[PATCH v5 5/5] dm-crypt: Fix lost ioprio when queuing write bios

Hongyu Jin posted 5 patches 2 years ago
There is a newer version of this series
[PATCH v5 5/5] dm-crypt: Fix lost ioprio when queuing write bios
Posted by Hongyu Jin 2 years ago
From: Hongyu Jin <hongyu.jin@unisoc.com>

The original submitting bio->bi_ioprio setting can be retained by
struct dm_crypt_io::base_bio, we set the original bio's ioprio to
the cloned bio for write.

Link: https://lore.kernel.org/dm-devel/alpine.LRH.2.11.1612141049250.13402@mail.ewheeler.net

Signed-off-by: Hongyu Jin <hongyu.jin@unisoc.com>
---
 drivers/md/dm-crypt.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/md/dm-crypt.c b/drivers/md/dm-crypt.c
index 6de107aff331..7149da6555b8 100644
--- a/drivers/md/dm-crypt.c
+++ b/drivers/md/dm-crypt.c
@@ -1683,6 +1683,7 @@ static struct bio *crypt_alloc_buffer(struct dm_crypt_io *io, unsigned int size)
 				 GFP_NOIO, &cc->bs);
 	clone->bi_private = io;
 	clone->bi_end_io = crypt_endio;
+	clone->bi_ioprio = io->base_bio->bi_ioprio;
 
 	remaining_size = size;
 
-- 
2.34.1
[PATCH v6 0/5] Fix I/O priority lost in device-mapper
Posted by Hongyu Jin 1 year, 12 months ago
From: Hongyu Jin <hongyu.jin@unisoc.com>

High-priority tasks get data from dm-verity devices via RT IO priority,
I/O will lose RT priority when reading FEC and hash values via kworker
submission IO during verification, and the verification phase may be
blocked by low-priority IO.

Dm-crypt has the same problem in the data writing process.

This is because io_context and blkcg are missing.

Move bio_set_ioprio() into submit_bio():
1. Only call bio_set_ioprio() once to set the priority of original bio,
   the bio that cloned and splited from original bio will auto inherit
   the priority of original bio in clone process.

2. Make the IO priority of the original bio to be passed to dm,
   and the dm target inherits the IO priority as needed.

All changes are based on master branch commit 2cf4f94d8e86 ("Merge tag
'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi")

Changes in v6:
  - Rebase patch and resolve conflict for patch 1, 3, 4
  - Modify patch 4: fec_read_parity() follow the priority of original
    bio
  - Update commit message
Changes in v5:
  - Rewrite patch 2, add ioprio parameter in dm_io();
  - Modify dm_io() in patch 3 
Changes in v4:
  - Modify commit message by Suggestion
  - Modify patch for dm-crypt
Changes in v3:
  - Split patch for device-mapper
  - Add patch to fix dm-crypy I/O priority question
  - Add block patch to review together
  - Fix some error in v2 patch
Changes in v2:
  - Add ioprio field in struct dm_io_region
  - Initial struct dm_io_region::ioprio to IOPRIO_DEFAULT
  - Add two interface

Hongyu Jin (5):
  block: Fix bio IO priority setting
  dm: Support I/O priority for dm_io()
  dm-bufio: Support I/O priority
  dm verity: Fix I/O priority lost when read FEC and hash
  dm-crypt: Fix lost ioprio when queuing write bios

 block/blk-core.c                              | 10 +++++
 block/blk-mq.c                                | 11 -----
 drivers/md/dm-bufio.c                         | 43 +++++++++++--------
 drivers/md/dm-crypt.c                         |  1 +
 drivers/md/dm-ebs-target.c                    |  8 ++--
 drivers/md/dm-integrity.c                     | 12 +++---
 drivers/md/dm-io.c                            | 23 +++++-----
 drivers/md/dm-kcopyd.c                        |  4 +-
 drivers/md/dm-log.c                           |  4 +-
 drivers/md/dm-raid1.c                         |  6 +--
 drivers/md/dm-snap-persistent.c               |  8 ++--
 drivers/md/dm-verity-fec.c                    | 20 ++++++---
 drivers/md/dm-verity-target.c                 | 13 ++++--
 drivers/md/dm-writecache.c                    |  8 ++--
 drivers/md/persistent-data/dm-block-manager.c |  6 +--
 include/linux/dm-bufio.h                      |  5 ++-
 include/linux/dm-io.h                         |  3 +-
 17 files changed, 105 insertions(+), 80 deletions(-)


base-commit: 2cf4f94d8e8646803f8fb0facf134b0cd7fb691a
-- 
2.34.1
[PATCH v6 1/5] block: Fix bio IO priority setting
Posted by Hongyu Jin 1 year, 12 months ago
From: Hongyu Jin <hongyu.jin@unisoc.com>

Move bio_set_ioprio() into submit_bio():
1. Only call bio_set_ioprio() once to set the priority of original bio,
   the bio that cloned and splited from original bio will auto inherit
   the priority of original bio in clone process.

2. The IO priority can be passed to module that implement
   struct gendisk::fops::submit_bio, help resolve some
   of the IO priority loss issues.

This patch depends on commit 82b74cac2849 ("blk-ioprio: Convert from
rqos policy to direct call")

Fixes: a78418e6a04c ("block: Always initialize bio IO priority on submit")

Co-developed-by: Yibin Ding <yibin.ding@unisoc.com>
Signed-off-by: Yibin Ding <yibin.ding@unisoc.com>
Signed-off-by: Hongyu Jin <hongyu.jin@unisoc.com>
---
 block/blk-core.c | 10 ++++++++++
 block/blk-mq.c   | 11 -----------
 2 files changed, 10 insertions(+), 11 deletions(-)

diff --git a/block/blk-core.c b/block/blk-core.c
index 2eca76ccf4ee..d707ec056f34 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -49,6 +49,7 @@
 #include "blk-pm.h"
 #include "blk-cgroup.h"
 #include "blk-throttle.h"
+#include "blk-ioprio.h"
 
 struct dentry *blk_debugfs_root;
 
@@ -817,6 +818,14 @@ void submit_bio_noacct(struct bio *bio)
 }
 EXPORT_SYMBOL(submit_bio_noacct);
 
+static void bio_set_ioprio(struct bio *bio)
+{
+	/* Nobody set ioprio so far? Initialize it based on task's nice value */
+	if (IOPRIO_PRIO_CLASS(bio->bi_ioprio) == IOPRIO_CLASS_NONE)
+		bio->bi_ioprio = get_current_ioprio();
+	blkcg_set_ioprio(bio);
+}
+
 /**
  * submit_bio - submit a bio to the block device layer for I/O
  * @bio: The &struct bio which describes the I/O
@@ -839,6 +848,7 @@ void submit_bio(struct bio *bio)
 		count_vm_events(PGPGOUT, bio_sectors(bio));
 	}
 
+	bio_set_ioprio(bio);
 	submit_bio_noacct(bio);
 }
 EXPORT_SYMBOL(submit_bio);
diff --git a/block/blk-mq.c b/block/blk-mq.c
index ac18f802c027..351e8283eda1 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -40,7 +40,6 @@
 #include "blk-stat.h"
 #include "blk-mq-sched.h"
 #include "blk-rq-qos.h"
-#include "blk-ioprio.h"
 
 static DEFINE_PER_CPU(struct llist_head, blk_cpu_done);
 static DEFINE_PER_CPU(call_single_data_t, blk_cpu_csd);
@@ -2919,14 +2918,6 @@ static bool blk_mq_can_use_cached_rq(struct request *rq, struct blk_plug *plug,
 	return true;
 }
 
-static void bio_set_ioprio(struct bio *bio)
-{
-	/* Nobody set ioprio so far? Initialize it based on task's nice value */
-	if (IOPRIO_PRIO_CLASS(bio->bi_ioprio) == IOPRIO_CLASS_NONE)
-		bio->bi_ioprio = get_current_ioprio();
-	blkcg_set_ioprio(bio);
-}
-
 /**
  * blk_mq_submit_bio - Create and send a request to block device.
  * @bio: Bio pointer.
@@ -2957,8 +2948,6 @@ void blk_mq_submit_bio(struct bio *bio)
 			return;
 	}
 
-	bio_set_ioprio(bio);
-
 	if (plug) {
 		rq = rq_list_peek(&plug->cached_rq);
 		if (rq && rq->q != q)
-- 
2.34.1
[PATCH v6 2/5] dm: Support I/O priority for dm_io()
Posted by Hongyu Jin 1 year, 12 months ago
From: Hongyu Jin <hongyu.jin@unisoc.com>

Some I/O will dispatch from kworker with different io_context settings
than the submitting task, we may need to specify a priority to avoid
losing priority.

Add I/O priority parameter for dm_io().

Co-developed-by: Yibin Ding <yibin.ding@unisoc.com>
Signed-off-by: Yibin Ding <yibin.ding@unisoc.com>
Signed-off-by: Hongyu Jin <hongyu.jin@unisoc.com>
---
 drivers/md/dm-bufio.c           |  6 +++---
 drivers/md/dm-integrity.c       | 10 +++++-----
 drivers/md/dm-io.c              | 23 +++++++++++++----------
 drivers/md/dm-kcopyd.c          |  4 ++--
 drivers/md/dm-log.c             |  4 ++--
 drivers/md/dm-raid1.c           |  6 +++---
 drivers/md/dm-snap-persistent.c |  4 ++--
 drivers/md/dm-writecache.c      |  8 ++++----
 include/linux/dm-io.h           |  3 ++-
 9 files changed, 36 insertions(+), 32 deletions(-)

diff --git a/drivers/md/dm-bufio.c b/drivers/md/dm-bufio.c
index f03d7dba270c..4f2808ef387f 100644
--- a/drivers/md/dm-bufio.c
+++ b/drivers/md/dm-bufio.c
@@ -1315,7 +1315,7 @@ static void use_dmio(struct dm_buffer *b, enum req_op op, sector_t sector,
 		io_req.mem.ptr.vma = (char *)b->data + offset;
 	}
 
-	r = dm_io(&io_req, 1, &region, NULL);
+	r = dm_io(&io_req, 1, &region, NULL, IOPRIO_DEFAULT);
 	if (unlikely(r))
 		b->end_io(b, errno_to_blk_status(r));
 }
@@ -2167,7 +2167,7 @@ int dm_bufio_issue_flush(struct dm_bufio_client *c)
 	if (WARN_ON_ONCE(dm_bufio_in_request()))
 		return -EINVAL;
 
-	return dm_io(&io_req, 1, &io_reg, NULL);
+	return dm_io(&io_req, 1, &io_reg, NULL, IOPRIO_DEFAULT);
 }
 EXPORT_SYMBOL_GPL(dm_bufio_issue_flush);
 
@@ -2191,7 +2191,7 @@ int dm_bufio_issue_discard(struct dm_bufio_client *c, sector_t block, sector_t c
 	if (WARN_ON_ONCE(dm_bufio_in_request()))
 		return -EINVAL; /* discards are optional */
 
-	return dm_io(&io_req, 1, &io_reg, NULL);
+	return dm_io(&io_req, 1, &io_reg, NULL, IOPRIO_DEFAULT);
 }
 EXPORT_SYMBOL_GPL(dm_bufio_issue_discard);
 
diff --git a/drivers/md/dm-integrity.c b/drivers/md/dm-integrity.c
index e85c688fd91e..9ffd093ad6cc 100644
--- a/drivers/md/dm-integrity.c
+++ b/drivers/md/dm-integrity.c
@@ -553,7 +553,7 @@ static int sync_rw_sb(struct dm_integrity_c *ic, blk_opf_t opf)
 		}
 	}
 
-	r = dm_io(&io_req, 1, &io_loc, NULL);
+	r = dm_io(&io_req, 1, &io_loc, NULL, IOPRIO_DEFAULT);
 	if (unlikely(r))
 		return r;
 
@@ -1071,7 +1071,7 @@ static void rw_journal_sectors(struct dm_integrity_c *ic, blk_opf_t opf,
 	io_loc.sector = ic->start + SB_SECTORS + sector;
 	io_loc.count = n_sectors;
 
-	r = dm_io(&io_req, 1, &io_loc, NULL);
+	r = dm_io(&io_req, 1, &io_loc, NULL, IOPRIO_DEFAULT);
 	if (unlikely(r)) {
 		dm_integrity_io_error(ic, (opf & REQ_OP_MASK) == REQ_OP_READ ?
 				      "reading journal" : "writing journal", r);
@@ -1188,7 +1188,7 @@ static void copy_from_journal(struct dm_integrity_c *ic, unsigned int section, u
 	io_loc.sector = target;
 	io_loc.count = n_sectors;
 
-	r = dm_io(&io_req, 1, &io_loc, NULL);
+	r = dm_io(&io_req, 1, &io_loc, NULL, IOPRIO_DEFAULT);
 	if (unlikely(r)) {
 		WARN_ONCE(1, "asynchronous dm_io failed: %d", r);
 		fn(-1UL, data);
@@ -1517,7 +1517,7 @@ static void dm_integrity_flush_buffers(struct dm_integrity_c *ic, bool flush_dat
 		fr.io_reg.count = 0,
 		fr.ic = ic;
 		init_completion(&fr.comp);
-		r = dm_io(&fr.io_req, 1, &fr.io_reg, NULL);
+		r = dm_io(&fr.io_req, 1, &fr.io_reg, NULL, IOPRIO_DEFAULT);
 		BUG_ON(r);
 	}
 
@@ -2739,7 +2739,7 @@ static void integrity_recalc(struct work_struct *w)
 	io_loc.sector = get_data_sector(ic, area, offset);
 	io_loc.count = n_sectors;
 
-	r = dm_io(&io_req, 1, &io_loc, NULL);
+	r = dm_io(&io_req, 1, &io_loc, NULL, IOPRIO_DEFAULT);
 	if (unlikely(r)) {
 		dm_integrity_io_error(ic, "reading data", r);
 		goto err;
diff --git a/drivers/md/dm-io.c b/drivers/md/dm-io.c
index f053ce245814..7409490259d1 100644
--- a/drivers/md/dm-io.c
+++ b/drivers/md/dm-io.c
@@ -305,7 +305,7 @@ static void km_dp_init(struct dpages *dp, void *data)
  */
 static void do_region(const blk_opf_t opf, unsigned int region,
 		      struct dm_io_region *where, struct dpages *dp,
-		      struct io *io)
+		      struct io *io, unsigned short ioprio)
 {
 	struct bio *bio;
 	struct page *page;
@@ -354,6 +354,7 @@ static void do_region(const blk_opf_t opf, unsigned int region,
 				       &io->client->bios);
 		bio->bi_iter.bi_sector = where->sector + (where->count - remaining);
 		bio->bi_end_io = endio;
+		bio->bi_ioprio = ioprio;
 		store_io_and_region_in_bio(bio, io, region);
 
 		if (op == REQ_OP_DISCARD || op == REQ_OP_WRITE_ZEROES) {
@@ -383,7 +384,7 @@ static void do_region(const blk_opf_t opf, unsigned int region,
 
 static void dispatch_io(blk_opf_t opf, unsigned int num_regions,
 			struct dm_io_region *where, struct dpages *dp,
-			struct io *io, int sync)
+			struct io *io, int sync, unsigned short ioprio)
 {
 	int i;
 	struct dpages old_pages = *dp;
@@ -400,7 +401,7 @@ static void dispatch_io(blk_opf_t opf, unsigned int num_regions,
 	for (i = 0; i < num_regions; i++) {
 		*dp = old_pages;
 		if (where[i].count || (opf & REQ_PREFLUSH))
-			do_region(opf, i, where + i, dp, io);
+			do_region(opf, i, where + i, dp, io, ioprio);
 	}
 
 	/*
@@ -425,7 +426,7 @@ static void sync_io_complete(unsigned long error, void *context)
 
 static int sync_io(struct dm_io_client *client, unsigned int num_regions,
 		   struct dm_io_region *where, blk_opf_t opf, struct dpages *dp,
-		   unsigned long *error_bits)
+		   unsigned long *error_bits, unsigned short ioprio)
 {
 	struct io *io;
 	struct sync_io sio;
@@ -447,7 +448,7 @@ static int sync_io(struct dm_io_client *client, unsigned int num_regions,
 	io->vma_invalidate_address = dp->vma_invalidate_address;
 	io->vma_invalidate_size = dp->vma_invalidate_size;
 
-	dispatch_io(opf, num_regions, where, dp, io, 1);
+	dispatch_io(opf, num_regions, where, dp, io, 1, ioprio);
 
 	wait_for_completion_io(&sio.wait);
 
@@ -459,7 +460,8 @@ static int sync_io(struct dm_io_client *client, unsigned int num_regions,
 
 static int async_io(struct dm_io_client *client, unsigned int num_regions,
 		    struct dm_io_region *where, blk_opf_t opf,
-		    struct dpages *dp, io_notify_fn fn, void *context)
+		    struct dpages *dp, io_notify_fn fn, void *context,
+		    unsigned short ioprio)
 {
 	struct io *io;
 
@@ -479,7 +481,7 @@ static int async_io(struct dm_io_client *client, unsigned int num_regions,
 	io->vma_invalidate_address = dp->vma_invalidate_address;
 	io->vma_invalidate_size = dp->vma_invalidate_size;
 
-	dispatch_io(opf, num_regions, where, dp, io, 0);
+	dispatch_io(opf, num_regions, where, dp, io, 0, ioprio);
 	return 0;
 }
 
@@ -521,7 +523,8 @@ static int dp_init(struct dm_io_request *io_req, struct dpages *dp,
 }
 
 int dm_io(struct dm_io_request *io_req, unsigned int num_regions,
-	  struct dm_io_region *where, unsigned long *sync_error_bits)
+	  struct dm_io_region *where, unsigned long *sync_error_bits,
+	  unsigned short ioprio)
 {
 	int r;
 	struct dpages dp;
@@ -532,11 +535,11 @@ int dm_io(struct dm_io_request *io_req, unsigned int num_regions,
 
 	if (!io_req->notify.fn)
 		return sync_io(io_req->client, num_regions, where,
-			       io_req->bi_opf, &dp, sync_error_bits);
+			       io_req->bi_opf, &dp, sync_error_bits, ioprio);
 
 	return async_io(io_req->client, num_regions, where,
 			io_req->bi_opf, &dp, io_req->notify.fn,
-			io_req->notify.context);
+			io_req->notify.context, ioprio);
 }
 EXPORT_SYMBOL(dm_io);
 
diff --git a/drivers/md/dm-kcopyd.c b/drivers/md/dm-kcopyd.c
index d01807c50f20..79c65c9ad5fa 100644
--- a/drivers/md/dm-kcopyd.c
+++ b/drivers/md/dm-kcopyd.c
@@ -578,9 +578,9 @@ static int run_io_job(struct kcopyd_job *job)
 	io_job_start(job->kc->throttle);
 
 	if (job->op == REQ_OP_READ)
-		r = dm_io(&io_req, 1, &job->source, NULL);
+		r = dm_io(&io_req, 1, &job->source, NULL, IOPRIO_DEFAULT);
 	else
-		r = dm_io(&io_req, job->num_dests, job->dests, NULL);
+		r = dm_io(&io_req, job->num_dests, job->dests, NULL, IOPRIO_DEFAULT);
 
 	return r;
 }
diff --git a/drivers/md/dm-log.c b/drivers/md/dm-log.c
index f9f84236dfcd..f7f9c2100937 100644
--- a/drivers/md/dm-log.c
+++ b/drivers/md/dm-log.c
@@ -300,7 +300,7 @@ static int rw_header(struct log_c *lc, enum req_op op)
 {
 	lc->io_req.bi_opf = op;
 
-	return dm_io(&lc->io_req, 1, &lc->header_location, NULL);
+	return dm_io(&lc->io_req, 1, &lc->header_location, NULL, IOPRIO_DEFAULT);
 }
 
 static int flush_header(struct log_c *lc)
@@ -313,7 +313,7 @@ static int flush_header(struct log_c *lc)
 
 	lc->io_req.bi_opf = REQ_OP_WRITE | REQ_PREFLUSH;
 
