From: Stefan Haberland sth@linux.ibm.com
commit 6f117cb854a44a79898d844e6ae3fd23bd94e786 upstream.
When requeueing all requests on the device request queue to the blocklayer we might get to an ERP (error recovery) request that is a copy of an original CQR.
Those requests do not have blocklayer request information or a pointer to the dasd_queue set. When trying to access those data it will lead to a null pointer dereference in dasd_requeue_all_requests().
Fix by checking if the request is an ERP request that can simply be ignored. The blocklayer request will be requeued by the original CQR that is on the device queue right behind the ERP request.
Fixes: 9487cfd3430d ("s390/dasd: fix handling of internal requests") Cc: stable@vger.kernel.org #4.16 Signed-off-by: Stefan Haberland sth@linux.ibm.com Reviewed-by: Jan Hoeppner hoeppner@linux.ibm.com Signed-off-by: Jens Axboe axboe@kernel.dk Signed-off-by: Greg Kroah-Hartman gregkh@linuxfoundation.org
--- drivers/s390/block/dasd.c | 6 ++++++ 1 file changed, 6 insertions(+)
--- a/drivers/s390/block/dasd.c +++ b/drivers/s390/block/dasd.c @@ -2891,6 +2891,12 @@ static int _dasd_requeue_request(struct
if (!block) return -EINVAL; + /* + * If the request is an ERP request there is nothing to requeue. + * This will be done with the remaining original request. + */ + if (cqr->refers) + return 0; spin_lock_irq(&cqr->dq->lock); req = (struct request *) cqr->callback_data; blk_mq_requeue_request(req, false);