[PATCH 0/2] block atomic writes tidy-ups/fix

[PATCH 0/2] block atomic writes tidy-ups/fix

[John Garry]

These two minor patches are tidy-ups for atomic write support.

Both are related to ignoring that REQ_ATOMIC can only be set for writes.

The block change could be considered a fix, as we are needlessly
checking for REQ_ATOMIC in fastpath.

Both changes can be picked up independently.

Thanks!

John Garry (2):
  scsi: sd: Don't check if a write for REQ_ATOMIC
  block: Don't check REQ_ATOMIC for reads

 block/blk-core.c  | 1 +
 drivers/scsi/sd.c | 2 +-
 2 files changed, 2 insertions(+), 1 deletion(-)

-- 
2.31.1

[PATCH 1/2] scsi: sd: Don't check if a write for REQ_ATOMIC

[John Garry]

Flag REQ_ATOMIC can only be set for writes, so don't check if the operation
is also a write in sd_setup_read_write_cmnd().

Fixes: bf4ae8f2e640 ("scsi: sd: Atomic write support")
Signed-off-by: John Garry <john.g.garry@oracle.com>
---
 drivers/scsi/sd.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c
index 8bb3a3611851..2ab47fb50c75 100644
--- a/drivers/scsi/sd.c
+++ b/drivers/scsi/sd.c
@@ -1382,7 +1382,7 @@ static blk_status_t sd_setup_read_write_cmnd(struct scsi_cmnd *cmd)
 	if (protect && sdkp->protection_type == T10_PI_TYPE2_PROTECTION) {
 		ret = sd_setup_rw32_cmnd(cmd, write, lba, nr_blocks,
 					 protect | fua, dld);
-	} else if (rq->cmd_flags & REQ_ATOMIC && write) {
+	} else if (rq->cmd_flags & REQ_ATOMIC) {
 		ret = sd_setup_atomic_cmnd(cmd, lba, nr_blocks,
 				sdkp->use_atomic_write_boundary,
 				protect | fua);
-- 
2.31.1

[PATCH 2/2] block: Don't check REQ_ATOMIC for reads

[John Garry]

We check in submit_bio_noacct() if flag REQ_ATOMIC is set for both read and
write operations, and then validate the atomic operation if set. Flag
REQ_ATOMIC can only be set for writes, so don't bother checking for reads.

Fixes: 9da3d1e912f3 ("block: Add core atomic write support")
Signed-off-by: John Garry <john.g.garry@oracle.com>
---
 block/blk-core.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/block/blk-core.c b/block/blk-core.c
index 1217c2cd66dd..bc5e8c5eaac9 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -799,6 +799,7 @@ void submit_bio_noacct(struct bio *bio)
 
 	switch (bio_op(bio)) {
 	case REQ_OP_READ:
+		break;
 	case REQ_OP_WRITE:
 		if (bio->bi_opf & REQ_ATOMIC) {
 			status = blk_validate_atomic_write_op_size(q, bio);
-- 
2.31.1

Re: [PATCH 2/2] block: Don't check REQ_ATOMIC for reads

[Kanchan Joshi]

Reviewed-by: Kanchan Joshi <joshi.k@samsung.com>

Re: [PATCH 1/2] scsi: sd: Don't check if a write for REQ_ATOMIC

[Kanchan Joshi]

Reviewed-by: Kanchan Joshi <joshi.k@samsung.com>

Re: [PATCH 0/2] block atomic writes tidy-ups/fix

[Martin K. Petersen]

John,

> These two minor patches are tidy-ups for atomic write support.

Applied to 6.12/scsi-staging, thanks!

-- 
Martin K. Petersen	Oracle Linux Engineering

Re: [PATCH 0/2] block atomic writes tidy-ups/fix

[Martin K. Petersen]

On Mon, 05 Aug 2024 11:33:13 +0000, John Garry wrote:

> These two minor patches are tidy-ups for atomic write support.
> 
> Both are related to ignoring that REQ_ATOMIC can only be set for writes.
> 
> The block change could be considered a fix, as we are needlessly
> checking for REQ_ATOMIC in fastpath.
> 
> [...]

Applied to 6.12/scsi-queue, thanks!

[1/2] scsi: sd: Don't check if a write for REQ_ATOMIC
      https://git.kernel.org/mkp/scsi/c/0c150b30d3d5
[2/2] block: Don't check REQ_ATOMIC for reads
      https://git.kernel.org/mkp/scsi/c/ea6787c695ab

-- 
Martin K. Petersen	Oracle Linux Engineering