[PATCH 1/6] pm: add suspend_mem and suspend_standby support

[rjw@sisk.pl (Rafael J. Wysocki)]

On Sunday 07 of October 2012 09:27:15 Jean-Christophe PLAGNIOL-VILLARD wrote:
> Today when we go to suspend we can not knwon at drivers level if we go in
> STANDBY or MEM. Fix this by introducing two new callback suspend_mem and
> suspend_standby.
> 
> If a bus does not need to care about this distinction we fallback to
> suspend. This will allow to do not update the current bus or drivers.
> 
> This is needed as example by at91 OHCI and atmel serial
> 
> Cc: Nicolas Ferre <nicolas.ferre@atmel.com>
> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
> Signed-off-by: Jean-Christophe PLAGNIOL-VILLARD <plagnioj@jcrosoft.com>
> Cc: linux-pm at vger.kernel.org

So for me, this patch is a total no-go.

Apart from the fact that I don't see any difference between .suspend()
and .suspend_mem(), it doesn't cover the late/early and noirq phases.

> ---
>  drivers/base/power/main.c |   12 ++++++++++++
>  include/linux/pm.h        |    9 +++++++++
>  kernel/power/suspend.c    |   16 ++++++++++++++--
>  3 files changed, 35 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/base/power/main.c b/drivers/base/power/main.c
> index a3c1404..581e135 100644
> --- a/drivers/base/power/main.c
> +++ b/drivers/base/power/main.c
> @@ -218,9 +218,21 @@ static void dpm_wait_for_children(struct device *dev, bool async)
>   */
>  static pm_callback_t pm_op(const struct dev_pm_ops *ops, pm_message_t state)
>  {
> +	pm_callback_t callback = NULL;
> +
>  	switch (state.event) {
>  #ifdef CONFIG_SUSPEND
>  	case PM_EVENT_SUSPEND:
> +		switch (state.param) {
> +		case PM_SUSPEND_MEM:
> +			callback = ops->suspend_mem;
> +			break;
> +		case PM_SUSPEND_STANDBY:
> +			callback = ops->suspend_standby;
> +			break;
> +		}
> +		if (callback)
> +			return callback;
>  		return ops->suspend;
>  	case PM_EVENT_RESUME:
>  		return ops->resume;
> diff --git a/include/linux/pm.h b/include/linux/pm.h
> index 007e687..aff344b 100644
> --- a/include/linux/pm.h
> +++ b/include/linux/pm.h
> @@ -49,6 +49,7 @@ extern const char power_group_name[];		/* = "power" */
>  
>  typedef struct pm_message {
>  	int event;
> +	int param;
>  } pm_message_t;

Please don't do that.  We have the struct platform_suspend_ops with callbacks
that take the state argument and we pass the information about the system sleep
state being entered in there.  The way this information is interpreted depends
on the platform and the actions of device drivers resulting from that must be
platform-dependent too.  Your design makes a very strong assumption that the
interpretation of "standby" by the platform and device drivers will always
be the same, which generally need not be the case.

[And what if the given platform actually has more than two sleep states?
Are you going to add a separate callback for each of them?]

Make your drivers ask the platform what power state to put the devices into
instead of doing this.
 
>  /**
> @@ -265,6 +266,8 @@ struct dev_pm_ops {
>  	int (*prepare)(struct device *dev);
>  	void (*complete)(struct device *dev);
>  	int (*suspend)(struct device *dev);
> +	int (*suspend_mem)(struct device *dev);
> +	int (*suspend_standby)(struct device *dev);
>  	int (*resume)(struct device *dev);
>  	int (*freeze)(struct device *dev);
>  	int (*thaw)(struct device *dev);
> @@ -295,8 +298,12 @@ struct dev_pm_ops {
>  	.thaw = resume_fn, \
>  	.poweroff = suspend_fn, \
>  	.restore = resume_fn,
> +#define SET_SYSTEM_SLEEP_STANDBY_MEM_PM_OPS(suspend_standby_fn, suspend_mem_fn) \
> +	.suspend_standby = suspend_standby_fn, \
> +	.suspend_mem = suspend_mem_fn,
>  #else
>  #define SET_SYSTEM_SLEEP_PM_OPS(suspend_fn, resume_fn)
> +#define SET_SYSTEM_SLEEP_STANDBY_MEM_PM_OPS(suspend_standby_fn, suspend_mem_fn)
>  #endif
>  
>  #ifdef CONFIG_PM_RUNTIME
> @@ -414,6 +421,8 @@ const struct dev_pm_ops name = { \
>  #define PMSG_FREEZE	((struct pm_message){ .event = PM_EVENT_FREEZE, })
>  #define PMSG_QUIESCE	((struct pm_message){ .event = PM_EVENT_QUIESCE, })
>  #define PMSG_SUSPEND	((struct pm_message){ .event = PM_EVENT_SUSPEND, })
> +#define PMSG_SUSPEND_MEM	((struct pm_message){ .event = PM_EVENT_SUSPEND, .param = PM_SUSPEND_MEM, })
> +#define PMSG_SUSPEND_STANDBY	((struct pm_message){ .event = PM_EVENT_SUSPEND, .param = PM_SUSPEND_STANDBY, })
>  #define PMSG_HIBERNATE	((struct pm_message){ .event = PM_EVENT_HIBERNATE, })
>  #define PMSG_RESUME	((struct pm_message){ .event = PM_EVENT_RESUME, })
>  #define PMSG_THAW	((struct pm_message){ .event = PM_EVENT_THAW, })
> diff --git a/kernel/power/suspend.c b/kernel/power/suspend.c
> index c8b7446..9505b55 100644
> --- a/kernel/power/suspend.c
> +++ b/kernel/power/suspend.c
> @@ -126,6 +126,18 @@ void __attribute__ ((weak)) arch_suspend_enable_irqs(void)
>  	local_irq_enable();
>  }
>  
> +static pm_message_t suspend_state_to_pm_state(suspend_state_t state)
> +{
> +	switch (state) {
> +	case PM_SUSPEND_STANDBY:
> +		return PMSG_SUSPEND_STANDBY;
> +	case PM_SUSPEND_MEM:
> +		return PMSG_SUSPEND_MEM;
> +	default:
> +		return PMSG_SUSPEND;
> +	}
> +}

Well, this is more than ugly.

> +
>  /**
>   * suspend_enter - Make the system enter the given sleep state.
>   * @state: System sleep state to enter.
> @@ -143,7 +155,7 @@ static int suspend_enter(suspend_state_t state, bool *wakeup)
>  			goto Platform_finish;
>  	}
>  
> -	error = dpm_suspend_end(PMSG_SUSPEND);
> +	error = dpm_suspend_end(suspend_state_to_pm_state(state));
>  	if (error) {
>  		printk(KERN_ERR "PM: Some devices failed to power down\n");
>  		goto Platform_finish;
> @@ -215,7 +227,7 @@ int suspend_devices_and_enter(suspend_state_t state)
>  	suspend_console();
>  	ftrace_stop();
>  	suspend_test_start();
> -	error = dpm_suspend_start(PMSG_SUSPEND);
> +	error = dpm_suspend_start(suspend_state_to_pm_state(state));
>  	if (error) {
>  		printk(KERN_ERR "PM: Some devices failed to suspend\n");
>  		goto Recover_platform;

Thanks,
Rafael


-- 
I speak only for myself.
Rafael J. Wysocki, Intel Open Source Technology Center.