On Sat, Sep 14, 2013 at 12:50 PM, Martin Wilck [off-list ref] wrote:

I just had the following weird effect under OpenSUSE 12.3, kernel
3.7.10-1.16. I am wondering if it's a known problem.

Creating a RAID5 array (format doesn't seem to matter) either with the
openSUSE native mdadm 3.2.6 or mdadm 3.3, and running mke2fs on it,
fails. The RAID array is broken after that, all disks faulty, any IO fails.

Further examination shows that the problem is related to mke2fs trying
to discard blocks. Running "-E nodiscard" or setting "discard=no" in
mke2fs.conf fixes the problem. Unfortnately, "discard" seems to be the
default.

Details under https://bugzilla.novell.com/show_bug.cgi?id=840413

Has anyone seen this before?

This sounds like the issue addressed by:

66c28f9 [SCSI] sd: Update WRITE SAME heuristics
5026d7a md/raid1,5,10: Disable WRITE SAME until a recovery strategy is in place