diff mbox series

[RFC,v1] ufs: poll HCS.UCRDY before issuing a UIC command

Message ID 1683620674-160173-1-git-send-email-kwmad.kim@samsung.com (mailing list archive)
State Superseded
Headers show
Series [RFC,v1] ufs: poll HCS.UCRDY before issuing a UIC command | expand

Commit Message

Kiwoong Kim May 9, 2023, 8:24 a.m. UTC
With auto hibern8 enabled, UIC could be working
for a while to process a hibern8 operation and HCI
reports UIC not ready for a short term through HCS.UCRDY.
And UFS driver can't recognize the operation.
UFSHCI spec specifies UCRDY like this:
whether the host controller is ready to process UIC COMMAND

The 'ready' could be seen as many different meanings. If the meaning
includes not processing any request from HCI, processing a hibern8
operation can be 'not ready'. In this situation, the driver needs to
wait until the operations is completed.

Signed-off-by: Kiwoong Kim <kwmad.kim@samsung.com>
---
 drivers/ufs/core/ufshcd.c | 13 ++++++++++++-
 1 file changed, 12 insertions(+), 1 deletion(-)

Comments

Avri Altman May 10, 2023, 10:14 a.m. UTC | #1
> With auto hibern8 enabled, UIC could be working
> for a while to process a hibern8 operation and HCI
> reports UIC not ready for a short term through HCS.UCRDY.
> And UFS driver can't recognize the operation.
> UFSHCI spec specifies UCRDY like this:
> whether the host controller is ready to process UIC COMMAND
> 
> The 'ready' could be seen as many different meanings. If the meaning
> includes not processing any request from HCI, processing a hibern8
> operation can be 'not ready'. In this situation, the driver needs to
> wait until the operations is completed.
> 
> Signed-off-by: Kiwoong Kim <kwmad.kim@samsung.com>
Is this replaces your previous suggestion - https://lore.kernel.org/lkml/1682385635-43601-1-git-send-email-kwmad.kim@samsung.com/
Or is it addressing another issue?

Thanks,
Avri

> ---
>  drivers/ufs/core/ufshcd.c | 13 ++++++++++++-
>  1 file changed, 12 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/ufs/core/ufshcd.c b/drivers/ufs/core/ufshcd.c
> index 96ce6af..fc79539 100644
> --- a/drivers/ufs/core/ufshcd.c
> +++ b/drivers/ufs/core/ufshcd.c
> @@ -2368,7 +2368,18 @@ static inline int ufshcd_hba_capabilities(struct
> ufs_hba *hba)
>   */
>  static inline bool ufshcd_ready_for_uic_cmd(struct ufs_hba *hba)
>  {
> -       return ufshcd_readl(hba, REG_CONTROLLER_STATUS) &
> UIC_COMMAND_READY;
> +       ktime_t timeout = ktime_add_ms(ktime_get(), UIC_CMD_TIMEOUT);
> +       u32 val = 0;
> +
> +       do {
> +               val = ufshcd_readl(hba, REG_CONTROLLER_STATUS) &
> +                       UIC_COMMAND_READY;
> +               if (val)
> +                       break;
> +               usleep_range(500, 1000);
> +       } while (ktime_before(ktime_get(), timeout));
> +
> +       return val ? true : false;
>  }
> 
>  /**
> --
> 2.7.4
Bart Van Assche May 10, 2023, 6:53 p.m. UTC | #2
On 5/9/23 01:24, Kiwoong Kim wrote:
> With auto hibern8 enabled, UIC could be working
> for a while to process a hibern8 operation and HCI
> reports UIC not ready for a short term through HCS.UCRDY.
> And UFS driver can't recognize the operation.
> UFSHCI spec specifies UCRDY like this:
> whether the host controller is ready to process UIC COMMAND
> 
> The 'ready' could be seen as many different meanings. If the meaning
> includes not processing any request from HCI, processing a hibern8
> operation can be 'not ready'. In this situation, the driver needs to
> wait until the operations is completed.
> 
> Signed-off-by: Kiwoong Kim <kwmad.kim@samsung.com>
> ---
>   drivers/ufs/core/ufshcd.c | 13 ++++++++++++-
>   1 file changed, 12 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/ufs/core/ufshcd.c b/drivers/ufs/core/ufshcd.c
> index 96ce6af..fc79539 100644
> --- a/drivers/ufs/core/ufshcd.c
> +++ b/drivers/ufs/core/ufshcd.c
> @@ -2368,7 +2368,18 @@ static inline int ufshcd_hba_capabilities(struct ufs_hba *hba)
>    */
>   static inline bool ufshcd_ready_for_uic_cmd(struct ufs_hba *hba)
>   {
> -	return ufshcd_readl(hba, REG_CONTROLLER_STATUS) & UIC_COMMAND_READY;
> +	ktime_t timeout = ktime_add_ms(ktime_get(), UIC_CMD_TIMEOUT);
> +	u32 val = 0;
> +
> +	do {
> +		val = ufshcd_readl(hba, REG_CONTROLLER_STATUS) &
> +			UIC_COMMAND_READY;
> +		if (val)
> +			break;
> +		usleep_range(500, 1000);
> +	} while (ktime_before(ktime_get(), timeout));
> +
> +	return val ? true : false;
>   }
>   
>   /**

Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Bao D. Nguyen May 10, 2023, 9:24 p.m. UTC | #3
On 5/9/2023 1:24 AM, Kiwoong Kim wrote:
> With auto hibern8 enabled, UIC could be working
> for a while to process a hibern8 operation and HCI
> reports UIC not ready for a short term through HCS.UCRDY.
> And UFS driver can't recognize the operation.
> UFSHCI spec specifies UCRDY like this:
> whether the host controller is ready to process UIC COMMAND
> 
> The 'ready' could be seen as many different meanings. If the meaning
> includes not processing any request from HCI, processing a hibern8
> operation can be 'not ready'. In this situation, the driver needs to
> wait until the operations is completed.
> 
> Signed-off-by: Kiwoong Kim <kwmad.kim@samsung.com>
> ---
>   drivers/ufs/core/ufshcd.c | 13 ++++++++++++-
>   1 file changed, 12 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/ufs/core/ufshcd.c b/drivers/ufs/core/ufshcd.c
> index 96ce6af..fc79539 100644
> --- a/drivers/ufs/core/ufshcd.c
> +++ b/drivers/ufs/core/ufshcd.c
> @@ -2368,7 +2368,18 @@ static inline int ufshcd_hba_capabilities(struct ufs_hba *hba)
>    */
>   static inline bool ufshcd_ready_for_uic_cmd(struct ufs_hba *hba)
>   {
> -	return ufshcd_readl(hba, REG_CONTROLLER_STATUS) & UIC_COMMAND_READY;
> +	ktime_t timeout = ktime_add_ms(ktime_get(), UIC_CMD_TIMEOUT);
> +	u32 val = 0;
> +
> +	do {
> +		val = ufshcd_readl(hba, REG_CONTROLLER_STATUS) &
> +			UIC_COMMAND_READY;
> +		if (val)
> +			break;
> +		usleep_range(500, 1000);
Hi Kiwoong,
It looks like you are sleeping while holding the 
spin_lock_irqsave(hba->host->host_lock, flags)
in ufshcd_send_uic_cmd()?


> +	} while (ktime_before(ktime_get(), timeout));
> +
> +	return val ? true : false;
>   }
>   
>   /**
Bart Van Assche May 10, 2023, 9:46 p.m. UTC | #4
On 5/10/23 14:24, Bao D. Nguyen wrote:
> It looks like you are sleeping while holding the 
> spin_lock_irqsave(hba->host->host_lock, flags)
> in ufshcd_send_uic_cmd()?
Right. Please drop my Reviewed-by.

Thanks,

Bart.
Kiwoong Kim May 10, 2023, 11:35 p.m. UTC | #5
> On 5/9/2023 1:24 AM, Kiwoong Kim wrote:
> > With auto hibern8 enabled, UIC could be working for a while to process
> > a hibern8 operation and HCI reports UIC not ready for a short term
> > through HCS.UCRDY.
> > And UFS driver can't recognize the operation.
> > UFSHCI spec specifies UCRDY like this:
> > whether the host controller is ready to process UIC COMMAND
> >
> > The 'ready' could be seen as many different meanings. If the meaning
> > includes not processing any request from HCI, processing a hibern8
> > operation can be 'not ready'. In this situation, the driver needs to
> > wait until the operations is completed.
> >
> > Signed-off-by: Kiwoong Kim <kwmad.kim@samsung.com>
> > ---
> >   drivers/ufs/core/ufshcd.c | 13 ++++++++++++-
> >   1 file changed, 12 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/ufs/core/ufshcd.c b/drivers/ufs/core/ufshcd.c
> > index 96ce6af..fc79539 100644
> > --- a/drivers/ufs/core/ufshcd.c
> > +++ b/drivers/ufs/core/ufshcd.c
> > @@ -2368,7 +2368,18 @@ static inline int ufshcd_hba_capabilities(struct
> ufs_hba *hba)
> >    */
> >   static inline bool ufshcd_ready_for_uic_cmd(struct ufs_hba *hba)
> >   {
> > -	return ufshcd_readl(hba, REG_CONTROLLER_STATUS) & UIC_COMMAND_READY;
> > +	ktime_t timeout = ktime_add_ms(ktime_get(), UIC_CMD_TIMEOUT);
> > +	u32 val = 0;
> > +
> > +	do {
> > +		val = ufshcd_readl(hba, REG_CONTROLLER_STATUS) &
> > +			UIC_COMMAND_READY;
> > +		if (val)
> > +			break;
> > +		usleep_range(500, 1000);
> Hi Kiwoong,
> It looks like you are sleeping while holding the spin_lock_irqsave(hba-
> >host->host_lock, flags) in ufshcd_send_uic_cmd()?

Right. Let me fix it.

