[PATCH v1 0/2] ufs: core: fix err handler mcq abort defect

[PATCH v1 0/2] ufs: core: fix err handler mcq abort defect

[peter.wang]

From: Peter Wang <peter.wang@mediatek.com>

This series fixes ufshcd_err_handler abort defect in mcq mode.

Peter Wang (2):
  ufs: core: complete scsi command after release
  ufs: core: force reset after mcq abort all

 drivers/ufs/core/ufshcd.c | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

-- 
2.45.2

[PATCH v1 1/2] ufs: core: complete scsi command after release

[peter.wang]

From: Peter Wang <peter.wang@mediatek.com>

When the error handler successfully aborts a MCQ request,
it only releases the command and does not notify the SCSI layer.
This may cause another abort after 30 seconds timeout.
This patch notifies the SCSI layer to requeue the request.

Below is error log
[   14.183804][   T74] ufshcd-mtk 112b0000.ufshci: ufshcd_err_handler started; HBA state eh_non_fatal; powered 1; shutting down 0; saved_err = 4; saved_uic_err = 64; force_reset = 0
[   14.256164][   T74] ufshcd-mtk 112b0000.ufshci: ufshcd_try_to_abort_task: cmd pending in the device. tag = 19
[   14.257511][   T74] ufshcd-mtk 112b0000.ufshci: Aborting tag 19 / CDB 0x35 succeeded
[   34.287949][    T8] ufshcd-mtk 112b0000.ufshci: ufshcd_abort: Device abort task at tag 19
[   34.290514][    T8] ufshcd-mtk 112b0000.ufshci: ufshcd_mcq_abort: skip abort. cmd at tag 19 already completed.

Fixes:93e6c0e19d5b ("scsi: ufs: core: Clear cmd if abort succeeds in MCQ mode")
Cc: <stable@vger.kernel.org> 6.6.x

Signed-off-by: Peter Wang <peter.wang@mediatek.com>
---
 drivers/ufs/core/ufshcd.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/drivers/ufs/core/ufshcd.c b/drivers/ufs/core/ufshcd.c
index 0b3d0c8e0dda..4bcd4e5b62bd 100644
--- a/drivers/ufs/core/ufshcd.c
+++ b/drivers/ufs/core/ufshcd.c
@@ -6482,8 +6482,12 @@ static bool ufshcd_abort_one(struct request *rq, void *priv)
 		if (!hwq)
 			return 0;
 		spin_lock_irqsave(&hwq->cq_lock, flags);
-		if (ufshcd_cmd_inflight(lrbp->cmd))
+		if (ufshcd_cmd_inflight(lrbp->cmd)) {
+			struct scsi_cmnd *cmd = lrbp->cmd;
+			set_host_byte(cmd, DID_REQUEUE);
 			ufshcd_release_scsi_cmd(hba, lrbp);
+			scsi_done(cmd);
+		}
 		spin_unlock_irqrestore(&hwq->cq_lock, flags);
 	}
 
-- 
2.45.2

Re: [PATCH v1 1/2] ufs: core: complete scsi command after release

[Bart Van Assche]

On 8/23/24 3:07 AM, peter.wang@mediatek.com wrote:
> diff --git a/drivers/ufs/core/ufshcd.c b/drivers/ufs/core/ufshcd.c
> index 0b3d0c8e0dda..4bcd4e5b62bd 100644
> --- a/drivers/ufs/core/ufshcd.c
> +++ b/drivers/ufs/core/ufshcd.c
> @@ -6482,8 +6482,12 @@ static bool ufshcd_abort_one(struct request *rq, void *priv)
>   		if (!hwq)
>   			return 0;
>   		spin_lock_irqsave(&hwq->cq_lock, flags);
> -		if (ufshcd_cmd_inflight(lrbp->cmd))
> +		if (ufshcd_cmd_inflight(lrbp->cmd)) {
> +			struct scsi_cmnd *cmd = lrbp->cmd;
> +			set_host_byte(cmd, DID_REQUEUE);
>   			ufshcd_release_scsi_cmd(hba, lrbp);
> +			scsi_done(cmd);
> +		}
>   		spin_unlock_irqrestore(&hwq->cq_lock, flags);
>   	}

Hmm ... isn't the ufshcd_complete_requests() call in ufshcd_abort_all()
expected to complete these commands? Can the above change lead to
scsi_done() being called twice, something that is not allowed?

