[GIT PULL FOR 3.9] Exynos SoC media drivers updates

[GIT PULL FOR 3.9] Exynos SoC media drivers updates

[Sylwester Nawrocki]

Hi Mauro,

Please pull the following for 3.9, it includes Exynos SoC drivers cleanups and
fixes. DMABUF exporting support for Exynos5 GScaler driver, device tree support
for Exynos MFC driver (platform bits for it got merged already for v3.8).

There is also included a patch removing deprecated image centering controls.

The following changes since commit 8cd7085ff460ead3aba6174052a408f4ad52ac36:

  [media] get_dvb_firmware: Fix the location of firmware for Terratec HTC
(2013-01-01 11:18:26 -0200)

are available in the git repository at:

  git://git.infradead.org/users/kmpark/linux-samsung media_for_v3.9

for you to fetch changes up to 36073ee2f7b3b5ae91900cb992b292404614243b:

  V4L: Remove deprecated image centering controls (2013-01-04 11:35:43 +0100)

----------------------------------------------------------------
Arun Kumar K (2):
      s5p-mfc: Add device tree support
      s5p-mfc: Flush DPB buffers during stream off

Kamil Debski (4):
      s5p-mfc: Move firmware allocation point to avoid allocation problems
      s5p-mfc: Correct check of vb2_dma_contig_init_ctx return value
      s5p-mfc: Change internal buffer allocation from vb2 ops to dma_alloc_coherent
      s5p-mfc: Context handling in open() bugfix

Sachin Kamat (9):
      s5p-tv: Add missing braces around sizeof in sdo_drv.c
      s5p-tv: Add missing braces around sizeof in mixer_video.c
      s5p-tv: Add missing braces around sizeof in mixer_reg.c
      s5p-tv: Add missing braces around sizeof in mixer_drv.c
      s5p-tv: Add missing braces around sizeof in hdmiphy_drv.c
      s5p-tv: Add missing braces around sizeof in hdmi_drv.c
      s5p-mfc: Remove redundant 'break'
      s5p-mfc: Fix a typo in error message in s5p_mfc_pm.c
      s5p-mfc: Fix an error check

Shaik Ameer Basha (1):
      exynos-gsc: Support dmabuf export buffer

Sylwester Nawrocki (5):
      s5p-fimc: Avoid possible NULL pointer dereference in set_fmt op
      s5p-fimc: Prevent potential buffer overflow
      s5p-fimc: Prevent AB-BA deadlock during links reconfiguration
      s5p-tv: Fix return value in sdo_probe() on error paths
      V4L: Remove deprecated image centering controls

Tomasz Stanislawski (1):
      s5p-tv: mixer: fix handling of VIDIOC_S_FMT

Tony Prisk (3):
      s5p-fimc: Fix incorrect usage of IS_ERR_OR_NULL
      s5p-tv: Fix incorrect usage of IS_ERR_OR_NULL
      s5p-g2d: Fix incorrect usage of IS_ERR_OR_NULL

Wei Yongjun (1):
      s5p-mfc: remove unused variable

 Documentation/DocBook/media/v4l/controls.xml    |   23 ---
 drivers/media/platform/exynos-gsc/gsc-m2m.c     |   12 +-
 drivers/media/platform/s5p-fimc/fimc-capture.c  |   58 +++++--
 drivers/media/platform/s5p-fimc/fimc-lite.c     |    6 +-
 drivers/media/platform/s5p-fimc/fimc-mdevice.c  |  100 +++++-------
 drivers/media/platform/s5p-g2d/g2d.c            |    4 +-
 drivers/media/platform/s5p-mfc/s5p_mfc.c        |  148 ++++++++++++-----
 drivers/media/platform/s5p-mfc/s5p_mfc_common.h |   31 ++--
 drivers/media/platform/s5p-mfc/s5p_mfc_ctrl.c   |  149 ++++++++---------
 drivers/media/platform/s5p-mfc/s5p_mfc_ctrl.h   |    3 +-
 drivers/media/platform/s5p-mfc/s5p_mfc_dec.c    |   15 +-
 drivers/media/platform/s5p-mfc/s5p_mfc_opr.c    |   30 ++++
 drivers/media/platform/s5p-mfc/s5p_mfc_opr.h    |    5 +
 drivers/media/platform/s5p-mfc/s5p_mfc_opr_v5.c |  197 ++++++++---------------
 drivers/media/platform/s5p-mfc/s5p_mfc_opr_v6.c |  148 ++++++-----------
 drivers/media/platform/s5p-mfc/s5p_mfc_pm.c     |    2 +-
 drivers/media/platform/s5p-tv/hdmi_drv.c        |   18 +--
 drivers/media/platform/s5p-tv/hdmiphy_drv.c     |    2 +-
 drivers/media/platform/s5p-tv/mixer_drv.c       |   14 +-
 drivers/media/platform/s5p-tv/mixer_reg.c       |    6 +-
 drivers/media/platform/s5p-tv/mixer_video.c     |   22 +--
 drivers/media/platform/s5p-tv/sdo_drv.c         |   29 ++--
 drivers/media/v4l2-core/v4l2-ctrls.c            |    2 -
 include/uapi/linux/v4l2-controls.h              |    4 -
 24 files changed, 503 insertions(+), 525 deletions(-)


