RFC: split scsi passthrough fields out of struct request

RFC: split scsi passthrough fields out of struct request

[Christoph Hellwig]

Hi all,

this series splits the support for SCSI passthrough commands from the
main struct request used all over the block layer into a separate
scsi_request structure that drivers that want to support SCSI passthough
need to embedded as the first thing into their request-private data,
similar to how we handle NVMe passthrough commands.

To support this I've added support for that the private data after
request structure to the legacy request path instead, so that it can
be treated the same way as the blk-mq path.  Compare to the current
scsi_cmnd allocator that actually is a major simplification.

There is one major show-stopper for the series as-is:  Request-based
device mapper currently allocate the request structures in the stacking
driver without a knowledge of the queue it's going to be submitte on
for the legacy request path.  It managed to avoid that issue for the
blk-mq path, but that currently can't be used on legacy request devices
for reasons I don't fully understand.  We'll need to figure out how
sort this out, but maybe that's a good opportunity to drop one or two
of the three different dm-mpath I/O paths? :)

Note that the first two patches are 4.10 material and have already
been submitted independently.

[PATCH 01/15] virtio_blk: avoid DMA to stack for the sense buffer

[Christoph Hellwig]

Most users of BLOCK_PC requests allocate the sense buffer on the stack,
so to avoid DMA to the stack copy them to a field in the heap allocated
virtblk_req structure.  Without that any attempt at SCSI passthrough I/O,
including the SG_IO ioctl from userspace will crash the kernel.  Note that
this includes running tools like hdparm even when the host does not have
SCSI passthrough enabled.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 drivers/block/virtio_blk.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c
index 5545a67..3c3b8f6 100644
--- a/drivers/block/virtio_blk.c
+++ b/drivers/block/virtio_blk.c
@@ -56,6 +56,7 @@ struct virtblk_req {
 	struct virtio_blk_outhdr out_hdr;
 	struct virtio_scsi_inhdr in_hdr;
 	u8 status;
+	u8 sense[SCSI_SENSE_BUFFERSIZE];
 	struct scatterlist sg[];
 };
 
@@ -102,7 +103,8 @@ static int __virtblk_add_req(struct virtqueue *vq,
 	}
 
 	if (type == cpu_to_virtio32(vq->vdev, VIRTIO_BLK_T_SCSI_CMD)) {
-		sg_init_one(&sense, vbr->req->sense, SCSI_SENSE_BUFFERSIZE);
+		memcpy(vbr->sense, vbr->req->sense, SCSI_SENSE_BUFFERSIZE);
+		sg_init_one(&sense, vbr->sense, SCSI_SENSE_BUFFERSIZE);
 		sgs[num_out + num_in++] = &sense;
 		sg_init_one(&inhdr, &vbr->in_hdr, sizeof(vbr->in_hdr));
 		sgs[num_out + num_in++] = &inhdr;
-- 
2.1.4

Re: [PATCH 01/15] virtio_blk: avoid DMA to stack for the sense buffer

[Johannes Thumshirn]

On Tue, Jan 10, 2017 at 04:06:06PM +0100, Christoph Hellwig wrote:
> Most users of BLOCK_PC requests allocate the sense buffer on the stack,
> so to avoid DMA to the stack copy them to a field in the heap allocated
> virtblk_req structure.  Without that any attempt at SCSI passthrough I/O,
> including the SG_IO ioctl from userspace will crash the kernel.  Note that
> this includes running tools like hdparm even when the host does not have
> SCSI passthrough enabled.
> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> ---

Isn't that one already queued in Jens' tree?

Byte,
	Johannes

-- 
Johannes Thumshirn                                          Storage
jthumshirn@suse.de                                +49 911 74053 689
SUSE LINUX GmbH, Maxfeldstr. 5, 90409 N�rnberg
GF: Felix Imend�rffer, Jane Smithard, Graham Norton
HRB 21284 (AG N�rnberg)
Key fingerprint = EC38 9CAB C2C4 F25D 8600 D0D0 0393 969D 2D76 0850

Re: [PATCH 01/15] virtio_blk: avoid DMA to stack for the sense buffer

[Christoph Hellwig]

On Wed, Jan 11, 2017 at 09:26:46AM +0100, Johannes Thumshirn wrote:
> Isn't that one already queued in Jens' tree?

Yes, it's now queued up.  Patch 2 was submitted as well and should
hopefully go into the next 4.10 RC.

[PATCH 02/15] nvme-rdma: fix nvme_rdma_queue_is_ready

[Christoph Hellwig]

Now that we don't abuse the cmd field in struct request for nvme command
passthrough this function needs to be converted to the proper accessor
as well.

Fixes: d49187e97e ("nvme: introduce struct nvme_request")
Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 drivers/nvme/host/rdma.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/nvme/host/rdma.c b/drivers/nvme/host/rdma.c
index f587af3..34e5648 100644
--- a/drivers/nvme/host/rdma.c
+++ b/drivers/nvme/host/rdma.c
@@ -1422,7 +1422,7 @@ static inline bool nvme_rdma_queue_is_ready(struct nvme_rdma_queue *queue,
 		struct request *rq)
 {
 	if (unlikely(!test_bit(NVME_RDMA_Q_LIVE, &queue->flags))) {
-		struct nvme_command *cmd = (struct nvme_command *)rq->cmd;
+		struct nvme_command *cmd = nvme_req(rq)->cmd;
 
 		if (rq->cmd_type != REQ_TYPE_DRV_PRIV ||
 		    cmd->common.opcode != nvme_fabrics_command ||
-- 
2.1.4

[PATCH 03/15] block: simplify blk_init_allocated_queue

[Christoph Hellwig]

Return an errno value instead of the passed in queue so that the callers
don't have to keep track of two queues, and move the assignment of the
request_fn and lock to the caller as passing them as argument doesn't
simplify anything.  While we're at it also remove two pointless NULL
assignments, given that the request structure is zeroed on allocation.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 block/blk-core.c       | 38 +++++++++++++++-----------------------
 drivers/md/dm-rq.c     |  3 ++-
 include/linux/blkdev.h |  3 +--
 3 files changed, 18 insertions(+), 26 deletions(-)

diff --git a/block/blk-core.c b/block/blk-core.c
index 61ba08c..761cee8 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -821,15 +821,19 @@ EXPORT_SYMBOL(blk_init_queue);
 struct request_queue *
 blk_init_queue_node(request_fn_proc *rfn, spinlock_t *lock, int node_id)
 {
-	struct request_queue *uninit_q, *q;
+	struct request_queue *q;
 
-	uninit_q = blk_alloc_queue_node(GFP_KERNEL, node_id);
-	if (!uninit_q)
+	q = blk_alloc_queue_node(GFP_KERNEL, node_id);
+	if (!q)
 		return NULL;
 
-	q = blk_init_allocated_queue(uninit_q, rfn, lock);
-	if (!q)
-		blk_cleanup_queue(uninit_q);
+	q->request_fn = rfn;
+	if (lock)
+		q->queue_lock = lock;
+	if (blk_init_allocated_queue(q) < 0) {
+		blk_cleanup_queue(q);
+		return NULL;
+	}
 
 	return q;
 }
@@ -837,30 +841,19 @@ EXPORT_SYMBOL(blk_init_queue_node);
 
 static blk_qc_t blk_queue_bio(struct request_queue *q, struct bio *bio);
 
-struct request_queue *
-blk_init_allocated_queue(struct request_queue *q, request_fn_proc *rfn,
-			 spinlock_t *lock)
-{
-	if (!q)
-		return NULL;
 
+int blk_init_allocated_queue(struct request_queue *q)
+{
 	q->fq = blk_alloc_flush_queue(q, NUMA_NO_NODE, 0);
 	if (!q->fq)
-		return NULL;
+		return -ENOMEM;
 
 	if (blk_init_rl(&q->root_rl, q, GFP_KERNEL))
 		goto fail;
 
 	INIT_WORK(&q->timeout_work, blk_timeout_work);
-	q->request_fn		= rfn;
-	q->prep_rq_fn		= NULL;
-	q->unprep_rq_fn		= NULL;
 	q->queue_flags		|= QUEUE_FLAG_DEFAULT;
 
-	/* Override internal queue lock with supplied lock pointer */
-	if (lock)
-		q->queue_lock		= lock;
-
 	/*
 	 * This also sets hw/phys segments, boundary and size
 	 */
@@ -878,13 +871,12 @@ blk_init_allocated_queue(struct request_queue *q, request_fn_proc *rfn,
 	}
 
 	mutex_unlock(&q->sysfs_lock);
-
-	return q;
+	return 0;
 
 fail:
 	blk_free_flush_queue(q->fq);
 	wbt_exit(q);
-	return NULL;
+	return -ENOMEM;
 }
 EXPORT_SYMBOL(blk_init_allocated_queue);
 
diff --git a/drivers/md/dm-rq.c b/drivers/md/dm-rq.c
index 9d7275f..93f6e9f 100644
--- a/drivers/md/dm-rq.c
+++ b/drivers/md/dm-rq.c
@@ -823,7 +823,8 @@ static void dm_old_request_fn(struct request_queue *q)
 int dm_old_init_request_queue(struct mapped_device *md)
 {
 	/* Fully initialize the queue */
-	if (!blk_init_allocated_queue(md->queue, dm_old_request_fn, NULL))
+	md->queue->request_fn = dm_old_request_fn;
+	if (blk_init_allocated_queue(md->queue) < 0)
 		return -EINVAL;
 
 	/* disable dm_old_request_fn's merge heuristic by default */
diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
index 8369564..9a6f3ab 100644
--- a/include/linux/blkdev.h
+++ b/include/linux/blkdev.h
@@ -1116,8 +1116,7 @@ extern void blk_unprep_request(struct request *);
 extern struct request_queue *blk_init_queue_node(request_fn_proc *rfn,
 					spinlock_t *lock, int node_id);
 extern struct request_queue *blk_init_queue(request_fn_proc *, spinlock_t *);
-extern struct request_queue *blk_init_allocated_queue(struct request_queue *,
-						      request_fn_proc *, spinlock_t *);
+extern int blk_init_allocated_queue(struct request_queue *);
 extern void blk_cleanup_queue(struct request_queue *);
 extern void blk_queue_make_request(struct request_queue *, make_request_fn *);
 extern void blk_queue_bounce_limit(struct request_queue *, u64);
-- 
2.1.4

[PATCH 04/15] block: allow specifying size for extra command data

[Christoph Hellwig]

This mirrors the blk-mq capabilities to allocate extra drivers-specific
data behind struct request by setting a cmd_size field, as well as having
a constructor / destructor for it.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 block/blk-core.c       | 59 ++++++++++++++++++++++++++++++++++++++++----------
 block/blk-flush.c      |  5 ++---
 block/blk-sysfs.c      |  7 ++++--
 include/linux/blkdev.h |  7 ++++++
 4 files changed, 61 insertions(+), 17 deletions(-)

diff --git a/block/blk-core.c b/block/blk-core.c
index 761cee8..7190bc1 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -604,17 +604,41 @@ void blk_cleanup_queue(struct request_queue *q)
 EXPORT_SYMBOL(blk_cleanup_queue);
 
 /* Allocate memory local to the request queue */
-static void *alloc_request_struct(gfp_t gfp_mask, void *data)
+static void *alloc_request_simple(gfp_t gfp_mask, void *data)
 {
-	int nid = (int)(long)data;
-	return kmem_cache_alloc_node(request_cachep, gfp_mask, nid);
+	struct request_queue *q = data;
+
+	return kmem_cache_alloc_node(request_cachep, gfp_mask, q->node);
 }
 
-static void free_request_struct(void *element, void *unused)
+static void free_request_simple(void *element, void *data)
 {
 	kmem_cache_free(request_cachep, element);
 }
 
+static void *alloc_request_size(gfp_t gfp_mask, void *data)
+{
+	struct request_queue *q = data;
+	struct request *rq;
+
+	rq = kmalloc_node(sizeof(struct request) + q->cmd_size, gfp_mask,
+			q->node);
+	if (rq && q->init_rq_fn && q->init_rq_fn(q, rq, gfp_mask) < 0) {
+		kfree(rq);
+		rq = NULL;
+	}
+	return rq;
+}
+
+static void free_request_size(void *element, void *data)
+{
+	struct request_queue *q = data;
+
+	if (q->exit_rq_fn)
+		q->exit_rq_fn(q, element);
+	kfree(element);
+}
+
 int blk_init_rl(struct request_list *rl, struct request_queue *q,
 		gfp_t gfp_mask)
 {
@@ -627,10 +651,15 @@ int blk_init_rl(struct request_list *rl, struct request_queue *q,
 	init_waitqueue_head(&rl->wait[BLK_RW_SYNC]);
 	init_waitqueue_head(&rl->wait[BLK_RW_ASYNC]);
 
-	rl->rq_pool = mempool_create_node(BLKDEV_MIN_RQ, alloc_request_struct,
-					  free_request_struct,
-					  (void *)(long)q->node, gfp_mask,
-					  q->node);
+	if (q->cmd_size) {
+		rl->rq_pool = mempool_create_node(BLKDEV_MIN_RQ,
+				alloc_request_size, free_request_size,
+				q, gfp_mask, q->node);
+	} else {
+		rl->rq_pool = mempool_create_node(BLKDEV_MIN_RQ,
+				alloc_request_simple, free_request_simple,
+				q, gfp_mask, q->node);
+	}
 	if (!rl->rq_pool)
 		return -ENOMEM;
 
@@ -844,12 +873,15 @@ static blk_qc_t blk_queue_bio(struct request_queue *q, struct bio *bio);
 
 int blk_init_allocated_queue(struct request_queue *q)
 {
-	q->fq = blk_alloc_flush_queue(q, NUMA_NO_NODE, 0);
+	q->fq = blk_alloc_flush_queue(q, NUMA_NO_NODE, q->cmd_size);
 	if (!q->fq)
 		return -ENOMEM;
 
+	if (q->init_rq_fn && q->init_rq_fn(q, q->fq->flush_rq, GFP_KERNEL))
+		goto out_free_flush_queue;
+
 	if (blk_init_rl(&q->root_rl, q, GFP_KERNEL))
-		goto fail;
+		goto out_exit_flush_rq;
 
 	INIT_WORK(&q->timeout_work, blk_timeout_work);
 	q->queue_flags		|= QUEUE_FLAG_DEFAULT;
@@ -867,13 +899,16 @@ int blk_init_allocated_queue(struct request_queue *q)
 	/* init elevator */
 	if (elevator_init(q, NULL)) {
 		mutex_unlock(&q->sysfs_lock);
-		goto fail;
+		goto out_exit_flush_rq;
 	}
 
 	mutex_unlock(&q->sysfs_lock);
 	return 0;
 
-fail:
+out_exit_flush_rq:
+	if (q->exit_rq_fn)
+		q->exit_rq_fn(q, q->fq->flush_rq);
+out_free_flush_queue:
 	blk_free_flush_queue(q->fq);
 	wbt_exit(q);
 	return -ENOMEM;
diff --git a/block/blk-flush.c b/block/blk-flush.c
index 20b7c7a..f390aea 100644
--- a/block/blk-flush.c
+++ b/block/blk-flush.c
@@ -545,11 +545,10 @@ struct blk_flush_queue *blk_alloc_flush_queue(struct request_queue *q,
 	if (!fq)
 		goto fail;
 
-	if (q->mq_ops) {
+	if (q->mq_ops)
 		spin_lock_init(&fq->mq_flush_lock);
-		rq_sz = round_up(rq_sz + cmd_size, cache_line_size());
-	}
 
+	rq_sz = round_up(rq_sz + cmd_size, cache_line_size());
 	fq->flush_rq = kzalloc_node(rq_sz, GFP_KERNEL, node);
 	if (!fq->flush_rq)
 		goto fail_rq;
diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c
index 1dbce05..894f773 100644
--- a/block/blk-sysfs.c
+++ b/block/blk-sysfs.c
@@ -814,10 +814,13 @@ static void blk_release_queue(struct kobject *kobj)
 	if (q->queue_tags)
 		__blk_queue_free_tags(q);
 
-	if (!q->mq_ops)
+	if (!q->mq_ops) {
+		if (q->exit_rq_fn)
+			q->exit_rq_fn(q, q->fq->flush_rq);
 		blk_free_flush_queue(q->fq);
-	else
+	} else {
 		blk_mq_release(q);
+	}
 
 	blk_trace_shutdown(q);
 
diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
index 9a6f3ab..be23859 100644
--- a/include/linux/blkdev.h
+++ b/include/linux/blkdev.h
@@ -271,6 +271,8 @@ typedef void (softirq_done_fn)(struct request *);
 typedef int (dma_drain_needed_fn)(struct request *);
 typedef int (lld_busy_fn) (struct request_queue *q);
 typedef int (bsg_job_fn) (struct bsg_job *);
+typedef int (init_rq_fn)(struct request_queue *, struct request *, gfp_t);
+typedef void (exit_rq_fn)(struct request_queue *, struct request *);
 
 enum blk_eh_timer_return {
 	BLK_EH_NOT_HANDLED,
@@ -406,6 +408,8 @@ struct request_queue {
 	rq_timed_out_fn		*rq_timed_out_fn;
 	dma_drain_needed_fn	*dma_drain_needed;
 	lld_busy_fn		*lld_busy_fn;
+	init_rq_fn		*init_rq_fn;
+	exit_rq_fn		*exit_rq_fn;
 
 	struct blk_mq_ops	*mq_ops;
 
@@ -570,6 +574,9 @@ struct request_queue {
 	struct bio_set		*bio_split;
 
 	bool			mq_sysfs_init_done;
+
+	size_t			cmd_size;
+	void			*rq_alloc_data;
 };
 
 #define QUEUE_FLAG_QUEUED	1	/* uses generic tag queueing */
-- 
2.1.4

[PATCH 05/15] dm: remove incomple BLOCK_PC support

[Christoph Hellwig]

DM tries to copy a few fields around for BLOCK_PC requests, but given
that no dm-target ever wires up scsi_cmd_ioctl BLOCK_PC can't actually
be sent to dm.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 drivers/md/dm-rq.c | 16 ----------------
 1 file changed, 16 deletions(-)

diff --git a/drivers/md/dm-rq.c b/drivers/md/dm-rq.c
index 93f6e9f..3f12916 100644
--- a/drivers/md/dm-rq.c
+++ b/drivers/md/dm-rq.c
@@ -270,19 +270,6 @@ static void dm_end_request(struct request *clone, int error)
 	struct mapped_device *md = tio->md;
 	struct request *rq = tio->orig;
 
-	if (rq->cmd_type == REQ_TYPE_BLOCK_PC) {
-		rq->errors = clone->errors;
-		rq->resid_len = clone->resid_len;
-
-		if (rq->sense)
-			/*
-			 * We are using the sense buffer of the original
-			 * request.
-			 * So setting the length of the sense data is enough.
-			 */
-			rq->sense_len = clone->sense_len;
-	}
-
 	free_rq_clone(clone);
 	rq_end_stats(md, rq);
 	if (!rq->q->mq_ops)
@@ -511,9 +498,6 @@ static int setup_clone(struct request *clone, struct request *rq,
 	if (r)
 		return r;
 
-	clone->cmd = rq->cmd;
-	clone->cmd_len = rq->cmd_len;
-	clone->sense = rq->sense;
 	clone->end_io = end_clone_request;
 	clone->end_io_data = tio;
 
-- 
2.1.4

Re: [PATCH 05/15] dm: remove incomple BLOCK_PC support

[Mike Snitzer]

On Tue, Jan 10 2017 at 10:06am -0500,
Christoph Hellwig <hch@lst.de> wrote:

> DM tries to copy a few fields around for BLOCK_PC requests, but given
> that no dm-target ever wires up scsi_cmd_ioctl BLOCK_PC can't actually
> be sent to dm.
> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> ---
>  drivers/md/dm-rq.c | 16 ----------------
>  1 file changed, 16 deletions(-)
> 
> diff --git a/drivers/md/dm-rq.c b/drivers/md/dm-rq.c
> index 93f6e9f..3f12916 100644
> --- a/drivers/md/dm-rq.c
> +++ b/drivers/md/dm-rq.c
> @@ -270,19 +270,6 @@ static void dm_end_request(struct request *clone, int error)
>  	struct mapped_device *md = tio->md;
>  	struct request *rq = tio->orig;
>  
> -	if (rq->cmd_type == REQ_TYPE_BLOCK_PC) {
> -		rq->errors = clone->errors;
> -		rq->resid_len = clone->resid_len;
> -
> -		if (rq->sense)
> -			/*
> -			 * We are using the sense buffer of the original
> -			 * request.
> -			 * So setting the length of the sense data is enough.
> -			 */
> -			rq->sense_len = clone->sense_len;
> -	}
> -
>  	free_rq_clone(clone);
>  	rq_end_stats(md, rq);
>  	if (!rq->q->mq_ops)
> @@ -511,9 +498,6 @@ static int setup_clone(struct request *clone, struct request *rq,
>  	if (r)
>  		return r;
>  
> -	clone->cmd = rq->cmd;
> -	clone->cmd_len = rq->cmd_len;
> -	clone->sense = rq->sense;
>  	clone->end_io = end_clone_request;
>  	clone->end_io_data = tio;
>  

I'm not following your reasoning.

dm_blk_ioctl calls __blkdev_driver_ioctl and will call scsi_cmd_ioctl
(sd_ioctl -> scsi_cmd_blk_ioctl -> scsi_cmd_ioctl) if DM's underlying
block device is a scsi device.

Re: [PATCH 05/15] dm: remove incomple BLOCK_PC support

[Christoph Hellwig]

On Wed, Jan 11, 2017 at 08:09:37PM -0500, Mike Snitzer wrote:
> I'm not following your reasoning.
> 
> dm_blk_ioctl calls __blkdev_driver_ioctl and will call scsi_cmd_ioctl
> (sd_ioctl -> scsi_cmd_blk_ioctl -> scsi_cmd_ioctl) if DM's underlying
> block device is a scsi device.

Yes, it it does.  But scsi_cmd_ioctl as called from sd_ioctl will
operate entirely on the SCSI request_queue - dm-mpath will never see
the BLOCK_PC request generated by it.

Re: [PATCH 05/15] dm: remove incomple BLOCK_PC support

[Mike Snitzer]

On Thu, Jan 12 2017 at  3:00am -0500,
Christoph Hellwig <hch@lst.de> wrote:

> On Wed, Jan 11, 2017 at 08:09:37PM -0500, Mike Snitzer wrote:
> > I'm not following your reasoning.
> > 
> > dm_blk_ioctl calls __blkdev_driver_ioctl and will call scsi_cmd_ioctl
> > (sd_ioctl -> scsi_cmd_blk_ioctl -> scsi_cmd_ioctl) if DM's underlying
> > block device is a scsi device.
> 
> Yes, it it does.  But scsi_cmd_ioctl as called from sd_ioctl will
> operate entirely on the SCSI request_queue - dm-mpath will never see
> the BLOCK_PC request generated by it.

I lost sight of the fact that BLOCK_PC requests are sent down via the
normal request submission (and not the ioctl path).  So my previous
reply wasn't relevant.

What is "incomplete" about request-based DM's BLOCK_PC support?

This code goes back to when request-based DM multipath was first
introduced via commit cec47e3d4a -- but I've never used the BLOCK_PC
requests for SCSI pass through myself.  I don't know who is using
it.. are you aware of some upper layer filesystem or userspace
submission path for these BLOCK_PC requests that they'd be passing
through DM?

I'm also missing how you're saying the new blk-mq request-based DM will
work with your new model.  I appreciate that we get the request from the
underlying blk-mq request_queue and it'll be properly sized.  But
wouldn't we need to pass data back up for these SCSI pass-through
requests?  So wouldn't the top-level multipath request_queue need to
setup cmd_size?

Sorry for the naive questions (that clearly speak to me not
understanding how this aspect of the block and SCSI code work).. but I'd
like to understand where DM will be lacking going forward.

Re: [PATCH 05/15] dm: remove incomple BLOCK_PC support

[Christoph Hellwig]

On Thu, Jan 12, 2017 at 05:28:45PM -0500, Mike Snitzer wrote:
> What is "incomplete" about request-based DM's BLOCK_PC support?

BLOCK_PC requests are always aomething issued by the driver itself
(for a broad defintion of the driver, aka everything under the block
layer that works together is a driver for this purpose, e.g. all
of the SCSI subsystem).  If a driver doesn't issue BLOCK_PC
requests itself or through library functions only called from the
driver (e.g. scsi_cmd_ioctl) it is incomplete because it can't
be used.

> I'm also missing how you're saying the new blk-mq request-based DM will
> work with your new model.  I appreciate that we get the request from the
> underlying blk-mq request_queue and it'll be properly sized.  But
> wouldn't we need to pass data back up for these SCSI pass-through
> requests?  So wouldn't the top-level multipath request_queue need to
> setup cmd_size?

As said above - supporting BLOCK_PC for dm does not make sense, as
it's an internal passthrough mechanism for driver internal use.
It just happened we standardized it at the block layer because SCSI
commands are a standard supported by a few different drivers, e.g.
SCSI itself, the old ide code for ATAPI and CCISS and virtio_blk
which primarily are block drivers but allow some SCSI passthrough.

To be honest I'd love to just fold BLOCK_PC into the SCSI layer sooner
or later - the old IDE code and CCISS should die off sooner or later,
and virtio_blk scsi passthrough was a horrible idea to start with
(and I say that as the person who had the idea back then and implemented
it..)

> Sorry for the naive questions (that clearly speak to me not
> understanding how this aspect of the block and SCSI code work).. but I'd
> like to understand where DM will be lacking going forward.

At least in terms of BLOCK_PC nothing will change for dm, the code
was simply dead on arrival.  Maybe I should change the subject
to say that more clearly.

[PATCH 06/15] scsi_dh_rdac: switch to scsi_execute_req_flags()

[Christoph Hellwig]

From: Hannes Reinecke <hare@suse.de>

Switch to scsi_execute_req_flags() and scsi_get_vpd_page() instead of
open-coding it.  Using scsi_execute_req_flags() will set REQ_QUIET and
REQ_PREEMPT, but this is okay as we're evaluating the errors anyway and
should be able to send the command even if the device is quiesced.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 drivers/scsi/device_handler/scsi_dh_rdac.c | 174 +++++++++--------------------
 1 file changed, 51 insertions(+), 123 deletions(-)

diff --git a/drivers/scsi/device_handler/scsi_dh_rdac.c b/drivers/scsi/device_handler/scsi_dh_rdac.c
index 00d9c32..b64eaae 100644
--- a/drivers/scsi/device_handler/scsi_dh_rdac.c
+++ b/drivers/scsi/device_handler/scsi_dh_rdac.c
@@ -205,7 +205,6 @@ struct rdac_dh_data {
 #define RDAC_NON_PREFERRED	1
 	char			preferred;
 
-	unsigned char		sense[SCSI_SENSE_BUFFERSIZE];
 	union			{
 		struct c2_inquiry c2;
 		struct c4_inquiry c4;
@@ -262,40 +261,12 @@ do { \
 		sdev_printk(KERN_INFO, sdev, RDAC_NAME ": " f "\n", ## arg); \
 } while (0);
 
-static struct request *get_rdac_req(struct scsi_device *sdev,
-			void *buffer, unsigned buflen, int rw)
+static unsigned int rdac_failover_get(struct rdac_controller *ctlr,
+				      struct list_head *list,
+				      unsigned char *cdb)
 {
-	struct request *rq;
-	struct request_queue *q = sdev->request_queue;
-
-	rq = blk_get_request(q, rw, GFP_NOIO);
-
-	if (IS_ERR(rq)) {
-		sdev_printk(KERN_INFO, sdev,
-				"get_rdac_req: blk_get_request failed.\n");
-		return NULL;
-	}
-	blk_rq_set_block_pc(rq);
-
-	if (buflen && blk_rq_map_kern(q, rq, buffer, buflen, GFP_NOIO)) {
-		blk_put_request(rq);
-		sdev_printk(KERN_INFO, sdev,
-				"get_rdac_req: blk_rq_map_kern failed.\n");
-		return NULL;
-	}
-
-	rq->cmd_flags |= REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT |
-			 REQ_FAILFAST_DRIVER;
-	rq->retries = RDAC_RETRIES;
-	rq->timeout = RDAC_TIMEOUT;
-
-	return rq;
-}
-
-static struct request *rdac_failover_get(struct scsi_device *sdev,
-			struct rdac_dh_data *h, struct list_head *list)
-{
-	struct request *rq;
+	struct scsi_device *sdev = ctlr->ms_sdev;
+	struct rdac_dh_data *h = sdev->handler_data;
 	struct rdac_mode_common *common;
 	unsigned data_size;
 	struct rdac_queue_data *qdata;
@@ -332,27 +303,17 @@ static struct request *rdac_failover_get(struct scsi_device *sdev,
 		lun_table[qdata->h->lun] = 0x81;
 	}
 
-	/* get request for block layer packet command */
-	rq = get_rdac_req(sdev, &h->ctlr->mode_select, data_size, WRITE);
-	if (!rq)
-		return NULL;
-
 	/* Prepare the command. */
 	if (h->ctlr->use_ms10) {
-		rq->cmd[0] = MODE_SELECT_10;
-		rq->cmd[7] = data_size >> 8;
-		rq->cmd[8] = data_size & 0xff;
+		cdb[0] = MODE_SELECT_10;
+		cdb[7] = data_size >> 8;
+		cdb[8] = data_size & 0xff;
 	} else {
-		rq->cmd[0] = MODE_SELECT;
-		rq->cmd[4] = data_size;
+		cdb[0] = MODE_SELECT;
+		cdb[4] = data_size;
 	}
-	rq->cmd_len = COMMAND_SIZE(rq->cmd[0]);
-
-	rq->sense = h->sense;
-	memset(rq->sense, 0, SCSI_SENSE_BUFFERSIZE);
-	rq->sense_len = 0;
 
-	return rq;
+	return data_size;
 }
 
 static void release_controller(struct kref *kref)
@@ -400,46 +361,14 @@ static struct rdac_controller *get_controller(int index, char *array_name,
 	return ctlr;
 }
 
-static int submit_inquiry(struct scsi_device *sdev, int page_code,
-			  unsigned int len, struct rdac_dh_data *h)
-{
-	struct request *rq;
-	struct request_queue *q = sdev->request_queue;
-	int err = SCSI_DH_RES_TEMP_UNAVAIL;
-
-	rq = get_rdac_req(sdev, &h->inq, len, READ);
-	if (!rq)
-		goto done;
-
-	/* Prepare the command. */
-	rq->cmd[0] = INQUIRY;
-	rq->cmd[1] = 1;
-	rq->cmd[2] = page_code;
-	rq->cmd[4] = len;
-	rq->cmd_len = COMMAND_SIZE(INQUIRY);
-
-	rq->sense = h->sense;
-	memset(rq->sense, 0, SCSI_SENSE_BUFFERSIZE);
-	rq->sense_len = 0;
-
-	err = blk_execute_rq(q, NULL, rq, 1);
-	if (err == -EIO)
-		err = SCSI_DH_IO;
-
-	blk_put_request(rq);
-done:
-	return err;
-}
-
 static int get_lun_info(struct scsi_device *sdev, struct rdac_dh_data *h,
 			char *array_name, u8 *array_id)
 {
-	int err, i;
-	struct c8_inquiry *inqp;
+	int err = SCSI_DH_IO, i;
+	struct c8_inquiry *inqp = &h->inq.c8;
 
-	err = submit_inquiry(sdev, 0xC8, sizeof(struct c8_inquiry), h);
-	if (err == SCSI_DH_OK) {
-		inqp = &h->inq.c8;
+	if (!scsi_get_vpd_page(sdev, 0xC8, (unsigned char *)inqp,
+			       sizeof(struct c8_inquiry))) {
 		if (inqp->page_code != 0xc8)
 			return SCSI_DH_NOSYS;
 		if (inqp->page_id[0] != 'e' || inqp->page_id[1] != 'd' ||
@@ -453,20 +382,20 @@ static int get_lun_info(struct scsi_device *sdev, struct rdac_dh_data *h,
 		*(array_name+ARRAY_LABEL_LEN-1) = '\0';
 		memset(array_id, 0, UNIQUE_ID_LEN);
 		memcpy(array_id, inqp->array_unique_id, inqp->array_uniq_id_len);
+		err = SCSI_DH_OK;
 	}
 	return err;
 }
 
 static int check_ownership(struct scsi_device *sdev, struct rdac_dh_data *h)
 {
-	int err, access_state;
+	int err = SCSI_DH_IO, access_state;
 	struct rdac_dh_data *tmp;
-	struct c9_inquiry *inqp;
+	struct c9_inquiry *inqp = &h->inq.c9;
 
 	h->state = RDAC_STATE_ACTIVE;
-	err = submit_inquiry(sdev, 0xC9, sizeof(struct c9_inquiry), h);
-	if (err == SCSI_DH_OK) {
-		inqp = &h->inq.c9;
+	if (!scsi_get_vpd_page(sdev, 0xC9, (unsigned char *)inqp,
+			       sizeof(struct c9_inquiry))) {
 		/* detect the operating mode */
 		if ((inqp->avte_cvp >> 5) & 0x1)
 			h->mode = RDAC_MODE_IOSHIP; /* LUN in IOSHIP mode */
@@ -501,6 +430,7 @@ static int check_ownership(struct scsi_device *sdev, struct rdac_dh_data *h)
 			tmp->sdev->access_state = access_state;
 		}
 		rcu_read_unlock();
+		err = SCSI_DH_OK;
 	}
 
 	return err;
@@ -509,12 +439,11 @@ static int check_ownership(struct scsi_device *sdev, struct rdac_dh_data *h)
 static int initialize_controller(struct scsi_device *sdev,
 		struct rdac_dh_data *h, char *array_name, u8 *array_id)
 {
-	int err, index;
-	struct c4_inquiry *inqp;
+	int err = SCSI_DH_IO, index;
+	struct c4_inquiry *inqp = &h->inq.c4;
 