-	return dm_io(&lc->io_req, 1, &null_location, NULL);
+	return dm_io(&lc->io_req, 1, &null_location, NULL, IOPRIO_DEFAULT);
 }
 
 static int read_header(struct log_c *log)
diff --git a/drivers/md/dm-raid1.c b/drivers/md/dm-raid1.c
index ddcb2bc4a617..9511dae5b556 100644
--- a/drivers/md/dm-raid1.c
+++ b/drivers/md/dm-raid1.c
@@ -278,7 +278,7 @@ static int mirror_flush(struct dm_target *ti)
 	}
 
 	error_bits = -1;
-	dm_io(&io_req, ms->nr_mirrors, io, &error_bits);
+	dm_io(&io_req, ms->nr_mirrors, io, &error_bits, IOPRIO_DEFAULT);
 	if (unlikely(error_bits != 0)) {
 		for (i = 0; i < ms->nr_mirrors; i++)
 			if (test_bit(i, &error_bits))
@@ -554,7 +554,7 @@ static void read_async_bio(struct mirror *m, struct bio *bio)
 
 	map_region(&io, m, bio);
 	bio_set_m(bio, m);
-	BUG_ON(dm_io(&io_req, 1, &io, NULL));
+	BUG_ON(dm_io(&io_req, 1, &io, NULL, IOPRIO_DEFAULT));
 }
 
 static inline int region_in_sync(struct mirror_set *ms, region_t region,
@@ -681,7 +681,7 @@ static void do_write(struct mirror_set *ms, struct bio *bio)
 	 */
 	bio_set_m(bio, get_default_mirror(ms));
 
-	BUG_ON(dm_io(&io_req, ms->nr_mirrors, io, NULL));
+	BUG_ON(dm_io(&io_req, ms->nr_mirrors, io, NULL, IOPRIO_DEFAULT));
 }
 
 static void do_writes(struct mirror_set *ms, struct bio_list *writes)
diff --git a/drivers/md/dm-snap-persistent.c b/drivers/md/dm-snap-persistent.c
index 15649921f2a9..568d10842b1f 100644
--- a/drivers/md/dm-snap-persistent.c
+++ b/drivers/md/dm-snap-persistent.c
@@ -223,7 +223,7 @@ static void do_metadata(struct work_struct *work)
 {
 	struct mdata_req *req = container_of(work, struct mdata_req, work);
 
-	req->result = dm_io(req->io_req, 1, req->where, NULL);
+	req->result = dm_io(req->io_req, 1, req->where, NULL, IOPRIO_DEFAULT);
 }
 
 /*
@@ -247,7 +247,7 @@ static int chunk_io(struct pstore *ps, void *area, chunk_t chunk, blk_opf_t opf,
 	struct mdata_req req;
 
 	if (!metadata)
-		return dm_io(&io_req, 1, &where, NULL);
+		return dm_io(&io_req, 1, &where, NULL, IOPRIO_DEFAULT);
 
 	req.where = &where;
 	req.io_req = &io_req;
diff --git a/drivers/md/dm-writecache.c b/drivers/md/dm-writecache.c
index 074cb785eafc..6a4279bfb1e7 100644
--- a/drivers/md/dm-writecache.c
+++ b/drivers/md/dm-writecache.c
@@ -531,7 +531,7 @@ static void ssd_commit_flushed(struct dm_writecache *wc, bool wait_for_ios)
 		req.notify.context = &endio;
 
 		/* writing via async dm-io (implied by notify.fn above) won't return an error */
-		(void) dm_io(&req, 1, &region, NULL);
+		(void) dm_io(&req, 1, &region, NULL, IOPRIO_DEFAULT);
 		i = j;
 	}
 
@@ -568,7 +568,7 @@ static void ssd_commit_superblock(struct dm_writecache *wc)
 	req.notify.fn = NULL;
 	req.notify.context = NULL;
 
-	r = dm_io(&req, 1, &region, NULL);
+	r = dm_io(&req, 1, &region, NULL, IOPRIO_DEFAULT);
 	if (unlikely(r))
 		writecache_error(wc, r, "error writing superblock");
 }
@@ -596,7 +596,7 @@ static void writecache_disk_flush(struct dm_writecache *wc, struct dm_dev *dev)
 	req.client = wc->dm_io;
 	req.notify.fn = NULL;
 
-	r = dm_io(&req, 1, &region, NULL);
+	r = dm_io(&req, 1, &region, NULL, IOPRIO_DEFAULT);
 	if (unlikely(r))
 		writecache_error(wc, r, "error flushing metadata: %d", r);
 }
@@ -990,7 +990,7 @@ static int writecache_read_metadata(struct dm_writecache *wc, sector_t n_sectors
 	req.client = wc->dm_io;
 	req.notify.fn = NULL;
 
-	return dm_io(&req, 1, &region, NULL);
+	return dm_io(&req, 1, &region, NULL, IOPRIO_DEFAULT);
 }
 
 static void writecache_resume(struct dm_target *ti)
diff --git a/include/linux/dm-io.h b/include/linux/dm-io.h
index 7595142f3fc5..7b2968612b7e 100644
--- a/include/linux/dm-io.h
+++ b/include/linux/dm-io.h
@@ -80,7 +80,8 @@ void dm_io_client_destroy(struct dm_io_client *client);
  * error occurred doing io to the corresponding region.
  */
 int dm_io(struct dm_io_request *io_req, unsigned int num_regions,
-	  struct dm_io_region *region, unsigned int long *sync_error_bits);
+	  struct dm_io_region *region, unsigned int long *sync_error_bits,
+	  unsigned short ioprio);
 
 #endif	/* __KERNEL__ */
 #endif	/* _LINUX_DM_IO_H */
-- 
2.34.1
[PATCH v6 3/5] dm-bufio: Support I/O priority
Posted by Hongyu Jin 1 year, 12 months ago
From: Hongyu Jin <hongyu.jin@unisoc.com>

Some I/O will dispatch from kworker with different io_context settings
than the submitting task, we may need to specify a priority to avoid
losing priority.

Add I/O priority parameter for dm_bufio_read() and
dm_bufio_prefetch().

Co-developed-by: Yibin Ding <yibin.ding@unisoc.com>
Signed-off-by: Yibin Ding <yibin.ding@unisoc.com>
Signed-off-by: Hongyu Jin <hongyu.jin@unisoc.com>
---
 drivers/md/dm-bufio.c                         | 39 +++++++++++--------
 drivers/md/dm-ebs-target.c                    |  8 ++--
 drivers/md/dm-integrity.c                     |  2 +-
 drivers/md/dm-snap-persistent.c               |  4 +-
 drivers/md/dm-verity-fec.c                    |  4 +-
 drivers/md/dm-verity-target.c                 |  5 ++-
 drivers/md/persistent-data/dm-block-manager.c |  6 +--
 include/linux/dm-bufio.h                      |  5 ++-
 8 files changed, 40 insertions(+), 33 deletions(-)

diff --git a/drivers/md/dm-bufio.c b/drivers/md/dm-bufio.c
index 4f2808ef387f..a6974ecab68e 100644
--- a/drivers/md/dm-bufio.c
+++ b/drivers/md/dm-bufio.c
@@ -1292,7 +1292,8 @@ static void dmio_complete(unsigned long error, void *context)
 }
 
 static void use_dmio(struct dm_buffer *b, enum req_op op, sector_t sector,
-		     unsigned int n_sectors, unsigned int offset)
+		     unsigned int n_sectors, unsigned int offset,
+		     unsigned short ioprio)
 {
 	int r;
 	struct dm_io_request io_req = {
@@ -1315,7 +1316,7 @@ static void use_dmio(struct dm_buffer *b, enum req_op op, sector_t sector,
 		io_req.mem.ptr.vma = (char *)b->data + offset;
 	}
 
-	r = dm_io(&io_req, 1, &region, NULL, IOPRIO_DEFAULT);
+	r = dm_io(&io_req, 1, &region, NULL, ioprio);
 	if (unlikely(r))
 		b->end_io(b, errno_to_blk_status(r));
 }
@@ -1331,7 +1332,8 @@ static void bio_complete(struct bio *bio)
 }
 
 static void use_bio(struct dm_buffer *b, enum req_op op, sector_t sector,
-		    unsigned int n_sectors, unsigned int offset)
+		    unsigned int n_sectors, unsigned int offset,
+		    unsigned short ioprio)
 {
 	struct bio *bio;
 	char *ptr;
@@ -1339,13 +1341,14 @@ static void use_bio(struct dm_buffer *b, enum req_op op, sector_t sector,
 
 	bio = bio_kmalloc(1, GFP_NOWAIT | __GFP_NORETRY | __GFP_NOWARN);
 	if (!bio) {
-		use_dmio(b, op, sector, n_sectors, offset);
+		use_dmio(b, op, sector, n_sectors, offset, ioprio);
 		return;
 	}
 	bio_init(bio, b->c->bdev, bio->bi_inline_vecs, 1, op);
 	bio->bi_iter.bi_sector = sector;
 	bio->bi_end_io = bio_complete;
 	bio->bi_private = b;
+	bio->bi_ioprio = ioprio;
 
 	ptr = (char *)b->data + offset;
 	len = n_sectors << SECTOR_SHIFT;
@@ -1368,7 +1371,7 @@ static inline sector_t block_to_sector(struct dm_bufio_client *c, sector_t block
 	return sector;
 }
 
-static void submit_io(struct dm_buffer *b, enum req_op op,
+static void submit_io(struct dm_buffer *b, enum req_op op, unsigned short ioprio,
 		      void (*end_io)(struct dm_buffer *, blk_status_t))
 {
 	unsigned int n_sectors;
@@ -1398,9 +1401,9 @@ static void submit_io(struct dm_buffer *b, enum req_op op,
 	}
 
 	if (b->data_mode != DATA_MODE_VMALLOC)
-		use_bio(b, op, sector, n_sectors, offset);
+		use_bio(b, op, sector, n_sectors, offset, ioprio);
 	else
-		use_dmio(b, op, sector, n_sectors, offset);
+		use_dmio(b, op, sector, n_sectors, offset, ioprio);
 }
 
 /*
@@ -1456,7 +1459,7 @@ static void __write_dirty_buffer(struct dm_buffer *b,
 	b->write_end = b->dirty_end;
 
 	if (!write_list)
-		submit_io(b, REQ_OP_WRITE, write_endio);
+		submit_io(b, REQ_OP_WRITE, IOPRIO_DEFAULT, write_endio);
 	else
 		list_add_tail(&b->write_list, write_list);
 }
@@ -1470,7 +1473,7 @@ static void __flush_write_list(struct list_head *write_list)
 		struct dm_buffer *b =
 			list_entry(write_list->next, struct dm_buffer, write_list);
 		list_del(&b->write_list);
-		submit_io(b, REQ_OP_WRITE, write_endio);
+		submit_io(b, REQ_OP_WRITE, IOPRIO_DEFAULT, write_endio);
 		cond_resched();
 	}
 	blk_finish_plug(&plug);
@@ -1852,7 +1855,8 @@ static void read_endio(struct dm_buffer *b, blk_status_t status)
  * and uses dm_bufio_mark_buffer_dirty to write new data back).
  */
 static void *new_read(struct dm_bufio_client *c, sector_t block,
-		      enum new_flag nf, struct dm_buffer **bp)
+		      enum new_flag nf, struct dm_buffer **bp,
+		      unsigned short ioprio)
 {
 	int need_submit = 0;
 	struct dm_buffer *b;
@@ -1905,7 +1909,7 @@ static void *new_read(struct dm_bufio_client *c, sector_t block,
 		return NULL;
 
 	if (need_submit)
-		submit_io(b, REQ_OP_READ, read_endio);
+		submit_io(b, REQ_OP_READ, ioprio, read_endio);
 
 	if (nf != NF_GET)	/* we already tested this condition above */
 		wait_on_bit_io(&b->state, B_READING, TASK_UNINTERRUPTIBLE);
@@ -1926,17 +1930,17 @@ static void *new_read(struct dm_bufio_client *c, sector_t block,
 void *dm_bufio_get(struct dm_bufio_client *c, sector_t block,
 		   struct dm_buffer **bp)
 {
-	return new_read(c, block, NF_GET, bp);
+	return new_read(c, block, NF_GET, bp, IOPRIO_DEFAULT);
 }
 EXPORT_SYMBOL_GPL(dm_bufio_get);
 
 void *dm_bufio_read(struct dm_bufio_client *c, sector_t block,
-		    struct dm_buffer **bp)
+		    struct dm_buffer **bp, unsigned short ioprio)
 {
 	if (WARN_ON_ONCE(dm_bufio_in_request()))
 		return ERR_PTR(-EINVAL);
 
-	return new_read(c, block, NF_READ, bp);
+	return new_read(c, block, NF_READ, bp, ioprio);
 }
 EXPORT_SYMBOL_GPL(dm_bufio_read);
 
@@ -1946,12 +1950,13 @@ void *dm_bufio_new(struct dm_bufio_client *c, sector_t block,
 	if (WARN_ON_ONCE(dm_bufio_in_request()))
 		return ERR_PTR(-EINVAL);
 
-	return new_read(c, block, NF_FRESH, bp);
+	return new_read(c, block, NF_FRESH, bp, IOPRIO_DEFAULT);
 }
 EXPORT_SYMBOL_GPL(dm_bufio_new);
 
 void dm_bufio_prefetch(struct dm_bufio_client *c,
-		       sector_t block, unsigned int n_blocks)
+		       sector_t block, unsigned int n_blocks,
+		       unsigned short ioprio)
 {
 	struct blk_plug plug;
 
@@ -1987,7 +1992,7 @@ void dm_bufio_prefetch(struct dm_bufio_client *c,
 			dm_bufio_unlock(c);
 
 			if (need_submit)
-				submit_io(b, REQ_OP_READ, read_endio);
+				submit_io(b, REQ_OP_READ, ioprio, read_endio);
 			dm_bufio_release(b);
 
 			cond_resched();
diff --git a/drivers/md/dm-ebs-target.c b/drivers/md/dm-ebs-target.c
index 435b45201f4d..8198c8a7b416 100644
--- a/drivers/md/dm-ebs-target.c
+++ b/drivers/md/dm-ebs-target.c
@@ -84,7 +84,7 @@ static int __ebs_rw_bvec(struct ebs_c *ec, enum req_op op, struct bio_vec *bv,
 
 		/* Avoid reading for writes in case bio vector's page overwrites block completely. */
 		if (op == REQ_OP_READ || buf_off || bv_len < dm_bufio_get_block_size(ec->bufio))
-			ba = dm_bufio_read(ec->bufio, block, &b);
+			ba = dm_bufio_read(ec->bufio, block, &b, IOPRIO_DEFAULT);
 		else
 			ba = dm_bufio_new(ec->bufio, block, &b);
 
@@ -194,13 +194,13 @@ static void __ebs_process_bios(struct work_struct *ws)
 	bio_list_for_each(bio, &bios) {
 		block1 = __sector_to_block(ec, bio->bi_iter.bi_sector);
 		if (bio_op(bio) == REQ_OP_READ)
-			dm_bufio_prefetch(ec->bufio, block1, __nr_blocks(ec, bio));
+			dm_bufio_prefetch(ec->bufio, block1, __nr_blocks(ec, bio), IOPRIO_DEFAULT);
 		else if (bio_op(bio) == REQ_OP_WRITE && !(bio->bi_opf & REQ_PREFLUSH)) {
 			block2 = __sector_to_block(ec, bio_end_sector(bio));
 			if (__block_mod(bio->bi_iter.bi_sector, ec->u_bs))
-				dm_bufio_prefetch(ec->bufio, block1, 1);
+				dm_bufio_prefetch(ec->bufio, block1, 1, IOPRIO_DEFAULT);
 			if (__block_mod(bio_end_sector(bio), ec->u_bs) && block2 != block1)
-				dm_bufio_prefetch(ec->bufio, block2, 1);
+				dm_bufio_prefetch(ec->bufio, block2, 1, IOPRIO_DEFAULT);
 		}
 	}
 
diff --git a/drivers/md/dm-integrity.c b/drivers/md/dm-integrity.c
index 9ffd093ad6cc..1e40e712bcd7 100644
--- a/drivers/md/dm-integrity.c
+++ b/drivers/md/dm-integrity.c
@@ -1418,7 +1418,7 @@ static int dm_integrity_rw_tag(struct dm_integrity_c *ic, unsigned char *tag, se
 		if (unlikely(r))
 			return r;
 
-		data = dm_bufio_read(ic->bufio, *metadata_block, &b);
+		data = dm_bufio_read(ic->bufio, *metadata_block, &b, IOPRIO_DEFAULT);
 		if (IS_ERR(data))
 			return PTR_ERR(data);
 
diff --git a/drivers/md/dm-snap-persistent.c b/drivers/md/dm-snap-persistent.c
index 568d10842b1f..a2072b95e28c 100644
--- a/drivers/md/dm-snap-persistent.c
+++ b/drivers/md/dm-snap-persistent.c
@@ -524,7 +524,7 @@ static int read_exceptions(struct pstore *ps,
 
 				if (unlikely(pf_chunk >= dm_bufio_get_device_size(client)))
 					break;
-				dm_bufio_prefetch(client, pf_chunk, 1);
+				dm_bufio_prefetch(client, pf_chunk, 1, IOPRIO_DEFAULT);
 				prefetch_area++;
 				if (unlikely(!prefetch_area))
 					break;
@@ -533,7 +533,7 @@ static int read_exceptions(struct pstore *ps,
 
 		chunk = area_location(ps, ps->current_area);
 
-		area = dm_bufio_read(client, chunk, &bp);
+		area = dm_bufio_read(client, chunk, &bp, IOPRIO_DEFAULT);
 		if (IS_ERR(area)) {
 			r = PTR_ERR(area);
 			goto ret_destroy_bufio;
diff --git a/drivers/md/dm-verity-fec.c b/drivers/md/dm-verity-fec.c
index b475200d8586..49db19e537f9 100644
--- a/drivers/md/dm-verity-fec.c
+++ b/drivers/md/dm-verity-fec.c
@@ -69,7 +69,7 @@ static u8 *fec_read_parity(struct dm_verity *v, u64 rsb, int index,
 	block = div64_u64_rem(position, v->fec->io_size, &rem);
 	*offset = (unsigned int)rem;
 
-	res = dm_bufio_read(v->fec->bufio, block, buf);
+	res = dm_bufio_read(v->fec->bufio, block, buf, IOPRIO_DEFAULT);
 	if (IS_ERR(res)) {
 		DMERR("%s: FEC %llu: parity read failed (block %llu): %ld",
 		      v->data_dev->name, (unsigned long long)rsb,
@@ -248,7 +248,7 @@ static int fec_read_bufs(struct dm_verity *v, struct dm_verity_io *io,
 			bufio = v->bufio;
 		}
 
-		bbuf = dm_bufio_read(bufio, block, &buf);
+		bbuf = dm_bufio_read(bufio, block, &buf, IOPRIO_DEFAULT);
 		if (IS_ERR(bbuf)) {
 			DMWARN_LIMIT("%s: FEC %llu: read failed (%llu): %ld",
 				     v->data_dev->name,
diff --git a/drivers/md/dm-verity-target.c b/drivers/md/dm-verity-target.c
index 14e58ae70521..4758bfe2c156 100644
--- a/drivers/md/dm-verity-target.c
+++ b/drivers/md/dm-verity-target.c
@@ -308,7 +308,7 @@ static int verity_verify_level(struct dm_verity *v, struct dm_verity_io *io,
 			return -EAGAIN;
 		}
 	} else
-		data = dm_bufio_read(v->bufio, hash_block, &buf);
+		data = dm_bufio_read(v->bufio, hash_block, &buf, IOPRIO_DEFAULT);
 
 	if (IS_ERR(data))
 		return PTR_ERR(data);
@@ -719,7 +719,8 @@ static void verity_prefetch_io(struct work_struct *work)
 		}
 no_prefetch_cluster:
 		dm_bufio_prefetch(v->bufio, hash_block_start,
-				  hash_block_end - hash_block_start + 1);
+				  hash_block_end - hash_block_start + 1,
+				  IOPRIO_DEFAULT);
 	}
 
 	kfree(pw);
diff --git a/drivers/md/persistent-data/dm-block-manager.c b/drivers/md/persistent-data/dm-block-manager.c
index 0e010e1204aa..86a4f73d2f3d 100644
--- a/drivers/md/persistent-data/dm-block-manager.c
+++ b/drivers/md/persistent-data/dm-block-manager.c
@@ -474,7 +474,7 @@ int dm_bm_read_lock(struct dm_block_manager *bm, dm_block_t b,
 	void *p;
 	int r;
 
-	p = dm_bufio_read(bm->bufio, b, (struct dm_buffer **) result);
+	p = dm_bufio_read(bm->bufio, b, (struct dm_buffer **) result, IOPRIO_DEFAULT);
 	if (IS_ERR(p))
 		return PTR_ERR(p);
 
@@ -510,7 +510,7 @@ int dm_bm_write_lock(struct dm_block_manager *bm,
 	if (dm_bm_is_read_only(bm))
 		return -EPERM;
 
-	p = dm_bufio_read(bm->bufio, b, (struct dm_buffer **) result);
+	p = dm_bufio_read(bm->bufio, b, (struct dm_buffer **) result, IOPRIO_DEFAULT);
 	if (IS_ERR(p))
 		return PTR_ERR(p);
 
@@ -624,7 +624,7 @@ EXPORT_SYMBOL_GPL(dm_bm_flush);
 
 void dm_bm_prefetch(struct dm_block_manager *bm, dm_block_t b)
 {
-	dm_bufio_prefetch(bm->bufio, b, 1);
+	dm_bufio_prefetch(bm->bufio, b, 1, IOPRIO_DEFAULT);
 }
 
 bool dm_bm_is_read_only(struct dm_block_manager *bm)
diff --git a/include/linux/dm-bufio.h b/include/linux/dm-bufio.h
index 75e7d8cbb532..256a246c7b97 100644
--- a/include/linux/dm-bufio.h
+++ b/include/linux/dm-bufio.h
@@ -62,7 +62,7 @@ void dm_bufio_set_sector_offset(struct dm_bufio_client *c, sector_t start);
  * it dirty.
  */
 void *dm_bufio_read(struct dm_bufio_client *c, sector_t block,
-		    struct dm_buffer **bp);
+		    struct dm_buffer **bp, unsigned short ioprio);
 
 /*
  * Like dm_bufio_read, but return buffer from cache, don't read
@@ -84,7 +84,8 @@ void *dm_bufio_new(struct dm_bufio_client *c, sector_t block,
  * I/O to finish.
  */
 void dm_bufio_prefetch(struct dm_bufio_client *c,
-		       sector_t block, unsigned int n_blocks);
+		       sector_t block, unsigned int n_blocks,
+		       unsigned short ioprio);
 
 /*
  * Release a reference obtained with dm_bufio_{read,get,new}. The data
-- 
2.34.1
[PATCH v6 4/5] dm verity: Fix I/O priority lost when read FEC and hash
Posted by Hongyu Jin 1 year, 12 months ago
From: Hongyu Jin <hongyu.jin@unisoc.com>

After obtaining the data, verification or error correction process may
trigger a new I/O that loses the priority of the original I/O, that is,
the verification of the higher priority IO may be blocked by the lower
priority IO.