Regards,
-- 
Sylwester Nawrocki
Samsung Poland R&D Center

Re: [GIT PULL FOR 3.9] Exynos SoC media drivers updates

[Sylwester Nawrocki]

On 01/04/2013 08:01 PM, Sylwester Nawrocki wrote:
> Hi Mauro,
>
> Please pull the following for 3.9, it includes Exynos SoC drivers cleanups and
> fixes. DMABUF exporting support for Exynos5 GScaler driver, device tree support
> for Exynos MFC driver (platform bits for it got merged already for v3.8).
>
> There is also included a patch removing deprecated image centering controls.
>
> The following changes since commit 8cd7085ff460ead3aba6174052a408f4ad52ac36:
>
>    [media] get_dvb_firmware: Fix the location of firmware for Terratec HTC
> (2013-01-01 11:18:26 -0200)
>
> are available in the git repository at:
>
>    git://git.infradead.org/users/kmpark/linux-samsung media_for_v3.9
>
> for you to fetch changes up to 36073ee2f7b3b5ae91900cb992b292404614243b:
>
>    V4L: Remove deprecated image centering controls (2013-01-04 11:35:43 +0100)
>
> ----------------------------------------------------------------
> Arun Kumar K (2):
>        s5p-mfc: Add device tree support
>        s5p-mfc: Flush DPB buffers during stream off
>
> Kamil Debski (4):
>        s5p-mfc: Move firmware allocation point to avoid allocation problems
>        s5p-mfc: Correct check of vb2_dma_contig_init_ctx return value
>        s5p-mfc: Change internal buffer allocation from vb2 ops to dma_alloc_coherent
>        s5p-mfc: Context handling in open() bugfix
>
> Sachin Kamat (9):
>        s5p-tv: Add missing braces around sizeof in sdo_drv.c
>        s5p-tv: Add missing braces around sizeof in mixer_video.c
>        s5p-tv: Add missing braces around sizeof in mixer_reg.c
>        s5p-tv: Add missing braces around sizeof in mixer_drv.c
>        s5p-tv: Add missing braces around sizeof in hdmiphy_drv.c
>        s5p-tv: Add missing braces around sizeof in hdmi_drv.c
>        s5p-mfc: Remove redundant 'break'
>        s5p-mfc: Fix a typo in error message in s5p_mfc_pm.c
>        s5p-mfc: Fix an error check
>
> Shaik Ameer Basha (1):
>        exynos-gsc: Support dmabuf export buffer
>
> Sylwester Nawrocki (5):
>        s5p-fimc: Avoid possible NULL pointer dereference in set_fmt op
>        s5p-fimc: Prevent potential buffer overflow
>        s5p-fimc: Prevent AB-BA deadlock during links reconfiguration
>        s5p-tv: Fix return value in sdo_probe() on error paths
>        V4L: Remove deprecated image centering controls
>
> Tomasz Stanislawski (1):
>        s5p-tv: mixer: fix handling of VIDIOC_S_FMT
>
> Tony Prisk (3):
>        s5p-fimc: Fix incorrect usage of IS_ERR_OR_NULL
>        s5p-tv: Fix incorrect usage of IS_ERR_OR_NULL
>        s5p-g2d: Fix incorrect usage of IS_ERR_OR_NULL
>
> Wei Yongjun (1):
>        s5p-mfc: remove unused variable

