Remaining drivers that aren't V4L2?

Remaining drivers that aren't V4L2?

[Devin Heitmueller]

Hello,

I know some months ago, there was some discussion about a few drivers
which were stragglers and had not been converted from V4L to V4L2.

Do we have a current list of driver which still haven't been converted?

I started doing some more tvtime work last night, and I would *love*
to drop V4L support (and *only* support V4L2 devices), since it would
make the code much cleaner, more reliable, and easier to test.

If there are only a few obscure webcams remaining, then I'm willing to
tell those users that they have to stick with whatever old version of
tvtime they've been using since the last release four years ago.

Devin

-- 
Devin J. Heitmueller - Kernel Labs
http://www.kernellabs.com

Re: Remaining drivers that aren't V4L2?

[Hans Verkuil]

On Friday 12 March 2010 21:11:49 Devin Heitmueller wrote:
> Hello,
> 
> I know some months ago, there was some discussion about a few drivers
> which were stragglers and had not been converted from V4L to V4L2.
> 
> Do we have a current list of driver which still haven't been converted?

These drivers are still v4l1:

arv
bw-qcam
c-qcam
cpia_pp
cpia_usb
ov511
se401
stradis
stv680
usbvideo
w9966

Some of these have counterparts in gspca these days so possibly some drivers
can be removed by now. Hans, can you point those out?

arv, bw-qcam, c-qcam, cpia_pp and stradis can probably be moved to staging
and if no one steps up then they can be dropped altogether.

According to my notes I should be able to test cpia_usb. I would have to
verify that, though. I think it is only used in a USB microscope. It is
effectively a webcam. I can also test usbvideo (USB 1 TV capture device).
The latter is probably the most important driver that needs converting,
because I think these are not uncommon.

However, I have no time to work on such a driver conversion. But if someone
is seriously willing to put time and effort in that, then I am willing to
mail the hardware.

> I started doing some more tvtime work last night, and I would *love*
> to drop V4L support (and *only* support V4L2 devices), since it would
> make the code much cleaner, more reliable, and easier to test.
> 
> If there are only a few obscure webcams remaining, then I'm willing to
> tell those users that they have to stick with whatever old version of
> tvtime they've been using since the last release four years ago.

To my knowledge the usbvideo driver is probably the least obscure device
that is still using V4L1.

Regards,

	Hans

-- 
Hans Verkuil - video4linux developer - sponsored by TANDBERG

Re: Remaining drivers that aren't V4L2?

[Michael Akey]

Hans Verkuil wrote:
> On Friday 12 March 2010 21:11:49 Devin Heitmueller wrote:
>   
>> Hello,
>>
>> I know some months ago, there was some discussion about a few drivers
>> which were stragglers and had not been converted from V4L to V4L2.
>>
>> Do we have a current list of driver which still haven't been converted?
>>     
>
> These drivers are still v4l1:
>
> arv
> bw-qcam
> c-qcam
> cpia_pp
> cpia_usb
> ov511
> se401
> stradis
> stv680
> usbvideo
> w9966
>
> Some of these have counterparts in gspca these days so possibly some drivers
> can be removed by now. Hans, can you point those out?
>
> arv, bw-qcam, c-qcam, cpia_pp and stradis can probably be moved to staging
> and if no one steps up then they can be dropped altogether.
>   

Does this mean that the bw-qcam driver will be removed in future 
revisions or does this mean it will just never be updated to v4l2?

> According to my notes I should be able to test cpia_usb. I would have to
> verify that, though. I think it is only used in a USB microscope. It is
> effectively a webcam. I can also test usbvideo (USB 1 TV capture device).
> The latter is probably the most important driver that needs converting,
> because I think these are not uncommon.
>
> However, I have no time to work on such a driver conversion. But if someone
> is seriously willing to put time and effort in that, then I am willing to
> mail the hardware.
>
>   
>> I started doing some more tvtime work last night, and I would *love*
>> to drop V4L support (and *only* support V4L2 devices), since it would
>> make the code much cleaner, more reliable, and easier to test.
>>
>> If there are only a few obscure webcams remaining, then I'm willing to
>> tell those users that they have to stick with whatever old version of
>> tvtime they've been using since the last release four years ago.
>>     
>
> To my knowledge the usbvideo driver is probably the least obscure device
> that is still using V4L1.
>
> Regards,
>
> 	Hans
>
>   

Re: Remaining drivers that aren't V4L2?

