On 1/05/12 8:18 PM, NeilBrown wrote:

On Tue, 1 May 2012 10:45:08 +0100 Brian Candler[off-list ref]  wrote:

quoted

On Sat, Apr 28, 2012 at 08:03:55AM +1000, NeilBrown wrote:

quoted

I'm afraid you've been bitten by a rather nasty bug which is present in 3.3
and got back ported to some -stable kernel.  The fix has been submitted and
should be appearing in -stable kernels soon (maybe already).

Do you happen to know if this bug was present in ubuntu 11.10, kernel
versions 3.0.0-15-server or 3.0.0-16-server ?

I don't keep track of what ubuntu (or anyone but suse) put in their kernels,
sorry.

I wondered about Ubuntu and checked the kernel package changelog:
http://changelogs.ubuntu.com/changelogs/pool/main/l/linux/linux_3.0.0-19.33/changelog

Cross-checking with the commit nominated by Jan Ceuleers to this list on 
30/04/12 (c744a65c1e2d59acc54333ce8) and the Ubuntu changelog (see grep 
below) - I don't believe Ubuntu 11.10 includes the bug.

Cheers,

lb


$ grep ' md' changelog
   * md/bitmap: ensure to load bitmap when creating via sysfs.
   * md/raid1,raid10: avoid deadlock during resync/recovery.
   * net: Fix driver name for mdio-gpio.c
   * tcp: md5: using remote adress for md5 lookup in rst packet
   * md/raid5: fix bug that could result in reads from a failed device.
   * md/raid5: abort any pending parity operations when array fails.
   * md: Avoid waking up a thread after it has been freed.
   * md/raid5: fix bug that could result in reads from a failed device.
   * tcp: properly handle md5sig_pool references
   * md/linear: avoid corrupting structure while waiting for rcu_free to
   * md: Fix handling for devices from 2TB to 4TB in 0.90 metadata.
   * winbond: include linux/delay.h for mdelay et al
   * md: fix 'degraded' calculation when starting a reshape.
   * md: fix small irregularity with start_ro module parameter
   * SCSI: st: fix mdata->page_order handling
   * md: Fix unfortunate interaction with evms
   * md/bitmap: protect against bitmap removal while being updated.
   * md: Fix "strchr" [drivers/md/dm-log-userspace.ko] undefined!

quoted

I saw failures to assemble RAID6 arrays after unclean shutdowns on both of
these, just in test environments.  If it happens again I'll send mdadm
--examine output. I just tried to replicate it and failed :-(

To replicate:

  - create an array.
  - stop the array
  - assemble the array with at least one missing device e.g.:
     mdadm -A /dev/md0 /dev/sda1 /dev/sdb1
    if it was a 3-device array
  - check in /proc/mdstat that it is listed as "inactivate"
  - reboot
  - now "mdadm -E" the devices.  If the raid level is -unknown- then the bug
    has hit.

NeilBrown

quoted

Regards,

Brian.
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html