From: Ewan D. Milne emilne@redhat.com
[ Upstream commit e5cc9002caafacbaa8dab878d17a313192c3b03b ]
The block layer code will split a large zeroout request into multiple bios and if WRITE SAME is disabled because the storage device reports that it does not support it (or support the length used), we can get an error message from the block layer despite the setting of RQF_QUIET on the first request. This is because more than one request may have already been submitted.
Fix this by setting RQF_QUIET when BLK_STS_TARGET is returned to fail the request early, we don't need to log a message because we did not actually submit the command to the device, and the block layer code will handle the error by submitting individual write bios.
Link: https://lore.kernel.org/r/20201207221021.28243-1-emilne@redhat.com Reviewed-by: Christoph Hellwig hch@lst.de Signed-off-by: Ewan D. Milne emilne@redhat.com Signed-off-by: Martin K. Petersen martin.petersen@oracle.com Signed-off-by: Sasha Levin sashal@kernel.org --- drivers/scsi/sd.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c index 656bcf4940d6d..fedb89d4ac3f0 100644 --- a/drivers/scsi/sd.c +++ b/drivers/scsi/sd.c @@ -986,8 +986,10 @@ static blk_status_t sd_setup_write_zeroes_cmnd(struct scsi_cmnd *cmd) } }
- if (sdp->no_write_same) + if (sdp->no_write_same) { + rq->rq_flags |= RQF_QUIET; return BLK_STS_TARGET; + }
if (sdkp->ws16 || lba > 0xffffffff || nr_blocks > 0xffff) return sd_setup_write_same16_cmnd(cmd, false);