switch block layer polling to a bio based model

switch block layer polling to a bio based model

[Christoph Hellwig]

Hi all,

This series clean up the block polling code a bit and changes the interface
to poll for a specific bio instead of a request_queue and cookie pair.

Polling for the bio itself leads to a few advantages:

  - the cookie construction can made entirely private in blk-mq.c
  - the caller does not need to remember the request_queue and cookie
    separately and thus sidesteps their lifetime issues
  - keeping the device and the cookie inside the bio allows to trivially
    support polling BIOs remapping by stacking drivers
  - a lot of code to propagate the cookie back up the submission path can
    removed entirely

The one major caveat is that this requires RCU freeing polled BIOs to make
sure the bio that contains the polling information is still alive when
io_uring tries to poll it through the iocb. For synchronous polling all the
callers have a bio reference anyway, so this is not an issue.

Git tree:

   git://git.infradead.org/users/hch/block.git bio-poll

Gitweb:

    http://git.infradead.org/users/hch/block.git/shortlog/refs/heads/bio-poll

Diffstat:
 arch/m68k/emu/nfblock.c             |    3 
 arch/xtensa/platforms/iss/simdisk.c |    3 
 block/bio.c                         |   29 ++++--
 block/blk-core.c                    |  105 ++++++++++++++++-------
 block/blk-merge.c                   |    2 
 block/blk-mq-debugfs.c              |    2 
 block/blk-mq.c                      |  163 ++++++++++++++----------------------
 block/blk-mq.h                      |    6 -
 block/blk-wbt.c                     |    4 
 drivers/block/brd.c                 |   12 +-
 drivers/block/drbd/drbd_int.h       |    2 
 drivers/block/drbd/drbd_req.c       |    3 
 drivers/block/n64cart.c             |   12 +-
 drivers/block/null_blk/main.c       |    3 
 drivers/block/pktcdvd.c             |    7 -
 drivers/block/ps3vram.c             |    6 -
 drivers/block/rsxx/dev.c            |    7 -
 drivers/block/umem.c                |    4 
 drivers/block/zram/zram_drv.c       |   10 --
 drivers/lightnvm/pblk-init.c        |    6 -
 drivers/md/bcache/request.c         |   13 +-
 drivers/md/bcache/request.h         |    4 
 drivers/md/dm.c                     |   28 ++----
 drivers/md/md.c                     |   10 --
 drivers/nvdimm/blk.c                |    5 -
 drivers/nvdimm/btt.c                |    5 -
 drivers/nvdimm/pmem.c               |    3 
 drivers/nvme/host/core.c            |    4 
 drivers/nvme/host/multipath.c       |    6 -
 drivers/nvme/host/nvme.h            |    2 
 drivers/s390/block/dcssblk.c        |    7 -
 drivers/s390/block/xpram.c          |    5 -
 fs/block_dev.c                      |   54 +++--------
 fs/btrfs/inode.c                    |    8 -
 fs/direct-io.c                      |   14 ---
 fs/ext4/file.c                      |    2 
 fs/gfs2/file.c                      |    4 
 fs/iomap/direct-io.c                |   56 ++++++------
 fs/xfs/xfs_file.c                   |    2 
 fs/zonefs/super.c                   |    2 
 include/linux/bio.h                 |    4 
 include/linux/blk-mq.h              |   15 ---
 include/linux/blk_types.h           |   40 ++------
 include/linux/blkdev.h              |    8 +
 include/linux/bvec.h                |    2 
 include/linux/fs.h                  |    6 -
 include/linux/iomap.h               |    3 
 mm/page_io.c                        |   10 --
 48 files changed, 318 insertions(+), 393 deletions(-)

[PATCH 01/12] direct-io: remove blk_poll support

[Christoph Hellwig]

The polling support in the legacy direct-io support is a little crufty.
It already doesn't support the asynchronous polling needed for io_uring
polling, and is hard to adopt to upcoming changes in the polling
interfaces.  Given that all the major file systems already use the iomap
direct I/O code, just drop the polling support.

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

diff --git a/fs/direct-io.c b/fs/direct-io.c
index b61491bf3166..237701c7e132 100644
--- a/fs/direct-io.c
+++ b/fs/direct-io.c
@@ -119,7 +119,6 @@ struct dio {
 	int flags;			/* doesn't change */
 	int op;
 	int op_flags;
-	blk_qc_t bio_cookie;
 	struct gendisk *bio_disk;
 	struct inode *inode;
 	loff_t i_size;			/* i_size when submitted */
@@ -438,11 +437,10 @@ static inline void dio_bio_submit(struct dio *dio, struct dio_submit *sdio)
 
 	dio->bio_disk = bio->bi_bdev->bd_disk;
 
-	if (sdio->submit_io) {
+	if (sdio->submit_io)
 		sdio->submit_io(bio, dio->inode, sdio->logical_offset_in_bio);
-		dio->bio_cookie = BLK_QC_T_NONE;
-	} else
-		dio->bio_cookie = submit_bio(bio);
+	else
+		submit_bio(bio);
 
 	sdio->bio = NULL;
 	sdio->boundary = 0;
@@ -481,9 +479,7 @@ static struct bio *dio_await_one(struct dio *dio)
 		__set_current_state(TASK_UNINTERRUPTIBLE);
 		dio->waiter = current;
 		spin_unlock_irqrestore(&dio->bio_lock, flags);
-		if (!(dio->iocb->ki_flags & IOCB_HIPRI) ||
-		    !blk_poll(dio->bio_disk->queue, dio->bio_cookie, true))
-			blk_io_schedule();
+		blk_io_schedule();
 		/* wake up sets us TASK_RUNNING */
 		spin_lock_irqsave(&dio->bio_lock, flags);
 		dio->waiter = NULL;
@@ -1213,8 +1209,6 @@ do_blockdev_direct_IO(struct kiocb *iocb, struct inode *inode,
 	} else {
 		dio->op = REQ_OP_READ;
 	}
-	if (iocb->ki_flags & IOCB_HIPRI)
-		dio->op_flags |= REQ_HIPRI;
 
 	/*
 	 * For AIO O_(D)SYNC writes we need to defer completions to a workqueue
-- 
2.30.1

[PATCH 02/12] block: don't try to poll multi-bio I/Os in __blkdev_direct_IO

[Christoph Hellwig]

If an iocb is split into multiple bios we can't poll for both.  So don't
bother to even try to poll in that case.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 fs/block_dev.c | 37 ++++++++++++++-----------------------
 1 file changed, 14 insertions(+), 23 deletions(-)

diff --git a/fs/block_dev.c b/fs/block_dev.c
index 363015fcffdb..ea5b4617ff86 100644
--- a/fs/block_dev.c
+++ b/fs/block_dev.c
@@ -373,7 +373,7 @@ static ssize_t __blkdev_direct_IO(struct kiocb *iocb, struct iov_iter *iter,
 	struct blk_plug plug;
 	struct blkdev_dio *dio;
 	struct bio *bio;
-	bool is_poll = (iocb->ki_flags & IOCB_HIPRI) != 0;
+	bool is_poll = (iocb->ki_flags & IOCB_HIPRI), do_poll = false;
 	bool is_read = (iov_iter_rw(iter) == READ), is_sync;
 	loff_t pos = iocb->ki_pos;
 	blk_qc_t qc = BLK_QC_T_NONE;
@@ -433,22 +433,9 @@ static ssize_t __blkdev_direct_IO(struct kiocb *iocb, struct iov_iter *iter,
 		pos += bio->bi_iter.bi_size;
 
 		nr_pages = bio_iov_vecs_to_alloc(iter, BIO_MAX_VECS);
-		if (!nr_pages) {
-			bool polled = false;
-
-			if (iocb->ki_flags & IOCB_HIPRI) {
-				bio_set_polled(bio, iocb);
-				polled = true;
-			}
-
-			qc = submit_bio(bio);
-
-			if (polled)
-				WRITE_ONCE(iocb->ki_cookie, qc);
-			break;
-		}
-
-		if (!dio->multi_bio) {
+		if (dio->multi_bio) {
+			atomic_inc(&dio->ref);
+		} else if (nr_pages) {
 			/*
 			 * AIO needs an extra reference to ensure the dio
 			 * structure which is embedded into the first bio
@@ -458,11 +445,16 @@ static ssize_t __blkdev_direct_IO(struct kiocb *iocb, struct iov_iter *iter,
 				bio_get(bio);
 			dio->multi_bio = true;
 			atomic_set(&dio->ref, 2);
-		} else {
-			atomic_inc(&dio->ref);
+		} else if (is_poll) {
+			bio_set_polled(bio, iocb);
+			do_poll = true;
+		}
+		qc = submit_bio(bio);
+		if (!nr_pages) {
+			if (do_poll)
+				WRITE_ONCE(iocb->ki_cookie, qc);
+			break;
 		}
-
-		submit_bio(bio);
 		bio = bio_alloc(GFP_KERNEL, nr_pages);
 	}
 
@@ -477,8 +469,7 @@ static ssize_t __blkdev_direct_IO(struct kiocb *iocb, struct iov_iter *iter,
 		if (!READ_ONCE(dio->waiter))
 			break;
 
-		if (!(iocb->ki_flags & IOCB_HIPRI) ||
-		    !blk_poll(bdev_get_queue(bdev), qc, true))
+		if (!do_poll || !blk_poll(bdev_get_queue(bdev), qc, true))
 			blk_io_schedule();
 	}
 	__set_current_state(TASK_RUNNING);
-- 
2.30.1

[PATCH 03/12] iomap: don't try to poll multi-bio I/Os in __iomap_dio_rw

[Christoph Hellwig]

If an iocb is split into multiple bios we can't poll for both.  So don't
bother to even try to poll in that case.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 fs/iomap/direct-io.c | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)

diff --git a/fs/iomap/direct-io.c b/fs/iomap/direct-io.c
index bdd0d89bbf0a..357419f39654 100644
--- a/fs/iomap/direct-io.c
+++ b/fs/iomap/direct-io.c
@@ -282,6 +282,13 @@ iomap_dio_bio_actor(struct inode *inode, loff_t pos, loff_t length,
 	if (!iov_iter_count(dio->submit.iter))
 		goto out;
 
+	/*
+	 * We can only poll for single bio I/Os.
+	 */
+	if (need_zeroout ||
+	    ((dio->flags & IOMAP_DIO_WRITE) && pos >= i_size_read(inode)))
+		dio->iocb->ki_flags &= ~IOCB_HIPRI;
+
 	if (need_zeroout) {
 		/* zero out from the start of the block to the write offset */
 		pad = pos & (fs_block_size - 1);
@@ -339,6 +346,11 @@ iomap_dio_bio_actor(struct inode *inode, loff_t pos, loff_t length,
 
 		nr_pages = bio_iov_vecs_to_alloc(dio->submit.iter,
 						 BIO_MAX_VECS);
+		/*
+		 * We can only poll for single bio I/Os.
+		 */
+		if (nr_pages)
+			dio->iocb->ki_flags &= ~IOCB_HIPRI;
 		iomap_dio_submit_bio(dio, iomap, bio, pos);
 		pos += n;
 	} while (nr_pages);
@@ -571,6 +583,11 @@ __iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
 			iov_iter_revert(iter, pos - dio->i_size);
 			break;
 		}
+
+		/*
+		 * We can only poll for single bio I/Os.
+		 */
+		iocb->ki_flags &= ~IOCB_HIPRI;
 	} while ((count = iov_iter_count(iter)) > 0);
 	blk_finish_plug(&plug);
 
-- 
2.30.1

[PATCH 04/12] blk-mq: factor out a "classic" poll helper

[Christoph Hellwig]

Factor the code to do the classic full metal polling out of blk_poll into
a separate blk_mq_poll_classic helper.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 block/blk-mq.c | 63 +++++++++++++++++++++++++++-----------------------
 1 file changed, 34 insertions(+), 29 deletions(-)

diff --git a/block/blk-mq.c b/block/blk-mq.c
index 927189a55575..4a72d3fa7964 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -3853,6 +3853,39 @@ static bool blk_mq_poll_hybrid(struct request_queue *q,
 	return blk_mq_poll_hybrid_sleep(q, rq);
 }
 
+static int blk_mq_poll_classic(struct request_queue *q, blk_qc_t qc, bool spin)
+{
+	struct blk_mq_hw_ctx *hctx = q->queue_hw_ctx[blk_qc_t_to_queue_num(qc)];
+	long state = current->state;
+
+	hctx->poll_considered++;
+
+	do {
+		int ret;
+
+		hctx->poll_invoked++;
+
+		ret = q->mq_ops->poll(hctx);
+		if (ret > 0) {
+			hctx->poll_success++;
+			__set_current_state(TASK_RUNNING);
+			return ret;
+		}
+
+		if (signal_pending_state(state, current))
+			__set_current_state(TASK_RUNNING);
+
+		if (current->state == TASK_RUNNING)
+			return 1;
+		if (ret < 0 || !spin)
+			break;
+		cpu_relax();
+	} while (!need_resched());
+
+	__set_current_state(TASK_RUNNING);
+	return 0;
+}
+
 /**
  * blk_poll - poll for IO completions
  * @q:  the queue
@@ -3868,7 +3901,6 @@ static bool blk_mq_poll_hybrid(struct request_queue *q,
 int blk_poll(struct request_queue *q, blk_qc_t cookie, bool spin)
 {
 	struct blk_mq_hw_ctx *hctx;
-	long state;
 
 	if (!blk_qc_t_valid(cookie) ||
 	    !test_bit(QUEUE_FLAG_POLL, &q->queue_flags))
@@ -3889,34 +3921,7 @@ int blk_poll(struct request_queue *q, blk_qc_t cookie, bool spin)
 	 */
 	if (spin && blk_mq_poll_hybrid(q, hctx, cookie))
 		return 1;
-
-	hctx->poll_considered++;
-
-	state = current->state;
-	do {
-		int ret;
-
-		hctx->poll_invoked++;
-
-		ret = q->mq_ops->poll(hctx);
-		if (ret > 0) {
-			hctx->poll_success++;
-			__set_current_state(TASK_RUNNING);
-			return ret;
-		}
-
-		if (signal_pending_state(state, current))
-			__set_current_state(TASK_RUNNING);
-
-		if (current->state == TASK_RUNNING)
-			return 1;
-		if (ret < 0 || !spin)
-			break;
-		cpu_relax();
-	} while (!need_resched());
-
-	__set_current_state(TASK_RUNNING);
-	return 0;
+	return blk_mq_poll_classic(q, cookie, spin);
 }
 EXPORT_SYMBOL_GPL(blk_poll);
 
-- 
2.30.1

[PATCH 05/12] blk-mq: factor out a blk_qc_to_hctx helper

[Christoph Hellwig]

Add a helper to get the hctx from a request_queue and cookie, and fold
the blk_qc_t_to_queue_num helper into it as no other callers are left.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 block/blk-mq.c            | 10 ++++++++--
 include/linux/blk_types.h |  5 -----
 2 files changed, 8 insertions(+), 7 deletions(-)

diff --git a/block/blk-mq.c b/block/blk-mq.c
index 4a72d3fa7964..3dde5e7b2251 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -63,6 +63,12 @@ static int blk_mq_poll_stats_bkt(const struct request *rq)
 	return bucket;
 }
 
+static inline struct blk_mq_hw_ctx *blk_qc_to_hctx(struct request_queue *q,
+		blk_qc_t qc)
+{
+	return q->queue_hw_ctx[(qc & ~BLK_QC_T_INTERNAL) >> BLK_QC_T_SHIFT];
+}
+
 /*
  * Check if any of the ctx, dispatch list or elevator
  * have pending work in this hardware queue.
@@ -3855,7 +3861,7 @@ static bool blk_mq_poll_hybrid(struct request_queue *q,
 
 static int blk_mq_poll_classic(struct request_queue *q, blk_qc_t qc, bool spin)
 {
-	struct blk_mq_hw_ctx *hctx = q->queue_hw_ctx[blk_qc_t_to_queue_num(qc)];
+	struct blk_mq_hw_ctx *hctx = blk_qc_to_hctx(q, qc);
 	long state = current->state;
 
 	hctx->poll_considered++;
@@ -3909,7 +3915,7 @@ int blk_poll(struct request_queue *q, blk_qc_t cookie, bool spin)
 	if (current->plug)
 		blk_flush_plug_list(current->plug, false);
 
-	hctx = q->queue_hw_ctx[blk_qc_t_to_queue_num(cookie)];
+	hctx = blk_qc_to_hctx(q, cookie);
 
 	/*
 	 * If we sleep, have the caller restart the poll loop to reset
diff --git a/include/linux/blk_types.h b/include/linux/blk_types.h
index db026b6ec15a..ed71a8d968f2 100644
--- a/include/linux/blk_types.h
+++ b/include/linux/blk_types.h
@@ -515,11 +515,6 @@ static inline bool blk_qc_t_valid(blk_qc_t cookie)
 	return cookie != BLK_QC_T_NONE;
 }
 
-static inline unsigned int blk_qc_t_to_queue_num(blk_qc_t cookie)
-{
-	return (cookie & ~BLK_QC_T_INTERNAL) >> BLK_QC_T_SHIFT;
-}
-
 static inline unsigned int blk_qc_t_to_tag(blk_qc_t cookie)
 {
 	return cookie & ((1u << BLK_QC_T_SHIFT) - 1);
-- 
2.30.1

[PATCH 06/12] blk-mq: refactor hybrid polling

[Christoph Hellwig]