Make the I/O of verification and error correction follow the
priority of original I/O.

Co-developed-by: Yibin Ding <yibin.ding@unisoc.com>
Signed-off-by: Yibin Ding <yibin.ding@unisoc.com>
Signed-off-by: Hongyu Jin <hongyu.jin@unisoc.com>
---
 drivers/md/dm-verity-fec.c    | 20 +++++++++++++++-----
 drivers/md/dm-verity-target.c | 12 ++++++++----
 2 files changed, 23 insertions(+), 9 deletions(-)

diff --git a/drivers/md/dm-verity-fec.c b/drivers/md/dm-verity-fec.c
index 49db19e537f9..a9e5402e3d43 100644
--- a/drivers/md/dm-verity-fec.c
+++ b/drivers/md/dm-verity-fec.c
@@ -18,6 +18,12 @@ bool verity_fec_is_enabled(struct dm_verity *v)
 	return v->fec && v->fec->dev;
 }
 
+static inline struct dm_verity_io *verity_io(struct dm_verity *v, struct dm_verity_fec_io *fio)
+{
+	return (struct dm_verity_io *)
+		((char *)fio + sizeof(struct dm_verity_fec_io) - v->ti->per_io_data_size);
+}
+
 /*
  * Return a pointer to dm_verity_fec_io after dm_verity_io and its variable
  * length fields.
@@ -60,7 +66,8 @@ static int fec_decode_rs8(struct dm_verity *v, struct dm_verity_fec_io *fio,
  * to the data block. Caller is responsible for releasing buf.
  */
 static u8 *fec_read_parity(struct dm_verity *v, u64 rsb, int index,
-			   unsigned int *offset, struct dm_buffer **buf)
+			   unsigned int *offset, struct dm_buffer **buf,
+			   unsigned short ioprio)
 {
 	u64 position, block, rem;
 	u8 *res;
@@ -69,7 +76,7 @@ static u8 *fec_read_parity(struct dm_verity *v, u64 rsb, int index,
 	block = div64_u64_rem(position, v->fec->io_size, &rem);
 	*offset = (unsigned int)rem;
 
-	res = dm_bufio_read(v->fec->bufio, block, buf, IOPRIO_DEFAULT);
+	res = dm_bufio_read(v->fec->bufio, block, buf, ioprio);
 	if (IS_ERR(res)) {
 		DMERR("%s: FEC %llu: parity read failed (block %llu): %ld",
 		      v->data_dev->name, (unsigned long long)rsb,
@@ -129,8 +136,10 @@ static int fec_decode_bufs(struct dm_verity *v, struct dm_verity_fec_io *fio,
 	struct dm_buffer *buf;
 	unsigned int n, i, offset;
 	u8 *par, *block;
+	struct dm_verity_io *io = verity_io(v, fio);
+	struct bio *bio = dm_bio_from_per_bio_data(io, v->ti->per_io_data_size);
 
-	par = fec_read_parity(v, rsb, block_offset, &offset, &buf);
+	par = fec_read_parity(v, rsb, block_offset, &offset, &buf, bio_prio(bio));
 	if (IS_ERR(par))
 		return PTR_ERR(par);
 
@@ -158,7 +167,7 @@ static int fec_decode_bufs(struct dm_verity *v, struct dm_verity_fec_io *fio,
 		if (offset >= v->fec->io_size) {
 			dm_bufio_release(buf);
 
-			par = fec_read_parity(v, rsb, block_offset, &offset, &buf);
+			par = fec_read_parity(v, rsb, block_offset, &offset, &buf, bio_prio(bio));
 			if (IS_ERR(par))
 				return PTR_ERR(par);
 		}
@@ -210,6 +219,7 @@ static int fec_read_bufs(struct dm_verity *v, struct dm_verity_io *io,
 	u8 *bbuf, *rs_block;
 	u8 want_digest[HASH_MAX_DIGESTSIZE];
 	unsigned int n, k;
+	struct bio *bio = dm_bio_from_per_bio_data(io, v->ti->per_io_data_size);
 
 	if (neras)
 		*neras = 0;
@@ -248,7 +258,7 @@ static int fec_read_bufs(struct dm_verity *v, struct dm_verity_io *io,
 			bufio = v->bufio;
 		}
 
-		bbuf = dm_bufio_read(bufio, block, &buf, IOPRIO_DEFAULT);
+		bbuf = dm_bufio_read(bufio, block, &buf, bio_prio(bio));
 		if (IS_ERR(bbuf)) {
 			DMWARN_LIMIT("%s: FEC %llu: read failed (%llu): %ld",
 				     v->data_dev->name,
diff --git a/drivers/md/dm-verity-target.c b/drivers/md/dm-verity-target.c
index 4758bfe2c156..8cbf81fc0031 100644
--- a/drivers/md/dm-verity-target.c
+++ b/drivers/md/dm-verity-target.c
@@ -51,6 +51,7 @@ static DEFINE_STATIC_KEY_FALSE(use_tasklet_enabled);
 struct dm_verity_prefetch_work {
 	struct work_struct work;
 	struct dm_verity *v;
+	unsigned short ioprio;
 	sector_t block;
 	unsigned int n_blocks;
 };
@@ -294,6 +295,7 @@ static int verity_verify_level(struct dm_verity *v, struct dm_verity_io *io,
 	int r;
 	sector_t hash_block;
 	unsigned int offset;
+	struct bio *bio = dm_bio_from_per_bio_data(io, v->ti->per_io_data_size);
 
 	verity_hash_at_level(v, block, level, &hash_block, &offset);
 
@@ -308,7 +310,7 @@ static int verity_verify_level(struct dm_verity *v, struct dm_verity_io *io,
 			return -EAGAIN;
 		}
 	} else
-		data = dm_bufio_read(v->bufio, hash_block, &buf, IOPRIO_DEFAULT);
+		data = dm_bufio_read(v->bufio, hash_block, &buf, bio_prio(bio));
 
 	if (IS_ERR(data))
 		return PTR_ERR(data);
@@ -720,13 +722,14 @@ static void verity_prefetch_io(struct work_struct *work)
 no_prefetch_cluster:
 		dm_bufio_prefetch(v->bufio, hash_block_start,
 				  hash_block_end - hash_block_start + 1,
-				  IOPRIO_DEFAULT);
+				  pw->ioprio);
 	}
 
 	kfree(pw);
 }
 
-static void verity_submit_prefetch(struct dm_verity *v, struct dm_verity_io *io)
+static void verity_submit_prefetch(struct dm_verity *v, struct dm_verity_io *io,
+				   unsigned short ioprio)
 {
 	sector_t block = io->block;
 	unsigned int n_blocks = io->n_blocks;
@@ -754,6 +757,7 @@ static void verity_submit_prefetch(struct dm_verity *v, struct dm_verity_io *io)
 	pw->v = v;
 	pw->block = block;
 	pw->n_blocks = n_blocks;
+	pw->ioprio = ioprio;
 	queue_work(v->verify_wq, &pw->work);
 }
 
@@ -796,7 +800,7 @@ static int verity_map(struct dm_target *ti, struct bio *bio)
 
 	verity_fec_init_io(io);
 
-	verity_submit_prefetch(v, io);
+	verity_submit_prefetch(v, io, bio_prio(bio));
 
 	submit_bio_noacct(bio);
 
-- 
2.34.1
Re: [PATCH v6 4/5] dm verity: Fix I/O priority lost when read FEC and hash
Posted by Eric Biggers 1 year, 12 months ago
On Wed, Dec 20, 2023 at 06:03:32PM +0800, Hongyu Jin wrote:
> +static inline struct dm_verity_io *verity_io(struct dm_verity *v, struct dm_verity_fec_io *fio)
> +{
> +	return (struct dm_verity_io *)
> +		((char *)fio + sizeof(struct dm_verity_fec_io) - v->ti->per_io_data_size);
> +}
[...]
> @@ -129,8 +136,10 @@ static int fec_decode_bufs(struct dm_verity *v, struct dm_verity_fec_io *fio,
>  	struct dm_buffer *buf;
>  	unsigned int n, i, offset;
>  	u8 *par, *block;
> +	struct dm_verity_io *io = verity_io(v, fio);
> +	struct bio *bio = dm_bio_from_per_bio_data(io, v->ti->per_io_data_size);

How about passing 'io' as an argument to fec_decode_bufs() instead of adding the
verity_io() helper function?

- Eric
[PATCH v6 5/5] dm-crypt: Fix lost ioprio when queuing write bios
Posted by Hongyu Jin 1 year, 12 months ago
From: Hongyu Jin <hongyu.jin@unisoc.com>

Since dm-crypt queues writes to a different kernel thread (workqueue),
the bios will dispatch from tasks with different io_context->ioprio
settings and blkcg than the submitting task, thus giving incorrect
ioprio to the io scheduler.

Get the original io priority setting via struct dm_crypt_io::base_bio
and set this priority to the bio for write.

Link: https://lore.kernel.org/dm-devel/alpine.LRH.2.11.1612141049250.13402@mail.ewheeler.net

Signed-off-by: Hongyu Jin <hongyu.jin@unisoc.com>
---
 drivers/md/dm-crypt.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/md/dm-crypt.c b/drivers/md/dm-crypt.c
index 2ae8560b6a14..ba6e794f7871 100644
--- a/drivers/md/dm-crypt.c
+++ b/drivers/md/dm-crypt.c
@@ -1683,6 +1683,7 @@ static struct bio *crypt_alloc_buffer(struct dm_crypt_io *io, unsigned int size)
 				 GFP_NOIO, &cc->bs);
 	clone->bi_private = io;
 	clone->bi_end_io = crypt_endio;
+	clone->bi_ioprio = io->base_bio->bi_ioprio;
 
 	remaining_size = size;
 
-- 
2.34.1
[PATCH v7 0/5] Fix I/O priority lost in device-mapper
Posted by Hongyu Jin 1 year, 12 months ago
From: Hongyu Jin <hongyu.jin@unisoc.com>

High-priority tasks get data from dm-verity devices via RT IO priority,
I/O will lose RT priority when reading FEC and hash values via kworker
submission IO during verification, and the verification phase may be
blocked by low-priority IO.

Dm-crypt has the same problem in the data writing process.

This is because io_context and blkcg are missing.

Move bio_set_ioprio() into submit_bio():
1. Only call bio_set_ioprio() once to set the priority of original bio,
   the bio that cloned and splited from original bio will auto inherit
   the priority of original bio in clone process.

2. Make the IO priority of the original bio to be passed to dm,
   and the dm target inherits the IO priority as needed.

All changes are based on master branch commit 2cf4f94d8e86 ("Merge tag
'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi")

Changes in v7:
  - Modify patch 4: change dm-verity-fec.c
Changes in v6:
  - Rebase patch and resolve conflict for patch 1, 3, 4
  - Modify patch 4: fec_read_parity() follow the priority of original
    bio
  - Update commit message
Changes in v5:
  - Rewrite patch 2, add ioprio parameter in dm_io();
  - Modify dm_io() in patch 3 
Changes in v4:
  - Modify commit message by Suggestion
  - Modify patch for dm-crypt
Changes in v3:
  - Split patch for device-mapper
  - Add patch to fix dm-crypy I/O priority question
  - Add block patch to review together
  - Fix some error in v2 patch
Changes in v2:
  - Add ioprio field in struct dm_io_region
  - Initial struct dm_io_region::ioprio to IOPRIO_DEFAULT
  - Add two interface

Hongyu Jin (5):
  block: Fix bio IO priority setting
  dm: Support I/O priority for dm_io()
  dm-bufio: Support I/O priority
  dm verity: Fix I/O priority lost when read FEC and hash
  dm-crypt: Fix lost ioprio when queuing write bios

 block/blk-core.c                              | 10 +++++
 block/blk-mq.c                                | 11 -----
 drivers/md/dm-bufio.c                         | 43 +++++++++++--------
 drivers/md/dm-crypt.c                         |  1 +
 drivers/md/dm-ebs-target.c                    |  8 ++--
 drivers/md/dm-integrity.c                     | 12 +++---
 drivers/md/dm-io.c                            | 23 +++++-----
 drivers/md/dm-kcopyd.c                        |  4 +-
 drivers/md/dm-log.c                           |  4 +-
 drivers/md/dm-raid1.c                         |  6 +--
 drivers/md/dm-snap-persistent.c               |  8 ++--
 drivers/md/dm-verity-fec.c                    | 21 +++++----
 drivers/md/dm-verity-target.c                 | 13 ++++--
 drivers/md/dm-writecache.c                    |  8 ++--
 drivers/md/persistent-data/dm-block-manager.c |  6 +--
 include/linux/dm-bufio.h                      |  5 ++-
 include/linux/dm-io.h                         |  3 +-
 17 files changed, 102 insertions(+), 84 deletions(-)


base-commit: 2cf4f94d8e8646803f8fb0facf134b0cd7fb691a
-- 
2.34.1
Re: [PATCH v7 0/5] Fix I/O priority lost in device-mapper
Posted by Eric Biggers 1 year, 11 months ago
On Thu, Dec 21, 2023 at 06:31:34PM +0800, Hongyu Jin wrote:
> Hongyu Jin (5):
>   block: Fix bio IO priority setting
>   dm: Support I/O priority for dm_io()
>   dm-bufio: Support I/O priority
>   dm verity: Fix I/O priority lost when read FEC and hash
>   dm-crypt: Fix lost ioprio when queuing write bios

Looks good,

Reviewed-by: Eric Biggers <ebiggers@google.com>

- Eric
[PATCH v7 1/5] block: Fix bio IO priority setting
Posted by Hongyu Jin 1 year, 12 months ago
From: Hongyu Jin <hongyu.jin@unisoc.com>

Move bio_set_ioprio() into submit_bio():
1. Only call bio_set_ioprio() once to set the priority of original bio,
   the bio that cloned and splited from original bio will auto inherit
   the priority of original bio in clone process.

2. The IO priority can be passed to module that implement
   struct gendisk::fops::submit_bio, help resolve some
   of the IO priority loss issues.

This patch depends on commit 82b74cac2849 ("blk-ioprio: Convert from
rqos policy to direct call")

Fixes: a78418e6a04c ("block: Always initialize bio IO priority on submit")

Co-developed-by: Yibin Ding <yibin.ding@unisoc.com>
Signed-off-by: Yibin Ding <yibin.ding@unisoc.com>
Signed-off-by: Hongyu Jin <hongyu.jin@unisoc.com>
---
 block/blk-core.c | 10 ++++++++++
 block/blk-mq.c   | 11 -----------
 2 files changed, 10 insertions(+), 11 deletions(-)

diff --git a/block/blk-core.c b/block/blk-core.c
index 2eca76ccf4ee..d707ec056f34 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -49,6 +49,7 @@
 #include "blk-pm.h"
 #include "blk-cgroup.h"
 #include "blk-throttle.h"
+#include "blk-ioprio.h"
 
 struct dentry *blk_debugfs_root;
 
@@ -817,6 +818,14 @@ void submit_bio_noacct(struct bio *bio)
 }
 EXPORT_SYMBOL(submit_bio_noacct);
 
+static void bio_set_ioprio(struct bio *bio)
+{
+	/* Nobody set ioprio so far? Initialize it based on task's nice value */
+	if (IOPRIO_PRIO_CLASS(bio->bi_ioprio) == IOPRIO_CLASS_NONE)
+		bio->bi_ioprio = get_current_ioprio();
+	blkcg_set_ioprio(bio);
+}
+
 /**
  * submit_bio - submit a bio to the block device layer for I/O
  * @bio: The &struct bio which describes the I/O
@@ -839,6 +848,7 @@ void submit_bio(struct bio *bio)
 		count_vm_events(PGPGOUT, bio_sectors(bio));
 	}
 
+	bio_set_ioprio(bio);
 	submit_bio_noacct(bio);
 }
 EXPORT_SYMBOL(submit_bio);
diff --git a/block/blk-mq.c b/block/blk-mq.c
index ac18f802c027..351e8283eda1 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -40,7 +40,6 @@
 #include "blk-stat.h"
 #include "blk-mq-sched.h"
 #include "blk-rq-qos.h"
-#include "blk-ioprio.h"
 
 static DEFINE_PER_CPU(struct llist_head, blk_cpu_done);
 static DEFINE_PER_CPU(call_single_data_t, blk_cpu_csd);
@@ -2919,14 +2918,6 @@ static bool blk_mq_can_use_cached_rq(struct request *rq, struct blk_plug *plug,
 	return true;
 }
 
-static void bio_set_ioprio(struct bio *bio)
-{
-	/* Nobody set ioprio so far? Initialize it based on task's nice value */
-	if (IOPRIO_PRIO_CLASS(bio->bi_ioprio) == IOPRIO_CLASS_NONE)
-		bio->bi_ioprio = get_current_ioprio();
-	blkcg_set_ioprio(bio);
-}
-
 /**
  * blk_mq_submit_bio - Create and send a request to block device.
  * @bio: Bio pointer.
@@ -2957,8 +2948,6 @@ void blk_mq_submit_bio(struct bio *bio)
 			return;
 	}
 
-	bio_set_ioprio(bio);
-
 	if (plug) {
 		rq = rq_list_peek(&plug->cached_rq);
 		if (rq && rq->q != q)
-- 
2.34.1
[PATCH v7 2/5] dm: Support I/O priority for dm_io()
Posted by Hongyu Jin 1 year, 12 months ago
From: Hongyu Jin <hongyu.jin@unisoc.com>

Some I/O will dispatch from kworker with different io_context settings
than the submitting task, we may need to specify a priority to avoid
losing priority.

