Re: [PATCH] Revert "raid5: make release_stripe lockless" because it causes test regression

On Mon, 25 Nov 2013, NeilBrown wrote:

On Fri, 22 Nov 2013 15:07:08 -0500 (EST) Mikulas Patocka
[off-list ref] wrote:

quoted

Revert 773ca82fa1ee58dd1bf88b6a5ca385ec83a2cac6 and
d265d9dc1d25a69affc21ae9fe5004b9d09c10ef.

The lvm testsuite (from 
ftp://sources.redhat.com/pub/lvm2/LVM2.2.02.104.tgz) fails with unkillable 
lvm and dmsetup processes on kernel 3.12. The testsuite passes on 3.11. 
Bisect shows that the failure is caused by patch 
773ca82fa1ee58dd1bf88b6a5ca385ec83a2cac6. When I revert this patch on 3.12 
kernel, the testsuite finishes and there are no unkillable processes.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Cc: stable@kernel.org	# 3.12

I certainly don't plan to revert these patches unless we fail to understand
and fix the problem.

Does commit ad4068de49862b083ac2a15bc50689bb30ce3e44 fix the problem?

Thanks,
NeilBrown

Yes. This patch, applied to 3.12, fixes it.

Mikulas