md/raid1: Don't use next_resync to determine how far resync has progressed
authorNeilBrown <neilb@suse.de>
Wed, 10 Sep 2014 05:56:57 +0000 (15:56 +1000)
committerNeilBrown <neilb@suse.de>
Mon, 22 Sep 2014 01:26:01 +0000 (11:26 +1000)
commit235549605eb7f1c5a37cef8b09d12e6d412c5cd6
tree13ea4c98190fd461cae935a0f5cf16cbc1403380
parent2f73d3c55d09ce60647b96ad2a9b539c95a530ee
md/raid1: Don't use next_resync to determine how far resync has progressed

next_resync is (approximately) the location for the next resync request.
However it does *not* reliably determine the earliest location
at which resync might be happening.
This is because resync requests can complete out of order, and
we only limit the number of current requests, not the distance
from the earliest pending request to the latest.

mddev->curr_resync_completed is a reliable indicator of the earliest
position at which resync could be happening.   It is updated less
frequently, but is actually reliable which is more important.

So use it to determine if a write request is before the region
being resynced and so safe from conflict.

This error can allow resync IO to interfere with normal IO which
could lead to data corruption. Hence: stable.

Fixes: 79ef3a8aa1cb1523cc231c9a90a278333c21f761
cc: stable@vger.kernel.org (v3.13+)
Signed-off-by: NeilBrown <neilb@suse.de>
drivers/md/raid1.c