-	err = submit_inquiry(sdev, 0xC4, sizeof(struct c4_inquiry), h);
-	if (err == SCSI_DH_OK) {
-		inqp = &h->inq.c4;
+	if (!scsi_get_vpd_page(sdev, 0xC4, (unsigned char *)inqp,
+			       sizeof(struct c4_inquiry))) {
 		/* get the controller index */
 		if (inqp->slot_id[1] == 0x31)
 			index = 0;
@@ -530,18 +459,18 @@ static int initialize_controller(struct scsi_device *sdev,
 			h->sdev = sdev;
 		}
 		spin_unlock(&list_lock);
+		err = SCSI_DH_OK;
 	}
 	return err;
 }
 
 static int set_mode_select(struct scsi_device *sdev, struct rdac_dh_data *h)
 {
-	int err;
-	struct c2_inquiry *inqp;
+	int err = SCSI_DH_IO;
+	struct c2_inquiry *inqp = &h->inq.c2;
 
-	err = submit_inquiry(sdev, 0xC2, sizeof(struct c2_inquiry), h);
-	if (err == SCSI_DH_OK) {
-		inqp = &h->inq.c2;
+	if (!scsi_get_vpd_page(sdev, 0xC2, (unsigned char *)inqp,
+			       sizeof(struct c2_inquiry))) {
 		/*
 		 * If more than MODE6_MAX_LUN luns are supported, use
 		 * mode select 10
@@ -550,36 +479,35 @@ static int set_mode_select(struct scsi_device *sdev, struct rdac_dh_data *h)
 			h->ctlr->use_ms10 = 1;
 		else
 			h->ctlr->use_ms10 = 0;
+		err = SCSI_DH_OK;
 	}
 	return err;
 }
 
 static int mode_select_handle_sense(struct scsi_device *sdev,
-					unsigned char *sensebuf)
+				    struct scsi_sense_hdr *sense_hdr)
 {
-	struct scsi_sense_hdr sense_hdr;
-	int err = SCSI_DH_IO, ret;
+	int err = SCSI_DH_IO;
 	struct rdac_dh_data *h = sdev->handler_data;
 
-	ret = scsi_normalize_sense(sensebuf, SCSI_SENSE_BUFFERSIZE, &sense_hdr);
-	if (!ret)
+	if (!scsi_sense_valid(sense_hdr))
 		goto done;
 
-	switch (sense_hdr.sense_key) {
+	switch (sense_hdr->sense_key) {
 	case NO_SENSE:
 	case ABORTED_COMMAND:
 	case UNIT_ATTENTION:
 		err = SCSI_DH_RETRY;
 		break;
 	case NOT_READY:
-		if (sense_hdr.asc == 0x04 && sense_hdr.ascq == 0x01)
+		if (sense_hdr->asc == 0x04 && sense_hdr->ascq == 0x01)
 			/* LUN Not Ready and is in the Process of Becoming
 			 * Ready
 			 */
 			err = SCSI_DH_RETRY;
 		break;
 	case ILLEGAL_REQUEST:
-		if (sense_hdr.asc == 0x91 && sense_hdr.ascq == 0x36)
+		if (sense_hdr->asc == 0x91 && sense_hdr->ascq == 0x36)
 			/*
 			 * Command Lock contention
 			 */
@@ -592,7 +520,7 @@ static int mode_select_handle_sense(struct scsi_device *sdev,
 	RDAC_LOG(RDAC_LOG_FAILOVER, sdev, "array %s, ctlr %d, "
 		"MODE_SELECT returned with sense %02x/%02x/%02x",
 		(char *) h->ctlr->array_name, h->ctlr->index,
-		sense_hdr.sense_key, sense_hdr.asc, sense_hdr.ascq);
+		sense_hdr->sense_key, sense_hdr->asc, sense_hdr->ascq);
 
 done:
 	return err;
@@ -602,13 +530,16 @@ static void send_mode_select(struct work_struct *work)
 {
 	struct rdac_controller *ctlr =
 		container_of(work, struct rdac_controller, ms_work);
-	struct request *rq;
 	struct scsi_device *sdev = ctlr->ms_sdev;
 	struct rdac_dh_data *h = sdev->handler_data;
-	struct request_queue *q = sdev->request_queue;
-	int err, retry_cnt = RDAC_RETRY_COUNT;
+	int err = SCSI_DH_OK, retry_cnt = RDAC_RETRY_COUNT;
 	struct rdac_queue_data *tmp, *qdata;
 	LIST_HEAD(list);
+	unsigned char cdb[COMMAND_SIZE(MODE_SELECT_10)];
+	struct scsi_sense_hdr sshdr;
+	unsigned int data_size;
+	u64 req_flags = REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT |
+		REQ_FAILFAST_DRIVER;
 
 	spin_lock(&ctlr->ms_lock);
 	list_splice_init(&ctlr->ms_head, &list);
@@ -616,21 +547,19 @@ static void send_mode_select(struct work_struct *work)
 	ctlr->ms_sdev = NULL;
 	spin_unlock(&ctlr->ms_lock);
 
-retry:
-	err = SCSI_DH_RES_TEMP_UNAVAIL;
-	rq = rdac_failover_get(sdev, h, &list);
-	if (!rq)
-		goto done;
+ retry:
+	data_size = rdac_failover_get(ctlr, &list, cdb);
 
 	RDAC_LOG(RDAC_LOG_FAILOVER, sdev, "array %s, ctlr %d, "
 		"%s MODE_SELECT command",
 		(char *) h->ctlr->array_name, h->ctlr->index,
 		(retry_cnt == RDAC_RETRY_COUNT) ? "queueing" : "retrying");
 
-	err = blk_execute_rq(q, NULL, rq, 1);
-	blk_put_request(rq);
-	if (err != SCSI_DH_OK) {
-		err = mode_select_handle_sense(sdev, h->sense);
+	if (scsi_execute_req_flags(sdev, cdb, DMA_TO_DEVICE,
+				   &h->ctlr->mode_select, data_size, &sshdr,
+				   RDAC_TIMEOUT * HZ,
+				   RDAC_RETRIES, NULL, req_flags, 0)) {
+		err = mode_select_handle_sense(sdev, &sshdr);
 		if (err == SCSI_DH_RETRY && retry_cnt--)
 			goto retry;
 		if (err == SCSI_DH_IMM_RETRY)
@@ -643,7 +572,6 @@ static void send_mode_select(struct work_struct *work)
 				(char *) h->ctlr->array_name, h->ctlr->index);
 	}
 
-done:
 	list_for_each_entry_safe(qdata, tmp, &list, entry) {
 		list_del(&qdata->entry);
 		if (err == SCSI_DH_OK)
-- 
2.1.4

[PATCH 07/15] scsi_dh_emc: switch to scsi_execute_req_flags()

[Christoph Hellwig]

From: Hannes Reinecke <hare@suse.de>

Switch to scsi_execute_req_flags() and scsi_get_vpd_page() instead of
open-coding it.  Using scsi_execute_req_flags() will set REQ_QUIET and
REQ_PREEMPT, but this is okay as we're evaluating the errors anyway and
should be able to send the command even if the device is quiesced.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 drivers/scsi/device_handler/scsi_dh_emc.c | 247 +++++++-----------------------
 1 file changed, 56 insertions(+), 191 deletions(-)

diff --git a/drivers/scsi/device_handler/scsi_dh_emc.c b/drivers/scsi/device_handler/scsi_dh_emc.c
index 5b80746..4a7679f 100644
--- a/drivers/scsi/device_handler/scsi_dh_emc.c
+++ b/drivers/scsi/device_handler/scsi_dh_emc.c
@@ -88,12 +88,6 @@ struct clariion_dh_data {
 	 */
 	unsigned char buffer[CLARIION_BUFFER_SIZE];
 	/*
-	 * SCSI sense buffer for commands -- assumes serial issuance
-	 * and completion sequence of all commands for same multipath.
-	 */
-	unsigned char sense[SCSI_SENSE_BUFFERSIZE];
-	unsigned int senselen;
-	/*
 	 * LUN state
 	 */
 	int lun_state;
@@ -116,44 +110,38 @@ struct clariion_dh_data {
 /*
  * Parse MODE_SELECT cmd reply.
  */
-static int trespass_endio(struct scsi_device *sdev, char *sense)
+static int trespass_endio(struct scsi_device *sdev,
+			  struct scsi_sense_hdr *sshdr)
 {
 	int err = SCSI_DH_IO;
-	struct scsi_sense_hdr sshdr;
-
-	if (!scsi_normalize_sense(sense, SCSI_SENSE_BUFFERSIZE, &sshdr)) {
-		sdev_printk(KERN_ERR, sdev, "%s: Found valid sense data 0x%2x, "
-			    "0x%2x, 0x%2x while sending CLARiiON trespass "
-			    "command.\n", CLARIION_NAME, sshdr.sense_key,
-			    sshdr.asc, sshdr.ascq);
 
-		if ((sshdr.sense_key == 0x05) && (sshdr.asc == 0x04) &&
-		     (sshdr.ascq == 0x00)) {
-			/*
-			 * Array based copy in progress -- do not send
-			 * mode_select or copy will be aborted mid-stream.
-			 */
-			sdev_printk(KERN_INFO, sdev, "%s: Array Based Copy in "
-				    "progress while sending CLARiiON trespass "
-				    "command.\n", CLARIION_NAME);
-			err = SCSI_DH_DEV_TEMP_BUSY;
-		} else if ((sshdr.sense_key == 0x02) && (sshdr.asc == 0x04) &&
-			    (sshdr.ascq == 0x03)) {
-			/*
-			 * LUN Not Ready - Manual Intervention Required
-			 * indicates in-progress ucode upgrade (NDU).
-			 */
-			sdev_printk(KERN_INFO, sdev, "%s: Detected in-progress "
-				    "ucode upgrade NDU operation while sending "
-				    "CLARiiON trespass command.\n", CLARIION_NAME);
-			err = SCSI_DH_DEV_TEMP_BUSY;
-		} else
-			err = SCSI_DH_DEV_FAILED;
-	} else {
-		sdev_printk(KERN_INFO, sdev,
-			    "%s: failed to send MODE SELECT, no sense available\n",
-			    CLARIION_NAME);
-	}
+	sdev_printk(KERN_ERR, sdev, "%s: Found valid sense data 0x%2x, "
+		    "0x%2x, 0x%2x while sending CLARiiON trespass "
+		    "command.\n", CLARIION_NAME, sshdr->sense_key,
+		    sshdr->asc, sshdr->ascq);
+
+	if (sshdr->sense_key == 0x05 && sshdr->asc == 0x04 &&
+	    sshdr->ascq == 0x00) {
+		/*
+		 * Array based copy in progress -- do not send
+		 * mode_select or copy will be aborted mid-stream.
+		 */
+		sdev_printk(KERN_INFO, sdev, "%s: Array Based Copy in "
+			    "progress while sending CLARiiON trespass "
+			    "command.\n", CLARIION_NAME);
+		err = SCSI_DH_DEV_TEMP_BUSY;
+	} else if (sshdr->sense_key == 0x02 && sshdr->asc == 0x04 &&
+		   sshdr->ascq == 0x03) {
+		/*
+		 * LUN Not Ready - Manual Intervention Required
+		 * indicates in-progress ucode upgrade (NDU).
+		 */
+		sdev_printk(KERN_INFO, sdev, "%s: Detected in-progress "
+			    "ucode upgrade NDU operation while sending "
+			    "CLARiiON trespass command.\n", CLARIION_NAME);
+		err = SCSI_DH_DEV_TEMP_BUSY;
+	} else
+		err = SCSI_DH_DEV_FAILED;
 	return err;
 }
 
@@ -257,103 +245,15 @@ static char * parse_sp_model(struct scsi_device *sdev, unsigned char *buffer)
 	return sp_model;
 }
 
-/*
- * Get block request for REQ_BLOCK_PC command issued to path.  Currently
- * limited to MODE_SELECT (trespass) and INQUIRY (VPD page 0xC0) commands.
- *
- * Uses data and sense buffers in hardware handler context structure and
- * assumes serial servicing of commands, both issuance and completion.
- */
-static struct request *get_req(struct scsi_device *sdev, int cmd,
-				unsigned char *buffer)
-{
-	struct request *rq;
-	int len = 0;
-
-	rq = blk_get_request(sdev->request_queue,
-			(cmd != INQUIRY) ? WRITE : READ, GFP_NOIO);
-	if (IS_ERR(rq)) {
-		sdev_printk(KERN_INFO, sdev, "get_req: blk_get_request failed");
-		return NULL;
-	}
-
-	blk_rq_set_block_pc(rq);
-	rq->cmd_len = COMMAND_SIZE(cmd);
-	rq->cmd[0] = cmd;
-
-	switch (cmd) {
-	case MODE_SELECT:
-		len = sizeof(short_trespass);
-		rq->cmd[1] = 0x10;
-		rq->cmd[4] = len;
-		break;
-	case MODE_SELECT_10:
-		len = sizeof(long_trespass);
-		rq->cmd[1] = 0x10;
-		rq->cmd[8] = len;
-		break;
-	case INQUIRY:
-		len = CLARIION_BUFFER_SIZE;
-		rq->cmd[4] = len;
-		memset(buffer, 0, len);
-		break;
-	default:
-		BUG_ON(1);
-		break;
-	}
-
-	rq->cmd_flags |= REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT |
-			 REQ_FAILFAST_DRIVER;
-	rq->timeout = CLARIION_TIMEOUT;
-	rq->retries = CLARIION_RETRIES;
-
-	if (blk_rq_map_kern(rq->q, rq, buffer, len, GFP_NOIO)) {
-		blk_put_request(rq);
-		return NULL;
-	}
-
-	return rq;
-}
-
-static int send_inquiry_cmd(struct scsi_device *sdev, int page,
-			    struct clariion_dh_data *csdev)
-{
-	struct request *rq = get_req(sdev, INQUIRY, csdev->buffer);
-	int err;
-
-	if (!rq)
-		return SCSI_DH_RES_TEMP_UNAVAIL;
-
-	rq->sense = csdev->sense;
-	memset(rq->sense, 0, SCSI_SENSE_BUFFERSIZE);
-	rq->sense_len = csdev->senselen = 0;
-
-	rq->cmd[0] = INQUIRY;
-	if (page != 0) {
-		rq->cmd[1] = 1;
-		rq->cmd[2] = page;
-	}
-	err = blk_execute_rq(sdev->request_queue, NULL, rq, 1);
-	if (err == -EIO) {
-		sdev_printk(KERN_INFO, sdev,
-			    "%s: failed to send %s INQUIRY: %x\n",
-			    CLARIION_NAME, page?"EVPD":"standard",
-			    rq->errors);
-		csdev->senselen = rq->sense_len;
-		err = SCSI_DH_IO;
-	}
-
-	blk_put_request(rq);
-
-	return err;
-}
-
 static int send_trespass_cmd(struct scsi_device *sdev,
 			    struct clariion_dh_data *csdev)
 {
-	struct request *rq;
 	unsigned char *page22;
-	int err, len, cmd;
+	unsigned char cdb[COMMAND_SIZE(MODE_SELECT)];
+	int err, res = SCSI_DH_OK, len;
+	struct scsi_sense_hdr sshdr;
+	u64 req_flags = REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT |
+		REQ_FAILFAST_DRIVER;
 
 	if (csdev->flags & CLARIION_SHORT_TRESPASS) {
 		page22 = short_trespass;
@@ -361,40 +261,37 @@ static int send_trespass_cmd(struct scsi_device *sdev,
 			/* Set Honor Reservations bit */
 			page22[6] |= 0x80;
 		len = sizeof(short_trespass);
-		cmd = MODE_SELECT;
+		cdb[0] = MODE_SELECT;
+		cdb[1] = 0x10;
+		cdb[4] = len;
 	} else {
 		page22 = long_trespass;
 		if (!(csdev->flags & CLARIION_HONOR_RESERVATIONS))
 			/* Set Honor Reservations bit */
 			page22[10] |= 0x80;
 		len = sizeof(long_trespass);
-		cmd = MODE_SELECT_10;
+		cdb[0] = MODE_SELECT_10;
+		cdb[8] = len;
 	}
 	BUG_ON((len > CLARIION_BUFFER_SIZE));
 	memcpy(csdev->buffer, page22, len);
 
-	rq = get_req(sdev, cmd, csdev->buffer);
-	if (!rq)
-		return SCSI_DH_RES_TEMP_UNAVAIL;
-
-	rq->sense = csdev->sense;
-	memset(rq->sense, 0, SCSI_SENSE_BUFFERSIZE);
-	rq->sense_len = csdev->senselen = 0;
-
-	err = blk_execute_rq(sdev->request_queue, NULL, rq, 1);
-	if (err == -EIO) {
-		if (rq->sense_len) {
-			err = trespass_endio(sdev, csdev->sense);
-		} else {
+	err = scsi_execute_req_flags(sdev, cdb, DMA_TO_DEVICE,
+				     csdev->buffer, len, &sshdr,
+				     CLARIION_TIMEOUT * HZ, CLARIION_RETRIES,
+				     NULL, req_flags, 0);
+	if (err) {
+		if (scsi_sense_valid(&sshdr))
+			res = trespass_endio(sdev, &sshdr);
+		else {
 			sdev_printk(KERN_INFO, sdev,
 				    "%s: failed to send MODE SELECT: %x\n",
-				    CLARIION_NAME, rq->errors);
+				    CLARIION_NAME, err);
+			res = SCSI_DH_IO;
 		}
 	}
 
-	blk_put_request(rq);
-
-	return err;
+	return res;
 }
 
 static int clariion_check_sense(struct scsi_device *sdev,
@@ -464,21 +361,7 @@ static int clariion_std_inquiry(struct scsi_device *sdev,
 	int err;
 	char *sp_model;
 
-	err = send_inquiry_cmd(sdev, 0, csdev);
-	if (err != SCSI_DH_OK && csdev->senselen) {
-		struct scsi_sense_hdr sshdr;
-
-		if (scsi_normalize_sense(csdev->sense, SCSI_SENSE_BUFFERSIZE,
-					 &sshdr)) {
-			sdev_printk(KERN_ERR, sdev, "%s: INQUIRY sense code "
-				    "%02x/%02x/%02x\n", CLARIION_NAME,
-				    sshdr.sense_key, sshdr.asc, sshdr.ascq);
-		}
-		err = SCSI_DH_IO;
-		goto out;
-	}
-
-	sp_model = parse_sp_model(sdev, csdev->buffer);
+	sp_model = parse_sp_model(sdev, sdev->inquiry);
 	if (!sp_model) {
 		err = SCSI_DH_DEV_UNSUPP;
 		goto out;
@@ -500,30 +383,12 @@ static int clariion_std_inquiry(struct scsi_device *sdev,
 static int clariion_send_inquiry(struct scsi_device *sdev,
 				 struct clariion_dh_data *csdev)
 {
-	int err, retry = CLARIION_RETRIES;
-
-retry:
-	err = send_inquiry_cmd(sdev, 0xC0, csdev);
-	if (err != SCSI_DH_OK && csdev->senselen) {
-		struct scsi_sense_hdr sshdr;
-
-		err = scsi_normalize_sense(csdev->sense, SCSI_SENSE_BUFFERSIZE,
-					   &sshdr);
-		if (!err)
-			return SCSI_DH_IO;
-
-		err = clariion_check_sense(sdev, &sshdr);
-		if (retry > 0 && err == ADD_TO_MLQUEUE) {
-			retry--;
-			goto retry;
-		}
-		sdev_printk(KERN_ERR, sdev, "%s: INQUIRY sense code "
-			    "%02x/%02x/%02x\n", CLARIION_NAME,
-			      sshdr.sense_key, sshdr.asc, sshdr.ascq);
-		err = SCSI_DH_IO;
-	} else {
+	int err = SCSI_DH_IO;
+
+	if (!scsi_get_vpd_page(sdev, 0xC0, csdev->buffer,
+			       CLARIION_BUFFER_SIZE))
 		err = parse_sp_info_reply(sdev, csdev);
-	}
+
 	return err;
 }
 
-- 
2.1.4

[PATCH 08/15] scsi_dh_hp_sw: switch to scsi_execute_req_flags()

[Christoph Hellwig]

From: Hannes Reinecke <hare@suse.de>

Switch to scsi_execute_req_flags() instead of using the block interface
directly.  This will set REQ_QUIET and REQ_PREEMPT, but this is okay as
we're evaluating the errors anyway and should be able to send the command
even if the device is quiesced.

Signed-off-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 drivers/scsi/device_handler/scsi_dh_hp_sw.c | 222 ++++++++--------------------
 1 file changed, 65 insertions(+), 157 deletions(-)

diff --git a/drivers/scsi/device_handler/scsi_dh_hp_sw.c b/drivers/scsi/device_handler/scsi_dh_hp_sw.c
index 308e871..be43c94 100644
--- a/drivers/scsi/device_handler/scsi_dh_hp_sw.c
+++ b/drivers/scsi/device_handler/scsi_dh_hp_sw.c
@@ -38,13 +38,10 @@
 #define HP_SW_PATH_PASSIVE		1
 
 struct hp_sw_dh_data {
-	unsigned char sense[SCSI_SENSE_BUFFERSIZE];
 	int path_state;
 	int retries;
 	int retry_cnt;
 	struct scsi_device *sdev;
-	activate_complete	callback_fn;
-	void			*callback_data;
 };
 
 static int hp_sw_start_stop(struct hp_sw_dh_data *);
@@ -56,43 +53,34 @@ static int hp_sw_start_stop(struct hp_sw_dh_data *);
  *
  * Returns SCSI_DH_DEV_OFFLINED if the sdev is on the passive path
  */
-static int tur_done(struct scsi_device *sdev, unsigned char *sense)
+static int tur_done(struct scsi_device *sdev, struct hp_sw_dh_data *h,
+		    struct scsi_sense_hdr *sshdr)
 {
-	struct scsi_sense_hdr sshdr;
-	int ret;
+	int ret = SCSI_DH_IO;
 
-	ret = scsi_normalize_sense(sense, SCSI_SENSE_BUFFERSIZE, &sshdr);
-	if (!ret) {
-		sdev_printk(KERN_WARNING, sdev,
-			    "%s: sending tur failed, no sense available\n",
-			    HP_SW_NAME);
-		ret = SCSI_DH_IO;
-		goto done;
-	}
-	switch (sshdr.sense_key) {
+	switch (sshdr->sense_key) {
 	case UNIT_ATTENTION:
 		ret = SCSI_DH_IMM_RETRY;
 		break;
 	case NOT_READY:
-		if ((sshdr.asc == 0x04) && (sshdr.ascq == 2)) {
+		if (sshdr->asc == 0x04 && sshdr->ascq == 2) {
 			/*
 			 * LUN not ready - Initialization command required
 			 *
 			 * This is the passive path
 			 */
-			ret = SCSI_DH_DEV_OFFLINED;
+			h->path_state = HP_SW_PATH_PASSIVE;
+			ret = SCSI_DH_OK;
 			break;
 		}
 		/* Fallthrough */
 	default:
 		sdev_printk(KERN_WARNING, sdev,
 			   "%s: sending tur failed, sense %x/%x/%x\n",
-			   HP_SW_NAME, sshdr.sense_key, sshdr.asc,
-			   sshdr.ascq);
+			   HP_SW_NAME, sshdr->sense_key, sshdr->asc,
+			   sshdr->ascq);
 		break;
 	}
-
-done:
 	return ret;
 }
 
@@ -105,131 +93,36 @@ static int tur_done(struct scsi_device *sdev, unsigned char *sense)
  */
 static int hp_sw_tur(struct scsi_device *sdev, struct hp_sw_dh_data *h)
 {
-	struct request *req;
-	int ret;
+	unsigned char cmd[6] = { TEST_UNIT_READY };
+	struct scsi_sense_hdr sshdr;
+	int ret = SCSI_DH_OK, res;
+	u64 req_flags = REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT |
+		REQ_FAILFAST_DRIVER;
 
 retry:
-	req = blk_get_request(sdev->request_queue, WRITE, GFP_NOIO);
-	if (IS_ERR(req))
-		return SCSI_DH_RES_TEMP_UNAVAIL;
-
-	blk_rq_set_block_pc(req);
-	req->cmd_flags |= REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT |
-			  REQ_FAILFAST_DRIVER;
-	req->cmd_len = COMMAND_SIZE(TEST_UNIT_READY);
-	req->cmd[0] = TEST_UNIT_READY;
-	req->timeout = HP_SW_TIMEOUT;
-	req->sense = h->sense;
-	memset(req->sense, 0, SCSI_SENSE_BUFFERSIZE);
-	req->sense_len = 0;
-
-	ret = blk_execute_rq(req->q, NULL, req, 1);
-	if (ret == -EIO) {
-		if (req->sense_len > 0) {
-			ret = tur_done(sdev, h->sense);
-		} else {
+	res = scsi_execute_req_flags(sdev, cmd, DMA_NONE, NULL, 0, &sshdr,
+				     HP_SW_TIMEOUT, HP_SW_RETRIES,
+				     NULL, req_flags, 0);
+	if (res) {
+		if (scsi_sense_valid(&sshdr))
+			ret = tur_done(sdev, h, &sshdr);
+		else {
 			sdev_printk(KERN_WARNING, sdev,
 				    "%s: sending tur failed with %x\n",
-				    HP_SW_NAME, req->errors);
+				    HP_SW_NAME, res);
 			ret = SCSI_DH_IO;
 		}
 	} else {
 		h->path_state = HP_SW_PATH_ACTIVE;
 		ret = SCSI_DH_OK;
 	}
-	if (ret == SCSI_DH_IMM_RETRY) {
-		blk_put_request(req);
+	if (ret == SCSI_DH_IMM_RETRY)
 		goto retry;
-	}
-	if (ret == SCSI_DH_DEV_OFFLINED) {
-		h->path_state = HP_SW_PATH_PASSIVE;
-		ret = SCSI_DH_OK;
-	}
-
-	blk_put_request(req);
 
 	return ret;
 }
 
 /*
- * start_done - Handle START STOP UNIT return status
- * @sdev: sdev the command has been sent to
- * @errors: blk error code
- */
-static int start_done(struct scsi_device *sdev, unsigned char *sense)
-{
-	struct scsi_sense_hdr sshdr;
-	int rc;
-
-	rc = scsi_normalize_sense(sense, SCSI_SENSE_BUFFERSIZE, &sshdr);
-	if (!rc) {
-		sdev_printk(KERN_WARNING, sdev,
-			    "%s: sending start_stop_unit failed, "
-			    "no sense available\n",
-			    HP_SW_NAME);
-		return SCSI_DH_IO;
-	}
-	switch (sshdr.sense_key) {
-	case NOT_READY:
-		if ((sshdr.asc == 0x04) && (sshdr.ascq == 3)) {
-			/*
-			 * LUN not ready - manual intervention required
-			 *
-			 * Switch-over in progress, retry.
-			 */
-			rc = SCSI_DH_RETRY;
-			break;
-		}
-		/* fall through */
-	default:
-		sdev_printk(KERN_WARNING, sdev,
-			   "%s: sending start_stop_unit failed, sense %x/%x/%x\n",
-			   HP_SW_NAME, sshdr.sense_key, sshdr.asc,
-			   sshdr.ascq);
-		rc = SCSI_DH_IO;
-	}
-
-	return rc;
-}
-
-static void start_stop_endio(struct request *req, int error)
-{
-	struct hp_sw_dh_data *h = req->end_io_data;
-	unsigned err = SCSI_DH_OK;
-
-	if (error || host_byte(req->errors) != DID_OK ||
-			msg_byte(req->errors) != COMMAND_COMPLETE) {
-		sdev_printk(KERN_WARNING, h->sdev,
-			    "%s: sending start_stop_unit failed with %x\n",
-			    HP_SW_NAME, req->errors);
-		err = SCSI_DH_IO;
-		goto done;
-	}
-
-	if (req->sense_len > 0) {
-		err = start_done(h->sdev, h->sense);
-		if (err == SCSI_DH_RETRY) {
-			err = SCSI_DH_IO;
-			if (--h->retry_cnt) {
-				blk_put_request(req);
-				err = hp_sw_start_stop(h);
-				if (err == SCSI_DH_OK)
-					return;
-			}
-		}
-	}
-done:
-	req->end_io_data = NULL;
-	__blk_put_request(req->q, req);
-	if (h->callback_fn) {
-		h->callback_fn(h->callback_data, err);
-		h->callback_fn = h->callback_data = NULL;
-	}
-	return;
-
-}
-
-/*
  * hp_sw_start_stop - Send START STOP UNIT command
  * @sdev: sdev command should be sent to
  *
@@ -237,26 +130,48 @@ static void start_stop_endio(struct request *req, int error)
  */
 static int hp_sw_start_stop(struct hp_sw_dh_data *h)
 {
-	struct request *req;
-
-	req = blk_get_request(h->sdev->request_queue, WRITE, GFP_ATOMIC);
-	if (IS_ERR(req))
-		return SCSI_DH_RES_TEMP_UNAVAIL;
-
-	blk_rq_set_block_pc(req);
-	req->cmd_flags |= REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT |
-			  REQ_FAILFAST_DRIVER;
-	req->cmd_len = COMMAND_SIZE(START_STOP);
-	req->cmd[0] = START_STOP;
-	req->cmd[4] = 1;	/* Start spin cycle */
-	req->timeout = HP_SW_TIMEOUT;
-	req->sense = h->sense;
-	memset(req->sense, 0, SCSI_SENSE_BUFFERSIZE);
-	req->sense_len = 0;
-	req->end_io_data = h;
+	unsigned char cmd[6] = { START_STOP, 0, 0, 0, 1, 0 };
+	struct scsi_sense_hdr sshdr;
+	struct scsi_device *sdev = h->sdev;
+	int res, rc = SCSI_DH_OK;
+	int retry_cnt = HP_SW_RETRIES;
+	u64 req_flags = REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT |
+		REQ_FAILFAST_DRIVER;
 
-	blk_execute_rq_nowait(req->q, NULL, req, 1, start_stop_endio);
-	return SCSI_DH_OK;
+retry:
+	res = scsi_execute_req_flags(sdev, cmd, DMA_NONE, NULL, 0, &sshdr,
+				     HP_SW_TIMEOUT, HP_SW_RETRIES,
+				     NULL, req_flags, 0);
+	if (res) {
+		if (!scsi_sense_valid(&sshdr)) {
+			sdev_printk(KERN_WARNING, sdev,
+				    "%s: sending start_stop_unit failed, "
+				    "no sense available\n", HP_SW_NAME);
+			return SCSI_DH_IO;
+		}
+		switch (sshdr.sense_key) {
+		case NOT_READY:
+			if (sshdr.asc == 0x04 && sshdr.ascq == 3) {
+				/*
+				 * LUN not ready - manual intervention required
+				 *
+				 * Switch-over in progress, retry.
+				 */
+				if (--retry_cnt)
+					goto retry;
+				rc = SCSI_DH_RETRY;
+				break;
+			}
+			/* fall through */
+		default:
+			sdev_printk(KERN_WARNING, sdev,
+				    "%s: sending start_stop_unit failed, "
+				    "sense %x/%x/%x\n", HP_SW_NAME,
+				    sshdr.sense_key, sshdr.asc, sshdr.ascq);
+			rc = SCSI_DH_IO;
+		}
+	}
+	return rc;
 }
 
 static int hp_sw_prep_fn(struct scsi_device *sdev, struct request *req)
@@ -290,15 +205,8 @@ static int hp_sw_activate(struct scsi_device *sdev,
 
 	ret = hp_sw_tur(sdev, h);
 
-	if (ret == SCSI_DH_OK && h->path_state == HP_SW_PATH_PASSIVE) {
-		h->retry_cnt = h->retries;
-		h->callback_fn = fn;
-		h->callback_data = data;
+	if (ret == SCSI_DH_OK && h->path_state == HP_SW_PATH_PASSIVE)
 		ret = hp_sw_start_stop(h);
-		if (ret == SCSI_DH_OK)
-			return 0;
-		h->callback_fn = h->callback_data = NULL;
-	}
 
 	if (fn)
 		fn(data, ret);
-- 
2.1.4

[PATCH 09/15] scsi: remove gfp_flags member in scsi_host_cmd_pool

[Christoph Hellwig]

When using the slab allocator we already decide at cache creation time if
an allocation comes from a GFP_DMA pool using the SLAB_CACHE_DMA flag,
and there is no point passing the kmalloc-family only GFP_DMA flag to
kmem_cache_alloc.  Drop all the infrastructure for doing so.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 drivers/scsi/scsi.c | 14 ++++----------
 1 file changed, 4 insertions(+), 10 deletions(-)

diff --git a/drivers/scsi/scsi.c b/drivers/scsi/scsi.c
index 75455d4..0f93892 100644
--- a/drivers/scsi/scsi.c
+++ b/drivers/scsi/scsi.c
@@ -105,7 +105,6 @@ struct scsi_host_cmd_pool {
 	char			*cmd_name;
 	char			*sense_name;
 	unsigned int		slab_flags;
-	gfp_t			gfp_mask;
 };
 
 static struct scsi_host_cmd_pool scsi_cmd_pool = {
@@ -118,7 +117,6 @@ static struct scsi_host_cmd_pool scsi_cmd_dma_pool = {
 	.cmd_name	= "scsi_cmd_cache(DMA)",
 	.sense_name	= "scsi_sense_cache(DMA)",
 	.slab_flags	= SLAB_HWCACHE_ALIGN|SLAB_CACHE_DMA,
-	.gfp_mask	= __GFP_DMA,
 };
 
 static DEFINE_MUTEX(host_cmd_pool_mutex);
@@ -156,12 +154,11 @@ scsi_host_alloc_command(struct Scsi_Host *shost, gfp_t gfp_mask)
 	struct scsi_host_cmd_pool *pool = shost->cmd_pool;
 	struct scsi_cmnd *cmd;
 
-	cmd = kmem_cache_zalloc(pool->cmd_slab, gfp_mask | pool->gfp_mask);
+	cmd = kmem_cache_zalloc(pool->cmd_slab, gfp_mask);
 	if (!cmd)
 		goto fail;
 
-	cmd->sense_buffer = kmem_cache_alloc(pool->sense_slab,
-					     gfp_mask | pool->gfp_mask);
+	cmd->sense_buffer = kmem_cache_alloc(pool->sense_slab, gfp_mask);
 	if (!cmd->sense_buffer)
 		goto fail_free_cmd;
 
@@ -327,10 +324,8 @@ scsi_alloc_host_cmd_pool(struct Scsi_Host *shost)
 	}
 
 	pool->slab_flags = SLAB_HWCACHE_ALIGN;
-	if (shost->unchecked_isa_dma) {
+	if (shost->unchecked_isa_dma)
 		pool->slab_flags |= SLAB_CACHE_DMA;
-		pool->gfp_mask = __GFP_DMA;
-	}
 
 	if (hostt->cmd_size)
 		hostt->cmd_pool = pool;
@@ -424,7 +419,6 @@ static void scsi_put_host_cmd_pool(struct Scsi_Host *shost)
  */
 int scsi_setup_command_freelist(struct Scsi_Host *shost)
 {
-	const gfp_t gfp_mask = shost->unchecked_isa_dma ? GFP_DMA : GFP_KERNEL;
 	struct scsi_cmnd *cmd;
 
 	spin_lock_init(&shost->free_list_lock);
@@ -437,7 +431,7 @@ int scsi_setup_command_freelist(struct Scsi_Host *shost)
 	/*
 	 * Get one backup command for this host.
 	 */
-	cmd = scsi_host_alloc_command(shost, gfp_mask);
+	cmd = scsi_host_alloc_command(shost, GFP_KERNEL);
 	if (!cmd) {
 		scsi_put_host_cmd_pool(shost);
 		shost->cmd_pool = NULL;
-- 
2.1.4

[PATCH 10/15] scsi: respect unchecked_isa_dma for blk-mq

[Christoph Hellwig]

Currently blk-mq always allocates the sense buffer using normal GFP_KERNEL
allocation.  Refactor the cmd pool code to split the cmd and sense allocation
and share the code to allocate the sense buffers as well as the sense buffer
slab caches between the legacy and blk-mq path.

