diff mbox

drm/hisilicon: Use drm_connector_register_all

Message ID 20160506083725.GD7026@nuc-i3427.alporthouse.com (mailing list archive)
State New, archived
Headers show

Commit Message

Chris Wilson May 6, 2016, 8:37 a.m. UTC
On Fri, May 06, 2016 at 10:28:41AM +0200, Daniel Vetter wrote:
> Also, the unbind function is totally not sufficient, and it's calling
> the deprecated drm_put_dev.


And you probably should markup all the others that are now deprecated and
being phased out.

Maybe behind drm_deprecated so that the spam is hidden behind another
kconfig option rather than enabled-by-buildbots?
-Chris

Comments

Daniel Vetter May 6, 2016, 8:39 a.m. UTC | #1
On Fri, May 06, 2016 at 09:37:25AM +0100, Chris Wilson wrote:
> On Fri, May 06, 2016 at 10:28:41AM +0200, Daniel Vetter wrote:
> > Also, the unbind function is totally not sufficient, and it's calling
> > the deprecated drm_put_dev.
> 
> diff --git a/include/drm/drmP.h b/include/drm/drmP.h
> index 360b2a74e1ef..7efc0b477e38 100644
> --- a/include/drm/drmP.h
> +++ b/include/drm/drmP.h
> @@ -1032,7 +1032,7 @@ extern void drm_vblank_post_modeset(struct drm_device *dev, unsigned int pipe);
>  extern struct drm_master *drm_master_get(struct drm_master *master);
>  extern void drm_master_put(struct drm_master **master);
>  
> -extern void drm_put_dev(struct drm_device *dev);
> +__deprecated extern void drm_put_dev(struct drm_device *dev);
>  extern void drm_unplug_dev(struct drm_device *dev);
>  extern unsigned int drm_debug;
>  extern bool drm_atomic;
> 
> And you probably should markup all the others that are now deprecated and
> being phased out.
> 
> Maybe behind drm_deprecated so that the spam is hidden behind another
> kconfig option rather than enabled-by-buildbots?

I thought about this, but the problem is that we still have plenty drivers
using them. And hiding behind a build option is a bit meh, or at least
we'd need me and Dave to enable that option to make sure nothing slips
through.

Dave, do you think such an option would be useful? You'd need to integrate
it into your build testing when pulling drivers to make sure we catch it
all?

Cheers, Daniel
Emil Velikov May 6, 2016, 1:40 p.m. UTC | #2
On 6 May 2016 at 09:39, Daniel Vetter <daniel@ffwll.ch> wrote:
> On Fri, May 06, 2016 at 09:37:25AM +0100, Chris Wilson wrote:
>> On Fri, May 06, 2016 at 10:28:41AM +0200, Daniel Vetter wrote:
>> > Also, the unbind function is totally not sufficient, and it's calling
>> > the deprecated drm_put_dev.
>>
Pretty much the example that I was thinking about when I said "use
check-in cocci scripts". The driver was done ~2 weeks before the
drm_connect_{un,}register_all helpers.

>> diff --git a/include/drm/drmP.h b/include/drm/drmP.h
>> index 360b2a74e1ef..7efc0b477e38 100644
>> --- a/include/drm/drmP.h
>> +++ b/include/drm/drmP.h
>> @@ -1032,7 +1032,7 @@ extern void drm_vblank_post_modeset(struct drm_device *dev, unsigned int pipe);
>>  extern struct drm_master *drm_master_get(struct drm_master *master);
>>  extern void drm_master_put(struct drm_master **master);
>>
>> -extern void drm_put_dev(struct drm_device *dev);
>> +__deprecated extern void drm_put_dev(struct drm_device *dev);
>>  extern void drm_unplug_dev(struct drm_device *dev);
>>  extern unsigned int drm_debug;
>>  extern bool drm_atomic;
>>
>> And you probably should markup all the others that are now deprecated and
>> being phased out.
>>
>> Maybe behind drm_deprecated so that the spam is hidden behind another
>> kconfig option rather than enabled-by-buildbots?
>
> I thought about this, but the problem is that we still have plenty drivers
> using them. And hiding behind a build option is a bit meh, or at least
> we'd need me and Dave to enable that option to make sure nothing slips
> through.
>
> Dave, do you think such an option would be useful? You'd need to integrate
> it into your build testing when pulling drivers to make sure we catch it
> all?
>
Another interesting topic that I might have mentioned ;-)

About the defaults on such a toggle:
 - enabled, shouldn't cause too much spam as the __deprecated
warnings, are off by default.
Pros: Inspire people (maintainers, others) to cleanup the low hanging fruit.
Cons: Might annoy Dave/Daniel and people looking at buildbot logs,
when they set the latter.

 - disabled, temporary enabled on pull req.
Pros: 'Status check' and a list of things to poke driver maintainers.
Cons: Maintainers might not have both toggles enabled, thus will
likely miss things.


Whichever it is it's up-to you guys. Personally I think that we do
want to annotate symbols as __deprecated with or without the extra
configure (drm toggle).

Regards,
Emil
diff mbox

Patch

diff --git a/include/drm/drmP.h b/include/drm/drmP.h
index 360b2a74e1ef..7efc0b477e38 100644
--- a/include/drm/drmP.h
+++ b/include/drm/drmP.h
@@ -1032,7 +1032,7 @@  extern void drm_vblank_post_modeset(struct drm_device *dev, unsigned int pipe);
 extern struct drm_master *drm_master_get(struct drm_master *master);
 extern void drm_master_put(struct drm_master **master);
 
-extern void drm_put_dev(struct drm_device *dev);
+__deprecated extern void drm_put_dev(struct drm_device *dev);
 extern void drm_unplug_dev(struct drm_device *dev);
 extern unsigned int drm_debug;
 extern bool drm_atomic;