[PATCH] clk: ti: clkctrl: Fix failed to enable error with double udelay timeout

[PATCH] clk: ti: clkctrl: Fix failed to enable error with double udelay timeout

[Tony Lindgren]

Commit 3d8598fb9c5a ("clk: ti: clkctrl: use fallback udelay approach if
timekeeping is suspended") added handling for cases when timekeeping is
suspended. But looks like we can still get occasional "failed to enable"
errors on the PM runtime resume path with udelay() returning faster than
expected.

With ti-sysc interconnect target module driver this leads into device
failure with PM runtime failing with "failed to enable" clkctrl error.

Let's fix the issue with a delay of two times the desired delay as in
often done for udelay() to account for the inaccuracy.

Fixes: 3d8598fb9c5a ("clk: ti: clkctrl: use fallback udelay approach if timekeeping is suspended")
Cc: Keerthy <j-keerthy@ti.com>
Cc: Tero Kristo <t-kristo@ti.com>
Signed-off-by: Tony Lindgren <tony@atomide.com>
---
 drivers/clk/ti/clkctrl.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/drivers/clk/ti/clkctrl.c b/drivers/clk/ti/clkctrl.c
--- a/drivers/clk/ti/clkctrl.c
+++ b/drivers/clk/ti/clkctrl.c
@@ -100,11 +100,12 @@ static bool _omap4_is_timeout(union omap4_timeout *time, u32 timeout)
 	 * can be from a timer that requires pm_runtime access, which
 	 * will eventually bring us here with timekeeping_suspended,
 	 * during both suspend entry and resume paths. This happens
-	 * at least on am43xx platform.
+	 * at least on am43xx platform. Account for flakeyness
+	 * with udelay() by multiplying the timeout value by 2.
 	 */
 	if (unlikely(_early_timeout || timekeeping_suspended)) {
 		if (time->cycles++ < timeout) {
-			udelay(1);
+			udelay(1 * 2);
 			return false;
 		}
 	} else {
-- 
2.23.0

Re: [PATCH] clk: ti: clkctrl: Fix failed to enable error with double udelay timeout

[Keerthy]

On 30/09/19 9:10 PM, Tony Lindgren wrote:
> Commit 3d8598fb9c5a ("clk: ti: clkctrl: use fallback udelay approach if
> timekeeping is suspended") added handling for cases when timekeeping is
> suspended. But looks like we can still get occasional "failed to enable"
> errors on the PM runtime resume path with udelay() returning faster than
> expected.
> 
> With ti-sysc interconnect target module driver this leads into device
> failure with PM runtime failing with "failed to enable" clkctrl error.
> 
> Let's fix the issue with a delay of two times the desired delay as in
> often done for udelay() to account for the inaccuracy.

Tested for DS0 and rtc+ddr modes on am43 and ds0 on am33.

Tested-by: Keerthy <j-keerthy@ti.com>

> 
> Fixes: 3d8598fb9c5a ("clk: ti: clkctrl: use fallback udelay approach if timekeeping is suspended")
> Cc: Keerthy <j-keerthy@ti.com>
> Cc: Tero Kristo <t-kristo@ti.com>
> Signed-off-by: Tony Lindgren <tony@atomide.com>
> ---
>   drivers/clk/ti/clkctrl.c | 5 +++--
>   1 file changed, 3 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/clk/ti/clkctrl.c b/drivers/clk/ti/clkctrl.c
> --- a/drivers/clk/ti/clkctrl.c
> +++ b/drivers/clk/ti/clkctrl.c
> @@ -100,11 +100,12 @@ static bool _omap4_is_timeout(union omap4_timeout *time, u32 timeout)
>   	 * can be from a timer that requires pm_runtime access, which
>   	 * will eventually bring us here with timekeeping_suspended,
>   	 * during both suspend entry and resume paths. This happens
> -	 * at least on am43xx platform.
> +	 * at least on am43xx platform. Account for flakeyness
> +	 * with udelay() by multiplying the timeout value by 2.
>   	 */
>   	if (unlikely(_early_timeout || timekeeping_suspended)) {
>   		if (time->cycles++ < timeout) {
> -			udelay(1);
> +			udelay(1 * 2);
>   			return false;
>   		}
>   	} else {
> 

Re: [PATCH] clk: ti: clkctrl: Fix failed to enable error with double udelay timeout

[Tony Lindgren]

* Keerthy <j-keerthy@ti.com> [191007 00:57]:
> 
> 
> On 30/09/19 9:10 PM, Tony Lindgren wrote:
> > Commit 3d8598fb9c5a ("clk: ti: clkctrl: use fallback udelay approach if
> > timekeeping is suspended") added handling for cases when timekeeping is
> > suspended. But looks like we can still get occasional "failed to enable"
> > errors on the PM runtime resume path with udelay() returning faster than
> > expected.
> > 
> > With ti-sysc interconnect target module driver this leads into device
> > failure with PM runtime failing with "failed to enable" clkctrl error.
> > 
> > Let's fix the issue with a delay of two times the desired delay as in
> > often done for udelay() to account for the inaccuracy.
> 
> Tested for DS0 and rtc+ddr modes on am43 and ds0 on am33.
> 
> Tested-by: Keerthy <j-keerthy@ti.com>

Thanks for testing. This one should be applied into v5.4-rc series
please if no more comments.