Note that this switches to lazy allocation of the sense slab caches - the
slab caches (not the actual allocations) won't be destroy until the scsi
module is unloaded instead of keeping track of hosts using them.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 drivers/scsi/hosts.c     |  4 ++++
 drivers/scsi/scsi.c      | 24 ++++---------------
 drivers/scsi/scsi_lib.c  | 62 +++++++++++++++++++++++++++++++++++++++++++++---
 drivers/scsi/scsi_priv.h |  5 ++++
 4 files changed, 73 insertions(+), 22 deletions(-)

diff --git a/drivers/scsi/hosts.c b/drivers/scsi/hosts.c
index 258a3f9..6d29c4a 100644
--- a/drivers/scsi/hosts.c
+++ b/drivers/scsi/hosts.c
@@ -213,6 +213,10 @@ int scsi_add_host_with_dma(struct Scsi_Host *shost, struct device *dev,
 		goto fail;
 	}
 
+	error = scsi_init_sense_cache(shost);
+	if (error)
+		goto fail;
+
 	if (shost_use_blk_mq(shost)) {
 		error = scsi_mq_setup_tags(shost);
 		if (error)
diff --git a/drivers/scsi/scsi.c b/drivers/scsi/scsi.c
index 0f93892..469aa0f 100644
--- a/drivers/scsi/scsi.c
+++ b/drivers/scsi/scsi.c
@@ -100,22 +100,18 @@ EXPORT_SYMBOL(scsi_sd_pm_domain);
 
 struct scsi_host_cmd_pool {
 	struct kmem_cache	*cmd_slab;
-	struct kmem_cache	*sense_slab;
 	unsigned int		users;
 	char			*cmd_name;
-	char			*sense_name;
 	unsigned int		slab_flags;
 };
 
 static struct scsi_host_cmd_pool scsi_cmd_pool = {
 	.cmd_name	= "scsi_cmd_cache",
-	.sense_name	= "scsi_sense_cache",
 	.slab_flags	= SLAB_HWCACHE_ALIGN,
 };
 
 static struct scsi_host_cmd_pool scsi_cmd_dma_pool = {
 	.cmd_name	= "scsi_cmd_cache(DMA)",
-	.sense_name	= "scsi_sense_cache(DMA)",
 	.slab_flags	= SLAB_HWCACHE_ALIGN|SLAB_CACHE_DMA,
 };
 
@@ -136,7 +132,7 @@ scsi_host_free_command(struct Scsi_Host *shost, struct scsi_cmnd *cmd)
 
 	if (cmd->prot_sdb)
 		kmem_cache_free(scsi_sdb_cache, cmd->prot_sdb);
-	kmem_cache_free(pool->sense_slab, cmd->sense_buffer);
+	scsi_free_sense_buffer(shost, cmd->sense_buffer);
 	kmem_cache_free(pool->cmd_slab, cmd);
 }
 
@@ -158,7 +154,8 @@ scsi_host_alloc_command(struct Scsi_Host *shost, gfp_t gfp_mask)
 	if (!cmd)
 		goto fail;
 
-	cmd->sense_buffer = kmem_cache_alloc(pool->sense_slab, gfp_mask);
+	cmd->sense_buffer = scsi_alloc_sense_buffer(shost, gfp_mask,
+			NUMA_NO_NODE);
 	if (!cmd->sense_buffer)
 		goto fail_free_cmd;
 
@@ -171,7 +168,7 @@ scsi_host_alloc_command(struct Scsi_Host *shost, gfp_t gfp_mask)
 	return cmd;
 
 fail_free_sense:
-	kmem_cache_free(pool->sense_slab, cmd->sense_buffer);
+	scsi_free_sense_buffer(shost, cmd->sense_buffer);
 fail_free_cmd:
 	kmem_cache_free(pool->cmd_slab, cmd);
 fail:
@@ -301,7 +298,6 @@ scsi_find_host_cmd_pool(struct Scsi_Host *shost)
 static void
 scsi_free_host_cmd_pool(struct scsi_host_cmd_pool *pool)
 {
-	kfree(pool->sense_name);
 	kfree(pool->cmd_name);
 	kfree(pool);
 }
@@ -317,8 +313,7 @@ scsi_alloc_host_cmd_pool(struct Scsi_Host *shost)
 		return NULL;
 
 	pool->cmd_name = kasprintf(GFP_KERNEL, "%s_cmd", hostt->proc_name);
-	pool->sense_name = kasprintf(GFP_KERNEL, "%s_sense", hostt->proc_name);
-	if (!pool->cmd_name || !pool->sense_name) {
+	if (!pool->cmd_name) {
 		scsi_free_host_cmd_pool(pool);
 		return NULL;
 	}
@@ -357,12 +352,6 @@ scsi_get_host_cmd_pool(struct Scsi_Host *shost)
 						   pool->slab_flags, NULL);
 		if (!pool->cmd_slab)
 			goto out_free_pool;
-
-		pool->sense_slab = kmem_cache_create(pool->sense_name,
-						     SCSI_SENSE_BUFFERSIZE, 0,
-						     pool->slab_flags, NULL);
-		if (!pool->sense_slab)
-			goto out_free_slab;
 	}
 
 	pool->users++;
@@ -371,8 +360,6 @@ scsi_get_host_cmd_pool(struct Scsi_Host *shost)
 	mutex_unlock(&host_cmd_pool_mutex);
 	return retval;
 
-out_free_slab:
-	kmem_cache_destroy(pool->cmd_slab);
 out_free_pool:
 	if (hostt->cmd_size) {
 		scsi_free_host_cmd_pool(pool);
@@ -398,7 +385,6 @@ static void scsi_put_host_cmd_pool(struct Scsi_Host *shost)
 
 	if (!--pool->users) {
 		kmem_cache_destroy(pool->cmd_slab);
-		kmem_cache_destroy(pool->sense_slab);
 		if (hostt->cmd_size) {
 			scsi_free_host_cmd_pool(pool);
 			hostt->cmd_pool = NULL;
diff --git a/drivers/scsi/scsi_lib.c b/drivers/scsi/scsi_lib.c
index c35b6de..e5ebb3a 100644
--- a/drivers/scsi/scsi_lib.c
+++ b/drivers/scsi/scsi_lib.c
@@ -39,6 +39,58 @@
 
 
 struct kmem_cache *scsi_sdb_cache;
+static struct kmem_cache *scsi_sense_cache;
+static struct kmem_cache *scsi_sense_isadma_cache;
+static DEFINE_MUTEX(scsi_sense_cache_mutex);
+
+static inline struct kmem_cache *
+scsi_select_sense_cache(struct Scsi_Host *shost)
+{
+	return shost->unchecked_isa_dma ?
+		scsi_sense_isadma_cache : scsi_sense_cache;
+}
+
+void scsi_free_sense_buffer(struct Scsi_Host *shost,
+		unsigned char *sense_buffer)
+{
+	kmem_cache_free(scsi_select_sense_cache(shost), sense_buffer);
+}
+
+unsigned char *scsi_alloc_sense_buffer(struct Scsi_Host *shost, gfp_t gfp_mask,
+		int numa_node)
+{
+	return kmem_cache_alloc_node(scsi_select_sense_cache(shost), gfp_mask,
+			numa_node);
+}
+
+int scsi_init_sense_cache(struct Scsi_Host *shost)
+{
+	struct kmem_cache *cache;
+	int ret = 0;
+
+	cache = scsi_select_sense_cache(shost);
+	if (cache)
+		return 0;
+
+	mutex_lock(&scsi_sense_cache_mutex);
+	if (shost->unchecked_isa_dma) {
+		scsi_sense_isadma_cache =
+			kmem_cache_create("scsi_sense_cache(DMA)",
+			SCSI_SENSE_BUFFERSIZE, 0,
+			SLAB_HWCACHE_ALIGN | SLAB_CACHE_DMA, NULL);
+		if (!scsi_sense_isadma_cache)
+			ret = -ENOMEM;
+	} else {
+		scsi_sense_cache =
+			kmem_cache_create("scsi_sense_cache",
+			SCSI_SENSE_BUFFERSIZE, 0, SLAB_HWCACHE_ALIGN, NULL);
+		if (!scsi_sense_cache)
+			ret = -ENOMEM;
+	}
+
+	mutex_unlock(&scsi_sense_cache_mutex);
+	return ret;
+}
 
 /*
  * When to reinvoke queueing after a resource shortage. It's 3 msecs to
@@ -1981,10 +2033,11 @@ static int scsi_init_request(void *data, struct request *rq,
 		unsigned int hctx_idx, unsigned int request_idx,
 		unsigned int numa_node)
 {
+	struct Scsi_Host *shost = data;
 	struct scsi_cmnd *cmd = blk_mq_rq_to_pdu(rq);
 
-	cmd->sense_buffer = kzalloc_node(SCSI_SENSE_BUFFERSIZE, GFP_KERNEL,
-			numa_node);
+	cmd->sense_buffer =
+		scsi_alloc_sense_buffer(shost, GFP_KERNEL, numa_node);
 	if (!cmd->sense_buffer)
 		return -ENOMEM;
 	return 0;
@@ -1993,9 +2046,10 @@ static int scsi_init_request(void *data, struct request *rq,
 static void scsi_exit_request(void *data, struct request *rq,
 		unsigned int hctx_idx, unsigned int request_idx)
 {
+	struct Scsi_Host *shost = data;
 	struct scsi_cmnd *cmd = blk_mq_rq_to_pdu(rq);
 
-	kfree(cmd->sense_buffer);
+	scsi_free_sense_buffer(shost, cmd->sense_buffer);
 }
 
 static int scsi_map_queues(struct blk_mq_tag_set *set)
@@ -2208,6 +2262,8 @@ int __init scsi_init_queue(void)
 
 void scsi_exit_queue(void)
 {
+	kmem_cache_destroy(scsi_sense_cache);
+	kmem_cache_destroy(scsi_sense_isadma_cache);
 	kmem_cache_destroy(scsi_sdb_cache);
 }
 
diff --git a/drivers/scsi/scsi_priv.h b/drivers/scsi/scsi_priv.h
index 193636a..1a712c6 100644
--- a/drivers/scsi/scsi_priv.h
+++ b/drivers/scsi/scsi_priv.h
@@ -30,6 +30,11 @@ extern void scsi_exit_hosts(void);
 
 /* scsi.c */
 extern bool scsi_use_blk_mq;
+void scsi_free_sense_buffer(struct Scsi_Host *shost,
+		unsigned char *sense_buffer);
+unsigned char *scsi_alloc_sense_buffer(struct Scsi_Host *shost, gfp_t gfp_mask,
+		int numa_node);
+int scsi_init_sense_cache(struct Scsi_Host *shost);
 extern int scsi_setup_command_freelist(struct Scsi_Host *shost);
 extern void scsi_destroy_command_freelist(struct Scsi_Host *shost);
 #ifdef CONFIG_SCSI_LOGGING
-- 
2.1.4

[PATCH 11/15] scsi: remove scsi_cmd_dma_pool

[Christoph Hellwig]

There is no need for GFP_DMA allocations of the scsi_cmnd structures
themselves, all that might be DMAed to or from is the actual payload,
or the sense buffers.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 drivers/scsi/scsi.c | 15 +--------------
 1 file changed, 1 insertion(+), 14 deletions(-)

diff --git a/drivers/scsi/scsi.c b/drivers/scsi/scsi.c
index 469aa0f..2e24f31 100644
--- a/drivers/scsi/scsi.c
+++ b/drivers/scsi/scsi.c
@@ -102,17 +102,10 @@ struct scsi_host_cmd_pool {
 	struct kmem_cache	*cmd_slab;
 	unsigned int		users;
 	char			*cmd_name;
-	unsigned int		slab_flags;
 };
 
 static struct scsi_host_cmd_pool scsi_cmd_pool = {
 	.cmd_name	= "scsi_cmd_cache",
-	.slab_flags	= SLAB_HWCACHE_ALIGN,
-};
-
-static struct scsi_host_cmd_pool scsi_cmd_dma_pool = {
-	.cmd_name	= "scsi_cmd_cache(DMA)",
-	.slab_flags	= SLAB_HWCACHE_ALIGN|SLAB_CACHE_DMA,
 };
 
 static DEFINE_MUTEX(host_cmd_pool_mutex);
@@ -290,8 +283,6 @@ scsi_find_host_cmd_pool(struct Scsi_Host *shost)
 {
 	if (shost->hostt->cmd_size)
 		return shost->hostt->cmd_pool;
-	if (shost->unchecked_isa_dma)
-		return &scsi_cmd_dma_pool;
 	return &scsi_cmd_pool;
 }
 
@@ -318,10 +309,6 @@ scsi_alloc_host_cmd_pool(struct Scsi_Host *shost)
 		return NULL;
 	}
 
-	pool->slab_flags = SLAB_HWCACHE_ALIGN;
-	if (shost->unchecked_isa_dma)
-		pool->slab_flags |= SLAB_CACHE_DMA;
-
 	if (hostt->cmd_size)
 		hostt->cmd_pool = pool;
 
@@ -349,7 +336,7 @@ scsi_get_host_cmd_pool(struct Scsi_Host *shost)
 
 	if (!pool->users) {
 		pool->cmd_slab = kmem_cache_create(pool->cmd_name, cmd_size, 0,
-						   pool->slab_flags, NULL);
+						   SLAB_HWCACHE_ALIGN, NULL);
 		if (!pool->cmd_slab)
 			goto out_free_pool;
 	}
-- 
2.1.4

[PATCH 12/15] scsi: remove __scsi_alloc_queue

[Christoph Hellwig]

Instead do an internal export of __scsi_init_queue for the transport
classes that export BSG nodes.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 drivers/scsi/scsi_lib.c             | 19 ++++---------------
 drivers/scsi/scsi_transport_fc.c    |  6 ++++--
 drivers/scsi/scsi_transport_iscsi.c |  3 ++-
 include/scsi/scsi_host.h            |  2 --
 include/scsi/scsi_transport.h       |  2 ++
 5 files changed, 12 insertions(+), 20 deletions(-)

diff --git a/drivers/scsi/scsi_lib.c b/drivers/scsi/scsi_lib.c
index e5ebb3a..898af1a 100644
--- a/drivers/scsi/scsi_lib.c
+++ b/drivers/scsi/scsi_lib.c
@@ -2082,7 +2082,7 @@ static u64 scsi_calculate_bounce_limit(struct Scsi_Host *shost)
 	return bounce_limit;
 }
 
-static void __scsi_init_queue(struct Scsi_Host *shost, struct request_queue *q)
+void __scsi_init_queue(struct Scsi_Host *shost, struct request_queue *q)
 {
 	struct device *dev = shost->dma_dev;
 
@@ -2117,28 +2117,17 @@ static void __scsi_init_queue(struct Scsi_Host *shost, struct request_queue *q)
 	 */
 	blk_queue_dma_alignment(q, 0x03);
 }
-
-struct request_queue *__scsi_alloc_queue(struct Scsi_Host *shost,
-					 request_fn_proc *request_fn)
-{
-	struct request_queue *q;
-
-	q = blk_init_queue(request_fn, NULL);
-	if (!q)
-		return NULL;
-	__scsi_init_queue(shost, q);
-	return q;
-}
-EXPORT_SYMBOL(__scsi_alloc_queue);
+EXPORT_SYMBOL_GPL(__scsi_init_queue);
 
 struct request_queue *scsi_alloc_queue(struct scsi_device *sdev)
 {
 	struct request_queue *q;
 
-	q = __scsi_alloc_queue(sdev->host, scsi_request_fn);
+	q = blk_init_queue(scsi_request_fn, NULL);
 	if (!q)
 		return NULL;
 
+	__scsi_init_queue(sdev->host, q);
 	blk_queue_prep_rq(q, scsi_prep_fn);
 	blk_queue_unprep_rq(q, scsi_unprep_fn);
 	blk_queue_softirq_done(q, scsi_softirq_done);
diff --git a/drivers/scsi/scsi_transport_fc.c b/drivers/scsi/scsi_transport_fc.c
index 03577bd..afcedec 100644
--- a/drivers/scsi/scsi_transport_fc.c
+++ b/drivers/scsi/scsi_transport_fc.c
@@ -3776,7 +3776,7 @@ fc_bsg_hostadd(struct Scsi_Host *shost, struct fc_host_attrs *fc_host)
 	snprintf(bsg_name, sizeof(bsg_name),
 		 "fc_host%d", shost->host_no);
 
-	q = __scsi_alloc_queue(shost, bsg_request_fn);
+	q = blk_init_queue(bsg_request_fn, NULL);
 	if (!q) {
 		dev_err(dev,
 			"fc_host%d: bsg interface failed to initialize - no request queue\n",
@@ -3784,6 +3784,7 @@ fc_bsg_hostadd(struct Scsi_Host *shost, struct fc_host_attrs *fc_host)
 		return -ENOMEM;
 	}
 
+	__scsi_init_queue(shost, q);
 	err = bsg_setup_queue(dev, q, bsg_name, fc_bsg_dispatch,
 				 i->f->dd_bsg_size);
 	if (err) {
@@ -3831,12 +3832,13 @@ fc_bsg_rportadd(struct Scsi_Host *shost, struct fc_rport *rport)
 	if (!i->f->bsg_request)
 		return -ENOTSUPP;
 
-	q = __scsi_alloc_queue(shost, bsg_request_fn);
+	q = blk_init_queue(bsg_request_fn, NULL);
 	if (!q) {
 		dev_err(dev, "bsg interface failed to initialize - no request queue\n");
 		return -ENOMEM;
 	}
 
+	__scsi_init_queue(shost, q);
 	err = bsg_setup_queue(dev, q, NULL, fc_bsg_dispatch, i->f->dd_bsg_size);
 	if (err) {
 		dev_err(dev, "failed to setup bsg queue\n");
diff --git a/drivers/scsi/scsi_transport_iscsi.c b/drivers/scsi/scsi_transport_iscsi.c
index 42bca61..04ebe6e 100644
--- a/drivers/scsi/scsi_transport_iscsi.c
+++ b/drivers/scsi/scsi_transport_iscsi.c
@@ -1544,10 +1544,11 @@ iscsi_bsg_host_add(struct Scsi_Host *shost, struct iscsi_cls_host *ihost)
 
 	snprintf(bsg_name, sizeof(bsg_name), "iscsi_host%d", shost->host_no);
 
-	q = __scsi_alloc_queue(shost, bsg_request_fn);
+	q = blk_init_queue(bsg_request_fn, NULL);
 	if (!q)
 		return -ENOMEM;
 
+	__scsi_init_queue(shost, q);
 	ret = bsg_setup_queue(dev, q, bsg_name, iscsi_bsg_host_dispatch, 0);
 	if (ret) {
 		shost_printk(KERN_ERR, shost, "bsg interface failed to "
diff --git a/include/scsi/scsi_host.h b/include/scsi/scsi_host.h
index 36680f1..f4964d7 100644
--- a/include/scsi/scsi_host.h
+++ b/include/scsi/scsi_host.h
@@ -826,8 +826,6 @@ extern void scsi_block_requests(struct Scsi_Host *);
 
 struct class_container;
 
-extern struct request_queue *__scsi_alloc_queue(struct Scsi_Host *shost,
-						void (*) (struct request_queue *));
 /*
  * These two functions are used to allocate and free a pseudo device
  * which will connect to the host adapter itself rather than any
diff --git a/include/scsi/scsi_transport.h b/include/scsi/scsi_transport.h
index 8129239..b6e07b5 100644
--- a/include/scsi/scsi_transport.h
+++ b/include/scsi/scsi_transport.h
@@ -119,4 +119,6 @@ scsi_transport_device_data(struct scsi_device *sdev)
 		+ shost->transportt->device_private_offset;
 }
 
+void __scsi_init_queue(struct Scsi_Host *shost, struct request_queue *q);
+
 #endif /* SCSI_TRANSPORT_H */
-- 
2.1.4

[PATCH 13/15] scsi: allocate scsi_cmnd structures as part of struct request

[Christoph Hellwig]

Rely on the new block layer functionality to allocate additional driver
specific data behind struct request instead of implementing it in SCSI
itѕelf.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 drivers/scsi/hosts.c      |  20 +--
 drivers/scsi/scsi.c       | 319 ----------------------------------------------
 drivers/scsi/scsi_error.c |  17 ++-
 drivers/scsi/scsi_lib.c   | 122 ++++++++++++------
 drivers/scsi/scsi_priv.h  |   8 +-
 include/scsi/scsi_host.h  |   3 -
 6 files changed, 95 insertions(+), 394 deletions(-)

diff --git a/drivers/scsi/hosts.c b/drivers/scsi/hosts.c
index 6d29c4a..831a1c8 100644
--- a/drivers/scsi/hosts.c
+++ b/drivers/scsi/hosts.c
@@ -230,19 +230,6 @@ int scsi_add_host_with_dma(struct Scsi_Host *shost, struct device *dev,
 		}
 	}
 
-	/*
-	 * Note that we allocate the freelist even for the MQ case for now,
-	 * as we need a command set aside for scsi_reset_provider.  Having
-	 * the full host freelist and one command available for that is a
-	 * little heavy-handed, but avoids introducing a special allocator
-	 * just for this.  Eventually the structure of scsi_reset_provider
-	 * will need a major overhaul.
-	 */
-	error = scsi_setup_command_freelist(shost);
-	if (error)
-		goto out_destroy_tags;
-
-
 	if (!shost->shost_gendev.parent)
 		shost->shost_gendev.parent = dev ? dev : &platform_bus;
 	if (!dma_dev)
@@ -262,7 +249,7 @@ int scsi_add_host_with_dma(struct Scsi_Host *shost, struct device *dev,
 
 	error = device_add(&shost->shost_gendev);
 	if (error)
-		goto out_destroy_freelist;
+		goto out_disable_runtime_pm;
 
 	scsi_host_set_state(shost, SHOST_RUNNING);
 	get_device(shost->shost_gendev.parent);
@@ -312,13 +299,11 @@ int scsi_add_host_with_dma(struct Scsi_Host *shost, struct device *dev,
 	device_del(&shost->shost_dev);
  out_del_gendev:
 	device_del(&shost->shost_gendev);
- out_destroy_freelist:
+ out_disable_runtime_pm:
 	device_disable_async_suspend(&shost->shost_gendev);
 	pm_runtime_disable(&shost->shost_gendev);
 	pm_runtime_set_suspended(&shost->shost_gendev);
 	pm_runtime_put_noidle(&shost->shost_gendev);
-	scsi_destroy_command_freelist(shost);
- out_destroy_tags:
 	if (shost_use_blk_mq(shost))
 		scsi_mq_destroy_tags(shost);
  fail:
@@ -359,7 +344,6 @@ static void scsi_host_dev_release(struct device *dev)
 		kfree(dev_name(&shost->shost_dev));
 	}
 
-	scsi_destroy_command_freelist(shost);
 	if (shost_use_blk_mq(shost)) {
 		if (shost->tag_set.tags)
 			scsi_mq_destroy_tags(shost);
diff --git a/drivers/scsi/scsi.c b/drivers/scsi/scsi.c
index 2e24f31..3d8d215 100644
--- a/drivers/scsi/scsi.c
+++ b/drivers/scsi/scsi.c
@@ -98,163 +98,6 @@ EXPORT_SYMBOL(scsi_sd_probe_domain);
 ASYNC_DOMAIN_EXCLUSIVE(scsi_sd_pm_domain);
 EXPORT_SYMBOL(scsi_sd_pm_domain);
 
-struct scsi_host_cmd_pool {
-	struct kmem_cache	*cmd_slab;
-	unsigned int		users;
-	char			*cmd_name;
-};
-
-static struct scsi_host_cmd_pool scsi_cmd_pool = {
-	.cmd_name	= "scsi_cmd_cache",
-};
-
-static DEFINE_MUTEX(host_cmd_pool_mutex);
-
-/**
- * scsi_host_free_command - internal function to release a command
- * @shost:	host to free the command for
- * @cmd:	command to release
- *
- * the command must previously have been allocated by
- * scsi_host_alloc_command.
- */
-static void
-scsi_host_free_command(struct Scsi_Host *shost, struct scsi_cmnd *cmd)
-{
-	struct scsi_host_cmd_pool *pool = shost->cmd_pool;
-
-	if (cmd->prot_sdb)
-		kmem_cache_free(scsi_sdb_cache, cmd->prot_sdb);
-	scsi_free_sense_buffer(shost, cmd->sense_buffer);
-	kmem_cache_free(pool->cmd_slab, cmd);
-}
-
-/**
- * scsi_host_alloc_command - internal function to allocate command
- * @shost:	SCSI host whose pool to allocate from
- * @gfp_mask:	mask for the allocation
- *
- * Returns a fully allocated command with sense buffer and protection
- * data buffer (where applicable) or NULL on failure
- */
-static struct scsi_cmnd *
-scsi_host_alloc_command(struct Scsi_Host *shost, gfp_t gfp_mask)
-{
-	struct scsi_host_cmd_pool *pool = shost->cmd_pool;
-	struct scsi_cmnd *cmd;
-
-	cmd = kmem_cache_zalloc(pool->cmd_slab, gfp_mask);
-	if (!cmd)
-		goto fail;
-
-	cmd->sense_buffer = scsi_alloc_sense_buffer(shost, gfp_mask,
-			NUMA_NO_NODE);
-	if (!cmd->sense_buffer)
-		goto fail_free_cmd;
-
-	if (scsi_host_get_prot(shost) >= SHOST_DIX_TYPE0_PROTECTION) {
-		cmd->prot_sdb = kmem_cache_zalloc(scsi_sdb_cache, gfp_mask);
-		if (!cmd->prot_sdb)
-			goto fail_free_sense;
-	}
-
-	return cmd;
-
-fail_free_sense:
-	scsi_free_sense_buffer(shost, cmd->sense_buffer);
-fail_free_cmd:
-	kmem_cache_free(pool->cmd_slab, cmd);
-fail:
-	return NULL;
-}
-
-/**
- * __scsi_get_command - Allocate a struct scsi_cmnd
- * @shost: host to transmit command
- * @gfp_mask: allocation mask
- *
- * Description: allocate a struct scsi_cmd from host's slab, recycling from the
- *              host's free_list if necessary.
- */
-static struct scsi_cmnd *
-__scsi_get_command(struct Scsi_Host *shost, gfp_t gfp_mask)
-{
-	struct scsi_cmnd *cmd = scsi_host_alloc_command(shost, gfp_mask);
-
-	if (unlikely(!cmd)) {
-		unsigned long flags;
-
-		spin_lock_irqsave(&shost->free_list_lock, flags);
-		if (likely(!list_empty(&shost->free_list))) {
-			cmd = list_entry(shost->free_list.next,
-					 struct scsi_cmnd, list);
-			list_del_init(&cmd->list);
-		}
-		spin_unlock_irqrestore(&shost->free_list_lock, flags);
-
-		if (cmd) {
-			void *buf, *prot;
-
-			buf = cmd->sense_buffer;
-			prot = cmd->prot_sdb;
-
-			memset(cmd, 0, sizeof(*cmd));
-
-			cmd->sense_buffer = buf;
-			cmd->prot_sdb = prot;
-		}
-	}
-
-	return cmd;
-}
-
-/**
- * scsi_get_command - Allocate and setup a scsi command block
- * @dev: parent scsi device
- * @gfp_mask: allocator flags
- *
- * Returns:	The allocated scsi command structure.
- */
-struct scsi_cmnd *scsi_get_command(struct scsi_device *dev, gfp_t gfp_mask)
-{
-	struct scsi_cmnd *cmd = __scsi_get_command(dev->host, gfp_mask);
-	unsigned long flags;
-
-	if (unlikely(cmd == NULL))
-		return NULL;
-
-	cmd->device = dev;
-	INIT_LIST_HEAD(&cmd->list);
-	INIT_DELAYED_WORK(&cmd->abort_work, scmd_eh_abort_handler);
-	spin_lock_irqsave(&dev->list_lock, flags);
-	list_add_tail(&cmd->list, &dev->cmd_list);
-	spin_unlock_irqrestore(&dev->list_lock, flags);
-	cmd->jiffies_at_alloc = jiffies;
-	return cmd;
-}
-
-/**
- * __scsi_put_command - Free a struct scsi_cmnd
- * @shost: dev->host
- * @cmd: Command to free
- */
-static void __scsi_put_command(struct Scsi_Host *shost, struct scsi_cmnd *cmd)
-{
-	unsigned long flags;
-
-	if (unlikely(list_empty(&shost->free_list))) {
-		spin_lock_irqsave(&shost->free_list_lock, flags);
-		if (list_empty(&shost->free_list)) {
-			list_add(&cmd->list, &shost->free_list);
-			cmd = NULL;
-		}
-		spin_unlock_irqrestore(&shost->free_list_lock, flags);
-	}
-
-	if (likely(cmd != NULL))
-		scsi_host_free_command(shost, cmd);
-}
-
 /**
  * scsi_put_command - Free a scsi command block
  * @cmd: command block to free
@@ -274,168 +117,6 @@ void scsi_put_command(struct scsi_cmnd *cmd)
 	spin_unlock_irqrestore(&cmd->device->list_lock, flags);
 
 	BUG_ON(delayed_work_pending(&cmd->abort_work));
-
-	__scsi_put_command(cmd->device->host, cmd);
-}
-
-static struct scsi_host_cmd_pool *
-scsi_find_host_cmd_pool(struct Scsi_Host *shost)
-{
-	if (shost->hostt->cmd_size)
-		return shost->hostt->cmd_pool;
-	return &scsi_cmd_pool;
-}
-
-static void
-scsi_free_host_cmd_pool(struct scsi_host_cmd_pool *pool)
-{
-	kfree(pool->cmd_name);
-	kfree(pool);
-}
-
-static struct scsi_host_cmd_pool *
-scsi_alloc_host_cmd_pool(struct Scsi_Host *shost)
-{
-	struct scsi_host_template *hostt = shost->hostt;
-	struct scsi_host_cmd_pool *pool;
-
-	pool = kzalloc(sizeof(*pool), GFP_KERNEL);
-	if (!pool)
-		return NULL;
-
-	pool->cmd_name = kasprintf(GFP_KERNEL, "%s_cmd", hostt->proc_name);
-	if (!pool->cmd_name) {
-		scsi_free_host_cmd_pool(pool);
-		return NULL;
-	}
-
-	if (hostt->cmd_size)
-		hostt->cmd_pool = pool;
-
-	return pool;
-}
-
-static struct scsi_host_cmd_pool *
-scsi_get_host_cmd_pool(struct Scsi_Host *shost)
-{
-	struct scsi_host_template *hostt = shost->hostt;
-	struct scsi_host_cmd_pool *retval = NULL, *pool;
-	size_t cmd_size = sizeof(struct scsi_cmnd) + hostt->cmd_size;
-
-	/*
-	 * Select a command slab for this host and create it if not
-	 * yet existent.
-	 */
-	mutex_lock(&host_cmd_pool_mutex);
-	pool = scsi_find_host_cmd_pool(shost);
-	if (!pool) {
-		pool = scsi_alloc_host_cmd_pool(shost);
-		if (!pool)
-			goto out;
-	}
-
-	if (!pool->users) {
-		pool->cmd_slab = kmem_cache_create(pool->cmd_name, cmd_size, 0,
-						   SLAB_HWCACHE_ALIGN, NULL);
-		if (!pool->cmd_slab)
-			goto out_free_pool;
-	}
-
-	pool->users++;
-	retval = pool;
-out:
-	mutex_unlock(&host_cmd_pool_mutex);
-	return retval;
-
-out_free_pool:
-	if (hostt->cmd_size) {
-		scsi_free_host_cmd_pool(pool);
-		hostt->cmd_pool = NULL;
-	}
-	goto out;
-}
-
-static void scsi_put_host_cmd_pool(struct Scsi_Host *shost)
-{
-	struct scsi_host_template *hostt = shost->hostt;
-	struct scsi_host_cmd_pool *pool;
-
-	mutex_lock(&host_cmd_pool_mutex);
-	pool = scsi_find_host_cmd_pool(shost);
-
-	/*
-	 * This may happen if a driver has a mismatched get and put
-	 * of the command pool; the driver should be implicated in
-	 * the stack trace
-	 */
-	BUG_ON(pool->users == 0);
-
-	if (!--pool->users) {
-		kmem_cache_destroy(pool->cmd_slab);
-		if (hostt->cmd_size) {
-			scsi_free_host_cmd_pool(pool);
-			hostt->cmd_pool = NULL;
-		}
-	}
-	mutex_unlock(&host_cmd_pool_mutex);
-}
-
-/**
- * scsi_setup_command_freelist - Setup the command freelist for a scsi host.
- * @shost: host to allocate the freelist for.
- *
- * Description: The command freelist protects against system-wide out of memory
- * deadlock by preallocating one SCSI command structure for each host, so the
- * system can always write to a swap file on a device associated with that host.
- *
- * Returns:	Nothing.
- */
-int scsi_setup_command_freelist(struct Scsi_Host *shost)
-{
-	struct scsi_cmnd *cmd;
-
-	spin_lock_init(&shost->free_list_lock);
-	INIT_LIST_HEAD(&shost->free_list);
-
-	shost->cmd_pool = scsi_get_host_cmd_pool(shost);
-	if (!shost->cmd_pool)
-		return -ENOMEM;
-
-	/*
-	 * Get one backup command for this host.
-	 */
-	cmd = scsi_host_alloc_command(shost, GFP_KERNEL);
-	if (!cmd) {
-		scsi_put_host_cmd_pool(shost);
-		shost->cmd_pool = NULL;
-		return -ENOMEM;
-	}
-	list_add(&cmd->list, &shost->free_list);
-	return 0;
-}
-
-/**
- * scsi_destroy_command_freelist - Release the command freelist for a scsi host.
- * @shost: host whose freelist is going to be destroyed
- */
-void scsi_destroy_command_freelist(struct Scsi_Host *shost)
-{
-	/*
-	 * If cmd_pool is NULL the free list was not initialized, so
-	 * do not attempt to release resources.
-	 */
-	if (!shost->cmd_pool)
-		return;
-
-	while (!list_empty(&shost->free_list)) {
-		struct scsi_cmnd *cmd;
-
-		cmd = list_entry(shost->free_list.next, struct scsi_cmnd, list);
-		list_del_init(&cmd->list);
-		scsi_host_free_command(shost, cmd);
-	}
-	shost->cmd_pool = NULL;
-	scsi_put_host_cmd_pool(shost);
 }
 
