[PATCH v2 0/3] drm/cma: use prime helpers instead GEM CMA specific dma_buf functionality

[PATCH v2 0/3] drm/cma: use prime helpers instead GEM CMA specific dma_buf functionality

[Joonyoung Shim]

Hello,

This is the second version patchset.

GEM CMA supports dma_buf but it needs GEM CMA specific functionality for
dma_buf. We can use prime helpers for dma_buf by commit
89177644a7b6306e6084a89eab7e290f4bfef397 "drm: add prime helpers", so
this patchset is to replace from using GEM CMA specific functions to
using prime helpers.


To Laurent,

It is merged a patch to cache mapping from DRM Prime, can this patchset
get your ack?


Changes from v1:
- rebased from drm-next branch of 
git://people.freedesktop.org/~airlied/linux
- fix to check whether gem_prime_mmap field is NULL
   from "drm: add mmap function to prime helpers" commit


Joonyoung Shim (3):
       drm: add mmap function to prime helpers
       drm/cma: add low-level hook functions to use prime helpers
       drm/cma: remove GEM CMA specific dma_buf functionality

  drivers/gpu/drm/drm_gem_cma_helper.c | 293 
+++++------------------------------
  drivers/gpu/drm/drm_prime.c          |   8 +-
  include/drm/drmP.h                   |   2 +
  include/drm/drm_gem_cma_helper.h     |  13 +-
  4 files changed, 60 insertions(+), 256 deletions(-)

Re: [PATCH v2 0/3] drm/cma: use prime helpers instead GEM CMA specific dma_buf functionality

[Laurent Pinchart]

Hi Joonyoung,

Thank you for the patches.

On Friday 28 June 2013 14:24:43 Joonyoung Shim wrote:
> Hello,
> 
> This is the second version patchset.
> 
> GEM CMA supports dma_buf but it needs GEM CMA specific functionality for
> dma_buf. We can use prime helpers for dma_buf by commit
> 89177644a7b6306e6084a89eab7e290f4bfef397 "drm: add prime helpers", so
> this patchset is to replace from using GEM CMA specific functions to
> using prime helpers.
> 
> 
> To Laurent,
> 
> It is merged a patch to cache mapping from DRM Prime, can this patchset
> get your ack?

There you go (and sorry for the late reply)

Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>

By the way, between the initial version of the GEM CMA PRIME patch and the 
version that got merged in v3.10, commit 
011c2282c74db120f01a8414edc66c3f217f5511 ("drm: prime: fix refcounting on the 
dmabuf import error path") was introduced. The GEM CMA PRIME code in v3.10 
thus has a refcounting bug :-(

Should this patch set go to -stable, or should we cook up a special fix ?

> Changes from v1:
> - rebased from drm-next branch of
> git://people.freedesktop.org/~airlied/linux
> - fix to check whether gem_prime_mmap field is NULL
>    from "drm: add mmap function to prime helpers" commit
> 
> 
> Joonyoung Shim (3):
>        drm: add mmap function to prime helpers
>        drm/cma: add low-level hook functions to use prime helpers
>        drm/cma: remove GEM CMA specific dma_buf functionality
> 
>   drivers/gpu/drm/drm_gem_cma_helper.c | 293 ++++---------------------------
>   drivers/gpu/drm/drm_prime.c          |   8 +-
>   include/drm/drmP.h                   |   2 +
>   include/drm/drm_gem_cma_helper.h     |  13 +-
>   4 files changed, 60 insertions(+), 256 deletions(-)

-- 
Regards,

Laurent Pinchart

Re: [PATCH v2 0/3] drm/cma: use prime helpers instead GEM CMA specific dma_buf functionality

[Joonyoung Shim]

On 07/04/2013 07:11 AM, Laurent Pinchart wrote:
> Hi Joonyoung,
>
> Thank you for the patches.
>
> On Friday 28 June 2013 14:24:43 Joonyoung Shim wrote:
>> Hello,
>>
>> This is the second version patchset.
>>
>> GEM CMA supports dma_buf but it needs GEM CMA specific functionality for
>> dma_buf. We can use prime helpers for dma_buf by commit
>> 89177644a7b6306e6084a89eab7e290f4bfef397 "drm: add prime helpers", so
>> this patchset is to replace from using GEM CMA specific functions to
>> using prime helpers.
>>
>>
>> To Laurent,
>>
>> It is merged a patch to cache mapping from DRM Prime, can this patchset
>> get your ack?
> There you go (and sorry for the late reply)
>
> Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
>
> By the way, between the initial version of the GEM CMA PRIME patch and the
> version that got merged in v3.10, commit
> 011c2282c74db120f01a8414edc66c3f217f5511 ("drm: prime: fix refcounting on the
> dmabuf import error path") was introduced. The GEM CMA PRIME code in v3.10
> thus has a refcounting bug :-(
>
> Should this patch set go to -stable, or should we cook up a special fix ?

