Re: mdadm-3.2.5: segfault in "--grow --continue"

[NeilBrown <neilb@suse.de>]

[-- Attachment #1: Type: text/plain, Size: 2385 bytes --]

On Tue, 10 Jul 2012 12:46:19 +0200 Sebastian Hegler
<sebastian.hegler@tu-dresden.de> wrote:

> Hi!
> 
> I had to shut down a server with a raid grow operation (cleanly). After some hassles I got the array assembled again (manually), but it would not continue to grow. When trying 
> 
> mdadm --grow --continue /dev/md127 
> 
> (as per the manpage) I receive a segfault. Compiling from source, and using gdb, I see:
> 
> root@kuiper:~/mdadm-3.2.5# gdb ./mdadm 
> [SNIP]
> Reading symbols from /root/mdadm-3.2.5/mdadm...done.
> (gdb) set args --grow --continue /dev/md127 
> (gdb) run 
> Starting program: /root/mdadm-3.2.5/mdadm --grow --continue /dev/md127 
> 
> Program received signal SIGSEGV, Segmentation fault.
> Grow_continue_command (devname=0x7fffffffe8cc "/dev/md127", fd=7, backup_file=0x0, verbose=0) at Grow.c:4118
> 4118		if (verify_reshape_position(content,
> (gdb) bt
> #0  Grow_continue_command (devname=0x7fffffffe8cc "/dev/md127", fd=7, backup_file=0x0, verbose=0) at Grow.c:4118
> #1  0x0000000000407ac2 in main (argc=4, argv=0x7fffffffe678) at mdadm.c:1701
> 
> 
> 
> The same bug is present in the git repository, but another location:
> 
> root@kuiper:~/mdadm.git# gdb ./mdadm 
> [SNIP]
> Reading symbols from /root/mdadm.git/mdadm...done.
> (gdb) set args --grow --continue /dev/md127 
> (gdb) run 
> Starting program: /root/mdadm.git/mdadm --grow --continue /dev/md127 
> 
> Program received signal SIGSEGV, Segmentation fault.
> Grow_continue_command (devname=0x7fffffffe8d0 "/dev/md127", fd=7, backup_file=0x0, verbose=0) at Grow.c:4086
> 4086		if (verify_reshape_position(content,
> (gdb) bt
> #0  Grow_continue_command (devname=0x7fffffffe8d0 "/dev/md127", fd=7, backup_file=0x0, verbose=0) at Grow.c:4086
> #1  0x0000000000406c39 in main (argc=4, argv=0x7fffffffe678) at mdadm.c:1447
> 
> 
> I also filed a bug against Debian's BTS, but I did not receive a bug number yet. In the meantime, I'd be very happy about any information on how to get my RAID array back into growing. I'm not on the list, please CC me.
> 


--grow --continue should only be needed if the array was assembled with
--freeze-reshape (I think).

Normally, the reshape should continue automatically.

cat /proc/mdstat
mdadm -E ....
mdadm -D ...

would be helpful, maybe together with kernel logs if there are any.

NeilBrown

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 828 bytes --]