Comment # 5 on bug 904097 from Takashi Iwai

You may try the following:
- Run "sysctl kernel.sysrq=1"
  (and/or edit /etc/sysctl.conf and reboot for the next boot)
- Trigger Alt-SysRq-9 once.  This will change the loglevel to 9 (full).
- When the problem happens, trigger Alt-SysRq-T, Alt-SysRq-S, Alt-SysRq-U and
Alt-SysRq-B.  This will dump the full stack traces, sync the disk, umount the
disks and reboot.
- Check the previous kernel messages.

If the stalling place is what the kernel warning below indiciates, you might
need to wait 1 minute for timeout.  If the stack traces can be caught by the
action above, we can see a bit more details.