I'm not sure it's better to choose which way.

Dave, how we should do about that problem?

Re: [PATCH v2 0/3] drm/cma: use prime helpers instead GEM CMA specific dma_buf functionality

[Dave Airlie]

On Thu, Jul 4, 2013 at 5:14 PM, Joonyoung Shim <jy0922.shim@samsung.com> wrote:
> On 07/04/2013 07:11 AM, Laurent Pinchart wrote:
>>
>> Hi Joonyoung,
>>
>> Thank you for the patches.
>>
>> On Friday 28 June 2013 14:24:43 Joonyoung Shim wrote:
>>>
>>> Hello,
>>>
>>> This is the second version patchset.
>>>
>>> GEM CMA supports dma_buf but it needs GEM CMA specific functionality for
>>> dma_buf. We can use prime helpers for dma_buf by commit
>>> 89177644a7b6306e6084a89eab7e290f4bfef397 "drm: add prime helpers", so
>>> this patchset is to replace from using GEM CMA specific functions to
>>> using prime helpers.
>>>
>>>
>>> To Laurent,
>>>
>>> It is merged a patch to cache mapping from DRM Prime, can this patchset
>>> get your ack?
>>
>> There you go (and sorry for the late reply)
>>
>> Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
>>
>> By the way, between the initial version of the GEM CMA PRIME patch and the
>> version that got merged in v3.10, commit
>> 011c2282c74db120f01a8414edc66c3f217f5511 ("drm: prime: fix refcounting on
>> the
>> dmabuf import error path") was introduced. The GEM CMA PRIME code in v3.10
>> thus has a refcounting bug :-(
>>
>> Should this patch set go to -stable, or should we cook up a special fix ?
>
>
> I'm not sure it's better to choose which way.
>
> Dave, how we should do about that problem?

I think a special fix for stable once we get these merged to Linus.

Dave.

Re: [PATCH v2 0/3] drm/cma: use prime helpers instead GEM CMA specific dma_buf functionality

[Joonyoung Shim]

On 07/05/2013 02:38 PM, Dave Airlie wrote:
> On Thu, Jul 4, 2013 at 5:14 PM, Joonyoung Shim <jy0922.shim@samsung.com> wrote:
>> On 07/04/2013 07:11 AM, Laurent Pinchart wrote:
>>> Hi Joonyoung,
>>>
>>> Thank you for the patches.
>>>
>>> On Friday 28 June 2013 14:24:43 Joonyoung Shim wrote:
>>>> Hello,
>>>>
>>>> This is the second version patchset.
>>>>
>>>> GEM CMA supports dma_buf but it needs GEM CMA specific functionality for
>>>> dma_buf. We can use prime helpers for dma_buf by commit
>>>> 89177644a7b6306e6084a89eab7e290f4bfef397 "drm: add prime helpers", so
>>>> this patchset is to replace from using GEM CMA specific functions to
>>>> using prime helpers.
>>>>
>>>>
>>>> To Laurent,
>>>>
>>>> It is merged a patch to cache mapping from DRM Prime, can this patchset
>>>> get your ack?
>>> There you go (and sorry for the late reply)
>>>
>>> Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
>>>
>>> By the way, between the initial version of the GEM CMA PRIME patch and the
>>> version that got merged in v3.10, commit
>>> 011c2282c74db120f01a8414edc66c3f217f5511 ("drm: prime: fix refcounting on
>>> the
>>> dmabuf import error path") was introduced. The GEM CMA PRIME code in v3.10
>>> thus has a refcounting bug :-(
>>>
>>> Should this patch set go to -stable, or should we cook up a special fix ?
>>
>> I'm not sure it's better to choose which way.
>>
>> Dave, how we should do about that problem?
> I think a special fix for stable once we get these merged to Linus.
>

OK, i will post a patch for a special fix.

Thanks.

