Re: Future of the SAA7146 drivers

[Soeren Moch <smoch@web.de>]

Hi Stefan, Hans,

On 02.02.23 00:12, Stefan Herdler wrote:
> Hi Hans, Sören,
>
> On 01/02/23, 10:15 Hans Verkuil wrote:
>
> [...]
>>
>> Once it is converted to vb2 the driver can stay.
>>
>> Note that the driver might need a bit more work: we use the
>> v4l2-compliance
>> utility to test V4L2 API compliance of a driver, and after the vb2
>> conversion the driver should pass this test. So the compliance test
>> might
>> find some other things that do not work as they should, and it would be
>> really good to clean that up as well. Usually the things it finds are
>> pretty
>> easy to fix.
>>
> For the records, as long I remember it:
> The "Buget Patch" driver is superfluous and can be removed.
> This driver is for an experimental hardware-mod which never really
> worked. No such hardware was ever produced.
> I was really surprised to see it.
I own such card, as I wrote earlier. The budget patch works great and is
necessary for such cards. Please keep it.

Regards,
Soeren
>>>
>>> There is a long lasting controversy about the deprecated 3
>>> DVB-API-files
>>> for the av7110 driver.
>>> Sören stated he is tired about that discussions and looking for
>>> permanent solution without having to rewrite the whole driver.
>>>
>>> In this case I'm just the messenger, but I would also prefer a
>>> permanent
>>> solution.
>>> That is why I made the suggestion with the driver specific UAPI file.
>>> The av7110 driver is only driver using this API-part.
>>> And the modification is fairly easy and the impact minimal.
>>>
>>> May that be a compromise all parties can live with?
>>
>> Moving it to a public av7110.h header makes sense to me.
>
> I've done this already.
> Primary for my self to see if and how it may work.
> I'll attach the patch, maybe it is helpful.
>
> Theoretically there shouldn't be a difference whether the headers are in
> 1 or 3 files.
> Unused symbols should be removed by the compiler while optimization the
> result should be identical.
> I was just curious and decided to try it.
>
> Unfortunately I couldn't test it yet. But hopefully next weekend.
> There seems to be an issue with new kernel and my Nvidia graphics-card.
> But the kernel builds and boots into text console.
>
>>
>> It's a nice-to-have in my view, and moving it to a driver-specific API
>> should avoid future discussions. The fact remains that as long as
>> people use this API we can't remove it, however much we would want to.
>>
>> I think the main problem has always been that we are surprised at how
>> many people still use these cards. It's very similar in that respect
>> to the bttv driver: very old, but still in use.
>
> I own one too.
> Such cards can still be handy for digitizing family videos and stuff.
>
>>
>> The difference appears to be that the use of saa7146 cards is confined
>> to a specific region (esp. Germany), whereas bttv is in use worldwide.
>
> Until recently I wasn't aware that there are no saa7146 cards for ATSC
> nor ISDB.
>
>>
>> Because of that we just miss that it is still in use.
>>
>> Having an active maintainer should help with that.
>>
>> Regards,
>>
>>     Hans
>
>
>
>
>
> On 01/02/23, 14:51 Hans Verkuil wrote:
>
> [...]
>>
>> It's the analog video streaming that uses vb2, so being able to test
>> that
>> is critical.
>>
> The DVB data always uses one specific of the 2 video ports. The other
> one is not able to handle DVB-data.
> It is always the path through BRS and FIFO3, if I remember correctly.
>
> The analog video data has to use the remaining video port and FIFO1 if a
> DVB-Tuner is present.
> Input data format is YUV and output RGB. At least at the FF cards.
>
> (I may have confused the Ports, but in general it should be correct.)
>
> I don't have a combined budget-card and don't know how the driver
> handles analog inputs. But because of this hardware limitation there
> shouldn't be much difference between the decoded video of the FF and the
> video-input of a combined DVB-budget.
> Maybe testing on a FF is sufficient?
>
>
>> So I decided to do this differently:
>>
>> 1) I'll revert the move of saa7146 to staging, it will go back to
>>     mainline. av7110 stays in staging for now (that might change, I
>>     just don't want to make more changes than strictly necessary).
>>
>> 2) I will do the vb2 conversion. I have the analog video hardware,
>>     so I can test this.
>>
>> I didn't want to spend time on that originally, but since these drivers
>> are still in use, it is probably best if I bite the bullet and just
>> do it.
>>
>> I'm now almost done with the vb2 conversion of cx18, and it was about
>> 2 days work, which isn't that bad. I'll try to get this saa7146 vb2
>> conversion done this month.
>>
>> The PR reverting this has just been posted:
>>
>> https://patchwork.linuxtv.org/project/linux-media/patch/5902a4f2-da31-816c-f3cf-020340dbaddf@xs4all.nl/
>>
>>
>
> Great news, thanks!
>
> Regards
>
> Stefan
>
>> Regards,
>>
>>     Hans