 #ifdef CONFIG_SCSI_LOGGING
diff --git a/drivers/scsi/scsi_error.c b/drivers/scsi/scsi_error.c
index 996e134..7c08460 100644
--- a/drivers/scsi/scsi_error.c
+++ b/drivers/scsi/scsi_error.c
@@ -2331,7 +2331,7 @@ scsi_ioctl_reset(struct scsi_device *dev, int __user *arg)
 {
 	struct scsi_cmnd *scmd;
 	struct Scsi_Host *shost = dev->host;
-	struct request req;
+	struct request *rq;
 	unsigned long flags;
 	int error = 0, rtn, val;
 
@@ -2346,14 +2346,16 @@ scsi_ioctl_reset(struct scsi_device *dev, int __user *arg)
 		return -EIO;
 
 	error = -EIO;
-	scmd = scsi_get_command(dev, GFP_KERNEL);
-	if (!scmd)
+	rq = kzalloc(sizeof(struct request) + sizeof(struct scsi_cmnd) +
+			shost->hostt->cmd_size, GFP_KERNEL);
+	if (!rq)
 		goto out_put_autopm_host;
+	blk_rq_init(NULL, rq);
 
-	blk_rq_init(NULL, &req);
-	scmd->request = &req;
-
-	scmd->cmnd = req.cmd;
+	scmd = (struct scsi_cmnd *)(rq + 1);
+	scsi_init_command(dev, scmd);
+	scmd->request = rq;
+	scmd->cmnd = rq->cmd;
 
 	scmd->scsi_done		= scsi_reset_provider_done_command;
 	memset(&scmd->sdb, 0, sizeof(scmd->sdb));
@@ -2413,6 +2415,7 @@ scsi_ioctl_reset(struct scsi_device *dev, int __user *arg)
 	scsi_run_host_queues(shost);
 
 	scsi_put_command(scmd);
+	kfree(rq);
 
 out_put_autopm_host:
 	scsi_autopm_put_host(shost);
diff --git a/drivers/scsi/scsi_lib.c b/drivers/scsi/scsi_lib.c
index 898af1a..3c11892 100644
--- a/drivers/scsi/scsi_lib.c
+++ b/drivers/scsi/scsi_lib.c
@@ -37,8 +37,7 @@
 #include "scsi_priv.h"
 #include "scsi_logging.h"
 
-
-struct kmem_cache *scsi_sdb_cache;
+static struct kmem_cache *scsi_sdb_cache;
 static struct kmem_cache *scsi_sense_cache;
 static struct kmem_cache *scsi_sense_isadma_cache;
 static DEFINE_MUTEX(scsi_sense_cache_mutex);
@@ -50,14 +49,14 @@ scsi_select_sense_cache(struct Scsi_Host *shost)
 		scsi_sense_isadma_cache : scsi_sense_cache;
 }
 
-void scsi_free_sense_buffer(struct Scsi_Host *shost,
+static void scsi_free_sense_buffer(struct Scsi_Host *shost,
 		unsigned char *sense_buffer)
 {
 	kmem_cache_free(scsi_select_sense_cache(shost), sense_buffer);
 }
 
-unsigned char *scsi_alloc_sense_buffer(struct Scsi_Host *shost, gfp_t gfp_mask,
-		int numa_node)
+static unsigned char *scsi_alloc_sense_buffer(struct Scsi_Host *shost,
+	gfp_t gfp_mask, int numa_node)
 {
 	return kmem_cache_alloc_node(scsi_select_sense_cache(shost), gfp_mask,
 			numa_node);
@@ -697,14 +696,13 @@ static bool scsi_end_request(struct request *req, int error,
 
 		if (bidi_bytes)
 			scsi_release_bidi_buffers(cmd);
+		scsi_release_buffers(cmd);
+		scsi_put_command(cmd);
 
 		spin_lock_irqsave(q->queue_lock, flags);
 		blk_finish_request(req, error);
 		spin_unlock_irqrestore(q->queue_lock, flags);
 
-		scsi_release_buffers(cmd);
-
-		scsi_put_command(cmd);
 		scsi_run_queue(q);
 	}
 
@@ -1161,34 +1159,22 @@ int scsi_init_io(struct scsi_cmnd *cmd)
 }
 EXPORT_SYMBOL(scsi_init_io);
 
-static struct scsi_cmnd *scsi_get_cmd_from_req(struct scsi_device *sdev,
-		struct request *req)
+void scsi_init_command(struct scsi_device *dev, struct scsi_cmnd *cmd)
 {
-	struct scsi_cmnd *cmd;
-
-	if (!req->special) {
-		/* Bail if we can't get a reference to the device */
-		if (!get_device(&sdev->sdev_gendev))
-			return NULL;
-
-		cmd = scsi_get_command(sdev, GFP_ATOMIC);
-		if (unlikely(!cmd)) {
-			put_device(&sdev->sdev_gendev);
-			return NULL;
-		}
-		req->special = cmd;
-	} else {
-		cmd = req->special;
-	}
-
-	/* pull a tag out of the request if we have one */
-	cmd->tag = req->tag;
-	cmd->request = req;
+	void *buf = cmd->sense_buffer;
+	void *prot = cmd->prot_sdb;
+	unsigned long flags;
 
-	cmd->cmnd = req->cmd;
-	cmd->prot_op = SCSI_PROT_NORMAL;
+	memset(cmd, 0, sizeof(*cmd));
+	cmd->device = dev;
+	cmd->sense_buffer = buf;
+	cmd->prot_sdb = prot;
+	INIT_DELAYED_WORK(&cmd->abort_work, scmd_eh_abort_handler);
+	cmd->jiffies_at_alloc = jiffies;
 
-	return cmd;
+	spin_lock_irqsave(&dev->list_lock, flags);
+	list_add_tail(&cmd->list, &dev->cmd_list);
+	spin_unlock_irqrestore(&dev->list_lock, flags);
 }
 
 static int scsi_setup_blk_pc_cmnd(struct scsi_device *sdev, struct request *req)
@@ -1349,19 +1335,29 @@ scsi_prep_return(struct request_queue *q, struct request *req, int ret)
 static int scsi_prep_fn(struct request_queue *q, struct request *req)
 {
 	struct scsi_device *sdev = q->queuedata;
-	struct scsi_cmnd *cmd;
+	struct scsi_cmnd *cmd = blk_mq_rq_to_pdu(req);
 	int ret;
 
 	ret = scsi_prep_state_check(sdev, req);
 	if (ret != BLKPREP_OK)
 		goto out;
 
-	cmd = scsi_get_cmd_from_req(sdev, req);
-	if (unlikely(!cmd)) {
-		ret = BLKPREP_DEFER;
-		goto out;
+	if (!req->special) {
+		/* Bail if we can't get a reference to the device */
+		if (unlikely(!get_device(&sdev->sdev_gendev))) {
+			ret = BLKPREP_DEFER;
+			goto out;
+		}
+
+		scsi_init_command(sdev, cmd);
+		req->special = cmd;
 	}
 
+	cmd->tag = req->tag;
+	cmd->request = req;
+	cmd->cmnd = req->cmd;
+	cmd->prot_op = SCSI_PROT_NORMAL;
+
 	ret = scsi_setup_cmnd(sdev, req);
 out:
 	return scsi_prep_return(q, req, ret);
@@ -2119,15 +2115,61 @@ void __scsi_init_queue(struct Scsi_Host *shost, struct request_queue *q)
 }
 EXPORT_SYMBOL_GPL(__scsi_init_queue);
 
+static int scsi_init_rq(struct request_queue *q, struct request *rq, gfp_t gfp)
+{
+	struct Scsi_Host *shost = q->rq_alloc_data;
+	struct scsi_cmnd *cmd = blk_mq_rq_to_pdu(rq);
+
+	memset(cmd, 0, sizeof(*cmd));
+
+	cmd->sense_buffer = scsi_alloc_sense_buffer(shost, gfp, NUMA_NO_NODE);
+	if (!cmd->sense_buffer)
+		goto fail;
+
+	if (scsi_host_get_prot(shost) >= SHOST_DIX_TYPE0_PROTECTION) {
+		cmd->prot_sdb = kmem_cache_zalloc(scsi_sdb_cache, gfp);
+		if (!cmd->prot_sdb)
+			goto fail_free_sense;
+	}
+
+	return 0;
+
+fail_free_sense:
+	scsi_free_sense_buffer(shost, cmd->sense_buffer);
+fail:
+	return -ENOMEM;
+}
+
+static void scsi_exit_rq(struct request_queue *q, struct request *rq)
+{
+	struct Scsi_Host *shost = q->rq_alloc_data;
+	struct scsi_cmnd *cmd = blk_mq_rq_to_pdu(rq);
+
+	if (cmd->prot_sdb)
+		kmem_cache_free(scsi_sdb_cache, cmd->prot_sdb);
+	scsi_free_sense_buffer(shost, cmd->sense_buffer);
+}
+
 struct request_queue *scsi_alloc_queue(struct scsi_device *sdev)
 {
+	struct Scsi_Host *shost = sdev->host;
 	struct request_queue *q;
 
-	q = blk_init_queue(scsi_request_fn, NULL);
+	q = blk_alloc_queue_node(GFP_KERNEL, NUMA_NO_NODE);
 	if (!q)
 		return NULL;
+	q->cmd_size = sizeof(struct scsi_cmnd) + shost->hostt->cmd_size;
+	q->rq_alloc_data = shost;
+	q->request_fn = scsi_request_fn;
+	q->init_rq_fn = scsi_init_rq;
+	q->exit_rq_fn = scsi_exit_rq;
+
+	if (blk_init_allocated_queue(q) < 0) {
+		blk_cleanup_queue(q);
+		return NULL;
+	}
 
-	__scsi_init_queue(sdev->host, q);
+	__scsi_init_queue(shost, q);
 	blk_queue_prep_rq(q, scsi_prep_fn);
 	blk_queue_unprep_rq(q, scsi_unprep_fn);
 	blk_queue_softirq_done(q, scsi_softirq_done);
diff --git a/drivers/scsi/scsi_priv.h b/drivers/scsi/scsi_priv.h
index 1a712c6..99bfc98 100644
--- a/drivers/scsi/scsi_priv.h
+++ b/drivers/scsi/scsi_priv.h
@@ -30,13 +30,8 @@ extern void scsi_exit_hosts(void);
 
 /* scsi.c */
 extern bool scsi_use_blk_mq;
-void scsi_free_sense_buffer(struct Scsi_Host *shost,
-		unsigned char *sense_buffer);
-unsigned char *scsi_alloc_sense_buffer(struct Scsi_Host *shost, gfp_t gfp_mask,
-		int numa_node);
 int scsi_init_sense_cache(struct Scsi_Host *shost);
-extern int scsi_setup_command_freelist(struct Scsi_Host *shost);
-extern void scsi_destroy_command_freelist(struct Scsi_Host *shost);
+void scsi_init_command(struct scsi_device *dev, struct scsi_cmnd *cmd);
 #ifdef CONFIG_SCSI_LOGGING
 void scsi_log_send(struct scsi_cmnd *cmd);
 void scsi_log_completion(struct scsi_cmnd *cmd, int disposition);
@@ -101,7 +96,6 @@ extern void scsi_exit_queue(void);
 extern void scsi_evt_thread(struct work_struct *work);
 struct request_queue;
 struct request;
-extern struct kmem_cache *scsi_sdb_cache;
 
 /* scsi_proc.c */
 #ifdef CONFIG_SCSI_PROC_FS
diff --git a/include/scsi/scsi_host.h b/include/scsi/scsi_host.h
index f4964d7..3cd8c3b 100644
--- a/include/scsi/scsi_host.h
+++ b/include/scsi/scsi_host.h
@@ -551,9 +551,6 @@ struct Scsi_Host {
 	struct list_head	__devices;
 	struct list_head	__targets;
 	
-	struct scsi_host_cmd_pool *cmd_pool;
-	spinlock_t		free_list_lock;
-	struct list_head	free_list; /* backup store of cmd structs */
 	struct list_head	starved_list;
 
 	spinlock_t		default_lock;
-- 
2.1.4

[PATCH 14/15] block/bsg: move queue creation into bsg_setup_queue

[Christoph Hellwig]

Simply the boilerplate code needed for bsg nodes a bit.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 block/bsg-lib.c                     | 21 +++++++++++----------
 drivers/scsi/scsi_transport_fc.c    | 36 ++++++++----------------------------
 drivers/scsi/scsi_transport_iscsi.c | 15 ++++-----------
 include/linux/bsg-lib.h             |  5 ++---
 4 files changed, 25 insertions(+), 52 deletions(-)

diff --git a/block/bsg-lib.c b/block/bsg-lib.c
index 9d652a9..c74acf4 100644
--- a/block/bsg-lib.c
+++ b/block/bsg-lib.c
@@ -177,7 +177,7 @@ static int bsg_create_job(struct device *dev, struct request *req)
  *
  * Drivers/subsys should pass this to the queue init function.
  */
-void bsg_request_fn(struct request_queue *q)
+static void bsg_request_fn(struct request_queue *q)
 	__releases(q->queue_lock)
 	__acquires(q->queue_lock)
 {
@@ -214,24 +214,24 @@ void bsg_request_fn(struct request_queue *q)
 	put_device(dev);
 	spin_lock_irq(q->queue_lock);
 }
-EXPORT_SYMBOL_GPL(bsg_request_fn);
 
 /**
  * bsg_setup_queue - Create and add the bsg hooks so we can receive requests
  * @dev: device to attach bsg device to
- * @q: request queue setup by caller
  * @name: device to give bsg device
  * @job_fn: bsg job handler
  * @dd_job_size: size of LLD data needed for each job
- *
- * The caller should have setup the reuqest queue with bsg_request_fn
- * as the request_fn.
  */
-int bsg_setup_queue(struct device *dev, struct request_queue *q,
-		    char *name, bsg_job_fn *job_fn, int dd_job_size)
+struct request_queue *bsg_setup_queue(struct device *dev, char *name,
+		bsg_job_fn *job_fn, int dd_job_size)
 {
+	struct request_queue *q;
 	int ret;
 
+	q = blk_init_queue(bsg_request_fn, NULL);
+	if (!q)
+		return ERR_PTR(-ENOMEM);
+
 	q->queuedata = dev;
 	q->bsg_job_size = dd_job_size;
 	q->bsg_job_fn = job_fn;
@@ -243,9 +243,10 @@ int bsg_setup_queue(struct device *dev, struct request_queue *q,
 	if (ret) {
 		printk(KERN_ERR "%s: bsg interface failed to "
 		       "initialize - register queue\n", dev->kobj.name);
-		return ret;
+		blk_cleanup_queue(q);
+		return ERR_PTR(ret);
 	}
 
-	return 0;
+	return q;
 }
 EXPORT_SYMBOL_GPL(bsg_setup_queue);
diff --git a/drivers/scsi/scsi_transport_fc.c b/drivers/scsi/scsi_transport_fc.c
index afcedec..13dcb9b 100644
--- a/drivers/scsi/scsi_transport_fc.c
+++ b/drivers/scsi/scsi_transport_fc.c
@@ -3765,7 +3765,6 @@ fc_bsg_hostadd(struct Scsi_Host *shost, struct fc_host_attrs *fc_host)
 	struct device *dev = &shost->shost_gendev;
 	struct fc_internal *i = to_fc_internal(shost->transportt);
 	struct request_queue *q;
-	int err;
 	char bsg_name[20];
 
 	fc_host->rqst_q = NULL;
@@ -3776,24 +3775,14 @@ fc_bsg_hostadd(struct Scsi_Host *shost, struct fc_host_attrs *fc_host)
 	snprintf(bsg_name, sizeof(bsg_name),
 		 "fc_host%d", shost->host_no);
 
-	q = blk_init_queue(bsg_request_fn, NULL);
-	if (!q) {
-		dev_err(dev,
-			"fc_host%d: bsg interface failed to initialize - no request queue\n",
-			shost->host_no);
-		return -ENOMEM;
-	}
-
-	__scsi_init_queue(shost, q);
-	err = bsg_setup_queue(dev, q, bsg_name, fc_bsg_dispatch,
-				 i->f->dd_bsg_size);
-	if (err) {
+	q = bsg_setup_queue(dev, bsg_name, fc_bsg_dispatch, i->f->dd_bsg_size);
+	if (IS_ERR(q)) {
 		dev_err(dev,
 			"fc_host%d: bsg interface failed to initialize - setup queue\n",
 			shost->host_no);
-		blk_cleanup_queue(q);
-		return err;
+		return PTR_ERR(q);
 	}
+	__scsi_init_queue(shost, q);
 	blk_queue_rq_timed_out(q, fc_bsg_job_timeout);
 	blk_queue_rq_timeout(q, FC_DEFAULT_BSG_TIMEOUT);
 	fc_host->rqst_q = q;
@@ -3825,27 +3814,18 @@ fc_bsg_rportadd(struct Scsi_Host *shost, struct fc_rport *rport)
 	struct device *dev = &rport->dev;
 	struct fc_internal *i = to_fc_internal(shost->transportt);
 	struct request_queue *q;
-	int err;
 
 	rport->rqst_q = NULL;
 
 	if (!i->f->bsg_request)
 		return -ENOTSUPP;
 
-	q = blk_init_queue(bsg_request_fn, NULL);
-	if (!q) {
-		dev_err(dev, "bsg interface failed to initialize - no request queue\n");
-		return -ENOMEM;
-	}
-
-	__scsi_init_queue(shost, q);
-	err = bsg_setup_queue(dev, q, NULL, fc_bsg_dispatch, i->f->dd_bsg_size);
-	if (err) {
+	q = bsg_setup_queue(dev, NULL, fc_bsg_dispatch, i->f->dd_bsg_size);
+	if (IS_ERR(q)) {
 		dev_err(dev, "failed to setup bsg queue\n");
-		blk_cleanup_queue(q);
-		return err;
+		return PTR_ERR(q);
 	}
-
+	__scsi_init_queue(shost, q);
 	blk_queue_prep_rq(q, fc_bsg_rport_prep);
 	blk_queue_rq_timed_out(q, fc_bsg_job_timeout);
 	blk_queue_rq_timeout(q, BLK_DEFAULT_SG_TIMEOUT);
diff --git a/drivers/scsi/scsi_transport_iscsi.c b/drivers/scsi/scsi_transport_iscsi.c
index 04ebe6e..568c9f2 100644
--- a/drivers/scsi/scsi_transport_iscsi.c
+++ b/drivers/scsi/scsi_transport_iscsi.c
@@ -1537,25 +1537,18 @@ iscsi_bsg_host_add(struct Scsi_Host *shost, struct iscsi_cls_host *ihost)
 	struct iscsi_internal *i = to_iscsi_internal(shost->transportt);
 	struct request_queue *q;
 	char bsg_name[20];
-	int ret;
 
 	if (!i->iscsi_transport->bsg_request)
 		return -ENOTSUPP;
 
 	snprintf(bsg_name, sizeof(bsg_name), "iscsi_host%d", shost->host_no);
-
-	q = blk_init_queue(bsg_request_fn, NULL);
-	if (!q)
-		return -ENOMEM;
-
-	__scsi_init_queue(shost, q);
-	ret = bsg_setup_queue(dev, q, bsg_name, iscsi_bsg_host_dispatch, 0);
-	if (ret) {
+	q = bsg_setup_queue(dev, bsg_name, iscsi_bsg_host_dispatch, 0);
+	if (IS_ERR(q)) {
 		shost_printk(KERN_ERR, shost, "bsg interface failed to "
 			     "initialize - no request queue\n");
-		blk_cleanup_queue(q);
-		return ret;
+		return PTR_ERR(q);
 	}
+	__scsi_init_queue(shost, q);
 
 	ihost->bsg_q = q;
 	return 0;
diff --git a/include/linux/bsg-lib.h b/include/linux/bsg-lib.h
index 657a718..e34dde2 100644
--- a/include/linux/bsg-lib.h
+++ b/include/linux/bsg-lib.h
@@ -66,9 +66,8 @@ struct bsg_job {
 
 void bsg_job_done(struct bsg_job *job, int result,
 		  unsigned int reply_payload_rcv_len);
-int bsg_setup_queue(struct device *dev, struct request_queue *q, char *name,
-		    bsg_job_fn *job_fn, int dd_job_size);
-void bsg_request_fn(struct request_queue *q);
+struct request_queue *bsg_setup_queue(struct device *dev, char *name,
+		bsg_job_fn *job_fn, int dd_job_size);
 void bsg_job_put(struct bsg_job *job);
 int __must_check bsg_job_get(struct bsg_job *job);
 
-- 
2.1.4

Re: [PATCH 14/15] block/bsg: move queue creation into bsg_setup_queue

[Johannes Thumshirn]

On Tue, Jan 10, 2017 at 04:06:19PM +0100, Christoph Hellwig wrote:
> Simply the boilerplate code needed for bsg nodes a bit.
> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> ---

that reminds me of posting my SAS bsg-lib patch...

Anyways looks good,
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>

-- 
Johannes Thumshirn                                          Storage
jthumshirn@suse.de                                +49 911 74053 689
SUSE LINUX GmbH, Maxfeldstr. 5, 90409 N�rnberg
GF: Felix Imend�rffer, Jane Smithard, Graham Norton
HRB 21284 (AG N�rnberg)
Key fingerprint = EC38 9CAB C2C4 F25D 8600 D0D0 0393 969D 2D76 0850

Re: [PATCH 14/15] block/bsg: move queue creation into bsg_setup_queue

[Christoph Hellwig]

On Wed, Jan 11, 2017 at 09:42:44AM +0100, Johannes Thumshirn wrote:
> On Tue, Jan 10, 2017 at 04:06:19PM +0100, Christoph Hellwig wrote:
> > Simply the boilerplate code needed for bsg nodes a bit.
> > 
> > Signed-off-by: Christoph Hellwig <hch@lst.de>
> > ---
> 
> that reminds me of posting my SAS bsg-lib patch...

Yes.  Having SAS use bsg-lib, and bsg-lib switched away from abusing
struct request_queue would make this series a lot cleaner.

So maybe we should get that into the scsi tree for 4.10 together
with the prep patches in this series as a priority and defer the actual
struct request changes once again.  That should also give us some more
time to sort out the dm-mpath story..

Re: [PATCH 14/15] block/bsg: move queue creation into bsg_setup_queue

[Johannes Thumshirn]

On Wed, Jan 11, 2017 at 09:45:12AM +0100, Christoph Hellwig wrote:
> On Wed, Jan 11, 2017 at 09:42:44AM +0100, Johannes Thumshirn wrote:
> > On Tue, Jan 10, 2017 at 04:06:19PM +0100, Christoph Hellwig wrote:
> > > Simply the boilerplate code needed for bsg nodes a bit.
> > > 
> > > Signed-off-by: Christoph Hellwig <hch@lst.de>
> > > ---
> > 
> > that reminds me of posting my SAS bsg-lib patch...
> 
> Yes.  Having SAS use bsg-lib, and bsg-lib switched away from abusing
> struct request_queue would make this series a lot cleaner.
> 
> So maybe we should get that into the scsi tree for 4.10 together
> with the prep patches in this series as a priority and defer the actual
> struct request changes once again.  That should also give us some more
> time to sort out the dm-mpath story..

I'll dig it up and RFC post it. It's currently untested though as I
currently don't have a SMP capable SAS HBA here.

Do you have an mptXsas available? 

Byte,
	Johannes
-- 
Johannes Thumshirn                                          Storage
jthumshirn@suse.de                                +49 911 74053 689
SUSE LINUX GmbH, Maxfeldstr. 5, 90409 N�rnberg
GF: Felix Imend�rffer, Jane Smithard, Graham Norton
HRB 21284 (AG N�rnberg)
Key fingerprint = EC38 9CAB C2C4 F25D 8600 D0D0 0393 969D 2D76 0850

Re: [PATCH 14/15] block/bsg: move queue creation into bsg_setup_queue

[Christoph Hellwig]

On Wed, Jan 11, 2017 at 09:56:01AM +0100, Johannes Thumshirn wrote:
> I'll dig it up and RFC post it. It's currently untested though as I
> currently don't have a SMP capable SAS HBA here.
> 
> Do you have an mptXsas available? 

Unfortunately not.  But I think Hannes has, he has recently played
around with the interrupt handling for the driver.

Re: [PATCH 14/15] block/bsg: move queue creation into bsg_setup_queue

[Hannes Reinecke]

On 01/11/2017 09:45 AM, Christoph Hellwig wrote:
> On Wed, Jan 11, 2017 at 09:42:44AM +0100, Johannes Thumshirn wrote:
>> On Tue, Jan 10, 2017 at 04:06:19PM +0100, Christoph Hellwig wrote:
>>> Simply the boilerplate code needed for bsg nodes a bit.
>>>
>>> Signed-off-by: Christoph Hellwig <hch@lst.de>
>>> ---
>>
>> that reminds me of posting my SAS bsg-lib patch...
> 
> Yes.  Having SAS use bsg-lib, and bsg-lib switched away from abusing
> struct request_queue would make this series a lot cleaner.
> 
> So maybe we should get that into the scsi tree for 4.10 together
> with the prep patches in this series as a priority and defer the actual
> struct request changes once again.  That should also give us some more
> time to sort out the dm-mpath story..
I'd advocate to discuss this at LSF.
Now that Mike moved the bio-based mpath stuff back in things got even
more complex.

I'll be posting a patchset for reimplementing multipath as a stand-alone
driver shortly; that'll give us a good starting point on how we want
multipath to evolve.

Who knows; we might even manage to move multipath out of device-mapper
altogether.
That would make Mike very happy, and I wouldn't mind, either :-)

Cheers,

Hannes
-- 
Dr. Hannes Reinecke		   Teamlead Storage & Networking
hare@suse.de			               +49 911 74053 688
SUSE LINUX GmbH, Maxfeldstr. 5, 90409 N�rnberg
GF: F. Imend�rffer, J. Smithard, J. Guild, D. Upmanyu, G. Norton
HRB 21284 (AG N�rnberg)

Re: [PATCH 14/15] block/bsg: move queue creation into bsg_setup_queue

[Christoph Hellwig]

On Wed, Jan 11, 2017 at 09:59:17AM +0100, Hannes Reinecke wrote:
> I'd advocate to discuss this at LSF.
> Now that Mike moved the bio-based mpath stuff back in things got even
> more complex.

Yeah.  If we'd _only_ have bio based support it would simplify things
a lot, but as a third parallel path it's not exactly making things easier.

> I'll be posting a patchset for reimplementing multipath as a stand-alone
> driver shortly; that'll give us a good starting point on how we want
> multipath to evolve.
> 
> Who knows; we might even manage to move multipath out of device-mapper
> altogether.
> That would make Mike very happy, and I wouldn't mind, either :-)

Heh.  I'm curious how you want to do that while keeping existing setups
working, though.

Re: [PATCH 14/15] block/bsg: move queue creation into bsg_setup_queue

[Hannes Reinecke]

On 01/11/2017 10:01 AM, Christoph Hellwig wrote:
> On Wed, Jan 11, 2017 at 09:59:17AM +0100, Hannes Reinecke wrote:
>> I'd advocate to discuss this at LSF.
>> Now that Mike moved the bio-based mpath stuff back in things got even
>> more complex.
> 
> Yeah.  If we'd _only_ have bio based support it would simplify things
> a lot, but as a third parallel path it's not exactly making things easier.
> 
>> I'll be posting a patchset for reimplementing multipath as a stand-alone
>> driver shortly; that'll give us a good starting point on how we want
>> multipath to evolve.
>>
>> Who knows; we might even manage to move multipath out of device-mapper
>> altogether.
>> That would make Mike very happy, and I wouldn't mind, either :-)
> 
> Heh.  I'm curious how you want to do that while keeping existing setups
> working, though.

which will become challenging, indeed.

ATM it's just a testbed on how things could work; we've got most of the
required infrastucture in the kernel nowadays, so that we can just drop
most of the complexity from the present multipath-tools mess.

In the end it might even boil down to update the existing device-mapper
multipath implementation. We'll see.

Cheers,

Hannes
-- 
Dr. Hannes Reinecke		   Teamlead Storage & Networking
hare@suse.de			               +49 911 74053 688
SUSE LINUX GmbH, Maxfeldstr. 5, 90409 N�rnberg
GF: F. Imend�rffer, J. Smithard, J. Guild, D. Upmanyu, G. Norton
HRB 21284 (AG N�rnberg)

Re: [PATCH 14/15] block/bsg: move queue creation into bsg_setup_queue

[Mike Snitzer]

On Wed, Jan 11 2017 at  4:37am -0500,
Hannes Reinecke <hare@suse.de> wrote:

> On 01/11/2017 10:01 AM, Christoph Hellwig wrote:
> > On Wed, Jan 11, 2017 at 09:59:17AM +0100, Hannes Reinecke wrote:
> >> I'd advocate to discuss this at LSF.
> >> Now that Mike moved the bio-based mpath stuff back in things got even
> >> more complex.
> > 
> > Yeah.  If we'd _only_ have bio based support it would simplify things
> > a lot, but as a third parallel path it's not exactly making things easier.
> > 
> >> I'll be posting a patchset for reimplementing multipath as a stand-alone
> >> driver shortly; that'll give us a good starting point on how we want
> >> multipath to evolve.
> >>
> >> Who knows; we might even manage to move multipath out of device-mapper
> >> altogether.
> >> That would make Mike very happy, and I wouldn't mind, either :-)
> > 
> > Heh.  I'm curious how you want to do that while keeping existing setups
> > working, though.
> 
> which will become challenging, indeed.
> 
> ATM it's just a testbed on how things could work; we've got most of the
> required infrastucture in the kernel nowadays, so that we can just drop
> most of the complexity from the present multipath-tools mess.
> 
> In the end it might even boil down to update the existing device-mapper
> multipath implementation. We'll see.

Honestly, at this point I'm pretty confident you guys will never
actually produce a new NIH multipath.  You'll just threaten to for
_years_.

Joking aside, I think DM multipath would benefit from a more focused
separation from all the SCSI cruft that wormed its way in.  Factoring
that stuff out so that multipath can be more easily adapted/optimized
for use with leaner transports like NVMe-over-fabrics.

But NVMe-over-fabrics setups are still pretty rare (only a handful of
vendors are stepping up with requirements to support it) so details on
what change is needed to DM multipath is largely unknown to me.  But may
become clearer in the coming month.

Mike

Re: [PATCH 14/15] block/bsg: move queue creation into bsg_setup_queue

[Mike Snitzer]

On Wed, Jan 11 2017 at  3:45am -0500,
Christoph Hellwig <hch@lst.de> wrote:

> On Wed, Jan 11, 2017 at 09:42:44AM +0100, Johannes Thumshirn wrote:
> > On Tue, Jan 10, 2017 at 04:06:19PM +0100, Christoph Hellwig wrote:
> > > Simply the boilerplate code needed for bsg nodes a bit.
> > > 
> > > Signed-off-by: Christoph Hellwig <hch@lst.de>
> > > ---
> > 
> > that reminds me of posting my SAS bsg-lib patch...
> 
> Yes.  Having SAS use bsg-lib, and bsg-lib switched away from abusing
> struct request_queue would make this series a lot cleaner.
> 
> So maybe we should get that into the scsi tree for 4.10 together
> with the prep patches in this series as a priority and defer the actual
> struct request changes once again.  That should also give us some more
> time to sort out the dm-mpath story..

I'm not aware of the story you're referring to.  I'm missing the actual
challenge you're facing.

But I've seen you reference the need to stop multipath from allocating
its own requests.  Are you referring to old request_fn request-based
multipath's clone_old_rq:alloc_old_clone_request?

Or how blk-mq request-based multipath gets a request from the blk-mq tag
space (via blk_mq_alloc_request)?

Or both?  How is that holding you back?