Re: [PATCH v2 0/3] drm/cma: use prime helpers instead GEM CMA specific dma_buf functionality

[Laurent Pinchart]

Hi Joonyoung,

On Friday 05 July 2013 15:30:25 Joonyoung Shim wrote:
> On 07/05/2013 02:38 PM, Dave Airlie wrote:
> > On Thu, Jul 4, 2013 at 5:14 PM, Joonyoung Shim wrote:
> >> On 07/04/2013 07:11 AM, Laurent Pinchart wrote:
> >>> On Friday 28 June 2013 14:24:43 Joonyoung Shim wrote:
> >>>> Hello,
> >>>> 
> >>>> This is the second version patchset.
> >>>> 
> >>>> GEM CMA supports dma_buf but it needs GEM CMA specific functionality
> >>>> for dma_buf. We can use prime helpers for dma_buf by commit
> >>>> 89177644a7b6306e6084a89eab7e290f4bfef397 "drm: add prime helpers", so
> >>>> this patchset is to replace from using GEM CMA specific functions to
> >>>> using prime helpers.
> >>>> 
> >>>> 
> >>>> To Laurent,
> >>>> 
> >>>> It is merged a patch to cache mapping from DRM Prime, can this patchset
> >>>> get your ack?
> >>> 
> >>> There you go (and sorry for the late reply)
> >>> 
> >>> Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
> >>> 
> >>> By the way, between the initial version of the GEM CMA PRIME patch and
> >>> the version that got merged in v3.10, commit
> >>> 011c2282c74db120f01a8414edc66c3f217f5511 ("drm: prime: fix refcounting
> >>> on the dmabuf import error path") was introduced. The GEM CMA PRIME code
> >>> in v3.10 thus has a refcounting bug :-(
> >>> 
> >>> Should this patch set go to -stable, or should we cook up a special fix
> >>> ?
> >> 
> >> I'm not sure it's better to choose which way.
> >> 
> >> Dave, how we should do about that problem?
> > 
> > I think a special fix for stable once we get these merged to Linus.
> 
> OK, i will post a patch for a special fix.

Thank you.

We need to fix the problem for v3.11 as well, by applying either this set or 
the fix you have just sent. I have no strong preference regarding which 
patches to pick.

-- 
Regards,

Laurent Pinchart

Re: [PATCH v2 0/3] drm/cma: use prime helpers instead GEM CMA specific dma_buf functionality

[Joonyoung Shim]

On 07/05/2013 10:03 PM, Laurent Pinchart wrote:
> Hi Joonyoung,
>
> On Friday 05 July 2013 15:30:25 Joonyoung Shim wrote:
>> On 07/05/2013 02:38 PM, Dave Airlie wrote:
>>> On Thu, Jul 4, 2013 at 5:14 PM, Joonyoung Shim wrote:
>>>> On 07/04/2013 07:11 AM, Laurent Pinchart wrote:
>>>>> On Friday 28 June 2013 14:24:43 Joonyoung Shim wrote:
>>>>>> Hello,
>>>>>>
>>>>>> This is the second version patchset.
>>>>>>
>>>>>> GEM CMA supports dma_buf but it needs GEM CMA specific functionality
>>>>>> for dma_buf. We can use prime helpers for dma_buf by commit
>>>>>> 89177644a7b6306e6084a89eab7e290f4bfef397 "drm: add prime helpers", so
>>>>>> this patchset is to replace from using GEM CMA specific functions to
>>>>>> using prime helpers.
>>>>>>
>>>>>>
>>>>>> To Laurent,
>>>>>>
>>>>>> It is merged a patch to cache mapping from DRM Prime, can this patchset
>>>>>> get your ack?
>>>>> There you go (and sorry for the late reply)
>>>>>
>>>>> Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
>>>>>
>>>>> By the way, between the initial version of the GEM CMA PRIME patch and
>>>>> the version that got merged in v3.10, commit
>>>>> 011c2282c74db120f01a8414edc66c3f217f5511 ("drm: prime: fix refcounting
>>>>> on the dmabuf import error path") was introduced. The GEM CMA PRIME code
>>>>> in v3.10 thus has a refcounting bug :-(
>>>>>
>>>>> Should this patch set go to -stable, or should we cook up a special fix
>>>>> ?
>>>> I'm not sure it's better to choose which way.
>>>>
>>>> Dave, how we should do about that problem?
>>> I think a special fix for stable once we get these merged to Linus.
>> OK, i will post a patch for a special fix.
> Thank you.
>
> We need to fix the problem for v3.11 as well, by applying either this set or
> the fix you have just sent. I have no strong preference regarding which
> patches to pick.
>

I also don't care, but i can report v3 patchset based on a patch for a
special fix for commit consistency.