Add I/O priority parameter for dm_io().

Co-developed-by: Yibin Ding <yibin.ding@unisoc.com>
Signed-off-by: Yibin Ding <yibin.ding@unisoc.com>
Signed-off-by: Hongyu Jin <hongyu.jin@unisoc.com>
---
 drivers/md/dm-bufio.c           |  6 +++---
 drivers/md/dm-integrity.c       | 10 +++++-----
 drivers/md/dm-io.c              | 23 +++++++++++++----------
 drivers/md/dm-kcopyd.c          |  4 ++--
 drivers/md/dm-log.c             |  4 ++--
 drivers/md/dm-raid1.c           |  6 +++---
 drivers/md/dm-snap-persistent.c |  4 ++--
 drivers/md/dm-writecache.c      |  8 ++++----
 include/linux/dm-io.h           |  3 ++-
 9 files changed, 36 insertions(+), 32 deletions(-)

diff --git a/drivers/md/dm-bufio.c b/drivers/md/dm-bufio.c
index f03d7dba270c..4f2808ef387f 100644
--- a/drivers/md/dm-bufio.c
+++ b/drivers/md/dm-bufio.c
@@ -1315,7 +1315,7 @@ static void use_dmio(struct dm_buffer *b, enum req_op op, sector_t sector,
 		io_req.mem.ptr.vma = (char *)b->data + offset;
 	}
 
-	r = dm_io(&io_req, 1, &region, NULL);
+	r = dm_io(&io_req, 1, &region, NULL, IOPRIO_DEFAULT);
 	if (unlikely(r))
 		b->end_io(b, errno_to_blk_status(r));
 }
@@ -2167,7 +2167,7 @@ int dm_bufio_issue_flush(struct dm_bufio_client *c)
 	if (WARN_ON_ONCE(dm_bufio_in_request()))
 		return -EINVAL;
 
-	return dm_io(&io_req, 1, &io_reg, NULL);
+	return dm_io(&io_req, 1, &io_reg, NULL, IOPRIO_DEFAULT);
 }
 EXPORT_SYMBOL_GPL(dm_bufio_issue_flush);
 
@@ -2191,7 +2191,7 @@ int dm_bufio_issue_discard(struct dm_bufio_client *c, sector_t block, sector_t c
 	if (WARN_ON_ONCE(dm_bufio_in_request()))
 		return -EINVAL; /* discards are optional */
 
-	return dm_io(&io_req, 1, &io_reg, NULL);
+	return dm_io(&io_req, 1, &io_reg, NULL, IOPRIO_DEFAULT);
 }
 EXPORT_SYMBOL_GPL(dm_bufio_issue_discard);
 
diff --git a/drivers/md/dm-integrity.c b/drivers/md/dm-integrity.c
index e85c688fd91e..9ffd093ad6cc 100644
--- a/drivers/md/dm-integrity.c
+++ b/drivers/md/dm-integrity.c
@@ -553,7 +553,7 @@ static int sync_rw_sb(struct dm_integrity_c *ic, blk_opf_t opf)
 		}
 	}
 
-	r = dm_io(&io_req, 1, &io_loc, NULL);
+	r = dm_io(&io_req, 1, &io_loc, NULL, IOPRIO_DEFAULT);
 	if (unlikely(r))
 		return r;
 
@@ -1071,7 +1071,7 @@ static void rw_journal_sectors(struct dm_integrity_c *ic, blk_opf_t opf,
 	io_loc.sector = ic->start + SB_SECTORS + sector;
 	io_loc.count = n_sectors;
 
-	r = dm_io(&io_req, 1, &io_loc, NULL);
+	r = dm_io(&io_req, 1, &io_loc, NULL, IOPRIO_DEFAULT);
 	if (unlikely(r)) {
 		dm_integrity_io_error(ic, (opf & REQ_OP_MASK) == REQ_OP_READ ?
 				      "reading journal" : "writing journal", r);
@@ -1188,7 +1188,7 @@ static void copy_from_journal(struct dm_integrity_c *ic, unsigned int section, u
 	io_loc.sector = target;
 	io_loc.count = n_sectors;
 
-	r = dm_io(&io_req, 1, &io_loc, NULL);
+	r = dm_io(&io_req, 1, &io_loc, NULL, IOPRIO_DEFAULT);
 	if (unlikely(r)) {
 		WARN_ONCE(1, "asynchronous dm_io failed: %d", r);
 		fn(-1UL, data);
@@ -1517,7 +1517,7 @@ static void dm_integrity_flush_buffers(struct dm_integrity_c *ic, bool flush_dat
 		fr.io_reg.count = 0,
 		fr.ic = ic;
 		init_completion(&fr.comp);
-		r = dm_io(&fr.io_req, 1, &fr.io_reg, NULL);
+		r = dm_io(&fr.io_req, 1, &fr.io_reg, NULL, IOPRIO_DEFAULT);
 		BUG_ON(r);
 	}
 
@@ -2739,7 +2739,7 @@ static void integrity_recalc(struct work_struct *w)
 	io_loc.sector = get_data_sector(ic, area, offset);
 	io_loc.count = n_sectors;
 
-	r = dm_io(&io_req, 1, &io_loc, NULL);
+	r = dm_io(&io_req, 1, &io_loc, NULL, IOPRIO_DEFAULT);
 	if (unlikely(r)) {
 		dm_integrity_io_error(ic, "reading data", r);
 		goto err;
diff --git a/drivers/md/dm-io.c b/drivers/md/dm-io.c
index f053ce245814..7409490259d1 100644
--- a/drivers/md/dm-io.c
+++ b/drivers/md/dm-io.c
@@ -305,7 +305,7 @@ static void km_dp_init(struct dpages *dp, void *data)
  */
 static void do_region(const blk_opf_t opf, unsigned int region,
 		      struct dm_io_region *where, struct dpages *dp,
-		      struct io *io)
+		      struct io *io, unsigned short ioprio)
 {
 	struct bio *bio;
 	struct page *page;
@@ -354,6 +354,7 @@ static void do_region(const blk_opf_t opf, unsigned int region,
 				       &io->client->bios);
 		bio->bi_iter.bi_sector = where->sector + (where->count - remaining);
 		bio->bi_end_io = endio;
+		bio->bi_ioprio = ioprio;
 		store_io_and_region_in_bio(bio, io, region);
 
 		if (op == REQ_OP_DISCARD || op == REQ_OP_WRITE_ZEROES) {
@@ -383,7 +384,7 @@ static void do_region(const blk_opf_t opf, unsigned int region,
 
 static void dispatch_io(blk_opf_t opf, unsigned int num_regions,
 			struct dm_io_region *where, struct dpages *dp,
-			struct io *io, int sync)
+			struct io *io, int sync, unsigned short ioprio)
 {
 	int i;
 	struct dpages old_pages = *dp;
@@ -400,7 +401,7 @@ static void dispatch_io(blk_opf_t opf, unsigned int num_regions,
 	for (i = 0; i < num_regions; i++) {
 		*dp = old_pages;
 		if (where[i].count || (opf & REQ_PREFLUSH))
-			do_region(opf, i, where + i, dp, io);
+			do_region(opf, i, where + i, dp, io, ioprio);
 	}
 
 	/*
@@ -425,7 +426,7 @@ static void sync_io_complete(unsigned long error, void *context)
 
 static int sync_io(struct dm_io_client *client, unsigned int num_regions,
 		   struct dm_io_region *where, blk_opf_t opf, struct dpages *dp,
-		   unsigned long *error_bits)
+		   unsigned long *error_bits, unsigned short ioprio)
 {
 	struct io *io;
 	struct sync_io sio;
@@ -447,7 +448,7 @@ static int sync_io(struct dm_io_client *client, unsigned int num_regions,
 	io->vma_invalidate_address = dp->vma_invalidate_address;
 	io->vma_invalidate_size = dp->vma_invalidate_size;
 
-	dispatch_io(opf, num_regions, where, dp, io, 1);
+	dispatch_io(opf, num_regions, where, dp, io, 1, ioprio);
 
 	wait_for_completion_io(&sio.wait);
 
@@ -459,7 +460,8 @@ static int sync_io(struct dm_io_client *client, unsigned int num_regions,
 
 static int async_io(struct dm_io_client *client, unsigned int num_regions,
 		    struct dm_io_region *where, blk_opf_t opf,
-		    struct dpages *dp, io_notify_fn fn, void *context)
+		    struct dpages *dp, io_notify_fn fn, void *context,
+		    unsigned short ioprio)
 {
 	struct io *io;
 
@@ -479,7 +481,7 @@ static int async_io(struct dm_io_client *client, unsigned int num_regions,
 	io->vma_invalidate_address = dp->vma_invalidate_address;
 	io->vma_invalidate_size = dp->vma_invalidate_size;
 
-	dispatch_io(opf, num_regions, where, dp, io, 0);
+	dispatch_io(opf, num_regions, where, dp, io, 0, ioprio);
 	return 0;
 }
 
@@ -521,7 +523,8 @@ static int dp_init(struct dm_io_request *io_req, struct dpages *dp,
 }
 
 int dm_io(struct dm_io_request *io_req, unsigned int num_regions,
-	  struct dm_io_region *where, unsigned long *sync_error_bits)
+	  struct dm_io_region *where, unsigned long *sync_error_bits,
+	  unsigned short ioprio)
 {
 	int r;
 	struct dpages dp;
@@ -532,11 +535,11 @@ int dm_io(struct dm_io_request *io_req, unsigned int num_regions,
 
 	if (!io_req->notify.fn)
 		return sync_io(io_req->client, num_regions, where,
-			       io_req->bi_opf, &dp, sync_error_bits);
+			       io_req->bi_opf, &dp, sync_error_bits, ioprio);
 
 	return async_io(io_req->client, num_regions, where,
 			io_req->bi_opf, &dp, io_req->notify.fn,
-			io_req->notify.context);
+			io_req->notify.context, ioprio);
 }
 EXPORT_SYMBOL(dm_io);
 
diff --git a/drivers/md/dm-kcopyd.c b/drivers/md/dm-kcopyd.c
index d01807c50f20..79c65c9ad5fa 100644
--- a/drivers/md/dm-kcopyd.c
+++ b/drivers/md/dm-kcopyd.c
@@ -578,9 +578,9 @@ static int run_io_job(struct kcopyd_job *job)
 	io_job_start(job->kc->throttle);
 
 	if (job->op == REQ_OP_READ)
-		r = dm_io(&io_req, 1, &job->source, NULL);
+		r = dm_io(&io_req, 1, &job->source, NULL, IOPRIO_DEFAULT);
 	else
-		r = dm_io(&io_req, job->num_dests, job->dests, NULL);
+		r = dm_io(&io_req, job->num_dests, job->dests, NULL, IOPRIO_DEFAULT);
 
 	return r;
 }
diff --git a/drivers/md/dm-log.c b/drivers/md/dm-log.c
index f9f84236dfcd..f7f9c2100937 100644
--- a/drivers/md/dm-log.c
+++ b/drivers/md/dm-log.c
@@ -300,7 +300,7 @@ static int rw_header(struct log_c *lc, enum req_op op)
 {
 	lc->io_req.bi_opf = op;
 
-	return dm_io(&lc->io_req, 1, &lc->header_location, NULL);
+	return dm_io(&lc->io_req, 1, &lc->header_location, NULL, IOPRIO_DEFAULT);
 }
 
 static int flush_header(struct log_c *lc)
@@ -313,7 +313,7 @@ static int flush_header(struct log_c *lc)
 
 	lc->io_req.bi_opf = REQ_OP_WRITE | REQ_PREFLUSH;
 
-	return dm_io(&lc->io_req, 1, &null_location, NULL);
+	return dm_io(&lc->io_req, 1, &null_location, NULL, IOPRIO_DEFAULT);
 }
 
 static int read_header(struct log_c *log)
diff --git a/drivers/md/dm-raid1.c b/drivers/md/dm-raid1.c
index ddcb2bc4a617..9511dae5b556 100644
--- a/drivers/md/dm-raid1.c
+++ b/drivers/md/dm-raid1.c
@@ -278,7 +278,7 @@ static int mirror_flush(struct dm_target *ti)
 	}
 
 	error_bits = -1;
-	dm_io(&io_req, ms->nr_mirrors, io, &error_bits);
+	dm_io(&io_req, ms->nr_mirrors, io, &error_bits, IOPRIO_DEFAULT);
 	if (unlikely(error_bits != 0)) {
 		for (i = 0; i < ms->nr_mirrors; i++)
 			if (test_bit(i, &error_bits))
@@ -554,7 +554,7 @@ static void read_async_bio(struct mirror *m, struct bio *bio)
 
 	map_region(&io, m, bio);
 	bio_set_m(bio, m);
-	BUG_ON(dm_io(&io_req, 1, &io, NULL));
+	BUG_ON(dm_io(&io_req, 1, &io, NULL, IOPRIO_DEFAULT));
 }
 
 static inline int region_in_sync(struct mirror_set *ms, region_t region,
@@ -681,7 +681,7 @@ static void do_write(struct mirror_set *ms, struct bio *bio)
 	 */
 	bio_set_m(bio, get_default_mirror(ms));
 
-	BUG_ON(dm_io(&io_req, ms->nr_mirrors, io, NULL));
+	BUG_ON(dm_io(&io_req, ms->nr_mirrors, io, NULL, IOPRIO_DEFAULT));
 }
 
 static void do_writes(struct mirror_set *ms, struct bio_list *writes)
diff --git a/drivers/md/dm-snap-persistent.c b/drivers/md/dm-snap-persistent.c
index 15649921f2a9..568d10842b1f 100644
--- a/drivers/md/dm-snap-persistent.c
+++ b/drivers/md/dm-snap-persistent.c
@@ -223,7 +223,7 @@ static void do_metadata(struct work_struct *work)
 {
 	struct mdata_req *req = container_of(work, struct mdata_req, work);
 
-	req->result = dm_io(req->io_req, 1, req->where, NULL);
+	req->result = dm_io(req->io_req, 1, req->where, NULL, IOPRIO_DEFAULT);
 }
 
 /*
@@ -247,7 +247,7 @@ static int chunk_io(struct pstore *ps, void *area, chunk_t chunk, blk_opf_t opf,
 	struct mdata_req req;
 
 	if (!metadata)
-		return dm_io(&io_req, 1, &where, NULL);
+		return dm_io(&io_req, 1, &where, NULL, IOPRIO_DEFAULT);
 
 	req.where = &where;
 	req.io_req = &io_req;
diff --git a/drivers/md/dm-writecache.c b/drivers/md/dm-writecache.c
index 074cb785eafc..6a4279bfb1e7 100644
--- a/drivers/md/dm-writecache.c
+++ b/drivers/md/dm-writecache.c
@@ -531,7 +531,7 @@ static void ssd_commit_flushed(struct dm_writecache *wc, bool wait_for_ios)
 		req.notify.context = &endio;
 
 		/* writing via async dm-io (implied by notify.fn above) won't return an error */
-		(void) dm_io(&req, 1, &region, NULL);
+		(void) dm_io(&req, 1, &region, NULL, IOPRIO_DEFAULT);
 		i = j;
 	}
 
@@ -568,7 +568,7 @@ static void ssd_commit_superblock(struct dm_writecache *wc)
 	req.notify.fn = NULL;
 	req.notify.context = NULL;
 
-	r = dm_io(&req, 1, &region, NULL);
+	r = dm_io(&req, 1, &region, NULL, IOPRIO_DEFAULT);
 	if (unlikely(r))
 		writecache_error(wc, r, "error writing superblock");
 }
@@ -596,7 +596,7 @@ static void writecache_disk_flush(struct dm_writecache *wc, struct dm_dev *dev)
 	req.client = wc->dm_io;
 	req.notify.fn = NULL;
 
-	r = dm_io(&req, 1, &region, NULL);
+	r = dm_io(&req, 1, &region, NULL, IOPRIO_DEFAULT);
 	if (unlikely(r))
 		writecache_error(wc, r, "error flushing metadata: %d", r);
 }
@@ -990,7 +990,7 @@ static int writecache_read_metadata(struct dm_writecache *wc, sector_t n_sectors
 	req.client = wc->dm_io;
 	req.notify.fn = NULL;
 
-	return dm_io(&req, 1, &region, NULL);
+	return dm_io(&req, 1, &region, NULL, IOPRIO_DEFAULT);
 }
 
 static void writecache_resume(struct dm_target *ti)
diff --git a/include/linux/dm-io.h b/include/linux/dm-io.h
index 7595142f3fc5..7b2968612b7e 100644
--- a/include/linux/dm-io.h
+++ b/include/linux/dm-io.h
@@ -80,7 +80,8 @@ void dm_io_client_destroy(struct dm_io_client *client);
  * error occurred doing io to the corresponding region.
  */
 int dm_io(struct dm_io_request *io_req, unsigned int num_regions,
-	  struct dm_io_region *region, unsigned int long *sync_error_bits);
+	  struct dm_io_region *region, unsigned int long *sync_error_bits,
+	  unsigned short ioprio);
 
 #endif	/* __KERNEL__ */
 #endif	/* _LINUX_DM_IO_H */
-- 
2.34.1
[PATCH v7 3/5] dm-bufio: Support I/O priority
Posted by Hongyu Jin 1 year, 12 months ago
From: Hongyu Jin <hongyu.jin@unisoc.com>

Some I/O will dispatch from kworker with different io_context settings
than the submitting task, we may need to specify a priority to avoid
losing priority.

Add I/O priority parameter for dm_bufio_read() and
dm_bufio_prefetch().

