| Message ID | 20231214122919.985087-1-yebin10@huawei.com (mailing list archive) |
|---|---|
| State | Superseded |
| Headers | show |
| Series | scsi: core: add CMD_LAST flag for error handle scsi command | expand |
Looks good:
Reviewed-by: Christoph Hellwig <hch@lst.de>
diff --git a/drivers/scsi/scsi_error.c b/drivers/scsi/scsi_error.c index 1bac12ef238e..9e79047a1250 100644 --- a/drivers/scsi/scsi_error.c +++ b/drivers/scsi/scsi_error.c @@ -1147,6 +1147,7 @@ static enum scsi_disposition scsi_send_eh_cmnd(struct scsi_cmnd *scmd, const unsigned long stall_for = msecs_to_jiffies(100); int rtn; + scmd->flags |= SCMD_LAST; retry: scsi_eh_prep_cmnd(scmd, &ses, cmnd, cmnd_size, sense_bytes); shost->eh_action = &done;
SCSI error handle will send scsi command bypass block layer. After commit 8930a6c20791 scsi support request batching. Some LLD only writing the hardware doorbell when necessary, after the last request was prepared. For scsi error handle, each command waits synchronously. So each scsi command is both the beginning and the end. So add CMD_LAST flag for error handle scsi command. Fixes: 8930a6c20791 ("scsi: core: add support for request batching") Signed-off-by: Ye Bin <yebin10@huawei.com> --- drivers/scsi/scsi_error.c | 1 + 1 file changed, 1 insertion(+)