Related patchwork commands:

pwclient update -s 'accepted' 15333
pwclient update -s 'accepted' 15565
pwclient update -s 'accepted' 16071
pwclient update -s 'accepted' 16072
pwclient update -s 'accepted' 16073
pwclient update -s 'accepted' 15657
pwclient update -s 'accepted' 15656
pwclient update -s 'accepted' 15658
pwclient update -s 'accepted' 15659
pwclient update -s 'accepted' 15660
pwclient update -s 'accepted' 15661
pwclient update -s 'accepted' 16013
pwclient update -s 'superseded' 16059
pwclient update -s 'accepted' 16060
pwclient update -s 'accepted' 16080
pwclient update -s 'accepted' 16081
pwclient update -s 'accepted' 16084
pwclient update -s 'accepted' 15647
pwclient update -s 'superseded' 16083
pwclient update -s 'accepted' 15765

Re: [GIT PULL FOR 3.9] Exynos SoC media drivers updates

[Mauro Carvalho Chehab]

Em Fri, 04 Jan 2013 23:39:12 +0100
Sylwester Nawrocki <sylvester.nawrocki@gmail.com> escreveu:


> > Tomasz Stanislawski (1):
> >        s5p-tv: mixer: fix handling of VIDIOC_S_FMT

I'll drop this one for now. Devin raised a point: such changes would break
existing applications.

So, we'll need to revisit this topic before changing the drivers.

Btw, I failed to find the corresponding patch at patchwork:
	http://patchwork.linuxtv.org/project/linux-media/list/?state=*&q=VIDIOC_S_FMT

So, its status update may be wrong after flushing your pwclient commands.

Regards,
Mauro

Re: [GIT PULL FOR 3.9] Exynos SoC media drivers updates

[Mauro Carvalho Chehab]

Em Fri, 04 Jan 2013 23:39:12 +0100
Sylwester Nawrocki <sylvester.nawrocki@gmail.com> escreveu:

> On 01/04/2013 08:01 PM, Sylwester Nawrocki wrote:
> > Hi Mauro,
> >
> > Please pull the following for 3.9, it includes Exynos SoC drivers cleanups and
> > fixes. DMABUF exporting support for Exynos5 GScaler driver, device tree support
> > for Exynos MFC driver (platform bits for it got merged already for v3.8).
> >
> > There is also included a patch removing deprecated image centering controls.
> >
> > The following changes since commit 8cd7085ff460ead3aba6174052a408f4ad52ac36:
> >
> >    [media] get_dvb_firmware: Fix the location of firmware for Terratec HTC
> > (2013-01-01 11:18:26 -0200)
> >
> > are available in the git repository at:
> >
> >    git://git.infradead.org/users/kmpark/linux-samsung media_for_v3.9
> >
> > for you to fetch changes up to 36073ee2f7b3b5ae91900cb992b292404614243b:
> >
> >    V4L: Remove deprecated image centering controls (2013-01-04 11:35:43 +0100)
> >
> > ----------------------------------------------------------------
> > Arun Kumar K (2):
> >        s5p-mfc: Add device tree support
> >        s5p-mfc: Flush DPB buffers during stream off
> >
> > Kamil Debski (4):
> >        s5p-mfc: Move firmware allocation point to avoid allocation problems
> >        s5p-mfc: Correct check of vb2_dma_contig_init_ctx return value
> >        s5p-mfc: Change internal buffer allocation from vb2 ops to dma_alloc_coherent
> >        s5p-mfc: Context handling in open() bugfix
> >
> > Sachin Kamat (9):
> >        s5p-tv: Add missing braces around sizeof in sdo_drv.c
> >        s5p-tv: Add missing braces around sizeof in mixer_video.c
> >        s5p-tv: Add missing braces around sizeof in mixer_reg.c
> >        s5p-tv: Add missing braces around sizeof in mixer_drv.c
> >        s5p-tv: Add missing braces around sizeof in hdmiphy_drv.c
> >        s5p-tv: Add missing braces around sizeof in hdmi_drv.c
> >        s5p-mfc: Remove redundant 'break'
> >        s5p-mfc: Fix a typo in error message in s5p_mfc_pm.c
> >        s5p-mfc: Fix an error check
> >
> > Shaik Ameer Basha (1):
> >        exynos-gsc: Support dmabuf export buffer
> >
> > Sylwester Nawrocki (5):
> >        s5p-fimc: Avoid possible NULL pointer dereference in set_fmt op
> >        s5p-fimc: Prevent potential buffer overflow
> >        s5p-fimc: Prevent AB-BA deadlock during links reconfiguration
> >        s5p-tv: Fix return value in sdo_probe() on error paths
> >        V4L: Remove deprecated image centering controls
> >
> > Tomasz Stanislawski (1):
> >        s5p-tv: mixer: fix handling of VIDIOC_S_FMT
> >
> > Tony Prisk (3):
> >        s5p-fimc: Fix incorrect usage of IS_ERR_OR_NULL
> >        s5p-tv: Fix incorrect usage of IS_ERR_OR_NULL
> >        s5p-g2d: Fix incorrect usage of IS_ERR_OR_NULL
> >
> > Wei Yongjun (1):
> >        s5p-mfc: remove unused variable
> 
> Related patchwork commands:
> 
> pwclient update -s 'accepted' 15333
> pwclient update -s 'accepted' 15565
> pwclient update -s 'accepted' 16071
> pwclient update -s 'accepted' 16072
> pwclient update -s 'accepted' 16073
> pwclient update -s 'accepted' 15657
> pwclient update -s 'accepted' 15656
> pwclient update -s 'accepted' 15658
> pwclient update -s 'accepted' 15659
> pwclient update -s 'accepted' 15660
> pwclient update -s 'accepted' 15661
> pwclient update -s 'accepted' 16013
> pwclient update -s 'superseded' 16059
> pwclient update -s 'accepted' 16060
> pwclient update -s 'accepted' 16080
> pwclient update -s 'accepted' 16081
> pwclient update -s 'accepted' 16084
> pwclient update -s 'accepted' 15647
> pwclient update -s 'superseded' 16083
> pwclient update -s 'accepted' 15765

