[PATCH] PM / core: Propagate dev->power.wakeup_path when no callbacks

[PATCH] PM / core: Propagate dev->power.wakeup_path when no callbacks

[Ulf Hansson]

The dev->power.direct_complete flag may become set in device_prepare() in
case the device don't have any PM callbacks (dev->power.no_pm_callbacks is
set). This leads to a broken behaviour, when there is child having wakeup
enabled and relies on its parent to be used in the wakeup path.

More precisely, when the direct complete path becomes selected for the
child in __device_suspend(), the propagation of the dev->power.wakeup_path
becomes skipped as well.

Let's address this problem, by checking if the device is a part the wakeup
path or has wakeup enabled, then prevent the direct complete path from
being used.

Reported-by: Loic Pallardy <loic.pallardy@st.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
---

More background:

This problem was reported by Loic Pallardy, offlist, while he was working
on enabling wakeup for a tty serial console driver.

When I looked more closely, I noticed that uart_suspend_port() calls
device_may_wakeup() for the tty child devices, and then also the used serial
driver check its device (parent) for device_may_wakeup(). To me this looks like
workarounds to fix a behaviour that really should be dealt with from the PM
core, no matter of whether the child have PM callbacks assigned or not.

In other words, it seems like the serial driver(s) should be checking the
wakeup_path flag for the parent, solely, instead.

I haven't digested further behaviours for other subsystem, but recently
reviewed a patch for a gpio driver [1], that seems to be suffering from the
similar problems.

Kind regards
Ulf Hansson

[1]
https://lkml.org/lkml/2019/4/4/1283

---
 drivers/base/power/main.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/drivers/base/power/main.c b/drivers/base/power/main.c
index 41eba82ee7b9..f9cfdeee8288 100644
--- a/drivers/base/power/main.c
+++ b/drivers/base/power/main.c
@@ -1747,6 +1747,10 @@ static int __device_suspend(struct device *dev, pm_message_t state, bool async)
 	if (dev->power.syscore)
 		goto Complete;
 
+	/* Avoid direct_complete, to let wakeup_path being propagated. */
+	if (device_may_wakeup(dev) || dev->power.wakeup_path)
+		dev->power.direct_complete = false;
+
 	if (dev->power.direct_complete) {
 		if (pm_runtime_status_suspended(dev)) {
 			pm_runtime_disable(dev);
-- 
2.17.1

Re: [PATCH] PM / core: Propagate dev->power.wakeup_path when no callbacks

[Alexandre Torgue]

Hi Ulf

On 4/10/19 11:55 AM, Ulf Hansson wrote:
> The dev->power.direct_complete flag may become set in device_prepare() in
> case the device don't have any PM callbacks (dev->power.no_pm_callbacks is
> set). This leads to a broken behaviour, when there is child having wakeup
> enabled and relies on its parent to be used in the wakeup path.
> 
> More precisely, when the direct complete path becomes selected for the
> child in __device_suspend(), the propagation of the dev->power.wakeup_path
> becomes skipped as well.
> 
> Let's address this problem, by checking if the device is a part the wakeup
> path or has wakeup enabled, then prevent the direct complete path from
> being used.
> 
> Reported-by: Loic Pallardy <loic.pallardy@st.com>
> Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
> ---
> 

Thanks Ulf for this patch. It'll avoid to have dirty hack in serial 
suspend callback (at least for stm32). I just tested it on stm32 kernel 
v4.19 (which embeds all our genpd based power features). Replacing
device_may_wakeup(dev) by
device_may_wakeup(dev) || dev->power.wakeup_path) then dirty hack to 
test ttydev wakeup flag in stm32 usart driver is no more needed.

