blk-flush: move the queue kick into
authorJeff Moyer <jmoyer@redhat.com>
Mon, 17 Oct 2011 10:57:23 +0000 (12:57 +0200)
committerGreg Kroah-Hartman <gregkh@suse.de>
Fri, 11 Nov 2011 17:44:36 +0000 (09:44 -0800)
commite94d1d00463f6fdcef956828ff698a2bce3f7a8f
tree711c335eda09b34a63e7ff3969c79710ce68ffde
parentd16a63a89dcb0cece02469b3bb17062c3d48656f
blk-flush: move the queue kick into

commit e67b77c791ca2778198c9e7088f3266ed2da7a55 upstream.

A dm-multipath user reported[1] a problem when trying to boot
a kernel with commit 4853abaae7e4a2af938115ce9071ef8684fb7af4
(block: fix flush machinery for stacking drivers with differring
flush flags) applied.  It turns out that an empty flush request
can be sent into blk_insert_flush.  When the BUG_ON was fixed
to allow for this, I/O on the underlying device would stall.  The
reason is that blk_insert_cloned_request does not kick the queue.
In the aforementioned commit, I had added a special case to
kick the queue if data was sent down but the queue flags did
not require a flush.  A better solution is to push the queue
kick up into blk_insert_cloned_request.

This patch, along with a follow-on which fixes the BUG_ON, fixes
the issue reported.

[1] http://www.redhat.com/archives/dm-devel/2011-September/msg00154.html

Reported-by: Christophe Saout <christophe@saout.de>
Signed-off-by: Jeff Moyer <jmoyer@redhat.com>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
block/blk-core.c
block/blk-flush.c