Re: [PATCH 14/15] block/bsg: move queue creation into bsg_setup_queue

[Christoph Hellwig]

On Wed, Jan 11, 2017 at 05:01:22PM -0500, Mike Snitzer wrote:
> But I've seen you reference the need to stop multipath from allocating
> its own requests.  Are you referring to old request_fn request-based
> multipath's clone_old_rq:alloc_old_clone_request?

Yes, that one is the issue.  It allocates a struct request "blind",
that is without known what queue it goes to.  With this queue (or blk-mq
for that matter) we need to know the queue, because the request structures
might have additional data behind it and require additional initialization
for drivers that require per-request data.  We make use of the per-request
data for SCSI passthrough in this patch.

> Or how blk-mq request-based multipath gets a request from the blk-mq tag
> space (via blk_mq_alloc_request)?

That's fine because it works on the queue of the device that I/O is
submitted to.

[PATCH 15/15] block: split scsi_request out of struct request

[Christoph Hellwig]

And require all drivers that want to support BLOCK_PC to allocate it
as the first thing of their private data.  To support this the legacy
IDE and BSG code is switched to set cmd_size on their queues to let
the block layer allocate the additional space.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 block/blk-core.c                         | 31 -----------
 block/blk-exec.c                         | 12 -----
 block/blk-mq.c                           | 10 ----
 block/bsg-lib.c                          | 34 ++++++++----
 block/bsg.c                              | 47 ++++++++---------
 block/scsi_ioctl.c                       | 87 ++++++++++++++++++------------
 drivers/ata/libata-scsi.c                |  2 +-
 drivers/block/cciss.c                    | 28 +++++-----
 drivers/block/pktcdvd.c                  |  6 +--
 drivers/block/virtio_blk.c               | 11 ++--
 drivers/cdrom/cdrom.c                    | 32 +++++------
 drivers/ide/ide-atapi.c                  | 43 ++++++++-------
 drivers/ide/ide-cd.c                     | 91 +++++++++++++++-----------------
 drivers/ide/ide-cd_ioctl.c               |  1 +
 drivers/ide/ide-cd_verbose.c             |  6 +--
 drivers/ide/ide-devsets.c                |  9 ++--
 drivers/ide/ide-disk.c                   |  1 +
 drivers/ide/ide-eh.c                     |  2 +-
 drivers/ide/ide-floppy.c                 |  4 +-
 drivers/ide/ide-io.c                     |  3 +-
 drivers/ide/ide-ioctls.c                 |  6 ++-
 drivers/ide/ide-park.c                   | 12 +++--
 drivers/ide/ide-pm.c                     |  2 +
 drivers/ide/ide-probe.c                  | 36 +++++++++++--
 drivers/ide/ide-tape.c                   | 32 +++++------
 drivers/ide/ide-taskfile.c               |  1 +
 drivers/ide/sis5513.c                    |  2 +-
 drivers/message/fusion/mptsas.c          |  8 +--
 drivers/scsi/libfc/fc_lport.c            |  2 +-
 drivers/scsi/libsas/sas_expander.c       |  8 +--
 drivers/scsi/libsas/sas_host_smp.c       | 38 ++++++-------
 drivers/scsi/mpt3sas/mpt3sas_transport.c |  8 +--
 drivers/scsi/osd/osd_initiator.c         | 19 +++----
 drivers/scsi/osst.c                      | 15 +++---
 drivers/scsi/qla2xxx/qla_bsg.c           |  2 +-
 drivers/scsi/qla2xxx/qla_isr.c           |  6 ++-
 drivers/scsi/qla2xxx/qla_mr.c            |  2 +-
 drivers/scsi/scsi_error.c                | 22 ++++----
 drivers/scsi/scsi_lib.c                  | 48 +++++++++--------
 drivers/scsi/scsi_transport_sas.c        |  5 ++
 drivers/scsi/sd.c                        |  4 +-
 drivers/scsi/sg.c                        | 30 ++++++-----
 drivers/scsi/st.c                        | 22 ++++----
 drivers/target/target_core_pscsi.c       | 11 ++--
 fs/nfsd/blocklayout.c                    | 17 +++---
 include/linux/blkdev.h                   | 11 ----
 include/linux/blktrace_api.h             |  3 +-
 include/linux/ide.h                      |  8 ++-
 include/scsi/scsi_cmnd.h                 |  2 +
 include/scsi/scsi_request.h              | 28 ++++++++++
 kernel/trace/blktrace.c                  | 32 +++++------
 51 files changed, 483 insertions(+), 419 deletions(-)
 create mode 100644 include/scsi/scsi_request.h

diff --git a/block/blk-core.c b/block/blk-core.c
index 7190bc1..9888819 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -131,8 +131,6 @@ void blk_rq_init(struct request_queue *q, struct request *rq)
 	rq->__sector = (sector_t) -1;
 	INIT_HLIST_NODE(&rq->hash);
 	RB_CLEAR_NODE(&rq->rb_node);
-	rq->cmd = rq->__cmd;
-	rq->cmd_len = BLK_MAX_CDB;
 	rq->tag = -1;
 	rq->start_time = jiffies;
 	set_start_time_ns(rq);
@@ -158,8 +156,6 @@ static void req_bio_endio(struct request *rq, struct bio *bio,
 
 void blk_dump_rq_flags(struct request *rq, char *msg)
 {
-	int bit;
-
 	printk(KERN_INFO "%s: dev %s: type=%x, flags=%llx\n", msg,
 		rq->rq_disk ? rq->rq_disk->disk_name : "?", rq->cmd_type,
 		(unsigned long long) rq->cmd_flags);
@@ -169,13 +165,6 @@ void blk_dump_rq_flags(struct request *rq, char *msg)
 	       blk_rq_sectors(rq), blk_rq_cur_sectors(rq));
 	printk(KERN_INFO "  bio %p, biotail %p, len %u\n",
 	       rq->bio, rq->biotail, blk_rq_bytes(rq));
-
-	if (rq->cmd_type == REQ_TYPE_BLOCK_PC) {
-		printk(KERN_INFO "  cdb: ");
-		for (bit = 0; bit < BLK_MAX_CDB; bit++)
-			printk("%02x ", rq->cmd[bit]);
-		printk("\n");
-	}
 }
 EXPORT_SYMBOL(blk_dump_rq_flags);
 
@@ -1348,18 +1337,6 @@ struct request *blk_get_request(struct request_queue *q, int rw, gfp_t gfp_mask)
 EXPORT_SYMBOL(blk_get_request);
 
 /**
- * blk_rq_set_block_pc - initialize a request to type BLOCK_PC
- * @rq:		request to be initialized
- *
- */
-void blk_rq_set_block_pc(struct request *rq)
-{
-	rq->cmd_type = REQ_TYPE_BLOCK_PC;
-	memset(rq->__cmd, 0, sizeof(rq->__cmd));
-}
-EXPORT_SYMBOL(blk_rq_set_block_pc);
-
-/**
  * blk_requeue_request - put a request back on queue
  * @q:		request queue where request should be inserted
  * @rq:		request to be inserted
@@ -2491,14 +2468,6 @@ void blk_start_request(struct request *req)
 		wbt_issue(req->q->rq_wb, &req->issue_stat);
 	}
 
-	/*
-	 * We are now handing the request to the hardware, initialize
-	 * resid_len to full count and add the timeout handler.
-	 */
-	req->resid_len = blk_rq_bytes(req);
-	if (unlikely(blk_bidi_rq(req)))
-		req->next_rq->resid_len = blk_rq_bytes(req->next_rq);
-
 	BUG_ON(test_bit(REQ_ATOM_COMPLETE, &req->atomic_flags));
 	blk_add_timer(req);
 }
diff --git a/block/blk-exec.c b/block/blk-exec.c
index 3ecb00a..3b83903 100644
--- a/block/blk-exec.c
+++ b/block/blk-exec.c
@@ -100,16 +100,9 @@ int blk_execute_rq(struct request_queue *q, struct gendisk *bd_disk,
 		   struct request *rq, int at_head)
 {
 	DECLARE_COMPLETION_ONSTACK(wait);
-	char sense[SCSI_SENSE_BUFFERSIZE];
 	int err = 0;
 	unsigned long hang_check;
 
-	if (!rq->sense) {
-		memset(sense, 0, sizeof(sense));
-		rq->sense = sense;
-		rq->sense_len = 0;
-	}
-
 	rq->end_io_data = &wait;
 	blk_execute_rq_nowait(q, bd_disk, rq, at_head, blk_end_sync_rq);
 
@@ -123,11 +116,6 @@ int blk_execute_rq(struct request_queue *q, struct gendisk *bd_disk,
 	if (rq->errors)
 		err = -EIO;
 
-	if (rq->sense == sense)	{
-		rq->sense = NULL;
-		rq->sense_len = 0;
-	}
-
 	return err;
 }
 EXPORT_SYMBOL(blk_execute_rq);
diff --git a/block/blk-mq.c b/block/blk-mq.c
index a8e67a1..1baa0cd 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -196,13 +196,7 @@ static void blk_mq_rq_ctx_init(struct request_queue *q, struct blk_mq_ctx *ctx,
 	rq->special = NULL;
 	/* tag was already set */
 	rq->errors = 0;
-
-	rq->cmd = rq->__cmd;
-
 	rq->extra_len = 0;
-	rq->sense_len = 0;
-	rq->resid_len = 0;
-	rq->sense = NULL;
 
 	INIT_LIST_HEAD(&rq->timeout_list);
 	rq->timeout = 0;
@@ -469,10 +463,6 @@ void blk_mq_start_request(struct request *rq)
 
 	trace_block_rq_issue(q, rq);
 
-	rq->resid_len = blk_rq_bytes(rq);
-	if (unlikely(blk_bidi_rq(rq)))
-		rq->next_rq->resid_len = blk_rq_bytes(rq->next_rq);
-
 	if (test_bit(QUEUE_FLAG_STATS, &q->queue_flags)) {
 		blk_stat_set_issue_time(&rq->issue_stat);
 		rq->rq_flags |= RQF_STATS;
diff --git a/block/bsg-lib.c b/block/bsg-lib.c
index c74acf4..cd15f9d 100644
--- a/block/bsg-lib.c
+++ b/block/bsg-lib.c
@@ -71,22 +71,24 @@ void bsg_job_done(struct bsg_job *job, int result,
 {
 	struct request *req = job->req;
 	struct request *rsp = req->next_rq;
+	struct scsi_request *rq = scsi_req(req);
 	int err;
 
 	err = job->req->errors = result;
 	if (err < 0)
 		/* we're only returning the result field in the reply */
-		job->req->sense_len = sizeof(u32);
+		rq->sense_len = sizeof(u32);
 	else
-		job->req->sense_len = job->reply_len;
+		rq->sense_len = job->reply_len;
 	/* we assume all request payload was transferred, residual == 0 */
-	req->resid_len = 0;
+	rq->resid_len = 0;
 
 	if (rsp) {
-		WARN_ON(reply_payload_rcv_len > rsp->resid_len);
+		WARN_ON(reply_payload_rcv_len > scsi_req(rsp)->resid_len);
 
 		/* set reply (bidi) residual */
-		rsp->resid_len -= min(reply_payload_rcv_len, rsp->resid_len);
+		scsi_req(rsp)->resid_len -=
+			min(reply_payload_rcv_len, scsi_req(rsp)->resid_len);
 	}
 	blk_complete_request(req);
 }
@@ -113,6 +115,7 @@ static int bsg_map_buffer(struct bsg_buffer *buf, struct request *req)
 	if (!buf->sg_list)
 		return -ENOMEM;
 	sg_init_table(buf->sg_list, req->nr_phys_segments);
+	scsi_req(req)->resid_len = blk_rq_bytes(req);
 	buf->sg_cnt = blk_rq_map_sg(req->q, req, buf->sg_list);
 	buf->payload_len = blk_rq_bytes(req);
 	return 0;
@@ -127,6 +130,7 @@ static int bsg_create_job(struct device *dev, struct request *req)
 {
 	struct request *rsp = req->next_rq;
 	struct request_queue *q = req->q;
+	struct scsi_request *rq = scsi_req(req);
 	struct bsg_job *job;
 	int ret;
 
@@ -140,9 +144,9 @@ static int bsg_create_job(struct device *dev, struct request *req)
 	job->req = req;
 	if (q->bsg_job_size)
 		job->dd_data = (void *)&job[1];
-	job->request = req->cmd;
-	job->request_len = req->cmd_len;
-	job->reply = req->sense;
+	job->request = rq->cmd;
+	job->request_len = rq->cmd_len;
+	job->reply = rq->sense;
 	job->reply_len = SCSI_SENSE_BUFFERSIZE;	/* Size of sense buffer
 						 * allocated */
 	if (req->bio) {
@@ -228,9 +232,15 @@ struct request_queue *bsg_setup_queue(struct device *dev, char *name,
 	struct request_queue *q;
 	int ret;
 
-	q = blk_init_queue(bsg_request_fn, NULL);
+	q = blk_alloc_queue(GFP_KERNEL);
 	if (!q)
 		return ERR_PTR(-ENOMEM);
+	q->cmd_size = sizeof(struct scsi_request);
+	q->request_fn = bsg_request_fn;
+
+	ret = blk_init_allocated_queue(q);
+	if (ret)
+		goto out_cleanup_queue;
 
 	q->queuedata = dev;
 	q->bsg_job_size = dd_job_size;
@@ -243,10 +253,12 @@ struct request_queue *bsg_setup_queue(struct device *dev, char *name,
 	if (ret) {
 		printk(KERN_ERR "%s: bsg interface failed to "
 		       "initialize - register queue\n", dev->kobj.name);
-		blk_cleanup_queue(q);
-		return ERR_PTR(ret);
+		goto out_cleanup_queue;
 	}
 
 	return q;
+out_cleanup_queue:
+	blk_cleanup_queue(q);
+	return ERR_PTR(ret);
 }
 EXPORT_SYMBOL_GPL(bsg_setup_queue);
diff --git a/block/bsg.c b/block/bsg.c
index a57046d..e34c332 100644
--- a/block/bsg.c
+++ b/block/bsg.c
@@ -85,7 +85,6 @@ struct bsg_command {
 	struct bio *bidi_bio;
 	int err;
 	struct sg_io_v4 hdr;
-	char sense[SCSI_SENSE_BUFFERSIZE];
 };
 
 static void bsg_free_command(struct bsg_command *bc)
@@ -140,18 +139,20 @@ static int blk_fill_sgv4_hdr_rq(struct request_queue *q, struct request *rq,
 				struct sg_io_v4 *hdr, struct bsg_device *bd,
 				fmode_t has_write_perm)
 {
+	struct scsi_request *req = scsi_req(rq);
+
 	if (hdr->request_len > BLK_MAX_CDB) {
-		rq->cmd = kzalloc(hdr->request_len, GFP_KERNEL);
-		if (!rq->cmd)
+		req->cmd = kzalloc(hdr->request_len, GFP_KERNEL);
+		if (!req->cmd)
 			return -ENOMEM;
 	}
 
-	if (copy_from_user(rq->cmd, (void __user *)(unsigned long)hdr->request,
+	if (copy_from_user(req->cmd, (void __user *)(unsigned long)hdr->request,
 			   hdr->request_len))
 		return -EFAULT;
 
 	if (hdr->subprotocol == BSG_SUB_PROTOCOL_SCSI_CMD) {
-		if (blk_verify_command(rq->cmd, has_write_perm))
+		if (blk_verify_command(req->cmd, has_write_perm))
 			return -EPERM;
 	} else if (!capable(CAP_SYS_RAWIO))
 		return -EPERM;
@@ -159,7 +160,7 @@ static int blk_fill_sgv4_hdr_rq(struct request_queue *q, struct request *rq,
 	/*
 	 * fill in request structure
 	 */
-	rq->cmd_len = hdr->request_len;
+	req->cmd_len = hdr->request_len;
 
 	rq->timeout = msecs_to_jiffies(hdr->timeout);
 	if (!rq->timeout)
@@ -205,8 +206,7 @@ bsg_validate_sgv4_hdr(struct sg_io_v4 *hdr, int *rw)
  * map sg_io_v4 to a request.
  */
 static struct request *
-bsg_map_hdr(struct bsg_device *bd, struct sg_io_v4 *hdr, fmode_t has_write_perm,
-	    u8 *sense)
+bsg_map_hdr(struct bsg_device *bd, struct sg_io_v4 *hdr, fmode_t has_write_perm)
 {
 	struct request_queue *q = bd->queue;
 	struct request *rq, *next_rq = NULL;
@@ -236,7 +236,7 @@ bsg_map_hdr(struct bsg_device *bd, struct sg_io_v4 *hdr, fmode_t has_write_perm,
 	rq = blk_get_request(q, rw, GFP_KERNEL);
 	if (IS_ERR(rq))
 		return rq;
-	blk_rq_set_block_pc(rq);
+	scsi_req_init(rq);
 
 	ret = blk_fill_sgv4_hdr_rq(q, rq, hdr, bd, has_write_perm);
 	if (ret)
@@ -280,13 +280,9 @@ bsg_map_hdr(struct bsg_device *bd, struct sg_io_v4 *hdr, fmode_t has_write_perm,
 			goto out;
 	}
 
-	rq->sense = sense;
-	rq->sense_len = 0;
-
 	return rq;
 out:
-	if (rq->cmd != rq->__cmd)
-		kfree(rq->cmd);
+	scsi_req_free_cmd(scsi_req(rq));
 	blk_put_request(rq);
 	if (next_rq) {
 		blk_rq_unmap_user(next_rq->bio);
@@ -393,6 +389,7 @@ static struct bsg_command *bsg_get_done_cmd(struct bsg_device *bd)
 static int blk_complete_sgv4_hdr_rq(struct request *rq, struct sg_io_v4 *hdr,
 				    struct bio *bio, struct bio *bidi_bio)
 {
+	struct scsi_request *req = scsi_req(rq);
 	int ret = 0;
 
 	dprintk("rq %p bio %p 0x%x\n", rq, bio, rq->errors);
@@ -407,12 +404,12 @@ static int blk_complete_sgv4_hdr_rq(struct request *rq, struct sg_io_v4 *hdr,
 		hdr->info |= SG_INFO_CHECK;
 	hdr->response_len = 0;
 
-	if (rq->sense_len && hdr->response) {
+	if (req->sense_len && hdr->response) {
 		int len = min_t(unsigned int, hdr->max_response_len,
-					rq->sense_len);
+					req->sense_len);
 
 		ret = copy_to_user((void __user *)(unsigned long)hdr->response,
-				   rq->sense, len);
+				   req->sense, len);
 		if (!ret)
 			hdr->response_len = len;
 		else
@@ -420,14 +417,14 @@ static int blk_complete_sgv4_hdr_rq(struct request *rq, struct sg_io_v4 *hdr,
 	}
 
 	if (rq->next_rq) {
-		hdr->dout_resid = rq->resid_len;
-		hdr->din_resid = rq->next_rq->resid_len;
+		hdr->dout_resid = req->resid_len;
+		hdr->din_resid = scsi_req(rq->next_rq)->resid_len;
 		blk_rq_unmap_user(bidi_bio);
 		blk_put_request(rq->next_rq);
 	} else if (rq_data_dir(rq) == READ)
-		hdr->din_resid = rq->resid_len;
+		hdr->din_resid = req->resid_len;
 	else
-		hdr->dout_resid = rq->resid_len;
+		hdr->dout_resid = req->resid_len;
 
 	/*
 	 * If the request generated a negative error number, return it
@@ -439,8 +436,7 @@ static int blk_complete_sgv4_hdr_rq(struct request *rq, struct sg_io_v4 *hdr,
 		ret = rq->errors;
 
 	blk_rq_unmap_user(bio);
-	if (rq->cmd != rq->__cmd)
-		kfree(rq->cmd);
+	scsi_req_free_cmd(req);
 	blk_put_request(rq);
 
 	return ret;
@@ -625,7 +621,7 @@ static int __bsg_write(struct bsg_device *bd, const char __user *buf,
 		/*
 		 * get a request, fill in the blanks, and add to request queue
 		 */
-		rq = bsg_map_hdr(bd, &bc->hdr, has_write_perm, bc->sense);
+		rq = bsg_map_hdr(bd, &bc->hdr, has_write_perm);
 		if (IS_ERR(rq)) {
 			ret = PTR_ERR(rq);
 			rq = NULL;
@@ -911,12 +907,11 @@ static long bsg_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
 		struct bio *bio, *bidi_bio = NULL;
 		struct sg_io_v4 hdr;
 		int at_head;
-		u8 sense[SCSI_SENSE_BUFFERSIZE];
 
 		if (copy_from_user(&hdr, uarg, sizeof(hdr)))
 			return -EFAULT;
 
-		rq = bsg_map_hdr(bd, &hdr, file->f_mode & FMODE_WRITE, sense);
+		rq = bsg_map_hdr(bd, &hdr, file->f_mode & FMODE_WRITE);
 		if (IS_ERR(rq))
 			return PTR_ERR(rq);
 
diff --git a/block/scsi_ioctl.c b/block/scsi_ioctl.c
index c2b6492..e542144 100644
--- a/block/scsi_ioctl.c
+++ b/block/scsi_ioctl.c
@@ -230,15 +230,17 @@ EXPORT_SYMBOL(blk_verify_command);
 static int blk_fill_sghdr_rq(struct request_queue *q, struct request *rq,
 			     struct sg_io_hdr *hdr, fmode_t mode)
 {
-	if (copy_from_user(rq->cmd, hdr->cmdp, hdr->cmd_len))
+	struct scsi_request *req = scsi_req(rq);
+
+	if (copy_from_user(req->cmd, hdr->cmdp, hdr->cmd_len))
 		return -EFAULT;
-	if (blk_verify_command(rq->cmd, mode & FMODE_WRITE))
+	if (blk_verify_command(req->cmd, mode & FMODE_WRITE))
 		return -EPERM;
 
 	/*
 	 * fill in request structure
 	 */
-	rq->cmd_len = hdr->cmd_len;
+	req->cmd_len = hdr->cmd_len;
 
 	rq->timeout = msecs_to_jiffies(hdr->timeout);
 	if (!rq->timeout)
@@ -254,6 +256,7 @@ static int blk_fill_sghdr_rq(struct request_queue *q, struct request *rq,
 static int blk_complete_sghdr_rq(struct request *rq, struct sg_io_hdr *hdr,
 				 struct bio *bio)
 {
+	struct scsi_request *req = scsi_req(rq);
 	int r, ret = 0;
 
 	/*
@@ -267,13 +270,13 @@ static int blk_complete_sghdr_rq(struct request *rq, struct sg_io_hdr *hdr,
 	hdr->info = 0;
 	if (hdr->masked_status || hdr->host_status || hdr->driver_status)
 		hdr->info |= SG_INFO_CHECK;
-	hdr->resid = rq->resid_len;
+	hdr->resid = req->resid_len;
 	hdr->sb_len_wr = 0;
 
-	if (rq->sense_len && hdr->sbp) {
-		int len = min((unsigned int) hdr->mx_sb_len, rq->sense_len);
+	if (req->sense_len && hdr->sbp) {
+		int len = min((unsigned int) hdr->mx_sb_len, req->sense_len);
 
-		if (!copy_to_user(hdr->sbp, rq->sense, len))
+		if (!copy_to_user(hdr->sbp, req->sense, len))
 			hdr->sb_len_wr = len;
 		else
 			ret = -EFAULT;
@@ -294,7 +297,7 @@ static int sg_io(struct request_queue *q, struct gendisk *bd_disk,
 	int writing = 0;
 	int at_head = 0;
 	struct request *rq;
-	char sense[SCSI_SENSE_BUFFERSIZE];
+	struct scsi_request *req;
 	struct bio *bio;
 
 	if (hdr->interface_id != 'S')
@@ -321,11 +324,12 @@ static int sg_io(struct request_queue *q, struct gendisk *bd_disk,
 	rq = blk_get_request(q, writing ? WRITE : READ, GFP_KERNEL);
 	if (IS_ERR(rq))
 		return PTR_ERR(rq);
-	blk_rq_set_block_pc(rq);
+	req = scsi_req(rq);
+	scsi_req_init(rq);
 
 	if (hdr->cmd_len > BLK_MAX_CDB) {
-		rq->cmd = kzalloc(hdr->cmd_len, GFP_KERNEL);
-		if (!rq->cmd)
+		req->cmd = kzalloc(hdr->cmd_len, GFP_KERNEL);
+		if (!req->cmd)
 			goto out_put_request;
 	}
 
@@ -357,9 +361,6 @@ static int sg_io(struct request_queue *q, struct gendisk *bd_disk,
 		goto out_free_cdb;
 
 	bio = rq->bio;
-	memset(sense, 0, sizeof(sense));
-	rq->sense = sense;
-	rq->sense_len = 0;
 	rq->retries = 0;
 
 	start_time = jiffies;
@@ -375,8 +376,7 @@ static int sg_io(struct request_queue *q, struct gendisk *bd_disk,
 	ret = blk_complete_sghdr_rq(rq, hdr, bio);
 
 out_free_cdb:
-	if (rq->cmd != rq->__cmd)
-		kfree(rq->cmd);
+	scsi_req_free_cmd(req);
 out_put_request:
 	blk_put_request(rq);
 	return ret;
@@ -420,9 +420,10 @@ int sg_scsi_ioctl(struct request_queue *q, struct gendisk *disk, fmode_t mode,
 		struct scsi_ioctl_command __user *sic)
 {
 	struct request *rq;
+	struct scsi_request *req;
 	int err;
 	unsigned int in_len, out_len, bytes, opcode, cmdlen;
-	char *buffer = NULL, sense[SCSI_SENSE_BUFFERSIZE];
+	char *buffer = NULL;
 
 	if (!sic)
 		return -EINVAL;
@@ -452,7 +453,8 @@ int sg_scsi_ioctl(struct request_queue *q, struct gendisk *disk, fmode_t mode,
 		err = PTR_ERR(rq);
 		goto error_free_buffer;
 	}
-	blk_rq_set_block_pc(rq);
+	req = scsi_req(rq);
+	scsi_req_init(rq);
 
 	cmdlen = COMMAND_SIZE(opcode);
 
@@ -460,14 +462,14 @@ int sg_scsi_ioctl(struct request_queue *q, struct gendisk *disk, fmode_t mode,
 	 * get command and data to send to device, if any
 	 */
 	err = -EFAULT;
-	rq->cmd_len = cmdlen;
-	if (copy_from_user(rq->cmd, sic->data, cmdlen))
+	req->cmd_len = cmdlen;
+	if (copy_from_user(req->cmd, sic->data, cmdlen))
 		goto error;
 
 	if (in_len && copy_from_user(buffer, sic->data + cmdlen, in_len))
 		goto error;
 
-	err = blk_verify_command(rq->cmd, mode & FMODE_WRITE);
+	err = blk_verify_command(req->cmd, mode & FMODE_WRITE);
 	if (err)
 		goto error;
 
@@ -503,18 +505,14 @@ int sg_scsi_ioctl(struct request_queue *q, struct gendisk *disk, fmode_t mode,
 		goto error;
 	}
 
-	memset(sense, 0, sizeof(sense));
-	rq->sense = sense;
-	rq->sense_len = 0;
-
 	blk_execute_rq(q, disk, rq, 0);
 
 	err = rq->errors & 0xff;	/* only 8 bit SCSI status */
 	if (err) {
-		if (rq->sense_len && rq->sense) {
-			bytes = (OMAX_SB_LEN > rq->sense_len) ?
-				rq->sense_len : OMAX_SB_LEN;
-			if (copy_to_user(sic->data, rq->sense, bytes))
+		if (req->sense_len && req->sense) {
+			bytes = (OMAX_SB_LEN > req->sense_len) ?
+				req->sense_len : OMAX_SB_LEN;
+			if (copy_to_user(sic->data, req->sense, bytes))
 				err = -EFAULT;
 		}
 	} else {
@@ -542,11 +540,11 @@ static int __blk_send_generic(struct request_queue *q, struct gendisk *bd_disk,
 	rq = blk_get_request(q, WRITE, __GFP_RECLAIM);
 	if (IS_ERR(rq))
 		return PTR_ERR(rq);
-	blk_rq_set_block_pc(rq);
+	scsi_req_init(rq);
 	rq->timeout = BLK_DEFAULT_SG_TIMEOUT;
-	rq->cmd[0] = cmd;
-	rq->cmd[4] = data;
-	rq->cmd_len = 6;
+	scsi_req(rq)->cmd[0] = cmd;
+	scsi_req(rq)->cmd[4] = data;
+	scsi_req(rq)->cmd_len = 6;
 	err = blk_execute_rq(q, bd_disk, rq, 0);
 	blk_put_request(rq);
 
@@ -743,6 +741,29 @@ int scsi_cmd_blk_ioctl(struct block_device *bd, fmode_t mode,
 }
 EXPORT_SYMBOL(scsi_cmd_blk_ioctl);
 
+/**
+ * scsi_req_init - initialize a request to type BLOCK_PC
+ * @rq:		request to be initialized
+ *
+ */
+void scsi_req_init(struct request *rq)
+{
+	struct scsi_request *req = scsi_req(rq);
+
+	rq->cmd_type = REQ_TYPE_BLOCK_PC;
+	memset(req->__cmd, 0, sizeof(req->__cmd));
+	req->cmd = req->__cmd;
+	req->cmd_len = BLK_MAX_CDB;
+	req->sense_len = 0;
+}
+EXPORT_SYMBOL(scsi_req_init);
+
+int scsi_cmd_buf_len(struct request *rq)
+{
+	return scsi_req(rq)->cmd_len * 3;
+}
+EXPORT_SYMBOL(scsi_cmd_buf_len);
+
 static int __init blk_scsi_ioctl_init(void)
 {
 	blk_set_cmd_filter_defaults(&blk_default_cmd_filter);
diff --git a/drivers/ata/libata-scsi.c b/drivers/ata/libata-scsi.c
index 1f863e7..6abd739 100644
--- a/drivers/ata/libata-scsi.c
+++ b/drivers/ata/libata-scsi.c
@@ -1271,7 +1271,7 @@ static int atapi_drain_needed(struct request *rq)
 	if (!blk_rq_bytes(rq) || op_is_write(req_op(rq)))
 		return 0;
 
-	return atapi_cmd_type(rq->cmd[0]) == ATAPI_MISC;
+	return atapi_cmd_type(scsi_req(rq)->cmd[0]) == ATAPI_MISC;
 }
 
 static int ata_scsi_dev_config(struct scsi_device *sdev,
diff --git a/drivers/block/cciss.c b/drivers/block/cciss.c
index e5c5b8e..b93bb73 100644
--- a/drivers/block/cciss.c
+++ b/drivers/block/cciss.c
@@ -52,6 +52,7 @@
 #include <scsi/scsi.h>
 #include <scsi/sg.h>
 #include <scsi/scsi_ioctl.h>
+#include <scsi/scsi_request.h>
 #include <linux/cdrom.h>
 #include <linux/scatterlist.h>
 #include <linux/kthread.h>
@@ -1854,7 +1855,7 @@ static void cciss_softirq_done(struct request *rq)
 
 	/* set the residual count for pc requests */
 	if (rq->cmd_type == REQ_TYPE_BLOCK_PC)
-		rq->resid_len = c->err_info->ResidualCnt;
+		scsi_req(rq)->resid_len = c->err_info->ResidualCnt;
 
 	blk_end_request_all(rq, (rq->errors == 0) ? 0 : -EIO);
 
@@ -1941,9 +1942,16 @@ static void cciss_get_serial_no(ctlr_info_t *h, int logvol,
 static int cciss_add_disk(ctlr_info_t *h, struct gendisk *disk,
 				int drv_index)
 {
-	disk->queue = blk_init_queue(do_cciss_request, &h->lock);
+	disk->queue = blk_alloc_queue(GFP_KERNEL);
 	if (!disk->queue)
 		goto init_queue_failure;
+
+	disk->queue->cmd_size = sizeof(struct scsi_request);
+	disk->queue->request_fn = do_cciss_request;
+	disk->queue->queue_lock = &h->lock;
+	if (blk_init_allocated_queue(disk->queue) < 0)
+		goto cleanup_queue;
+
 	sprintf(disk->disk_name, "cciss/c%dd%d", h->ctlr, drv_index);
 	disk->major = h->major;
 	disk->first_minor = drv_index << NWD_SHIFT;
@@ -3111,15 +3119,7 @@ static inline int evaluate_target_status(ctlr_info_t *h,
 		return error_value;
 	}
 
-	/* SG_IO or similar, copy sense data back */
-	if (cmd->rq->sense) {
-		if (cmd->rq->sense_len > cmd->err_info->SenseLen)
-			cmd->rq->sense_len = cmd->err_info->SenseLen;
-		memcpy(cmd->rq->sense, cmd->err_info->SenseInfo,
-			cmd->rq->sense_len);
-	} else
-		cmd->rq->sense_len = 0;
-
+	scsi_req(cmd->rq)->sense_len = cmd->err_info->SenseLen;
 	return error_value;
 }
 
@@ -3150,7 +3150,6 @@ static inline void complete_command(ctlr_info_t *h, CommandList_struct *cmd,
 			dev_warn(&h->pdev->dev, "cmd %p has"
 			       " completed with data underrun "
 			       "reported\n", cmd);
-			cmd->rq->resid_len = cmd->err_info->ResidualCnt;
 		}
 		break;
 	case CMD_DATA_OVERRUN:
@@ -3426,8 +3425,9 @@ static void do_cciss_request(struct request_queue *q)
 			c->Request.CDB[14] = c->Request.CDB[15] = 0;
 		}
 	} else if (creq->cmd_type == REQ_TYPE_BLOCK_PC) {
-		c->Request.CDBLen = creq->cmd_len;
-		memcpy(c->Request.CDB, creq->cmd, BLK_MAX_CDB);
+		c->Request.CDBLen = scsi_req(creq)->cmd_len;
+		memcpy(c->Request.CDB, scsi_req(creq)->cmd, BLK_MAX_CDB);
+		scsi_req(creq)->sense = c->err_info->SenseInfo;
 	} else {
 		dev_warn(&h->pdev->dev, "bad request type %d\n",
 			creq->cmd_type);
diff --git a/drivers/block/pktcdvd.c b/drivers/block/pktcdvd.c
index 1b94c1c..918a92c 100644
--- a/drivers/block/pktcdvd.c
+++ b/drivers/block/pktcdvd.c
@@ -707,7 +707,7 @@ static int pkt_generic_packet(struct pktcdvd_device *pd, struct packet_command *
 			     WRITE : READ, __GFP_RECLAIM);
 	if (IS_ERR(rq))
 		return PTR_ERR(rq);
-	blk_rq_set_block_pc(rq);
+	scsi_req_init(rq);
 
 	if (cgc->buflen) {
 		ret = blk_rq_map_kern(q, rq, cgc->buffer, cgc->buflen,
@@ -716,8 +716,8 @@ static int pkt_generic_packet(struct pktcdvd_device *pd, struct packet_command *
 			goto out;
 	}
 