Refactor the hybrid polling logic by moving the request lookup into a
helper, and the check if hybrid polling is enabled into blk_poll, and
then merging blk_mq_poll_hybrid and blk_mq_poll_hybrid_sleep into a
single function.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 block/blk-mq.c | 65 +++++++++++++++++++-------------------------------
 1 file changed, 25 insertions(+), 40 deletions(-)

diff --git a/block/blk-mq.c b/block/blk-mq.c
index 3dde5e7b2251..15a8bfd27aa2 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -69,6 +69,14 @@ static inline struct blk_mq_hw_ctx *blk_qc_to_hctx(struct request_queue *q,
 	return q->queue_hw_ctx[(qc & ~BLK_QC_T_INTERNAL) >> BLK_QC_T_SHIFT];
 }
 
+static inline struct request *blk_qc_to_rq(struct blk_mq_hw_ctx *hctx,
+		blk_qc_t qc)
+{
+	if (blk_qc_t_is_internal(qc))
+		return blk_mq_tag_to_rq(hctx->sched_tags, blk_qc_t_to_tag(qc));
+	return blk_mq_tag_to_rq(hctx->tags, blk_qc_t_to_tag(qc));
+}
+
 /*
  * Check if any of the ctx, dispatch list or elevator
  * have pending work in this hardware queue.
@@ -3781,15 +3789,20 @@ static unsigned long blk_mq_poll_nsecs(struct request_queue *q,
 	return ret;
 }
 
-static bool blk_mq_poll_hybrid_sleep(struct request_queue *q,
-				     struct request *rq)
+static bool blk_mq_poll_hybrid(struct request_queue *q, blk_qc_t qc)
 {
+	struct blk_mq_hw_ctx *hctx = blk_qc_to_hctx(q, qc);
+	struct request *rq = blk_qc_to_rq(hctx, qc);
 	struct hrtimer_sleeper hs;
 	enum hrtimer_mode mode;
 	unsigned int nsecs;
 	ktime_t kt;
 
-	if (rq->rq_flags & RQF_MQ_POLL_SLEPT)
+	/*
+	 * If a request has completed on queue that uses an I/O scheduler, we
+	 * won't get back a request from blk_qc_to_rq.
+	 */
+	if (!rq || (rq->rq_flags & RQF_MQ_POLL_SLEPT))
 		return false;
 
 	/*
@@ -3834,31 +3847,6 @@ static bool blk_mq_poll_hybrid_sleep(struct request_queue *q,
 	return true;
 }
 
-static bool blk_mq_poll_hybrid(struct request_queue *q,
-			       struct blk_mq_hw_ctx *hctx, blk_qc_t cookie)
-{
-	struct request *rq;
-
-	if (q->poll_nsec == BLK_MQ_POLL_CLASSIC)
-		return false;
-
-	if (!blk_qc_t_is_internal(cookie))
-		rq = blk_mq_tag_to_rq(hctx->tags, blk_qc_t_to_tag(cookie));
-	else {
-		rq = blk_mq_tag_to_rq(hctx->sched_tags, blk_qc_t_to_tag(cookie));
-		/*
-		 * With scheduling, if the request has completed, we'll
-		 * get a NULL return here, as we clear the sched tag when
-		 * that happens. The request still remains valid, like always,
-		 * so we should be safe with just the NULL check.
-		 */
-		if (!rq)
-			return false;
-	}
-
-	return blk_mq_poll_hybrid_sleep(q, rq);
-}
-
 static int blk_mq_poll_classic(struct request_queue *q, blk_qc_t qc, bool spin)
 {
 	struct blk_mq_hw_ctx *hctx = blk_qc_to_hctx(q, qc);
@@ -3906,8 +3894,6 @@ static int blk_mq_poll_classic(struct request_queue *q, blk_qc_t qc, bool spin)
  */
 int blk_poll(struct request_queue *q, blk_qc_t cookie, bool spin)
 {
-	struct blk_mq_hw_ctx *hctx;
-
 	if (!blk_qc_t_valid(cookie) ||
 	    !test_bit(QUEUE_FLAG_POLL, &q->queue_flags))
 		return 0;
@@ -3915,18 +3901,17 @@ int blk_poll(struct request_queue *q, blk_qc_t cookie, bool spin)
 	if (current->plug)
 		blk_flush_plug_list(current->plug, false);
 
-	hctx = blk_qc_to_hctx(q, cookie);
-
 	/*
-	 * If we sleep, have the caller restart the poll loop to reset
-	 * the state. Like for the other success return cases, the
-	 * caller is responsible for checking if the IO completed. If
-	 * the IO isn't complete, we'll get called again and will go
-	 * straight to the busy poll loop. If specified not to spin,
-	 * we also should not sleep.
+	 * If we sleep, have the caller restart the poll loop to reset the
+	 * state.  Like for the other success return cases, the caller is
+	 * responsible for checking if the IO completed.  If the IO isn't
+	 * complete, we'll get called again and will go straight to the busy
+	 * poll loop. If specified not to spin, we also should not sleep.
 	 */
-	if (spin && blk_mq_poll_hybrid(q, hctx, cookie))
-		return 1;
+	if (spin && q->poll_nsec != BLK_MQ_POLL_CLASSIC) {
+		if (blk_mq_poll_hybrid(q, cookie))
+			return 1;
+	}
 	return blk_mq_poll_classic(q, cookie, spin);
 }
 EXPORT_SYMBOL_GPL(blk_poll);
-- 
2.30.1

[PATCH 07/12] blk-mq: remove blk_qc_t_to_tag and blk_qc_t_is_internal

[Christoph Hellwig]

Merge both functions into their only caller to keep the blk-mq tag to
blk_qc_t mapping as private as possible in blk-mq.c.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 block/blk-mq.c            |  8 +++++---
 include/linux/blk_types.h | 10 ----------
 2 files changed, 5 insertions(+), 13 deletions(-)

diff --git a/block/blk-mq.c b/block/blk-mq.c
index 15a8bfd27aa2..846e1c51128d 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -72,9 +72,11 @@ static inline struct blk_mq_hw_ctx *blk_qc_to_hctx(struct request_queue *q,
 static inline struct request *blk_qc_to_rq(struct blk_mq_hw_ctx *hctx,
 		blk_qc_t qc)
 {
-	if (blk_qc_t_is_internal(qc))
-		return blk_mq_tag_to_rq(hctx->sched_tags, blk_qc_t_to_tag(qc));
-	return blk_mq_tag_to_rq(hctx->tags, blk_qc_t_to_tag(qc));
+	unsigned int tag = qc & ((1U << BLK_QC_T_SHIFT) - 1);
+
+	if (qc & BLK_QC_T_INTERNAL)
+		return blk_mq_tag_to_rq(hctx->sched_tags, tag);
+	return blk_mq_tag_to_rq(hctx->tags, tag);
 }
 
 /*
diff --git a/include/linux/blk_types.h b/include/linux/blk_types.h
index ed71a8d968f2..4c28ab22f93d 100644
--- a/include/linux/blk_types.h
+++ b/include/linux/blk_types.h
@@ -515,16 +515,6 @@ static inline bool blk_qc_t_valid(blk_qc_t cookie)
 	return cookie != BLK_QC_T_NONE;
 }
 
-static inline unsigned int blk_qc_t_to_tag(blk_qc_t cookie)
-{
-	return cookie & ((1u << BLK_QC_T_SHIFT) - 1);
-}
-
-static inline bool blk_qc_t_is_internal(blk_qc_t cookie)
-{
-	return (cookie & BLK_QC_T_INTERNAL) != 0;
-}
-
 struct blk_rq_stat {
 	u64 mean;
 	u64 min;
-- 
2.30.1

[PATCH 08/12] blk-mq: remove blk_qc_t_valid

[Christoph Hellwig]

Move the trivial check into the only caller.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 block/blk-mq.c            | 2 +-
 include/linux/blk_types.h | 5 -----
 2 files changed, 1 insertion(+), 6 deletions(-)

diff --git a/block/blk-mq.c b/block/blk-mq.c
index 846e1c51128d..40b20f7986c1 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -3896,7 +3896,7 @@ static int blk_mq_poll_classic(struct request_queue *q, blk_qc_t qc, bool spin)
  */
 int blk_poll(struct request_queue *q, blk_qc_t cookie, bool spin)
 {
-	if (!blk_qc_t_valid(cookie) ||
+	if (cookie == BLK_QC_T_NONE ||
 	    !test_bit(QUEUE_FLAG_POLL, &q->queue_flags))
 		return 0;
 
diff --git a/include/linux/blk_types.h b/include/linux/blk_types.h
index 4c28ab22f93d..d0cf835d3b50 100644
--- a/include/linux/blk_types.h
+++ b/include/linux/blk_types.h
@@ -510,11 +510,6 @@ typedef unsigned int blk_qc_t;
 #define BLK_QC_T_SHIFT		16
 #define BLK_QC_T_INTERNAL	(1U << 31)
 
-static inline bool blk_qc_t_valid(blk_qc_t cookie)
-{
-	return cookie != BLK_QC_T_NONE;
-}
-
 struct blk_rq_stat {
 	u64 mean;
 	u64 min;
-- 
2.30.1

[PATCH 09/12] block: rename REQ_HIPRI to REQ_POLLED

[Christoph Hellwig]

Unlike the RWF_HIPRI userspace ABI which is intentionally kept vague,
the bio flag is specіfic to the polling implementation, so rename and
document it properly.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 block/blk-core.c          | 2 +-
 block/blk-merge.c         | 2 +-
 block/blk-mq-debugfs.c    | 2 +-
 block/blk-mq.c            | 4 ++--
 block/blk-mq.h            | 4 ++--
 block/blk-wbt.c           | 4 ++--
 drivers/nvme/host/core.c  | 2 +-
 include/linux/bio.h       | 2 +-
 include/linux/blk_types.h | 4 ++--
 mm/page_io.c              | 2 +-
 10 files changed, 14 insertions(+), 14 deletions(-)

diff --git a/block/blk-core.c b/block/blk-core.c
index 9bcdae93f6d4..adfab5976be0 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -837,7 +837,7 @@ static noinline_for_stack bool submit_bio_checks(struct bio *bio)
 	}
 
 	if (!test_bit(QUEUE_FLAG_POLL, &q->queue_flags))
-		bio->bi_opf &= ~REQ_HIPRI;
+		bio->bi_opf &= ~REQ_POLLED;
 
 	switch (bio_op(bio)) {
 	case REQ_OP_DISCARD:
diff --git a/block/blk-merge.c b/block/blk-merge.c
index ffb4aa0ea68b..ef54c2e68657 100644
--- a/block/blk-merge.c
+++ b/block/blk-merge.c
@@ -285,7 +285,7 @@ static struct bio *blk_bio_segment_split(struct request_queue *q,
 	 * iopoll in direct IO routine. Given performance gain of iopoll for
 	 * big IO can be trival, disable iopoll when split needed.
 	 */
-	bio->bi_opf &= ~REQ_HIPRI;
+	bio->bi_opf &= ~REQ_POLLED;
 
 	return bio_split(bio, sectors, GFP_NOIO, bs);
 }
diff --git a/block/blk-mq-debugfs.c b/block/blk-mq-debugfs.c
index 8ea96d83d599..cd1b34e2c6e7 100644
--- a/block/blk-mq-debugfs.c
+++ b/block/blk-mq-debugfs.c
@@ -286,7 +286,7 @@ static const char *const cmd_flag_name[] = {
 	CMD_FLAG_NAME(BACKGROUND),
 	CMD_FLAG_NAME(NOWAIT),
 	CMD_FLAG_NAME(NOUNMAP),
-	CMD_FLAG_NAME(HIPRI),
+	CMD_FLAG_NAME(POLLED),
 };
 #undef CMD_FLAG_NAME
 
diff --git a/block/blk-mq.c b/block/blk-mq.c
index 40b20f7986c1..c252d42b456f 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -667,7 +667,7 @@ bool blk_mq_complete_request_remote(struct request *rq)
 	 * For a polled request, always complete locallly, it's pointless
 	 * to redirect the completion.
 	 */
-	if (rq->cmd_flags & REQ_HIPRI)
+	if (rq->cmd_flags & REQ_POLLED)
 		return false;
 
 	if (blk_mq_complete_need_ipi(rq)) {
@@ -2201,7 +2201,7 @@ blk_qc_t blk_mq_submit_bio(struct bio *bio)
 
 	rq_qos_throttle(q, bio);
 
-	hipri = bio->bi_opf & REQ_HIPRI;
+	hipri = bio->bi_opf & REQ_POLLED;
 
 	data.cmd_flags = bio->bi_opf;
 	rq = __blk_mq_alloc_request(&data);
diff --git a/block/blk-mq.h b/block/blk-mq.h
index 3616453ca28c..38eac0434a52 100644
--- a/block/blk-mq.h
+++ b/block/blk-mq.h
@@ -108,9 +108,9 @@ static inline struct blk_mq_hw_ctx *blk_mq_map_queue(struct request_queue *q,
 	enum hctx_type type = HCTX_TYPE_DEFAULT;
 
 	/*
-	 * The caller ensure that if REQ_HIPRI, poll must be enabled.
+	 * The caller ensure that if REQ_POLLED, poll must be enabled.
 	 */
-	if (flags & REQ_HIPRI)
+	if (flags & REQ_POLLED)
 		type = HCTX_TYPE_POLL;
 	else if ((flags & REQ_OP_MASK) == REQ_OP_READ)
 		type = HCTX_TYPE_READ;
diff --git a/block/blk-wbt.c b/block/blk-wbt.c
index 42aed0160f86..75787460726d 100644
--- a/block/blk-wbt.c
+++ b/block/blk-wbt.c
@@ -445,7 +445,7 @@ static bool close_io(struct rq_wb *rwb)
 		time_before(now, rwb->last_comp + HZ / 10);
 }
 
-#define REQ_HIPRIO	(REQ_SYNC | REQ_META | REQ_PRIO)
+#define REQ_POLLEDO	(REQ_SYNC | REQ_META | REQ_PRIO)
 
 static inline unsigned int get_limit(struct rq_wb *rwb, unsigned long rw)
 {
@@ -469,7 +469,7 @@ static inline unsigned int get_limit(struct rq_wb *rwb, unsigned long rw)
 	 * the idle limit, or go to normal if we haven't had competing
 	 * IO for a bit.
 	 */
-	if ((rw & REQ_HIPRIO) || wb_recent_wait(rwb) || current_is_kswapd())
+	if ((rw & REQ_POLLEDO) || wb_recent_wait(rwb) || current_is_kswapd())
 		limit = rwb->rq_depth.max_depth;
 	else if ((rw & REQ_BACKGROUND) || close_io(rwb)) {
 		/*
diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
index 0896e21642be..723a4c382385 100644
--- a/drivers/nvme/host/core.c
+++ b/drivers/nvme/host/core.c
@@ -958,7 +958,7 @@ static void nvme_execute_rq_polled(struct request_queue *q,
 
 	WARN_ON_ONCE(!test_bit(QUEUE_FLAG_POLL, &q->queue_flags));
 
-	rq->cmd_flags |= REQ_HIPRI;
+	rq->cmd_flags |= REQ_POLLED;
 	rq->end_io_data = &wait;
 	blk_execute_rq_nowait(bd_disk, rq, at_head, nvme_end_sync_rq);
 
diff --git a/include/linux/bio.h b/include/linux/bio.h
index a0b4cfdf62a4..439a70bc42e2 100644
--- a/include/linux/bio.h
+++ b/include/linux/bio.h
@@ -817,7 +817,7 @@ static inline int bio_integrity_add_page(struct bio *bio, struct page *page,
  */
 static inline void bio_set_polled(struct bio *bio, struct kiocb *kiocb)
 {
-	bio->bi_opf |= REQ_HIPRI;
+	bio->bi_opf |= REQ_POLLED;
 	if (!is_sync_kiocb(kiocb))
 		bio->bi_opf |= REQ_NOWAIT;
 }
diff --git a/include/linux/blk_types.h b/include/linux/blk_types.h
index d0cf835d3b50..ac60432752e3 100644
--- a/include/linux/blk_types.h
+++ b/include/linux/blk_types.h
@@ -392,7 +392,7 @@ enum req_flag_bits {
 	/* command specific flags for REQ_OP_WRITE_ZEROES: */
 	__REQ_NOUNMAP,		/* do not free blocks when zeroing */
 
-	__REQ_HIPRI,
+	__REQ_POLLED,		/* caller polls for completion using blk_poll */
 
 	/* for driver use */
 	__REQ_DRV,
@@ -417,7 +417,7 @@ enum req_flag_bits {
 #define REQ_CGROUP_PUNT		(1ULL << __REQ_CGROUP_PUNT)
 
 #define REQ_NOUNMAP		(1ULL << __REQ_NOUNMAP)
-#define REQ_HIPRI		(1ULL << __REQ_HIPRI)
+#define REQ_POLLED		(1ULL << __REQ_POLLED)
 
 #define REQ_DRV			(1ULL << __REQ_DRV)
 #define REQ_SWAP		(1ULL << __REQ_SWAP)
diff --git a/mm/page_io.c b/mm/page_io.c
index c493ce9ebcf5..dd86ea217da2 100644
--- a/mm/page_io.c
+++ b/mm/page_io.c
@@ -416,7 +416,7 @@ int swap_readpage(struct page *page, bool synchronous)
 	 * attempt to access it in the page fault retry time check.
 	 */
 	if (synchronous) {
-		bio->bi_opf |= REQ_HIPRI;
+		bio->bi_opf |= REQ_POLLED;
 		get_task_struct(current);
 		bio->bi_private = current;
 	}
-- 
2.30.1

