[patch 1/2]block, cfq: fix empty queue crash caused by request merge

[patch 1/2]block, cfq: fix empty queue crash caused by request merge

[Shaohua Li]

All requests of a queue could be merged to other requests of other queue.
Such queue will not have request in it, but it's in service tree. This
will cause kernel oops.
I encounter a BUG_ON() in cfq_dispatch_request() with next patch, but the
issue should exist without the patch.

Signed-off-by: Shaohua Li <shaohua.li@intel.com>
---
 block/cfq-iosched.c |   12 ++++++++++++
 1 file changed, 12 insertions(+)

Index: linux/block/cfq-iosched.c
===================================================================
--- linux.orig/block/cfq-iosched.c	2011-12-16 09:01:57.000000000 +0800
+++ linux/block/cfq-iosched.c	2011-12-16 09:07:44.000000000 +0800
@@ -1655,6 +1655,8 @@ cfq_merged_requests(struct request_queue
 		    struct request *next)
 {
 	struct cfq_queue *cfqq = RQ_CFQQ(rq);
+	struct cfq_data *cfqd = q->elevator->elevator_data;
+
 	/*
 	 * reposition in fifo if next is older than rq
 	 */
@@ -1669,6 +1671,16 @@ cfq_merged_requests(struct request_queue
 	cfq_remove_request(next);
 	cfq_blkiocg_update_io_merged_stats(&(RQ_CFQG(rq))->blkg,
 					rq_data_dir(next), rq_is_sync(next));
+
+	cfqq = RQ_CFQQ(next);
+	/*
+	 * all requests of this queue are merged to other queues, delete it
+	 * from the service tree. If it's the active_queue,
+	 * cfq_dispatch_requests() will choose to expire it or do idle
+	 */
+	if (cfq_cfqq_on_rr(cfqq) && RB_EMPTY_ROOT(&cfqq->sort_list) &&
+	    cfqq != cfqd->active_queue)
+		cfq_del_cfqq_rr(cfqd, cfqq);
 }
 
 static int cfq_allow_merge(struct request_queue *q, struct request *rq,

Re: [patch 1/2]block, cfq: fix empty queue crash caused by request merge

[Jens Axboe]

On 2011-12-16 02:52, Shaohua Li wrote:
> All requests of a queue could be merged to other requests of other queue.
> Such queue will not have request in it, but it's in service tree. This
> will cause kernel oops.
> I encounter a BUG_ON() in cfq_dispatch_request() with next patch, but the
> issue should exist without the patch.

Good catch, yes I think this could happen if you just happen to have a
very deliberate/unlucky set of processes and timings. Applied.


-- 
Jens Axboe