-	rq->cmd_len = COMMAND_SIZE(cgc->cmd[0]);
-	memcpy(rq->cmd, cgc->cmd, CDROM_PACKET_SIZE);
+	scsi_req(rq)->cmd_len = COMMAND_SIZE(cgc->cmd[0]);
+	memcpy(scsi_req(rq)->cmd, cgc->cmd, CDROM_PACKET_SIZE);
 
 	rq->timeout = 60*HZ;
 	if (cgc->quiet)
diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c
index 3c3b8f6..938ec4f 100644
--- a/drivers/block/virtio_blk.c
+++ b/drivers/block/virtio_blk.c
@@ -52,6 +52,7 @@ struct virtio_blk {
 };
 
 struct virtblk_req {
+	struct scsi_request sreq;	/* for SCSI passthrough */
 	struct request *req;
 	struct virtio_blk_outhdr out_hdr;
 	struct virtio_scsi_inhdr in_hdr;
@@ -91,7 +92,7 @@ static int __virtblk_add_req(struct virtqueue *vq,
 	 * inhdr with additional status information.
 	 */
 	if (type == cpu_to_virtio32(vq->vdev, VIRTIO_BLK_T_SCSI_CMD)) {
-		sg_init_one(&cmd, vbr->req->cmd, vbr->req->cmd_len);
+		sg_init_one(&cmd, vbr->sreq.cmd, vbr->sreq.cmd_len);
 		sgs[num_out++] = &cmd;
 	}
 
@@ -103,7 +104,6 @@ static int __virtblk_add_req(struct virtqueue *vq,
 	}
 
 	if (type == cpu_to_virtio32(vq->vdev, VIRTIO_BLK_T_SCSI_CMD)) {
-		memcpy(vbr->sense, vbr->req->sense, SCSI_SENSE_BUFFERSIZE);
 		sg_init_one(&sense, vbr->sense, SCSI_SENSE_BUFFERSIZE);
 		sgs[num_out + num_in++] = &sense;
 		sg_init_one(&inhdr, &vbr->in_hdr, sizeof(vbr->in_hdr));
@@ -123,8 +123,10 @@ static inline void virtblk_request_done(struct request *req)
 	int error = virtblk_result(vbr);
 
 	if (req->cmd_type == REQ_TYPE_BLOCK_PC) {
-		req->resid_len = virtio32_to_cpu(vblk->vdev, vbr->in_hdr.residual);
-		req->sense_len = virtio32_to_cpu(vblk->vdev, vbr->in_hdr.sense_len);
+		scsi_req(req)->resid_len =
+			virtio32_to_cpu(vblk->vdev, vbr->in_hdr.residual);
+		vbr->sreq.sense_len =
+			virtio32_to_cpu(vblk->vdev, vbr->in_hdr.sense_len);
 		req->errors = virtio32_to_cpu(vblk->vdev, vbr->in_hdr.errors);
 	} else if (req->cmd_type == REQ_TYPE_DRV_PRIV) {
 		req->errors = (error != 0);
@@ -538,6 +540,7 @@ static int virtblk_init_request(void *data, struct request *rq,
 	struct virtio_blk *vblk = data;
 	struct virtblk_req *vbr = blk_mq_rq_to_pdu(rq);
 
+	vbr->sreq.sense = vbr->sense;
 	sg_init_table(vbr->sg, vblk->sg_elems);
 	return 0;
 }
diff --git a/drivers/cdrom/cdrom.c b/drivers/cdrom/cdrom.c
index 59cca72..36f5237 100644
--- a/drivers/cdrom/cdrom.c
+++ b/drivers/cdrom/cdrom.c
@@ -281,8 +281,8 @@
 #include <linux/fcntl.h>
 #include <linux/blkdev.h>
 #include <linux/times.h>
-
 #include <linux/uaccess.h>
+#include <scsi/scsi_request.h>
 
 /* used to tell the module to turn on full debugging messages */
 static bool debug;
@@ -2172,6 +2172,7 @@ static int cdrom_read_cdda_bpc(struct cdrom_device_info *cdi, __u8 __user *ubuf,
 {
 	struct request_queue *q = cdi->disk->queue;
 	struct request *rq;
+	struct scsi_request *req;
 	struct bio *bio;
 	unsigned int len;
 	int nr, ret = 0;
@@ -2195,7 +2196,8 @@ static int cdrom_read_cdda_bpc(struct cdrom_device_info *cdi, __u8 __user *ubuf,
 			ret = PTR_ERR(rq);
 			break;
 		}
-		blk_rq_set_block_pc(rq);
+		req = scsi_req(rq);
+		scsi_req_init(rq);
 
 		ret = blk_rq_map_user(q, rq, NULL, ubuf, len, GFP_KERNEL);
 		if (ret) {
@@ -2203,23 +2205,23 @@ static int cdrom_read_cdda_bpc(struct cdrom_device_info *cdi, __u8 __user *ubuf,
 			break;
 		}
 
-		rq->cmd[0] = GPCMD_READ_CD;
-		rq->cmd[1] = 1 << 2;
-		rq->cmd[2] = (lba >> 24) & 0xff;
-		rq->cmd[3] = (lba >> 16) & 0xff;
-		rq->cmd[4] = (lba >>  8) & 0xff;
-		rq->cmd[5] = lba & 0xff;
-		rq->cmd[6] = (nr >> 16) & 0xff;
-		rq->cmd[7] = (nr >>  8) & 0xff;
-		rq->cmd[8] = nr & 0xff;
-		rq->cmd[9] = 0xf8;
-
-		rq->cmd_len = 12;
+		req->cmd[0] = GPCMD_READ_CD;
+		req->cmd[1] = 1 << 2;
+		req->cmd[2] = (lba >> 24) & 0xff;
+		req->cmd[3] = (lba >> 16) & 0xff;
+		req->cmd[4] = (lba >>  8) & 0xff;
+		req->cmd[5] = lba & 0xff;
+		req->cmd[6] = (nr >> 16) & 0xff;
+		req->cmd[7] = (nr >>  8) & 0xff;
+		req->cmd[8] = nr & 0xff;
+		req->cmd[9] = 0xf8;
+
+		req->cmd_len = 12;
 		rq->timeout = 60 * HZ;
 		bio = rq->bio;
 
 		if (blk_execute_rq(q, cdi->disk, rq, 0)) {
-			struct request_sense *s = rq->sense;
+			struct request_sense *s = req->sense;
 			ret = -EIO;
 			cdi->last_sense = s->sense_key;
 		}
diff --git a/drivers/ide/ide-atapi.c b/drivers/ide/ide-atapi.c
index f90ea22..7c826ec 100644
--- a/drivers/ide/ide-atapi.c
+++ b/drivers/ide/ide-atapi.c
@@ -93,6 +93,7 @@ int ide_queue_pc_tail(ide_drive_t *drive, struct gendisk *disk,
 	int error;
 
 	rq = blk_get_request(drive->queue, READ, __GFP_RECLAIM);
+	scsi_req_init(rq);
 	rq->cmd_type = REQ_TYPE_DRV_PRIV;
 	rq->special = (char *)pc;
 
@@ -103,9 +104,9 @@ int ide_queue_pc_tail(ide_drive_t *drive, struct gendisk *disk,
 			goto put_req;
 	}
 
-	memcpy(rq->cmd, pc->c, 12);
+	memcpy(scsi_req(rq)->cmd, pc->c, 12);
 	if (drive->media == ide_tape)
-		rq->cmd[13] = REQ_IDETAPE_PC1;
+		scsi_req(rq)->cmd[13] = REQ_IDETAPE_PC1;
 	error = blk_execute_rq(drive->queue, disk, rq, 0);
 put_req:
 	blk_put_request(rq);
@@ -171,7 +172,8 @@ EXPORT_SYMBOL_GPL(ide_create_request_sense_cmd);
 void ide_prep_sense(ide_drive_t *drive, struct request *rq)
 {
 	struct request_sense *sense = &drive->sense_data;
-	struct request *sense_rq = &drive->sense_rq;
+	struct request *sense_rq = drive->sense_rq;
+	struct scsi_request *req = scsi_req(sense_rq);
 	unsigned int cmd_len, sense_len;
 	int err;
 
@@ -197,6 +199,7 @@ void ide_prep_sense(ide_drive_t *drive, struct request *rq)
 	memset(sense, 0, sizeof(*sense));
 
 	blk_rq_init(rq->q, sense_rq);
+	scsi_req_init(sense_rq);
 
 	err = blk_rq_map_kern(drive->queue, sense_rq, sense, sense_len,
 			      GFP_NOIO);
@@ -208,13 +211,13 @@ void ide_prep_sense(ide_drive_t *drive, struct request *rq)
 	}
 
 	sense_rq->rq_disk = rq->rq_disk;
-	sense_rq->cmd[0] = GPCMD_REQUEST_SENSE;
-	sense_rq->cmd[4] = cmd_len;
 	sense_rq->cmd_type = REQ_TYPE_ATA_SENSE;
 	sense_rq->rq_flags |= RQF_PREEMPT;
 
+	req->cmd[0] = GPCMD_REQUEST_SENSE;
+	req->cmd[4] = cmd_len;
 	if (drive->media == ide_tape)
-		sense_rq->cmd[13] = REQ_IDETAPE_PC1;
+		req->cmd[13] = REQ_IDETAPE_PC1;
 
 	drive->sense_rq_armed = true;
 }
@@ -229,12 +232,12 @@ int ide_queue_sense_rq(ide_drive_t *drive, void *special)
 		return -ENOMEM;
 	}
 
-	drive->sense_rq.special = special;
+	drive->sense_rq->special = special;
 	drive->sense_rq_armed = false;
 
 	drive->hwif->rq = NULL;
 
-	elv_add_request(drive->queue, &drive->sense_rq, ELEVATOR_INSERT_FRONT);
+	elv_add_request(drive->queue, drive->sense_rq, ELEVATOR_INSERT_FRONT);
 	return 0;
 }
 EXPORT_SYMBOL_GPL(ide_queue_sense_rq);
@@ -247,14 +250,14 @@ EXPORT_SYMBOL_GPL(ide_queue_sense_rq);
 void ide_retry_pc(ide_drive_t *drive)
 {
 	struct request *failed_rq = drive->hwif->rq;
-	struct request *sense_rq = &drive->sense_rq;
+	struct request *sense_rq = drive->sense_rq;
 	struct ide_atapi_pc *pc = &drive->request_sense_pc;
 
 	(void)ide_read_error(drive);
 
 	/* init pc from sense_rq */
 	ide_init_pc(pc);
-	memcpy(pc->c, sense_rq->cmd, 12);
+	memcpy(pc->c, scsi_req(sense_rq)->cmd, 12);
 
 	if (drive->media == ide_tape)
 		drive->atapi_flags |= IDE_AFLAG_IGNORE_DSC;
@@ -286,7 +289,7 @@ int ide_cd_expiry(ide_drive_t *drive)
 	 * commands/drives support that. Let ide_timer_expiry keep polling us
 	 * for these.
 	 */
-	switch (rq->cmd[0]) {
+	switch (scsi_req(rq)->cmd[0]) {
 	case GPCMD_BLANK:
 	case GPCMD_FORMAT_UNIT:
 	case GPCMD_RESERVE_RZONE_TRACK:
@@ -297,7 +300,7 @@ int ide_cd_expiry(ide_drive_t *drive)
 	default:
 		if (!(rq->rq_flags & RQF_QUIET))
 			printk(KERN_INFO PFX "cmd 0x%x timed out\n",
-					 rq->cmd[0]);
+					 scsi_req(rq)->cmd[0]);
 		wait = 0;
 		break;
 	}
@@ -420,7 +423,7 @@ static ide_startstop_t ide_pc_intr(ide_drive_t *drive)
 						     ? "write" : "read");
 			pc->flags |= PC_FLAG_DMA_ERROR;
 		} else
-			rq->resid_len = 0;
+			scsi_req(rq)->resid_len = 0;
 		debug_log("%s: DMA finished\n", drive->name);
 	}
 
@@ -436,7 +439,7 @@ static ide_startstop_t ide_pc_intr(ide_drive_t *drive)
 		local_irq_enable_in_hardirq();
 
 		if (drive->media == ide_tape &&
-		    (stat & ATA_ERR) && rq->cmd[0] == REQUEST_SENSE)
+		    (stat & ATA_ERR) && scsi_req(rq)->cmd[0] == REQUEST_SENSE)
 			stat &= ~ATA_ERR;
 
 		if ((stat & ATA_ERR) || (pc->flags & PC_FLAG_DMA_ERROR)) {
@@ -446,7 +449,7 @@ static ide_startstop_t ide_pc_intr(ide_drive_t *drive)
 			if (drive->media != ide_tape)
 				pc->rq->errors++;
 
-			if (rq->cmd[0] == REQUEST_SENSE) {
+			if (scsi_req(rq)->cmd[0] == REQUEST_SENSE) {
 				printk(KERN_ERR PFX "%s: I/O error in request "
 						"sense command\n", drive->name);
 				return ide_do_reset(drive);
@@ -512,7 +515,7 @@ static ide_startstop_t ide_pc_intr(ide_drive_t *drive)
 	ide_pio_bytes(drive, cmd, write, done);
 
 	/* Update transferred byte count */
-	rq->resid_len -= done;
+	scsi_req(rq)->resid_len -= done;
 
 	bcount -= done;
 
@@ -520,7 +523,7 @@ static ide_startstop_t ide_pc_intr(ide_drive_t *drive)
 		ide_pad_transfer(drive, write, bcount);
 
 	debug_log("[cmd %x] transferred %d bytes, padded %d bytes, resid: %u\n",
-		  rq->cmd[0], done, bcount, rq->resid_len);
+		  rq->cmd[0], done, bcount, scsi_req(rq)->resid_len);
 
 	/* And set the interrupt handler again */
 	ide_set_handler(drive, ide_pc_intr, timeout);
@@ -603,7 +606,7 @@ static ide_startstop_t ide_transfer_pc(ide_drive_t *drive)
 
 	if (dev_is_idecd(drive)) {
 		/* ATAPI commands get padded out to 12 bytes minimum */
-		cmd_len = COMMAND_SIZE(rq->cmd[0]);
+		cmd_len = COMMAND_SIZE(scsi_req(rq)->cmd[0]);
 		if (cmd_len < ATAPI_MIN_CDB_BYTES)
 			cmd_len = ATAPI_MIN_CDB_BYTES;
 
@@ -650,7 +653,7 @@ static ide_startstop_t ide_transfer_pc(ide_drive_t *drive)
 
 	/* Send the actual packet */
 	if ((drive->atapi_flags & IDE_AFLAG_ZIP_DRIVE) == 0)
-		hwif->tp_ops->output_data(drive, NULL, rq->cmd, cmd_len);
+		hwif->tp_ops->output_data(drive, NULL, scsi_req(rq)->cmd, cmd_len);
 
 	/* Begin DMA, if necessary */
 	if (dev_is_idecd(drive)) {
@@ -695,7 +698,7 @@ ide_startstop_t ide_issue_pc(ide_drive_t *drive, struct ide_cmd *cmd)
 							     bytes, 63 * 1024));
 
 		/* We haven't transferred any data yet */
-		rq->resid_len = bcount;
+		scsi_req(rq)->resid_len = bcount;
 
 		if (pc->flags & PC_FLAG_DMA_ERROR) {
 			pc->flags &= ~PC_FLAG_DMA_ERROR;
diff --git a/drivers/ide/ide-cd.c b/drivers/ide/ide-cd.c
index 9cbd217..6eb9872 100644
--- a/drivers/ide/ide-cd.c
+++ b/drivers/ide/ide-cd.c
@@ -121,7 +121,7 @@ static int cdrom_log_sense(ide_drive_t *drive, struct request *rq)
 		 * don't log START_STOP unit with LoEj set, since we cannot
 		 * reliably check if drive can auto-close
 		 */
-		if (rq->cmd[0] == GPCMD_START_STOP_UNIT && sense->asc == 0x24)
+		if (scsi_req(rq)->cmd[0] == GPCMD_START_STOP_UNIT && sense->asc == 0x24)
 			break;
 		log = 1;
 		break;
@@ -163,7 +163,7 @@ static void cdrom_analyze_sense_data(ide_drive_t *drive,
 	 * toc has not been recorded yet, it will fail with 05/24/00 (which is a
 	 * confusing error)
 	 */
-	if (failed_command && failed_command->cmd[0] == GPCMD_READ_TOC_PMA_ATIP)
+	if (failed_command && scsi_req(failed_command)->cmd[0] == GPCMD_READ_TOC_PMA_ATIP)
 		if (sense->sense_key == 0x05 && sense->asc == 0x24)
 			return;
 
@@ -219,15 +219,12 @@ static void ide_cd_complete_failed_rq(ide_drive_t *drive, struct request *rq)
 	void *sense = bio_data(rq->bio);
 
 	if (failed) {
-		if (failed->sense) {
-			/*
-			 * Sense is always read into drive->sense_data.
-			 * Copy back if the failed request has its
-			 * sense pointer set.
-			 */
-			memcpy(failed->sense, sense, 18);
-			failed->sense_len = rq->sense_len;
-		}
+		/*
+		 * Sense is always read into drive->sense_data, copy back to the
+		 * original request.
+		 */
+		memcpy(scsi_req(failed)->sense, sense, 18);
+		scsi_req(failed)->sense_len = scsi_req(rq)->sense_len;
 		cdrom_analyze_sense_data(drive, failed);
 
 		if (ide_end_rq(drive, failed, -EIO, blk_rq_bytes(failed)))
@@ -338,7 +335,7 @@ static int cdrom_decode_status(ide_drive_t *drive, u8 stat)
 		 *
 		 * cdrom_log_sense() knows this!
 		 */
-		if (rq->cmd[0] == GPCMD_START_STOP_UNIT)
+		if (scsi_req(rq)->cmd[0] == GPCMD_START_STOP_UNIT)
 			break;
 		/* fall-through */
 	case DATA_PROTECT:
@@ -414,7 +411,7 @@ static void ide_cd_request_sense_fixup(ide_drive_t *drive, struct ide_cmd *cmd)
 	 * Some of the trailing request sense fields are optional,
 	 * and some drives don't send them.  Sigh.
 	 */
-	if (rq->cmd[0] == GPCMD_REQUEST_SENSE &&
+	if (scsi_req(rq)->cmd[0] == GPCMD_REQUEST_SENSE &&
 	    cmd->nleft > 0 && cmd->nleft <= 5)
 		cmd->nleft = 0;
 }
@@ -425,12 +422,8 @@ int ide_cd_queue_pc(ide_drive_t *drive, const unsigned char *cmd,
 		    req_flags_t rq_flags)
 {
 	struct cdrom_info *info = drive->driver_data;
-	struct request_sense local_sense;
 	int retries = 10;
-	req_flags_t flags = 0;
-
-	if (!sense)
-		sense = &local_sense;
+	bool failed;
 
 	ide_debug_log(IDE_DBG_PC, "cmd[0]: 0x%x, write: 0x%x, timeout: %d, "
 				  "rq_flags: 0x%x",
@@ -440,12 +433,12 @@ int ide_cd_queue_pc(ide_drive_t *drive, const unsigned char *cmd,
 	do {
 		struct request *rq;
 		int error;
+		bool delay = false;
 
 		rq = blk_get_request(drive->queue, write, __GFP_RECLAIM);
-
-		memcpy(rq->cmd, cmd, BLK_MAX_CDB);
+		scsi_req_init(rq);
+		memcpy(scsi_req(rq)->cmd, cmd, BLK_MAX_CDB);
 		rq->cmd_type = REQ_TYPE_ATA_PC;
-		rq->sense = sense;
 		rq->rq_flags |= rq_flags;
 		rq->timeout = timeout;
 		if (buffer) {
@@ -460,21 +453,21 @@ int ide_cd_queue_pc(ide_drive_t *drive, const unsigned char *cmd,
 		error = blk_execute_rq(drive->queue, info->disk, rq, 0);
 
 		if (buffer)
-			*bufflen = rq->resid_len;
-
-		flags = rq->rq_flags;
-		blk_put_request(rq);
+			*bufflen = scsi_req(rq)->resid_len;
+		if (sense)
+			memcpy(sense, scsi_req(rq)->sense, sizeof(*sense));
 
 		/*
 		 * FIXME: we should probably abort/retry or something in case of
 		 * failure.
 		 */
-		if (flags & RQF_FAILED) {
+		failed = (rq->rq_flags & RQF_FAILED) != 0;
+		if (failed) {
 			/*
 			 * The request failed.  Retry if it was due to a unit
 			 * attention status (usually means media was changed).
 			 */
-			struct request_sense *reqbuf = sense;
+			struct request_sense *reqbuf = scsi_req(rq)->sense;
 
 			if (reqbuf->sense_key == UNIT_ATTENTION)
 				cdrom_saw_media_change(drive);
@@ -485,19 +478,20 @@ int ide_cd_queue_pc(ide_drive_t *drive, const unsigned char *cmd,
 				 * a disk.  Retry, but wait a little to give
 				 * the drive time to complete the load.
 				 */
-				ssleep(2);
+				delay = true;
 			} else {
 				/* otherwise, don't retry */
 				retries = 0;
 			}
 			--retries;
 		}
-
-		/* end of retry loop */
-	} while ((flags & RQF_FAILED) && retries >= 0);
+		blk_put_request(rq);
+		if (delay)
+			ssleep(2);
+	} while (failed && retries >= 0);
 
 	/* return an error if the command failed */
-	return (flags & RQF_FAILED) ? -EIO : 0;
+	return failed ? -EIO : 0;
 }
 
 /*
@@ -636,7 +630,7 @@ static ide_startstop_t cdrom_newpc_intr(ide_drive_t *drive)
 		len -= blen;
 
 		if (sense && write == 0)
-			rq->sense_len += blen;
+			scsi_req(rq)->sense_len += blen;
 	}
 
 	/* pad, if necessary */
@@ -664,7 +658,7 @@ static ide_startstop_t cdrom_newpc_intr(ide_drive_t *drive)
 
 out_end:
 	if (rq->cmd_type == REQ_TYPE_BLOCK_PC && rc == 0) {
-		rq->resid_len = 0;
+		scsi_req(rq)->resid_len = 0;
 		blk_end_request_all(rq, 0);
 		hwif->rq = NULL;
 	} else {
@@ -685,9 +679,9 @@ static ide_startstop_t cdrom_newpc_intr(ide_drive_t *drive)
 
 		/* make sure it's fully ended */
 		if (rq->cmd_type != REQ_TYPE_FS) {
-			rq->resid_len -= cmd->nbytes - cmd->nleft;
+			scsi_req(rq)->resid_len -= cmd->nbytes - cmd->nleft;
 			if (uptodate == 0 && (cmd->tf_flags & IDE_TFLAG_WRITE))
-				rq->resid_len += cmd->last_xfer_len;
+				scsi_req(rq)->resid_len += cmd->last_xfer_len;
 		}
 
 		ide_complete_rq(drive, uptodate ? 0 : -EIO, blk_rq_bytes(rq));
@@ -1312,28 +1306,29 @@ static int ide_cdrom_prep_fs(struct request_queue *q, struct request *rq)
 	int hard_sect = queue_logical_block_size(q);
 	long block = (long)blk_rq_pos(rq) / (hard_sect >> 9);
 	unsigned long blocks = blk_rq_sectors(rq) / (hard_sect >> 9);
+	struct scsi_request *req = scsi_req(rq);
 
-	memset(rq->cmd, 0, BLK_MAX_CDB);
+	memset(req->cmd, 0, BLK_MAX_CDB);
 
 	if (rq_data_dir(rq) == READ)
-		rq->cmd[0] = GPCMD_READ_10;
+		req->cmd[0] = GPCMD_READ_10;
 	else
-		rq->cmd[0] = GPCMD_WRITE_10;
+		req->cmd[0] = GPCMD_WRITE_10;
 
 	/*
 	 * fill in lba
 	 */
-	rq->cmd[2] = (block >> 24) & 0xff;
-	rq->cmd[3] = (block >> 16) & 0xff;
-	rq->cmd[4] = (block >>  8) & 0xff;
-	rq->cmd[5] = block & 0xff;
+	req->cmd[2] = (block >> 24) & 0xff;
+	req->cmd[3] = (block >> 16) & 0xff;
+	req->cmd[4] = (block >>  8) & 0xff;
+	req->cmd[5] = block & 0xff;
 
 	/*
 	 * and transfer length
 	 */
-	rq->cmd[7] = (blocks >> 8) & 0xff;
-	rq->cmd[8] = blocks & 0xff;
-	rq->cmd_len = 10;
+	req->cmd[7] = (blocks >> 8) & 0xff;
+	req->cmd[8] = blocks & 0xff;
+	req->cmd_len = 10;
 	return BLKPREP_OK;
 }
 
@@ -1343,7 +1338,7 @@ static int ide_cdrom_prep_fs(struct request_queue *q, struct request *rq)
  */
 static int ide_cdrom_prep_pc(struct request *rq)
 {
-	u8 *c = rq->cmd;
+	u8 *c = scsi_req(rq)->cmd;
 
 	/* transform 6-byte read/write commands to the 10-byte version */
 	if (c[0] == READ_6 || c[0] == WRITE_6) {
@@ -1354,7 +1349,7 @@ static int ide_cdrom_prep_pc(struct request *rq)
 		c[2] = 0;
 		c[1] &= 0xe0;
 		c[0] += (READ_10 - READ_6);
-		rq->cmd_len = 10;
+		scsi_req(rq)->cmd_len = 10;
 		return BLKPREP_OK;
 	}
 
diff --git a/drivers/ide/ide-cd_ioctl.c b/drivers/ide/ide-cd_ioctl.c
index f085e3a..da0aa01 100644
--- a/drivers/ide/ide-cd_ioctl.c
+++ b/drivers/ide/ide-cd_ioctl.c
@@ -304,6 +304,7 @@ int ide_cdrom_reset(struct cdrom_device_info *cdi)
 	int ret;
 
 	rq = blk_get_request(drive->queue, READ, __GFP_RECLAIM);
+	scsi_req_init(rq);
 	rq->cmd_type = REQ_TYPE_DRV_PRIV;
 	rq->rq_flags = RQF_QUIET;
 	ret = blk_execute_rq(drive->queue, cd->disk, rq, 0);
diff --git a/drivers/ide/ide-cd_verbose.c b/drivers/ide/ide-cd_verbose.c
index f079ca2..58a6feb 100644
--- a/drivers/ide/ide-cd_verbose.c
+++ b/drivers/ide/ide-cd_verbose.c
@@ -315,12 +315,12 @@ void ide_cd_log_error(const char *name, struct request *failed_command,
 		while (hi > lo) {
 			mid = (lo + hi) / 2;
 			if (packet_command_texts[mid].packet_command ==
-			    failed_command->cmd[0]) {
+			    scsi_req(failed_command)->cmd[0]) {
 				s = packet_command_texts[mid].text;
 				break;
 			}
 			if (packet_command_texts[mid].packet_command >
-			    failed_command->cmd[0])
+			    scsi_req(failed_command)->cmd[0])
 				hi = mid;
 			else
 				lo = mid + 1;
@@ -329,7 +329,7 @@ void ide_cd_log_error(const char *name, struct request *failed_command,
 		printk(KERN_ERR "  The failed \"%s\" packet command "
 				"was: \n  \"", s);
 		for (i = 0; i < BLK_MAX_CDB; i++)
-			printk(KERN_CONT "%02x ", failed_command->cmd[i]);
+			printk(KERN_CONT "%02x ", scsi_req(failed_command)->cmd[i]);
 		printk(KERN_CONT "\"\n");
 	}
 
diff --git a/drivers/ide/ide-devsets.c b/drivers/ide/ide-devsets.c
index 0dd43b4..fd56c9d 100644
--- a/drivers/ide/ide-devsets.c
+++ b/drivers/ide/ide-devsets.c
@@ -166,10 +166,11 @@ int ide_devset_execute(ide_drive_t *drive, const struct ide_devset *setting,
 		return setting->set(drive, arg);
 
 	rq = blk_get_request(q, READ, __GFP_RECLAIM);
+	scsi_req_init(rq);
 	rq->cmd_type = REQ_TYPE_DRV_PRIV;
-	rq->cmd_len = 5;
-	rq->cmd[0] = REQ_DEVSET_EXEC;
-	*(int *)&rq->cmd[1] = arg;
+	scsi_req(rq)->cmd_len = 5;
+	scsi_req(rq)->cmd[0] = REQ_DEVSET_EXEC;
+	*(int *)&scsi_req(rq)->cmd[1] = arg;
 	rq->special = setting->set;
 
 	if (blk_execute_rq(q, NULL, rq, 0))
@@ -183,7 +184,7 @@ ide_startstop_t ide_do_devset(ide_drive_t *drive, struct request *rq)
 {
 	int err, (*setfunc)(ide_drive_t *, int) = rq->special;
 
-	err = setfunc(drive, *(int *)&rq->cmd[1]);
+	err = setfunc(drive, *(int *)&scsi_req(rq)->cmd[1]);
 	if (err)
 		rq->errors = err;
 	ide_complete_rq(drive, err, blk_rq_bytes(rq));
diff --git a/drivers/ide/ide-disk.c b/drivers/ide/ide-disk.c
index 5ceace5..3437c5b 100644
--- a/drivers/ide/ide-disk.c
+++ b/drivers/ide/ide-disk.c
@@ -478,6 +478,7 @@ static int set_multcount(ide_drive_t *drive, int arg)
 		return -EBUSY;
 
 	rq = blk_get_request(drive->queue, READ, __GFP_RECLAIM);
+	scsi_req_init(rq);
 	rq->cmd_type = REQ_TYPE_ATA_TASKFILE;
 
 	drive->mult_req = arg;
diff --git a/drivers/ide/ide-eh.c b/drivers/ide/ide-eh.c
index d6da011..35e5b89 100644
--- a/drivers/ide/ide-eh.c
+++ b/drivers/ide/ide-eh.c
@@ -148,7 +148,7 @@ static inline void ide_complete_drive_reset(ide_drive_t *drive, int err)
 	struct request *rq = drive->hwif->rq;
 
 	if (rq && rq->cmd_type == REQ_TYPE_DRV_PRIV &&
-	    rq->cmd[0] == REQ_DRIVE_RESET) {
+	    scsi_req(rq)->cmd[0] == REQ_DRIVE_RESET) {
 		if (err <= 0 && rq->errors == 0)
 			rq->errors = -EIO;
 		ide_complete_rq(drive, err ? err : 0, blk_rq_bytes(rq));
diff --git a/drivers/ide/ide-floppy.c b/drivers/ide/ide-floppy.c
index f079d8d..3bd678a 100644
--- a/drivers/ide/ide-floppy.c
+++ b/drivers/ide/ide-floppy.c
@@ -203,7 +203,7 @@ static void idefloppy_create_rw_cmd(ide_drive_t *drive,
 	put_unaligned(cpu_to_be16(blocks), (unsigned short *)&pc->c[7]);
 	put_unaligned(cpu_to_be32(block), (unsigned int *) &pc->c[2]);
 
-	memcpy(rq->cmd, pc->c, 12);
+	memcpy(scsi_req(rq)->cmd, pc->c, 12);
 
 	pc->rq = rq;
 	if (cmd == WRITE)
@@ -216,7 +216,7 @@ static void idefloppy_blockpc_cmd(struct ide_disk_obj *floppy,
 		struct ide_atapi_pc *pc, struct request *rq)
 {
 	ide_init_pc(pc);
-	memcpy(pc->c, rq->cmd, sizeof(pc->c));
+	memcpy(pc->c, scsi_req(rq)->cmd, sizeof(pc->c));
 	pc->rq = rq;
 	if (blk_rq_bytes(rq)) {
 		pc->flags |= PC_FLAG_DMA_OK;
diff --git a/drivers/ide/ide-io.c b/drivers/ide/ide-io.c
index 201e43f..3378503 100644
--- a/drivers/ide/ide-io.c
+++ b/drivers/ide/ide-io.c
@@ -279,7 +279,7 @@ static ide_startstop_t execute_drive_cmd (ide_drive_t *drive,
 
 static ide_startstop_t ide_special_rq(ide_drive_t *drive, struct request *rq)
 {
-	u8 cmd = rq->cmd[0];
+	u8 cmd = scsi_req(rq)->cmd[0];
 
 	switch (cmd) {
 	case REQ_PARK_HEADS:
@@ -545,6 +545,7 @@ void do_ide_request(struct request_queue *q)
 			goto plug_device;
 		}
 