[PATCH 10/12] block: RCU free polled bios

[Christoph Hellwig]

Free the actual memory used for the bio using RCU for polled bios.
This will allow to store all the polling information in the bio and
thus simplify passing the cookie and allow for polling stacked devices.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 block/bio.c               | 28 +++++++++++++++++++---------
 include/linux/blk_types.h |  6 +++++-
 2 files changed, 24 insertions(+), 10 deletions(-)

diff --git a/block/bio.c b/block/bio.c
index 303298996afe..7296abe293de 100644
--- a/block/bio.c
+++ b/block/bio.c
@@ -215,25 +215,35 @@ void bio_uninit(struct bio *bio)
 }
 EXPORT_SYMBOL(bio_uninit);
 
+static inline void bio_mempool_free(struct bio *bio)
+{
+	struct bio_set *bs = bio->bi_pool;
+	void *p = bio;
+
+	p -= bs->front_pad;
+	mempool_free(p, &bs->bio_pool);
+}
+
+static void bio_free_rcu(struct rcu_head *head)
+{
+	bio_mempool_free(container_of(head, struct bio, bi_rcu_free));
+}
+
 static void bio_free(struct bio *bio)
 {
 	struct bio_set *bs = bio->bi_pool;
-	void *p;
 
 	bio_uninit(bio);
 
 	if (bs) {
 		bvec_free(&bs->bvec_pool, bio->bi_io_vec, bio->bi_max_vecs);
-
-		/*
-		 * If we have front padding, adjust the bio pointer before freeing
-		 */
-		p = bio;
-		p -= bs->front_pad;
-
-		mempool_free(p, &bs->bio_pool);
+		if (bio->bi_opf & REQ_POLLED)
+			call_rcu(&bio->bi_rcu_free, bio_free_rcu);
+		else
+			bio_mempool_free(bio);
 	} else {
 		/* Bio was allocated by bio_kmalloc() */
+		WARN_ON_ONCE(bio->bi_opf & REQ_POLLED);
 		kfree(bio);
 	}
 }
diff --git a/include/linux/blk_types.h b/include/linux/blk_types.h
index ac60432752e3..183a76bf24b7 100644
--- a/include/linux/blk_types.h
+++ b/include/linux/blk_types.h
@@ -10,6 +10,7 @@
 #include <linux/bvec.h>
 #include <linux/device.h>
 #include <linux/ktime.h>
+#include <linux/rcupdate.h>
 
 struct bio_set;
 struct bio;
@@ -233,7 +234,10 @@ struct bio {
 	blk_status_t		bi_status;
 	atomic_t		__bi_remaining;
 
-	struct bvec_iter	bi_iter;
+	union {
+		struct bvec_iter	bi_iter;
+		struct rcu_head		bi_rcu_free;
+	};
 
 	bio_end_io_t		*bi_end_io;
 
-- 
2.30.1

[PATCH 11/12] block: define 'struct bvec_iter' as packed

[Christoph Hellwig]

From: Ming Lei <ming.lei@redhat.com>

'struct bvec_iter' is embedded into 'struct bio', define it as packed
so that we can get one extra 4bytes for other uses without expanding
bio.

'struct bvec_iter' is often allocated on stack, so making it packed
doesn't affect performance. Also I have run io_uring on both
nvme/null_blk, and not observe performance effect in this way.

Suggested-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 include/linux/bvec.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/include/linux/bvec.h b/include/linux/bvec.h
index ff832e698efb..a0c4f41dfc83 100644
--- a/include/linux/bvec.h
+++ b/include/linux/bvec.h
@@ -43,7 +43,7 @@ struct bvec_iter {
 
 	unsigned int            bi_bvec_done;	/* number of bytes completed in
 						   current bvec */
-};
+} __packed;
 
 struct bvec_iter_all {
 	struct bio_vec	bv;
-- 
2.30.1

[PATCH 12/12] block: switch polling to be bio based

[Christoph Hellwig]

Replace the blk_poll interface that requires the caller to keep a queue
and cookie from the submissions with polling based on the bio.

Polling for the bio itself leads to a few advantages:

 - the cookie construction can made entirely private in blk-mq.c
 - the caller does not need to remember the request_queue and cookie
   separately and thus sidesteps their lifetime issues
 - keeping the device and the cookie inside the bio allows to trivially
   support polling BIOs remapping by stacking drivers
 - a lot of code to propagate the cookie back up the submission path can
   be removed entirely.

Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 arch/m68k/emu/nfblock.c             |   3 +-
 arch/xtensa/platforms/iss/simdisk.c |   3 +-
 block/bio.c                         |   1 +
 block/blk-core.c                    | 103 +++++++++++++++++++---------
 block/blk-mq.c                      |  75 +++++++-------------
 block/blk-mq.h                      |   2 +
 drivers/block/brd.c                 |  12 ++--
 drivers/block/drbd/drbd_int.h       |   2 +-
 drivers/block/drbd/drbd_req.c       |   3 +-
 drivers/block/n64cart.c             |  12 ++--
 drivers/block/null_blk/main.c       |   3 +-
 drivers/block/pktcdvd.c             |   7 +-
 drivers/block/ps3vram.c             |   6 +-
 drivers/block/rsxx/dev.c            |   7 +-
 drivers/block/umem.c                |   4 +-
 drivers/block/zram/zram_drv.c       |  10 +--
 drivers/lightnvm/pblk-init.c        |   6 +-
 drivers/md/bcache/request.c         |  13 ++--
 drivers/md/bcache/request.h         |   4 +-
 drivers/md/dm.c                     |  28 +++-----
 drivers/md/md.c                     |  10 ++-
 drivers/nvdimm/blk.c                |   5 +-
 drivers/nvdimm/btt.c                |   5 +-
 drivers/nvdimm/pmem.c               |   3 +-
 drivers/nvme/host/core.c            |   2 +-
 drivers/nvme/host/multipath.c       |   6 +-
 drivers/nvme/host/nvme.h            |   2 +-
 drivers/s390/block/dcssblk.c        |   7 +-
 drivers/s390/block/xpram.c          |   5 +-
 fs/block_dev.c                      |  25 +++----
 fs/btrfs/inode.c                    |   8 +--
 fs/ext4/file.c                      |   2 +-
 fs/gfs2/file.c                      |   4 +-
 fs/iomap/direct-io.c                |  39 ++++-------
 fs/xfs/xfs_file.c                   |   2 +-
 fs/zonefs/super.c                   |   2 +-
 include/linux/bio.h                 |   2 +-
 include/linux/blk-mq.h              |  15 +---
 include/linux/blk_types.h           |  12 ++--
 include/linux/blkdev.h              |   8 ++-
 include/linux/fs.h                  |   6 +-
 include/linux/iomap.h               |   3 +-
 mm/page_io.c                        |   8 +--
 43 files changed, 208 insertions(+), 277 deletions(-)

diff --git a/arch/m68k/emu/nfblock.c b/arch/m68k/emu/nfblock.c
index ba808543161a..dd36808f0d5e 100644
--- a/arch/m68k/emu/nfblock.c
+++ b/arch/m68k/emu/nfblock.c
@@ -59,7 +59,7 @@ struct nfhd_device {
 	struct gendisk *disk;
 };
 
-static blk_qc_t nfhd_submit_bio(struct bio *bio)
+static void nfhd_submit_bio(struct bio *bio)
 {
 	struct nfhd_device *dev = bio->bi_bdev->bd_disk->private_data;
 	struct bio_vec bvec;
@@ -77,7 +77,6 @@ static blk_qc_t nfhd_submit_bio(struct bio *bio)
 		sec += len;
 	}
 	bio_endio(bio);
-	return BLK_QC_T_NONE;
 }
 
 static int nfhd_getgeo(struct block_device *bdev, struct hd_geometry *geo)
diff --git a/arch/xtensa/platforms/iss/simdisk.c b/arch/xtensa/platforms/iss/simdisk.c
index fc09be7b1347..182825d639e2 100644
--- a/arch/xtensa/platforms/iss/simdisk.c
+++ b/arch/xtensa/platforms/iss/simdisk.c
@@ -101,7 +101,7 @@ static void simdisk_transfer(struct simdisk *dev, unsigned long sector,
 	spin_unlock(&dev->lock);
 }
 
-static blk_qc_t simdisk_submit_bio(struct bio *bio)
+static void simdisk_submit_bio(struct bio *bio)
 {
 	struct simdisk *dev = bio->bi_bdev->bd_disk->private_data;
 	struct bio_vec bvec;
@@ -119,7 +119,6 @@ static blk_qc_t simdisk_submit_bio(struct bio *bio)
 	}
 
 	bio_endio(bio);
-	return BLK_QC_T_NONE;
 }
 
 static int simdisk_open(struct block_device *bdev, fmode_t mode)
diff --git a/block/bio.c b/block/bio.c
index 7296abe293de..484b6d786857 100644
--- a/block/bio.c
+++ b/block/bio.c
@@ -259,6 +259,7 @@ void bio_init(struct bio *bio, struct bio_vec *table,
 	memset(bio, 0, sizeof(*bio));
 	atomic_set(&bio->__bi_remaining, 1);
 	atomic_set(&bio->__bi_cnt, 1);
+	bio->bi_cookie = BLK_QC_T_NONE;
 
 	bio->bi_io_vec = table;
 	bio->bi_max_vecs = max_vecs;
diff --git a/block/blk-core.c b/block/blk-core.c
index adfab5976be0..77fdb00fcad3 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -910,18 +910,18 @@ static noinline_for_stack bool submit_bio_checks(struct bio *bio)
 	return false;
 }
 
-static blk_qc_t __submit_bio(struct bio *bio)
+static void __submit_bio(struct bio *bio)
 {
 	struct gendisk *disk = bio->bi_bdev->bd_disk;
-	blk_qc_t ret = BLK_QC_T_NONE;
 
 	if (blk_crypto_bio_prep(&bio)) {
-		if (!disk->fops->submit_bio)
-			return blk_mq_submit_bio(bio);
-		ret = disk->fops->submit_bio(bio);
+		if (!disk->fops->submit_bio) {
+			blk_mq_submit_bio(bio);
+			return;
+		}
+		disk->fops->submit_bio(bio);
 	}
 	blk_queue_exit(disk->queue);
-	return ret;
 }
 
 /*
@@ -943,10 +943,9 @@ static blk_qc_t __submit_bio(struct bio *bio)
  * bio_list_on_stack[1] contains bios that were submitted before the current
  *	->submit_bio_bio, but that haven't been processed yet.
  */
-static blk_qc_t __submit_bio_noacct(struct bio *bio)
+static void __submit_bio_noacct(struct bio *bio)
 {
 	struct bio_list bio_list_on_stack[2];
-	blk_qc_t ret = BLK_QC_T_NONE;
 
 	BUG_ON(bio->bi_next);
 
@@ -966,7 +965,7 @@ static blk_qc_t __submit_bio_noacct(struct bio *bio)
 		bio_list_on_stack[1] = bio_list_on_stack[0];
 		bio_list_init(&bio_list_on_stack[0]);
 
-		ret = __submit_bio(bio);
+		__submit_bio(bio);
 
 		/*
 		 * Sort new bios into those for a lower level and those for the
@@ -989,13 +988,11 @@ static blk_qc_t __submit_bio_noacct(struct bio *bio)
 	} while ((bio = bio_list_pop(&bio_list_on_stack[0])));
 
 	current->bio_list = NULL;
-	return ret;
 }
 
-static blk_qc_t __submit_bio_noacct_mq(struct bio *bio)
+static void __submit_bio_noacct_mq(struct bio *bio)
 {
 	struct bio_list bio_list[2] = { };
-	blk_qc_t ret = BLK_QC_T_NONE;
 
 	current->bio_list = bio_list;
 
@@ -1007,15 +1004,13 @@ static blk_qc_t __submit_bio_noacct_mq(struct bio *bio)
 
 		if (!blk_crypto_bio_prep(&bio)) {
 			blk_queue_exit(disk->queue);
-			ret = BLK_QC_T_NONE;
 			continue;
 		}
 
-		ret = blk_mq_submit_bio(bio);
+		blk_mq_submit_bio(bio);
 	} while ((bio = bio_list_pop(&bio_list[0])));
 
 	current->bio_list = NULL;
-	return ret;
 }
 
 /**
@@ -1027,10 +1022,10 @@ static blk_qc_t __submit_bio_noacct_mq(struct bio *bio)
  * systems and other upper level users of the block layer should use
  * submit_bio() instead.
  */
-blk_qc_t submit_bio_noacct(struct bio *bio)
+void submit_bio_noacct(struct bio *bio)
 {
 	if (!submit_bio_checks(bio))
-		return BLK_QC_T_NONE;
+		return;
 
 	/*
 	 * We only want one ->submit_bio to be active at a time, else stack
@@ -1038,14 +1033,12 @@ blk_qc_t submit_bio_noacct(struct bio *bio)
 	 * to collect a list of requests submited by a ->submit_bio method while
 	 * it is active, and then process them after it returned.
 	 */
-	if (current->bio_list) {
+	if (current->bio_list)
 		bio_list_add(&current->bio_list[0], bio);
-		return BLK_QC_T_NONE;
-	}
-
-	if (!bio->bi_bdev->bd_disk->fops->submit_bio)
-		return __submit_bio_noacct_mq(bio);
-	return __submit_bio_noacct(bio);
+	else if (!bio->bi_bdev->bd_disk->fops->submit_bio)
+		__submit_bio_noacct_mq(bio);
+	else
+		__submit_bio_noacct(bio);
 }
 EXPORT_SYMBOL(submit_bio_noacct);
 
@@ -1062,10 +1055,10 @@ EXPORT_SYMBOL(submit_bio_noacct);
  * in @bio.  The bio must NOT be touched by thecaller until ->bi_end_io() has
  * been called.
  */
-blk_qc_t submit_bio(struct bio *bio)
+void submit_bio(struct bio *bio)
 {
 	if (blkcg_punt_bio_submit(bio))
-		return BLK_QC_T_NONE;
+		return;
 
 	/*
 	 * If it's a regular read/write or a barrier with data attached,
@@ -1106,19 +1099,67 @@ blk_qc_t submit_bio(struct bio *bio)
 	if (unlikely(bio_op(bio) == REQ_OP_READ &&
 	    bio_flagged(bio, BIO_WORKINGSET))) {
 		unsigned long pflags;
-		blk_qc_t ret;
 
 		psi_memstall_enter(&pflags);
-		ret = submit_bio_noacct(bio);
+		submit_bio_noacct(bio);
 		psi_memstall_leave(&pflags);
-
-		return ret;
+		return;
 	}
 
-	return submit_bio_noacct(bio);
+	submit_bio_noacct(bio);
 }
 EXPORT_SYMBOL(submit_bio);
 
+/**
+ * bio_poll - poll for BIO completions
+ * @bio: bio to poll for
+ *
+ * Poll for completions on queue associated with the bio. Returns number of
+ * completed entries found. If @spin is true, then bio_poll will continue
+ * looping until at least one completion is found, unless the task is
+ * otherwise marked running (or we need to reschedule).
+ *
+ * Note: the caller must either be the context that submitted @bio, or
+ * be in a RCU critical section to prevent freeing of @bio.
+ */
+int bio_poll(struct bio *bio, bool spin)
+{
+	struct request_queue *q = bio->bi_bdev->bd_disk->queue;
+	blk_qc_t cookie = READ_ONCE(bio->bi_cookie);
+
+	if (cookie == BLK_QC_T_NONE ||
+	    !test_bit(QUEUE_FLAG_POLL, &q->queue_flags))
+		return 0;
+
+	if (current->plug)
+		blk_flush_plug_list(current->plug, false);
+
+	/* not yet implemented, so this should not happen */
+	if (WARN_ON_ONCE(!queue_is_mq(q)))
+		return 0;
+	return blk_mq_poll(q, cookie, spin);
+}
+EXPORT_SYMBOL_GPL(bio_poll);
+
+/*
+ * Helper to implements file_operations.iopoll.  Requires the bio to be stored
+ * in iocb->private, and cleared before freeing the bio.
+ */
+int iocb_bio_iopoll(struct kiocb *kiocb, bool spin)
+{
+	struct bio *bio;
+	int ret = 0;
+
+	rcu_read_lock();
+	bio = READ_ONCE(kiocb->private);
+	if (bio)
+		ret = bio_poll(bio, spin);
+	rcu_read_unlock();
+
+	return ret;
+}
+EXPORT_SYMBOL_GPL(iocb_bio_iopoll);
+
 /**
  * blk_cloned_rq_check_limits - Helper function to check a cloned request
  *                              for the new queue limits
diff --git a/block/blk-mq.c b/block/blk-mq.c
index c252d42b456f..a2d3f72f1984 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -63,6 +63,9 @@ static int blk_mq_poll_stats_bkt(const struct request *rq)
 	return bucket;
 }
 
+#define BLK_QC_T_SHIFT		16
+#define BLK_QC_T_INTERNAL	(1U << 31)
+
 static inline struct blk_mq_hw_ctx *blk_qc_to_hctx(struct request_queue *q,
 		blk_qc_t qc)
 {
@@ -79,6 +82,13 @@ static inline struct request *blk_qc_to_rq(struct blk_mq_hw_ctx *hctx,
 	return blk_mq_tag_to_rq(hctx->tags, tag);
 }
 