[Devin Heitmueller]

On Fri, Mar 12, 2010 at 5:20 PM, Michael Akey <akeym@onid.orst.edu> wrote:
> Hans Verkuil wrote:
>> These drivers are still v4l1:
>>
>> arv
>> bw-qcam
>> c-qcam
>> cpia_pp
>> cpia_usb
>> ov511
>> se401
>> stradis
>> stv680
>> usbvideo
>> w9966
>>
>> Some of these have counterparts in gspca these days so possibly some
>> drivers
>> can be removed by now. Hans, can you point those out?
>>
>> arv, bw-qcam, c-qcam, cpia_pp and stradis can probably be moved to staging
>> and if no one steps up then they can be dropped altogether.
>>
>
> Does this mean that the bw-qcam driver will be removed in future revisions
> or does this mean it will just never be updated to v4l2?

Hans is suggesting that support for those devices would be dropped
entirely if no developer steps up to convert them to v4l2.

The problem is that supporting the long deprecated API is a burden
that makes it much harder to extend certain aspects of the internal
code.  If we were able to drop those devices, it would be much easier
to improve all the other drivers (of which the *vast* majority have
been converted to v4l2).

It's been over ten years since v4l2 came out.  If nobody has converted
those drivers to v4l2, then it's safe to say it's probably never going
to happen.

Devin

-- 
Devin J. Heitmueller - Kernel Labs
http://www.kernellabs.com

Re: Remaining drivers that aren't V4L2?

[Hans Verkuil]

On Friday 12 March 2010 23:20:44 Michael Akey wrote:
> Hans Verkuil wrote:
> > On Friday 12 March 2010 21:11:49 Devin Heitmueller wrote:
> >   
> >> Hello,
> >>
> >> I know some months ago, there was some discussion about a few drivers
> >> which were stragglers and had not been converted from V4L to V4L2.
> >>
> >> Do we have a current list of driver which still haven't been converted?
> >>     
> >
> > These drivers are still v4l1:
> >
> > arv
> > bw-qcam
> > c-qcam
> > cpia_pp
> > cpia_usb
> > ov511
> > se401
> > stradis
> > stv680
> > usbvideo
> > w9966
> >
> > Some of these have counterparts in gspca these days so possibly some drivers
> > can be removed by now. Hans, can you point those out?
> >
> > arv, bw-qcam, c-qcam, cpia_pp and stradis can probably be moved to staging
> > and if no one steps up then they can be dropped altogether.
> >   
> 
> Does this mean that the bw-qcam driver will be removed in future 
> revisions or does this mean it will just never be updated to v4l2?

Removal. At least, that is what I would propose. Greg KH has proposed some time
ago to use the staging tree not only for incoming but also for outgoing drivers.

And drivers that have not seen any development for years and that nobody seems
to be using and where the hardware is obsolete are definitely candidates for
this removal process.

I suspect that the only two drivers that we might need to keep are usbvideo and
cpia_usb. The latter is still used in hardware you can buy today, the same may
also be true of the first, and products supported by the usbvideo driver are
certainly still out there.

If we are indeed left with just two V4L1 drivers that cannot easily be removed,
then we should perhaps try to convert them after all, even though it is hard to
be motivated to do that work.

I definitely agree with Devin that it would be really great to finally remove
the V4L1 support completely from the kernel.

Regards,

	Hans

> 
> > According to my notes I should be able to test cpia_usb. I would have to
> > verify that, though. I think it is only used in a USB microscope. It is
> > effectively a webcam. I can also test usbvideo (USB 1 TV capture device).
> > The latter is probably the most important driver that needs converting,
> > because I think these are not uncommon.
> >
> > However, I have no time to work on such a driver conversion. But if someone
> > is seriously willing to put time and effort in that, then I am willing to
> > mail the hardware.
> >
> >   
> >> I started doing some more tvtime work last night, and I would *love*
> >> to drop V4L support (and *only* support V4L2 devices), since it would
> >> make the code much cleaner, more reliable, and easier to test.
> >>
> >> If there are only a few obscure webcams remaining, then I'm willing to
> >> tell those users that they have to stick with whatever old version of
> >> tvtime they've been using since the last release four years ago.
> >>     
> >
> > To my knowledge the usbvideo driver is probably the least obscure device
> > that is still using V4L1.
> >
> > Regards,
> >
> > 	Hans
> >
> >   
> 
> 

-- 
Hans Verkuil - video4linux developer - sponsored by TANDBERG

Re: Remaining drivers that aren't V4L2?

