@@ -670,6 +670,7 @@ struct bio *bio_clone_bioset(struct bio *bio_src, gfp_t gfp_mask,
switch (bio_op(bio)) {
case REQ_OP_DISCARD:
case REQ_OP_SECURE_ERASE:
+ case REQ_OP_WRITE_ZEROES:
break;
case REQ_OP_WRITE_SAME:
bio->bi_io_vec[bio->bi_vcnt++] = bio_src->bi_io_vec[0];
@@ -1942,6 +1942,10 @@ static inline int bio_check_eod(struct bio *bio, unsigned int nr_sectors)
if (!bdev_is_zoned(bio->bi_bdev))
goto not_supported;
break;
+ case REQ_OP_WRITE_ZEROES:
+ if (!blk_queue_write_zeroes(q))
+ goto not_supported;
+ break;
default:
break;
}
@@ -227,6 +227,55 @@ int blkdev_issue_write_same(struct block_device *bdev, sector_t sector,
EXPORT_SYMBOL(blkdev_issue_write_same);
/**
+ * __blkdev_issue_write_zeroes - generate number of bios with WRITE ZEROES
+ * @bdev: blockdev to issue
+ * @sector: start sector
+ * @nr_sects: number of sectors to write
+ * @gfp_mask: memory allocation flags (for bio_alloc)
+ * @biop: pointer to anchor bio
+ *
+ * Description:
+ * Generate and issue number of bios(REQ_OP_WRITE_ZEROES) with zerofiled pages.
+ */
+static int __blkdev_issue_write_zeroes(struct block_device *bdev,
+ sector_t sector, sector_t nr_sects, gfp_t gfp_mask,
+ struct bio **biop)
+{
+ struct bio *bio = *biop;
+ unsigned int max_write_zeroes_sectors;
+ struct request_queue *q = bdev_get_queue(bdev);
+
+ if (!q)
+ return -ENXIO;
+
+ if (!blk_queue_write_zeroes(q))
+ return -EOPNOTSUPP;
+
+ /* Ensure that max_write_zeroes_sectors doesn't overflow bi_size */
+ max_write_zeroes_sectors = UINT_MAX >> 9;
+
+ while (nr_sects) {
+ bio = next_bio(bio, 0, gfp_mask);
+ bio->bi_iter.bi_sector = sector;
+ bio->bi_bdev = bdev;
+ bio_set_op_attrs(bio, REQ_OP_WRITE_ZEROES, 0);
+
+ if (nr_sects > max_write_zeroes_sectors) {
+ bio->bi_iter.bi_size = max_write_zeroes_sectors << 9;
+ nr_sects -= max_write_zeroes_sectors;
+ sector += max_write_zeroes_sectors;
+ } else {
+ bio->bi_iter.bi_size = nr_sects << 9;
+ nr_sects = 0;
+ }
+ cond_resched();
+ }
+
+ *biop = bio;
+ return 0;
+}
+
+/**
* __blkdev_issue_zeroout - generate number of zero filed write bios
* @bdev: blockdev to issue
* @sector: start sector
@@ -259,6 +308,11 @@ int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
goto out;
}
+ ret = __blkdev_issue_write_zeroes(bdev, sector, nr_sects, gfp_mask,
+ biop);
+ if (ret == 0 || (ret && ret != -EOPNOTSUPP))
+ goto out;
+
ret = __blkdev_issue_write_same(bdev, sector, nr_sects, gfp_mask,
ZERO_PAGE(0), biop);
if (ret == 0 || (ret && ret != -EOPNOTSUPP))
@@ -304,8 +358,8 @@ int __blkdev_issue_zeroout(struct block_device *bdev, sector_t sector,
* the discard request fail, if the discard flag is not set, or if
* discard_zeroes_data is not supported, this function will resort to
* zeroing the blocks manually, thus provisioning (allocating,
- * anchoring) them. If the block device supports the WRITE SAME command
- * blkdev_issue_zeroout() will use it to optimize the process of
+ * anchoring) them. If the block device supports WRITE ZEROES or WRITE SAME
+ * command(s), blkdev_issue_zeroout() will use it to optimize the process of
* clearing the block range. Otherwise the zeroing will be performed
* using regular WRITE calls.
*/
@@ -199,6 +199,10 @@ void blk_queue_split(struct request_queue *q, struct bio **bio,
case REQ_OP_SECURE_ERASE:
split = blk_bio_discard_split(q, *bio, bs, &nsegs);
break;
+ case REQ_OP_WRITE_ZEROES:
+ split = NULL;
+ nsegs = (*bio)->bi_phys_segments;
+ break;
case REQ_OP_WRITE_SAME:
split = blk_bio_write_same_split(q, *bio, bs, &nsegs);
break;
@@ -241,11 +245,15 @@ static unsigned int __blk_recalc_rq_segments(struct request_queue *q,
* This should probably be returning 0, but blk_add_request_payload()
* (Christoph!!!!)
