On Thu, 12 Oct 2017, Mika Kahola <[email protected]> wrote:
> CI system spotted an error on CFL system when running IGT tests with
> display disabled. In this case, the 'INTEL_INFO(dev_priv)->num_pipes'
> is set to 0. This will cause that we prematurely return from
> 'get_saved_enc()'.
>
> To fix this issue, the patch introduces a 'max_pipes' variable which caches
> the maximum number of available pipes.

This is not the right fix. The audio component init should be skipped
and thus all component calls blocked when num_pipes == 0. It's just that
the num_pipes = 0 happens way too late when the display is fused off or
disabled via modparam.

This makes me wonder how many things we actually screw up because
intel_device_info_runtime_init() happens too late.

> Reference: https://bugs.freedesktop.org/show_bug.cgi?id=103206

If this patch fixes the bug, we use Bugzilla: tag. Reference: is for
other references, e.g. the referenced bug is related to the patch at
hand somehow, but this does not fix the bug.

BR,
Jani.


> Signed-off-by: Mika Kahola <[email protected]>
> ---
>  drivers/gpu/drm/i915/i915_drv.h          | 1 +
>  drivers/gpu/drm/i915/intel_audio.c       | 2 +-
>  drivers/gpu/drm/i915/intel_device_info.c | 2 ++
>  3 files changed, 4 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
> index 770305b..6229ff8 100644
> --- a/drivers/gpu/drm/i915/i915_drv.h
> +++ b/drivers/gpu/drm/i915/i915_drv.h
> @@ -868,6 +868,7 @@ struct intel_device_info {
>       u8 num_pipes;
>       u8 num_sprites[I915_MAX_PIPES];
>       u8 num_scalers[I915_MAX_PIPES];
> +     u8 max_pipes;
>  
>       unsigned int page_sizes; /* page sizes supported by the HW */
>  
> diff --git a/drivers/gpu/drm/i915/intel_audio.c 
> b/drivers/gpu/drm/i915/intel_audio.c
> index 0ddba16..147bd3d 100644
> --- a/drivers/gpu/drm/i915/intel_audio.c
> +++ b/drivers/gpu/drm/i915/intel_audio.c
> @@ -754,7 +754,7 @@ static struct intel_encoder *get_saved_enc(struct 
> drm_i915_private *dev_priv,
>  {
>       struct intel_encoder *encoder;
>  
> -     if (WARN_ON(pipe >= INTEL_INFO(dev_priv)->num_pipes))
> +     if (WARN_ON(pipe >= INTEL_INFO(dev_priv)->max_pipes))
>               return NULL;
>  
>       /* MST */
> diff --git a/drivers/gpu/drm/i915/intel_device_info.c 
> b/drivers/gpu/drm/i915/intel_device_info.c
> index 875d428..ba1a807 100644
> --- a/drivers/gpu/drm/i915/intel_device_info.c
> +++ b/drivers/gpu/drm/i915/intel_device_info.c
> @@ -376,6 +376,8 @@ void intel_device_info_runtime_init(struct 
> drm_i915_private *dev_priv)
>                       info->num_sprites[pipe] = 1;
>       }
>  
> +     info->max_pipes = info->num_pipes;
> +
>       if (i915_modparams.disable_display) {
>               DRM_INFO("Display disabled (module parameter)\n");
>               info->num_pipes = 0;

-- 
Jani Nikula, Intel Open Source Technology Center
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

Reply via email to