Those status updates were missing:

pwclient update -s 'superseded' 14608
pwclient update -s 'superseded' 15188
pwclient update -s 'accepted' 16058
pwclient update -s 'accepted' 16108

> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-media" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html


-- 

Cheers,
Mauro

Re: [GIT PULL FOR 3.9] Exynos SoC media drivers updates

[Sylwester Nawrocki]

On 01/06/2013 12:32 PM, Mauro Carvalho Chehab wrote:
> Em Fri, 04 Jan 2013 23:39:12 +0100
> Sylwester Nawrocki<sylvester.nawrocki@gmail.com>  escreveu:
>
>
>>> Tomasz Stanislawski (1):
>>>         s5p-tv: mixer: fix handling of VIDIOC_S_FMT
>
> I'll drop this one for now. Devin raised a point: such changes would break
> existing applications.
>
> So, we'll need to revisit this topic before changing the drivers.
>
> Btw, I failed to find the corresponding patch at patchwork:
> 	http://patchwork.linuxtv.org/project/linux-media/list/?state=*&q=VIDIOC_S_FMT
>
> So, its status update may be wrong after flushing your pwclient commands.

Hmm, I got this patch from Tomasz by e-mail and added it to the pull 
request.
I think it wasn't sent to the mailing list, but I noticed it only after
sending you the pull requests, when was preparing the pwclient commands.
I've just posted it now, sorry. The link is here:
http://patchwork.linuxtv.org/patch/16143

Tomasz created this patch specifically for the purpose of format negotiation
in video pipeline in the application we used to test various scenarios with
DMABUF. I agree this patch has a potential of breaking buggy user space
applications. I can't see other solution for it right now, there seems even
to be no possibility to return some flag in VIDIOC_S_FMT indicating that
format has been modified and is valid, when -EINVAL was returned. This 
sounds
ugly anyway, but could ensure backward compatibility for applications that
exppect EINVAL when format has been changed. BTW, I wonder if it is only 
fourcc,
or other format parameters as well - like width, height, some applications
expect to get EINVAL when those have changed.


Regards,
Sylwester

Re: [GIT PULL FOR 3.9] Exynos SoC media drivers updates

[Mauro Carvalho Chehab]

Em Sun, 06 Jan 2013 13:34:53 +0100
Sylwester Nawrocki <sylvester.nawrocki@gmail.com> escreveu:

