block: Introduce new bio_split()

The new bio_split() can split arbitrary bios - it's not restricted to
single page bios, like the old bio_split() (previously renamed to
bio_pair_split()). It also has different semantics - it doesn't allocate
a struct bio_pair, leaving it up to the caller to handle completions.

Then convert the existing bio_pair_split() users to the new bio_split()
- and also nvme, which was open coding bio splitting.

(We have to take that BUG_ON() out of bio_integrity_trim() because this
bio_split() needs to use it, and there's no reason it has to be used on
bios marked as cloned; BIO_CLONED doesn't seem to have clearly
documented semantics anyways.)

Signed-off-by: Kent Overstreet <kmo@daterainc.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Martin K. Petersen <martin.petersen@oracle.com>
Cc: Matthew Wilcox <matthew.r.wilcox@intel.com>
Cc: Keith Busch <keith.busch@intel.com>
Cc: Vishal Verma <vishal.l.verma@intel.com>
Cc: Jiri Kosina <jkosina@suse.cz>
Cc: Neil Brown <neilb@suse.de>

2 files changed, 83 insertions, 159 deletions

diff --git a/drivers/block/nvme-core.c b/drivers/block/nvme-core.c
index 5539d29..1f14ac4 100644
--- a/drivers/block/nvme-core.c
+++ b/drivers/block/nvme-core.c
@@ -441,104 +441,19 @@ int nvme_setup_prps(struct nvme_dev *dev, struct nvme_common_command *cmd,
 	return total_len;
 }
 
-struct nvme_bio_pair {
-	struct bio b1, b2, *parent;
-	struct bio_vec *bv1, *bv2;
-	int err;
-	atomic_t cnt;
-};
-
-static void nvme_bio_pair_endio(struct bio *bio, int err)
-{
-	struct nvme_bio_pair *bp = bio->bi_private;
-
-	if (err)
-		bp->err = err;
-
-	if (atomic_dec_and_test(&bp->cnt)) {
-		bio_endio(bp->parent, bp->err);
-		kfree(bp->bv1);
-		kfree(bp->bv2);
-		kfree(bp);
-	}
-}
-
-static struct nvme_bio_pair *nvme_bio_split(struct bio *bio, int idx,
-							int len, int offset)
-{
-	struct nvme_bio_pair *bp;
-
-	BUG_ON(len > bio->bi_iter.bi_size);
-	BUG_ON(idx > bio->bi_vcnt);
-
-	bp = kmalloc(sizeof(*bp), GFP_ATOMIC);
-	if (!bp)
-		return NULL;
-	bp->err = 0;
-
-	bp->b1 = *bio;
-	bp->b2 = *bio;
-
-	bp->b1.bi_iter.bi_size = len;
-	bp->b2.bi_iter.bi_size -= len;
-	bp->b1.bi_vcnt = idx;
-	bp->b2.bi_iter.bi_idx = idx;
-	bp->b2.bi_iter.bi_sector += len >> 9;
-
-	if (offset) {
-		bp->bv1 = kmalloc(bio->bi_max_vecs * sizeof(struct bio_vec),
-								GFP_ATOMIC);
-		if (!bp->bv1)
-			goto split_fail_1;
-
-		bp->bv2 = kmalloc(bio->bi_max_vecs * sizeof(struct bio_vec),
-								GFP_ATOMIC);
-		if (!bp->bv2)
-			goto split_fail_2;
-
-		memcpy(bp->bv1, bio->bi_io_vec,
-			bio->bi_max_vecs * sizeof(struct bio_vec));
-		memcpy(bp->bv2, bio->bi_io_vec,
-			bio->bi_max_vecs * sizeof(struct bio_vec));
-
-		bp->b1.bi_io_vec = bp->bv1;
-		bp->b2.bi_io_vec = bp->bv2;
-		bp->b2.bi_io_vec[idx].bv_offset += offset;
-		bp->b2.bi_io_vec[idx].bv_len -= offset;
-		bp->b1.bi_io_vec[idx].bv_len = offset;
-		bp->b1.bi_vcnt++;
-	} else
-		bp->bv1 = bp->bv2 = NULL;
-
-	bp->b1.bi_private = bp;
-	bp->b2.bi_private = bp;
-
-	bp->b1.bi_end_io = nvme_bio_pair_endio;
-	bp->b2.bi_end_io = nvme_bio_pair_endio;
-
-	bp->parent = bio;
-	atomic_set(&bp->cnt, 2);
-
-	return bp;
-
- split_fail_2:
-	kfree(bp->bv1);
- split_fail_1:
-	kfree(bp);
-	return NULL;
-}
-
 static int nvme_split_and_submit(struct bio *bio, struct nvme_queue *nvmeq,
-						int idx, int len, int offset)
+				 int len)
 {
-	struct nvme_bio_pair *bp = nvme_bio_split(bio, idx, len, offset);
-	if (!bp)
+	struct bio *split = bio_split(bio, len >> 9, GFP_ATOMIC, NULL);
+	if (!split)
 		return -ENOMEM;
 