[Mauro Carvalho Chehab]

Hans Verkuil wrote:
> On Friday 12 March 2010 21:11:49 Devin Heitmueller wrote:
>> Hello,
>>
>> I know some months ago, there was some discussion about a few drivers
>> which were stragglers and had not been converted from V4L to V4L2.
>>
>> Do we have a current list of driver which still haven't been converted?
> 
> These drivers are still v4l1:
> 
> arv
> bw-qcam
> c-qcam
> cpia_pp
> cpia_usb
> ov511
> se401
> stradis
> stv680
> usbvideo
> w9966

All the above are webcam drivers. I doubt that those drivers would work
with tvtime: this software were meant to test the Vector's deinterlacing
algorithms, so it requires some specific video formats/resolutions found on TV
and require 25 or 30 fps, as far as I remember. For example, It doesn't support
QCIF/QVGA cameras.

If you want to extend tvtime to use webcams, some work is needed. Probably the easiest
way would be to use libv4l, that also does the V4L1 conversion, if needed. This may
actually make sense even for a few TV cards like em28xx, where you could use a bayer
format with a lower color depth and/or lower resolution, in order to allow viewing two 
simultaneous streams.

So, I suggest you to just drop V4L1 from tvtime and convert it to use libv4l (the conversion
is trivial: just replace open/close/ioctl from the V4L2 driver to the libv4l ones). This will
allow you to drop the old V4L1 driver from it, and, if you decide later to accept other
resolutions and make it more webcam friendly, you'll just need to allow tvtime to accept
other video resolutions and disable the de-interlacing setup if a webcam is detected.

-- 

Cheers,
Mauro

Re: Remaining drivers that aren't V4L2?

[Devin Heitmueller]

On Fri, Mar 12, 2010 at 5:52 PM, Mauro Carvalho Chehab
<mchehab@redhat.com> wrote:
> All the above are webcam drivers. I doubt that those drivers would work
> with tvtime: this software were meant to test the Vector's deinterlacing
> algorithms, so it requires some specific video formats/resolutions found on TV
> and require 25 or 30 fps, as far as I remember. For example, It doesn't support
> QCIF/QVGA cameras.

Yup, I was indeed aware that tvtime doesn't really work with webcams.
I wanted to see the list of remaining drivers, and now that I see the
list (and also came to the conclusion that they were all webcams), I
feel much more comfortable just dropping V4L1 support.

> If you want to extend tvtime to use webcams, some work is needed. Probably the easiest
> way would be to use libv4l, that also does the V4L1 conversion, if needed. This may
> actually make sense even for a few TV cards like em28xx, where you could use a bayer
> format with a lower color depth and/or lower resolution, in order to allow viewing two
> simultaneous streams.
>
> So, I suggest you to just drop V4L1 from tvtime and convert it to use libv4l (the conversion
> is trivial: just replace open/close/ioctl from the V4L2 driver to the libv4l ones). This will
> allow you to drop the old V4L1 driver from it, and, if you decide later to accept other
> resolutions and make it more webcam friendly, you'll just need to allow tvtime to accept
> other video resolutions and disable the de-interlacing setup if a webcam is detected.

I have actually been considering converting tvtime to using libv4l for
a while now, as I need it to support cards that use the HM12
pixelformat (such as the HVR-1600).  I wanted to rip out the V4L1
support first to make the conversion more straightforward.

It really isn't my goal to make tvtime support webcams, although they
might just start to work as an unintended side-effect.

Devin

-- 
Devin J. Heitmueller - Kernel Labs
http://www.kernellabs.com

Re: Remaining drivers that aren't V4L2?

[Mauro Carvalho Chehab]

Devin Heitmueller wrote:
> On Fri, Mar 12, 2010 at 5:52 PM, Mauro Carvalho Chehab
> <mchehab@redhat.com> wrote:

> Yup, I was indeed aware that tvtime doesn't really work with webcams.
> I wanted to see the list of remaining drivers, and now that I see the
> list (and also came to the conclusion that they were all webcams), I
> feel much more comfortable just dropping V4L1 support.