+static inline blk_qc_t blk_rq_to_qc(struct request *rq)
+{
+	return (rq->mq_hctx->queue_num << BLK_QC_T_SHIFT) |
+		(rq->tag != -1 ?
+		 rq->tag : (rq->internal_tag | BLK_QC_T_INTERNAL));
+}
+
 /*
  * Check if any of the ctx, dispatch list or elevator
  * have pending work in this hardware queue.
@@ -747,6 +757,8 @@ void blk_mq_start_request(struct request *rq)
 	if (blk_integrity_rq(rq) && req_op(rq) == REQ_OP_WRITE)
 		q->integrity.profile->prepare_fn(rq);
 #endif
+	if (rq->bio && rq->bio->bi_opf & REQ_POLLED)
+	        WRITE_ONCE(rq->bio->bi_cookie, blk_rq_to_qc(rq));
 }
 EXPORT_SYMBOL(blk_mq_start_request);
 
@@ -1981,19 +1993,15 @@ static void blk_mq_bio_to_request(struct request *rq, struct bio *bio,
 }
 
 static blk_status_t __blk_mq_issue_directly(struct blk_mq_hw_ctx *hctx,
-					    struct request *rq,
-					    blk_qc_t *cookie, bool last)
+					    struct request *rq, bool last)
 {
 	struct request_queue *q = rq->q;
 	struct blk_mq_queue_data bd = {
 		.rq = rq,
 		.last = last,
 	};
-	blk_qc_t new_cookie;
 	blk_status_t ret;
 
-	new_cookie = request_to_qc_t(hctx, rq);
-
 	/*
 	 * For OK queue, we are done. For error, caller may kill it.
 	 * Any other error (busy), just add it to our list as we
@@ -2003,7 +2011,6 @@ static blk_status_t __blk_mq_issue_directly(struct blk_mq_hw_ctx *hctx,
 	switch (ret) {
 	case BLK_STS_OK:
 		blk_mq_update_dispatch_busy(hctx, false);
-		*cookie = new_cookie;
 		break;
 	case BLK_STS_RESOURCE:
 	case BLK_STS_DEV_RESOURCE:
@@ -2012,7 +2019,6 @@ static blk_status_t __blk_mq_issue_directly(struct blk_mq_hw_ctx *hctx,
 		break;
 	default:
 		blk_mq_update_dispatch_busy(hctx, false);
-		*cookie = BLK_QC_T_NONE;
 		break;
 	}
 
@@ -2021,7 +2027,6 @@ static blk_status_t __blk_mq_issue_directly(struct blk_mq_hw_ctx *hctx,
 
 static blk_status_t __blk_mq_try_issue_directly(struct blk_mq_hw_ctx *hctx,
 						struct request *rq,
-						blk_qc_t *cookie,
 						bool bypass_insert, bool last)
 {
 	struct request_queue *q = rq->q;
@@ -2051,7 +2056,7 @@ static blk_status_t __blk_mq_try_issue_directly(struct blk_mq_hw_ctx *hctx,
 		goto insert;
 	}
 
-	return __blk_mq_issue_directly(hctx, rq, cookie, last);
+	return __blk_mq_issue_directly(hctx, rq, last);
 insert:
 	if (bypass_insert)
 		return BLK_STS_RESOURCE;
@@ -2065,7 +2070,6 @@ static blk_status_t __blk_mq_try_issue_directly(struct blk_mq_hw_ctx *hctx,
  * blk_mq_try_issue_directly - Try to send a request directly to device driver.
  * @hctx: Pointer of the associated hardware queue.
  * @rq: Pointer to request to be sent.
- * @cookie: Request queue cookie.
  *
  * If the device has enough resources to accept a new request now, send the
  * request directly to device driver. Else, insert at hctx->dispatch queue, so
@@ -2073,7 +2077,7 @@ static blk_status_t __blk_mq_try_issue_directly(struct blk_mq_hw_ctx *hctx,
  * queue have higher priority.
  */
 static void blk_mq_try_issue_directly(struct blk_mq_hw_ctx *hctx,
-		struct request *rq, blk_qc_t *cookie)
+		struct request *rq)
 {
 	blk_status_t ret;
 	int srcu_idx;
@@ -2082,7 +2086,7 @@ static void blk_mq_try_issue_directly(struct blk_mq_hw_ctx *hctx,
 
 	hctx_lock(hctx, &srcu_idx);
 
-	ret = __blk_mq_try_issue_directly(hctx, rq, cookie, false, true);
+	ret = __blk_mq_try_issue_directly(hctx, rq, false, true);
 	if (ret == BLK_STS_RESOURCE || ret == BLK_STS_DEV_RESOURCE)
 		blk_mq_request_bypass_insert(rq, false, true);
 	else if (ret != BLK_STS_OK)
@@ -2095,11 +2099,10 @@ blk_status_t blk_mq_request_issue_directly(struct request *rq, bool last)
 {
 	blk_status_t ret;
 	int srcu_idx;
-	blk_qc_t unused_cookie;
 	struct blk_mq_hw_ctx *hctx = rq->mq_hctx;
 
 	hctx_lock(hctx, &srcu_idx);
-	ret = __blk_mq_try_issue_directly(hctx, rq, &unused_cookie, true, last);
+	ret = __blk_mq_try_issue_directly(hctx, rq, true, last);
 	hctx_unlock(hctx, srcu_idx);
 
 	return ret;
@@ -2167,10 +2170,8 @@ static void blk_add_rq_to_plug(struct blk_plug *plug, struct request *rq)
  *
  * It will not queue the request if there is an error with the bio, or at the
  * request creation.
- *
- * Returns: Request queue cookie.
  */
-blk_qc_t blk_mq_submit_bio(struct bio *bio)
+void blk_mq_submit_bio(struct bio *bio)
 {
 	struct request_queue *q = bio->bi_bdev->bd_disk->queue;
 	const int is_sync = op_is_sync(bio->bi_opf);
@@ -2182,9 +2183,7 @@ blk_qc_t blk_mq_submit_bio(struct bio *bio)
 	struct blk_plug *plug;
 	struct request *same_queue_rq = NULL;
 	unsigned int nr_segs;
-	blk_qc_t cookie;
 	blk_status_t ret;
-	bool hipri;
 
 	blk_queue_bounce(q, &bio);
 	__blk_queue_split(&bio, &nr_segs);
@@ -2201,8 +2200,6 @@ blk_qc_t blk_mq_submit_bio(struct bio *bio)
 
 	rq_qos_throttle(q, bio);
 
-	hipri = bio->bi_opf & REQ_POLLED;
-
 	data.cmd_flags = bio->bi_opf;
 	rq = __blk_mq_alloc_request(&data);
 	if (unlikely(!rq)) {
@@ -2216,8 +2213,6 @@ blk_qc_t blk_mq_submit_bio(struct bio *bio)
 
 	rq_qos_track(q, rq, bio);
 
-	cookie = request_to_qc_t(data.hctx, rq);
-
 	blk_mq_bio_to_request(rq, bio, nr_segs);
 
 	ret = blk_crypto_init_request(rq);
@@ -2225,7 +2220,7 @@ blk_qc_t blk_mq_submit_bio(struct bio *bio)
 		bio->bi_status = ret;
 		bio_endio(bio);
 		blk_mq_free_request(rq);
-		return BLK_QC_T_NONE;
+		return;
 	}
 
 	plug = blk_mq_plug(q, bio);
@@ -2280,8 +2275,7 @@ blk_qc_t blk_mq_submit_bio(struct bio *bio)
 		if (same_queue_rq) {
 			data.hctx = same_queue_rq->mq_hctx;
 			trace_block_unplug(q, 1, true);
-			blk_mq_try_issue_directly(data.hctx, same_queue_rq,
-					&cookie);
+			blk_mq_try_issue_directly(data.hctx, same_queue_rq);
 		}
 	} else if ((q->nr_hw_queues > 1 && is_sync) ||
 			!data.hctx->dispatch_busy) {
@@ -2289,18 +2283,15 @@ blk_qc_t blk_mq_submit_bio(struct bio *bio)
 		 * There is no scheduler and we can try to send directly
 		 * to the hardware.
 		 */
-		blk_mq_try_issue_directly(data.hctx, rq, &cookie);
+		blk_mq_try_issue_directly(data.hctx, rq);
 	} else {
 		/* Default case. */
 		blk_mq_sched_insert_request(rq, false, true, true);
 	}
 
-	if (!hipri)
-		return BLK_QC_T_NONE;
-	return cookie;
+	return;
 queue_exit:
 	blk_queue_exit(q);
-	return BLK_QC_T_NONE;
 }
 
 void blk_mq_free_rqs(struct blk_mq_tag_set *set, struct blk_mq_tags *tags,
@@ -3882,27 +3873,8 @@ static int blk_mq_poll_classic(struct request_queue *q, blk_qc_t qc, bool spin)
 	return 0;
 }
 
-/**
- * blk_poll - poll for IO completions
- * @q:  the queue
- * @cookie: cookie passed back at IO submission time
- * @spin: whether to spin for completions
- *
- * Description:
- *    Poll for completions on the passed in queue. Returns number of
- *    completed entries found. If @spin is true, then blk_poll will continue
- *    looping until at least one completion is found, unless the task is
- *    otherwise marked running (or we need to reschedule).
- */
-int blk_poll(struct request_queue *q, blk_qc_t cookie, bool spin)
+int blk_mq_poll(struct request_queue *q, blk_qc_t cookie, bool spin)
 {
-	if (cookie == BLK_QC_T_NONE ||
-	    !test_bit(QUEUE_FLAG_POLL, &q->queue_flags))
-		return 0;
-
-	if (current->plug)
-		blk_flush_plug_list(current->plug, false);
-
 	/*
 	 * If we sleep, have the caller restart the poll loop to reset the
 	 * state.  Like for the other success return cases, the caller is
@@ -3916,7 +3888,6 @@ int blk_poll(struct request_queue *q, blk_qc_t cookie, bool spin)
 	}
 	return blk_mq_poll_classic(q, cookie, spin);
 }
-EXPORT_SYMBOL_GPL(blk_poll);
 
 unsigned int blk_mq_rq_cpu(struct request *rq)
 {
diff --git a/block/blk-mq.h b/block/blk-mq.h
index 38eac0434a52..17c5287d1590 100644
--- a/block/blk-mq.h
+++ b/block/blk-mq.h
@@ -37,6 +37,8 @@ struct blk_mq_ctx {
 	struct kobject		kobj;
 } ____cacheline_aligned_in_smp;
 
+void blk_mq_submit_bio(struct bio *bio);
+int blk_mq_poll(struct request_queue *q, blk_qc_t cookie, bool spin);
 void blk_mq_exit_queue(struct request_queue *q);
 int blk_mq_update_nr_requests(struct request_queue *q, unsigned int nr);
 void blk_mq_wake_waiters(struct request_queue *q);
diff --git a/drivers/block/brd.c b/drivers/block/brd.c
index 18bf99906662..0b1004ce3329 100644
--- a/drivers/block/brd.c
+++ b/drivers/block/brd.c
@@ -282,7 +282,7 @@ static int brd_do_bvec(struct brd_device *brd, struct page *page,
 	return err;
 }
 
-static blk_qc_t brd_submit_bio(struct bio *bio)
+static void brd_submit_bio(struct bio *bio)
 {
 	struct brd_device *brd = bio->bi_bdev->bd_disk->private_data;
 	sector_t sector = bio->bi_iter.bi_sector;
@@ -299,16 +299,14 @@ static blk_qc_t brd_submit_bio(struct bio *bio)
 
 		err = brd_do_bvec(brd, bvec.bv_page, len, bvec.bv_offset,
 				  bio_op(bio), sector);
-		if (err)
-			goto io_error;
+		if (err) {
+			bio_io_error(bio);
+			return;
+		}
 		sector += len >> SECTOR_SHIFT;
 	}
 
 	bio_endio(bio);
-	return BLK_QC_T_NONE;
-io_error:
-	bio_io_error(bio);
-	return BLK_QC_T_NONE;
 }
 
 static int brd_rw_page(struct block_device *bdev, sector_t sector,
diff --git a/drivers/block/drbd/drbd_int.h b/drivers/block/drbd/drbd_int.h
index 5d9181382ce1..6674a0b88341 100644
--- a/drivers/block/drbd/drbd_int.h
+++ b/drivers/block/drbd/drbd_int.h
@@ -1448,7 +1448,7 @@ extern void conn_free_crypto(struct drbd_connection *connection);
 /* drbd_req */
 extern void do_submit(struct work_struct *ws);
 extern void __drbd_make_request(struct drbd_device *, struct bio *);
-extern blk_qc_t drbd_submit_bio(struct bio *bio);
+void drbd_submit_bio(struct bio *bio);
 extern int drbd_read_remote(struct drbd_device *device, struct drbd_request *req);
 extern int is_valid_ar_handle(struct drbd_request *, sector_t);
 
diff --git a/drivers/block/drbd/drbd_req.c b/drivers/block/drbd/drbd_req.c
index 9398c2c2cb2d..df70b33216e0 100644
--- a/drivers/block/drbd/drbd_req.c
+++ b/drivers/block/drbd/drbd_req.c
@@ -1596,7 +1596,7 @@ void do_submit(struct work_struct *ws)
 	}
 }
 
-blk_qc_t drbd_submit_bio(struct bio *bio)
+void drbd_submit_bio(struct bio *bio)
 {
 	struct drbd_device *device = bio->bi_bdev->bd_disk->private_data;
 
@@ -1609,7 +1609,6 @@ blk_qc_t drbd_submit_bio(struct bio *bio)
 
 	inc_ap_bio(device);
 	__drbd_make_request(device, bio);
-	return BLK_QC_T_NONE;
 }
 
 static bool net_timeout_reached(struct drbd_request *net_req,
diff --git a/drivers/block/n64cart.c b/drivers/block/n64cart.c
index 47bdf324e962..94f3c43d6da2 100644
--- a/drivers/block/n64cart.c
+++ b/drivers/block/n64cart.c
@@ -84,7 +84,7 @@ static bool n64cart_do_bvec(struct device *dev, struct bio_vec *bv, u32 pos)
 	return true;
 }
 
-static blk_qc_t n64cart_submit_bio(struct bio *bio)
+static void n64cart_submit_bio(struct bio *bio)
 {
 	struct bio_vec bvec;
 	struct bvec_iter iter;
@@ -92,16 +92,14 @@ static blk_qc_t n64cart_submit_bio(struct bio *bio)
 	u32 pos = bio->bi_iter.bi_sector << SECTOR_SHIFT;
 
 	bio_for_each_segment(bvec, bio, iter) {
-		if (!n64cart_do_bvec(dev, &bvec, pos))
-			goto io_error;
+		if (!n64cart_do_bvec(dev, &bvec, pos)) {
+			bio_io_error(bio);
+			return;
+		}
 		pos += bvec.bv_len;
 	}
 
 	bio_endio(bio);
-	return BLK_QC_T_NONE;
-io_error:
-	bio_io_error(bio);
-	return BLK_QC_T_NONE;
 }
 
 static const struct block_device_operations n64cart_fops = {
diff --git a/drivers/block/null_blk/main.c b/drivers/block/null_blk/main.c
index d6c821d48090..29a86322cc9f 100644
--- a/drivers/block/null_blk/main.c
+++ b/drivers/block/null_blk/main.c
@@ -1416,7 +1416,7 @@ static struct nullb_queue *nullb_to_queue(struct nullb *nullb)
 	return &nullb->queues[index];
 }
 
-static blk_qc_t null_submit_bio(struct bio *bio)
+static void null_submit_bio(struct bio *bio)
 {
 	sector_t sector = bio->bi_iter.bi_sector;
 	sector_t nr_sectors = bio_sectors(bio);
@@ -1428,7 +1428,6 @@ static blk_qc_t null_submit_bio(struct bio *bio)
 	cmd->bio = bio;
 
 	null_handle_cmd(cmd, sector, nr_sectors, bio_op(bio));
-	return BLK_QC_T_NONE;
 }
 
 static bool should_timeout_request(struct request *rq)
diff --git a/drivers/block/pktcdvd.c b/drivers/block/pktcdvd.c
index bd3556585122..628ed0d3f2e4 100644
--- a/drivers/block/pktcdvd.c
+++ b/drivers/block/pktcdvd.c
@@ -2402,7 +2402,7 @@ static void pkt_make_request_write(struct request_queue *q, struct bio *bio)
 	}
 }
 
-static blk_qc_t pkt_submit_bio(struct bio *bio)
+static void pkt_submit_bio(struct bio *bio)
 {
 	struct pktcdvd_device *pd;
 	char b[BDEVNAME_SIZE];
@@ -2425,7 +2425,7 @@ static blk_qc_t pkt_submit_bio(struct bio *bio)
 	 */
 	if (bio_data_dir(bio) == READ) {
 		pkt_make_request_read(pd, bio);
-		return BLK_QC_T_NONE;
+		return;
 	}
 
 	if (!test_bit(PACKET_WRITABLE, &pd->flags)) {
@@ -2457,10 +2457,9 @@ static blk_qc_t pkt_submit_bio(struct bio *bio)
 		pkt_make_request_write(bio->bi_bdev->bd_disk->queue, split);
 	} while (split != bio);
 
-	return BLK_QC_T_NONE;
+	return;
 end_io:
 	bio_io_error(bio);
-	return BLK_QC_T_NONE;
 }
 
 static void pkt_init_queue(struct pktcdvd_device *pd)
diff --git a/drivers/block/ps3vram.c b/drivers/block/ps3vram.c
index 1d738999fb69..382be2f07c49 100644
--- a/drivers/block/ps3vram.c
+++ b/drivers/block/ps3vram.c
@@ -579,7 +579,7 @@ static struct bio *ps3vram_do_bio(struct ps3_system_bus_device *dev,
 	return next;
 }
 
