On Wed, Jun 08, 2016 at 12:49:29PM +0200, Lukas Wunner wrote:
> The PM core introduced the ability to keep devices runtime suspended
> during the entire system suspend/resume process with commit aae4518b3124
> ("PM / sleep: Mechanism to avoid resuming runtime-suspended devices
> unnecessarily"). Before this so-called "direct-complete" procedure was
> introduced, devices were always runtime resumed only to be immediately
> put to sleep again using their ->suspend hook. Direct-complete is
> enabled by returning a positive value from the ->prepare hook. The PCI
> core usually does this automatically.
>
> Direct-complete is only available for a device if all children use it as
> well. Currently we cannot support direct-complete for DRM drivers
> because the DRM core automatically registers multiple DRM minors which
> belong to device class drm_class, and drm_class uses a struct dev_pm_ops
> which lacks the ->prepare callback.
>
> While this could be solved by adding the missing ->prepare callback,
> closer inspection shows that there are no DRM drivers left which declare
> the legacy ->suspend and ->resume callbacks in their drm_driver struct.
> The last ones to remove them were i915 with commit 1751fcf9f92e
> ("drm/i915: Fix module initialisation, v2.") and exynos with commit
> e7fefb1d5af5 ("drm/exynos: remove legacy ->suspend()/resume()").
>
> Consequently the struct dev_pm_ops of drm_class is now dead code. Remove
> it. If no dev_pm_ops is declared for a device, the PM core automatically
> enables direct-complete for it, thereby making that mechanism available
> to the parent DRM PCI devices.
>
> Signed-off-by: Lukas Wunner <lukas@xxxxxxxxx>
wow, awesome!
/me pops some champange
This is excellent really. Applied to drm-misc, but I'll wait with pushing
out since I want to roll forward to the latest drm-next that Dave promised
to push out tomorrow.
-Daniel
> ---
> drivers/gpu/drm/drm_drv.c | 2 --
> drivers/gpu/drm/drm_sysfs.c | 71 ---------------------------------------------
> include/drm/drmP.h | 2 --
> 3 files changed, 75 deletions(-)
>
> diff --git a/drivers/gpu/drm/drm_drv.c b/drivers/gpu/drm/drm_drv.c
> index bff8922..8b2582a 100644
> --- a/drivers/gpu/drm/drm_drv.c
> +++ b/drivers/gpu/drm/drm_drv.c
> @@ -605,8 +605,6 @@ struct drm_device *drm_dev_alloc(struct drm_driver *driver,
> ret = drm_minor_alloc(dev, DRM_MINOR_CONTROL);
> if (ret)
> goto err_minors;
> -
> - WARN_ON(driver->suspend || driver->resume);
> }
>
> if (drm_core_check_feature(dev, DRIVER_RENDER)) {
> diff --git a/drivers/gpu/drm/drm_sysfs.c b/drivers/gpu/drm/drm_sysfs.c
> index fa7fadc..32dd821 100644
> --- a/drivers/gpu/drm/drm_sysfs.c
> +++ b/drivers/gpu/drm/drm_sysfs.c
> @@ -32,75 +32,6 @@ static struct device_type drm_sysfs_device_minor = {
>
> struct class *drm_class;
>
> -/**
> - * __drm_class_suspend - internal DRM class suspend routine
> - * @dev: Linux device to suspend
> - * @state: power state to enter
> - *
> - * Just figures out what the actual struct drm_device associated with
> - * @dev is and calls its suspend hook, if present.
> - */
> -static int __drm_class_suspend(struct device *dev, pm_message_t state)
> -{
> - if (dev->type == &drm_sysfs_device_minor) {
> - struct drm_minor *drm_minor = to_drm_minor(dev);
> - struct drm_device *drm_dev = drm_minor->dev;
> -
> - if (drm_minor->type == DRM_MINOR_LEGACY &&
> - !drm_core_check_feature(drm_dev, DRIVER_MODESET) &&
> - drm_dev->driver->suspend)
> - return drm_dev->driver->suspend(drm_dev, state);
> - }
> - return 0;
> -}
> -
> -/**
> - * drm_class_suspend - internal DRM class suspend hook. Simply calls
> - * __drm_class_suspend() with the correct pm state.
> - * @dev: Linux device to suspend
> - */
> -static int drm_class_suspend(struct device *dev)
> -{
> - return __drm_class_suspend(dev, PMSG_SUSPEND);
> -}
> -
> -/**
> - * drm_class_freeze - internal DRM class freeze hook. Simply calls
> - * __drm_class_suspend() with the correct pm state.
> - * @dev: Linux device to freeze
> - */
> -static int drm_class_freeze(struct device *dev)
> -{
> - return __drm_class_suspend(dev, PMSG_FREEZE);
> -}
> -
> -/**
> - * drm_class_resume - DRM class resume hook
> - * @dev: Linux device to resume
> - *
> - * Just figures out what the actual struct drm_device associated with
> - * @dev is and calls its resume hook, if present.
> - */
> -static int drm_class_resume(struct device *dev)
> -{
> - if (dev->type == &drm_sysfs_device_minor) {
> - struct drm_minor *drm_minor = to_drm_minor(dev);
> - struct drm_device *drm_dev = drm_minor->dev;
> -
> - if (drm_minor->type == DRM_MINOR_LEGACY &&
> - !drm_core_check_feature(drm_dev, DRIVER_MODESET) &&
> - drm_dev->driver->resume)
> - return drm_dev->driver->resume(drm_dev);
> - }
> - return 0;
> -}
> -
> -static const struct dev_pm_ops drm_class_dev_pm_ops = {
> - .suspend = drm_class_suspend,
> - .resume = drm_class_resume,
> - .freeze = drm_class_freeze,
> -};
> -
> static char *drm_devnode(struct device *dev, umode_t *mode)
> {
> return kasprintf(GFP_KERNEL, "dri/%s", dev_name(dev));
> @@ -131,8 +62,6 @@ int drm_sysfs_init(void)
> if (IS_ERR(drm_class))
> return PTR_ERR(drm_class);
>
> - drm_class->pm = &drm_class_dev_pm_ops;
> -
> err = class_create_file(drm_class, &class_attr_version.attr);
> if (err) {
> class_destroy(drm_class);
> diff --git a/include/drm/drmP.h b/include/drm/drmP.h
> index c5d2950..fc40543 100644
> --- a/include/drm/drmP.h
> +++ b/include/drm/drmP.h
> @@ -416,8 +416,6 @@ struct drm_driver {
> void (*postclose) (struct drm_device *, struct drm_file *);
> void (*lastclose) (struct drm_device *);
> int (*unload) (struct drm_device *);
> - int (*suspend) (struct drm_device *, pm_message_t state);
> - int (*resume) (struct drm_device *);
> int (*dma_ioctl) (struct drm_device *dev, void *data, struct drm_file *file_priv);
> int (*dma_quiescent) (struct drm_device *);
> int (*context_dtor) (struct drm_device *dev, int context);
> --
> 2.8.1
>
> _______________________________________________
> dri-devel mailing list
> dri-devel@xxxxxxxxxxxxxxxxxxxxx
> https://lists.freedesktop.org/mailman/listinfo/dri-devel
--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
dri-devel mailing list
dri-devel@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/dri-devel