*/
- if (bio_op(bio) == REQ_OP_DISCARD || bio_op(bio) == REQ_OP_SECURE_ERASE)
- return 1;
-
- if (bio_op(bio) == REQ_OP_WRITE_SAME)
+ switch (bio_op(bio)) {
+ case REQ_OP_DISCARD:
+ case REQ_OP_SECURE_ERASE:
+ case REQ_OP_WRITE_SAME:
+ case REQ_OP_WRITE_ZEROES:
return 1;
+ default:
+ break;
+ }
fbio = bio;
cluster = blk_queue_cluster(q);
@@ -416,6 +424,7 @@ static int __blk_bios_map_sg(struct request_queue *q, struct bio *bio,
switch (bio_op(bio)) {
case REQ_OP_DISCARD:
case REQ_OP_SECURE_ERASE:
+ case REQ_OP_WRITE_ZEROES:
/*
* This is a hack - drivers should be neither modifying the
* biovec, nor relying on bi_vcnt - but because of
@@ -575,9 +575,10 @@ static inline bool wbt_should_throttle(struct rq_wb *rwb, struct bio *bio)
const int op = bio_op(bio);
/*
- * If not a WRITE (or a discard), do nothing
+ * If not a WRITE (or a discard or write zeroes), do nothing
*/
- if (!(op == REQ_OP_WRITE || op == REQ_OP_DISCARD))
+ if (!(op == REQ_OP_WRITE || op == REQ_OP_DISCARD ||
+ op == REQ_OP_WRITE_ZEROES))
return false;
/*
@@ -76,7 +76,8 @@ static inline bool bio_has_data(struct bio *bio)
if (bio &&
bio->bi_iter.bi_size &&
bio_op(bio) != REQ_OP_DISCARD &&
- bio_op(bio) != REQ_OP_SECURE_ERASE)
+ bio_op(bio) != REQ_OP_SECURE_ERASE &&
+ bio_op(bio) != REQ_OP_WRITE_ZEROES)
return true;
return false;
@@ -86,7 +87,8 @@ static inline bool bio_no_advance_iter(struct bio *bio)
{
return bio_op(bio) == REQ_OP_DISCARD ||
bio_op(bio) == REQ_OP_SECURE_ERASE ||
- bio_op(bio) == REQ_OP_WRITE_SAME;
+ bio_op(bio) == REQ_OP_WRITE_SAME ||
+ bio_op(bio) == REQ_OP_WRITE_ZEROES;
}
static inline bool bio_mergeable(struct bio *bio)
@@ -188,18 +190,19 @@ static inline unsigned bio_segments(struct bio *bio)
struct bvec_iter iter;
/*
- * We special case discard/write same, because they interpret bi_size
- * differently:
+ * We special case discard/write same/write zeroes, because they
+ * interpret bi_size differently:
*/
- if (bio_op(bio) == REQ_OP_DISCARD)
- return 1;
-
- if (bio_op(bio) == REQ_OP_SECURE_ERASE)
- return 1;
-
- if (bio_op(bio) == REQ_OP_WRITE_SAME)
+ switch (bio_op(bio)) {
+ case REQ_OP_DISCARD:
+ case REQ_OP_SECURE_ERASE:
+ case REQ_OP_WRITE_SAME:
+ case REQ_OP_WRITE_ZEROES:
return 1;
+ default:
+ break;
+ }
bio_for_each_segment(bv, bio, iter)
segs++;
@@ -159,6 +159,8 @@ enum req_opf {
REQ_OP_ZONE_RESET = 6,
/* write the same sector many times */
REQ_OP_WRITE_SAME = 7,
+ /* write the zero filled sector many times */
+ REQ_OP_WRITE_ZEROES = 8,
REQ_OP_LAST,
};
@@ -595,6 +595,7 @@ struct request_queue {
#define QUEUE_FLAG_FLUSH_NQ 25 /* flush not queueuable */
#define QUEUE_FLAG_DAX 26 /* device supports DAX */
#define QUEUE_FLAG_STATS 27 /* track rq completion times */
+#define QUEUE_FLAG_WRITE_ZEROES 28 /* device supports write zeroes */
#define QUEUE_FLAG_DEFAULT ((1 << QUEUE_FLAG_IO_STAT) | \
(1 << QUEUE_FLAG_STACKABLE) | \
@@ -685,6 +686,8 @@ static inline void queue_flag_clear(unsigned int flag, struct request_queue *q)
#define blk_queue_secure_erase(q) \
(test_bit(QUEUE_FLAG_SECERASE, &(q)->queue_flags))
#define blk_queue_dax(q) test_bit(QUEUE_FLAG_DAX, &(q)->queue_flags)
+#define blk_queue_write_zeroes(q) \
+ (test_bit(QUEUE_FLAG_WRITE_ZEROES, &(q)->queue_flags))
#define blk_noretry_request(rq) \
((rq)->cmd_flags & (REQ_FAILFAST_DEV|REQ_FAILFAST_TRANSPORT| \
@@ -773,6 +776,9 @@ static inline bool rq_mergeable(struct request *rq)
if (req_op(rq) == REQ_OP_FLUSH)
return false;
+ if (req_op(rq) == REQ_OP_WRITE_ZEROES)
+ return false;
+
if (rq->cmd_flags & REQ_NOMERGE_FLAGS)
return false;
if (rq->rq_flags & RQF_NOMERGE_FLAGS)
This adds a new block layer operation to zero out a range of LBAs. This allows to implement zeroing for devices that don't use either discard with a predictable zero pattern or WRITE SAME of zeroes. The prominent example of that is NVMe with the Write Zeroes command, but in the future this should also help with improving the way zeroing discards work. Signed-off-by: Chaitanya Kulkarni <chaitanya.kulkarni@hgst.com> --- block/bio.c | 1 + block/blk-core.c | 4 ++++ block/blk-lib.c | 58 +++++++++++++++++++++++++++++++++++++++++++++-- block/blk-merge.c | 17 ++++++++++---- block/blk-wbt.c | 5 ++-- include/linux/bio.h | 25 +++++++++++--------- include/linux/blk_types.h | 2 ++ include/linux/blkdev.h | 6 +++++ 8 files changed, 99 insertions(+), 19 deletions(-)