-static blk_qc_t ps3vram_submit_bio(struct bio *bio)
+static void ps3vram_submit_bio(struct bio *bio)
 {
 	struct ps3_system_bus_device *dev = bio->bi_bdev->bd_disk->private_data;
 	struct ps3vram_priv *priv = ps3_system_bus_get_drvdata(dev);
@@ -595,13 +595,11 @@ static blk_qc_t ps3vram_submit_bio(struct bio *bio)
 	spin_unlock_irq(&priv->lock);
 
 	if (busy)
-		return BLK_QC_T_NONE;
+		return;
 
 	do {
 		bio = ps3vram_do_bio(dev, bio);
 	} while (bio);
-
-	return BLK_QC_T_NONE;
 }
 
 static const struct block_device_operations ps3vram_fops = {
diff --git a/drivers/block/rsxx/dev.c b/drivers/block/rsxx/dev.c
index 9a28322a8cd8..a4f5a09d0937 100644
--- a/drivers/block/rsxx/dev.c
+++ b/drivers/block/rsxx/dev.c
@@ -50,7 +50,7 @@ struct rsxx_bio_meta {
 
 static struct kmem_cache *bio_meta_pool;
 
-static blk_qc_t rsxx_submit_bio(struct bio *bio);
+static void rsxx_submit_bio(struct bio *bio);
 
 /*----------------- Block Device Operations -----------------*/
 static int rsxx_blkdev_ioctl(struct block_device *bdev,
@@ -120,7 +120,7 @@ static void bio_dma_done_cb(struct rsxx_cardinfo *card,
 	}
 }
 
-static blk_qc_t rsxx_submit_bio(struct bio *bio)
+static void rsxx_submit_bio(struct bio *bio)
 {
 	struct rsxx_cardinfo *card = bio->bi_bdev->bd_disk->private_data;
 	struct rsxx_bio_meta *bio_meta;
@@ -169,7 +169,7 @@ static blk_qc_t rsxx_submit_bio(struct bio *bio)
 	if (st)
 		goto queue_err;
 
-	return BLK_QC_T_NONE;
+	return;
 
 queue_err:
 	kmem_cache_free(bio_meta_pool, bio_meta);
@@ -177,7 +177,6 @@ static blk_qc_t rsxx_submit_bio(struct bio *bio)
 	if (st)
 		bio->bi_status = st;
 	bio_endio(bio);
-	return BLK_QC_T_NONE;
 }
 
 /*----------------- Device Setup -------------------*/
diff --git a/drivers/block/umem.c b/drivers/block/umem.c
index 664280f23bee..36610a4de792 100644
--- a/drivers/block/umem.c
+++ b/drivers/block/umem.c
@@ -519,7 +519,7 @@ static int mm_check_plugged(struct cardinfo *card)
 	return !!blk_check_plugged(mm_unplug, card, sizeof(struct blk_plug_cb));
 }
 
-static blk_qc_t mm_submit_bio(struct bio *bio)
+static void mm_submit_bio(struct bio *bio)
 {
 	struct cardinfo *card = bio->bi_bdev->bd_disk->private_data;
 
@@ -536,8 +536,6 @@ static blk_qc_t mm_submit_bio(struct bio *bio)
 	if (op_is_sync(bio->bi_opf) || !mm_check_plugged(card))
 		activate(card);
 	spin_unlock_irq(&card->lock);
-
-	return BLK_QC_T_NONE;
 }
 
 static irqreturn_t mm_interrupt(int irq, void *__card)
diff --git a/drivers/block/zram/zram_drv.c b/drivers/block/zram/zram_drv.c
index cf8deecc39ef..51dab0d4bedf 100644
--- a/drivers/block/zram/zram_drv.c
+++ b/drivers/block/zram/zram_drv.c
@@ -1598,22 +1598,18 @@ static void __zram_make_request(struct zram *zram, struct bio *bio)
 /*
  * Handler function for all zram I/O requests.
  */
-static blk_qc_t zram_submit_bio(struct bio *bio)
+static void zram_submit_bio(struct bio *bio)
 {
 	struct zram *zram = bio->bi_bdev->bd_disk->private_data;
 
 	if (!valid_io_request(zram, bio->bi_iter.bi_sector,
 					bio->bi_iter.bi_size)) {
 		atomic64_inc(&zram->stats.invalid_io);
-		goto error;
+		bio_io_error(bio);
+		return;
 	}
 
 	__zram_make_request(zram, bio);
-	return BLK_QC_T_NONE;
-
-error:
-	bio_io_error(bio);
-	return BLK_QC_T_NONE;
 }
 
 static void zram_slot_free_notify(struct block_device *bdev,
diff --git a/drivers/lightnvm/pblk-init.c b/drivers/lightnvm/pblk-init.c
index 5924f09c217b..6716003d4cb6 100644
--- a/drivers/lightnvm/pblk-init.c
+++ b/drivers/lightnvm/pblk-init.c
@@ -47,7 +47,7 @@ static struct pblk_global_caches pblk_caches = {
 
 struct bio_set pblk_bio_set;
 
-static blk_qc_t pblk_submit_bio(struct bio *bio)
+static void pblk_submit_bio(struct bio *bio)
 {
 	struct pblk *pblk = bio->bi_bdev->bd_disk->queue->queuedata;
 
@@ -55,7 +55,7 @@ static blk_qc_t pblk_submit_bio(struct bio *bio)
 		pblk_discard(pblk, bio);
 		if (!(bio->bi_opf & REQ_PREFLUSH)) {
 			bio_endio(bio);
-			return BLK_QC_T_NONE;
+			return;
 		}
 	}
 
@@ -75,8 +75,6 @@ static blk_qc_t pblk_submit_bio(struct bio *bio)
 
 		pblk_write_to_cache(pblk, bio, PBLK_IOTYPE_USER);
 	}
-
-	return BLK_QC_T_NONE;
 }
 
 static const struct block_device_operations pblk_bops = {
diff --git a/drivers/md/bcache/request.c b/drivers/md/bcache/request.c
index 29c231758293..b549147a567f 100644
--- a/drivers/md/bcache/request.c
+++ b/drivers/md/bcache/request.c
@@ -1169,7 +1169,7 @@ static void quit_max_writeback_rate(struct cache_set *c,
 
 /* Cached devices - read & write stuff */
 
-blk_qc_t cached_dev_submit_bio(struct bio *bio)
+void cached_dev_submit_bio(struct bio *bio)
 {
 	struct search *s;
 	struct block_device *orig_bdev = bio->bi_bdev;
@@ -1182,7 +1182,7 @@ blk_qc_t cached_dev_submit_bio(struct bio *bio)
 		     dc->io_disable)) {
 		bio->bi_status = BLK_STS_IOERR;
 		bio_endio(bio);
-		return BLK_QC_T_NONE;
+		return;
 	}
 
 	if (likely(d->c)) {
@@ -1228,8 +1228,6 @@ blk_qc_t cached_dev_submit_bio(struct bio *bio)
 	} else
 		/* I/O request sent to backing device */
 		detached_dev_do_request(d, bio, orig_bdev, start_time);
-
-	return BLK_QC_T_NONE;
 }
 
 static int cached_dev_ioctl(struct bcache_device *d, fmode_t mode,
@@ -1279,7 +1277,7 @@ static void flash_dev_nodata(struct closure *cl)
 	continue_at(cl, search_free, NULL);
 }
 
-blk_qc_t flash_dev_submit_bio(struct bio *bio)
+void flash_dev_submit_bio(struct bio *bio)
 {
 	struct search *s;
 	struct closure *cl;
@@ -1288,7 +1286,7 @@ blk_qc_t flash_dev_submit_bio(struct bio *bio)
 	if (unlikely(d->c && test_bit(CACHE_SET_IO_DISABLE, &d->c->flags))) {
 		bio->bi_status = BLK_STS_IOERR;
 		bio_endio(bio);
-		return BLK_QC_T_NONE;
+		return;
 	}
 
 	s = search_alloc(bio, d, bio->bi_bdev, bio_start_io_acct(bio));
@@ -1304,7 +1302,7 @@ blk_qc_t flash_dev_submit_bio(struct bio *bio)
 		continue_at_nobarrier(&s->cl,
 				      flash_dev_nodata,
 				      bcache_wq);
-		return BLK_QC_T_NONE;
+		return;
 	} else if (bio_data_dir(bio)) {
 		bch_keybuf_check_overlapping(&s->iop.c->moving_gc_keys,
 					&KEY(d->id, bio->bi_iter.bi_sector, 0),
@@ -1320,7 +1318,6 @@ blk_qc_t flash_dev_submit_bio(struct bio *bio)
 	}
 
 	continue_at(cl, search_free, NULL);
-	return BLK_QC_T_NONE;
 }
 
 static int flash_dev_ioctl(struct bcache_device *d, fmode_t mode,
diff --git a/drivers/md/bcache/request.h b/drivers/md/bcache/request.h
index 82b38366a95d..38ab4856eaab 100644
--- a/drivers/md/bcache/request.h
+++ b/drivers/md/bcache/request.h
@@ -37,10 +37,10 @@ unsigned int bch_get_congested(const struct cache_set *c);
 void bch_data_insert(struct closure *cl);
 
 void bch_cached_dev_request_init(struct cached_dev *dc);
-blk_qc_t cached_dev_submit_bio(struct bio *bio);
+void cached_dev_submit_bio(struct bio *bio);
 
 void bch_flash_dev_request_init(struct bcache_device *d);
-blk_qc_t flash_dev_submit_bio(struct bio *bio);
+void flash_dev_submit_bio(struct bio *bio);
 
 extern struct kmem_cache *bch_search_cache;
 
diff --git a/drivers/md/dm.c b/drivers/md/dm.c
index 50b693d776d6..0e3a9c8ac2ff 100644
--- a/drivers/md/dm.c
+++ b/drivers/md/dm.c
@@ -1294,14 +1294,13 @@ static noinline void __set_swap_bios_limit(struct mapped_device *md, int latch)
 	mutex_unlock(&md->swap_bios_lock);
 }
 
-static blk_qc_t __map_bio(struct dm_target_io *tio)
+static void __map_bio(struct dm_target_io *tio)
 {
 	int r;
 	sector_t sector;
 	struct bio *clone = &tio->clone;
 	struct dm_io *io = tio->io;
 	struct dm_target *ti = tio->ti;
-	blk_qc_t ret = BLK_QC_T_NONE;
 
 	clone->bi_end_io = clone_endio;
 
@@ -1328,7 +1327,7 @@ static blk_qc_t __map_bio(struct dm_target_io *tio)
 	case DM_MAPIO_REMAPPED:
 		/* the bio has been remapped so dispatch it */
 		trace_block_bio_remap(clone, bio_dev(io->orig_bio), sector);
-		ret = submit_bio_noacct(clone);
+		submit_bio_noacct(clone);
 		break;
 	case DM_MAPIO_KILL:
 		if (unlikely(swap_bios_limit(ti, clone))) {
@@ -1350,8 +1349,6 @@ static blk_qc_t __map_bio(struct dm_target_io *tio)
 		DMWARN("unimplemented target map return value: %d", r);
 		BUG();
 	}
-
-	return ret;
 }
 
 static void bio_setup_sector(struct bio *bio, sector_t sector, unsigned len)
@@ -1438,7 +1435,7 @@ static void alloc_multiple_bios(struct bio_list *blist, struct clone_info *ci,
 	}
 }
 
-static blk_qc_t __clone_and_map_simple_bio(struct clone_info *ci,
+static void __clone_and_map_simple_bio(struct clone_info *ci,
 					   struct dm_target_io *tio, unsigned *len)
 {
 	struct bio *clone = &tio->clone;
@@ -1448,8 +1445,7 @@ static blk_qc_t __clone_and_map_simple_bio(struct clone_info *ci,
 	__bio_clone_fast(clone, ci->bio);
 	if (len)
 		bio_setup_sector(clone, ci->sector, *len);
-
-	return __map_bio(tio);
+	__map_bio(tio);
 }
 
 static void __send_duplicate_bios(struct clone_info *ci, struct dm_target *ti,
@@ -1463,7 +1459,7 @@ static void __send_duplicate_bios(struct clone_info *ci, struct dm_target *ti,
 
 	while ((bio = bio_list_pop(&blist))) {
 		tio = container_of(bio, struct dm_target_io, clone);
-		(void) __clone_and_map_simple_bio(ci, tio, len);
+		__clone_and_map_simple_bio(ci, tio, len);
 	}
 }
 
@@ -1507,7 +1503,7 @@ static int __clone_and_map_data_bio(struct clone_info *ci, struct dm_target *ti,
 		free_tio(tio);
 		return r;
 	}
-	(void) __map_bio(tio);
+	__map_bio(tio);
 
 	return 0;
 }
@@ -1622,11 +1618,10 @@ static void init_clone_info(struct clone_info *ci, struct mapped_device *md,
 /*
  * Entry point to split a bio into clones and submit them to the targets.
  */
-static blk_qc_t __split_and_process_bio(struct mapped_device *md,
+static void __split_and_process_bio(struct mapped_device *md,
 					struct dm_table *map, struct bio *bio)
 {
 	struct clone_info ci;
-	blk_qc_t ret = BLK_QC_T_NONE;
 	int error = 0;
 
 	init_clone_info(&ci, md, map, bio);
@@ -1670,7 +1665,7 @@ static blk_qc_t __split_and_process_bio(struct mapped_device *md,
 
 				bio_chain(b, bio);
 				trace_block_split(b, bio->bi_iter.bi_sector);
-				ret = submit_bio_noacct(bio);
+				submit_bio_noacct(bio);
 				break;
 			}
 		}
@@ -1678,13 +1673,11 @@ static blk_qc_t __split_and_process_bio(struct mapped_device *md,
 
 	/* drop the extra reference count */
 	dec_pending(ci.io, errno_to_blk_status(error));
-	return ret;
 }
 
-static blk_qc_t dm_submit_bio(struct bio *bio)
+static void dm_submit_bio(struct bio *bio)
 {
 	struct mapped_device *md = bio->bi_bdev->bd_disk->private_data;
-	blk_qc_t ret = BLK_QC_T_NONE;
 	int srcu_idx;
 	struct dm_table *map;
 
@@ -1714,10 +1707,9 @@ static blk_qc_t dm_submit_bio(struct bio *bio)
 	if (is_abnormal_io(bio))
 		blk_queue_split(&bio);
 
-	ret = __split_and_process_bio(md, map, bio);
+	__split_and_process_bio(md, map, bio);
 out:
 	dm_put_live_table(md, srcu_idx);
-	return ret;
 }
 
 /*-----------------------------------------------------------------
diff --git a/drivers/md/md.c b/drivers/md/md.c
index 21da0c48f6c2..2953eb308055 100644
--- a/drivers/md/md.c
+++ b/drivers/md/md.c
@@ -465,19 +465,19 @@ static void md_end_io(struct bio *bio)
 		bio->bi_end_io(bio);
 }
 
-static blk_qc_t md_submit_bio(struct bio *bio)
+static void md_submit_bio(struct bio *bio)
 {
 	const int rw = bio_data_dir(bio);
 	struct mddev *mddev = bio->bi_bdev->bd_disk->private_data;
 
 	if (mddev == NULL || mddev->pers == NULL) {
 		bio_io_error(bio);
-		return BLK_QC_T_NONE;
+		return;
 	}
 
 	if (unlikely(test_bit(MD_BROKEN, &mddev->flags)) && (rw == WRITE)) {
 		bio_io_error(bio);
-		return BLK_QC_T_NONE;
+		return;
 	}
 
 	blk_queue_split(&bio);
@@ -486,7 +486,7 @@ static blk_qc_t md_submit_bio(struct bio *bio)
 		if (bio_sectors(bio) != 0)
 			bio->bi_status = BLK_STS_IOERR;
 		bio_endio(bio);
-		return BLK_QC_T_NONE;
+		return;
 	}
 
 	if (bio->bi_end_io != md_end_io) {
@@ -508,8 +508,6 @@ static blk_qc_t md_submit_bio(struct bio *bio)
 	bio->bi_opf &= ~REQ_NOMERGE;
 
 	md_handle_request(mddev, bio);
-
-	return BLK_QC_T_NONE;
 }
 
 /* mddev_suspend makes sure no new requests are submitted
diff --git a/drivers/nvdimm/blk.c b/drivers/nvdimm/blk.c
index 7b9556291eb1..90c202cd39dd 100644
--- a/drivers/nvdimm/blk.c
+++ b/drivers/nvdimm/blk.c
@@ -162,7 +162,7 @@ static int nsblk_do_bvec(struct nd_namespace_blk *nsblk,
 	return err;
 }
 
-static blk_qc_t nd_blk_submit_bio(struct bio *bio)
+static void nd_blk_submit_bio(struct bio *bio)
 {
 	struct bio_integrity_payload *bip;
 	struct nd_namespace_blk *nsblk = bio->bi_bdev->bd_disk->private_data;
@@ -173,7 +173,7 @@ static blk_qc_t nd_blk_submit_bio(struct bio *bio)
 	bool do_acct;
 
 	if (!bio_integrity_prep(bio))
-		return BLK_QC_T_NONE;
+		return;
 
 	bip = bio_integrity(bio);
 	rw = bio_data_dir(bio);
@@ -199,7 +199,6 @@ static blk_qc_t nd_blk_submit_bio(struct bio *bio)
 		bio_end_io_acct(bio, start);
 
 	bio_endio(bio);
-	return BLK_QC_T_NONE;
 }
 
 static int nsblk_rw_bytes(struct nd_namespace_common *ndns,
diff --git a/drivers/nvdimm/btt.c b/drivers/nvdimm/btt.c
index 41aa1f01fc07..cd3509e69974 100644
--- a/drivers/nvdimm/btt.c
+++ b/drivers/nvdimm/btt.c
@@ -1439,7 +1439,7 @@ static int btt_do_bvec(struct btt *btt, struct bio_integrity_payload *bip,
 	return ret;
 }
 
