memcpy () at ../sysdeps/x86_64/memcpy.S:267
> #1  0x0000000000411aba in shm_queue_push (type=7007, node=<value optimized out>, buf=<value optimized out>, 
>     buf_len=<value optimized out>) at cluster/local.c:129
> #2  add_event (type=7007, node=<value optimized out>, buf=<value optimized out>, buf_len=<value optimized out>)
>     at cluster/local.c:266
> #3  0x0000000000411bbc in local_block () at cluster/local.c:343
> #4  0x0000000000404c37 in queue_cluster_request (req=0x16af5d0) at group.c:337
> #5  0x0000000000408ab2 in finish_rx (fd=<value optimized out>, events=1, data=<value optimized out>) at request.c:595
> #6  do_client_rx (fd=<value optimized out>, events=1, data=<value optimized out>) at request.c:603
> #7  client_handler (fd=<value optimized out>, events=1, data=<value optimized out>) at request.c:797
> #8  0x0000000000419c06 in event_loop (timeout=<value optimized out>) at event.c:180
> #9  0x0000000000404245 in main (argc=<value optimized out>, argv=0x7fffe55e77f8) at sheep.c:396

It seems that the memmapped area is an invalid address.  I have no
idea why this problem happens.

I've tested 026 for more than ten hours on several machines, but I
couldn't reproduce this problem.  Doesn't it happen with the current
master branch?  If it happens, I'd like to apply the series to move
forward.  I'll give a fix after I reproduce this crash problem.

Thanks,

Kazutaka



More information about the sheepdog mailing list