Re: [Intel-gfx] [PATCH 1/5] drm/i915: convert to using the drm_dbg_kms() macro.

[Jani Nikula]

On Wed, 08 Jan 2020, Chris Wilson <chris@xxxxxxxxxxxxxxxxxx> wrote:
> Quoting Jani Nikula (2020-01-08 09:40:40)
>> On Wed, 08 Jan 2020, Joonas Lahtinen <joonas.lahtinen@xxxxxxxxxxxxxxx> wrote:
>> > Quoting Wambui Karuga (2020-01-07 17:13:29)
>> >> Convert the use of the DRM_DEBUG_KMS() logging macro to the new struct
>> >> drm_device based drm_dbg_kms() logging macro in i915/intel_pch.c.
>> >> 
>> >> Signed-off-by: Wambui Karuga <wambui.karugax@xxxxxxxxx>
>> >> ---
>> >>  drivers/gpu/drm/i915/intel_pch.c | 46 +++++++++++++++++---------------
>> >>  1 file changed, 24 insertions(+), 22 deletions(-)
>> >> 
>> >> diff --git a/drivers/gpu/drm/i915/intel_pch.c b/drivers/gpu/drm/i915/intel_pch.c
>> >> index 43b68b5fc562..4ed60e1f01db 100644
>> >> --- a/drivers/gpu/drm/i915/intel_pch.c
>> >> +++ b/drivers/gpu/drm/i915/intel_pch.c
>> >> @@ -12,90 +12,91 @@ intel_pch_type(const struct drm_i915_private *dev_priv, unsigned short id)
>> >>  {
>> >>         switch (id) {
>> >>         case INTEL_PCH_IBX_DEVICE_ID_TYPE:
>> >> -               DRM_DEBUG_KMS("Found Ibex Peak PCH\n");
>> >> +               drm_dbg_kms(&dev_priv->drm, "Found Ibex Peak PCH\n");
>> >
>> > Did we at some point consider i915_dbg_kms alias? That would just take
>> > dev_priv (or i915, as it's called in newer code). It would shorten many
>> > of the statements.
>> >
>> > i915_dbg_kms(dev_priv, ...) or i915_dbg_kms(i915, ...)
>> 
>> I'd rather use the common drm logging macros. I thought about adding
>> i915 specific ones only if the drm device specific logging macros
>> weren't going to be merged.
>
> Why do they even exist? Why isn't it enough to do
> #define drm_info(drm, fmt, ...) dev_info(&(drm)->dev, fmt, ##__VA_ARGS) ?

It *is* enough to do that, and that's essentially what the new macros
do, just with an extra helper macro in between.

> #define i915_info(i915, fmt, ...) drm_info(&(i915)->drm, fmt, ##__VA_ARGS)
>
> The lea for &i915->drm.dev is the same as the mov, so we shave off an
> unnecessary wrapper.

I was hoping to avoid having our own aliases and abstractions of
everything, and instead making the drm macros do what we want. I mean I
could've just ignored drm completely, add our own hacks and convert the
driver...

Sure, there's the boilerplate of dereferencing &i915->drm, but in many
places we already have struct drm_device * available too.

BR,
Jani.

-- 
Jani Nikula, Intel Open Source Graphics Center