[PATCH 0/3] blk-mq ->queue_rq error return fixes

[PATCH 0/3] blk-mq ->queue_rq error return fixes

[Christoph Hellwig]

[PATCH 1/3] blk-mq: dont assume rq->errors is set when returning an error from ->queue_rq

[Christoph Hellwig]

[-- Attachment #1: 0001-blk-mq-don-t-assume-rq-errors-is-set-when-returning-.patch --]
[-- Type: text/plain, Size: 903 bytes --]

rq->errors never has been part of the communication protocol between drivers
and the block stack and most drivers will not have initialized it.

Return -EIO to upper layers when the driver returns BLK_MQ_RQ_QUEUE_ERROR
unconditionally.  If a driver want to return a different error it can easily
do so by returning success after calling blk_mq_end_io itself.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 block/blk-mq.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/block/blk-mq.c b/block/blk-mq.c
index a59b056..0480710 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -605,8 +605,8 @@ static void __blk_mq_run_hw_queue(struct blk_mq_hw_ctx *hctx)
 			break;
 		default:
 			pr_err("blk-mq: bad return on queue: %d\n", ret);
-			rq->errors = -EIO;
 		case BLK_MQ_RQ_QUEUE_ERROR:
+			rq->errors = -EIO;
 			blk_mq_end_io(rq, rq->errors);
 			break;
 		}
-- 
1.7.10.4

[PATCH 2/3] blk-mq: pair blk_mq_start_request / blk_mq_requeue_request

[Christoph Hellwig]

[-- Attachment #1: 0002-blk-mq-pair-blk_mq_start_request-blk_mq_requeue_requ.patch --]
[-- Type: text/plain, Size: 3102 bytes --]

Make sure we have a proper pairing between starting and requeueing
requests.  Move the dma drain and REQ_END setup into blk_mq_start_request,
and make sure blk_mq_requeue_request properly undoes them, giving us
a pair of function to prepare and unprepare a request without leaving
side effects.

Together this ensures we always clean up properly after
BLK_MQ_RQ_QUEUE_BUSY returns from ->queue_rq.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 block/blk-mq.c |   49 ++++++++++++++++++++++++++-----------------------
 1 file changed, 26 insertions(+), 23 deletions(-)

diff --git a/block/blk-mq.c b/block/blk-mq.c
index 0480710..1fa9dd1 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -377,7 +377,7 @@ void blk_mq_complete_request(struct request *rq)
 }
 EXPORT_SYMBOL(blk_mq_complete_request);
 
-static void blk_mq_start_request(struct request *rq)
+static void blk_mq_start_request(struct request *rq, bool last)
 {
 	struct request_queue *q = rq->q;
 
@@ -390,6 +390,25 @@ static void blk_mq_start_request(struct request *rq)
 	 */
 	rq->deadline = jiffies + q->rq_timeout;
 	set_bit(REQ_ATOM_STARTED, &rq->atomic_flags);
+
+	if (q->dma_drain_size && blk_rq_bytes(rq)) {
+		/*
+		 * Make sure space for the drain appears.  We know we can do
+		 * this because max_hw_segments has been adjusted to be one
+		 * fewer than the device can handle.
+		 */
+		rq->nr_phys_segments++;
+	}
+
+	/*
+	 * Flag the last request in the series so that drivers know when IO
+	 * should be kicked off, if they don't do it on a per-request basis.
+	 *
+	 * Note: the flag isn't the only condition drivers should do kick off.
+	 * If drive is busy, the last request might not have the bit set.
+	 */
+	if (last)
+		rq->cmd_flags |= REQ_END;
 }
 
 static void blk_mq_requeue_request(struct request *rq)
@@ -398,6 +417,11 @@ static void blk_mq_requeue_request(struct request *rq)
 
 	trace_block_rq_requeue(q, rq);
 	clear_bit(REQ_ATOM_STARTED, &rq->atomic_flags);
+
+	rq->cmd_flags &= ~REQ_END;
+
+	if (q->dma_drain_size && blk_rq_bytes(rq))
+		rq->nr_phys_segments--;
 }
 
 struct blk_mq_timeout_data {
@@ -565,29 +589,8 @@ static void __blk_mq_run_hw_queue(struct blk_mq_hw_ctx *hctx)
 
 		rq = list_first_entry(&rq_list, struct request, queuelist);
 		list_del_init(&rq->queuelist);
-		blk_mq_start_request(rq);
 
-		if (q->dma_drain_size && blk_rq_bytes(rq)) {
-			/*
-			 * make sure space for the drain appears we
-			 * know we can do this because max_hw_segments
-			 * has been adjusted to be one fewer than the
-			 * device can handle
-			 */
-			rq->nr_phys_segments++;
-		}
-
-		/*
-		 * Last request in the series. Flag it as such, this
-		 * enables drivers to know when IO should be kicked off,
-		 * if they don't do it on a per-request basis.
-		 *
-		 * Note: the flag isn't the only condition drivers
-		 * should do kick off. If drive is busy, the last
-		 * request might not have the bit set.
-		 */
-		if (list_empty(&rq_list))
-			rq->cmd_flags |= REQ_END;
+		blk_mq_start_request(rq, list_empty(&rq_list));
 
 		ret = q->mq_ops->queue_rq(hctx, rq);
 		switch (ret) {
-- 
1.7.10.4

[PATCH 3/3] 0001-blk-mq-don-t-assume-rq-errors-is-set-when-returning-.patch 0002-blk-mq-pair-blk_mq_start_request-blk_mq_requeue_requ.patch series

[Christoph Hellwig]

[-- Attachment #1: series --]
[-- Type: text/plain, Size: 1 bytes --]

Re: [PATCH 3/3] 0001-blk-mq-don-t-assume-rq-errors-is-set-when-returning-.patch 0002-blk-mq-pair-blk_mq_start_request-blk_mq_requeue_requ.patch series

[Jens Axboe]

On Tue, Feb 11 2014, Christoph Hellwig wrote:
> 

looks like the last one got botched?

-- 
Jens Axboe

Re: [PATCH 3/3] 0001-blk-mq-don-t-assume-rq-errors-is-set-when-returning-.patch 0002-blk-mq-pair-blk_mq_start_request-blk_mq_requeue_requ.patch series

[Christoph Hellwig]

On Tue, Feb 11, 2014 at 09:31:04AM -0700, Jens Axboe wrote:
> On Tue, Feb 11 2014, Christoph Hellwig wrote:
> > 
> 
> looks like the last one got botched?

That was quilt sending along the series file which I created by doing

ls > series

just ignore it, hopefully one day quilt and I will get along fine..

Re: [PATCH 3/3] 0001-blk-mq-don-t-assume-rq-errors-is-set-when-returning-.patch 0002-blk-mq-pair-blk_mq_start_request-blk_mq_requeue_requ.patch series

[Jens Axboe]

On Tue, Feb 11 2014, Christoph Hellwig wrote:
> On Tue, Feb 11, 2014 at 09:31:04AM -0700, Jens Axboe wrote:
> > On Tue, Feb 11 2014, Christoph Hellwig wrote:
> > > 
> > 
> > looks like the last one got botched?
> 
> That was quilt sending along the series file which I created by doing
> 
> ls > series
> 
> just ignore it, hopefully one day quilt and I will get along fine..

Ignored, applied the other two, thanks!

-- 
Jens Axboe