diff mbox series

[v2,1/2] ufs: core: fix the issue of ICU failure

Message ID 20240902021805.1125-2-peter.wang@mediatek.com (mailing list archive)
State Superseded
Headers show
Series fix MCQ abort defect | expand

Commit Message

Peter Wang (王信友) Sept. 2, 2024, 2:18 a.m. UTC
From: Peter Wang <peter.wang@mediatek.com>

When setting the ICU bit without using read-modify-write,
SQRTCy will restart SQ again and receive an RTC return
error code 2 (Failure - SQ not stopped).

Additionally, the error log has been modified so that
this type of error can be observed.

Fixes: ab248643d3d6 ("scsi: ufs: core: Add error handling for MCQ mode")
Cc: stable@vger.kernel.org
Signed-off-by: Peter Wang <peter.wang@mediatek.com>
---
 drivers/ufs/core/ufs-mcq.c | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

Comments

Peter Wang (王信友) Sept. 3, 2024, 12:56 p.m. UTC | #1
On Mon, 2024-09-02 at 10:18 +0800, peter.wang@mediatek.com wrote:
> From: Peter Wang <peter.wang@mediatek.com>
> 
> When setting the ICU bit without using read-modify-write,
> SQRTCy will restart SQ again and receive an RTC return
> error code 2 (Failure - SQ not stopped).
> 
> Additionally, the error log has been modified so that
> this type of error can be observed.
> 
> Fixes: ab248643d3d6 ("scsi: ufs: core: Add error handling for MCQ
> mode")
> Cc: stable@vger.kernel.org
> Signed-off-by: Peter Wang <peter.wang@mediatek.com>
> ---
>  drivers/ufs/core/ufs-mcq.c | 9 +++++----
>  1 file changed, 5 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/ufs/core/ufs-mcq.c b/drivers/ufs/core/ufs-mcq.c
> index 5891cdacd0b3..afd9541f4bd8 100644
> --- a/drivers/ufs/core/ufs-mcq.c
> +++ b/drivers/ufs/core/ufs-mcq.c
> @@ -570,15 +570,16 @@ int ufshcd_mcq_sq_cleanup(struct ufs_hba *hba,
> int task_tag)
>  	writel(nexus, opr_sqd_base + REG_SQCTI);
>  
>  	/* SQRTCy.ICU = 1 */
> -	writel(SQ_ICU, opr_sqd_base + REG_SQRTC);
> +	writel(readl(opr_sqd_base + REG_SQRTC) | SQ_ICU,
> +		opr_sqd_base + REG_SQRTC);
> 

Hi Bao,

Could you help review this patch series?
It is a trivial MCQ spec violate bug fixed.

Thanks.
Peter


>  
>  	/* Poll SQRTSy.CUS = 1. Return result from SQRTSy.RTC */
>  	reg = opr_sqd_base + REG_SQRTS;
>  	err = read_poll_timeout(readl, val, val & SQ_CUS, 20,
>  				MCQ_POLL_US, false, reg);
> -	if (err)
> -		dev_err(hba->dev, "%s: failed. hwq=%d, tag=%d
> err=%ld\n",
> -			__func__, id, task_tag,
> +	if (err || FIELD_GET(SQ_ICU_ERR_CODE_MASK, readl(reg)))
> +		dev_err(hba->dev, "%s: failed. hwq=%d, tag=%d err=%d
> RTC=%ld\n",
> +			__func__, id, task_tag, err,
>  			FIELD_GET(SQ_ICU_ERR_CODE_MASK, readl(reg)));
>  
>  	if (ufshcd_mcq_sq_start(hba, hwq))
Bart Van Assche Sept. 5, 2024, 9:06 p.m. UTC | #2
On 9/1/24 7:18 PM, peter.wang@mediatek.com wrote:
>   	/* SQRTCy.ICU = 1 */

Feel free to leave out the above comment since it duplicates the code
below this comment. A comment that explains that "ICU = Initiate
Cleanup" probably would be appropriate.

> -	writel(SQ_ICU, opr_sqd_base + REG_SQRTC);
> +	writel(readl(opr_sqd_base + REG_SQRTC) | SQ_ICU,
> +		opr_sqd_base + REG_SQRTC);

Is this perhaps an open-coded version of ufshcd_rmwl()?

>   
>   	/* Poll SQRTSy.CUS = 1. Return result from SQRTSy.RTC */
>   	reg = opr_sqd_base + REG_SQRTS;
>   	err = read_poll_timeout(readl, val, val & SQ_CUS, 20,
>   				MCQ_POLL_US, false, reg);
> -	if (err)
> -		dev_err(hba->dev, "%s: failed. hwq=%d, tag=%d err=%ld\n",
> -			__func__, id, task_tag,
> +	if (err || FIELD_GET(SQ_ICU_ERR_CODE_MASK, readl(reg)))
> +		dev_err(hba->dev, "%s: failed. hwq=%d, tag=%d err=%d RTC=%ld\n",
> +			__func__, id, task_tag, err,
>   			FIELD_GET(SQ_ICU_ERR_CODE_MASK, readl(reg)));

In the above code the expression "FIELD_GET(SQ_ICU_ERR_CODE_MASK, 
readl(reg))" occurs twice. Please consider storing that expression in
a variable such that this expression only occurs once.