So you can add my Tested-by

Regards
Alex


> More background:
> 
> This problem was reported by Loic Pallardy, offlist, while he was working
> on enabling wakeup for a tty serial console driver.
> 
> When I looked more closely, I noticed that uart_suspend_port() calls
> device_may_wakeup() for the tty child devices, and then also the used serial
> driver check its device (parent) for device_may_wakeup(). To me this looks like
> workarounds to fix a behaviour that really should be dealt with from the PM
> core, no matter of whether the child have PM callbacks assigned or not.
> 
> In other words, it seems like the serial driver(s) should be checking the
> wakeup_path flag for the parent, solely, instead.
> 
> I haven't digested further behaviours for other subsystem, but recently
> reviewed a patch for a gpio driver [1], that seems to be suffering from the
> similar problems.
> 
> Kind regards
> Ulf Hansson
> 
> [1]
> https://lkml.org/lkml/2019/4/4/1283
> 
> ---
>   drivers/base/power/main.c | 4 ++++
>   1 file changed, 4 insertions(+)
> 
> diff --git a/drivers/base/power/main.c b/drivers/base/power/main.c
> index 41eba82ee7b9..f9cfdeee8288 100644
> --- a/drivers/base/power/main.c
> +++ b/drivers/base/power/main.c
> @@ -1747,6 +1747,10 @@ static int __device_suspend(struct device *dev, pm_message_t state, bool async)
>   	if (dev->power.syscore)
>   		goto Complete;
>   
> +	/* Avoid direct_complete, to let wakeup_path being propagated. */
> +	if (device_may_wakeup(dev) || dev->power.wakeup_path)
> +		dev->power.direct_complete = false;
> +
>   	if (dev->power.direct_complete) {
>   		if (pm_runtime_status_suspended(dev)) {
>   			pm_runtime_disable(dev);
> 

Re: [PATCH] PM / core: Propagate dev->power.wakeup_path when no callbacks

[Rafael J. Wysocki]

On Wednesday, April 10, 2019 11:55:16 AM CEST Ulf Hansson wrote:
> The dev->power.direct_complete flag may become set in device_prepare() in
> case the device don't have any PM callbacks (dev->power.no_pm_callbacks is
> set). This leads to a broken behaviour, when there is child having wakeup
> enabled and relies on its parent to be used in the wakeup path.
> 
> More precisely, when the direct complete path becomes selected for the
> child in __device_suspend(), the propagation of the dev->power.wakeup_path
> becomes skipped as well.
> 
> Let's address this problem, by checking if the device is a part the wakeup
> path or has wakeup enabled, then prevent the direct complete path from
> being used.
> 
> Reported-by: Loic Pallardy <loic.pallardy@st.com>
> Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
> ---
> 
> More background:
> 
> This problem was reported by Loic Pallardy, offlist, while he was working
> on enabling wakeup for a tty serial console driver.
> 
> When I looked more closely, I noticed that uart_suspend_port() calls
> device_may_wakeup() for the tty child devices, and then also the used serial
> driver check its device (parent) for device_may_wakeup(). To me this looks like
> workarounds to fix a behaviour that really should be dealt with from the PM
> core, no matter of whether the child have PM callbacks assigned or not.
> 
> In other words, it seems like the serial driver(s) should be checking the
> wakeup_path flag for the parent, solely, instead.
> 
> I haven't digested further behaviours for other subsystem, but recently
> reviewed a patch for a gpio driver [1], that seems to be suffering from the
> similar problems.
> 
> Kind regards
> Ulf Hansson
> 
> [1]
> https://lkml.org/lkml/2019/4/4/1283
> 
> ---
>  drivers/base/power/main.c | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/drivers/base/power/main.c b/drivers/base/power/main.c
> index 41eba82ee7b9..f9cfdeee8288 100644
> --- a/drivers/base/power/main.c
> +++ b/drivers/base/power/main.c
> @@ -1747,6 +1747,10 @@ static int __device_suspend(struct device *dev, pm_message_t state, bool async)
>  	if (dev->power.syscore)
>  		goto Complete;
>  
> +	/* Avoid direct_complete, to let wakeup_path being propagated. */
> +	if (device_may_wakeup(dev) || dev->power.wakeup_path)
> +		dev->power.direct_complete = false;
> +
>  	if (dev->power.direct_complete) {
>  		if (pm_runtime_status_suspended(dev)) {
>  			pm_runtime_disable(dev);
> 

Applied, thanks!