[PATCH] scsi: ufs: Enable WB flush during suspend only if WB is enabled

[PATCH] scsi: ufs: Enable WB flush during suspend only if WB is enabled

[Stanley Chu]

WriteBootser flush during suspend is not necessary to be enabled if
WriteBooster feature is disabled. Simply adding a check to prevent
unexpected power drain.

Signed-off-by: Stanley Chu <stanley.chu@mediatek.com>
---
 drivers/scsi/ufs/ufshcd.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c
index 4879e87577e1..89fa8b9ac11d 100644
--- a/drivers/scsi/ufs/ufshcd.c
+++ b/drivers/scsi/ufs/ufshcd.c
@@ -5458,7 +5458,7 @@ static bool ufshcd_wb_need_flush(struct ufs_hba *hba)
 	u32 avail_buf;
 	u8 index;
 
-	if (!ufshcd_is_wb_allowed(hba))
+	if (!ufshcd_is_wb_allowed(hba) || !hba->wb_enabled)
 		return false;
 	/*
 	 * The ufs device needs the vcc to be ON to flush.
-- 
2.18.0

Re: [PATCH] scsi: ufs: Enable WB flush during suspend only if WB is enabled

[Bean Huo]

On Mon, 2020-12-07 at 13:50 +0800, Stanley Chu wrote:
> WriteBootser flush during suspend is not necessary to be enabled if
> WriteBooster feature is disabled. Simply adding a check to prevent
> unexpected power drain.
> 
> Signed-off-by: Stanley Chu <stanley.chu@mediatek.com>
> ---
>  drivers/scsi/ufs/ufshcd.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c
> index 4879e87577e1..89fa8b9ac11d 100644
> --- a/drivers/scsi/ufs/ufshcd.c
> +++ b/drivers/scsi/ufs/ufshcd.c
> @@ -5458,7 +5458,7 @@ static bool ufshcd_wb_need_flush(struct ufs_hba
> *hba)
>         u32 avail_buf;
>         u8 index;
>  
> -       if (!ufshcd_is_wb_allowed(hba))
> +       if (!ufshcd_is_wb_allowed(hba) || !hba->wb_enabled)
>                 return false;
>         /*
>          * The ufs device needs the vcc to be ON to flush.


Hi Stanley

In the 3.1 Spec:

"If the fWriteBoosterEn flag is set to zero, data written to any
logical unit is written in normal storage.
If the fWriteBoosterEn flag is set to one and the device is configured
in “shared buffer” mode, data written to any logical unit is written in
the shared WriteBooster Buffer."

so, IMO, fWriteBoosterEn is independant with WB buffer flush.

as for the flush:

"There are two methods for flushing data from the WriteBooster Buffer
to the normal storage: one is using an explicit flush command, the
other enabling the flushing during link hibernate state. If the
fWriteBoosterBufferFlushEn flag is set to one, the device shall flush
the data stored in the WriteBooster Buffer to the normal storage. If
fWriteBoosterBufferFlushDuringHibernate is set to one, the device
flushes the WriteBooster Buffer data automatically whenever the link
enters the hibernate (HIBERN8) state."

IMO, for the flush, it is controlled by fWriteBoosterBufferFlushEn and
fWriteBoosterBufferFlushDuringHibernate.

how do you understand the above two paragraphs from Spec?


thanks,
Bean

Re: [PATCH] scsi: ufs: Enable WB flush during suspend only if WB is enabled

[Stanley Chu]

Hi Bean,

On Mon, 2020-12-07 at 11:59 +0100, Bean Huo wrote:
> On Mon, 2020-12-07 at 13:50 +0800, Stanley Chu wrote:
> > WriteBootser flush during suspend is not necessary to be enabled if
> > WriteBooster feature is disabled. Simply adding a check to prevent
> > unexpected power drain.
> > 
> > Signed-off-by: Stanley Chu <stanley.chu@mediatek.com>
> > ---
> >  drivers/scsi/ufs/ufshcd.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c
> > index 4879e87577e1..89fa8b9ac11d 100644
> > --- a/drivers/scsi/ufs/ufshcd.c
> > +++ b/drivers/scsi/ufs/ufshcd.c
> > @@ -5458,7 +5458,7 @@ static bool ufshcd_wb_need_flush(struct ufs_hba
> > *hba)
> >         u32 avail_buf;
> >         u8 index;
> >  
> > -       if (!ufshcd_is_wb_allowed(hba))
> > +       if (!ufshcd_is_wb_allowed(hba) || !hba->wb_enabled)
> >                 return false;
> >         /*
> >          * The ufs device needs the vcc to be ON to flush.
> 
> 
> Hi Stanley
> 
> In the 3.1 Spec:
> 
> "If the fWriteBoosterEn flag is set to zero, data written to any
> logical unit is written in normal storage.
> If the fWriteBoosterEn flag is set to one and the device is configured
> in “shared buffer” mode, data written to any logical unit is written in
> the shared WriteBooster Buffer."
> 
> so, IMO, fWriteBoosterEn is independant with WB buffer flush.
> 
> as for the flush:
> 
> "There are two methods for flushing data from the WriteBooster Buffer
> to the normal storage: one is using an explicit flush command, the
> other enabling the flushing during link hibernate state. If the
> fWriteBoosterBufferFlushEn flag is set to one, the device shall flush
> the data stored in the WriteBooster Buffer to the normal storage. If
> fWriteBoosterBufferFlushDuringHibernate is set to one, the device
> flushes the WriteBooster Buffer data automatically whenever the link
> enters the hibernate (HIBERN8) state."
> 
> IMO, for the flush, it is controlled by fWriteBoosterBufferFlushEn and
> fWriteBoosterBufferFlushDuringHibernate.
> 
> how do you understand the above two paragraphs from Spec?

Thanks for your review and feedback! : )

Actually this patch is not motivated by any limitation in spec. I was
thinking that if host disables WriteBooster, which may imply that host
does not want any WB related operations in device during the disabled
period.

However I may be wrong because host may only want not consuming any WB
buffer in device during the disabled time, but still want the "flush"
operation in device to clean WB buffer as quickly as possible to fulfill
future high-throughput requirement after WB is re-enabled.

So, I would drop this patch.

Thanks,
Stanley Chu
> 
> 
> thanks,
> Bean
> 
>