+	bio_chain(split, bio);
+
 	if (bio_list_empty(&nvmeq->sq_cong))
 		add_wait_queue(&nvmeq->sq_full, &nvmeq->sq_cong_wait);
-	bio_list_add(&nvmeq->sq_cong, &bp->b1);
-	bio_list_add(&nvmeq->sq_cong, &bp->b2);
+	bio_list_add(&nvmeq->sq_cong, split);
+	bio_list_add(&nvmeq->sq_cong, bio);
 
 	return 0;
 }
@@ -568,8 +483,7 @@ static int nvme_map_bio(struct nvme_queue *nvmeq, struct nvme_iod *iod,
 		} else {
 			if (!first && BIOVEC_NOT_VIRT_MERGEABLE(&bvprv, &bvec))
 				return nvme_split_and_submit(bio, nvmeq,
-							     iter.bi_idx,
-							     length, 0);
+							     length);
 
 			sg = sg ? sg + 1 : iod->sg;
 			sg_set_page(sg, bvec.bv_page,
@@ -578,9 +492,7 @@ static int nvme_map_bio(struct nvme_queue *nvmeq, struct nvme_iod *iod,
 		}
 
 		if (split_len - length < bvec.bv_len)
-			return nvme_split_and_submit(bio, nvmeq, iter.bi_idx,
-						     split_len,
-						     split_len - length);
+			return nvme_split_and_submit(bio, nvmeq, split_len);
 		length += bvec.bv_len;
 		bvprv = bvec;
 		first = 0;
diff --git a/drivers/block/pktcdvd.c b/drivers/block/pktcdvd.c
index 28789b8..3dda09a 100644
--- a/drivers/block/pktcdvd.c
+++ b/drivers/block/pktcdvd.c
@@ -2338,75 +2338,29 @@ static void pkt_end_io_read_cloned(struct bio *bio, int err)
 	pkt_bio_finished(pd);
 }
 
-static void pkt_make_request(struct request_queue *q, struct bio *bio)
+static void pkt_make_request_read(struct pktcdvd_device *pd, struct bio *bio)
 {
-	struct pktcdvd_device *pd;
-	char b[BDEVNAME_SIZE];
+	struct bio *cloned_bio = bio_clone(bio, GFP_NOIO);
+	struct packet_stacked_data *psd = mempool_alloc(psd_pool, GFP_NOIO);
+
+	psd->pd = pd;
+	psd->bio = bio;
+	cloned_bio->bi_bdev = pd->bdev;
+	cloned_bio->bi_private = psd;
+	cloned_bio->bi_end_io = pkt_end_io_read_cloned;
+	pd->stats.secs_r += bio_sectors(bio);
+	pkt_queue_bio(pd, cloned_bio);
+}
+
+static void pkt_make_request_write(struct request_queue *q, struct bio *bio)
+{
+	struct pktcdvd_device *pd = q->queuedata;
 	sector_t zone;
 	struct packet_data *pkt;
 	int was_empty, blocked_bio;
 	struct pkt_rb_node *node;
 