Yep. For its target, V4L1 is not needed anymore.

> I have actually been considering converting tvtime to using libv4l for
> a while now, as I need it to support cards that use the HM12
> pixelformat (such as the HVR-1600).  I wanted to rip out the V4L1
> support first to make the conversion more straightforward.
> 
> It really isn't my goal to make tvtime support webcams, although they
> might just start to work as an unintended side-effect.

If you take the right decisions with tvtime, you'll end to have it allowing
webcams, even unintentionally.

For example, on a quick brainstorming, I see some interesting features for
the tvtime todo list (just my $0,01 cents):

1) alsa support (undergoing/finished work?);

2) allow adjusting frame rate, to better support environments where the bus 
bandwidth is limited;

3) support for other video formats;

4) a generic control interface (a qv4l2 like interface);

5) support to adjust the resolution based on the screen size (as xawtv does);

6) allow tvtime to record programs;

7) direct access to IR event interface, in order to better work with IR's without
needing any extra software like lirc;

8) allow changing video standard without needing to restart tvtime. It is very common
on some Countries the usage of two or more simultaneous standard - For example, almost
all DVD/STB devices in Brazil outputs in NTSC, while TV is broadcasted in PAL-M. As far
as I know, the other Countries where Analog TV is still the main/only broadcast standard
have similar issues.

For sure the top priority is Alsa, but, at the end of the day, by handling (some) of the
above requirements, its usage with webcams will make more sense, and you won't need to
spend any time specifically devoted to webcam support.

Also, assuming that analog TV will end broadcasting some day in the world, the usage for
a tvtime-like application will likely be for video surveillance and other webcam usages.

So, in brief, I think a webcam support side-effect is a good thing.

-- 

Cheers,
Mauro

Re: Remaining drivers that aren't V4L2?

[Hans de Goede]

Hi,

On 03/12/2010 10:42 PM, Hans Verkuil wrote:
> On Friday 12 March 2010 21:11:49 Devin Heitmueller wrote:
>> Hello,
>>
>> I know some months ago, there was some discussion about a few drivers
>> which were stragglers and had not been converted from V4L to V4L2.
>>
>> Do we have a current list of driver which still haven't been converted?
>

As Hans Verkuil already said I've been working on steadily converting
v4l1 usb device drivers to gspca sub drivers, removing a lot of
redundant code and making them v4l2.

> These drivers are still v4l1:
>
> arv
> bw-qcam
> c-qcam
> cpia_pp

> cpia_usb

This one has a gspca subdriver now, and has been marked as
deprecated in 2.6.34, and scheduled for removal.

> ov511

This one has a gspca subdriver now, and has been marked as
deprecated in 2.6.34, and scheduled for removal.

> se401

Hans Verkuil gave me a camera with such a chip, I've been
meaning to work on this for a while. Note that the v4l1 driver
is completely broken (hanhs the machine) which complicates writing
a v4l2 driver, as I either need to first fix the v4l1 driver,
or just copy do a v4l2 driver based on the info in the v4l1 driver,
without having a driver to compare with.

> stradis

> stv680

This one has a gspca subdriver now, and has been marked as
deprecated in 2.6.34, and scheduled for removal.

> usbvideo

This actually is a framework for usb video devices a bit like
gspca one could say. It supports the following devices:

"USB 3com HomeConnect (aka vicam)"
"USB IBM (Xirlink) C-it Camera"
"USB Konica Webcam support"
"USB Logitech Quickcam Messenger"

Of which the Logitech Quickcam Messenger has a gspca subdriver
now, and is scheduled for removal.

> w9966
>
> Some of these have counterparts in gspca these days so possibly some drivers
> can be removed by now. Hans, can you point those out?
>

See above. Note in order to finish the conversion of drivers to
v4l1 besides time (which can be made every now and then) I really
really need hardware access, so if anyone has one of the usbvideo supported
devices lying around, and is willing to ship it to me, please drop
me a private mail!


> arv, bw-qcam, c-qcam, cpia_pp and stradis can probably be moved to staging
> and if no one steps up then they can be dropped altogether.
>

Ack!