Co-developed-by: Yibin Ding <yibin.ding@unisoc.com>
Signed-off-by: Yibin Ding <yibin.ding@unisoc.com>
Signed-off-by: Hongyu Jin <hongyu.jin@unisoc.com>
---
 drivers/md/dm-bufio.c                         | 39 +++++++++++--------
 drivers/md/dm-ebs-target.c                    |  8 ++--
 drivers/md/dm-integrity.c                     |  2 +-
 drivers/md/dm-snap-persistent.c               |  4 +-
 drivers/md/dm-verity-fec.c                    |  4 +-
 drivers/md/dm-verity-target.c                 |  5 ++-
 drivers/md/persistent-data/dm-block-manager.c |  6 +--
 include/linux/dm-bufio.h                      |  5 ++-
 8 files changed, 40 insertions(+), 33 deletions(-)

diff --git a/drivers/md/dm-bufio.c b/drivers/md/dm-bufio.c
index 4f2808ef387f..a6974ecab68e 100644
--- a/drivers/md/dm-bufio.c
+++ b/drivers/md/dm-bufio.c
@@ -1292,7 +1292,8 @@ static void dmio_complete(unsigned long error, void *context)
 }
 
 static void use_dmio(struct dm_buffer *b, enum req_op op, sector_t sector,
-		     unsigned int n_sectors, unsigned int offset)
+		     unsigned int n_sectors, unsigned int offset,
+		     unsigned short ioprio)
 {
 	int r;
 	struct dm_io_request io_req = {
@@ -1315,7 +1316,7 @@ static void use_dmio(struct dm_buffer *b, enum req_op op, sector_t sector,
 		io_req.mem.ptr.vma = (char *)b->data + offset;
 	}
 
-	r = dm_io(&io_req, 1, &region, NULL, IOPRIO_DEFAULT);
+	r = dm_io(&io_req, 1, &region, NULL, ioprio);
 	if (unlikely(r))
 		b->end_io(b, errno_to_blk_status(r));
 }
@@ -1331,7 +1332,8 @@ static void bio_complete(struct bio *bio)
 }
 
 static void use_bio(struct dm_buffer *b, enum req_op op, sector_t sector,
-		    unsigned int n_sectors, unsigned int offset)
+		    unsigned int n_sectors, unsigned int offset,
+		    unsigned short ioprio)
 {
 	struct bio *bio;
 	char *ptr;
@@ -1339,13 +1341,14 @@ static void use_bio(struct dm_buffer *b, enum req_op op, sector_t sector,
 
 	bio = bio_kmalloc(1, GFP_NOWAIT | __GFP_NORETRY | __GFP_NOWARN);
 	if (!bio) {
-		use_dmio(b, op, sector, n_sectors, offset);
+		use_dmio(b, op, sector, n_sectors, offset, ioprio);
 		return;
 	}
 	bio_init(bio, b->c->bdev, bio->bi_inline_vecs, 1, op);
 	bio->bi_iter.bi_sector = sector;
 	bio->bi_end_io = bio_complete;
 	bio->bi_private = b;
+	bio->bi_ioprio = ioprio;
 
 	ptr = (char *)b->data + offset;
 	len = n_sectors << SECTOR_SHIFT;
@@ -1368,7 +1371,7 @@ static inline sector_t block_to_sector(struct dm_bufio_client *c, sector_t block
 	return sector;
 }
 
-static void submit_io(struct dm_buffer *b, enum req_op op,
+static void submit_io(struct dm_buffer *b, enum req_op op, unsigned short ioprio,
 		      void (*end_io)(struct dm_buffer *, blk_status_t))
 {
 	unsigned int n_sectors;
@@ -1398,9 +1401,9 @@ static void submit_io(struct dm_buffer *b, enum req_op op,
 	}
 
 	if (b->data_mode != DATA_MODE_VMALLOC)
-		use_bio(b, op, sector, n_sectors, offset);
+		use_bio(b, op, sector, n_sectors, offset, ioprio);
 	else
-		use_dmio(b, op, sector, n_sectors, offset);
+		use_dmio(b, op, sector, n_sectors, offset, ioprio);
 }
 
 /*
@@ -1456,7 +1459,7 @@ static void __write_dirty_buffer(struct dm_buffer *b,
 	b->write_end = b->dirty_end;
 
 	if (!write_list)
-		submit_io(b, REQ_OP_WRITE, write_endio);
+		submit_io(b, REQ_OP_WRITE, IOPRIO_DEFAULT, write_endio);
 	else
 		list_add_tail(&b->write_list, write_list);
 }
@@ -1470,7 +1473,7 @@ static void __flush_write_list(struct list_head *write_list)
 		struct dm_buffer *b =
 			list_entry(write_list->next, struct dm_buffer, write_list);
 		list_del(&b->write_list);
-		submit_io(b, REQ_OP_WRITE, write_endio);
+		submit_io(b, REQ_OP_WRITE, IOPRIO_DEFAULT, write_endio);
 		cond_resched();
 	}
 	blk_finish_plug(&plug);
@@ -1852,7 +1855,8 @@ static void read_endio(struct dm_buffer *b, blk_status_t status)
  * and uses dm_bufio_mark_buffer_dirty to write new data back).
  */
 static void *new_read(struct dm_bufio_client *c, sector_t block,
-		      enum new_flag nf, struct dm_buffer **bp)
+		      enum new_flag nf, struct dm_buffer **bp,
+		      unsigned short ioprio)
 {
 	int need_submit = 0;
 	struct dm_buffer *b;
@@ -1905,7 +1909,7 @@ static void *new_read(struct dm_bufio_client *c, sector_t block,
 		return NULL;
 
 	if (need_submit)
-		submit_io(b, REQ_OP_READ, read_endio);
+		submit_io(b, REQ_OP_READ, ioprio, read_endio);
 
 	if (nf != NF_GET)	/* we already tested this condition above */
 		wait_on_bit_io(&b->state, B_READING, TASK_UNINTERRUPTIBLE);
@@ -1926,17 +1930,17 @@ static void *new_read(struct dm_bufio_client *c, sector_t block,
 void *dm_bufio_get(struct dm_bufio_client *c, sector_t block,
 		   struct dm_buffer **bp)
 {
-	return new_read(c, block, NF_GET, bp);
+	return new_read(c, block, NF_GET, bp, IOPRIO_DEFAULT);
 }
 EXPORT_SYMBOL_GPL(dm_bufio_get);
 
 void *dm_bufio_read(struct dm_bufio_client *c, sector_t block,
-		    struct dm_buffer **bp)
+		    struct dm_buffer **bp, unsigned short ioprio)
 {
 	if (WARN_ON_ONCE(dm_bufio_in_request()))
 		return ERR_PTR(-EINVAL);
 
-	return new_read(c, block, NF_READ, bp);
+	return new_read(c, block, NF_READ, bp, ioprio);
 }
 EXPORT_SYMBOL_GPL(dm_bufio_read);
 
@@ -1946,12 +1950,13 @@ void *dm_bufio_new(struct dm_bufio_client *c, sector_t block,
 	if (WARN_ON_ONCE(dm_bufio_in_request()))
 		return ERR_PTR(-EINVAL);
 
-	return new_read(c, block, NF_FRESH, bp);
+	return new_read(c, block, NF_FRESH, bp, IOPRIO_DEFAULT);
 }
 EXPORT_SYMBOL_GPL(dm_bufio_new);
 
 void dm_bufio_prefetch(struct dm_bufio_client *c,
-		       sector_t block, unsigned int n_blocks)
+		       sector_t block, unsigned int n_blocks,
+		       unsigned short ioprio)
 {
 	struct blk_plug plug;
 
@@ -1987,7 +1992,7 @@ void dm_bufio_prefetch(struct dm_bufio_client *c,
 			dm_bufio_unlock(c);
 
 			if (need_submit)
-				submit_io(b, REQ_OP_READ, read_endio);
+				submit_io(b, REQ_OP_READ, ioprio, read_endio);
 			dm_bufio_release(b);
 
 			cond_resched();
diff --git a/drivers/md/dm-ebs-target.c b/drivers/md/dm-ebs-target.c
index 435b45201f4d..8198c8a7b416 100644
--- a/drivers/md/dm-ebs-target.c
+++ b/drivers/md/dm-ebs-target.c
@@ -84,7 +84,7 @@ static int __ebs_rw_bvec(struct ebs_c *ec, enum req_op op, struct bio_vec *bv,
 
 		/* Avoid reading for writes in case bio vector's page overwrites block completely. */
 		if (op == REQ_OP_READ || buf_off || bv_len < dm_bufio_get_block_size(ec->bufio))
-			ba = dm_bufio_read(ec->bufio, block, &b);
+			ba = dm_bufio_read(ec->bufio, block, &b, IOPRIO_DEFAULT);
 		else
 			ba = dm_bufio_new(ec->bufio, block, &b);
 
@@ -194,13 +194,13 @@ static void __ebs_process_bios(struct work_struct *ws)
 	bio_list_for_each(bio, &bios) {
 		block1 = __sector_to_block(ec, bio->bi_iter.bi_sector);
 		if (bio_op(bio) == REQ_OP_READ)
-			dm_bufio_prefetch(ec->bufio, block1, __nr_blocks(ec, bio));
+			dm_bufio_prefetch(ec->bufio, block1, __nr_blocks(ec, bio), IOPRIO_DEFAULT);
 		else if (bio_op(bio) == REQ_OP_WRITE && !(bio->bi_opf & REQ_PREFLUSH)) {
 			block2 = __sector_to_block(ec, bio_end_sector(bio));
 			if (__block_mod(bio->bi_iter.bi_sector, ec->u_bs))
-				dm_bufio_prefetch(ec->bufio, block1, 1);
+				dm_bufio_prefetch(ec->bufio, block1, 1, IOPRIO_DEFAULT);
 			if (__block_mod(bio_end_sector(bio), ec->u_bs) && block2 != block1)
-				dm_bufio_prefetch(ec->bufio, block2, 1);
+				dm_bufio_prefetch(ec->bufio, block2, 1, IOPRIO_DEFAULT);
 		}
 	}
 
diff --git a/drivers/md/dm-integrity.c b/drivers/md/dm-integrity.c
index 9ffd093ad6cc..1e40e712bcd7 100644
--- a/drivers/md/dm-integrity.c
+++ b/drivers/md/dm-integrity.c
@@ -1418,7 +1418,7 @@ static int dm_integrity_rw_tag(struct dm_integrity_c *ic, unsigned char *tag, se
 		if (unlikely(r))
 			return r;
 
-		data = dm_bufio_read(ic->bufio, *metadata_block, &b);
+		data = dm_bufio_read(ic->bufio, *metadata_block, &b, IOPRIO_DEFAULT);
 		if (IS_ERR(data))
 			return PTR_ERR(data);
 
diff --git a/drivers/md/dm-snap-persistent.c b/drivers/md/dm-snap-persistent.c
index 568d10842b1f..a2072b95e28c 100644
--- a/drivers/md/dm-snap-persistent.c
+++ b/drivers/md/dm-snap-persistent.c
@@ -524,7 +524,7 @@ static int read_exceptions(struct pstore *ps,
 
 				if (unlikely(pf_chunk >= dm_bufio_get_device_size(client)))
 					break;
-				dm_bufio_prefetch(client, pf_chunk, 1);
+				dm_bufio_prefetch(client, pf_chunk, 1, IOPRIO_DEFAULT);
 				prefetch_area++;
 				if (unlikely(!prefetch_area))
 					break;
@@ -533,7 +533,7 @@ static int read_exceptions(struct pstore *ps,
 
 		chunk = area_location(ps, ps->current_area);
 
-		area = dm_bufio_read(client, chunk, &bp);
+		area = dm_bufio_read(client, chunk, &bp, IOPRIO_DEFAULT);
 		if (IS_ERR(area)) {
 			r = PTR_ERR(area);
 			goto ret_destroy_bufio;
diff --git a/drivers/md/dm-verity-fec.c b/drivers/md/dm-verity-fec.c
index b475200d8586..49db19e537f9 100644
--- a/drivers/md/dm-verity-fec.c
+++ b/drivers/md/dm-verity-fec.c
@@ -69,7 +69,7 @@ static u8 *fec_read_parity(struct dm_verity *v, u64 rsb, int index,
 	block = div64_u64_rem(position, v->fec->io_size, &rem);
 	*offset = (unsigned int)rem;
 
-	res = dm_bufio_read(v->fec->bufio, block, buf);
+	res = dm_bufio_read(v->fec->bufio, block, buf, IOPRIO_DEFAULT);
 	if (IS_ERR(res)) {
 		DMERR("%s: FEC %llu: parity read failed (block %llu): %ld",
 		      v->data_dev->name, (unsigned long long)rsb,
@@ -248,7 +248,7 @@ static int fec_read_bufs(struct dm_verity *v, struct dm_verity_io *io,
 			bufio = v->bufio;
 		}
 
-		bbuf = dm_bufio_read(bufio, block, &buf);
+		bbuf = dm_bufio_read(bufio, block, &buf, IOPRIO_DEFAULT);
 		if (IS_ERR(bbuf)) {
 			DMWARN_LIMIT("%s: FEC %llu: read failed (%llu): %ld",
 				     v->data_dev->name,
diff --git a/drivers/md/dm-verity-target.c b/drivers/md/dm-verity-target.c
index 14e58ae70521..4758bfe2c156 100644
--- a/drivers/md/dm-verity-target.c
+++ b/drivers/md/dm-verity-target.c
@@ -308,7 +308,7 @@ static int verity_verify_level(struct dm_verity *v, struct dm_verity_io *io,
 			return -EAGAIN;
 		}
 	} else
-		data = dm_bufio_read(v->bufio, hash_block, &buf);
+		data = dm_bufio_read(v->bufio, hash_block, &buf, IOPRIO_DEFAULT);
 
 	if (IS_ERR(data))
 		return PTR_ERR(data);
@@ -719,7 +719,8 @@ static void verity_prefetch_io(struct work_struct *work)
 		}
 no_prefetch_cluster:
 		dm_bufio_prefetch(v->bufio, hash_block_start,
-				  hash_block_end - hash_block_start + 1);
+				  hash_block_end - hash_block_start + 1,
+				  IOPRIO_DEFAULT);
 	}
 
 	kfree(pw);
diff --git a/drivers/md/persistent-data/dm-block-manager.c b/drivers/md/persistent-data/dm-block-manager.c
index 0e010e1204aa..86a4f73d2f3d 100644
--- a/drivers/md/persistent-data/dm-block-manager.c
+++ b/drivers/md/persistent-data/dm-block-manager.c
@@ -474,7 +474,7 @@ int dm_bm_read_lock(struct dm_block_manager *bm, dm_block_t b,
 	void *p;
 	int r;
 
-	p = dm_bufio_read(bm->bufio, b, (struct dm_buffer **) result);
+	p = dm_bufio_read(bm->bufio, b, (struct dm_buffer **) result, IOPRIO_DEFAULT);
 	if (IS_ERR(p))
 		return PTR_ERR(p);
 
@@ -510,7 +510,7 @@ int dm_bm_write_lock(struct dm_block_manager *bm,
 	if (dm_bm_is_read_only(bm))
 		return -EPERM;
 
-	p = dm_bufio_read(bm->bufio, b, (struct dm_buffer **) result);
+	p = dm_bufio_read(bm->bufio, b, (struct dm_buffer **) result, IOPRIO_DEFAULT);
 	if (IS_ERR(p))
 		return PTR_ERR(p);
 
@@ -624,7 +624,7 @@ EXPORT_SYMBOL_GPL(dm_bm_flush);
 
 void dm_bm_prefetch(struct dm_block_manager *bm, dm_block_t b)
 {
-	dm_bufio_prefetch(bm->bufio, b, 1);
+	dm_bufio_prefetch(bm->bufio, b, 1, IOPRIO_DEFAULT);
 }
 
 bool dm_bm_is_read_only(struct dm_block_manager *bm)
diff --git a/include/linux/dm-bufio.h b/include/linux/dm-bufio.h
index 75e7d8cbb532..256a246c7b97 100644
--- a/include/linux/dm-bufio.h
+++ b/include/linux/dm-bufio.h
@@ -62,7 +62,7 @@ void dm_bufio_set_sector_offset(struct dm_bufio_client *c, sector_t start);
  * it dirty.
  */
 void *dm_bufio_read(struct dm_bufio_client *c, sector_t block,
-		    struct dm_buffer **bp);
+		    struct dm_buffer **bp, unsigned short ioprio);
 
 /*
  * Like dm_bufio_read, but return buffer from cache, don't read
@@ -84,7 +84,8 @@ void *dm_bufio_new(struct dm_bufio_client *c, sector_t block,
  * I/O to finish.
  */
 void dm_bufio_prefetch(struct dm_bufio_client *c,
-		       sector_t block, unsigned int n_blocks);
+		       sector_t block, unsigned int n_blocks,
+		       unsigned short ioprio);
 
 /*
  * Release a reference obtained with dm_bufio_{read,get,new}. The data
-- 
2.34.1
[PATCH v7 4/5] dm verity: Fix I/O priority lost when read FEC and hash
Posted by Hongyu Jin 1 year, 12 months ago
From: Hongyu Jin <hongyu.jin@unisoc.com>

After obtaining the data, verification or error correction process may
trigger a new I/O that loses the priority of the original I/O, that is,
the verification of the higher priority IO may be blocked by the lower
priority IO.

Make the I/O of verification and error correction follow the
priority of original I/O.

Co-developed-by: Yibin Ding <yibin.ding@unisoc.com>
Signed-off-by: Yibin Ding <yibin.ding@unisoc.com>
Signed-off-by: Hongyu Jin <hongyu.jin@unisoc.com>
---
 drivers/md/dm-verity-fec.c    | 21 ++++++++++++---------
 drivers/md/dm-verity-target.c | 12 ++++++++----
 2 files changed, 20 insertions(+), 13 deletions(-)

diff --git a/drivers/md/dm-verity-fec.c b/drivers/md/dm-verity-fec.c
index 49db19e537f9..066521de08da 100644
--- a/drivers/md/dm-verity-fec.c
+++ b/drivers/md/dm-verity-fec.c
@@ -60,7 +60,8 @@ static int fec_decode_rs8(struct dm_verity *v, struct dm_verity_fec_io *fio,
  * to the data block. Caller is responsible for releasing buf.
  */
 static u8 *fec_read_parity(struct dm_verity *v, u64 rsb, int index,
-			   unsigned int *offset, struct dm_buffer **buf)
+			   unsigned int *offset, struct dm_buffer **buf,
+			   unsigned short ioprio)
 {
 	u64 position, block, rem;
 	u8 *res;
@@ -69,7 +70,7 @@ static u8 *fec_read_parity(struct dm_verity *v, u64 rsb, int index,
 	block = div64_u64_rem(position, v->fec->io_size, &rem);
 	*offset = (unsigned int)rem;
 
-	res = dm_bufio_read(v->fec->bufio, block, buf, IOPRIO_DEFAULT);
+	res = dm_bufio_read(v->fec->bufio, block, buf, ioprio);
 	if (IS_ERR(res)) {
 		DMERR("%s: FEC %llu: parity read failed (block %llu): %ld",
 		      v->data_dev->name, (unsigned long long)rsb,
@@ -121,16 +122,17 @@ static inline unsigned int fec_buffer_rs_index(unsigned int i, unsigned int j)
  * Decode all RS blocks from buffers and copy corrected bytes into fio->output
  * starting from block_offset.
  */
-static int fec_decode_bufs(struct dm_verity *v, struct dm_verity_fec_io *fio,
-			   u64 rsb, int byte_index, unsigned int block_offset,
-			   int neras)
+static int fec_decode_bufs(struct dm_verity *v, struct dm_verity_io *io,
+			   struct dm_verity_fec_io *fio, u64 rsb, int byte_index,
+			   unsigned int block_offset, int neras)
 {
 	int r, corrected = 0, res;
 	struct dm_buffer *buf;
 	unsigned int n, i, offset;
 	u8 *par, *block;
+	struct bio *bio = dm_bio_from_per_bio_data(io, v->ti->per_io_data_size);
 
-	par = fec_read_parity(v, rsb, block_offset, &offset, &buf);
+	par = fec_read_parity(v, rsb, block_offset, &offset, &buf, bio_prio(bio));
 	if (IS_ERR(par))
 		return PTR_ERR(par);
 
@@ -158,7 +160,7 @@ static int fec_decode_bufs(struct dm_verity *v, struct dm_verity_fec_io *fio,
 		if (offset >= v->fec->io_size) {
 			dm_bufio_release(buf);
 
-			par = fec_read_parity(v, rsb, block_offset, &offset, &buf);
+			par = fec_read_parity(v, rsb, block_offset, &offset, &buf, bio_prio(bio));
 			if (IS_ERR(par))
 				return PTR_ERR(par);
 		}
@@ -210,6 +212,7 @@ static int fec_read_bufs(struct dm_verity *v, struct dm_verity_io *io,
 	u8 *bbuf, *rs_block;
 	u8 want_digest[HASH_MAX_DIGESTSIZE];
 	unsigned int n, k;
+	struct bio *bio = dm_bio_from_per_bio_data(io, v->ti->per_io_data_size);
 
 	if (neras)
 		*neras = 0;
@@ -248,7 +251,7 @@ static int fec_read_bufs(struct dm_verity *v, struct dm_verity_io *io,
 			bufio = v->bufio;
 		}
 
-		bbuf = dm_bufio_read(bufio, block, &buf, IOPRIO_DEFAULT);
+		bbuf = dm_bufio_read(bufio, block, &buf, bio_prio(bio));
 		if (IS_ERR(bbuf)) {
 			DMWARN_LIMIT("%s: FEC %llu: read failed (%llu): %ld",
 				     v->data_dev->name,
@@ -377,7 +380,7 @@ static int fec_decode_rsb(struct dm_verity *v, struct dm_verity_io *io,
 		if (unlikely(r < 0))
 			return r;
 
-		r = fec_decode_bufs(v, fio, rsb, r, pos, neras);
+		r = fec_decode_bufs(v, io, fio, rsb, r, pos, neras);
 		if (r < 0)
 			return r;
 
diff --git a/drivers/md/dm-verity-target.c b/drivers/md/dm-verity-target.c
index 4758bfe2c156..8cbf81fc0031 100644
--- a/drivers/md/dm-verity-target.c
+++ b/drivers/md/dm-verity-target.c
@@ -51,6 +51,7 @@ static DEFINE_STATIC_KEY_FALSE(use_tasklet_enabled);
 struct dm_verity_prefetch_work {
 	struct work_struct work;
 	struct dm_verity *v;
+	unsigned short ioprio;
 	sector_t block;
 	unsigned int n_blocks;
 };
@@ -294,6 +295,7 @@ static int verity_verify_level(struct dm_verity *v, struct dm_verity_io *io,
 	int r;
 	sector_t hash_block;
 	unsigned int offset;
+	struct bio *bio = dm_bio_from_per_bio_data(io, v->ti->per_io_data_size);
 
 	verity_hash_at_level(v, block, level, &hash_block, &offset);
 
@@ -308,7 +310,7 @@ static int verity_verify_level(struct dm_verity *v, struct dm_verity_io *io,
 			return -EAGAIN;
 		}
 	} else
-		data = dm_bufio_read(v->bufio, hash_block, &buf, IOPRIO_DEFAULT);
+		data = dm_bufio_read(v->bufio, hash_block, &buf, bio_prio(bio));
 
 	if (IS_ERR(data))
 		return PTR_ERR(data);
@@ -720,13 +722,14 @@ static void verity_prefetch_io(struct work_struct *work)
 no_prefetch_cluster:
 		dm_bufio_prefetch(v->bufio, hash_block_start,
 				  hash_block_end - hash_block_start + 1,
-				  IOPRIO_DEFAULT);
+				  pw->ioprio);
 	}
 
 	kfree(pw);
 }
 
-static void verity_submit_prefetch(struct dm_verity *v, struct dm_verity_io *io)
+static void verity_submit_prefetch(struct dm_verity *v, struct dm_verity_io *io,
+				   unsigned short ioprio)
 {
 	sector_t block = io->block;
 	unsigned int n_blocks = io->n_blocks;
@@ -754,6 +757,7 @@ static void verity_submit_prefetch(struct dm_verity *v, struct dm_verity_io *io)
 	pw->v = v;
 	pw->block = block;
 	pw->n_blocks = n_blocks;
+	pw->ioprio = ioprio;
 	queue_work(v->verify_wq, &pw->work);
 }
 
@@ -796,7 +800,7 @@ static int verity_map(struct dm_target *ti, struct bio *bio)
 
 	verity_fec_init_io(io);
 
-	verity_submit_prefetch(v, io);
+	verity_submit_prefetch(v, io, bio_prio(bio));
 
 	submit_bio_noacct(bio);
 
-- 
2.34.1
[PATCH v7 5/5] dm-crypt: Fix lost ioprio when queuing write bios
Posted by Hongyu Jin 1 year, 12 months ago
From: Hongyu Jin <hongyu.jin@unisoc.com>

Since dm-crypt queues writes to a different kernel thread (workqueue),
the bios will dispatch from tasks with different io_context->ioprio
settings and blkcg than the submitting task, thus giving incorrect
ioprio to the io scheduler.