-	pd = q->queuedata;
-	if (!pd) {
-		pr_err("%s incorrect request queue\n",
-		       bdevname(bio->bi_bdev, b));
-		goto end_io;
-	}
-
-	/*
-	 * Clone READ bios so we can have our own bi_end_io callback.
-	 */
-	if (bio_data_dir(bio) == READ) {
-		struct bio *cloned_bio = bio_clone(bio, GFP_NOIO);
-		struct packet_stacked_data *psd = mempool_alloc(psd_pool, GFP_NOIO);
-
-		psd->pd = pd;
-		psd->bio = bio;
-		cloned_bio->bi_bdev = pd->bdev;
-		cloned_bio->bi_private = psd;
-		cloned_bio->bi_end_io = pkt_end_io_read_cloned;
-		pd->stats.secs_r += bio_sectors(bio);
-		pkt_queue_bio(pd, cloned_bio);
-		return;
-	}
-
-	if (!test_bit(PACKET_WRITABLE, &pd->flags)) {
-		pkt_notice(pd, "WRITE for ro device (%llu)\n",
-			   (unsigned long long)bio->bi_iter.bi_sector);
-		goto end_io;
-	}
-
-	if (!bio->bi_iter.bi_size || (bio->bi_iter.bi_size % CD_FRAMESIZE)) {
-		pkt_err(pd, "wrong bio size\n");
-		goto end_io;
-	}
-
-	blk_queue_bounce(q, &bio);
-
 	zone = get_zone(bio->bi_iter.bi_sector, pd);
-	pkt_dbg(2, pd, "start = %6llx stop = %6llx\n",
-		(unsigned long long)bio->bi_iter.bi_sector,
-		(unsigned long long)bio_end_sector(bio));
-
-	/* Check if we have to split the bio */
-	{
-		struct bio_pair *bp;
-		sector_t last_zone;
-		int first_sectors;
-
-		last_zone = get_zone(bio_end_sector(bio) - 1, pd);
-		if (last_zone != zone) {
-			BUG_ON(last_zone != zone + pd->settings.size);
-			first_sectors = last_zone - bio->bi_iter.bi_sector;
-			bp = bio_pair_split(bio, first_sectors);
-			BUG_ON(!bp);
-			pkt_make_request(q, &bp->bio1);
-			pkt_make_request(q, &bp->bio2);
-			bio_pair_release(bp);
-			return;
-		}
-	}
 
 	/*
 	 * If we find a matching packet in state WAITING or READ_WAIT, we can
@@ -2480,6 +2434,64 @@ static void pkt_make_request(struct request_queue *q, struct bio *bio)
 		 */
 		wake_up(&pd->wqueue);
 	}
+}
+
+static void pkt_make_request(struct request_queue *q, struct bio *bio)
+{
+	struct pktcdvd_device *pd;
+	char b[BDEVNAME_SIZE];
+	struct bio *split;
+
+	pd = q->queuedata;
+	if (!pd) {
+		pr_err("%s incorrect request queue\n",
+		       bdevname(bio->bi_bdev, b));
+		goto end_io;
+	}
+
+	pkt_dbg(2, pd, "start = %6llx stop = %6llx\n",
+		(unsigned long long)bio->bi_iter.bi_sector,
+		(unsigned long long)bio_end_sector(bio));
+
+	/*
+	 * Clone READ bios so we can have our own bi_end_io callback.
+	 */
+	if (bio_data_dir(bio) == READ) {
+		pkt_make_request_read(pd, bio);
+		return;
+	}
+
+	if (!test_bit(PACKET_WRITABLE, &pd->flags)) {
+		pkt_notice(pd, "WRITE for ro device (%llu)\n",
+			   (unsigned long long)bio->bi_iter.bi_sector);
+		goto end_io;
+	}
+
+	if (!bio->bi_iter.bi_size || (bio->bi_iter.bi_size % CD_FRAMESIZE)) {
+		pkt_err(pd, "wrong bio size\n");
+		goto end_io;
+	}
+
+	blk_queue_bounce(q, &bio);
+
+	do {
+		sector_t zone = get_zone(bio->bi_iter.bi_sector, pd);
+		sector_t last_zone = get_zone(bio_end_sector(bio) - 1, pd);
+
+		if (last_zone != zone) {
+			BUG_ON(last_zone != zone + pd->settings.size);
+
+			split = bio_split(bio, last_zone -
+					  bio->bi_iter.bi_sector,
+					  GFP_NOIO, fs_bio_set);
+			bio_chain(split, bio);
+		} else {
+			split = bio;
+		}
+
+		pkt_make_request_write(q, split);
+	} while (split != bio);
+
 	return;
 end_io:
 	bio_io_error(bio);