> According to my notes I should be able to test cpia_usb. I would have to
> verify that, though. I think it is only used in a USB microscope. It is
> effectively a webcam. I can also test usbvideo (USB 1 TV capture device).
> The latter is probably the most important driver that needs converting,
> because I think these are not uncommon.
>

You gave your cpia1 camera to me, so now I have 2 to test with, 1 from
creative and the brandless one from you. Also note that this indeed is
used in microscopes, but also in regular webcams, Either way the cpia1
is supported in gspca now (for the usb version).

> However, I have no time to work on such a driver conversion. But if someone
> is seriously willing to put time and effort in that, then I am willing to
> mail the hardware.
>

You already did that, you gave me a cpia1, stv0680 and an se401 camera :)

>> I started doing some more tvtime work last night, and I would *love*
>> to drop V4L support (and *only* support V4L2 devices), since it would
>> make the code much cleaner, more reliable, and easier to test.
>>
>> If there are only a few obscure webcams remaining, then I'm willing to
>> tell those users that they have to stick with whatever old version of
>> tvtime they've been using since the last release four years ago.
>
> To my knowledge the usbvideo driver is probably the least obscure device
> that is still using V4L1.

I think you are confusing the usbvideo driver with the v4l2 usbvision
driver, which indeed gets used a lot in usb tv devices.

I think it is ok to drop v4l1 support from tvtime.

Regards,

Hans

Re: Remaining drivers that aren't V4L2?

[Hans Verkuil]

On Saturday 13 March 2010 07:33:57 Hans de Goede wrote:
> > To my knowledge the usbvideo driver is probably the least obscure device
> > that is still using V4L1.
> 
> I think you are confusing the usbvideo driver with the v4l2 usbvision
> driver, which indeed gets used a lot in usb tv devices.

You are correct. I confused those two. Sorry about that.

> 
> I think it is ok to drop v4l1 support from tvtime.

I agree.

Regards,

	Hans

-- 
Hans Verkuil - video4linux developer - sponsored by TANDBERG

Re: Remaining drivers that aren't V4L2?

[Devin Heitmueller]

On Sat, Mar 13, 2010 at 1:33 AM, Hans de Goede <hdegoede@redhat.com> wrote:
>> usbvideo
>
> This actually is a framework for usb video devices a bit like
> gspca one could say. It supports the following devices:
>
> "USB 3com HomeConnect (aka vicam)"
> "USB IBM (Xirlink) C-it Camera"
> "USB Konica Webcam support"
> "USB Logitech Quickcam Messenger"
>
> Of which the Logitech Quickcam Messenger has a gspca subdriver
> now, and is scheduled for removal.

Now that I see the product list, I realize that I actually have a 3com
HomeConnect kicking around in a box.  So if nobody gets around to it,
I could probably kill a few hours and do the conversion (given that
was a fairly popular product at the time).

Or would it be better to convert the products to gpsca (I don't
actually know/understand if that's possible at this point)?

Devin

-- 
Devin J. Heitmueller - Kernel Labs
http://www.kernellabs.com

Re: Remaining drivers that aren't V4L2?

[Hans de Goede]

Hi,

On 03/13/2010 03:23 PM, Devin Heitmueller wrote:
> On Sat, Mar 13, 2010 at 1:33 AM, Hans de Goede<hdegoede@redhat.com>  wrote:
>>> usbvideo
>>
>> This actually is a framework for usb video devices a bit like
>> gspca one could say. It supports the following devices:
>>
>> "USB 3com HomeConnect (aka vicam)"
>> "USB IBM (Xirlink) C-it Camera"
>> "USB Konica Webcam support"
>> "USB Logitech Quickcam Messenger"
>>
>> Of which the Logitech Quickcam Messenger has a gspca subdriver
>> now, and is scheduled for removal.
>
> Now that I see the product list, I realize that I actually have a 3com
> HomeConnect kicking around in a box.  So if nobody gets around to it,
> I could probably kill a few hours and do the conversion (given that
> was a fairly popular product at the time).
>
> Or would it be better to convert the products to gpsca (I don't
> actually know/understand if that's possible at this point)?
>

It would be much better to change it into a gspca subdriver, gspca is
a generic framework for usb webcams, and as such has a lot of code
which all these devices need shared in place, making the subdrivers
quite small, and nice to write as you can focus on the actual
camera specifics instead of on things like getting locking in case of
hot unplug while an app is streaming right.

Regards,

Hans