According to below the patch applied, lrbp->cmd do not will NULL
after call release_scsi_cmd()
So check the rq->state unconditionally even if it is completed normally
in ufshcd_cmd_inflight()

If occurred abort status and tm_cmd timeout, will run err_handler
for re-init UFS.
Then err_handler will check pending request for clearing cmd.
At that time, check if the state of rq is not MQ_RQ_IDLE.
In other words, check if it is MQ_RQ_COMPLETE or MQ_RQ_IN_FLIGHT.
If rq->state is MQ_RQ_COMPLETE, it is already completed in the block,
so there is no need to process the queue.

Link: https://lore.kernel.org/linux-scsi/20230517223157.1068210-3-bvanassche@acm.org/
Signed-off-by: SEO HOYOUNG <hy50.seo@samsung.com>
---
 drivers/ufs/core/ufshcd.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/ufs/core/ufshcd.c b/drivers/ufs/core/ufshcd.c
index 21429eec1b82..3f47ea584cb1 100644
--- a/drivers/ufs/core/ufshcd.c
+++ b/drivers/ufs/core/ufshcd.c
@@ -3088,7 +3088,7 @@ bool ufshcd_cmd_inflight(struct scsi_cmnd *cmd)
 		return false;
 
 	rq = scsi_cmd_to_rq(cmd);
-	if (!blk_mq_request_started(rq))
+	if (blk_mq_rq_state(rq) != MQ_RQ_IN_FLIGHT)
 		return false;
 
 	return true;
-- 
2.26.0

On 4/5/24 00:16, SEO HOYOUNG wrote:
> According to below the patch applied, lrbp->cmd do not will NULL
> after call release_scsi_cmd()
> So check the rq->state unconditionally even if it is completed normally
> in ufshcd_cmd_inflight()
> 
> If occurred abort status and tm_cmd timeout, will run err_handler
> for re-init UFS.
> Then err_handler will check pending request for clearing cmd.
> At that time, check if the state of rq is not MQ_RQ_IDLE.
> In other words, check if it is MQ_RQ_COMPLETE or MQ_RQ_IN_FLIGHT.
> If rq->state is MQ_RQ_COMPLETE, it is already completed in the block,
> so there is no need to process the queue.

The above description does not explain the motivation for this patch
very well. How about the following description?

"ufshcd_cmd_inflight() is used to check whether or not a command is
in progress. Make it skip commands that have already completed by
changing the !blk_mq_request_started(rq) check into blk_mq_rq_state(rq) 
!= MQ_RQ_IN_FLIGHT. We cannot rely on lrbp->cmd since lrbp->cmd is not
cleared when a command completes."

> diff --git a/drivers/ufs/core/ufshcd.c b/drivers/ufs/core/ufshcd.c
> index 21429eec1b82..3f47ea584cb1 100644
> --- a/drivers/ufs/core/ufshcd.c
> +++ b/drivers/ufs/core/ufshcd.c
> @@ -3088,7 +3088,7 @@ bool ufshcd_cmd_inflight(struct scsi_cmnd *cmd)
>   		return false;
>   
>   	rq = scsi_cmd_to_rq(cmd);
> -	if (!blk_mq_request_started(rq))
> +	if (blk_mq_rq_state(rq) != MQ_RQ_IN_FLIGHT)
>   		return false;
>   
>   	return true;

Please convert the two return statements into a single return statement,
e.g. as follows:

	return cmd &&
	    blk_mq_rq_state(scsi_cmd_to_rq(cmd)) == MQ_RQ_IN_FLIGHT;

Thanks,

Bart.