Thanks.
Kiwoong Kim

> 
> 
> > +	} while (ktime_before(ktime_get(), timeout));
> > +
> > +	return val ? true : false;
> >   }
> >
> >   /**
Adrian Hunter May 11, 2023, 5:22 a.m. UTC | #6
On 9/05/23 11:24, Kiwoong Kim wrote:
> With auto hibern8 enabled, UIC could be working
> for a while to process a hibern8 operation and HCI
> reports UIC not ready for a short term through HCS.UCRDY.
> And UFS driver can't recognize the operation.
> UFSHCI spec specifies UCRDY like this:
> whether the host controller is ready to process UIC COMMAND
> 
> The 'ready' could be seen as many different meanings. If the meaning
> includes not processing any request from HCI, processing a hibern8
> operation can be 'not ready'. In this situation, the driver needs to
> wait until the operations is completed.
> 
> Signed-off-by: Kiwoong Kim <kwmad.kim@samsung.com>
> ---
>  drivers/ufs/core/ufshcd.c | 13 ++++++++++++-
>  1 file changed, 12 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/ufs/core/ufshcd.c b/drivers/ufs/core/ufshcd.c
> index 96ce6af..fc79539 100644
> --- a/drivers/ufs/core/ufshcd.c
> +++ b/drivers/ufs/core/ufshcd.c
> @@ -2368,7 +2368,18 @@ static inline int ufshcd_hba_capabilities(struct ufs_hba *hba)
>   */
>  static inline bool ufshcd_ready_for_uic_cmd(struct ufs_hba *hba)
>  {
> -	return ufshcd_readl(hba, REG_CONTROLLER_STATUS) & UIC_COMMAND_READY;
> +	ktime_t timeout = ktime_add_ms(ktime_get(), UIC_CMD_TIMEOUT);
> +	u32 val = 0;
> +
> +	do {
> +		val = ufshcd_readl(hba, REG_CONTROLLER_STATUS) &
> +			UIC_COMMAND_READY;
> +		if (val)
> +			break;
> +		usleep_range(500, 1000);

Pedantically, the sleep probably needs to be less than the
auto-hibernate idle timer period?

> +	} while (ktime_before(ktime_get(), timeout));

read_poll_timeout() would be a better choice for I/O polling

> +
> +	return val ? true : false;
>  }
>  
>  /**
Kiwoong Kim May 12, 2023, 12:56 a.m. UTC | #7
> Pedantically, the sleep probably needs to be less than the auto-hibernate
> idle timer period?
> 
> > +	} while (ktime_before(ktime_get(), timeout));
> 
> read_poll_timeout() would be a better choice for I/O polling

It makes sense. Got it. Thanks.
Kiwoong Kim May 16, 2023, 3:28 a.m. UTC | #8
> > +	do {
> > +		val = ufshcd_readl(hba, REG_CONTROLLER_STATUS) &
> > +			UIC_COMMAND_READY;
> > +		if (val)
> > +			break;
> > +		usleep_range(500, 1000);
> Hi Kiwoong,
> It looks like you are sleeping while holding the spin_lock_irqsave(hba-
> >host->host_lock, flags) in ufshcd_send_uic_cmd()?

You're right. Let me fix it.
Kiwoong Kim May 16, 2023, 3:30 a.m. UTC | #9
> > With auto hibern8 enabled, UIC could be working for a while to process
> > a hibern8 operation and HCI reports UIC not ready for a short term
> > through HCS.UCRDY.
> > And UFS driver can't recognize the operation.
> > UFSHCI spec specifies UCRDY like this:
> > whether the host controller is ready to process UIC COMMAND
> >
> > The 'ready' could be seen as many different meanings. If the meaning
> > includes not processing any request from HCI, processing a hibern8
> > operation can be 'not ready'. In this situation, the driver needs to
> > wait until the operations is completed.
> >
> > Signed-off-by: Kiwoong Kim <kwmad.kim@samsung.com>
> Is this replaces your previous suggestion -
> https://lore.kernel.org/lkml/1682385635-43601-1-git-send-email-
> kwmad.kim@samsung.com/
> Or is it addressing another issue?
> 
> Thanks,
> Avri
> 

No, they are different. This is about reporting UIC ready and what you mentioned
is about reporting UIC busy in a certain race condition.
diff mbox series

Patch

diff --git a/drivers/ufs/core/ufshcd.c b/drivers/ufs/core/ufshcd.c
index 96ce6af..fc79539 100644
--- a/drivers/ufs/core/ufshcd.c
+++ b/drivers/ufs/core/ufshcd.c
@@ -2368,7 +2368,18 @@  static inline int ufshcd_hba_capabilities(struct ufs_hba *hba)
  */
 static inline bool ufshcd_ready_for_uic_cmd(struct ufs_hba *hba)
 {
-	return ufshcd_readl(hba, REG_CONTROLLER_STATUS) & UIC_COMMAND_READY;
+	ktime_t timeout = ktime_add_ms(ktime_get(), UIC_CMD_TIMEOUT);
+	u32 val = 0;
+
+	do {
+		val = ufshcd_readl(hba, REG_CONTROLLER_STATUS) &
+			UIC_COMMAND_READY;
+		if (val)
+			break;
+		usleep_range(500, 1000);
+	} while (ktime_before(ktime_get(), timeout));
+
+	return val ? true : false;
 }
 
 /**