> On 01/06/2013 12:32 PM, Mauro Carvalho Chehab wrote:
> > Em Fri, 04 Jan 2013 23:39:12 +0100
> > Sylwester Nawrocki<sylvester.nawrocki@gmail.com>  escreveu:
> >
> >
> >>> Tomasz Stanislawski (1):
> >>>         s5p-tv: mixer: fix handling of VIDIOC_S_FMT
> >
> > I'll drop this one for now. Devin raised a point: such changes would break
> > existing applications.
> >
> > So, we'll need to revisit this topic before changing the drivers.
> >
> > Btw, I failed to find the corresponding patch at patchwork:
> > 	http://patchwork.linuxtv.org/project/linux-media/list/?state=*&q=VIDIOC_S_FMT
> >
> > So, its status update may be wrong after flushing your pwclient commands.
> 
> Hmm, I got this patch from Tomasz by e-mail and added it to the pull 
> request.
> I think it wasn't sent to the mailing list, but I noticed it only after
> sending you the pull requests, when was preparing the pwclient commands.
> I've just posted it now, sorry. The link is here:
> http://patchwork.linuxtv.org/patch/16143
> 
> Tomasz created this patch specifically for the purpose of format negotiation
> in video pipeline in the application we used to test various scenarios with
> DMABUF. I agree this patch has a potential of breaking buggy user space
> applications. I can't see other solution for it right now, there seems even
> to be no possibility to return some flag in VIDIOC_S_FMT indicating that
> format has been modified and is valid, when -EINVAL was returned. This 
> sounds
> ugly anyway, but could ensure backward compatibility for applications that
> exppect EINVAL when format has been changed. BTW, I wonder if it is only 
> fourcc,
> or other format parameters as well - like width, height, some applications
> expect to get EINVAL when those have changed.

The patch makes the driver compliant to v4l-compilance, as its behavior asks
for such change, after some discussions we had this year in San Diego. At that
time, we all believed that such change were safe.

However, we can't do it like proposed there (and on other patches from Hans).

The fact is that tvtime and mythtv applications (maybe more) will fail 
if the returned format is different than the requested ones, as they 
don't check for the returned value.

As no regressions on userspace are allowed, we need to re-discuss this issue.

While this doesn't happen, I'll postpone such patches.

Comments/suggestions are welcome.

Regards,
Mauro

Re: [GIT PULL FOR 3.9] Exynos SoC media drivers updates

[Hans Verkuil]

On Sun January 6 2013 13:41:57 Mauro Carvalho Chehab wrote:
> Em Sun, 06 Jan 2013 13:34:53 +0100
> Sylwester Nawrocki <sylvester.nawrocki@gmail.com> escreveu:
> 
> > On 01/06/2013 12:32 PM, Mauro Carvalho Chehab wrote:
> > > Em Fri, 04 Jan 2013 23:39:12 +0100
> > > Sylwester Nawrocki<sylvester.nawrocki@gmail.com>  escreveu:
> > >
> > >
> > >>> Tomasz Stanislawski (1):
> > >>>         s5p-tv: mixer: fix handling of VIDIOC_S_FMT
> > >
> > > I'll drop this one for now. Devin raised a point: such changes would break
> > > existing applications.
> > >
> > > So, we'll need to revisit this topic before changing the drivers.
> > >
> > > Btw, I failed to find the corresponding patch at patchwork:
> > > 	http://patchwork.linuxtv.org/project/linux-media/list/?state=*&q=VIDIOC_S_FMT
> > >
> > > So, its status update may be wrong after flushing your pwclient commands.
> > 
> > Hmm, I got this patch from Tomasz by e-mail and added it to the pull 
> > request.
> > I think it wasn't sent to the mailing list, but I noticed it only after
> > sending you the pull requests, when was preparing the pwclient commands.
> > I've just posted it now, sorry. The link is here:
> > http://patchwork.linuxtv.org/patch/16143
> > 
> > Tomasz created this patch specifically for the purpose of format negotiation
> > in video pipeline in the application we used to test various scenarios with
> > DMABUF. I agree this patch has a potential of breaking buggy user space
> > applications. I can't see other solution for it right now, there seems even
> > to be no possibility to return some flag in VIDIOC_S_FMT indicating that
> > format has been modified and is valid, when -EINVAL was returned. This 
> > sounds
> > ugly anyway, but could ensure backward compatibility for applications that
> > exppect EINVAL when format has been changed. BTW, I wonder if it is only 
> > fourcc,
> > or other format parameters as well - like width, height, some applications
> > expect to get EINVAL when those have changed.
> 
> The patch makes the driver compliant to v4l-compilance, as its behavior asks
> for such change, after some discussions we had this year in San Diego. At that
> time, we all believed that such change were safe.
> 
> However, we can't do it like proposed there (and on other patches from Hans).
> 
> The fact is that tvtime and mythtv applications (maybe more) will fail 
> if the returned format is different than the requested ones, as they 
> don't check for the returned value.
> 
> As no regressions on userspace are allowed, we need to re-discuss this issue.
> 
> While this doesn't happen, I'll postpone such patches.

This is a video output device. So this patch will never affect tvtime/mythtv/etc.
I have no problem with this change being merged.