-static blk_qc_t btt_submit_bio(struct bio *bio)
+static void btt_submit_bio(struct bio *bio)
 {
 	struct bio_integrity_payload *bip = bio_integrity(bio);
 	struct btt *btt = bio->bi_bdev->bd_disk->private_data;
@@ -1450,7 +1450,7 @@ static blk_qc_t btt_submit_bio(struct bio *bio)
 	bool do_acct;
 
 	if (!bio_integrity_prep(bio))
-		return BLK_QC_T_NONE;
+		return;
 
 	do_acct = blk_queue_io_stat(bio->bi_bdev->bd_disk->queue);
 	if (do_acct)
@@ -1482,7 +1482,6 @@ static blk_qc_t btt_submit_bio(struct bio *bio)
 		bio_end_io_acct(bio, start);
 
 	bio_endio(bio);
-	return BLK_QC_T_NONE;
 }
 
 static int btt_rw_page(struct block_device *bdev, sector_t sector,
diff --git a/drivers/nvdimm/pmem.c b/drivers/nvdimm/pmem.c
index b8a85bfb2e95..55393ef11055 100644
--- a/drivers/nvdimm/pmem.c
+++ b/drivers/nvdimm/pmem.c
@@ -188,7 +188,7 @@ static blk_status_t pmem_do_write(struct pmem_device *pmem,
 	return rc;
 }
 
-static blk_qc_t pmem_submit_bio(struct bio *bio)
+static void pmem_submit_bio(struct bio *bio)
 {
 	int ret = 0;
 	blk_status_t rc = 0;
@@ -227,7 +227,6 @@ static blk_qc_t pmem_submit_bio(struct bio *bio)
 		bio->bi_status = errno_to_blk_status(ret);
 
 	bio_endio(bio);
-	return BLK_QC_T_NONE;
 }
 
 static int pmem_rw_page(struct block_device *bdev, sector_t sector,
diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
index 723a4c382385..b0da256e7c48 100644
--- a/drivers/nvme/host/core.c
+++ b/drivers/nvme/host/core.c
@@ -963,7 +963,7 @@ static void nvme_execute_rq_polled(struct request_queue *q,
 	blk_execute_rq_nowait(bd_disk, rq, at_head, nvme_end_sync_rq);
 
 	while (!completion_done(&wait)) {
-		blk_poll(q, request_to_qc_t(rq->mq_hctx, rq), true);
+		bio_poll(rq->bio, true);
 		cond_resched();
 	}
 }
diff --git a/drivers/nvme/host/multipath.c b/drivers/nvme/host/multipath.c
index a1d476e1ac02..7ccea6063cb5 100644
--- a/drivers/nvme/host/multipath.c
+++ b/drivers/nvme/host/multipath.c
@@ -294,12 +294,11 @@ static bool nvme_available_path(struct nvme_ns_head *head)
 	return false;
 }
 
-blk_qc_t nvme_ns_head_submit_bio(struct bio *bio)
+void nvme_ns_head_submit_bio(struct bio *bio)
 {
 	struct nvme_ns_head *head = bio->bi_bdev->bd_disk->private_data;
 	struct device *dev = disk_to_dev(head->disk);
 	struct nvme_ns *ns;
-	blk_qc_t ret = BLK_QC_T_NONE;
 	int srcu_idx;
 
 	/*
@@ -316,7 +315,7 @@ blk_qc_t nvme_ns_head_submit_bio(struct bio *bio)
 		bio->bi_opf |= REQ_NVME_MPATH;
 		trace_block_bio_remap(bio, disk_devt(ns->head->disk),
 				      bio->bi_iter.bi_sector);
-		ret = submit_bio_noacct(bio);
+		submit_bio_noacct(bio);
 	} else if (nvme_available_path(head)) {
 		dev_warn_ratelimited(dev, "no usable path - requeuing I/O\n");
 
@@ -331,7 +330,6 @@ blk_qc_t nvme_ns_head_submit_bio(struct bio *bio)
 	}
 
 	srcu_read_unlock(&head->srcu, srcu_idx);
-	return ret;
 }
 
 static void nvme_requeue_work(struct work_struct *work)
diff --git a/drivers/nvme/host/nvme.h b/drivers/nvme/host/nvme.h
index 07b34175c6ce..29368b930fe0 100644
--- a/drivers/nvme/host/nvme.h
+++ b/drivers/nvme/host/nvme.h
@@ -674,7 +674,7 @@ void nvme_mpath_stop(struct nvme_ctrl *ctrl);
 bool nvme_mpath_clear_current_path(struct nvme_ns *ns);
 void nvme_mpath_clear_ctrl_paths(struct nvme_ctrl *ctrl);
 struct nvme_ns *nvme_find_path(struct nvme_ns_head *head);
-blk_qc_t nvme_ns_head_submit_bio(struct bio *bio);
+void nvme_ns_head_submit_bio(struct bio *bio);
 
 static inline void nvme_mpath_check_last_path(struct nvme_ns *ns)
 {
diff --git a/drivers/s390/block/dcssblk.c b/drivers/s390/block/dcssblk.c
index da33cb4cba28..f17582ade60b 100644
--- a/drivers/s390/block/dcssblk.c
+++ b/drivers/s390/block/dcssblk.c
@@ -31,7 +31,7 @@
 
 static int dcssblk_open(struct block_device *bdev, fmode_t mode);
 static void dcssblk_release(struct gendisk *disk, fmode_t mode);
-static blk_qc_t dcssblk_submit_bio(struct bio *bio);
+static void dcssblk_submit_bio(struct bio *bio);
 static long dcssblk_dax_direct_access(struct dax_device *dax_dev, pgoff_t pgoff,
 		long nr_pages, void **kaddr, pfn_t *pfn);
 
@@ -865,7 +865,7 @@ dcssblk_release(struct gendisk *disk, fmode_t mode)
 	up_write(&dcssblk_devices_sem);
 }
 
-static blk_qc_t
+static void
 dcssblk_submit_bio(struct bio *bio)
 {
 	struct dcssblk_dev_info *dev_info;
@@ -919,10 +919,9 @@ dcssblk_submit_bio(struct bio *bio)
 		bytes_done += bvec.bv_len;
 	}
 	bio_endio(bio);
-	return BLK_QC_T_NONE;
+	return;
 fail:
 	bio_io_error(bio);
-	return BLK_QC_T_NONE;
 }
 
 static long
diff --git a/drivers/s390/block/xpram.c b/drivers/s390/block/xpram.c
index d1ed39162943..bb129e31df77 100644
--- a/drivers/s390/block/xpram.c
+++ b/drivers/s390/block/xpram.c
@@ -182,7 +182,7 @@ static unsigned long xpram_highest_page_index(void)
 /*
  * Block device make request function.
  */
-static blk_qc_t xpram_submit_bio(struct bio *bio)
+static void xpram_submit_bio(struct bio *bio)
 {
 	xpram_device_t *xdev = bio->bi_bdev->bd_disk->private_data;
 	struct bio_vec bvec;
@@ -224,10 +224,9 @@ static blk_qc_t xpram_submit_bio(struct bio *bio)
 		}
 	}
 	bio_endio(bio);
-	return BLK_QC_T_NONE;
+	return;
 fail:
 	bio_io_error(bio);
-	return BLK_QC_T_NONE;
 }
 
 static int xpram_getgeo(struct block_device *bdev, struct hd_geometry *geo)
diff --git a/fs/block_dev.c b/fs/block_dev.c
index ea5b4617ff86..95f047060df5 100644
--- a/fs/block_dev.c
+++ b/fs/block_dev.c
@@ -239,7 +239,6 @@ __blkdev_direct_IO_simple(struct kiocb *iocb, struct iov_iter *iter,
 	bool should_dirty = false;
 	struct bio bio;
 	ssize_t ret;
-	blk_qc_t qc;
 
 	if ((pos | iov_iter_alignment(iter)) &
 	    (bdev_logical_block_size(bdev) - 1))
@@ -278,13 +277,12 @@ __blkdev_direct_IO_simple(struct kiocb *iocb, struct iov_iter *iter,
 	if (iocb->ki_flags & IOCB_HIPRI)
 		bio_set_polled(&bio, iocb);
 
-	qc = submit_bio(&bio);
+	submit_bio(&bio);
 	for (;;) {
 		set_current_state(TASK_UNINTERRUPTIBLE);
 		if (!READ_ONCE(bio.bi_private))
 			break;
-		if (!(iocb->ki_flags & IOCB_HIPRI) ||
-		    !blk_poll(bdev_get_queue(bdev), qc, true))
+		if (!(iocb->ki_flags & IOCB_HIPRI) || !bio_poll(&bio, true))
 			blk_io_schedule();
 	}
 	__set_current_state(TASK_RUNNING);
@@ -317,14 +315,6 @@ struct blkdev_dio {
 
 static struct bio_set blkdev_dio_pool;
 
-static int blkdev_iopoll(struct kiocb *kiocb, bool wait)
-{
-	struct block_device *bdev = I_BDEV(kiocb->ki_filp->f_mapping->host);
-	struct request_queue *q = bdev_get_queue(bdev);
-
-	return blk_poll(q, READ_ONCE(kiocb->ki_cookie), wait);
-}
-
 static void blkdev_bio_end_io(struct bio *bio)
 {
 	struct blkdev_dio *dio = bio->bi_private;
@@ -338,6 +328,8 @@ static void blkdev_bio_end_io(struct bio *bio)
 			struct kiocb *iocb = dio->iocb;
 			ssize_t ret;
 
+			WRITE_ONCE(iocb->private, NULL);
+
 			if (likely(!dio->bio.bi_status)) {
 				ret = dio->size;
 				iocb->ki_pos += ret;
@@ -376,7 +368,6 @@ static ssize_t __blkdev_direct_IO(struct kiocb *iocb, struct iov_iter *iter,
 	bool is_poll = (iocb->ki_flags & IOCB_HIPRI), do_poll = false;
 	bool is_read = (iov_iter_rw(iter) == READ), is_sync;
 	loff_t pos = iocb->ki_pos;
-	blk_qc_t qc = BLK_QC_T_NONE;
 	int ret = 0;
 
 	if ((pos | iov_iter_alignment(iter)) &
@@ -449,10 +440,10 @@ static ssize_t __blkdev_direct_IO(struct kiocb *iocb, struct iov_iter *iter,
 			bio_set_polled(bio, iocb);
 			do_poll = true;
 		}
-		qc = submit_bio(bio);
+		submit_bio(bio);
 		if (!nr_pages) {
 			if (do_poll)
-				WRITE_ONCE(iocb->ki_cookie, qc);
+				WRITE_ONCE(iocb->private, bio);
 			break;
 		}
 		bio = bio_alloc(GFP_KERNEL, nr_pages);
@@ -469,7 +460,7 @@ static ssize_t __blkdev_direct_IO(struct kiocb *iocb, struct iov_iter *iter,
 		if (!READ_ONCE(dio->waiter))
 			break;
 
-		if (!do_poll || !blk_poll(bdev_get_queue(bdev), qc, true))
+		if (!do_poll || !bio_poll(bio, true))
 			blk_io_schedule();
 	}
 	__set_current_state(TASK_RUNNING);
@@ -1823,7 +1814,7 @@ const struct file_operations def_blk_fops = {
 	.llseek		= block_llseek,
 	.read_iter	= blkdev_read_iter,
 	.write_iter	= blkdev_write_iter,
-	.iopoll		= blkdev_iopoll,
+	.iopoll		= iocb_bio_iopoll,
 	.mmap		= generic_file_mmap,
 	.fsync		= blkdev_fsync,
 	.unlocked_ioctl	= block_ioctl,
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 7cdf65be3707..e9e39209c326 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -8095,7 +8095,7 @@ static struct btrfs_dio_private *btrfs_create_dio_private(struct bio *dio_bio,
 	return dip;
 }
 
-static blk_qc_t btrfs_submit_direct(struct inode *inode, struct iomap *iomap,
+static void btrfs_submit_direct(struct inode *inode, struct iomap *iomap,
 		struct bio *dio_bio, loff_t file_offset)
 {
 	const bool write = (btrfs_op(dio_bio) == BTRFS_MAP_WRITE);
@@ -8124,7 +8124,7 @@ static blk_qc_t btrfs_submit_direct(struct inode *inode, struct iomap *iomap,
 		}
 		dio_bio->bi_status = BLK_STS_RESOURCE;
 		bio_endio(dio_bio);
-		return BLK_QC_T_NONE;
+		return;
 	}
 
 	if (!write) {
@@ -8222,15 +8222,13 @@ static blk_qc_t btrfs_submit_direct(struct inode *inode, struct iomap *iomap,
 
 		free_extent_map(em);
 	} while (submit_len > 0);
-	return BLK_QC_T_NONE;
+	return;
 
 out_err_em:
 	free_extent_map(em);
 out_err:
 	dip->dio_bio->bi_status = status;
 	btrfs_dio_private_put(dip);
-
-	return BLK_QC_T_NONE;
 }
 
 const struct iomap_ops btrfs_dio_iomap_ops = {
diff --git a/fs/ext4/file.c b/fs/ext4/file.c
index 194f5d00fa32..a3b093c9d2a4 100644
--- a/fs/ext4/file.c
+++ b/fs/ext4/file.c
@@ -896,7 +896,7 @@ const struct file_operations ext4_file_operations = {
 	.llseek		= ext4_llseek,
 	.read_iter	= ext4_file_read_iter,
 	.write_iter	= ext4_file_write_iter,
-	.iopoll		= iomap_dio_iopoll,
+	.iopoll		= iocb_bio_iopoll,
 	.unlocked_ioctl = ext4_ioctl,
 #ifdef CONFIG_COMPAT
 	.compat_ioctl	= ext4_compat_ioctl,
diff --git a/fs/gfs2/file.c b/fs/gfs2/file.c
index 2d500f90cdac..a2d120430ea9 100644
--- a/fs/gfs2/file.c
+++ b/fs/gfs2/file.c
@@ -1359,7 +1359,7 @@ const struct file_operations gfs2_file_fops = {
 	.llseek		= gfs2_llseek,
 	.read_iter	= gfs2_file_read_iter,
 	.write_iter	= gfs2_file_write_iter,
-	.iopoll		= iomap_dio_iopoll,
+	.iopoll		= iocb_bio_iopoll,
 	.unlocked_ioctl	= gfs2_ioctl,
 	.compat_ioctl	= gfs2_compat_ioctl,
 	.mmap		= gfs2_mmap,
@@ -1392,7 +1392,7 @@ const struct file_operations gfs2_file_fops_nolock = {
 	.llseek		= gfs2_llseek,
 	.read_iter	= gfs2_file_read_iter,
 	.write_iter	= gfs2_file_write_iter,
-	.iopoll		= iomap_dio_iopoll,
+	.iopoll		= iocb_bio_iopoll,
 	.unlocked_ioctl	= gfs2_ioctl,
 	.compat_ioctl	= gfs2_compat_ioctl,
 	.mmap		= gfs2_mmap,
diff --git a/fs/iomap/direct-io.c b/fs/iomap/direct-io.c
index 357419f39654..19c98ee947ea 100644
--- a/fs/iomap/direct-io.c
+++ b/fs/iomap/direct-io.c
@@ -38,8 +38,7 @@ struct iomap_dio {
 		struct {
 			struct iov_iter		*iter;
 			struct task_struct	*waiter;
-			struct request_queue	*last_queue;
-			blk_qc_t		cookie;
+			struct bio		*poll_bio;
 		} submit;
 
 		/* used for aio completion: */
@@ -49,31 +48,21 @@ struct iomap_dio {
 	};
 };
 
-int iomap_dio_iopoll(struct kiocb *kiocb, bool spin)
-{
-	struct request_queue *q = READ_ONCE(kiocb->private);
-
-	if (!q)
-		return 0;
-	return blk_poll(q, READ_ONCE(kiocb->ki_cookie), spin);
-}
-EXPORT_SYMBOL_GPL(iomap_dio_iopoll);
-
 static void iomap_dio_submit_bio(struct iomap_dio *dio, struct iomap *iomap,
 		struct bio *bio, loff_t pos)
 {
 	atomic_inc(&dio->ref);
 
-	if (dio->iocb->ki_flags & IOCB_HIPRI)
+	if (dio->iocb->ki_flags & IOCB_HIPRI) {
 		bio_set_polled(bio, dio->iocb);
+		dio->submit.poll_bio = bio;
+	}
 
-	dio->submit.last_queue = bdev_get_queue(iomap->bdev);
 	if (dio->dops && dio->dops->submit_io)
-		dio->submit.cookie = dio->dops->submit_io(
-				file_inode(dio->iocb->ki_filp),
-				iomap, bio, pos);
+		dio->dops->submit_io(file_inode(dio->iocb->ki_filp), iomap, bio,
+				     pos);
 	else
-		dio->submit.cookie = submit_bio(bio);
+		submit_bio(bio);
 }
 
 ssize_t iomap_dio_complete(struct iomap_dio *dio)
@@ -166,9 +155,11 @@ static void iomap_dio_bio_end_io(struct bio *bio)
 		} else if (dio->flags & IOMAP_DIO_WRITE) {
 			struct inode *inode = file_inode(dio->iocb->ki_filp);
 
