Re: [PATCH] PM / Suspend: Add more suspend notifier events

["Rafael J. Wysocki" <rjw@sisk.pl>]

On Friday, March 18, 2011, Daniel Drake wrote:
> Without becoming a full platform_suspend_ops implementation, the core
> olpc driver needs to be called during late suspend to apply a wakeup mask
> that will have been set by individual drivers' suspend routines.
> 
> Add a notifier here, and a symmetric one to happen during resume.

Why not to use a struct syscore_ops instead?

Rafael


> Signed-off-by: Daniel Drake <dsd@laptop.org>
> ---
>  include/linux/notifier.h |   10 ++++++----
>  kernel/power/suspend.c   |    5 +++++
>  2 files changed, 11 insertions(+), 4 deletions(-)
> 
> To further clarify how this will be used by upcoming patches:
> 
> OLPC's EC needs to be run with a specific event mask during runtime,
> and another one (based on configured wakeup events) during suspend.
> 
> The exact details of the in-suspend mask we are interested in cannot be
> computed until individual drivers suspend routines are run.
> 
> Therefore we need a way to hook in generic code that runs after normal drivers
> suspend routines, in process context with interrupts still enabled, to apply
> the in-suspend event mask.
> 
> This patch will add the olpc_ec_wakeup_set() API for setting the appropriate
> bits in the in-suspend event mask (stored only in memory to begin with), and
> uses the new notifiers to program that mask to the hardware at the appropriate
> time:
> http://dev.laptop.org/~dsd/20110114/0006-olpc-EC-SCI-wakeup-mask-functionality.patch
> 
> This patch uses the olpc_ec_wakeup_set() API to compute and set the suspend
> preference during olpc-battery's suspend routine:
> http://dev.laptop.org/~dsd/20110114/0013-olpc_battery-add-wakeup-support.patch
> (other drivers will follow this pattern too)
> 
> diff --git a/include/linux/notifier.h b/include/linux/notifier.h
> index 2026f9e..4519d4f 100644
> --- a/include/linux/notifier.h
> +++ b/include/linux/notifier.h
> @@ -254,10 +254,12 @@ static inline int notifier_to_errno(int ret)
>  /* Hibernation and suspend events */
>  #define PM_HIBERNATION_PREPARE	0x0001 /* Going to hibernate */
>  #define PM_POST_HIBERNATION	0x0002 /* Hibernation finished */
> -#define PM_SUSPEND_PREPARE	0x0003 /* Going to suspend the system */
> -#define PM_POST_SUSPEND		0x0004 /* Suspend finished */
> -#define PM_RESTORE_PREPARE	0x0005 /* Going to restore a saved image */
> -#define PM_POST_RESTORE		0x0006 /* Restore failed */
> +#define PM_SUSPEND_PREPARE	0x0003 /* Preparing to suspend the system */
> +#define PM_DO_SUSPEND		0x0004 /* Going to suspend the system */
> +#define PM_DO_RESUME		0x0005 /* Going to resume the system */
> +#define PM_POST_SUSPEND		0x0006 /* Suspend finished */
> +#define PM_RESTORE_PREPARE	0x0007 /* Going to restore a saved image */
> +#define PM_POST_RESTORE		0x0008 /* Restore failed */
>  
>  /* Console keyboard events.
>   * Note: KBD_KEYCODE is always sent before KBD_UNBOUND_KEYCODE, KBD_UNICODE and
> diff --git a/kernel/power/suspend.c b/kernel/power/suspend.c
> index 2814c32..35acfc5 100644
> --- a/kernel/power/suspend.c
> +++ b/kernel/power/suspend.c
> @@ -224,9 +224,14 @@ int suspend_devices_and_enter(suspend_state_t state)
>  	if (suspend_test(TEST_DEVICES))
>  		goto Recover_platform;
>  
> +	error = pm_notifier_call_chain(PM_DO_SUSPEND);
> +	if (error)
> +		goto Recover_platform;
> +
>  	suspend_enter(state);
>  
>   Resume_devices:
> +	pm_notifier_call_chain(PM_DO_RESUME);
>  	suspend_test_start();
>  	dpm_resume_end(PMSG_RESUME);
>  	suspend_test_finish("resume devices");
>