Bart.

Re: [PATCH v1 1/2] ufs: core: complete scsi command after release

[Peter Wang (王信友)]

On Fri, 2024-08-23 at 09:20 -0700, Bart Van Assche wrote:
>  	 
> External email : Please do not click links or open attachments until
> you have verified the sender or the content.
>  On 8/23/24 3:07 AM, peter.wang@mediatek.com wrote:
> > diff --git a/drivers/ufs/core/ufshcd.c b/drivers/ufs/core/ufshcd.c
> > index 0b3d0c8e0dda..4bcd4e5b62bd 100644
> > --- a/drivers/ufs/core/ufshcd.c
> > +++ b/drivers/ufs/core/ufshcd.c
> > @@ -6482,8 +6482,12 @@ static bool ufshcd_abort_one(struct request
> *rq, void *priv)
> >   if (!hwq)
> >   return 0;
> >   spin_lock_irqsave(&hwq->cq_lock, flags);
> > -if (ufshcd_cmd_inflight(lrbp->cmd))
> > +if (ufshcd_cmd_inflight(lrbp->cmd)) {
> > +struct scsi_cmnd *cmd = lrbp->cmd;
> > +set_host_byte(cmd, DID_REQUEUE);
> >   ufshcd_release_scsi_cmd(hba, lrbp);
> > +scsi_done(cmd);
> > +}
> >   spin_unlock_irqrestore(&hwq->cq_lock, flags);
> >   }
> 
> Hmm ... isn't the ufshcd_complete_requests() call in
> ufshcd_abort_all()
> expected to complete these commands? Can the above change lead to
> scsi_done() being called twice, something that is not allowed?
> 
> Bart.
> 

Hi Bart,

ufshcd_complete_requests call in ufshcd_abort_all() with 
force_compl = false in mcq mode, which only check cq is empty or not.
For already abort cmd, no response will come back and no cq 
slot insert by hardware, which we need release this cmd and scsi done,
same as force_compl = true does.

Thanks
Peter

[PATCH v1 2/2] ufs: core: force reset after mcq abort all

[peter.wang]

From: Peter Wang <peter.wang@mediatek.com>

In mcq mode gerneal case, cq (head/tail) poniter is same as
sq pointer (head/tail) if the hwq is empty.

But if command send to device and abort it,
no response return and cq point will less than sq.
In this case will have unpredictable error.
This patch force reset for this case.

Below is error log
[   34.976612][    C3] ufshcd-mtk 112b0000.ufshci: OCS error from controller = 3 for tag 19

Signed-off-by: Peter Wang <peter.wang@mediatek.com>
---
 drivers/ufs/core/ufshcd.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/ufs/core/ufshcd.c b/drivers/ufs/core/ufshcd.c
index 4bcd4e5b62bd..d9ef8f0279da 100644
--- a/drivers/ufs/core/ufshcd.c
+++ b/drivers/ufs/core/ufshcd.c
@@ -6519,6 +6519,8 @@ static bool ufshcd_abort_all(struct ufs_hba *hba)
 	/* Complete the requests that are cleared by s/w */
 	ufshcd_complete_requests(hba, false);
 
+	if (is_mcq_enabled(hba))
+		return true;
 	return ret != 0;
 }
 
-- 
2.45.2

Re: [PATCH v1 2/2] ufs: core: force reset after mcq abort all

[Bart Van Assche]

On 8/23/24 3:07 AM, peter.wang@mediatek.com wrote:
> In mcq mode gerneal case, cq (head/tail) poniter is same as
> sq pointer (head/tail) if the hwq is empty.

poniter -> pointer

Thanks,

Bart.

Re: [PATCH v1 2/2] ufs: core: force reset after mcq abort all

[Peter Wang (王信友)]

On Fri, 2024-08-23 at 09:21 -0700, Bart Van Assche wrote:
>  	 
> External email : Please do not click links or open attachments until
> you have verified the sender or the content.
>  On 8/23/24 3:07 AM, peter.wang@mediatek.com wrote:
> > In mcq mode gerneal case, cq (head/tail) poniter is same as
> > sq pointer (head/tail) if the hwq is empty.
> 
> poniter -> pointer
> 
> Thanks,
> 
> Bart.

Hi Bart,

Will fix typo next version.

Thanks.
Peter