+		scsi_req(rq)->resid_len = blk_rq_bytes(rq);
 		hwif->rq = rq;
 
 		spin_unlock_irq(&hwif->lock);
diff --git a/drivers/ide/ide-ioctls.c b/drivers/ide/ide-ioctls.c
index d05db24..a5d22c6 100644
--- a/drivers/ide/ide-ioctls.c
+++ b/drivers/ide/ide-ioctls.c
@@ -126,6 +126,7 @@ static int ide_cmd_ioctl(ide_drive_t *drive, unsigned long arg)
 		struct request *rq;
 
 		rq = blk_get_request(drive->queue, READ, __GFP_RECLAIM);
+		scsi_req_init(rq);
 		rq->cmd_type = REQ_TYPE_ATA_TASKFILE;
 		err = blk_execute_rq(drive->queue, NULL, rq, 0);
 		blk_put_request(rq);
@@ -222,9 +223,10 @@ static int generic_drive_reset(ide_drive_t *drive)
 	int ret = 0;
 
 	rq = blk_get_request(drive->queue, READ, __GFP_RECLAIM);
+	scsi_req_init(rq);
 	rq->cmd_type = REQ_TYPE_DRV_PRIV;
-	rq->cmd_len = 1;
-	rq->cmd[0] = REQ_DRIVE_RESET;
+	scsi_req(rq)->cmd_len = 1;
+	scsi_req(rq)->cmd[0] = REQ_DRIVE_RESET;
 	if (blk_execute_rq(drive->queue, NULL, rq, 1))
 		ret = rq->errors;
 	blk_put_request(rq);
diff --git a/drivers/ide/ide-park.c b/drivers/ide/ide-park.c
index 2d7dca5..c37604a 100644
--- a/drivers/ide/ide-park.c
+++ b/drivers/ide/ide-park.c
@@ -32,8 +32,9 @@ static void issue_park_cmd(ide_drive_t *drive, unsigned long timeout)
 	spin_unlock_irq(&hwif->lock);
 
 	rq = blk_get_request(q, READ, __GFP_RECLAIM);
-	rq->cmd[0] = REQ_PARK_HEADS;
-	rq->cmd_len = 1;
+	scsi_req_init(rq);
+	scsi_req(rq)->cmd[0] = REQ_PARK_HEADS;
+	scsi_req(rq)->cmd_len = 1;
 	rq->cmd_type = REQ_TYPE_DRV_PRIV;
 	rq->special = &timeout;
 	rc = blk_execute_rq(q, NULL, rq, 1);
@@ -46,11 +47,12 @@ static void issue_park_cmd(ide_drive_t *drive, unsigned long timeout)
 	 * timeout has expired, so power management will be reenabled.
 	 */
 	rq = blk_get_request(q, READ, GFP_NOWAIT);
+	scsi_req_init(rq);
 	if (IS_ERR(rq))
 		goto out;
 
-	rq->cmd[0] = REQ_UNPARK_HEADS;
-	rq->cmd_len = 1;
+	scsi_req(rq)->cmd[0] = REQ_UNPARK_HEADS;
+	scsi_req(rq)->cmd_len = 1;
 	rq->cmd_type = REQ_TYPE_DRV_PRIV;
 	elv_add_request(q, rq, ELEVATOR_INSERT_FRONT);
 
@@ -64,7 +66,7 @@ ide_startstop_t ide_do_park_unpark(ide_drive_t *drive, struct request *rq)
 	struct ide_taskfile *tf = &cmd.tf;
 
 	memset(&cmd, 0, sizeof(cmd));
-	if (rq->cmd[0] == REQ_PARK_HEADS) {
+	if (scsi_req(rq)->cmd[0] == REQ_PARK_HEADS) {
 		drive->sleep = *(unsigned long *)rq->special;
 		drive->dev_flags |= IDE_DFLAG_SLEEPING;
 		tf->command = ATA_CMD_IDLEIMMEDIATE;
diff --git a/drivers/ide/ide-pm.c b/drivers/ide/ide-pm.c
index a015acd..f6767ab 100644
--- a/drivers/ide/ide-pm.c
+++ b/drivers/ide/ide-pm.c
@@ -19,6 +19,7 @@ int generic_ide_suspend(struct device *dev, pm_message_t mesg)
 
 	memset(&rqpm, 0, sizeof(rqpm));
 	rq = blk_get_request(drive->queue, READ, __GFP_RECLAIM);
+	scsi_req_init(rq);
 	rq->cmd_type = REQ_TYPE_ATA_PM_SUSPEND;
 	rq->special = &rqpm;
 	rqpm.pm_step = IDE_PM_START_SUSPEND;
@@ -89,6 +90,7 @@ int generic_ide_resume(struct device *dev)
 
 	memset(&rqpm, 0, sizeof(rqpm));
 	rq = blk_get_request(drive->queue, READ, __GFP_RECLAIM);
+	scsi_req_init(rq);
 	rq->cmd_type = REQ_TYPE_ATA_PM_RESUME;
 	rq->rq_flags |= RQF_PREEMPT;
 	rq->special = &rqpm;
diff --git a/drivers/ide/ide-probe.c b/drivers/ide/ide-probe.c
index 330e319..a74ae8df 100644
--- a/drivers/ide/ide-probe.c
+++ b/drivers/ide/ide-probe.c
@@ -741,6 +741,14 @@ static void ide_port_tune_devices(ide_hwif_t *hwif)
 	}
 }
 
+static int ide_init_rq(struct request_queue *q, struct request *rq, gfp_t gfp)
+{
+	struct ide_request *req = blk_mq_rq_to_pdu(rq);
+
+	req->sreq.sense = req->sense;
+	return 0;
+}
+
 /*
  * init request queue
  */
@@ -758,11 +766,18 @@ static int ide_init_queue(ide_drive_t *drive)
 	 *	limits and LBA48 we could raise it but as yet
 	 *	do not.
 	 */
-
-	q = blk_init_queue_node(do_ide_request, NULL, hwif_to_node(hwif));
+	q = blk_alloc_queue_node(GFP_KERNEL, hwif_to_node(hwif));
 	if (!q)
 		return 1;
 
+	q->request_fn = do_ide_request;
+	q->init_rq_fn = ide_init_rq;
+	q->cmd_size = sizeof(struct ide_request);
+	if (blk_init_allocated_queue(q) < 0) {
+		blk_cleanup_queue(q);
+		return 1;
+	}
+
 	q->queuedata = drive;
 	blk_queue_segment_boundary(q, 0xffff);
 
@@ -1131,10 +1146,12 @@ static void ide_port_init_devices_data(ide_hwif_t *hwif)
 	ide_port_for_each_dev(i, drive, hwif) {
 		u8 j = (hwif->index * MAX_DRIVES) + i;
 		u16 *saved_id = drive->id;
+		struct request *saved_sense_rq = drive->sense_rq;
 
 		memset(drive, 0, sizeof(*drive));
 		memset(saved_id, 0, SECTOR_SIZE);
 		drive->id = saved_id;
+		drive->sense_rq = saved_sense_rq;
 
 		drive->media			= ide_disk;
 		drive->select			= (i << 4) | ATA_DEVICE_OBS;
@@ -1241,6 +1258,7 @@ static void ide_port_free_devices(ide_hwif_t *hwif)
 	int i;
 
 	ide_port_for_each_dev(i, drive, hwif) {
+		kfree(drive->sense_rq);
 		kfree(drive->id);
 		kfree(drive);
 	}
@@ -1248,11 +1266,10 @@ static void ide_port_free_devices(ide_hwif_t *hwif)
 
 static int ide_port_alloc_devices(ide_hwif_t *hwif, int node)
 {
+	ide_drive_t *drive;
 	int i;
 
 	for (i = 0; i < MAX_DRIVES; i++) {
-		ide_drive_t *drive;
-
 		drive = kzalloc_node(sizeof(*drive), GFP_KERNEL, node);
 		if (drive == NULL)
 			goto out_nomem;
@@ -1267,12 +1284,21 @@ static int ide_port_alloc_devices(ide_hwif_t *hwif, int node)
 		 */
 		drive->id = kzalloc_node(SECTOR_SIZE, GFP_KERNEL, node);
 		if (drive->id == NULL)
-			goto out_nomem;
+			goto out_free_drive;
+
+		drive->sense_rq = kmalloc(sizeof(struct request) +
+				sizeof(struct ide_request), GFP_KERNEL);
+		if (!drive->sense_rq)
+			goto out_free_id;
 
 		hwif->devices[i] = drive;
 	}
 	return 0;
 
+out_free_id:
+	kfree(drive->id);
+out_free_drive:
+	kfree(drive);
 out_nomem:
 	ide_port_free_devices(hwif);
 	return -ENOMEM;
diff --git a/drivers/ide/ide-tape.c b/drivers/ide/ide-tape.c
index 9ecf4e3..f6bc1e2 100644
--- a/drivers/ide/ide-tape.c
+++ b/drivers/ide/ide-tape.c
@@ -282,7 +282,7 @@ static void idetape_analyze_error(ide_drive_t *drive)
 
 	/* correct remaining bytes to transfer */
 	if (pc->flags & PC_FLAG_DMA_ERROR)
-		rq->resid_len = tape->blk_size * get_unaligned_be32(&sense[3]);
+		scsi_req(rq)->resid_len = tape->blk_size * get_unaligned_be32(&sense[3]);
 
 	/*
 	 * If error was the result of a zero-length read or write command,
@@ -316,7 +316,7 @@ static void idetape_analyze_error(ide_drive_t *drive)
 			pc->flags |= PC_FLAG_ABORT;
 		}
 		if (!(pc->flags & PC_FLAG_ABORT) &&
-		    (blk_rq_bytes(rq) - rq->resid_len))
+		    (blk_rq_bytes(rq) - scsi_req(rq)->resid_len))
 			pc->retries = IDETAPE_MAX_PC_RETRIES + 1;
 	}
 }
@@ -348,7 +348,7 @@ static int ide_tape_callback(ide_drive_t *drive, int dsc)
 					"itself - Aborting request!\n");
 	} else if (pc->c[0] == READ_6 || pc->c[0] == WRITE_6) {
 		unsigned int blocks =
-			(blk_rq_bytes(rq) - rq->resid_len) / tape->blk_size;
+			(blk_rq_bytes(rq) - scsi_req(rq)->resid_len) / tape->blk_size;
 
 		tape->avg_size += blocks * tape->blk_size;
 
@@ -560,7 +560,7 @@ static void ide_tape_create_rw_cmd(idetape_tape_t *tape,
 		pc->flags |= PC_FLAG_WRITING;
 	}
 
-	memcpy(rq->cmd, pc->c, 12);
+	memcpy(scsi_req(rq)->cmd, pc->c, 12);
 }
 
 static ide_startstop_t idetape_do_request(ide_drive_t *drive,
@@ -570,10 +570,11 @@ static ide_startstop_t idetape_do_request(ide_drive_t *drive,
 	idetape_tape_t *tape = drive->driver_data;
 	struct ide_atapi_pc *pc = NULL;
 	struct ide_cmd cmd;
+	struct scsi_request *req = scsi_req(rq);
 	u8 stat;
 
 	ide_debug_log(IDE_DBG_RQ, "cmd: 0x%x, sector: %llu, nr_sectors: %u",
-		      rq->cmd[0], (unsigned long long)blk_rq_pos(rq),
+		      req->cmd[0], (unsigned long long)blk_rq_pos(rq),
 		      blk_rq_sectors(rq));
 
 	BUG_ON(!(rq->cmd_type == REQ_TYPE_DRV_PRIV ||
@@ -592,7 +593,7 @@ static ide_startstop_t idetape_do_request(ide_drive_t *drive,
 	stat = hwif->tp_ops->read_status(hwif);
 
 	if ((drive->dev_flags & IDE_DFLAG_DSC_OVERLAP) == 0 &&
-	    (rq->cmd[13] & REQ_IDETAPE_PC2) == 0)
+	    (req->cmd[13] & REQ_IDETAPE_PC2) == 0)
 		drive->atapi_flags |= IDE_AFLAG_IGNORE_DSC;
 
 	if (drive->dev_flags & IDE_DFLAG_POST_RESET) {
@@ -609,7 +610,7 @@ static ide_startstop_t idetape_do_request(ide_drive_t *drive,
 		} else if (time_after(jiffies, tape->dsc_timeout)) {
 			printk(KERN_ERR "ide-tape: %s: DSC timeout\n",
 				tape->name);
-			if (rq->cmd[13] & REQ_IDETAPE_PC2) {
+			if (req->cmd[13] & REQ_IDETAPE_PC2) {
 				idetape_media_access_finished(drive);
 				return ide_stopped;
 			} else {
@@ -626,23 +627,23 @@ static ide_startstop_t idetape_do_request(ide_drive_t *drive,
 		tape->postponed_rq = false;
 	}
 
-	if (rq->cmd[13] & REQ_IDETAPE_READ) {
+	if (req->cmd[13] & REQ_IDETAPE_READ) {
 		pc = &tape->queued_pc;
 		ide_tape_create_rw_cmd(tape, pc, rq, READ_6);
 		goto out;
 	}
-	if (rq->cmd[13] & REQ_IDETAPE_WRITE) {
+	if (req->cmd[13] & REQ_IDETAPE_WRITE) {
 		pc = &tape->queued_pc;
 		ide_tape_create_rw_cmd(tape, pc, rq, WRITE_6);
 		goto out;
 	}
-	if (rq->cmd[13] & REQ_IDETAPE_PC1) {
+	if (req->cmd[13] & REQ_IDETAPE_PC1) {
 		pc = (struct ide_atapi_pc *)rq->special;
-		rq->cmd[13] &= ~(REQ_IDETAPE_PC1);
-		rq->cmd[13] |= REQ_IDETAPE_PC2;
+		req->cmd[13] &= ~(REQ_IDETAPE_PC1);
+		req->cmd[13] |= REQ_IDETAPE_PC2;
 		goto out;
 	}
-	if (rq->cmd[13] & REQ_IDETAPE_PC2) {
+	if (req->cmd[13] & REQ_IDETAPE_PC2) {
 		idetape_media_access_finished(drive);
 		return ide_stopped;
 	}
@@ -853,8 +854,9 @@ static int idetape_queue_rw_tail(ide_drive_t *drive, int cmd, int size)
 	BUG_ON(size < 0 || size % tape->blk_size);
 
 	rq = blk_get_request(drive->queue, READ, __GFP_RECLAIM);
+	scsi_req_init(rq);
 	rq->cmd_type = REQ_TYPE_DRV_PRIV;
-	rq->cmd[13] = cmd;
+	scsi_req(rq)->cmd[13] = cmd;
 	rq->rq_disk = tape->disk;
 	rq->__sector = tape->first_frame;
 
@@ -868,7 +870,7 @@ static int idetape_queue_rw_tail(ide_drive_t *drive, int cmd, int size)
 	blk_execute_rq(drive->queue, tape->disk, rq, 0);
 
 	/* calculate the number of transferred bytes and update buffer state */
-	size -= rq->resid_len;
+	size -= scsi_req(rq)->resid_len;
 	tape->cur = tape->buf;
 	if (cmd == REQ_IDETAPE_READ)
 		tape->valid = size;
diff --git a/drivers/ide/ide-taskfile.c b/drivers/ide/ide-taskfile.c
index a716693..a393e13 100644
--- a/drivers/ide/ide-taskfile.c
+++ b/drivers/ide/ide-taskfile.c
@@ -431,6 +431,7 @@ int ide_raw_taskfile(ide_drive_t *drive, struct ide_cmd *cmd, u8 *buf,
 	int rw = !(cmd->tf_flags & IDE_TFLAG_WRITE) ? READ : WRITE;
 
 	rq = blk_get_request(drive->queue, rw, __GFP_RECLAIM);
+	scsi_req_init(rq);
 	rq->cmd_type = REQ_TYPE_ATA_TASKFILE;
 
 	/*
diff --git a/drivers/ide/sis5513.c b/drivers/ide/sis5513.c
index 247853e..c3062b5 100644
--- a/drivers/ide/sis5513.c
+++ b/drivers/ide/sis5513.c
@@ -54,7 +54,7 @@
 #define DRV_NAME "sis5513"
 
 /* registers layout and init values are chipset family dependent */
-
+#undef ATA_16
 #define ATA_16		0x01
 #define ATA_33		0x02
 #define ATA_66		0x03
diff --git a/drivers/message/fusion/mptsas.c b/drivers/message/fusion/mptsas.c
index 7ee1667..b8c4b2b 100644
--- a/drivers/message/fusion/mptsas.c
+++ b/drivers/message/fusion/mptsas.c
@@ -2320,10 +2320,10 @@ static int mptsas_smp_handler(struct Scsi_Host *shost, struct sas_rphy *rphy,
 		SmpPassthroughReply_t *smprep;
 
 		smprep = (SmpPassthroughReply_t *)ioc->sas_mgmt.reply;
-		memcpy(req->sense, smprep, sizeof(*smprep));
-		req->sense_len = sizeof(*smprep);
-		req->resid_len = 0;
-		rsp->resid_len -= smprep->ResponseDataLength;
+		memcpy(scsi_req(req)->sense, smprep, sizeof(*smprep));
+		scsi_req(req)->sense_len = sizeof(*smprep);
+		scsi_req(req)->resid_len = 0;
+		scsi_req(rsp)->resid_len -= smprep->ResponseDataLength;
 	} else {
 		printk(MYIOC_s_ERR_FMT
 		    "%s: smp passthru reply failed to be returned\n",
diff --git a/drivers/scsi/libfc/fc_lport.c b/drivers/scsi/libfc/fc_lport.c
index 919736a..aa76f36 100644
--- a/drivers/scsi/libfc/fc_lport.c
+++ b/drivers/scsi/libfc/fc_lport.c
@@ -2095,7 +2095,7 @@ int fc_lport_bsg_request(struct bsg_job *job)
 
 	bsg_reply->reply_payload_rcv_len = 0;
 	if (rsp)
-		rsp->resid_len = job->reply_payload.payload_len;
+		scsi_req(rsp)->resid_len = job->reply_payload.payload_len;
 
 	mutex_lock(&lport->lp_mutex);
 
diff --git a/drivers/scsi/libsas/sas_expander.c b/drivers/scsi/libsas/sas_expander.c
index 022bb6e..570b2cb 100644
--- a/drivers/scsi/libsas/sas_expander.c
+++ b/drivers/scsi/libsas/sas_expander.c
@@ -2174,12 +2174,12 @@ int sas_smp_handler(struct Scsi_Host *shost, struct sas_rphy *rphy,
 			       bio_data(rsp->bio), blk_rq_bytes(rsp));
 	if (ret > 0) {
 		/* positive number is the untransferred residual */
-		rsp->resid_len = ret;
-		req->resid_len = 0;
+		scsi_req(rsp)->resid_len = ret;
+		scsi_req(req)->resid_len = 0;
 		ret = 0;
 	} else if (ret == 0) {
-		rsp->resid_len = 0;
-		req->resid_len = 0;
+		scsi_req(rsp)->resid_len = 0;
+		scsi_req(req)->resid_len = 0;
 	}
 
 	return ret;
diff --git a/drivers/scsi/libsas/sas_host_smp.c b/drivers/scsi/libsas/sas_host_smp.c
index d247925..45cbbc4 100644
--- a/drivers/scsi/libsas/sas_host_smp.c
+++ b/drivers/scsi/libsas/sas_host_smp.c
@@ -274,15 +274,15 @@ int sas_smp_host_handler(struct Scsi_Host *shost, struct request *req,
 
 	switch (req_data[1]) {
 	case SMP_REPORT_GENERAL:
-		req->resid_len -= 8;
-		rsp->resid_len -= 32;
+		scsi_req(req)->resid_len -= 8;
+		scsi_req(rsp)->resid_len -= 32;
 		resp_data[2] = SMP_RESP_FUNC_ACC;
 		resp_data[9] = sas_ha->num_phys;
 		break;
 
 	case SMP_REPORT_MANUF_INFO:
-		req->resid_len -= 8;
-		rsp->resid_len -= 64;
+		scsi_req(req)->resid_len -= 8;
+		scsi_req(rsp)->resid_len -= 64;
 		resp_data[2] = SMP_RESP_FUNC_ACC;
 		memcpy(resp_data + 12, shost->hostt->name,
 		       SAS_EXPANDER_VENDOR_ID_LEN);
@@ -295,13 +295,13 @@ int sas_smp_host_handler(struct Scsi_Host *shost, struct request *req,
 		break;
 
 	case SMP_DISCOVER:
-		req->resid_len -= 16;
-		if ((int)req->resid_len < 0) {
-			req->resid_len = 0;
+		scsi_req(req)->resid_len -= 16;
+		if ((int)scsi_req(req)->resid_len < 0) {
+			scsi_req(req)->resid_len = 0;
 			error = -EINVAL;
 			goto out;
 		}
-		rsp->resid_len -= 56;
+		scsi_req(rsp)->resid_len -= 56;
 		sas_host_smp_discover(sas_ha, resp_data, req_data[9]);
 		break;
 
@@ -311,13 +311,13 @@ int sas_smp_host_handler(struct Scsi_Host *shost, struct request *req,
 		break;
 
 	case SMP_REPORT_PHY_SATA:
-		req->resid_len -= 16;
-		if ((int)req->resid_len < 0) {
-			req->resid_len = 0;
+		scsi_req(req)->resid_len -= 16;
+		if ((int)scsi_req(req)->resid_len < 0) {
+			scsi_req(req)->resid_len = 0;
 			error = -EINVAL;
 			goto out;
 		}
-		rsp->resid_len -= 60;
+		scsi_req(rsp)->resid_len -= 60;
 		sas_report_phy_sata(sas_ha, resp_data, req_data[9]);
 		break;
 
@@ -331,15 +331,15 @@ int sas_smp_host_handler(struct Scsi_Host *shost, struct request *req,
 		int to_write = req_data[4];
 
 		if (blk_rq_bytes(req) < base_frame_size + to_write * 4 ||
-		    req->resid_len < base_frame_size + to_write * 4) {
+		    scsi_req(req)->resid_len < base_frame_size + to_write * 4) {
 			resp_data[2] = SMP_RESP_INV_FRM_LEN;
 			break;
 		}
 
 		to_write = sas_host_smp_write_gpio(sas_ha, resp_data, req_data[2],
 						   req_data[3], to_write, &req_data[8]);
-		req->resid_len -= base_frame_size + to_write * 4;
-		rsp->resid_len -= 8;
+		scsi_req(req)->resid_len -= base_frame_size + to_write * 4;
+		scsi_req(rsp)->resid_len -= 8;
 		break;
 	}
 
@@ -348,13 +348,13 @@ int sas_smp_host_handler(struct Scsi_Host *shost, struct request *req,
 		break;
 
 	case SMP_PHY_CONTROL:
-		req->resid_len -= 44;
-		if ((int)req->resid_len < 0) {
-			req->resid_len = 0;
+		scsi_req(req)->resid_len -= 44;
+		if ((int)scsi_req(req)->resid_len < 0) {
+			scsi_req(req)->resid_len = 0;
 			error = -EINVAL;
 			goto out;
 		}
-		rsp->resid_len -= 8;
+		scsi_req(rsp)->resid_len -= 8;
 		sas_phy_control(sas_ha, req_data[9], req_data[10],
 				req_data[32] >> 4, req_data[33] >> 4,
 				resp_data);
diff --git a/drivers/scsi/mpt3sas/mpt3sas_transport.c b/drivers/scsi/mpt3sas/mpt3sas_transport.c
index 7f1d578..e7a7a70 100644
--- a/drivers/scsi/mpt3sas/mpt3sas_transport.c
+++ b/drivers/scsi/mpt3sas/mpt3sas_transport.c
@@ -2057,10 +2057,10 @@ _transport_smp_handler(struct Scsi_Host *shost, struct sas_rphy *rphy,
 		    ioc->name, __func__,
 		    le16_to_cpu(mpi_reply->ResponseDataLength)));
 
-		memcpy(req->sense, mpi_reply, sizeof(*mpi_reply));
-		req->sense_len = sizeof(*mpi_reply);
-		req->resid_len = 0;
-		rsp->resid_len -=
+		memcpy(scsi_req(req)->sense, mpi_reply, sizeof(*mpi_reply));
+		scsi_req(req)->sense_len = sizeof(*mpi_reply);
+		scsi_req(req)->resid_len = 0;
+		scsi_req(rsp)->resid_len -=
 		    le16_to_cpu(mpi_reply->ResponseDataLength);
 
 		/* check if the resp needs to be copied from the allocated
diff --git a/drivers/scsi/osd/osd_initiator.c b/drivers/scsi/osd/osd_initiator.c
index ef99f62..fcb040e 100644
--- a/drivers/scsi/osd/osd_initiator.c
+++ b/drivers/scsi/osd/osd_initiator.c
@@ -48,6 +48,7 @@
 #include <scsi/osd_sense.h>
 
 #include <scsi/scsi_device.h>
+#include <scsi/scsi_request.h>
 
 #include "osd_debug.h"
 
@@ -477,11 +478,13 @@ static void _set_error_resid(struct osd_request *or, struct request *req,
 {
 	or->async_error = error;
 	or->req_errors = req->errors ? : error;
-	or->sense_len = req->sense_len;
+	or->sense_len = scsi_req(req)->sense_len;
+	if (or->sense_len)
+		memcpy(or->sense, scsi_req(req)->sense, or->sense_len);
 	if (or->out.req)
-		or->out.residual = or->out.req->resid_len;
+		or->out.residual = scsi_req(or->out.req)->resid_len;
 	if (or->in.req)
-		or->in.residual = or->in.req->resid_len;
+		or->in.residual = scsi_req(or->in.req)->resid_len;
 }
 
 int osd_execute_request(struct osd_request *or)
@@ -1565,7 +1568,7 @@ static struct request *_make_request(struct request_queue *q, bool has_write,
 	req = blk_get_request(q, has_write ? WRITE : READ, flags);
 	if (IS_ERR(req))
 		return req;
-	blk_rq_set_block_pc(req);
+	scsi_req_init(req);
 
 	for_each_bio(bio) {
 		struct bio *bounce_bio = bio;
@@ -1599,8 +1602,6 @@ static int _init_blk_request(struct osd_request *or,
 
 	req->timeout = or->timeout;
 	req->retries = or->retries;
-	req->sense = or->sense;
-	req->sense_len = 0;
 
 	if (has_out) {
 		or->out.req = req;
@@ -1612,7 +1613,7 @@ static int _init_blk_request(struct osd_request *or,
 				ret = PTR_ERR(req);
 				goto out;
 			}
-			blk_rq_set_block_pc(req);
+			scsi_req_init(req);
 			or->in.req = or->request->next_rq = req;
 		}
 	} else if (has_in)
@@ -1699,8 +1700,8 @@ int osd_finalize_request(struct osd_request *or,
 
 	osd_sec_sign_cdb(&or->cdb, cap_key);
 
-	or->request->cmd = or->cdb.buff;
-	or->request->cmd_len = _osd_req_cdb_len(or);
+	scsi_req(or->request)->cmd = or->cdb.buff;
+	scsi_req(or->request)->cmd_len = _osd_req_cdb_len(or);
 
 	return 0;
 }
diff --git a/drivers/scsi/osst.c b/drivers/scsi/osst.c
index e8196c5..d314aa5 100644
--- a/drivers/scsi/osst.c
+++ b/drivers/scsi/osst.c
@@ -322,6 +322,7 @@ static int osst_chk_result(struct osst_tape * STp, struct osst_request * SRpnt)
 /* Wakeup from interrupt */
 static void osst_end_async(struct request *req, int update)
 {
+	struct scsi_request *rq = scsi_req(req);
 	struct osst_request *SRpnt = req->end_io_data;
 	struct osst_tape *STp = SRpnt->stp;
 	struct rq_map_data *mdata = &SRpnt->stp->buffer->map_data;
@@ -330,6 +331,8 @@ static void osst_end_async(struct request *req, int update)
 #if DEBUG
 	STp->write_pending = 0;
 #endif
+	if (rq->sense_len)
+		memcpy(SRpnt->sense, rq->sense, SCSI_SENSE_BUFFERSIZE);
 	if (SRpnt->waiting)
 		complete(SRpnt->waiting);
 
@@ -357,6 +360,7 @@ static int osst_execute(struct osst_request *SRpnt, const unsigned char *cmd,
 			int use_sg, int timeout, int retries)
 {
 	struct request *req;
+	struct scsi_request *rq;
 	struct page **pages = NULL;
 	struct rq_map_data *mdata = &SRpnt->stp->buffer->map_data;
 
@@ -367,7 +371,8 @@ static int osst_execute(struct osst_request *SRpnt, const unsigned char *cmd,
 	if (IS_ERR(req))
 		return DRIVER_ERROR << 24;
 
-	blk_rq_set_block_pc(req);
+	rq = scsi_req(req);
+	scsi_req_init(req);
 	req->rq_flags |= RQF_QUIET;
 
 	SRpnt->bio = NULL;
@@ -404,11 +409,9 @@ static int osst_execute(struct osst_request *SRpnt, const unsigned char *cmd,
 			goto free_req;
 	}
 
-	req->cmd_len = cmd_len;
-	memset(req->cmd, 0, BLK_MAX_CDB); /* ATAPI hates garbage after CDB */
-	memcpy(req->cmd, cmd, req->cmd_len);
-	req->sense = SRpnt->sense;
-	req->sense_len = 0;
+	rq->cmd_len = cmd_len;
+	memset(rq->cmd, 0, BLK_MAX_CDB); /* ATAPI hates garbage after CDB */
+	memcpy(rq->cmd, cmd, rq->cmd_len);
 	req->timeout = timeout;
 	req->retries = retries;
 	req->end_io_data = SRpnt;
diff --git a/drivers/scsi/qla2xxx/qla_bsg.c b/drivers/scsi/qla2xxx/qla_bsg.c
index 1bf8061..40ca75b 100644
--- a/drivers/scsi/qla2xxx/qla_bsg.c
+++ b/drivers/scsi/qla2xxx/qla_bsg.c
@@ -921,7 +921,7 @@ qla2x00_process_loopback(struct bsg_job *bsg_job)
 
 	bsg_job->reply_len = sizeof(struct fc_bsg_reply) +
 	    sizeof(response) + sizeof(uint8_t);
-	fw_sts_ptr = ((uint8_t *)bsg_job->req->sense) +
+	fw_sts_ptr = ((uint8_t *)scsi_req(bsg_job->req)->sense) +
 	    sizeof(struct fc_bsg_reply);
 	memcpy(fw_sts_ptr, response, sizeof(response));
 	fw_sts_ptr += sizeof(response);
diff --git a/drivers/scsi/qla2xxx/qla_isr.c b/drivers/scsi/qla2xxx/qla_isr.c
index 5093ca9..759f281 100644
--- a/drivers/scsi/qla2xxx/qla_isr.c
+++ b/drivers/scsi/qla2xxx/qla_isr.c
@@ -1472,7 +1472,8 @@ qla24xx_els_ct_entry(scsi_qla_host_t *vha, struct req_que *req,
 			    type, sp->handle, comp_status, fw_status[1], fw_status[2],
 			    le16_to_cpu(((struct els_sts_entry_24xx *)
 				pkt)->total_byte_count));
-			fw_sts_ptr = ((uint8_t*)bsg_job->req->sense) + sizeof(struct fc_bsg_reply);
+			fw_sts_ptr = ((uint8_t*)scsi_req(bsg_job->req)->sense) +
+				sizeof(struct fc_bsg_reply);
 			memcpy( fw_sts_ptr, fw_status, sizeof(fw_status));
 		}
 		else {
@@ -1486,7 +1487,8 @@ qla24xx_els_ct_entry(scsi_qla_host_t *vha, struct req_que *req,
 				    pkt)->error_subcode_2));
 			res = DID_ERROR << 16;
 			bsg_reply->reply_payload_rcv_len = 0;
-			fw_sts_ptr = ((uint8_t*)bsg_job->req->sense) + sizeof(struct fc_bsg_reply);
+			fw_sts_ptr = ((uint8_t*)scsi_req(bsg_job->req)->sense) +
+					sizeof(struct fc_bsg_reply);
 			memcpy( fw_sts_ptr, fw_status, sizeof(fw_status));
 		}
 		ql_dump_buffer(ql_dbg_user + ql_dbg_buffer, vha, 0x5056,
diff --git a/drivers/scsi/qla2xxx/qla_mr.c b/drivers/scsi/qla2xxx/qla_mr.c
index 02f1de1..96c33e29 100644
--- a/drivers/scsi/qla2xxx/qla_mr.c
+++ b/drivers/scsi/qla2xxx/qla_mr.c
@@ -2244,7 +2244,7 @@ qlafx00_ioctl_iosb_entry(scsi_qla_host_t *vha, struct req_que *req,
 		memcpy(fstatus.reserved_3,
 		    pkt->reserved_2, 20 * sizeof(uint8_t));
 
-		fw_sts_ptr = ((uint8_t *)bsg_job->req->sense) +
+		fw_sts_ptr = ((uint8_t *)scsi_req(bsg_job->req)->sense) +
 		    sizeof(struct fc_bsg_reply);
 
 		memcpy(fw_sts_ptr, (uint8_t *)&fstatus,
diff --git a/drivers/scsi/scsi_error.c b/drivers/scsi/scsi_error.c
index 7c08460..4b40f74 100644
--- a/drivers/scsi/scsi_error.c
+++ b/drivers/scsi/scsi_error.c
@@ -1968,6 +1968,7 @@ static void eh_lock_door_done(struct request *req, int uptodate)
 static void scsi_eh_lock_door(struct scsi_device *sdev)
 {
 	struct request *req;
+	struct scsi_request *rq;
 
 	/*
 	 * blk_get_request with GFP_KERNEL (__GFP_RECLAIM) sleeps until a
@@ -1976,17 +1977,16 @@ static void scsi_eh_lock_door(struct scsi_device *sdev)
 	req = blk_get_request(sdev->request_queue, READ, GFP_KERNEL);
 	if (IS_ERR(req))
 		return;
+	rq = scsi_req(req);
+	scsi_req_init(req);
 