+			WRITE_ONCE(dio->iocb->private, NULL);
 			INIT_WORK(&dio->aio.work, iomap_dio_complete_work);
 			queue_work(inode->i_sb->s_dio_done_wq, &dio->aio.work);
 		} else {
+			WRITE_ONCE(dio->iocb->private, NULL);
 			iomap_dio_complete_work(&dio->aio.work);
 		}
 	}
@@ -492,8 +483,7 @@ __iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
 
 	dio->submit.iter = iter;
 	dio->submit.waiter = current;
-	dio->submit.cookie = BLK_QC_T_NONE;
-	dio->submit.last_queue = NULL;
+	dio->submit.poll_bio = NULL;
 
 	if (iov_iter_rw(iter) == READ) {
 		if (pos >= dio->i_size)
@@ -601,8 +591,7 @@ __iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
 	if (dio->flags & IOMAP_DIO_WRITE_FUA)
 		dio->flags &= ~IOMAP_DIO_NEED_SYNC;
 
-	WRITE_ONCE(iocb->ki_cookie, dio->submit.cookie);
-	WRITE_ONCE(iocb->private, dio->submit.last_queue);
+	WRITE_ONCE(iocb->private, dio->submit.poll_bio);
 
 	/*
 	 * We are about to drop our additional submission reference, which
@@ -629,10 +618,8 @@ __iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
 			if (!READ_ONCE(dio->submit.waiter))
 				break;
 
-			if (!(iocb->ki_flags & IOCB_HIPRI) ||
-			    !dio->submit.last_queue ||
-			    !blk_poll(dio->submit.last_queue,
-					 dio->submit.cookie, true))
+			if (!dio->submit.poll_bio ||
+			    !bio_poll(dio->submit.poll_bio, true))
 				blk_io_schedule();
 		}
 		__set_current_state(TASK_RUNNING);
diff --git a/fs/xfs/xfs_file.c b/fs/xfs/xfs_file.c
index a007ca0711d9..6bebee96ab65 100644
--- a/fs/xfs/xfs_file.c
+++ b/fs/xfs/xfs_file.c
@@ -1429,7 +1429,7 @@ const struct file_operations xfs_file_operations = {
 	.write_iter	= xfs_file_write_iter,
 	.splice_read	= generic_file_splice_read,
 	.splice_write	= iter_file_splice_write,
-	.iopoll		= iomap_dio_iopoll,
+	.iopoll		= iocb_bio_iopoll,
 	.unlocked_ioctl	= xfs_file_ioctl,
 #ifdef CONFIG_COMPAT
 	.compat_ioctl	= xfs_file_compat_ioctl,
diff --git a/fs/zonefs/super.c b/fs/zonefs/super.c
index 049e36c69ed7..633e4890dfe4 100644
--- a/fs/zonefs/super.c
+++ b/fs/zonefs/super.c
@@ -1143,7 +1143,7 @@ static const struct file_operations zonefs_file_operations = {
 	.write_iter	= zonefs_file_write_iter,
 	.splice_read	= generic_file_splice_read,
 	.splice_write	= iter_file_splice_write,
-	.iopoll		= iomap_dio_iopoll,
+	.iopoll		= iocb_bio_iopoll,
 };
 
 static struct kmem_cache *zonefs_inode_cachep;
diff --git a/include/linux/bio.h b/include/linux/bio.h
index 439a70bc42e2..37a63c0b5c39 100644
--- a/include/linux/bio.h
+++ b/include/linux/bio.h
@@ -426,7 +426,7 @@ static inline struct bio *bio_alloc(gfp_t gfp_mask, unsigned short nr_iovecs)
 	return bio_alloc_bioset(gfp_mask, nr_iovecs, &fs_bio_set);
 }
 
-extern blk_qc_t submit_bio(struct bio *);
+void submit_bio(struct bio *bio);
 
 extern void bio_endio(struct bio *);
 
diff --git a/include/linux/blk-mq.h b/include/linux/blk-mq.h
index 2c473c9b8990..69f509bf6761 100644
--- a/include/linux/blk-mq.h
+++ b/include/linux/blk-mq.h
@@ -148,9 +148,9 @@ struct blk_mq_hw_ctx {
 	/** @kobj: Kernel object for sysfs. */
 	struct kobject		kobj;
 
-	/** @poll_considered: Count times blk_poll() was called. */
+	/** @poll_considered: Count times blk_mq_poll() was called. */
 	unsigned long		poll_considered;
-	/** @poll_invoked: Count how many requests blk_poll() polled. */
+	/** @poll_invoked: Count how many requests blk_mq_poll() polled. */
 	unsigned long		poll_invoked;
 	/** @poll_success: Count how many polled requests were completed. */
 	unsigned long		poll_success;
@@ -586,16 +586,6 @@ static inline void *blk_mq_rq_to_pdu(struct request *rq)
 	for ((i) = 0; (i) < (hctx)->nr_ctx &&				\
 	     ({ ctx = (hctx)->ctxs[(i)]; 1; }); (i)++)
 
-static inline blk_qc_t request_to_qc_t(struct blk_mq_hw_ctx *hctx,
-		struct request *rq)
-{
-	if (rq->tag != -1)
-		return rq->tag | (hctx->queue_num << BLK_QC_T_SHIFT);
-
-	return rq->internal_tag | (hctx->queue_num << BLK_QC_T_SHIFT) |
-			BLK_QC_T_INTERNAL;
-}
-
 static inline void blk_mq_cleanup_rq(struct request *rq)
 {
 	if (rq->q->mq_ops->cleanup_rq)
@@ -614,7 +604,6 @@ static inline void blk_rq_bio_prep(struct request *rq, struct bio *bio,
 		rq->rq_disk = bio->bi_bdev->bd_disk;
 }
 
-blk_qc_t blk_mq_submit_bio(struct bio *bio);
 void blk_mq_hctx_set_fq_lock_class(struct blk_mq_hw_ctx *hctx,
 		struct lock_class_key *key);
 
diff --git a/include/linux/blk_types.h b/include/linux/blk_types.h
index 183a76bf24b7..e0ca0cb9c94a 100644
--- a/include/linux/blk_types.h
+++ b/include/linux/blk_types.h
@@ -217,6 +217,9 @@ static inline void bio_issue_init(struct bio_issue *issue,
 			((u64)size << BIO_ISSUE_SIZE_SHIFT));
 }
 
+typedef unsigned int blk_qc_t;
+#define BLK_QC_T_NONE		-1U
+
 /*
  * main unit of I/O for the block layer and lower layers (ie drivers and
  * stacking drivers)
@@ -238,9 +241,9 @@ struct bio {
 		struct bvec_iter	bi_iter;
 		struct rcu_head		bi_rcu_free;
 	};
+	blk_qc_t		bi_cookie;
 
 	bio_end_io_t		*bi_end_io;
-
 	void			*bi_private;
 #ifdef CONFIG_BLK_CGROUP
 	/*
@@ -396,7 +399,7 @@ enum req_flag_bits {
 	/* command specific flags for REQ_OP_WRITE_ZEROES: */
 	__REQ_NOUNMAP,		/* do not free blocks when zeroing */
 
-	__REQ_POLLED,		/* caller polls for completion using blk_poll */
+	__REQ_POLLED,		/* caller polls for completion using bio_poll */
 
 	/* for driver use */
 	__REQ_DRV,
@@ -509,11 +512,6 @@ static inline int op_stat_group(unsigned int op)
 	return op_is_write(op);
 }
 
-typedef unsigned int blk_qc_t;
-#define BLK_QC_T_NONE		-1U
-#define BLK_QC_T_SHIFT		16
-#define BLK_QC_T_INTERNAL	(1U << 31)
-
 struct blk_rq_stat {
 	u64 mean;
 	u64 min;
diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
index f2e77ba97550..d395c23206a4 100644
--- a/include/linux/blkdev.h
+++ b/include/linux/blkdev.h
@@ -38,6 +38,7 @@ struct sg_io_hdr;
 struct bsg_job;
 struct blkcg_gq;
 struct blk_flush_queue;
+struct kiocb;
 struct pr_ops;
 struct rq_qos;
 struct blk_queue_stats;
@@ -904,7 +905,7 @@ static inline void rq_flush_dcache_pages(struct request *rq)
 
 extern int blk_register_queue(struct gendisk *disk);
 extern void blk_unregister_queue(struct gendisk *disk);
-blk_qc_t submit_bio_noacct(struct bio *bio);
+void submit_bio_noacct(struct bio *bio);
 extern void blk_rq_init(struct request_queue *q, struct request *rq);
 extern void blk_put_request(struct request *);
 extern struct request *blk_get_request(struct request_queue *, unsigned int op,
@@ -950,7 +951,8 @@ extern const char *blk_op_str(unsigned int op);
 int blk_status_to_errno(blk_status_t status);
 blk_status_t errno_to_blk_status(int errno);
 
-int blk_poll(struct request_queue *q, blk_qc_t cookie, bool spin);
+int bio_poll(struct bio *bio, bool spin);
+int iocb_bio_iopoll(struct kiocb *kiocb, bool spin);
 
 static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
 {
@@ -1855,7 +1857,7 @@ static inline void blk_ksm_unregister(struct request_queue *q) { }
 
 
 struct block_device_operations {
-	blk_qc_t (*submit_bio) (struct bio *bio);
+	void (*submit_bio)(struct bio *bio);
 	int (*open) (struct block_device *, fmode_t);
 	void (*release) (struct gendisk *, fmode_t);
 	int (*rw_page)(struct block_device *, sector_t, struct page *, unsigned int);
diff --git a/include/linux/fs.h b/include/linux/fs.h
index ec8f3ddf4a6a..279522d05a9d 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -331,11 +331,7 @@ struct kiocb {
 	int			ki_flags;
 	u16			ki_hint;
 	u16			ki_ioprio; /* See linux/ioprio.h */
-	union {
-		unsigned int		ki_cookie; /* for ->iopoll */
-		struct wait_page_queue	*ki_waitq; /* for async buffered IO */
-	};
-
+	struct wait_page_queue	*ki_waitq; /* for async buffered IO */
 	randomized_struct_fields_end
 };
 
diff --git a/include/linux/iomap.h b/include/linux/iomap.h
index d202fd2d0f91..6f94da85163d 100644
--- a/include/linux/iomap.h
+++ b/include/linux/iomap.h
@@ -254,7 +254,7 @@ int iomap_writepages(struct address_space *mapping,
 struct iomap_dio_ops {
 	int (*end_io)(struct kiocb *iocb, ssize_t size, int error,
 		      unsigned flags);
-	blk_qc_t (*submit_io)(struct inode *inode, struct iomap *iomap,
+	void (*submit_io)(struct inode *inode, struct iomap *iomap,
 			struct bio *bio, loff_t file_offset);
 };
 
@@ -278,7 +278,6 @@ struct iomap_dio *__iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
 		const struct iomap_ops *ops, const struct iomap_dio_ops *dops,
 		unsigned int dio_flags);
 ssize_t iomap_dio_complete(struct iomap_dio *dio);
-int iomap_dio_iopoll(struct kiocb *kiocb, bool spin);
 
 #ifdef CONFIG_SWAP
 struct file;
diff --git a/mm/page_io.c b/mm/page_io.c
index dd86ea217da2..5b5259fc2c2a 100644
--- a/mm/page_io.c
+++ b/mm/page_io.c
@@ -358,8 +358,6 @@ int swap_readpage(struct page *page, bool synchronous)
 	struct bio *bio;
 	int ret = 0;
 	struct swap_info_struct *sis = page_swap_info(page);
-	blk_qc_t qc;
-	struct gendisk *disk;
 	unsigned long pflags;
 
 	VM_BUG_ON_PAGE(!PageSwapCache(page) && !synchronous, page);
@@ -409,8 +407,6 @@ int swap_readpage(struct page *page, bool synchronous)
 	bio->bi_iter.bi_sector = swap_page_sector(page);
 	bio->bi_end_io = end_swap_bio_read;
 	bio_add_page(bio, page, thp_size(page), 0);
-
-	disk = bio->bi_bdev->bd_disk;
 	/*
 	 * Keep this task valid during swap readpage because the oom killer may
 	 * attempt to access it in the page fault retry time check.
@@ -422,13 +418,13 @@ int swap_readpage(struct page *page, bool synchronous)
 	}
 	count_vm_event(PSWPIN);
 	bio_get(bio);
-	qc = submit_bio(bio);
+	submit_bio(bio);
 	while (synchronous) {
 		set_current_state(TASK_UNINTERRUPTIBLE);
 		if (!READ_ONCE(bio->bi_private))
 			break;
 
-		if (!blk_poll(disk->queue, qc, true))
+		if (!bio_poll(bio, true))
 			blk_io_schedule();
 	}
 	__set_current_state(TASK_RUNNING);
-- 
2.30.1

Re: switch block layer polling to a bio based model

[Jens Axboe]

On 4/26/21 7:48 AM, Christoph Hellwig wrote:
> Hi all,
> 
> This series clean up the block polling code a bit and changes the interface
> to poll for a specific bio instead of a request_queue and cookie pair.
> 
> Polling for the bio itself leads to a few advantages:
> 
>   - the cookie construction can made entirely private in blk-mq.c
>   - the caller does not need to remember the request_queue and cookie
>     separately and thus sidesteps their lifetime issues
>   - keeping the device and the cookie inside the bio allows to trivially
>     support polling BIOs remapping by stacking drivers
>   - a lot of code to propagate the cookie back up the submission path can
>     removed entirely
> 
> The one major caveat is that this requires RCU freeing polled BIOs to make
> sure the bio that contains the polling information is still alive when
> io_uring tries to poll it through the iocb. For synchronous polling all the
> callers have a bio reference anyway, so this is not an issue.

Was curious about this separately, so ran a quick test on it. Running polled
IO on a fast device, performance drops about 10% with this applied. Outside
of that, we have ksoftirqd using 5-7% of CPU continually, just doing frees:

+   45.33%  ksoftirqd/0  [kernel.vmlinux]  [k] __slab_free
+   15.91%  ksoftirqd/0  [kernel.vmlinux]  [k] kmem_cache_free
+   12.66%  ksoftirqd/0  [kernel.vmlinux]  [k] rcu_cblist_dequeue
+    8.39%  ksoftirqd/0  [kernel.vmlinux]  [k] rcu_core
+    4.75%  ksoftirqd/0  [kernel.vmlinux]  [k] free_one_page
+    3.27%  ksoftirqd/0  [kernel.vmlinux]  [k] bio_free_rcu
+    1.98%  ksoftirqd/0  [kernel.vmlinux]  [k] mempool_free_slab

This all means that we go from 2.97M IOPS to 2.70M IOPS in that
particular test (QD=128, async polled).

I was separately curious about this as I have a (as of yet unposted)
patchset that recycles bio allocations, as we spend quite a bit of time
doing that for high rate polled IO. It's good for taking the above 2.97M
IOPS to 3.2-3.3M IOPS, and it'd obviously be a bit more problematic with
required RCU freeing of bio's. Even without the alloc cache, using RCU
will ruin any potential cache locality on back-to-back bio free + bio
alloc.

-- 
Jens Axboe

Re: switch block layer polling to a bio based model

[Christoph Hellwig]

On Mon, Apr 26, 2021 at 08:57:31AM -0600, Jens Axboe wrote:
> I was separately curious about this as I have a (as of yet unposted)
> patchset that recycles bio allocations, as we spend quite a bit of time
> doing that for high rate polled IO. It's good for taking the above 2.97M
> IOPS to 3.2-3.3M IOPS, and it'd obviously be a bit more problematic with
> required RCU freeing of bio's. Even without the alloc cache, using RCU
> will ruin any potential cache locality on back-to-back bio free + bio
> alloc.

That sucks indeed.  How do you recycle the bios?  If we make sure the
bio is only ever recycled as a bio and bi_bdev remaings valid long
enough we might not need the rcu free.  Even without your recycling
we could probably do something nasty using SLAB_TYPESAFE_BY_RCU.

Re: switch block layer polling to a bio based model

[Jens Axboe]

On 4/26/21 9:06 AM, Christoph Hellwig wrote:
> On Mon, Apr 26, 2021 at 08:57:31AM -0600, Jens Axboe wrote:
>> I was separately curious about this as I have a (as of yet unposted)
>> patchset that recycles bio allocations, as we spend quite a bit of time
>> doing that for high rate polled IO. It's good for taking the above 2.97M
>> IOPS to 3.2-3.3M IOPS, and it'd obviously be a bit more problematic with
>> required RCU freeing of bio's. Even without the alloc cache, using RCU
>> will ruin any potential cache locality on back-to-back bio free + bio
>> alloc.
> 
> That sucks indeed.  How do you recycle the bios?  If we make sure the

Here's the series. It's not super clean (yet), but basically allows
users like io_uring to setup a bio cache, and pass that in through
iocb->ki_bi_cache. With that, we can recycle them instead of going
through free+alloc continually. If you look at profiles for high iops,
we're spending more time than desired doing just that.

https://git.kernel.dk/cgit/linux-block/log/?h=io_uring-bio-cache

> bio is only ever recycled as a bio and bi_bdev remaings valid long
> enough we might not need the rcu free.  Even without your recycling
> we could probably do something nasty using SLAB_TYPESAFE_BY_RCU.

