cciss: fix lost command issue
authorBud Brown <bud.brown@redhat.com>
Wed, 23 Mar 2011 19:47:11 +0000 (20:47 +0100)
committerAK <andi@firstfloor.org>
Thu, 31 Mar 2011 18:58:49 +0000 (11:58 -0700)
commitd6b391f4879ddcc92f3f46dca793aaaa2cb1a84e
tree8922bc3816c56031a842670ecd50d11b26e4b29d
parent7d90e50de05ac6589b9109039fa15b778522c62a
cciss: fix lost command issue

commit 1ddd5049545e0aa1a0ed19bca4d9c9c3ce1ac8a2 upstream.

Under certain workloads a command may seem to get lost. IOW, the Smart Array
thinks all commands have been completed but we still have commands in our
completion queue. This may lead to system instability, filesystems going
read-only, or even panics depending on the affected filesystem. We add an
extra read to force the write to complete.

Testing shows this extra read avoids the problem.

Signed-off-by: Mike Miller <mike.miller@hp.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: Andi Kleen <ak@linux.intel.com>
drivers/block/cciss.h