Virtualization - Dec 2021 - [PATCH] drm: Return error codes from struct drm_driver.gem_create_object

Hi

Am 06.12.21 um 11:42 schrieb Dan Carpenter:
> On Tue, Nov 30, 2021 at 10:52:55AM +0100, Thomas Zimmermann wrote:
>> GEM helper libraries use struct drm_driver.gem_create_object to let
>> drivers override GEM object allocation. On failure, the call returns
>> NULL.
>>
>> Change the semantics to make the calls return a pointer-encoded error.
>> This aligns the callback with its callers. Fixes the ingenic driver,
>> which already returns an error pointer.
>>
>> Also update the callers to handle the involved types more strictly.
>>
>> Signed-off-by: Thomas Zimmermann <tzimmermann at suse.de>
>> ---
>> There is an alternative patch at [1] that updates the value returned
>> by ingenics' gem_create_object to NULL. Fixing the interface to
return
>> an errno code is more consistent with the rest of the GEM functions.
>>
>> [1] https://lore.kernel.org/dri-devel/20211118111522.GD1147 at kili/
> 
> My fix was already applied and backported to -stable etc...  Your
> patch is not developed against a current tree so you broke it.
Do you have a specific link? I just checked the stable tree at [1] and 
there no trace of your patch.

Patches for DRM should go through through DRM trees; drm-misc-fixes in 
this case. Exceptions should at least be announce on dri-devel. Neither 
is the case here.

Best regards
Thomas

[1] 
https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/tree/drivers/gpu/drm/ingenic/ingenic-drm-drv.c
> 
> That's the tricky thing with changing the API because say people wrote
> their code last week where returning NULL was correct.  When they submit
> their driver upstream, everything will merge and build but it will break
> at runtime.
> 
> For now, it's only vc4_create_object() which is broken.
> 
> regards,
> dan carpenter
> 
-- 
Thomas Zimmermann
Graphics Driver Developer
SUSE Software Solutions Germany GmbH
Maxfeldstr. 5, 90409 N?rnberg, Germany
(HRB 36809, AG N?rnberg)
Gesch?ftsf?hrer: Ivo Totev
-------------- next part --------------
A non-text attachment was scrubbed...
Name: OpenPGP_signature
Type: application/pgp-signature
Size: 840 bytes
Desc: OpenPGP digital signature
URL:
<http://lists.linuxfoundation.org/pipermail/virtualization/attachments/20211206/0b37dd52/attachment.sig>

On Mon, Dec 06, 2021 at 12:16:24PM +0100, Thomas Zimmermann
wrote:
> Hi
> 
> Am 06.12.21 um 11:42 schrieb Dan Carpenter:
> > On Tue, Nov 30, 2021 at 10:52:55AM +0100, Thomas Zimmermann wrote:
> > > GEM helper libraries use struct drm_driver.gem_create_object to
let
> > > drivers override GEM object allocation. On failure, the call
returns
> > > NULL.
> > > 
> > > Change the semantics to make the calls return a pointer-encoded
error.
> > > This aligns the callback with its callers. Fixes the ingenic
driver,
> > > which already returns an error pointer.
> > > 
> > > Also update the callers to handle the involved types more
strictly.
> > > 
> > > Signed-off-by: Thomas Zimmermann <tzimmermann at suse.de>
> > > ---
> > > There is an alternative patch at [1] that updates the value
returned
> > > by ingenics' gem_create_object to NULL. Fixing the interface
to return
> > > an errno code is more consistent with the rest of the GEM
functions.
> > > 
> > > [1] https://lore.kernel.org/dri-devel/20211118111522.GD1147 at
kili/
> > 
> > My fix was already applied and backported to -stable etc...  Your
> > patch is not developed against a current tree so you broke it.
> 
> Do you have a specific link? I just checked the stable tree at [1] and
there
> no trace of your patch.
It's in 5.15.6 and probably all the other supported -stable trees.

https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/tree/drivers/gpu/drm/vc4/vc4_bo.c?h=v5.15.6#n387
> 
> Patches for DRM should go through through DRM trees; drm-misc-fixes in this
> case. Exceptions should at least be announce on dri-devel. Neither is the
> case here.
Yeah.  That's a good question.  I don't know, because I just work
against linux-next...

regards,
dan carpenter