raid5: fix 2 bugs in ops_complete_biofill
authorDan Williams <dan.j.williams@intel.com>
Mon, 24 Sep 2007 17:06:13 +0000 (10:06 -0700)
committerDan Williams <dan.j.williams@intel.com>
Mon, 24 Sep 2007 20:23:35 +0000 (13:23 -0700)
commite4d84909dd48b5e5806a5d18b881e1ca1610ba9b
tree85aa712ebc034a31b9a2a5db2f0e7ce4951cb440
parent6247cdc2cd334dad0ea5428245a7d8f4b075f21e
raid5: fix 2 bugs in ops_complete_biofill

1/ ops_complete_biofill tried to avoid calling handle_stripe since all the
state necessary to return read completions is available.  However the
process of determining whether more read requests are pending requires
locking the stripe (to block add_stripe_bio from updating dev->toead).
ops_complete_biofill can run in tasklet context, so rather than upgrading
all the stripe locks from spin_lock to spin_lock_bh this patch just
unconditionally reschedules handle_stripe after completing the read
request.

2/ ops_complete_biofill needlessly qualified processing R5_Wantfill with
dev->toread.  The result being that the 'biofill' pending bit is cleared
before handling the pending read-completions on dev->read.  R5_Wantfill can
be unconditionally handled because the 'biofill' pending bit prevents new
R5_Wantfill requests from being seen by ops_run_biofill and
ops_complete_biofill.

Found-by: Yuri Tikhonov <yur@emcraft.com>
[neilb@suse.de: simpler fix for bug 1 than moving code]
Signed-off-by: NeilBrown <neilb@suse.de>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
drivers/md/raid5.c