-	blk_rq_set_block_pc(req);
-
-	req->cmd[0] = ALLOW_MEDIUM_REMOVAL;
-	req->cmd[1] = 0;
-	req->cmd[2] = 0;
-	req->cmd[3] = 0;
-	req->cmd[4] = SCSI_REMOVAL_PREVENT;
-	req->cmd[5] = 0;
-
-	req->cmd_len = COMMAND_SIZE(req->cmd[0]);
+	rq->cmd[0] = ALLOW_MEDIUM_REMOVAL;
+	rq->cmd[1] = 0;
+	rq->cmd[2] = 0;
+	rq->cmd[3] = 0;
+	rq->cmd[4] = SCSI_REMOVAL_PREVENT;
+	rq->cmd[5] = 0;
+	rq->cmd_len = COMMAND_SIZE(rq->cmd[0]);
 
 	req->rq_flags |= RQF_QUIET;
 	req->timeout = 10 * HZ;
@@ -2355,7 +2355,7 @@ scsi_ioctl_reset(struct scsi_device *dev, int __user *arg)
 	scmd = (struct scsi_cmnd *)(rq + 1);
 	scsi_init_command(dev, scmd);
 	scmd->request = rq;
-	scmd->cmnd = rq->cmd;
+	scmd->cmnd = scsi_req(rq)->cmd;
 
 	scmd->scsi_done		= scsi_reset_provider_done_command;
 	memset(&scmd->sdb, 0, sizeof(scmd->sdb));
diff --git a/drivers/scsi/scsi_lib.c b/drivers/scsi/scsi_lib.c
index 3c11892..f7ed71b 100644
--- a/drivers/scsi/scsi_lib.c
+++ b/drivers/scsi/scsi_lib.c
@@ -220,21 +220,21 @@ static int __scsi_execute(struct scsi_device *sdev, const unsigned char *cmd,
 {
 	struct request *req;
 	int write = (data_direction == DMA_TO_DEVICE);
+	struct scsi_request *rq;
 	int ret = DRIVER_ERROR << 24;
 
 	req = blk_get_request(sdev->request_queue, write, __GFP_RECLAIM);
 	if (IS_ERR(req))
 		return ret;
-	blk_rq_set_block_pc(req);
+	rq = scsi_req(req);
+	scsi_req_init(req);
 
 	if (bufflen &&	blk_rq_map_kern(sdev->request_queue, req,
 					buffer, bufflen, __GFP_RECLAIM))
 		goto out;
 
-	req->cmd_len = COMMAND_SIZE(cmd[0]);
-	memcpy(req->cmd, cmd, req->cmd_len);
-	req->sense = sense;
-	req->sense_len = 0;
+	rq->cmd_len = COMMAND_SIZE(cmd[0]);
+	memcpy(rq->cmd, cmd, rq->cmd_len);
 	req->retries = retries;
 	req->timeout = timeout;
 	req->cmd_flags |= flags;
@@ -251,11 +251,13 @@ static int __scsi_execute(struct scsi_device *sdev, const unsigned char *cmd,
 	 * is invalid.  Prevent the garbage from being misinterpreted
 	 * and prevent security leaks by zeroing out the excess data.
 	 */
-	if (unlikely(req->resid_len > 0 && req->resid_len <= bufflen))
-		memset(buffer + (bufflen - req->resid_len), 0, req->resid_len);
+	if (unlikely(rq->resid_len > 0 && rq->resid_len <= bufflen))
+		memset(buffer + (bufflen - rq->resid_len), 0, rq->resid_len);
 
 	if (resid)
-		*resid = req->resid_len;
+		*resid = rq->resid_len;
+	if (rq->sense_len)
+		memcpy(sense, rq->sense, SCSI_SENSE_BUFFERSIZE);
 	ret = req->errors;
  out:
 	blk_put_request(req);
@@ -806,16 +808,13 @@ void scsi_io_completion(struct scsi_cmnd *cmd, unsigned int good_bytes)
 
 	if (req->cmd_type == REQ_TYPE_BLOCK_PC) { /* SG_IO ioctl from block level */
 		if (result) {
-			if (sense_valid && req->sense) {
+			if (sense_valid) {
 				/*
 				 * SG_IO wants current and deferred errors
 				 */
-				int len = 8 + cmd->sense_buffer[7];
-
-				if (len > SCSI_SENSE_BUFFERSIZE)
-					len = SCSI_SENSE_BUFFERSIZE;
-				memcpy(req->sense, cmd->sense_buffer,  len);
-				req->sense_len = len;
+				scsi_req(req)->sense_len =
+					min(8 + cmd->sense_buffer[7],
+					    SCSI_SENSE_BUFFERSIZE);
 			}
 			if (!sense_deferred)
 				error = __scsi_error_from_host_byte(cmd, result);
@@ -825,14 +824,14 @@ void scsi_io_completion(struct scsi_cmnd *cmd, unsigned int good_bytes)
 		 */
 		req->errors = cmd->result;
 
-		req->resid_len = scsi_get_resid(cmd);
+		scsi_req(req)->resid_len = scsi_get_resid(cmd);
 
 		if (scsi_bidi_cmnd(cmd)) {
 			/*
 			 * Bidi commands Must be complete as a whole,
 			 * both sides at once.
 			 */
-			req->next_rq->resid_len = scsi_in(cmd)->resid;
+			scsi_req(req->next_rq)->resid_len = scsi_in(cmd)->resid;
 			if (scsi_end_request(req, 0, blk_rq_bytes(req),
 					blk_rq_bytes(req->next_rq)))
 				BUG();
@@ -1165,7 +1164,8 @@ void scsi_init_command(struct scsi_device *dev, struct scsi_cmnd *cmd)
 	void *prot = cmd->prot_sdb;
 	unsigned long flags;
 
-	memset(cmd, 0, sizeof(*cmd));
+	memset((char *)cmd + sizeof(cmd->req), 0,
+		sizeof(*cmd) - sizeof(cmd->req));
 	cmd->device = dev;
 	cmd->sense_buffer = buf;
 	cmd->prot_sdb = prot;
@@ -1197,7 +1197,8 @@ static int scsi_setup_blk_pc_cmnd(struct scsi_device *sdev, struct request *req)
 		memset(&cmd->sdb, 0, sizeof(cmd->sdb));
 	}
 
-	cmd->cmd_len = req->cmd_len;
+	cmd->cmd_len = scsi_req(req)->cmd_len;
+	cmd->cmnd = scsi_req(req)->cmd;
 	cmd->transfersize = blk_rq_bytes(req);
 	cmd->allowed = req->retries;
 	return BLKPREP_OK;
@@ -1217,6 +1218,7 @@ static int scsi_setup_fs_cmnd(struct scsi_device *sdev, struct request *req)
 			return ret;
 	}
 
+	cmd->cmnd = scsi_req(req)->cmd = scsi_req(req)->__cmd;
 	memset(cmd->cmnd, 0, BLK_MAX_CDB);
 	return scsi_cmd_to_driver(cmd)->init_command(cmd);
 }
@@ -1355,7 +1357,6 @@ static int scsi_prep_fn(struct request_queue *q, struct request *req)
 
 	cmd->tag = req->tag;
 	cmd->request = req;
-	cmd->cmnd = req->cmd;
 	cmd->prot_op = SCSI_PROT_NORMAL;
 
 	ret = scsi_setup_cmnd(sdev, req);
@@ -1874,7 +1875,8 @@ static int scsi_mq_prep_fn(struct request *req)
 	unsigned char *sense_buf = cmd->sense_buffer;
 	struct scatterlist *sg;
 
-	memset(cmd, 0, sizeof(struct scsi_cmnd));
+	memset((char *)cmd + sizeof(cmd->req), 0,
+		sizeof(*cmd) - sizeof(cmd->req));
 
 	req->special = cmd;
 
@@ -1884,7 +1886,6 @@ static int scsi_mq_prep_fn(struct request *req)
 
 	cmd->tag = req->tag;
 
-	cmd->cmnd = req->cmd;
 	cmd->prot_op = SCSI_PROT_NORMAL;
 
 	INIT_LIST_HEAD(&cmd->list);
@@ -1959,7 +1960,6 @@ static int scsi_queue_rq(struct blk_mq_hw_ctx *hctx,
 	if (!scsi_host_queue_ready(q, shost, sdev))
 		goto out_dec_target_busy;
 
-
 	if (!(req->rq_flags & RQF_DONTPREP)) {
 		ret = prep_to_mq(scsi_mq_prep_fn(req));
 		if (ret != BLK_MQ_RQ_QUEUE_OK)
@@ -2036,6 +2036,7 @@ static int scsi_init_request(void *data, struct request *rq,
 		scsi_alloc_sense_buffer(shost, GFP_KERNEL, numa_node);
 	if (!cmd->sense_buffer)
 		return -ENOMEM;
+	cmd->req.sense = cmd->sense_buffer;
 	return 0;
 }
 
@@ -2125,6 +2126,7 @@ static int scsi_init_rq(struct request_queue *q, struct request *rq, gfp_t gfp)
 	cmd->sense_buffer = scsi_alloc_sense_buffer(shost, gfp, NUMA_NO_NODE);
 	if (!cmd->sense_buffer)
 		goto fail;
+	cmd->req.sense = cmd->sense_buffer;
 
 	if (scsi_host_get_prot(shost) >= SHOST_DIX_TYPE0_PROTECTION) {
 		cmd->prot_sdb = kmem_cache_zalloc(scsi_sdb_cache, gfp);
diff --git a/drivers/scsi/scsi_transport_sas.c b/drivers/scsi/scsi_transport_sas.c
index 60b651b..126a5ee 100644
--- a/drivers/scsi/scsi_transport_sas.c
+++ b/drivers/scsi/scsi_transport_sas.c
@@ -33,6 +33,7 @@
 #include <linux/bsg.h>
 
 #include <scsi/scsi.h>
+#include <scsi/scsi_request.h>
 #include <scsi/scsi_device.h>
 #include <scsi/scsi_host.h>
 #include <scsi/scsi_transport.h>
@@ -177,6 +178,10 @@ static void sas_smp_request(struct request_queue *q, struct Scsi_Host *shost,
 	while ((req = blk_fetch_request(q)) != NULL) {
 		spin_unlock_irq(q->queue_lock);
 
+		scsi_req(req)->resid_len = blk_rq_bytes(req);
+		if (req->next_rq)
+			scsi_req(req->next_rq)->resid_len =
+				blk_rq_bytes(req->next_rq);
 		handler = to_sas_internal(shost->transportt)->f->smp_handler;
 		ret = handler(shost, rphy, req);
 		req->errors = ret;
diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c
index b193304..1f80e31 100644
--- a/drivers/scsi/sd.c
+++ b/drivers/scsi/sd.c
@@ -781,7 +781,7 @@ static int sd_setup_discard_cmnd(struct scsi_cmnd *cmd)
 	rq->special_vec.bv_len = len;
 
 	rq->rq_flags |= RQF_SPECIAL_PAYLOAD;
-	rq->resid_len = len;
+	scsi_req(rq)->resid_len = len;
 
 	ret = scsi_init_io(cmd);
 out:
@@ -1179,7 +1179,7 @@ static void sd_uninit_command(struct scsi_cmnd *SCpnt)
 	if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
 		__free_page(rq->special_vec.bv_page);
 
-	if (SCpnt->cmnd != rq->cmd) {
+	if (SCpnt->cmnd != scsi_req(rq)->cmd) {
 		mempool_free(SCpnt->cmnd, sd_cdb_pool);
 		SCpnt->cmnd = NULL;
 		SCpnt->cmd_len = 0;
diff --git a/drivers/scsi/sg.c b/drivers/scsi/sg.c
index dbe5b4b..226a8de 100644
--- a/drivers/scsi/sg.c
+++ b/drivers/scsi/sg.c
@@ -781,9 +781,7 @@ sg_common_write(Sg_fd * sfp, Sg_request * srp,
 	}
 	if (atomic_read(&sdp->detaching)) {
 		if (srp->bio) {
-			if (srp->rq->cmd != srp->rq->__cmd)
-				kfree(srp->rq->cmd);
-
+			scsi_req_free_cmd(scsi_req(srp->rq));
 			blk_end_request_all(srp->rq, -EIO);
 			srp->rq = NULL;
 		}
@@ -1279,6 +1277,7 @@ static void
 sg_rq_end_io(struct request *rq, int uptodate)
 {
 	struct sg_request *srp = rq->end_io_data;
+	struct scsi_request *req = scsi_req(rq);
 	Sg_device *sdp;
 	Sg_fd *sfp;
 	unsigned long iflags;
@@ -1297,9 +1296,9 @@ sg_rq_end_io(struct request *rq, int uptodate)
 	if (unlikely(atomic_read(&sdp->detaching)))
 		pr_info("%s: device detaching\n", __func__);
 
-	sense = rq->sense;
+	sense = req->sense;
 	result = rq->errors;
-	resid = rq->resid_len;
+	resid = req->resid_len;
 
 	SCSI_LOG_TIMEOUT(4, sg_printk(KERN_INFO, sdp,
 				      "sg_cmd_done: pack_id=%d, res=0x%x\n",
@@ -1333,6 +1332,10 @@ sg_rq_end_io(struct request *rq, int uptodate)
 			sdp->device->changed = 1;
 		}
 	}
+
+	if (req->sense_len)
+		memcpy(srp->sense_b, req->sense, SCSI_SENSE_BUFFERSIZE);
+
 	/* Rely on write phase to clean out srp status values, so no "else" */
 
 	/*
@@ -1342,8 +1345,7 @@ sg_rq_end_io(struct request *rq, int uptodate)
 	 * blk_rq_unmap_user() can be called from user context.
 	 */
 	srp->rq = NULL;
-	if (rq->cmd != rq->__cmd)
-		kfree(rq->cmd);
+	scsi_req_free_cmd(scsi_req(rq));
 	__blk_put_request(rq->q, rq);
 
 	write_lock_irqsave(&sfp->rq_list_lock, iflags);
@@ -1658,6 +1660,7 @@ sg_start_req(Sg_request *srp, unsigned char *cmd)
 {
 	int res;
 	struct request *rq;
+	struct scsi_request *req;
 	Sg_fd *sfp = srp->parentfp;
 	sg_io_hdr_t *hp = &srp->header;
 	int dxfer_len = (int) hp->dxfer_len;
@@ -1700,17 +1703,17 @@ sg_start_req(Sg_request *srp, unsigned char *cmd)
 		kfree(long_cmdp);
 		return PTR_ERR(rq);
 	}
+	req = scsi_req(rq);
 
-	blk_rq_set_block_pc(rq);
+	scsi_req_init(rq);
 
 	if (hp->cmd_len > BLK_MAX_CDB)
-		rq->cmd = long_cmdp;
-	memcpy(rq->cmd, cmd, hp->cmd_len);
-	rq->cmd_len = hp->cmd_len;
+		req->cmd = long_cmdp;
+	memcpy(req->cmd, cmd, hp->cmd_len);
+	req->cmd_len = hp->cmd_len;
 
 	srp->rq = rq;
 	rq->end_io_data = srp;
-	rq->sense = srp->sense_b;
 	rq->retries = SG_DEFAULT_RETRIES;
 
 	if ((dxfer_len <= 0) || (dxfer_dir == SG_DXFER_NONE))
@@ -1786,8 +1789,7 @@ sg_finish_rem_req(Sg_request *srp)
 		ret = blk_rq_unmap_user(srp->bio);
 
 	if (srp->rq) {
-		if (srp->rq->cmd != srp->rq->__cmd)
-			kfree(srp->rq->cmd);
+		scsi_req_free_cmd(scsi_req(srp->rq));
 		blk_put_request(srp->rq);
 	}
 
diff --git a/drivers/scsi/st.c b/drivers/scsi/st.c
index 5f35b86..4af9001 100644
--- a/drivers/scsi/st.c
+++ b/drivers/scsi/st.c
@@ -475,7 +475,7 @@ static void st_do_stats(struct scsi_tape *STp, struct request *req)
 	ktime_t now;
 
 	now = ktime_get();
-	if (req->cmd[0] == WRITE_6) {
+	if (scsi_req(req)->cmd[0] == WRITE_6) {
 		now = ktime_sub(now, STp->stats->write_time);
 		atomic64_add(ktime_to_ns(now), &STp->stats->tot_write_time);
 		atomic64_add(ktime_to_ns(now), &STp->stats->tot_io_time);
@@ -489,7 +489,7 @@ static void st_do_stats(struct scsi_tape *STp, struct request *req)
 		} else
 			atomic64_add(atomic_read(&STp->stats->last_write_size),
 				&STp->stats->write_byte_cnt);
-	} else if (req->cmd[0] == READ_6) {
+	} else if (scsi_req(req)->cmd[0] == READ_6) {
 		now = ktime_sub(now, STp->stats->read_time);
 		atomic64_add(ktime_to_ns(now), &STp->stats->tot_read_time);
 		atomic64_add(ktime_to_ns(now), &STp->stats->tot_io_time);
@@ -514,15 +514,18 @@ static void st_do_stats(struct scsi_tape *STp, struct request *req)
 static void st_scsi_execute_end(struct request *req, int uptodate)
 {
 	struct st_request *SRpnt = req->end_io_data;
+	struct scsi_request *rq = scsi_req(req);
 	struct scsi_tape *STp = SRpnt->stp;
 	struct bio *tmp;
 
 	STp->buffer->cmdstat.midlevel_result = SRpnt->result = req->errors;
-	STp->buffer->cmdstat.residual = req->resid_len;
+	STp->buffer->cmdstat.residual = rq->resid_len;
 
 	st_do_stats(STp, req);
 
 	tmp = SRpnt->bio;
+	if (rq->sense_len)
+		memcpy(SRpnt->sense, rq->sense, SCSI_SENSE_BUFFERSIZE);
 	if (SRpnt->waiting)
 		complete(SRpnt->waiting);
 
@@ -535,6 +538,7 @@ static int st_scsi_execute(struct st_request *SRpnt, const unsigned char *cmd,
 			   int timeout, int retries)
 {
 	struct request *req;
+	struct scsi_request *rq;
 	struct rq_map_data *mdata = &SRpnt->stp->buffer->map_data;
 	int err = 0;
 	int write = (data_direction == DMA_TO_DEVICE);
@@ -544,8 +548,8 @@ static int st_scsi_execute(struct st_request *SRpnt, const unsigned char *cmd,
 			      GFP_KERNEL);
 	if (IS_ERR(req))
 		return DRIVER_ERROR << 24;
-
-	blk_rq_set_block_pc(req);
+	rq = scsi_req(req);
+	scsi_req_init(req);
 	req->rq_flags |= RQF_QUIET;
 
 	mdata->null_mapped = 1;
@@ -571,11 +575,9 @@ static int st_scsi_execute(struct st_request *SRpnt, const unsigned char *cmd,
 	}
 
 	SRpnt->bio = req->bio;
-	req->cmd_len = COMMAND_SIZE(cmd[0]);
-	memset(req->cmd, 0, BLK_MAX_CDB);
-	memcpy(req->cmd, cmd, req->cmd_len);
-	req->sense = SRpnt->sense;
-	req->sense_len = 0;
+	rq->cmd_len = COMMAND_SIZE(cmd[0]);
+	memset(rq->cmd, 0, BLK_MAX_CDB);
+	memcpy(rq->cmd, cmd, rq->cmd_len);
 	req->timeout = timeout;
 	req->retries = retries;
 	req->end_io_data = SRpnt;
diff --git a/drivers/target/target_core_pscsi.c b/drivers/target/target_core_pscsi.c
index 04d7aa7..e52f4e1 100644
--- a/drivers/target/target_core_pscsi.c
+++ b/drivers/target/target_core_pscsi.c
@@ -1013,7 +1013,7 @@ pscsi_execute_cmd(struct se_cmd *cmd)
 		goto fail;
 	}
 
-	blk_rq_set_block_pc(req);
+	scsi_req_init(req);
 
 	if (sgl) {
 		ret = pscsi_map_sg(cmd, sgl, sgl_nents, req);
@@ -1023,10 +1023,8 @@ pscsi_execute_cmd(struct se_cmd *cmd)
 
 	req->end_io = pscsi_req_done;
 	req->end_io_data = cmd;
-	req->cmd_len = scsi_command_size(pt->pscsi_cdb);
-	req->cmd = &pt->pscsi_cdb[0];
-	req->sense = &pt->pscsi_sense[0];
-	req->sense_len = 0;
+	scsi_req(req)->cmd_len = scsi_command_size(pt->pscsi_cdb);
+	scsi_req(req)->cmd = &pt->pscsi_cdb[0];
 	if (pdv->pdv_sd->type == TYPE_DISK)
 		req->timeout = PS_TIMEOUT_DISK;
 	else
@@ -1075,7 +1073,7 @@ static void pscsi_req_done(struct request *req, int uptodate)
 	struct pscsi_plugin_task *pt = cmd->priv;
 
 	pt->pscsi_result = req->errors;
-	pt->pscsi_resid = req->resid_len;
+	pt->pscsi_resid = scsi_req(req)->resid_len;
 
 	cmd->scsi_status = status_byte(pt->pscsi_result) << 1;
 	if (cmd->scsi_status) {
@@ -1096,6 +1094,7 @@ static void pscsi_req_done(struct request *req, int uptodate)
 		break;
 	}
 
+	memcpy(pt->pscsi_sense, scsi_req(req)->sense, TRANSPORT_SENSE_BUFFER);
 	__blk_put_request(req->q, req);
 	kfree(pt);
 }
diff --git a/fs/nfsd/blocklayout.c b/fs/nfsd/blocklayout.c
index 0780ff8..930d98c 100644
--- a/fs/nfsd/blocklayout.c
+++ b/fs/nfsd/blocklayout.c
@@ -10,6 +10,7 @@
 #include <linux/nfsd/debug.h>
 #include <scsi/scsi_proto.h>
 #include <scsi/scsi_common.h>
+#include <scsi/scsi_request.h>
 
 #include "blocklayoutxdr.h"
 #include "pnfs.h"
@@ -213,6 +214,7 @@ static int nfsd4_scsi_identify_device(struct block_device *bdev,
 {
 	struct request_queue *q = bdev->bd_disk->queue;
 	struct request *rq;
+	struct scsi_request *req;
 	size_t bufflen = 252, len, id_len;
 	u8 *buf, *d, type, assoc;
 	int error;
@@ -226,18 +228,19 @@ static int nfsd4_scsi_identify_device(struct block_device *bdev,
 		error = -ENOMEM;
 		goto out_free_buf;
 	}
-	blk_rq_set_block_pc(rq);
+	req = scsi_req(rq);
+	scsi_req_init(rq);
 
 	error = blk_rq_map_kern(q, rq, buf, bufflen, GFP_KERNEL);
 	if (error)
 		goto out_put_request;
 
-	rq->cmd[0] = INQUIRY;
-	rq->cmd[1] = 1;
-	rq->cmd[2] = 0x83;
-	rq->cmd[3] = bufflen >> 8;
-	rq->cmd[4] = bufflen & 0xff;
-	rq->cmd_len = COMMAND_SIZE(INQUIRY);
+	req->cmd[0] = INQUIRY;
+	req->cmd[1] = 1;
+	req->cmd[2] = 0x83;
+	req->cmd[3] = bufflen >> 8;
+	req->cmd[4] = bufflen & 0xff;
+	req->cmd_len = COMMAND_SIZE(INQUIRY);
 
 	error = blk_execute_rq(rq->q, NULL, rq, 1);
 	if (error) {
diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
index be23859..499728d 100644
--- a/include/linux/blkdev.h
+++ b/include/linux/blkdev.h
@@ -225,17 +225,7 @@ struct request {
 	int tag;
 	int errors;
 
-	/*
-	 * when request is used as a packet command carrier
-	 */
-	unsigned char __cmd[BLK_MAX_CDB];
-	unsigned char *cmd;
-	unsigned short cmd_len;
-
 	unsigned int extra_len;	/* length of alignment and padding */
-	unsigned int sense_len;
-	unsigned int resid_len;	/* residual count */
-	void *sense;
 
 	unsigned long deadline;
 	struct list_head timeout_list;
@@ -917,7 +907,6 @@ extern void blk_rq_init(struct request_queue *q, struct request *rq);
 extern void blk_put_request(struct request *);
 extern void __blk_put_request(struct request_queue *, struct request *);
 extern struct request *blk_get_request(struct request_queue *, int, gfp_t);
-extern void blk_rq_set_block_pc(struct request *);
 extern void blk_requeue_request(struct request_queue *, struct request *);
 extern int blk_lld_busy(struct request_queue *q);
 extern int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
diff --git a/include/linux/blktrace_api.h b/include/linux/blktrace_api.h
index e417f08..36767ec 100644
--- a/include/linux/blktrace_api.h
+++ b/include/linux/blktrace_api.h
@@ -112,9 +112,10 @@ struct compat_blk_user_trace_setup {
 
 #if defined(CONFIG_EVENT_TRACING) && defined(CONFIG_BLOCK)
 
+int scsi_cmd_buf_len(struct request *rq);
 static inline int blk_cmd_buf_len(struct request *rq)
 {
-	return (rq->cmd_type == REQ_TYPE_BLOCK_PC) ? rq->cmd_len * 3 : 1;
+	return (rq->cmd_type == REQ_TYPE_BLOCK_PC) ? scsi_cmd_buf_len(rq) : 1;
 }
 
 extern void blk_dump_cmd(char *buf, struct request *rq);
diff --git a/include/linux/ide.h b/include/linux/ide.h
index a633898..086fbe1 100644
--- a/include/linux/ide.h
+++ b/include/linux/ide.h
@@ -20,6 +20,7 @@
 #include <linux/mutex.h>
 /* for request_sense */
 #include <linux/cdrom.h>
+#include <scsi/scsi_cmnd.h>
 #include <asm/byteorder.h>
 #include <asm/io.h>
 
@@ -52,6 +53,11 @@ enum ata_cmd_type_bits {
 	((rq)->cmd_type == REQ_TYPE_ATA_PM_SUSPEND || \
 	 (rq)->cmd_type == REQ_TYPE_ATA_PM_RESUME)
 
+struct ide_request {
+	struct scsi_request sreq;
+	u8 sense[SCSI_SENSE_BUFFERSIZE];
+};
+
 /* Error codes returned in rq->errors to the higher part of the driver. */
 enum {
 	IDE_DRV_ERROR_GENERAL	= 101,
@@ -579,7 +585,7 @@ struct ide_drive_s {
 
 	/* current sense rq and buffer */
 	bool sense_rq_armed;
-	struct request sense_rq;
+	struct request *sense_rq;
 	struct request_sense sense_data;
 };
 
diff --git a/include/scsi/scsi_cmnd.h b/include/scsi/scsi_cmnd.h
index 9fc1aec..f708f1a 100644
--- a/include/scsi/scsi_cmnd.h
+++ b/include/scsi/scsi_cmnd.h
@@ -8,6 +8,7 @@
 #include <linux/timer.h>
 #include <linux/scatterlist.h>
 #include <scsi/scsi_device.h>
+#include <scsi/scsi_request.h>
 
 struct Scsi_Host;
 struct scsi_driver;
@@ -57,6 +58,7 @@ struct scsi_pointer {
 #define SCMD_TAGGED		(1 << 0)
 
 struct scsi_cmnd {
+	struct scsi_request req;
 	struct scsi_device *device;
 	struct list_head list;  /* scsi_cmnd participates in queue lists */
 	struct list_head eh_entry; /* entry for the host eh_cmd_q */
diff --git a/include/scsi/scsi_request.h b/include/scsi/scsi_request.h
new file mode 100644
index 0000000..c4d86c6
--- /dev/null
+++ b/include/scsi/scsi_request.h
@@ -0,0 +1,28 @@
+#ifndef _SCSI_SCSI_REQUEST_H
+#define _SCSI_SCSI_REQUEST_H
+
+#include <linux/blk-mq.h>
+
+struct scsi_request {
+	unsigned char	__cmd[BLK_MAX_CDB];
+	unsigned char	*cmd;
+	unsigned short	cmd_len;
+	unsigned int	sense_len;
+	unsigned int	resid_len;	/* residual count */
+	void		*sense;
+};
+
+static inline struct scsi_request *scsi_req(struct request *rq)
+{
+	return blk_mq_rq_to_pdu(rq);
+}
+
+void scsi_req_init(struct request *);
+
+static inline void scsi_req_free_cmd(struct scsi_request *req)
+{
+	if (req->cmd != req->__cmd)
+		kfree(req->cmd);
+}
+
+#endif /* _SCSI_SCSI_REQUEST_H */
diff --git a/kernel/trace/blktrace.c b/kernel/trace/blktrace.c
index 95cecbf..69c36e6a 100644
--- a/kernel/trace/blktrace.c
+++ b/kernel/trace/blktrace.c
@@ -27,6 +27,7 @@
 #include <linux/time.h>
 #include <linux/uaccess.h>
 #include <linux/list.h>
+#include <scsi/scsi_request.h>
 
 #include <trace/events/block.h>
 
@@ -715,7 +716,8 @@ static void blk_add_trace_rq(struct request_queue *q, struct request *rq,
 	if (rq->cmd_type == REQ_TYPE_BLOCK_PC) {
 		what |= BLK_TC_ACT(BLK_TC_PC);
 		__blk_add_trace(bt, 0, nr_bytes, req_op(rq), rq->cmd_flags,
-				what, rq->errors, rq->cmd_len, rq->cmd);
+				what, rq->errors,
+				scsi_req(rq)->cmd_len, scsi_req(rq)->cmd);
 	} else  {
 		what |= BLK_TC_ACT(BLK_TC_FS);
 		__blk_add_trace(bt, blk_rq_pos(rq), nr_bytes, req_op(rq),
@@ -1751,32 +1753,32 @@ void blk_trace_remove_sysfs(struct device *dev)
 #endif /* CONFIG_BLK_DEV_IO_TRACE */
 
 #ifdef CONFIG_EVENT_TRACING
-
-void blk_dump_cmd(char *buf, struct request *rq)
+static void blk_dump_scsi_cmd(char *buf, struct scsi_request *req)
 {
 	int i, end;
-	int len = rq->cmd_len;
-	unsigned char *cmd = rq->cmd;
 
-	if (rq->cmd_type != REQ_TYPE_BLOCK_PC) {
-		buf[0] = '\0';
-		return;
-	}
-
-	for (end = len - 1; end >= 0; end--)
-		if (cmd[end])
+	for (end = req->cmd_len - 1; end >= 0; end--)
+		if (req->cmd[end])
 			break;
 	end++;
 
-	for (i = 0; i < len; i++) {
-		buf += sprintf(buf, "%s%02x", i == 0 ? "" : " ", cmd[i]);
-		if (i == end && end != len - 1) {
+	for (i = 0; i < req->cmd_len; i++) {
+		buf += sprintf(buf, "%s%02x", i == 0 ? "" : " ", req->cmd[i]);
+		if (i == end && end != req->cmd_len - 1) {
 			sprintf(buf, " ..");
 			break;
 		}
 	}
 }
 
+void blk_dump_cmd(char *buf, struct request *rq)
+{
+	if (rq->cmd_type == REQ_TYPE_BLOCK_PC)
+		blk_dump_scsi_cmd(buf, scsi_req(rq));
+	else
+		buf[0] = '\0';
+}
+
 void blk_fill_rwbs(char *rwbs, unsigned int op, int bytes)
 {
 	int i = 0;
-- 
2.1.4

Re: [PATCH 15/15] block: split scsi_request out of struct request

[Jens Axboe]

On Tue, Jan 10 2017, Christoph Hellwig wrote:
> And require all drivers that want to support BLOCK_PC to allocate it
> as the first thing of their private data.  To support this the legacy
> IDE and BSG code is switched to set cmd_size on their queues to let
> the block layer allocate the additional space.

> diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
> index be23859..499728d 100644
> --- a/include/linux/blkdev.h
> +++ b/include/linux/blkdev.h
> @@ -225,17 +225,7 @@ struct request {
>  	int tag;
>  	int errors;
>  
> -	/*
> -	 * when request is used as a packet command carrier
> -	 */
> -	unsigned char __cmd[BLK_MAX_CDB];
> -	unsigned char *cmd;
> -	unsigned short cmd_len;
> -
>  	unsigned int extra_len;	/* length of alignment and padding */
> -	unsigned int sense_len;
> -	unsigned int resid_len;	/* residual count */
> -	void *sense;

Oh yes, my mouth is watering. I've wanted to kill off ->cmd and friends
with fire for years.

On the block front, I think the solution is fine with mimicking the mq
solution of carrying an rq payload, and the implementation for the
legacy path looks fine.

-- 
Jens Axboe

Re: RFC: split scsi passthrough fields out of struct request

[Mike Snitzer]

On Tue, Jan 10 2017 at 10:06am -0500,
Christoph Hellwig <hch@lst.de> wrote:

> Hi all,
> 
> this series splits the support for SCSI passthrough commands from the
> main struct request used all over the block layer into a separate
> scsi_request structure that drivers that want to support SCSI passthough
> need to embedded as the first thing into their request-private data,
> similar to how we handle NVMe passthrough commands.
> 
> To support this I've added support for that the private data after
> request structure to the legacy request path instead, so that it can
> be treated the same way as the blk-mq path.  Compare to the current
> scsi_cmnd allocator that actually is a major simplification.
> 
> There is one major show-stopper for the series as-is:  Request-based
> device mapper currently allocate the request structures in the stacking
> driver without a knowledge of the queue it's going to be submitte on
> for the legacy request path.  It managed to avoid that issue for the
> blk-mq path, but that currently can't be used on legacy request devices
> for reasons I don't fully understand.  We'll need to figure out how
> sort this out, but maybe that's a good opportunity to drop one or two
> of the three different dm-mpath I/O paths? :)

Ah, should've read the 0th patch earlier..

I removed blk-mq on request_fn paths support because it was one of the
permutations that I felt least useful/stable (see commit c5248f79f3 "dm:
remove support for stacking dm-mq on .request_fn device(s)")

As for all of the different IO paths.  I've always liked the idea of
blk-mq ruling the world.  With Jens' blk-mq IO scheduling advances maybe
we're closer!

Re: RFC: split scsi passthrough fields out of struct request

[Christoph Hellwig]

On Wed, Jan 11, 2017 at 05:41:42PM -0500, Mike Snitzer wrote:
> I removed blk-mq on request_fn paths support because it was one of the
> permutations that I felt least useful/stable (see commit c5248f79f3 "dm:
> remove support for stacking dm-mq on .request_fn device(s)")
> 
> As for all of the different IO paths.  I've always liked the idea of
> blk-mq ruling the world.  With Jens' blk-mq IO scheduling advances maybe
> we're closer!

That removed blk-mq on top of request_fn code would do the right thing
for this series if it entirely replaced the old request_fn code in dm-mpath.
(as would the new bio code for that matter)