Regards,

	Hans

Re: [GIT PULL FOR 3.9] Exynos SoC media drivers updates

[Sylwester Nawrocki]

On 01/06/2013 01:05 PM, Mauro Carvalho Chehab wrote:
>> Related patchwork commands:
>>
>> pwclient update -s 'accepted' 15333
>> pwclient update -s 'accepted' 15565
>> pwclient update -s 'accepted' 16071
>> pwclient update -s 'accepted' 16072
>> pwclient update -s 'accepted' 16073
>> pwclient update -s 'accepted' 15657
>> pwclient update -s 'accepted' 15656
>> pwclient update -s 'accepted' 15658
>> pwclient update -s 'accepted' 15659
>> pwclient update -s 'accepted' 15660
>> pwclient update -s 'accepted' 15661
>> pwclient update -s 'accepted' 16013
>> pwclient update -s 'superseded' 16059
>> pwclient update -s 'accepted' 16060
>> pwclient update -s 'accepted' 16080
>> pwclient update -s 'accepted' 16081
>> pwclient update -s 'accepted' 16084
>> pwclient update -s 'accepted' 15647
>> pwclient update -s 'superseded' 16083
>> pwclient update -s 'accepted' 15765
>
> Those status updates were missing:
>
> pwclient update -s 'superseded' 14608
> pwclient update -s 'superseded' 15188
> pwclient update -s 'accepted' 16058

OK, sorry. Just learning to use the tools, will try to do better
next time. :)

> pwclient update -s 'accepted' 16108

And that's the pull request itself, I've obviously missed it.

--

Thanks,
Sylwester

Re: [GIT PULL FOR 3.9] Exynos SoC media drivers updates

[Mauro Carvalho Chehab]

Em Sun, 6 Jan 2013 13:53:52 +0100
Hans Verkuil <hverkuil@xs4all.nl> escreveu:

> On Sun January 6 2013 13:41:57 Mauro Carvalho Chehab wrote:
> > Em Sun, 06 Jan 2013 13:34:53 +0100
> > Sylwester Nawrocki <sylvester.nawrocki@gmail.com> escreveu:
> > 
> > > On 01/06/2013 12:32 PM, Mauro Carvalho Chehab wrote:
> > > > Em Fri, 04 Jan 2013 23:39:12 +0100
> > > > Sylwester Nawrocki<sylvester.nawrocki@gmail.com>  escreveu:
> > > >
> > > >
> > > >>> Tomasz Stanislawski (1):
> > > >>>         s5p-tv: mixer: fix handling of VIDIOC_S_FMT
> > > >
> > > > I'll drop this one for now. Devin raised a point: such changes would break
> > > > existing applications.
> > > >
> > > > So, we'll need to revisit this topic before changing the drivers.
> > > >
> > > > Btw, I failed to find the corresponding patch at patchwork:
> > > > 	http://patchwork.linuxtv.org/project/linux-media/list/?state=*&q=VIDIOC_S_FMT
> > > >
> > > > So, its status update may be wrong after flushing your pwclient commands.
> > > 
> > > Hmm, I got this patch from Tomasz by e-mail and added it to the pull 
> > > request.
> > > I think it wasn't sent to the mailing list, but I noticed it only after
> > > sending you the pull requests, when was preparing the pwclient commands.
> > > I've just posted it now, sorry. The link is here:
> > > http://patchwork.linuxtv.org/patch/16143
> > > 
> > > Tomasz created this patch specifically for the purpose of format negotiation
> > > in video pipeline in the application we used to test various scenarios with
> > > DMABUF. I agree this patch has a potential of breaking buggy user space
> > > applications. I can't see other solution for it right now, there seems even
> > > to be no possibility to return some flag in VIDIOC_S_FMT indicating that
> > > format has been modified and is valid, when -EINVAL was returned. This 
> > > sounds
> > > ugly anyway, but could ensure backward compatibility for applications that
> > > exppect EINVAL when format has been changed. BTW, I wonder if it is only 
> > > fourcc,
> > > or other format parameters as well - like width, height, some applications
> > > expect to get EINVAL when those have changed.
> > 
> > The patch makes the driver compliant to v4l-compilance, as its behavior asks
> > for such change, after some discussions we had this year in San Diego. At that
> > time, we all believed that such change were safe.
> > 
> > However, we can't do it like proposed there (and on other patches from Hans).
> > 
> > The fact is that tvtime and mythtv applications (maybe more) will fail 
> > if the returned format is different than the requested ones, as they 
> > don't check for the returned value.
> > 
> > As no regressions on userspace are allowed, we need to re-discuss this issue.
> > 
> > While this doesn't happen, I'll postpone such patches.
> 
> This is a video output device. So this patch will never affect tvtime/mythtv/etc.
> I have no problem with this change being merged.