Get the original io priority setting via struct dm_crypt_io::base_bio
and set this priority to the bio for write.

Link: https://lore.kernel.org/dm-devel/alpine.LRH.2.11.1612141049250.13402@mail.ewheeler.net

Signed-off-by: Hongyu Jin <hongyu.jin@unisoc.com>
---
 drivers/md/dm-crypt.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/md/dm-crypt.c b/drivers/md/dm-crypt.c
index 2ae8560b6a14..ba6e794f7871 100644
--- a/drivers/md/dm-crypt.c
+++ b/drivers/md/dm-crypt.c
@@ -1683,6 +1683,7 @@ static struct bio *crypt_alloc_buffer(struct dm_crypt_io *io, unsigned int size)
 				 GFP_NOIO, &cc->bs);
 	clone->bi_private = io;
 	clone->bi_end_io = crypt_endio;
+	clone->bi_ioprio = io->base_bio->bi_ioprio;
 
 	remaining_size = size;
 
-- 
2.34.1
[PATCH v8 0/5] Fix I/O priority lost in device-mapper
Posted by Hongyu Jin 1 year, 10 months ago
From: Hongyu Jin <hongyu.jin@unisoc.com>

High-priority tasks get data from dm-verity devices via RT IO priority,
I/O will lose RT priority when reading FEC and hash values via kworker
submission IO during verification, and the verification phase may be
blocked by low-priority IO.

Dm-crypt has the same problem in the data writing process.

This is because io_context and blkcg are missing.

Move bio_set_ioprio() into submit_bio():
1. Only call bio_set_ioprio() once to set the priority of original bio,
   the bio that cloned and splited from original bio will auto inherit
   the priority of original bio in clone process.

2. Make the IO priority of the original bio to be passed to dm,
   and the dm target inherits the IO priority as needed.

Changes in v8:
  - Rebase patch 1 on commit 7ed2632ec7d7
Changes in v7:
  - Modify patch 4: change dm-verity-fec.c
Changes in v6:
  - Rebase patch and resolve conflict for patch 1, 3, 4
  - Modify patch 4: fec_read_parity() follow the priority of original
    bio
  - Update commit message
Changes in v5:
  - Rewrite patch 2, add ioprio parameter in dm_io();
  - Modify dm_io() in patch 3
Changes in v4:
  - Modify commit message by Suggestion
  - Modify patch for dm-crypt
Changes in v3:
  - Split patch for device-mapper
  - Add patch to fix dm-crypy I/O priority question
  - Add block patch to review together
  - Fix some error in v2 patch
Changes in v2:
  - Add ioprio field in struct dm_io_region
  - Initial struct dm_io_region::ioprio to IOPRIO_DEFAULT
  - Add two interface


Hongyu Jin (5):
  block: Fix bio IO priority setting
  dm: Support I/O priority for dm_io()
  dm-bufio: Support I/O priority
  dm verity: Fix I/O priority lost when read FEC and hash
  dm-crypt: Fix lost ioprio when queuing write bios

 block/blk-core.c                              | 10 +++++
 block/blk-mq.c                                | 10 -----
 drivers/md/dm-bufio.c                         | 43 +++++++++++--------
 drivers/md/dm-crypt.c                         |  1 +
 drivers/md/dm-ebs-target.c                    |  8 ++--
 drivers/md/dm-integrity.c                     | 12 +++---
 drivers/md/dm-io.c                            | 23 +++++-----
 drivers/md/dm-kcopyd.c                        |  4 +-
 drivers/md/dm-log.c                           |  4 +-
 drivers/md/dm-raid1.c                         |  6 +--
 drivers/md/dm-snap-persistent.c               |  8 ++--
 drivers/md/dm-verity-fec.c                    | 21 +++++----
 drivers/md/dm-verity-target.c                 | 13 ++++--
 drivers/md/dm-writecache.c                    |  8 ++--
 drivers/md/persistent-data/dm-block-manager.c |  6 +--
 include/linux/dm-bufio.h                      |  5 ++-
 include/linux/dm-io.h                         |  3 +-
 17 files changed, 102 insertions(+), 83 deletions(-)


base-commit: 7ed2632ec7d72e926b9e8bcc9ad1bb0cd37274bf
-- 
2.34.1
Re: [PATCH v8 0/5] Fix I/O priority lost in device-mapper
Posted by Mikulas Patocka 1 year, 10 months ago
The patchset seems OK to me.

Reviewed-by: Mikulas Patocka <mpatocka@redhat.com>


On Wed, 24 Jan 2024, Hongyu Jin wrote:

> From: Hongyu Jin <hongyu.jin@unisoc.com>
> 
> High-priority tasks get data from dm-verity devices via RT IO priority,
> I/O will lose RT priority when reading FEC and hash values via kworker
> submission IO during verification, and the verification phase may be
> blocked by low-priority IO.
> 
> Dm-crypt has the same problem in the data writing process.
> 
> This is because io_context and blkcg are missing.
> 
> Move bio_set_ioprio() into submit_bio():
> 1. Only call bio_set_ioprio() once to set the priority of original bio,
>    the bio that cloned and splited from original bio will auto inherit
>    the priority of original bio in clone process.
> 
> 2. Make the IO priority of the original bio to be passed to dm,
>    and the dm target inherits the IO priority as needed.
> 
> Changes in v8:
>   - Rebase patch 1 on commit 7ed2632ec7d7
> Changes in v7:
>   - Modify patch 4: change dm-verity-fec.c
> Changes in v6:
>   - Rebase patch and resolve conflict for patch 1, 3, 4
>   - Modify patch 4: fec_read_parity() follow the priority of original
>     bio
>   - Update commit message
> Changes in v5:
>   - Rewrite patch 2, add ioprio parameter in dm_io();
>   - Modify dm_io() in patch 3
> Changes in v4:
>   - Modify commit message by Suggestion
>   - Modify patch for dm-crypt
> Changes in v3:
>   - Split patch for device-mapper
>   - Add patch to fix dm-crypy I/O priority question
>   - Add block patch to review together
>   - Fix some error in v2 patch
> Changes in v2:
>   - Add ioprio field in struct dm_io_region
>   - Initial struct dm_io_region::ioprio to IOPRIO_DEFAULT
>   - Add two interface
> 
> 
> Hongyu Jin (5):
>   block: Fix bio IO priority setting
>   dm: Support I/O priority for dm_io()
>   dm-bufio: Support I/O priority
>   dm verity: Fix I/O priority lost when read FEC and hash
>   dm-crypt: Fix lost ioprio when queuing write bios
> 
>  block/blk-core.c                              | 10 +++++
>  block/blk-mq.c                                | 10 -----
>  drivers/md/dm-bufio.c                         | 43 +++++++++++--------
>  drivers/md/dm-crypt.c                         |  1 +
>  drivers/md/dm-ebs-target.c                    |  8 ++--
>  drivers/md/dm-integrity.c                     | 12 +++---
>  drivers/md/dm-io.c                            | 23 +++++-----
>  drivers/md/dm-kcopyd.c                        |  4 +-
>  drivers/md/dm-log.c                           |  4 +-
>  drivers/md/dm-raid1.c                         |  6 +--
>  drivers/md/dm-snap-persistent.c               |  8 ++--
>  drivers/md/dm-verity-fec.c                    | 21 +++++----
>  drivers/md/dm-verity-target.c                 | 13 ++++--
>  drivers/md/dm-writecache.c                    |  8 ++--
>  drivers/md/persistent-data/dm-block-manager.c |  6 +--
>  include/linux/dm-bufio.h                      |  5 ++-
>  include/linux/dm-io.h                         |  3 +-
>  17 files changed, 102 insertions(+), 83 deletions(-)
> 
> 
> base-commit: 7ed2632ec7d72e926b9e8bcc9ad1bb0cd37274bf
> -- 
> 2.34.1
Re: [PATCH v8 0/5] Fix I/O priority lost in device-mapper
Posted by Mike Snitzer 1 year, 10 months ago
On Wed, Jan 24 2024 at 12:35P -0500,
Hongyu Jin <hongyu.jin.cn@gmail.com> wrote:

> From: Hongyu Jin <hongyu.jin@unisoc.com>
> 
> High-priority tasks get data from dm-verity devices via RT IO priority,
> I/O will lose RT priority when reading FEC and hash values via kworker
> submission IO during verification, and the verification phase may be
> blocked by low-priority IO.
> 
> Dm-crypt has the same problem in the data writing process.
> 
> This is because io_context and blkcg are missing.
> 
> Move bio_set_ioprio() into submit_bio():
> 1. Only call bio_set_ioprio() once to set the priority of original bio,
>    the bio that cloned and splited from original bio will auto inherit
>    the priority of original bio in clone process.
> 
> 2. Make the IO priority of the original bio to be passed to dm,
>    and the dm target inherits the IO priority as needed.
> 
> Changes in v8:
>   - Rebase patch 1 on commit 7ed2632ec7d7
> Changes in v7:
>   - Modify patch 4: change dm-verity-fec.c
> Changes in v6:
>   - Rebase patch and resolve conflict for patch 1, 3, 4
>   - Modify patch 4: fec_read_parity() follow the priority of original
>     bio
>   - Update commit message
> Changes in v5:
>   - Rewrite patch 2, add ioprio parameter in dm_io();
>   - Modify dm_io() in patch 3
> Changes in v4:
>   - Modify commit message by Suggestion
>   - Modify patch for dm-crypt
> Changes in v3:
>   - Split patch for device-mapper
>   - Add patch to fix dm-crypy I/O priority question
>   - Add block patch to review together
>   - Fix some error in v2 patch
> Changes in v2:
>   - Add ioprio field in struct dm_io_region
>   - Initial struct dm_io_region::ioprio to IOPRIO_DEFAULT
>   - Add two interface
> 
> 
> Hongyu Jin (5):
>   block: Fix bio IO priority setting
>   dm: Support I/O priority for dm_io()
>   dm-bufio: Support I/O priority
>   dm verity: Fix I/O priority lost when read FEC and hash
>   dm-crypt: Fix lost ioprio when queuing write bios

Sorry for the delay.. I've been consumed with other work.  I will look
at this patchset for consideration for the 6.9 merge window (we still
have time to make changes given we're now squarely in the 6.9
development window).  So I appreciate getting you feedback sooner
rather than later is both useful and important.

I see Eric provided his Reviewed-by for v7 -- that really helps.  BUT,
for some reason you didn't add his provided Reviewed-by to each commit
when you rebased with v8...

Mikulas, if you beat me to providing closer review: great.  If not,
that's cool.  That DM requires such care (with sprinkling changes
throughout DM core and targets) is unfortunate -- but	could be
unavoidable all things considered.  I will look closer "soon" (if not
this week then next).

Thanks for following up!

Mike
[PATCH v8 1/5] block: Fix bio IO priority setting
Posted by Hongyu Jin 1 year, 10 months ago
From: Hongyu Jin <hongyu.jin@unisoc.com>

Move bio_set_ioprio() into submit_bio():
1. Only call bio_set_ioprio() once to set the priority of original bio,
   the bio that cloned and splited from original bio will auto inherit
   the priority of original bio in clone process.

2. The IO priority can be passed to module that implement
   struct gendisk::fops::submit_bio, help resolve some
   of the IO priority loss issues.

This patch depends on commit 82b74cac2849 ("blk-ioprio: Convert from
rqos policy to direct call")

Fixes: a78418e6a04c ("block: Always initialize bio IO priority on submit")

Co-developed-by: Yibin Ding <yibin.ding@unisoc.com>
Signed-off-by: Yibin Ding <yibin.ding@unisoc.com>
Signed-off-by: Hongyu Jin <hongyu.jin@unisoc.com>
---
 block/blk-core.c | 10 ++++++++++
 block/blk-mq.c   | 10 ----------
 2 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/block/blk-core.c b/block/blk-core.c
index 11342af420d0..de771093b526 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -49,6 +49,7 @@
 #include "blk-pm.h"
 #include "blk-cgroup.h"
 #include "blk-throttle.h"
+#include "blk-ioprio.h"
 
 struct dentry *blk_debugfs_root;
 
@@ -833,6 +834,14 @@ void submit_bio_noacct(struct bio *bio)
 }
 EXPORT_SYMBOL(submit_bio_noacct);
 
+static void bio_set_ioprio(struct bio *bio)
+{
+	/* Nobody set ioprio so far? Initialize it based on task's nice value */
+	if (IOPRIO_PRIO_CLASS(bio->bi_ioprio) == IOPRIO_CLASS_NONE)
+		bio->bi_ioprio = get_current_ioprio();
+	blkcg_set_ioprio(bio);
+}
+
 /**
  * submit_bio - submit a bio to the block device layer for I/O
  * @bio: The &struct bio which describes the I/O
@@ -855,6 +864,7 @@ void submit_bio(struct bio *bio)
 		count_vm_events(PGPGOUT, bio_sectors(bio));
 	}
 
+	bio_set_ioprio(bio);
 	submit_bio_noacct(bio);
 }
 EXPORT_SYMBOL(submit_bio);
diff --git a/block/blk-mq.c b/block/blk-mq.c
index aa87fcfda1ec..2dc01551e27c 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -40,7 +40,6 @@
 #include "blk-stat.h"
 #include "blk-mq-sched.h"
 #include "blk-rq-qos.h"
-#include "blk-ioprio.h"
 
 static DEFINE_PER_CPU(struct llist_head, blk_cpu_done);
 static DEFINE_PER_CPU(call_single_data_t, blk_cpu_csd);
@@ -2944,14 +2943,6 @@ static bool blk_mq_use_cached_rq(struct request *rq, struct blk_plug *plug,
 	return true;
 }
 
-static void bio_set_ioprio(struct bio *bio)
-{
-	/* Nobody set ioprio so far? Initialize it based on task's nice value */
-	if (IOPRIO_PRIO_CLASS(bio->bi_ioprio) == IOPRIO_CLASS_NONE)
-		bio->bi_ioprio = get_current_ioprio();
-	blkcg_set_ioprio(bio);
-}
-
 /**
  * blk_mq_submit_bio - Create and send a request to block device.
  * @bio: Bio pointer.
@@ -2976,7 +2967,6 @@ void blk_mq_submit_bio(struct bio *bio)
 	blk_status_t ret;
 
 	bio = blk_queue_bounce(bio, q);
-	bio_set_ioprio(bio);
 
 	if (plug) {
 		rq = rq_list_peek(&plug->cached_rq);
-- 
2.34.1
[PATCH v8 2/5] dm: Support I/O priority for dm_io()
Posted by Hongyu Jin 1 year, 10 months ago
From: Hongyu Jin <hongyu.jin@unisoc.com>

Some I/O will dispatch from kworker with different io_context settings
than the submitting task, we may need to specify a priority to avoid
losing priority.

Add I/O priority parameter for dm_io().

Co-developed-by: Yibin Ding <yibin.ding@unisoc.com>
Signed-off-by: Yibin Ding <yibin.ding@unisoc.com>
Signed-off-by: Hongyu Jin <hongyu.jin@unisoc.com>
---
 drivers/md/dm-bufio.c           |  6 +++---
 drivers/md/dm-integrity.c       | 10 +++++-----
 drivers/md/dm-io.c              | 23 +++++++++++++----------
 drivers/md/dm-kcopyd.c          |  4 ++--
 drivers/md/dm-log.c             |  4 ++--
 drivers/md/dm-raid1.c           |  6 +++---
 drivers/md/dm-snap-persistent.c |  4 ++--
 drivers/md/dm-writecache.c      |  8 ++++----
 include/linux/dm-io.h           |  3 ++-
 9 files changed, 36 insertions(+), 32 deletions(-)

diff --git a/drivers/md/dm-bufio.c b/drivers/md/dm-bufio.c
index 13c65b7e1ed6..f5541b8f6320 100644
--- a/drivers/md/dm-bufio.c
+++ b/drivers/md/dm-bufio.c
@@ -1315,7 +1315,7 @@ static void use_dmio(struct dm_buffer *b, enum req_op op, sector_t sector,
 		io_req.mem.ptr.vma = (char *)b->data + offset;
 	}
 
-	r = dm_io(&io_req, 1, &region, NULL);
+	r = dm_io(&io_req, 1, &region, NULL, IOPRIO_DEFAULT);
 	if (unlikely(r))
 		b->end_io(b, errno_to_blk_status(r));
 }
@@ -2167,7 +2167,7 @@ int dm_bufio_issue_flush(struct dm_bufio_client *c)
 	if (WARN_ON_ONCE(dm_bufio_in_request()))
 		return -EINVAL;
 
-	return dm_io(&io_req, 1, &io_reg, NULL);
+	return dm_io(&io_req, 1, &io_reg, NULL, IOPRIO_DEFAULT);
 }
 EXPORT_SYMBOL_GPL(dm_bufio_issue_flush);
 
@@ -2191,7 +2191,7 @@ int dm_bufio_issue_discard(struct dm_bufio_client *c, sector_t block, sector_t c
 	if (WARN_ON_ONCE(dm_bufio_in_request()))
 		return -EINVAL; /* discards are optional */
 
-	return dm_io(&io_req, 1, &io_reg, NULL);
+	return dm_io(&io_req, 1, &io_reg, NULL, IOPRIO_DEFAULT);
 }
 EXPORT_SYMBOL_GPL(dm_bufio_issue_discard);
 
diff --git a/drivers/md/dm-integrity.c b/drivers/md/dm-integrity.c
index c5f03aab4552..ed45411eb68d 100644
--- a/drivers/md/dm-integrity.c
+++ b/drivers/md/dm-integrity.c
@@ -553,7 +553,7 @@ static int sync_rw_sb(struct dm_integrity_c *ic, blk_opf_t opf)
 		}
 	}
 
-	r = dm_io(&io_req, 1, &io_loc, NULL);
+	r = dm_io(&io_req, 1, &io_loc, NULL, IOPRIO_DEFAULT);
 	if (unlikely(r))
 		return r;
 
