[PATCH 5/6] scsi-mq: fix 'BUG_ON(count > sdb->table.nents)' in scsi_init_sgtable().

[PATCH 5/6] scsi-mq: fix 'BUG_ON(count > sdb->table.nents)' in scsi_init_sgtable().

[majianpeng]

Test cdrom on intel ahci, it will trigger this condition.
This is because scsi-mq don' test 'q->dma_drain_size &&
blk_rq_bytes(rq)' like in blk_peek_request().

Signed-off-by: Jianpeng Ma <majianpeng@gmail.com>
---
 drivers/scsi/scsi-mq.c |   10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/drivers/scsi/scsi-mq.c b/drivers/scsi/scsi-mq.c
index 933ba0f..78b05c1 100644
--- a/drivers/scsi/scsi-mq.c
+++ b/drivers/scsi/scsi-mq.c
@@ -35,6 +35,16 @@ static int scsi_mq_queue_rq(struct blk_mq_hw_ctx *hctx, struct request *rq)
 	sc->device = sdev;
 	sc->sense_buffer = sense_buf;
 	sc->mq_sgl = sg;
+
+	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++;
+	}
 	/*
 	 * Mark the end of the pre-allocated SGL based upon the
 	 * incoming number of physical segments from blk-mq
-- 
1.7.10.4

Re: [PATCH 5/6] scsi-mq: fix 'BUG_ON(count > sdb->table.nents)' in scsi_init_sgtable().

[Nicholas A. Bellinger]

On Fri, 2013-11-29 at 16:42 +0800, majianpeng wrote:
> Test cdrom on intel ahci, it will trigger this condition.
> This is because scsi-mq don' test 'q->dma_drain_size &&
> blk_rq_bytes(rq)' like in blk_peek_request().
> 
> Signed-off-by: Jianpeng Ma <majianpeng@gmail.com>
> ---
>  drivers/scsi/scsi-mq.c |   10 ++++++++++
>  1 file changed, 10 insertions(+)
> 

Applied, + squashed into current target-pending/scsi-mq code.

Thanks Jianpeng!

--nab

> diff --git a/drivers/scsi/scsi-mq.c b/drivers/scsi/scsi-mq.c
> index 933ba0f..78b05c1 100644
> --- a/drivers/scsi/scsi-mq.c
> +++ b/drivers/scsi/scsi-mq.c
> @@ -35,6 +35,16 @@ static int scsi_mq_queue_rq(struct blk_mq_hw_ctx *hctx, struct request *rq)
>  	sc->device = sdev;
>  	sc->sense_buffer = sense_buf;
>  	sc->mq_sgl = sg;
> +
> +	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++;
> +	}
>  	/*
>  	 * Mark the end of the pre-allocated SGL based upon the
>  	 * incoming number of physical segments from blk-mq