Thanks,

Bart.
Bao D. Nguyen Sept. 6, 2024, 8:39 p.m. UTC | #3
On 9/1/2024 7:18 PM, peter.wang@mediatek.com wrote:

>   	/* SQRTCy.ICU = 1 */
> -	writel(SQ_ICU, opr_sqd_base + REG_SQRTC);
> +	writel(readl(opr_sqd_base + REG_SQRTC) | SQ_ICU,
> +		opr_sqd_base + REG_SQRTC);
Hi Peter,
Instead of readl() here, how about write (SQ_STOP | SQ_ICU) to SQRTC?

Thanks, Bao
Peter Wang (王信友) Sept. 9, 2024, 3:38 a.m. UTC | #4
On Thu, 2024-09-05 at 14:06 -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 9/1/24 7:18 PM, peter.wang@mediatek.com wrote:
> >   /* SQRTCy.ICU = 1 */
> 
> Feel free to leave out the above comment since it duplicates the code
> below this comment. A comment that explains that "ICU = Initiate
> Cleanup" probably would be appropriate.
> 

Hi Bart,

Will change comment to  "/* Initiate Cleanup */"


> > -writel(SQ_ICU, opr_sqd_base + REG_SQRTC);
> > +writel(readl(opr_sqd_base + REG_SQRTC) | SQ_ICU,
> > +opr_sqd_base + REG_SQRTC);
> 
> Is this perhaps an open-coded version of ufshcd_rmwl()?
> 
> 

ufshcd_rmwl usage includes mmio_base, and it's not suitable for use
here.

> >   
> >   /* Poll SQRTSy.CUS = 1. Return result from SQRTSy.RTC */
> >   reg = opr_sqd_base + REG_SQRTS;
> >   err = read_poll_timeout(readl, val, val & SQ_CUS, 20,
> >   MCQ_POLL_US, false, reg);
> > -if (err)
> > -dev_err(hba->dev, "%s: failed. hwq=%d, tag=%d err=%ld\n",
> > -__func__, id, task_tag,
> > +if (err || FIELD_GET(SQ_ICU_ERR_CODE_MASK, readl(reg)))
> > +dev_err(hba->dev, "%s: failed. hwq=%d, tag=%d err=%d RTC=%ld\n",
> > +__func__, id, task_tag, err,
> >   FIELD_GET(SQ_ICU_ERR_CODE_MASK, readl(reg)));
> 
> In the above code the expression "FIELD_GET(SQ_ICU_ERR_CODE_MASK, 
> readl(reg))" occurs twice. Please consider storing that expression in
> a variable such that this expression only occurs once.
> 

Will use a variable.

Thanks.
Peter


> Thanks,
> 
> Bart.
>
Peter Wang (王信友) Sept. 9, 2024, 3:41 a.m. UTC | #5
On Fri, 2024-09-06 at 13:39 -0700, Bao D. Nguyen wrote:
>  	 
> External email : Please do not click links or open attachments until
> you have verified the sender or the content.
>  On 9/1/2024 7:18 PM, peter.wang@mediatek.com wrote:
> 
> >   /* SQRTCy.ICU = 1 */
> > -writel(SQ_ICU, opr_sqd_base + REG_SQRTC);
> > +writel(readl(opr_sqd_base + REG_SQRTC) | SQ_ICU,
> > +opr_sqd_base + REG_SQRTC);
> Hi Peter,
> Instead of readl() here, how about write (SQ_STOP | SQ_ICU) to SQRTC?
> 
> Thanks, Bao

Hi Bao,

My opinion is not to do it this way, because we don't know if 
the future specification of REG_SQRTC will add extra bits. 
Once added, such an approach would again become a bug in ufs driver.

Thanks
Peter
diff mbox series

Patch

diff --git a/drivers/ufs/core/ufs-mcq.c b/drivers/ufs/core/ufs-mcq.c
index 5891cdacd0b3..afd9541f4bd8 100644
--- a/drivers/ufs/core/ufs-mcq.c
+++ b/drivers/ufs/core/ufs-mcq.c
@@ -570,15 +570,16 @@  int ufshcd_mcq_sq_cleanup(struct ufs_hba *hba, int task_tag)
 	writel(nexus, opr_sqd_base + REG_SQCTI);
 
 	/* SQRTCy.ICU = 1 */
-	writel(SQ_ICU, opr_sqd_base + REG_SQRTC);
+	writel(readl(opr_sqd_base + REG_SQRTC) | SQ_ICU,
+		opr_sqd_base + REG_SQRTC);
 
 	/* Poll SQRTSy.CUS = 1. Return result from SQRTSy.RTC */
 	reg = opr_sqd_base + REG_SQRTS;
 	err = read_poll_timeout(readl, val, val & SQ_CUS, 20,
 				MCQ_POLL_US, false, reg);
-	if (err)
-		dev_err(hba->dev, "%s: failed. hwq=%d, tag=%d err=%ld\n",
-			__func__, id, task_tag,
+	if (err || FIELD_GET(SQ_ICU_ERR_CODE_MASK, readl(reg)))
+		dev_err(hba->dev, "%s: failed. hwq=%d, tag=%d err=%d RTC=%ld\n",
+			__func__, id, task_tag, err,
 			FIELD_GET(SQ_ICU_ERR_CODE_MASK, readl(reg)));
 
 	if (ufshcd_mcq_sq_start(hba, hwq))