@@ -1071,7 +1071,7 @@ static void rw_journal_sectors(struct dm_integrity_c *ic, blk_opf_t opf,
 	io_loc.sector = ic->start + SB_SECTORS + sector;
 	io_loc.count = n_sectors;
 
-	r = dm_io(&io_req, 1, &io_loc, NULL);
+	r = dm_io(&io_req, 1, &io_loc, NULL, IOPRIO_DEFAULT);
 	if (unlikely(r)) {
 		dm_integrity_io_error(ic, (opf & REQ_OP_MASK) == REQ_OP_READ ?
 				      "reading journal" : "writing journal", r);
@@ -1188,7 +1188,7 @@ static void copy_from_journal(struct dm_integrity_c *ic, unsigned int section, u
 	io_loc.sector = target;
 	io_loc.count = n_sectors;
 
-	r = dm_io(&io_req, 1, &io_loc, NULL);
+	r = dm_io(&io_req, 1, &io_loc, NULL, IOPRIO_DEFAULT);
 	if (unlikely(r)) {
 		WARN_ONCE(1, "asynchronous dm_io failed: %d", r);
 		fn(-1UL, data);
@@ -1517,7 +1517,7 @@ static void dm_integrity_flush_buffers(struct dm_integrity_c *ic, bool flush_dat
 		fr.io_reg.count = 0,
 		fr.ic = ic;
 		init_completion(&fr.comp);
-		r = dm_io(&fr.io_req, 1, &fr.io_reg, NULL);
+		r = dm_io(&fr.io_req, 1, &fr.io_reg, NULL, IOPRIO_DEFAULT);
 		BUG_ON(r);
 	}
 
@@ -2740,7 +2740,7 @@ static void integrity_recalc(struct work_struct *w)
 	io_loc.sector = get_data_sector(ic, area, offset);
 	io_loc.count = n_sectors;
 
-	r = dm_io(&io_req, 1, &io_loc, NULL);
+	r = dm_io(&io_req, 1, &io_loc, NULL, IOPRIO_DEFAULT);
 	if (unlikely(r)) {
 		dm_integrity_io_error(ic, "reading data", r);
 		goto err;
diff --git a/drivers/md/dm-io.c b/drivers/md/dm-io.c
index f053ce245814..7409490259d1 100644
--- a/drivers/md/dm-io.c
+++ b/drivers/md/dm-io.c
@@ -305,7 +305,7 @@ static void km_dp_init(struct dpages *dp, void *data)
  */
 static void do_region(const blk_opf_t opf, unsigned int region,
 		      struct dm_io_region *where, struct dpages *dp,
-		      struct io *io)
+		      struct io *io, unsigned short ioprio)
 {
 	struct bio *bio;
 	struct page *page;
@@ -354,6 +354,7 @@ static void do_region(const blk_opf_t opf, unsigned int region,
 				       &io->client->bios);
 		bio->bi_iter.bi_sector = where->sector + (where->count - remaining);
 		bio->bi_end_io = endio;
+		bio->bi_ioprio = ioprio;
 		store_io_and_region_in_bio(bio, io, region);
 
 		if (op == REQ_OP_DISCARD || op == REQ_OP_WRITE_ZEROES) {
@@ -383,7 +384,7 @@ static void do_region(const blk_opf_t opf, unsigned int region,
 
 static void dispatch_io(blk_opf_t opf, unsigned int num_regions,
 			struct dm_io_region *where, struct dpages *dp,
-			struct io *io, int sync)
+			struct io *io, int sync, unsigned short ioprio)
 {
 	int i;
 	struct dpages old_pages = *dp;
@@ -400,7 +401,7 @@ static void dispatch_io(blk_opf_t opf, unsigned int num_regions,
 	for (i = 0; i < num_regions; i++) {
 		*dp = old_pages;
 		if (where[i].count || (opf & REQ_PREFLUSH))
-			do_region(opf, i, where + i, dp, io);
+			do_region(opf, i, where + i, dp, io, ioprio);
 	}
 
 	/*
@@ -425,7 +426,7 @@ static void sync_io_complete(unsigned long error, void *context)
 
 static int sync_io(struct dm_io_client *client, unsigned int num_regions,
 		   struct dm_io_region *where, blk_opf_t opf, struct dpages *dp,
-		   unsigned long *error_bits)
+		   unsigned long *error_bits, unsigned short ioprio)
 {
 	struct io *io;
 	struct sync_io sio;
@@ -447,7 +448,7 @@ static int sync_io(struct dm_io_client *client, unsigned int num_regions,
 	io->vma_invalidate_address = dp->vma_invalidate_address;
 	io->vma_invalidate_size = dp->vma_invalidate_size;
 
-	dispatch_io(opf, num_regions, where, dp, io, 1);
+	dispatch_io(opf, num_regions, where, dp, io, 1, ioprio);
 
 	wait_for_completion_io(&sio.wait);
 
@@ -459,7 +460,8 @@ static int sync_io(struct dm_io_client *client, unsigned int num_regions,
 
 static int async_io(struct dm_io_client *client, unsigned int num_regions,
 		    struct dm_io_region *where, blk_opf_t opf,
-		    struct dpages *dp, io_notify_fn fn, void *context)
+		    struct dpages *dp, io_notify_fn fn, void *context,
+		    unsigned short ioprio)
 {
 	struct io *io;
 
@@ -479,7 +481,7 @@ static int async_io(struct dm_io_client *client, unsigned int num_regions,
 	io->vma_invalidate_address = dp->vma_invalidate_address;
 	io->vma_invalidate_size = dp->vma_invalidate_size;
 
-	dispatch_io(opf, num_regions, where, dp, io, 0);
+	dispatch_io(opf, num_regions, where, dp, io, 0, ioprio);
 	return 0;
 }
 
@@ -521,7 +523,8 @@ static int dp_init(struct dm_io_request *io_req, struct dpages *dp,
 }
 
 int dm_io(struct dm_io_request *io_req, unsigned int num_regions,
-	  struct dm_io_region *where, unsigned long *sync_error_bits)
+	  struct dm_io_region *where, unsigned long *sync_error_bits,
+	  unsigned short ioprio)
 {
 	int r;
 	struct dpages dp;
@@ -532,11 +535,11 @@ int dm_io(struct dm_io_request *io_req, unsigned int num_regions,
 
 	if (!io_req->notify.fn)
 		return sync_io(io_req->client, num_regions, where,
-			       io_req->bi_opf, &dp, sync_error_bits);
+			       io_req->bi_opf, &dp, sync_error_bits, ioprio);
 
 	return async_io(io_req->client, num_regions, where,
 			io_req->bi_opf, &dp, io_req->notify.fn,
-			io_req->notify.context);
+			io_req->notify.context, ioprio);
 }
 EXPORT_SYMBOL(dm_io);
 
diff --git a/drivers/md/dm-kcopyd.c b/drivers/md/dm-kcopyd.c
index 36bcfdccae04..6ea75436a433 100644
--- a/drivers/md/dm-kcopyd.c
+++ b/drivers/md/dm-kcopyd.c
@@ -578,9 +578,9 @@ static int run_io_job(struct kcopyd_job *job)
 	io_job_start(job->kc->throttle);
 
 	if (job->op == REQ_OP_READ)
-		r = dm_io(&io_req, 1, &job->source, NULL);
+		r = dm_io(&io_req, 1, &job->source, NULL, IOPRIO_DEFAULT);
 	else
-		r = dm_io(&io_req, job->num_dests, job->dests, NULL);
+		r = dm_io(&io_req, job->num_dests, job->dests, NULL, IOPRIO_DEFAULT);
 
 	return r;
 }
diff --git a/drivers/md/dm-log.c b/drivers/md/dm-log.c
index f9f84236dfcd..f7f9c2100937 100644
--- a/drivers/md/dm-log.c
+++ b/drivers/md/dm-log.c
@@ -300,7 +300,7 @@ static int rw_header(struct log_c *lc, enum req_op op)
 {
 	lc->io_req.bi_opf = op;
 
-	return dm_io(&lc->io_req, 1, &lc->header_location, NULL);
+	return dm_io(&lc->io_req, 1, &lc->header_location, NULL, IOPRIO_DEFAULT);
 }
 
 static int flush_header(struct log_c *lc)
@@ -313,7 +313,7 @@ static int flush_header(struct log_c *lc)
 
 	lc->io_req.bi_opf = REQ_OP_WRITE | REQ_PREFLUSH;
 
-	return dm_io(&lc->io_req, 1, &null_location, NULL);
+	return dm_io(&lc->io_req, 1, &null_location, NULL, IOPRIO_DEFAULT);
 }
 
 static int read_header(struct log_c *log)
diff --git a/drivers/md/dm-raid1.c b/drivers/md/dm-raid1.c
index ddcb2bc4a617..9511dae5b556 100644
--- a/drivers/md/dm-raid1.c
+++ b/drivers/md/dm-raid1.c
@@ -278,7 +278,7 @@ static int mirror_flush(struct dm_target *ti)
 	}
 
 	error_bits = -1;
-	dm_io(&io_req, ms->nr_mirrors, io, &error_bits);
+	dm_io(&io_req, ms->nr_mirrors, io, &error_bits, IOPRIO_DEFAULT);
 	if (unlikely(error_bits != 0)) {
 		for (i = 0; i < ms->nr_mirrors; i++)
 			if (test_bit(i, &error_bits))
@@ -554,7 +554,7 @@ static void read_async_bio(struct mirror *m, struct bio *bio)
 
 	map_region(&io, m, bio);
 	bio_set_m(bio, m);
-	BUG_ON(dm_io(&io_req, 1, &io, NULL));
+	BUG_ON(dm_io(&io_req, 1, &io, NULL, IOPRIO_DEFAULT));
 }
 
 static inline int region_in_sync(struct mirror_set *ms, region_t region,
@@ -681,7 +681,7 @@ static void do_write(struct mirror_set *ms, struct bio *bio)
 	 */
 	bio_set_m(bio, get_default_mirror(ms));
 
-	BUG_ON(dm_io(&io_req, ms->nr_mirrors, io, NULL));
+	BUG_ON(dm_io(&io_req, ms->nr_mirrors, io, NULL, IOPRIO_DEFAULT));
 }
 
 static void do_writes(struct mirror_set *ms, struct bio_list *writes)
diff --git a/drivers/md/dm-snap-persistent.c b/drivers/md/dm-snap-persistent.c
index 15649921f2a9..568d10842b1f 100644
--- a/drivers/md/dm-snap-persistent.c
+++ b/drivers/md/dm-snap-persistent.c
@@ -223,7 +223,7 @@ static void do_metadata(struct work_struct *work)
 {
 	struct mdata_req *req = container_of(work, struct mdata_req, work);
 
-	req->result = dm_io(req->io_req, 1, req->where, NULL);
+	req->result = dm_io(req->io_req, 1, req->where, NULL, IOPRIO_DEFAULT);
 }
 
 /*
@@ -247,7 +247,7 @@ static int chunk_io(struct pstore *ps, void *area, chunk_t chunk, blk_opf_t opf,
 	struct mdata_req req;
 
 	if (!metadata)
-		return dm_io(&io_req, 1, &where, NULL);
+		return dm_io(&io_req, 1, &where, NULL, IOPRIO_DEFAULT);
 
 	req.where = &where;
 	req.io_req = &io_req;
diff --git a/drivers/md/dm-writecache.c b/drivers/md/dm-writecache.c
index 074cb785eafc..6a4279bfb1e7 100644
--- a/drivers/md/dm-writecache.c
+++ b/drivers/md/dm-writecache.c
@@ -531,7 +531,7 @@ static void ssd_commit_flushed(struct dm_writecache *wc, bool wait_for_ios)
 		req.notify.context = &endio;
 
 		/* writing via async dm-io (implied by notify.fn above) won't return an error */
-		(void) dm_io(&req, 1, &region, NULL);
+		(void) dm_io(&req, 1, &region, NULL, IOPRIO_DEFAULT);
 		i = j;
 	}
 
@@ -568,7 +568,7 @@ static void ssd_commit_superblock(struct dm_writecache *wc)
 	req.notify.fn = NULL;
 	req.notify.context = NULL;
 
-	r = dm_io(&req, 1, &region, NULL);
+	r = dm_io(&req, 1, &region, NULL, IOPRIO_DEFAULT);
 	if (unlikely(r))
 		writecache_error(wc, r, "error writing superblock");
 }
@@ -596,7 +596,7 @@ static void writecache_disk_flush(struct dm_writecache *wc, struct dm_dev *dev)
 	req.client = wc->dm_io;
 	req.notify.fn = NULL;
 
-	r = dm_io(&req, 1, &region, NULL);
+	r = dm_io(&req, 1, &region, NULL, IOPRIO_DEFAULT);
 	if (unlikely(r))
 		writecache_error(wc, r, "error flushing metadata: %d", r);
 }
@@ -990,7 +990,7 @@ static int writecache_read_metadata(struct dm_writecache *wc, sector_t n_sectors
 	req.client = wc->dm_io;
 	req.notify.fn = NULL;
 
-	return dm_io(&req, 1, &region, NULL);
+	return dm_io(&req, 1, &region, NULL, IOPRIO_DEFAULT);
 }
 
 static void writecache_resume(struct dm_target *ti)
diff --git a/include/linux/dm-io.h b/include/linux/dm-io.h
index 7595142f3fc5..7b2968612b7e 100644
--- a/include/linux/dm-io.h
+++ b/include/linux/dm-io.h
@@ -80,7 +80,8 @@ void dm_io_client_destroy(struct dm_io_client *client);
  * error occurred doing io to the corresponding region.
  */
 int dm_io(struct dm_io_request *io_req, unsigned int num_regions,
-	  struct dm_io_region *region, unsigned int long *sync_error_bits);
+	  struct dm_io_region *region, unsigned int long *sync_error_bits,
+	  unsigned short ioprio);
 
 #endif	/* __KERNEL__ */
 #endif	/* _LINUX_DM_IO_H */
-- 
2.34.1
[PATCH v8 3/5] dm-bufio: Support I/O priority
Posted by Hongyu Jin 1 year, 10 months ago
From: Hongyu Jin <hongyu.jin@unisoc.com>

Some I/O will dispatch from kworker with different io_context settings
than the submitting task, we may need to specify a priority to avoid
losing priority.

Add I/O priority parameter for dm_bufio_read() and
dm_bufio_prefetch().

Co-developed-by: Yibin Ding <yibin.ding@unisoc.com>
Signed-off-by: Yibin Ding <yibin.ding@unisoc.com>
Signed-off-by: Hongyu Jin <hongyu.jin@unisoc.com>
---
 drivers/md/dm-bufio.c                         | 39 +++++++++++--------
 drivers/md/dm-ebs-target.c                    |  8 ++--
 drivers/md/dm-integrity.c                     |  2 +-
 drivers/md/dm-snap-persistent.c               |  4 +-
 drivers/md/dm-verity-fec.c                    |  4 +-
 drivers/md/dm-verity-target.c                 |  5 ++-
 drivers/md/persistent-data/dm-block-manager.c |  6 +--
 include/linux/dm-bufio.h                      |  5 ++-
 8 files changed, 40 insertions(+), 33 deletions(-)

diff --git a/drivers/md/dm-bufio.c b/drivers/md/dm-bufio.c
index f5541b8f6320..31b7351398a8 100644
--- a/drivers/md/dm-bufio.c
+++ b/drivers/md/dm-bufio.c
@@ -1292,7 +1292,8 @@ static void dmio_complete(unsigned long error, void *context)
 }
 
 static void use_dmio(struct dm_buffer *b, enum req_op op, sector_t sector,
-		     unsigned int n_sectors, unsigned int offset)
+		     unsigned int n_sectors, unsigned int offset,
+		     unsigned short ioprio)
 {
 	int r;
 	struct dm_io_request io_req = {
@@ -1315,7 +1316,7 @@ static void use_dmio(struct dm_buffer *b, enum req_op op, sector_t sector,
 		io_req.mem.ptr.vma = (char *)b->data + offset;
 	}
 
-	r = dm_io(&io_req, 1, &region, NULL, IOPRIO_DEFAULT);
+	r = dm_io(&io_req, 1, &region, NULL, ioprio);
 	if (unlikely(r))
 		b->end_io(b, errno_to_blk_status(r));
 }
@@ -1331,7 +1332,8 @@ static void bio_complete(struct bio *bio)
 }
 
 static void use_bio(struct dm_buffer *b, enum req_op op, sector_t sector,
-		    unsigned int n_sectors, unsigned int offset)
+		    unsigned int n_sectors, unsigned int offset,
+		    unsigned short ioprio)
 {
 	struct bio *bio;
 	char *ptr;
@@ -1339,13 +1341,14 @@ static void use_bio(struct dm_buffer *b, enum req_op op, sector_t sector,
 
 	bio = bio_kmalloc(1, GFP_NOWAIT | __GFP_NORETRY | __GFP_NOWARN);
 	if (!bio) {
-		use_dmio(b, op, sector, n_sectors, offset);
+		use_dmio(b, op, sector, n_sectors, offset, ioprio);
 		return;
 	}
 	bio_init(bio, b->c->bdev, bio->bi_inline_vecs, 1, op);
 	bio->bi_iter.bi_sector = sector;
 	bio->bi_end_io = bio_complete;
 	bio->bi_private = b;
+	bio->bi_ioprio = ioprio;
 
 	ptr = (char *)b->data + offset;
 	len = n_sectors << SECTOR_SHIFT;
@@ -1368,7 +1371,7 @@ static inline sector_t block_to_sector(struct dm_bufio_client *c, sector_t block
 	return sector;
 }
 
-static void submit_io(struct dm_buffer *b, enum req_op op,
+static void submit_io(struct dm_buffer *b, enum req_op op, unsigned short ioprio,
 		      void (*end_io)(struct dm_buffer *, blk_status_t))
 {
 	unsigned int n_sectors;
@@ -1398,9 +1401,9 @@ static void submit_io(struct dm_buffer *b, enum req_op op,
 	}
 
 	if (b->data_mode != DATA_MODE_VMALLOC)
-		use_bio(b, op, sector, n_sectors, offset);
+		use_bio(b, op, sector, n_sectors, offset, ioprio);
 	else
-		use_dmio(b, op, sector, n_sectors, offset);
+		use_dmio(b, op, sector, n_sectors, offset, ioprio);
 }
 
 /*
@@ -1456,7 +1459,7 @@ static void __write_dirty_buffer(struct dm_buffer *b,
 	b->write_end = b->dirty_end;
 
 	if (!write_list)
-		submit_io(b, REQ_OP_WRITE, write_endio);
+		submit_io(b, REQ_OP_WRITE, IOPRIO_DEFAULT, write_endio);
 	else
 		list_add_tail(&b->write_list, write_list);
 }
@@ -1470,7 +1473,7 @@ static void __flush_write_list(struct list_head *write_list)
 		struct dm_buffer *b =
 			list_entry(write_list->next, struct dm_buffer, write_list);
 		list_del(&b->write_list);
-		submit_io(b, REQ_OP_WRITE, write_endio);
+		submit_io(b, REQ_OP_WRITE, IOPRIO_DEFAULT, write_endio);
 		cond_resched();
 	}
 	blk_finish_plug(&plug);
@@ -1852,7 +1855,8 @@ static void read_endio(struct dm_buffer *b, blk_status_t status)
  * and uses dm_bufio_mark_buffer_dirty to write new data back).
  */
 static void *new_read(struct dm_bufio_client *c, sector_t block,
-		      enum new_flag nf, struct dm_buffer **bp)
+		      enum new_flag nf, struct dm_buffer **bp,
+		      unsigned short ioprio)
 {
 	int need_submit = 0;
 	struct dm_buffer *b;
@@ -1905,7 +1909,7 @@ static void *new_read(struct dm_bufio_client *c, sector_t block,
 		return NULL;
 
 	if (need_submit)
-		submit_io(b, REQ_OP_READ, read_endio);
+		submit_io(b, REQ_OP_READ, ioprio, read_endio);
 
 	if (nf != NF_GET)	/* we already tested this condition above */
 		wait_on_bit_io(&b->state, B_READING, TASK_UNINTERRUPTIBLE);
@@ -1926,17 +1930,17 @@ static void *new_read(struct dm_bufio_client *c, sector_t block,
 void *dm_bufio_get(struct dm_bufio_client *c, sector_t block,
 		   struct dm_buffer **bp)
 {
-	return new_read(c, block, NF_GET, bp);
+	return new_read(c, block, NF_GET, bp, IOPRIO_DEFAULT);
 }
 EXPORT_SYMBOL_GPL(dm_bufio_get);
 
 void *dm_bufio_read(struct dm_bufio_client *c, sector_t block,
-		    struct dm_buffer **bp)
+		    struct dm_buffer **bp, unsigned short ioprio)
 {
 	if (WARN_ON_ONCE(dm_bufio_in_request()))
 		return ERR_PTR(-EINVAL);
 
