On 5/22/25 00:31, Keith Busch wrote:
From: Keith Busch <kbusch@xxxxxxxxxx>
Various storage protocols can support offloading block data copies.
Enhance the block layer to know about the device's copying capabilities,
introduce the new REQ_OP_COPY operation, and provide the infrastructure
to iterate, split, and merge these kinds of requests.
A copy command must provide the device with a list of source LBAs and
their lengths, and a destination LBA. The 'struct bio' type doesn't
readily have a way to describe such a thing. But a copy request doesn't
use host memory for data, so the bio's bio_vec is unused space. This
patch adds a new purpose to the bio_vec where it can provide a vector of
sectors instead of memory pages.
Signed-off-by: Keith Busch <kbusch@xxxxxxxxxx>
---
block/bio.c | 25 ++++++++++++++
block/blk-core.c | 4 +++
block/blk-lib.c | 47 ++++++++++++++++++++++-----
block/blk-merge.c | 28 +++++++++++++++-
block/blk-sysfs.c | 9 ++++++
block/blk.h | 17 +++++++++-
include/linux/bio.h | 20 ++++++++++++
include/linux/blk-mq.h | 5 +++
include/linux/blk_types.h | 2 ++
include/linux/blkdev.h | 14 ++++++++
include/linux/bvec.h | 68 +++++++++++++++++++++++++++++++++++++--
11 files changed, 226 insertions(+), 13 deletions(-)
diff --git a/block/bio.c b/block/bio.c
index 3c0a558c90f52..9c73a895c987b 100644
--- a/block/bio.c
+++ b/block/bio.c
@@ -1156,6 +1156,31 @@ void bio_iov_bvec_set(struct bio *bio, const struct iov_iter *iter)
bio_set_flag(bio, BIO_CLONED);
}
+static int bvec_try_merge_copy_src(struct bio *bio, struct bio_vec *src)
+{
+ struct bio_vec *bv;
+
+ if (!bio->bi_vcnt)
+ return false;
+
+ bv = &bio->bi_io_vec[bio->bi_vcnt - 1];
+ if (bv->bv_sector + src->bv_sectors != src->bv_sector)
+ return false;
+
+ bv->bv_sectors += src->bv_sectors;
+ return true;
+}
+
+int bio_add_copy_src(struct bio *bio, struct bio_vec *src)
+{
+ if (bvec_try_merge_copy_src(bio, src))
+ return 0;
+ if (bio->bi_vcnt >= bio->bi_max_vecs)
+ return -EINVAL;
+ bio->bi_io_vec[bio->bi_vcnt++] = *src;
+ return 0;
+}
+
static unsigned int get_contig_folio_len(unsigned int *num_pages,
struct page **pages, unsigned int i,
struct folio *folio, size_t left,
diff --git a/block/blk-core.c b/block/blk-core.c
index b862c66018f25..cb3d9879e2d65 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -837,6 +837,10 @@ void submit_bio_noacct(struct bio *bio)
if (!bdev_max_discard_sectors(bdev))
goto not_supported;
break;
+ case REQ_OP_COPY:
+ if (!bdev_copy_sectors(bdev))
+ goto not_supported;
+ break;
case REQ_OP_SECURE_ERASE:
if (!bdev_max_secure_erase_sectors(bdev))
goto not_supported;
diff --git a/block/blk-lib.c b/block/blk-lib.c
index a819ded0ed3a9..a538acbaa2cd7 100644
--- a/block/blk-lib.c
+++ b/block/blk-lib.c
@@ -369,14 +369,7 @@ int blkdev_issue_secure_erase(struct block_device *bdev, sector_t sector,
}
EXPORT_SYMBOL(blkdev_issue_secure_erase);
-/**
- * blkdev_copy - copy source sectors to a destination on the same block device
- * @dst_sector: start sector of the destination to copy to
- * @src_sector: start sector of the source to copy from
- * @nr_sects: number of sectors to copy
- * @gfp: allocation flags to use
- */
-int blkdev_copy(struct block_device *bdev, sector_t dst_sector,
+static int __blkdev_copy(struct block_device *bdev, sector_t dst_sector,
sector_t src_sector, sector_t nr_sects, gfp_t gfp)
{
unsigned int nr_vecs = __blkdev_sectors_to_bio_pages(nr_sects);
That's a bit odd, renaming a just introduced function.
But if you must...
Otherwise looks good.
Cheers,
Hannes
--
Dr. Hannes Reinecke Kernel Storage Architect
hare@xxxxxxx +49 911 74053 688
SUSE Software Solutions GmbH, Frankenstr. 146, 90461 Nürnberg
HRB 36809 (AG Nürnberg), GF: I. Totev, A. McDonald, W. Knoblich