Ok. still, we need to double-check what happens on video output apps. I'll keep
it on hold for a while, tagged as "under review" at Sylwester's umbrella.

Regards,
Mauro

Re: [GIT PULL FOR 3.9] Exynos SoC media drivers updates

[Sylwester Nawrocki]

On 01/06/2013 01:53 PM, Hans Verkuil wrote:
>>>>>> Tomasz Stanislawski (1):
>>>>>>          s5p-tv: mixer: fix handling of VIDIOC_S_FMT
>>>>
>>>> I'll drop this one for now. Devin raised a point: such changes would break
>>>> existing applications.
>>>>
>>>> So, we'll need to revisit this topic before changing the drivers.
>>>>
>>>> Btw, I failed to find the corresponding patch at patchwork:
>>>> 	http://patchwork.linuxtv.org/project/linux-media/list/?state=*&q=VIDIOC_S_FMT
>>>>
>>>> So, its status update may be wrong after flushing your pwclient commands.
>>>
>>> Hmm, I got this patch from Tomasz by e-mail and added it to the pull
>>> request.
>>> I think it wasn't sent to the mailing list, but I noticed it only after
>>> sending you the pull requests, when was preparing the pwclient commands.
>>> I've just posted it now, sorry. The link is here:
>>> http://patchwork.linuxtv.org/patch/16143
>>>
>>> Tomasz created this patch specifically for the purpose of format negotiation
>>> in video pipeline in the application we used to test various scenarios with
>>> DMABUF. I agree this patch has a potential of breaking buggy user space
>>> applications. I can't see other solution for it right now, there seems even
>>> to be no possibility to return some flag in VIDIOC_S_FMT indicating that
>>> format has been modified and is valid, when -EINVAL was returned. This
>>> sounds
>>> ugly anyway, but could ensure backward compatibility for applications that
>>> exppect EINVAL when format has been changed. BTW, I wonder if it is only
>>> fourcc,
>>> or other format parameters as well - like width, height, some applications
>>> expect to get EINVAL when those have changed.
>>
>> The patch makes the driver compliant to v4l-compilance, as its behavior asks
>> for such change, after some discussions we had this year in San Diego. At that
>> time, we all believed that such change were safe.
>>
>> However, we can't do it like proposed there (and on other patches from Hans).
>>
>> The fact is that tvtime and mythtv applications (maybe more) will fail
>> if the returned format is different than the requested ones, as they
>> don't check for the returned value.
>>
>> As no regressions on userspace are allowed, we need to re-discuss this issue.
>>
>> While this doesn't happen, I'll postpone such patches.
>
> This is a video output device. So this patch will never affect tvtime/mythtv/etc.
> I have no problem with this change being merged.

TBH, I very much doubt anyone would complain in case of this driver. I'm not
certain if there is complete support for even one board in the mainline 
kernel,
likely only Origen A. AFAIK most applications use either Exynos DRM driver,
that has support for all features available in s5p-tv driver, or 
framebuffer
emulation on top of v4l2 output interface (there were in the past RFC 
patches
posted for vb2 adding FB emulation) is used. Although I agree with Mauro in
principle, I think chances of above patch causing any trouble to anyone are
close to zero.

--

Regards,
Sylwester

Re: [GIT PULL FOR 3.9] Exynos SoC media drivers updates

[Mauro Carvalho Chehab]

Em Sun, 06 Jan 2013 13:58:59 +0100
Sylwester Nawrocki <sylvester.nawrocki@gmail.com> escreveu:

> On 01/06/2013 01:05 PM, Mauro Carvalho Chehab wrote:
> >> Related patchwork commands:
> >>
> >> pwclient update -s 'accepted' 15333
> >> pwclient update -s 'accepted' 15565
> >> pwclient update -s 'accepted' 16071
> >> pwclient update -s 'accepted' 16072
> >> pwclient update -s 'accepted' 16073
> >> pwclient update -s 'accepted' 15657
> >> pwclient update -s 'accepted' 15656
> >> pwclient update -s 'accepted' 15658
> >> pwclient update -s 'accepted' 15659
> >> pwclient update -s 'accepted' 15660
> >> pwclient update -s 'accepted' 15661
> >> pwclient update -s 'accepted' 16013
> >> pwclient update -s 'superseded' 16059
> >> pwclient update -s 'accepted' 16060
> >> pwclient update -s 'accepted' 16080
> >> pwclient update -s 'accepted' 16081
> >> pwclient update -s 'accepted' 16084
> >> pwclient update -s 'accepted' 15647
> >> pwclient update -s 'superseded' 16083
> >> pwclient update -s 'accepted' 15765
> >
> > Those status updates were missing:
> >
> > pwclient update -s 'superseded' 14608
> > pwclient update -s 'superseded' 15188
> > pwclient update -s 'accepted' 16058
> 
> OK, sorry. Just learning to use the tools, will try to do better
> next time. :)

No problem.

> > pwclient update -s 'accepted' 16108
> 
> And that's the pull request itself, I've obviously missed it.

Yeah ;) Well, you shouldn't include it. Sorry to add it at the
list.

Regards,
Mauro

Re: [GIT PULL FOR 3.9] Exynos SoC media drivers updates

[Mauro Carvalho Chehab]

Em Sun, 06 Jan 2013 14:15:44 +0100
Sylwester Nawrocki <sylvester.nawrocki@gmail.com> escreveu:

> On 01/06/2013 01:53 PM, Hans Verkuil wrote:
> >>>>>> Tomasz Stanislawski (1):
> >>>>>>          s5p-tv: mixer: fix handling of VIDIOC_S_FMT
> >>>>
> >>>> I'll drop this one for now. Devin raised a point: such changes would break
> >>>> existing applications.
> >>>>
> >>>> So, we'll need to revisit this topic before changing the drivers.
> >>>>
> >>>> Btw, I failed to find the corresponding patch at patchwork:
> >>>> 	http://patchwork.linuxtv.org/project/linux-media/list/?state=*&q=VIDIOC_S_FMT
> >>>>
> >>>> So, its status update may be wrong after flushing your pwclient commands.
> >>>
> >>> Hmm, I got this patch from Tomasz by e-mail and added it to the pull
> >>> request.
> >>> I think it wasn't sent to the mailing list, but I noticed it only after
> >>> sending you the pull requests, when was preparing the pwclient commands.
> >>> I've just posted it now, sorry. The link is here:
> >>> http://patchwork.linuxtv.org/patch/16143
> >>>
> >>> Tomasz created this patch specifically for the purpose of format negotiation
> >>> in video pipeline in the application we used to test various scenarios with
> >>> DMABUF. I agree this patch has a potential of breaking buggy user space
> >>> applications. I can't see other solution for it right now, there seems even
> >>> to be no possibility to return some flag in VIDIOC_S_FMT indicating that
> >>> format has been modified and is valid, when -EINVAL was returned. This
> >>> sounds
> >>> ugly anyway, but could ensure backward compatibility for applications that
> >>> exppect EINVAL when format has been changed. BTW, I wonder if it is only
> >>> fourcc,
> >>> or other format parameters as well - like width, height, some applications
> >>> expect to get EINVAL when those have changed.
> >>
> >> The patch makes the driver compliant to v4l-compilance, as its behavior asks
> >> for such change, after some discussions we had this year in San Diego. At that
> >> time, we all believed that such change were safe.
> >>
> >> However, we can't do it like proposed there (and on other patches from Hans).
> >>
> >> The fact is that tvtime and mythtv applications (maybe more) will fail
> >> if the returned format is different than the requested ones, as they
> >> don't check for the returned value.
> >>
> >> As no regressions on userspace are allowed, we need to re-discuss this issue.
> >>
> >> While this doesn't happen, I'll postpone such patches.
> >
> > This is a video output device. So this patch will never affect tvtime/mythtv/etc.
> > I have no problem with this change being merged.
> 
> TBH, I very much doubt anyone would complain in case of this driver. I'm not
> certain if there is complete support for even one board in the mainline 
> kernel,
> likely only Origen A. AFAIK most applications use either Exynos DRM driver,
> that has support for all features available in s5p-tv driver, or 
> framebuffer
> emulation on top of v4l2 output interface (there were in the past RFC 
> patches
> posted for vb2 adding FB emulation) is used. Although I agree with Mauro in
> principle, I think chances of above patch causing any trouble to anyone are
> close to zero.

Yes, I see your point. Yet, it doesn't hurt to keep it on hold for a couple
weeks while we discuss it at the ML.

Regards,
Mauro