Re: 2.6.25-rc2 System no longer powers off after suspend-to-disk. Screen becomes green.

[Rafael J. Wysocki]

On Wednesday, 20 of February 2008, Jesse Barnes wrote:
> On Wednesday, February 20, 2008 11:18 am Jesse Barnes wrote:
> > On Wednesday, February 20, 2008 11:10 am Jeff Chua wrote:
> > > On Feb 21, 2008 2:53 AM, Jesse Barnes <jesse.barnes@xxxxxxxxx> wrote:
> > > > > So, next I'll try "shutdown" to see if it work. I was using
> > > > > "platform".
> > > >
> > > > Ok, that would be good to try.
> > >
> > > "shutdown" does power down properly. But still green on resume.
> >
> > Ok, so Linus' theory about something later in the resume path trying to
> > touch video is looking good.
> >
> > Rafael, is there anyway to prevent the device shutdown in the hibernate
> > path?
> 
> Given the way the PM core works, do we need to set a flag like this?  I really 
> hope there's a better way of doing this...

I think we should export the target sleep state somehow.

> diff --git a/drivers/char/drm/i915_drv.c b/drivers/char/drm/i915_drv.c
> index 4048f39..a2d6242 100644
> --- a/drivers/char/drm/i915_drv.c
> +++ b/drivers/char/drm/i915_drv.c
> @@ -238,6 +238,13 @@ static void i915_restore_vga(struct drm_device *dev)
>  
>  }
>  
> +/*
> + * If we're doing a suspend to disk, we don't want to power off the device.
> + * Unfortunately, the PM core doesn't tell us if we're headed for a regular
> + * S3 state or that it's about to shut down the machine, so we use this flag.
> + */
> +static int i915_hibernate;
> +
>  static int i915_suspend(struct drm_device *dev, pm_message_t state)
>  {
>  	struct drm_i915_private *dev_priv = dev->dev_private;
> @@ -252,6 +259,9 @@ static int i915_suspend(struct drm_device *dev, 
> pm_message_t state)
>  	if (state.event == PM_EVENT_PRETHAW)
>  		return 0;
>  
> +	if (state.event == PM_EVENT_FREEZE)
> +		i915_hibernate = 1;
> +
>  	pci_save_state(dev->pdev);
>  	pci_read_config_byte(dev->pdev, LBB, &dev_priv->saveLBB);
>  
> @@ -366,7 +376,7 @@ static int i915_suspend(struct drm_device *dev, 
> pm_message_t state)
>  
>  	i915_save_vga(dev);
>  
> -	if (state.event == PM_EVENT_SUSPEND) {
> +	if (!i915_hibernate) {
>  		/* Shut down the device */
>  		pci_disable_device(dev->pdev);
>  		pci_set_power_state(dev->pdev, PCI_D3hot);
> @@ -385,6 +395,8 @@ static int i915_resume(struct drm_device *dev)
>  	if (pci_enable_device(dev->pdev))
>  		return -1;
>  
> +	i915_hibernate = 0;
> +
>  	pci_write_config_byte(dev->pdev, LBB, dev_priv->saveLBB);
>  
>  	/* Pipe & plane A info */

Then, the .resume() called after the image creation will clear the flag and I
don't think it's safe to allow it to survive i915_resume() ...

Thanks,
Rafael
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/