Regards,

Tony

> > Fixes: 3d8598fb9c5a ("clk: ti: clkctrl: use fallback udelay approach if timekeeping is suspended")
> > Cc: Keerthy <j-keerthy@ti.com>
> > Cc: Tero Kristo <t-kristo@ti.com>
> > Signed-off-by: Tony Lindgren <tony@atomide.com>
> > ---
> >   drivers/clk/ti/clkctrl.c | 5 +++--
> >   1 file changed, 3 insertions(+), 2 deletions(-)
> > 
> > diff --git a/drivers/clk/ti/clkctrl.c b/drivers/clk/ti/clkctrl.c
> > --- a/drivers/clk/ti/clkctrl.c
> > +++ b/drivers/clk/ti/clkctrl.c
> > @@ -100,11 +100,12 @@ static bool _omap4_is_timeout(union omap4_timeout *time, u32 timeout)
> >   	 * can be from a timer that requires pm_runtime access, which
> >   	 * will eventually bring us here with timekeeping_suspended,
> >   	 * during both suspend entry and resume paths. This happens
> > -	 * at least on am43xx platform.
> > +	 * at least on am43xx platform. Account for flakeyness
> > +	 * with udelay() by multiplying the timeout value by 2.
> >   	 */
> >   	if (unlikely(_early_timeout || timekeeping_suspended)) {
> >   		if (time->cycles++ < timeout) {
> > -			udelay(1);
> > +			udelay(1 * 2);
> >   			return false;
> >   		}
> >   	} else {
> > 

Re: [PATCH] clk: ti: clkctrl: Fix failed to enable error with double udelay timeout

[Tero Kristo]

On 10/10/2019 17:05, Tony Lindgren wrote:
> * Keerthy <j-keerthy@ti.com> [191007 00:57]:
>>
>>
>> On 30/09/19 9:10 PM, Tony Lindgren wrote:
>>> Commit 3d8598fb9c5a ("clk: ti: clkctrl: use fallback udelay approach if
>>> timekeeping is suspended") added handling for cases when timekeeping is
>>> suspended. But looks like we can still get occasional "failed to enable"
>>> errors on the PM runtime resume path with udelay() returning faster than
>>> expected.
>>>
>>> With ti-sysc interconnect target module driver this leads into device
>>> failure with PM runtime failing with "failed to enable" clkctrl error.
>>>
>>> Let's fix the issue with a delay of two times the desired delay as in
>>> often done for udelay() to account for the inaccuracy.
>>
>> Tested for DS0 and rtc+ddr modes on am43 and ds0 on am33.
>>
>> Tested-by: Keerthy <j-keerthy@ti.com>
> 
> Thanks for testing. This one should be applied into v5.4-rc series
> please if no more comments.

Queued up for 5.4 fixes, thanks.

-Tero

> 
> Regards,
> 
> Tony
> 
>>> Fixes: 3d8598fb9c5a ("clk: ti: clkctrl: use fallback udelay approach if timekeeping is suspended")
>>> Cc: Keerthy <j-keerthy@ti.com>
>>> Cc: Tero Kristo <t-kristo@ti.com>
>>> Signed-off-by: Tony Lindgren <tony@atomide.com>
>>> ---
>>>    drivers/clk/ti/clkctrl.c | 5 +++--
>>>    1 file changed, 3 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/drivers/clk/ti/clkctrl.c b/drivers/clk/ti/clkctrl.c
>>> --- a/drivers/clk/ti/clkctrl.c
>>> +++ b/drivers/clk/ti/clkctrl.c
>>> @@ -100,11 +100,12 @@ static bool _omap4_is_timeout(union omap4_timeout *time, u32 timeout)
>>>    	 * can be from a timer that requires pm_runtime access, which
>>>    	 * will eventually bring us here with timekeeping_suspended,
>>>    	 * during both suspend entry and resume paths. This happens
>>> -	 * at least on am43xx platform.
>>> +	 * at least on am43xx platform. Account for flakeyness
>>> +	 * with udelay() by multiplying the timeout value by 2.
>>>    	 */
>>>    	if (unlikely(_early_timeout || timekeeping_suspended)) {
>>>    		if (time->cycles++ < timeout) {
>>> -			udelay(1);
>>> +			udelay(1 * 2);
>>>    			return false;
>>>    		}
>>>    	} else {
>>>

--
Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki

Re: [PATCH] clk: ti: clkctrl: Fix failed to enable error with double udelay timeout

[Stephen Boyd]

Quoting Tony Lindgren (2019-09-30 08:40:01)
> Commit 3d8598fb9c5a ("clk: ti: clkctrl: use fallback udelay approach if
> timekeeping is suspended") added handling for cases when timekeeping is
> suspended. But looks like we can still get occasional "failed to enable"
> errors on the PM runtime resume path with udelay() returning faster than
> expected.
> 
> With ti-sysc interconnect target module driver this leads into device
> failure with PM runtime failing with "failed to enable" clkctrl error.
> 
> Let's fix the issue with a delay of two times the desired delay as in
> often done for udelay() to account for the inaccuracy.
> 
> Fixes: 3d8598fb9c5a ("clk: ti: clkctrl: use fallback udelay approach if timekeeping is suspended")
> Cc: Keerthy <j-keerthy@ti.com>
> Cc: Tero Kristo <t-kristo@ti.com>
> Signed-off-by: Tony Lindgren <tony@atomide.com>
> ---

Applied to clk-fixes