-	return new_read(c, block, NF_READ, bp);
+	return new_read(c, block, NF_READ, bp, ioprio);
 }
 EXPORT_SYMBOL_GPL(dm_bufio_read);
 
@@ -1946,12 +1950,13 @@ void *dm_bufio_new(struct dm_bufio_client *c, sector_t block,
 	if (WARN_ON_ONCE(dm_bufio_in_request()))
 		return ERR_PTR(-EINVAL);
 
-	return new_read(c, block, NF_FRESH, bp);
+	return new_read(c, block, NF_FRESH, bp, IOPRIO_DEFAULT);
 }
 EXPORT_SYMBOL_GPL(dm_bufio_new);
 
 void dm_bufio_prefetch(struct dm_bufio_client *c,
-		       sector_t block, unsigned int n_blocks)
+		       sector_t block, unsigned int n_blocks,
+		       unsigned short ioprio)
 {
 	struct blk_plug plug;
 
@@ -1987,7 +1992,7 @@ void dm_bufio_prefetch(struct dm_bufio_client *c,
 			dm_bufio_unlock(c);
 
 			if (need_submit)
-				submit_io(b, REQ_OP_READ, read_endio);
+				submit_io(b, REQ_OP_READ, ioprio, read_endio);
 			dm_bufio_release(b);
 
 			cond_resched();
diff --git a/drivers/md/dm-ebs-target.c b/drivers/md/dm-ebs-target.c
index 435b45201f4d..8198c8a7b416 100644
--- a/drivers/md/dm-ebs-target.c
+++ b/drivers/md/dm-ebs-target.c
@@ -84,7 +84,7 @@ static int __ebs_rw_bvec(struct ebs_c *ec, enum req_op op, struct bio_vec *bv,
 
 		/* Avoid reading for writes in case bio vector's page overwrites block completely. */
 		if (op == REQ_OP_READ || buf_off || bv_len < dm_bufio_get_block_size(ec->bufio))
-			ba = dm_bufio_read(ec->bufio, block, &b);
+			ba = dm_bufio_read(ec->bufio, block, &b, IOPRIO_DEFAULT);
 		else
 			ba = dm_bufio_new(ec->bufio, block, &b);
 
@@ -194,13 +194,13 @@ static void __ebs_process_bios(struct work_struct *ws)
 	bio_list_for_each(bio, &bios) {
 		block1 = __sector_to_block(ec, bio->bi_iter.bi_sector);
 		if (bio_op(bio) == REQ_OP_READ)
-			dm_bufio_prefetch(ec->bufio, block1, __nr_blocks(ec, bio));
+			dm_bufio_prefetch(ec->bufio, block1, __nr_blocks(ec, bio), IOPRIO_DEFAULT);
 		else if (bio_op(bio) == REQ_OP_WRITE && !(bio->bi_opf & REQ_PREFLUSH)) {
 			block2 = __sector_to_block(ec, bio_end_sector(bio));
 			if (__block_mod(bio->bi_iter.bi_sector, ec->u_bs))
-				dm_bufio_prefetch(ec->bufio, block1, 1);
+				dm_bufio_prefetch(ec->bufio, block1, 1, IOPRIO_DEFAULT);
 			if (__block_mod(bio_end_sector(bio), ec->u_bs) && block2 != block1)
-				dm_bufio_prefetch(ec->bufio, block2, 1);
+				dm_bufio_prefetch(ec->bufio, block2, 1, IOPRIO_DEFAULT);
 		}
 	}
 
diff --git a/drivers/md/dm-integrity.c b/drivers/md/dm-integrity.c
index ed45411eb68d..42abc5c316f1 100644
--- a/drivers/md/dm-integrity.c
+++ b/drivers/md/dm-integrity.c
@@ -1418,7 +1418,7 @@ static int dm_integrity_rw_tag(struct dm_integrity_c *ic, unsigned char *tag, se
 		if (unlikely(r))
 			return r;
 
-		data = dm_bufio_read(ic->bufio, *metadata_block, &b);
+		data = dm_bufio_read(ic->bufio, *metadata_block, &b, IOPRIO_DEFAULT);
 		if (IS_ERR(data))
 			return PTR_ERR(data);
 
diff --git a/drivers/md/dm-snap-persistent.c b/drivers/md/dm-snap-persistent.c
index 568d10842b1f..a2072b95e28c 100644
--- a/drivers/md/dm-snap-persistent.c
+++ b/drivers/md/dm-snap-persistent.c
@@ -524,7 +524,7 @@ static int read_exceptions(struct pstore *ps,
 
 				if (unlikely(pf_chunk >= dm_bufio_get_device_size(client)))
 					break;
-				dm_bufio_prefetch(client, pf_chunk, 1);
+				dm_bufio_prefetch(client, pf_chunk, 1, IOPRIO_DEFAULT);
 				prefetch_area++;
 				if (unlikely(!prefetch_area))
 					break;
@@ -533,7 +533,7 @@ static int read_exceptions(struct pstore *ps,
 
 		chunk = area_location(ps, ps->current_area);
 
-		area = dm_bufio_read(client, chunk, &bp);
+		area = dm_bufio_read(client, chunk, &bp, IOPRIO_DEFAULT);
 		if (IS_ERR(area)) {
 			r = PTR_ERR(area);
 			goto ret_destroy_bufio;
diff --git a/drivers/md/dm-verity-fec.c b/drivers/md/dm-verity-fec.c
index b475200d8586..49db19e537f9 100644
--- a/drivers/md/dm-verity-fec.c
+++ b/drivers/md/dm-verity-fec.c
@@ -69,7 +69,7 @@ static u8 *fec_read_parity(struct dm_verity *v, u64 rsb, int index,
 	block = div64_u64_rem(position, v->fec->io_size, &rem);
 	*offset = (unsigned int)rem;
 
-	res = dm_bufio_read(v->fec->bufio, block, buf);
+	res = dm_bufio_read(v->fec->bufio, block, buf, IOPRIO_DEFAULT);
 	if (IS_ERR(res)) {
 		DMERR("%s: FEC %llu: parity read failed (block %llu): %ld",
 		      v->data_dev->name, (unsigned long long)rsb,
@@ -248,7 +248,7 @@ static int fec_read_bufs(struct dm_verity *v, struct dm_verity_io *io,
 			bufio = v->bufio;
 		}
 
-		bbuf = dm_bufio_read(bufio, block, &buf);
+		bbuf = dm_bufio_read(bufio, block, &buf, IOPRIO_DEFAULT);
 		if (IS_ERR(bbuf)) {
 			DMWARN_LIMIT("%s: FEC %llu: read failed (%llu): %ld",
 				     v->data_dev->name,
diff --git a/drivers/md/dm-verity-target.c b/drivers/md/dm-verity-target.c
index 14e58ae70521..4758bfe2c156 100644
--- a/drivers/md/dm-verity-target.c
+++ b/drivers/md/dm-verity-target.c
@@ -308,7 +308,7 @@ static int verity_verify_level(struct dm_verity *v, struct dm_verity_io *io,
 			return -EAGAIN;
 		}
 	} else
-		data = dm_bufio_read(v->bufio, hash_block, &buf);
+		data = dm_bufio_read(v->bufio, hash_block, &buf, IOPRIO_DEFAULT);
 
 	if (IS_ERR(data))
 		return PTR_ERR(data);
@@ -719,7 +719,8 @@ static void verity_prefetch_io(struct work_struct *work)
 		}
 no_prefetch_cluster:
 		dm_bufio_prefetch(v->bufio, hash_block_start,
-				  hash_block_end - hash_block_start + 1);
+				  hash_block_end - hash_block_start + 1,
+				  IOPRIO_DEFAULT);
 	}
 
 	kfree(pw);
diff --git a/drivers/md/persistent-data/dm-block-manager.c b/drivers/md/persistent-data/dm-block-manager.c
index 0e010e1204aa..86a4f73d2f3d 100644
--- a/drivers/md/persistent-data/dm-block-manager.c
+++ b/drivers/md/persistent-data/dm-block-manager.c
@@ -474,7 +474,7 @@ int dm_bm_read_lock(struct dm_block_manager *bm, dm_block_t b,
 	void *p;
 	int r;
 
-	p = dm_bufio_read(bm->bufio, b, (struct dm_buffer **) result);
+	p = dm_bufio_read(bm->bufio, b, (struct dm_buffer **) result, IOPRIO_DEFAULT);
 	if (IS_ERR(p))
 		return PTR_ERR(p);
 
@@ -510,7 +510,7 @@ int dm_bm_write_lock(struct dm_block_manager *bm,
 	if (dm_bm_is_read_only(bm))
 		return -EPERM;
 
-	p = dm_bufio_read(bm->bufio, b, (struct dm_buffer **) result);
+	p = dm_bufio_read(bm->bufio, b, (struct dm_buffer **) result, IOPRIO_DEFAULT);
 	if (IS_ERR(p))
 		return PTR_ERR(p);
 
@@ -624,7 +624,7 @@ EXPORT_SYMBOL_GPL(dm_bm_flush);
 
 void dm_bm_prefetch(struct dm_block_manager *bm, dm_block_t b)
 {
-	dm_bufio_prefetch(bm->bufio, b, 1);
+	dm_bufio_prefetch(bm->bufio, b, 1, IOPRIO_DEFAULT);
 }
 
 bool dm_bm_is_read_only(struct dm_block_manager *bm)
diff --git a/include/linux/dm-bufio.h b/include/linux/dm-bufio.h
index 75e7d8cbb532..256a246c7b97 100644
--- a/include/linux/dm-bufio.h
+++ b/include/linux/dm-bufio.h
@@ -62,7 +62,7 @@ void dm_bufio_set_sector_offset(struct dm_bufio_client *c, sector_t start);
  * it dirty.
  */
 void *dm_bufio_read(struct dm_bufio_client *c, sector_t block,
-		    struct dm_buffer **bp);
+		    struct dm_buffer **bp, unsigned short ioprio);
 
 /*
  * Like dm_bufio_read, but return buffer from cache, don't read
@@ -84,7 +84,8 @@ void *dm_bufio_new(struct dm_bufio_client *c, sector_t block,
  * I/O to finish.
  */
 void dm_bufio_prefetch(struct dm_bufio_client *c,
-		       sector_t block, unsigned int n_blocks);
+		       sector_t block, unsigned int n_blocks,
+		       unsigned short ioprio);
 
 /*
  * Release a reference obtained with dm_bufio_{read,get,new}. The data
-- 
2.34.1
[PATCH v8 4/5] dm verity: Fix I/O priority lost when read FEC and hash
Posted by Hongyu Jin 1 year, 10 months ago
From: Hongyu Jin <hongyu.jin@unisoc.com>

After obtaining the data, verification or error correction process may
trigger a new I/O that loses the priority of the original I/O, that is,
the verification of the higher priority IO may be blocked by the lower
priority IO.

Make the I/O of verification and error correction follow the
priority of original I/O.

Co-developed-by: Yibin Ding <yibin.ding@unisoc.com>
Signed-off-by: Yibin Ding <yibin.ding@unisoc.com>
Signed-off-by: Hongyu Jin <hongyu.jin@unisoc.com>
---
 drivers/md/dm-verity-fec.c    | 21 ++++++++++++---------
 drivers/md/dm-verity-target.c | 12 ++++++++----
 2 files changed, 20 insertions(+), 13 deletions(-)

diff --git a/drivers/md/dm-verity-fec.c b/drivers/md/dm-verity-fec.c
index 49db19e537f9..066521de08da 100644
--- a/drivers/md/dm-verity-fec.c
+++ b/drivers/md/dm-verity-fec.c
@@ -60,7 +60,8 @@ static int fec_decode_rs8(struct dm_verity *v, struct dm_verity_fec_io *fio,
  * to the data block. Caller is responsible for releasing buf.
  */
 static u8 *fec_read_parity(struct dm_verity *v, u64 rsb, int index,
-			   unsigned int *offset, struct dm_buffer **buf)
+			   unsigned int *offset, struct dm_buffer **buf,
+			   unsigned short ioprio)
 {
 	u64 position, block, rem;
 	u8 *res;
@@ -69,7 +70,7 @@ static u8 *fec_read_parity(struct dm_verity *v, u64 rsb, int index,
 	block = div64_u64_rem(position, v->fec->io_size, &rem);
 	*offset = (unsigned int)rem;
 
-	res = dm_bufio_read(v->fec->bufio, block, buf, IOPRIO_DEFAULT);
+	res = dm_bufio_read(v->fec->bufio, block, buf, ioprio);
 	if (IS_ERR(res)) {
 		DMERR("%s: FEC %llu: parity read failed (block %llu): %ld",
 		      v->data_dev->name, (unsigned long long)rsb,
@@ -121,16 +122,17 @@ static inline unsigned int fec_buffer_rs_index(unsigned int i, unsigned int j)
  * Decode all RS blocks from buffers and copy corrected bytes into fio->output
  * starting from block_offset.
  */
-static int fec_decode_bufs(struct dm_verity *v, struct dm_verity_fec_io *fio,
-			   u64 rsb, int byte_index, unsigned int block_offset,
-			   int neras)
+static int fec_decode_bufs(struct dm_verity *v, struct dm_verity_io *io,
+			   struct dm_verity_fec_io *fio, u64 rsb, int byte_index,
+			   unsigned int block_offset, int neras)
 {
 	int r, corrected = 0, res;
 	struct dm_buffer *buf;
 	unsigned int n, i, offset;
 	u8 *par, *block;
+	struct bio *bio = dm_bio_from_per_bio_data(io, v->ti->per_io_data_size);
 
-	par = fec_read_parity(v, rsb, block_offset, &offset, &buf);
+	par = fec_read_parity(v, rsb, block_offset, &offset, &buf, bio_prio(bio));
 	if (IS_ERR(par))
 		return PTR_ERR(par);
 
@@ -158,7 +160,7 @@ static int fec_decode_bufs(struct dm_verity *v, struct dm_verity_fec_io *fio,
 		if (offset >= v->fec->io_size) {
 			dm_bufio_release(buf);
 
-			par = fec_read_parity(v, rsb, block_offset, &offset, &buf);
+			par = fec_read_parity(v, rsb, block_offset, &offset, &buf, bio_prio(bio));
 			if (IS_ERR(par))
 				return PTR_ERR(par);
 		}
@@ -210,6 +212,7 @@ static int fec_read_bufs(struct dm_verity *v, struct dm_verity_io *io,
 	u8 *bbuf, *rs_block;
 	u8 want_digest[HASH_MAX_DIGESTSIZE];
 	unsigned int n, k;
+	struct bio *bio = dm_bio_from_per_bio_data(io, v->ti->per_io_data_size);
 
 	if (neras)
 		*neras = 0;
@@ -248,7 +251,7 @@ static int fec_read_bufs(struct dm_verity *v, struct dm_verity_io *io,
 			bufio = v->bufio;
 		}
 
-		bbuf = dm_bufio_read(bufio, block, &buf, IOPRIO_DEFAULT);
+		bbuf = dm_bufio_read(bufio, block, &buf, bio_prio(bio));
 		if (IS_ERR(bbuf)) {
 			DMWARN_LIMIT("%s: FEC %llu: read failed (%llu): %ld",
 				     v->data_dev->name,
@@ -377,7 +380,7 @@ static int fec_decode_rsb(struct dm_verity *v, struct dm_verity_io *io,
 		if (unlikely(r < 0))
 			return r;
 
-		r = fec_decode_bufs(v, fio, rsb, r, pos, neras);
+		r = fec_decode_bufs(v, io, fio, rsb, r, pos, neras);
 		if (r < 0)
 			return r;
 
diff --git a/drivers/md/dm-verity-target.c b/drivers/md/dm-verity-target.c
index 4758bfe2c156..8cbf81fc0031 100644
--- a/drivers/md/dm-verity-target.c
+++ b/drivers/md/dm-verity-target.c
@@ -51,6 +51,7 @@ static DEFINE_STATIC_KEY_FALSE(use_tasklet_enabled);
 struct dm_verity_prefetch_work {
 	struct work_struct work;
 	struct dm_verity *v;
+	unsigned short ioprio;
 	sector_t block;
 	unsigned int n_blocks;
 };
@@ -294,6 +295,7 @@ static int verity_verify_level(struct dm_verity *v, struct dm_verity_io *io,
 	int r;
 	sector_t hash_block;
 	unsigned int offset;
+	struct bio *bio = dm_bio_from_per_bio_data(io, v->ti->per_io_data_size);
 
 	verity_hash_at_level(v, block, level, &hash_block, &offset);
 
@@ -308,7 +310,7 @@ static int verity_verify_level(struct dm_verity *v, struct dm_verity_io *io,
 			return -EAGAIN;
 		}
 	} else
-		data = dm_bufio_read(v->bufio, hash_block, &buf, IOPRIO_DEFAULT);
+		data = dm_bufio_read(v->bufio, hash_block, &buf, bio_prio(bio));
 
 	if (IS_ERR(data))
 		return PTR_ERR(data);
@@ -720,13 +722,14 @@ static void verity_prefetch_io(struct work_struct *work)
 no_prefetch_cluster:
 		dm_bufio_prefetch(v->bufio, hash_block_start,
 				  hash_block_end - hash_block_start + 1,
-				  IOPRIO_DEFAULT);
+				  pw->ioprio);
 	}
 
 	kfree(pw);
 }
 
-static void verity_submit_prefetch(struct dm_verity *v, struct dm_verity_io *io)
+static void verity_submit_prefetch(struct dm_verity *v, struct dm_verity_io *io,
+				   unsigned short ioprio)
 {
 	sector_t block = io->block;
 	unsigned int n_blocks = io->n_blocks;
@@ -754,6 +757,7 @@ static void verity_submit_prefetch(struct dm_verity *v, struct dm_verity_io *io)
 	pw->v = v;
 	pw->block = block;
 	pw->n_blocks = n_blocks;
+	pw->ioprio = ioprio;
 	queue_work(v->verify_wq, &pw->work);
 }
 
@@ -796,7 +800,7 @@ static int verity_map(struct dm_target *ti, struct bio *bio)
 
 	verity_fec_init_io(io);
 
-	verity_submit_prefetch(v, io);
+	verity_submit_prefetch(v, io, bio_prio(bio));
 
 	submit_bio_noacct(bio);
 
-- 
2.34.1
[PATCH v8 5/5] dm-crypt: Fix lost ioprio when queuing write bios
Posted by Hongyu Jin 1 year, 10 months ago
From: Hongyu Jin <hongyu.jin@unisoc.com>

Since dm-crypt queues writes to a different kernel thread (workqueue),
the bios will dispatch from tasks with different io_context->ioprio
settings and blkcg than the submitting task, thus giving incorrect
ioprio to the io scheduler.

Get the original io priority setting via struct dm_crypt_io::base_bio
and set this priority to the bio for write.

Link: https://lore.kernel.org/dm-devel/alpine.LRH.2.11.1612141049250.13402@mail.ewheeler.net

Signed-off-by: Hongyu Jin <hongyu.jin@unisoc.com>
---
 drivers/md/dm-crypt.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/md/dm-crypt.c b/drivers/md/dm-crypt.c
index 855b482cbff1..e0804a86946f 100644
--- a/drivers/md/dm-crypt.c
+++ b/drivers/md/dm-crypt.c
@@ -1683,6 +1683,7 @@ static struct bio *crypt_alloc_buffer(struct dm_crypt_io *io, unsigned int size)
 				 GFP_NOIO, &cc->bs);
 	clone->bi_private = io;
 	clone->bi_end_io = crypt_endio;
+	clone->bi_ioprio = io->base_bio->bi_ioprio;
 
 	remaining_size = size;
 
-- 
2.34.1