It would not be hard to restrict to same bdev for the cache, just one
more check to do for recycling.

Note that the caching series _only_ supports polled IO for now, as
non-polled would require IRQ juggling for free+alloc and that will
definitely take some of the win away and maybe even render it moot.
Have yet to test that part out. Not a huge deal with the RCU free, as
you end up doing that purely for polled IO and hence wouldn't impact the
IRQ side of things negatively.

-- 
Jens Axboe

Re: [PATCH 12/12] block: switch polling to be bio based

[Ming Lei]

On Mon, Apr 26, 2021 at 03:48:21PM +0200, Christoph Hellwig wrote:
> Replace the blk_poll interface that requires the caller to keep a queue
> and cookie from the submissions with polling based on the bio.
> 
> Polling for the bio itself leads to a few advantages:
> 
>  - the cookie construction can made entirely private in blk-mq.c
>  - the caller does not need to remember the request_queue and cookie
>    separately and thus sidesteps their lifetime issues
>  - keeping the device and the cookie inside the bio allows to trivially
>    support polling BIOs remapping by stacking drivers
>  - a lot of code to propagate the cookie back up the submission path can
>    be removed entirely.
> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> ---
>  arch/m68k/emu/nfblock.c             |   3 +-
>  arch/xtensa/platforms/iss/simdisk.c |   3 +-
>  block/bio.c                         |   1 +
>  block/blk-core.c                    | 103 +++++++++++++++++++---------
>  block/blk-mq.c                      |  75 +++++++-------------
>  block/blk-mq.h                      |   2 +
>  drivers/block/brd.c                 |  12 ++--
>  drivers/block/drbd/drbd_int.h       |   2 +-
>  drivers/block/drbd/drbd_req.c       |   3 +-
>  drivers/block/n64cart.c             |  12 ++--
>  drivers/block/null_blk/main.c       |   3 +-
>  drivers/block/pktcdvd.c             |   7 +-
>  drivers/block/ps3vram.c             |   6 +-
>  drivers/block/rsxx/dev.c            |   7 +-
>  drivers/block/umem.c                |   4 +-
>  drivers/block/zram/zram_drv.c       |  10 +--
>  drivers/lightnvm/pblk-init.c        |   6 +-
>  drivers/md/bcache/request.c         |  13 ++--
>  drivers/md/bcache/request.h         |   4 +-
>  drivers/md/dm.c                     |  28 +++-----
>  drivers/md/md.c                     |  10 ++-
>  drivers/nvdimm/blk.c                |   5 +-
>  drivers/nvdimm/btt.c                |   5 +-
>  drivers/nvdimm/pmem.c               |   3 +-
>  drivers/nvme/host/core.c            |   2 +-
>  drivers/nvme/host/multipath.c       |   6 +-
>  drivers/nvme/host/nvme.h            |   2 +-
>  drivers/s390/block/dcssblk.c        |   7 +-
>  drivers/s390/block/xpram.c          |   5 +-
>  fs/block_dev.c                      |  25 +++----
>  fs/btrfs/inode.c                    |   8 +--
>  fs/ext4/file.c                      |   2 +-
>  fs/gfs2/file.c                      |   4 +-
>  fs/iomap/direct-io.c                |  39 ++++-------
>  fs/xfs/xfs_file.c                   |   2 +-
>  fs/zonefs/super.c                   |   2 +-
>  include/linux/bio.h                 |   2 +-
>  include/linux/blk-mq.h              |  15 +---
>  include/linux/blk_types.h           |  12 ++--
>  include/linux/blkdev.h              |   8 ++-
>  include/linux/fs.h                  |   6 +-
>  include/linux/iomap.h               |   3 +-
>  mm/page_io.c                        |   8 +--
>  43 files changed, 208 insertions(+), 277 deletions(-)
> 
> diff --git a/arch/m68k/emu/nfblock.c b/arch/m68k/emu/nfblock.c
> index ba808543161a..dd36808f0d5e 100644
> --- a/arch/m68k/emu/nfblock.c
> +++ b/arch/m68k/emu/nfblock.c
> @@ -59,7 +59,7 @@ struct nfhd_device {
>  	struct gendisk *disk;
>  };
>  
> -static blk_qc_t nfhd_submit_bio(struct bio *bio)
> +static void nfhd_submit_bio(struct bio *bio)
>  {
>  	struct nfhd_device *dev = bio->bi_bdev->bd_disk->private_data;
>  	struct bio_vec bvec;
> @@ -77,7 +77,6 @@ static blk_qc_t nfhd_submit_bio(struct bio *bio)
>  		sec += len;
>  	}
>  	bio_endio(bio);
> -	return BLK_QC_T_NONE;
>  }
>  
>  static int nfhd_getgeo(struct block_device *bdev, struct hd_geometry *geo)
> diff --git a/arch/xtensa/platforms/iss/simdisk.c b/arch/xtensa/platforms/iss/simdisk.c
> index fc09be7b1347..182825d639e2 100644
> --- a/arch/xtensa/platforms/iss/simdisk.c
> +++ b/arch/xtensa/platforms/iss/simdisk.c
> @@ -101,7 +101,7 @@ static void simdisk_transfer(struct simdisk *dev, unsigned long sector,
>  	spin_unlock(&dev->lock);
>  }
>  
> -static blk_qc_t simdisk_submit_bio(struct bio *bio)
> +static void simdisk_submit_bio(struct bio *bio)
>  {
>  	struct simdisk *dev = bio->bi_bdev->bd_disk->private_data;
>  	struct bio_vec bvec;
> @@ -119,7 +119,6 @@ static blk_qc_t simdisk_submit_bio(struct bio *bio)
>  	}
>  
>  	bio_endio(bio);
> -	return BLK_QC_T_NONE;
>  }
>  
>  static int simdisk_open(struct block_device *bdev, fmode_t mode)
> diff --git a/block/bio.c b/block/bio.c
> index 7296abe293de..484b6d786857 100644
> --- a/block/bio.c
> +++ b/block/bio.c
> @@ -259,6 +259,7 @@ void bio_init(struct bio *bio, struct bio_vec *table,
>  	memset(bio, 0, sizeof(*bio));
>  	atomic_set(&bio->__bi_remaining, 1);
>  	atomic_set(&bio->__bi_cnt, 1);
> +	bio->bi_cookie = BLK_QC_T_NONE;
>  
>  	bio->bi_io_vec = table;
>  	bio->bi_max_vecs = max_vecs;
> diff --git a/block/blk-core.c b/block/blk-core.c
> index adfab5976be0..77fdb00fcad3 100644
> --- a/block/blk-core.c
> +++ b/block/blk-core.c
> @@ -910,18 +910,18 @@ static noinline_for_stack bool submit_bio_checks(struct bio *bio)
>  	return false;
>  }
>  
> -static blk_qc_t __submit_bio(struct bio *bio)
> +static void __submit_bio(struct bio *bio)
>  {
>  	struct gendisk *disk = bio->bi_bdev->bd_disk;
> -	blk_qc_t ret = BLK_QC_T_NONE;
>  
>  	if (blk_crypto_bio_prep(&bio)) {
> -		if (!disk->fops->submit_bio)
> -			return blk_mq_submit_bio(bio);
> -		ret = disk->fops->submit_bio(bio);
> +		if (!disk->fops->submit_bio) {
> +			blk_mq_submit_bio(bio);
> +			return;
> +		}
> +		disk->fops->submit_bio(bio);
>  	}
>  	blk_queue_exit(disk->queue);
> -	return ret;
>  }
>  
>  /*
> @@ -943,10 +943,9 @@ static blk_qc_t __submit_bio(struct bio *bio)
>   * bio_list_on_stack[1] contains bios that were submitted before the current
>   *	->submit_bio_bio, but that haven't been processed yet.
>   */
> -static blk_qc_t __submit_bio_noacct(struct bio *bio)
> +static void __submit_bio_noacct(struct bio *bio)
>  {
>  	struct bio_list bio_list_on_stack[2];
> -	blk_qc_t ret = BLK_QC_T_NONE;
>  
>  	BUG_ON(bio->bi_next);
>  
> @@ -966,7 +965,7 @@ static blk_qc_t __submit_bio_noacct(struct bio *bio)
>  		bio_list_on_stack[1] = bio_list_on_stack[0];
>  		bio_list_init(&bio_list_on_stack[0]);
>  
> -		ret = __submit_bio(bio);
> +		__submit_bio(bio);
>  
>  		/*
>  		 * Sort new bios into those for a lower level and those for the
> @@ -989,13 +988,11 @@ static blk_qc_t __submit_bio_noacct(struct bio *bio)
>  	} while ((bio = bio_list_pop(&bio_list_on_stack[0])));
>  
>  	current->bio_list = NULL;
> -	return ret;
>  }
>  
> -static blk_qc_t __submit_bio_noacct_mq(struct bio *bio)
> +static void __submit_bio_noacct_mq(struct bio *bio)
>  {
>  	struct bio_list bio_list[2] = { };
> -	blk_qc_t ret = BLK_QC_T_NONE;
>  
>  	current->bio_list = bio_list;
>  
> @@ -1007,15 +1004,13 @@ static blk_qc_t __submit_bio_noacct_mq(struct bio *bio)
>  
>  		if (!blk_crypto_bio_prep(&bio)) {
>  			blk_queue_exit(disk->queue);
> -			ret = BLK_QC_T_NONE;
>  			continue;
>  		}
>  
> -		ret = blk_mq_submit_bio(bio);
> +		blk_mq_submit_bio(bio);
>  	} while ((bio = bio_list_pop(&bio_list[0])));
>  
>  	current->bio_list = NULL;
> -	return ret;
>  }
>  
>  /**
> @@ -1027,10 +1022,10 @@ static blk_qc_t __submit_bio_noacct_mq(struct bio *bio)
>   * systems and other upper level users of the block layer should use
>   * submit_bio() instead.
>   */
> -blk_qc_t submit_bio_noacct(struct bio *bio)
> +void submit_bio_noacct(struct bio *bio)
>  {
>  	if (!submit_bio_checks(bio))
> -		return BLK_QC_T_NONE;
> +		return;
>  
>  	/*
>  	 * We only want one ->submit_bio to be active at a time, else stack
> @@ -1038,14 +1033,12 @@ blk_qc_t submit_bio_noacct(struct bio *bio)
>  	 * to collect a list of requests submited by a ->submit_bio method while
>  	 * it is active, and then process them after it returned.
>  	 */
> -	if (current->bio_list) {
> +	if (current->bio_list)
>  		bio_list_add(&current->bio_list[0], bio);
> -		return BLK_QC_T_NONE;
> -	}
> -
> -	if (!bio->bi_bdev->bd_disk->fops->submit_bio)
> -		return __submit_bio_noacct_mq(bio);
> -	return __submit_bio_noacct(bio);
> +	else if (!bio->bi_bdev->bd_disk->fops->submit_bio)
> +		__submit_bio_noacct_mq(bio);
> +	else
> +		__submit_bio_noacct(bio);
>  }
>  EXPORT_SYMBOL(submit_bio_noacct);
>  
> @@ -1062,10 +1055,10 @@ EXPORT_SYMBOL(submit_bio_noacct);
>   * in @bio.  The bio must NOT be touched by thecaller until ->bi_end_io() has
>   * been called.
>   */
> -blk_qc_t submit_bio(struct bio *bio)
> +void submit_bio(struct bio *bio)
>  {
>  	if (blkcg_punt_bio_submit(bio))
> -		return BLK_QC_T_NONE;
> +		return;
>  
>  	/*
>  	 * If it's a regular read/write or a barrier with data attached,
> @@ -1106,19 +1099,67 @@ blk_qc_t submit_bio(struct bio *bio)
>  	if (unlikely(bio_op(bio) == REQ_OP_READ &&
>  	    bio_flagged(bio, BIO_WORKINGSET))) {
>  		unsigned long pflags;
> -		blk_qc_t ret;
>  
>  		psi_memstall_enter(&pflags);
> -		ret = submit_bio_noacct(bio);
> +		submit_bio_noacct(bio);
>  		psi_memstall_leave(&pflags);
> -
> -		return ret;
> +		return;
>  	}
>  
> -	return submit_bio_noacct(bio);
> +	submit_bio_noacct(bio);
>  }
>  EXPORT_SYMBOL(submit_bio);
>  
> +/**
> + * bio_poll - poll for BIO completions
> + * @bio: bio to poll for
> + *
> + * Poll for completions on queue associated with the bio. Returns number of
> + * completed entries found. If @spin is true, then bio_poll will continue
> + * looping until at least one completion is found, unless the task is
> + * otherwise marked running (or we need to reschedule).
> + *
> + * Note: the caller must either be the context that submitted @bio, or
> + * be in a RCU critical section to prevent freeing of @bio.
> + */
> +int bio_poll(struct bio *bio, bool spin)
> +{
> +	struct request_queue *q = bio->bi_bdev->bd_disk->queue;
> +	blk_qc_t cookie = READ_ONCE(bio->bi_cookie);
> +
> +	if (cookie == BLK_QC_T_NONE ||
> +	    !test_bit(QUEUE_FLAG_POLL, &q->queue_flags))
> +		return 0;
> +
> +	if (current->plug)
> +		blk_flush_plug_list(current->plug, false);
> +
> +	/* not yet implemented, so this should not happen */
> +	if (WARN_ON_ONCE(!queue_is_mq(q)))
> +		return 0;
> +	return blk_mq_poll(q, cookie, spin);
> +}
> +EXPORT_SYMBOL_GPL(bio_poll);
> +
> +/*
> + * Helper to implements file_operations.iopoll.  Requires the bio to be stored
> + * in iocb->private, and cleared before freeing the bio.
> + */
> +int iocb_bio_iopoll(struct kiocb *kiocb, bool spin)
> +{
> +	struct bio *bio;
> +	int ret = 0;
> +
> +	rcu_read_lock();
> +	bio = READ_ONCE(kiocb->private);
> +	if (bio)
> +		ret = bio_poll(bio, spin);
> +	rcu_read_unlock();

If bio_poll() sleeps in case of 'spin', rcu_read_lock() can't be used.
And the POLLED bio may be ended by iocb_bio_iopoll() explicitly via
one bio flag.

Thanks,
Ming

Re: switch block layer polling to a bio based model

[Christoph Hellwig]

On Mon, Apr 26, 2021 at 09:12:09AM -0600, Jens Axboe wrote:
> Here's the series. It's not super clean (yet), but basically allows
> users like io_uring to setup a bio cache, and pass that in through
> iocb->ki_bi_cache. With that, we can recycle them instead of going
> through free+alloc continually. If you look at profiles for high iops,
> we're spending more time than desired doing just that.
> 
> https://git.kernel.dk/cgit/linux-block/log/?h=io_uring-bio-cache

So where do you spend the cycles?  The do not memset the whole bio
optimization is pretty obvious and is someting we should do independent
of the allocator.

The other thing that sucks is the mempool implementation, as it forces
each allocation and free to do an indirect call.  I think it might be
worth to try to frontend it with a normal slab cache and only fall back
to the mempool if that fails.

Re: switch block layer polling to a bio based model

[Jens Axboe]

On 4/26/21 10:15 AM, Christoph Hellwig wrote:
> On Mon, Apr 26, 2021 at 09:12:09AM -0600, Jens Axboe wrote:
>> Here's the series. It's not super clean (yet), but basically allows
>> users like io_uring to setup a bio cache, and pass that in through
>> iocb->ki_bi_cache. With that, we can recycle them instead of going
>> through free+alloc continually. If you look at profiles for high iops,
>> we're spending more time than desired doing just that.
>>
>> https://git.kernel.dk/cgit/linux-block/log/?h=io_uring-bio-cache
> 
> So where do you spend the cycles?  The do not memset the whole bio
> optimization is pretty obvious and is someting we should do independent
> of the allocator.

memset is just a small optimization on top. If we look at current
profiles, the alloc+free looks something ala:

+    2.71%  io_uring  [kernel.vmlinux]  [k] bio_alloc_bioset
+    2.03%  io_uring  [kernel.vmlinux]  [k] kmem_cache_alloc

and

+    2.82%  io_uring  [kernel.vmlinux]  [k] __slab_free
+    1.73%  io_uring  [kernel.vmlinux]  [k] kmem_cache_free
     0.36%  io_uring  [kernel.vmlinux]  [k] mempool_free_slab
     0.27%  io_uring  [kernel.vmlinux]  [k] mempool_free

Which is a substantial amount of cycles that is needed just to
repeatedly use the same set of bios for doing IO. Using the caching
patchset, all of the above are completely eliminated, and the only thing
we dynamically allocate is a request which is a lot cheaper (ends up
being 1-2% for either kernel).

> The other thing that sucks is the mempool implementation, as it forces
> each allocation and free to do an indirect call.  I think it might be
> worth to try to frontend it with a normal slab cache and only fall back
> to the mempool if that fails.

Also minor I believe, but yes it'll eat cycles too. FWIW, the testing
above is done without RETPOLINE.

-- 
Jens Axboe

Re: switch block layer polling to a bio based model

[Christoph Hellwig]

Can you test the force pushed update?  This now avoids the RCU free
for each bio and just uses SLAB_TYPESAFE_BY_RCU instead.

Re: switch block layer polling to a bio based model

[Jens Axboe]

On 4/26/21 11:05 AM, Christoph Hellwig wrote:
> Can you test the force pushed update?  This now avoids the RCU free
> for each bio and just uses SLAB_TYPESAFE_BY_RCU instead.

That looks better, perf is now back where it is without the series.
At least according to usual variation within boots.

-- 
Jens Axboe