Hello List,
At some point between the version of v4l that is in kernel 2.6.16, and
the latest hg copy, the FM radio reception on my Hauppauge FM PCI card
stopped working. The last hg download that worked for me had this date
at the top of the changelog - Thu Apr 13 12:29:04 2006 -0400. the
download with
Fri Apr 28 10:14:45 2006 -0300 in the changelog it had stopped working -
it might have happened before that, this was just the next version I
downloaded.
The symptom is that the radio applications cannot tune into a signal,
they just get static. If I just use the version in the 2.6.16 kernel,
the radio applications tune into the signal and give good reception. So
its not a hardware/aerial issue - just the drivers.
I see the following displayed if I run gnomeradio (or radio) from the
command line:
VIDICGAUDI: Invalid argument
VIDICSAUDI: Invalid argument
I've included the dmesg loading the latest drivers below.
I have had a look at some of the code in the cx88 modules, but I can't
work out what might have changed to cause this.
If someone could point me at some specific stuff to try I'm happy to
experiment to fix it.
My thought so far is that this could be a v4l -v4l2 compatibility problem?
Thanks
Regards
Ian
ian2 ~ # uname -a
Linux ian2 2.6.16-gentoo-r6 #1 PREEMPT Wed May 3 21:04:22 BST 2006 i686
AMD Athlon(tm) XP 2800+ GNU/Linux
dmesg (using latest hg version - Fri May 5 09:53:05 2006 -0300):
May 6 13:17:51 ian2 kernel: Linux video capture interface: v1.00
May 6 13:17:51 ian2 kernel: cx2388x v4l2 driver version 0.0.5 loaded
May 6 13:17:51 ian2 kernel: ACPI: PCI Interrupt 0000:01:0a.0[A] -Link
[LNK1] -GSI 11 (level, low) -IRQ 11
May 6 13:17:51 ian2 kernel: CRE cx88[0]: subsystem: 0070:3401, board:
Hauppauge WinTV 34xxx models [card=1,autodetected]
May 6 13:17:51 ian2 kernel: TV tuner -1 at 0x1fe, Radio tuner -1 at 0x1fe
May 6 13:17:51 ian2 kernel: tda9887 2-0043: chip found @ 0x86 (cx88[0])
May 6 13:17:51 ian2 kernel: tveeprom 2-0050: Hauppauge model 34519, rev
J157, serial# 2906136
May 6 13:17:51 ian2 kernel: tveeprom 2-0050: tuner model is Philips
FM1216 ME MK3 (idx 57, type 38)
May 6 13:17:51 ian2 kernel: tveeprom 2-0050: TV standards PAL(B/G)
PAL(I) SECAM(L/L') PAL(D/D1/K) (eeprom 0x74)
May 6 13:17:51 ian2 kernel: tveeprom 2-0050: audio processor is CX881
(idx 31)
May 6 13:17:51 ian2 kernel: tveeprom 2-0050: has radio
May 6 13:17:51 ian2 kernel: cx88[0]: hauppauge eeprom: model=34519
May 6 13:17:51 ian2 kernel: input: cx88 IR (Hauppauge WinTV 34xxx as
/class/input/input6
May 6 13:17:51 ian2 kernel: cx88[0]/0: found at 0000:01:0a.0, rev: 5,
irq: 11, latency: 32, mmio: 0xe8000000
May 6 13:17:51 ian2 kernel: tuner 2-0061: chip found @ 0xc2 (cx88[0])
May 6 13:17:51 ian2 kernel: tuner 2-0061: type set to 38 (Philips
PAL/SECAM multi (FM1216ME MK3))
May 6 13:17:51 ian2 kernel: cx88[0]/0: registered device video0 [v4l2]
May 6 13:17:51 ian2 kernel: cx88[0]/0: registered device vbi0
May 6 13:17:51 ian2 kernel: cx88[0]/0: registered device radio0
May 6 13:17:51 ian2 kernel: cx2388x alsa driver version 0.0.5 loaded
May 6 13:17:51 ian2 kernel: ACPI: PCI Interrupt 0000:01:0a.1[A] -Link
[LNK1] -GSI 11 (level, low) -IRQ 11
May 6 13:17:52 ian2 kernel: cx88[0]/1: CX88x/0: ALSA support for
cx2388x boards

Hi Ian,

Am Samstag, den 06.05.2006, 13:42 +0100 schrieb Ian Pickworth:
Hello List,
At some point between the version of v4l that is in kernel 2.6.16, and
the latest hg copy, the FM radio reception on my Hauppauge FM PCI card
stopped working. The last hg download that worked for me had this date
at the top of the changelog - Thu Apr 13 12:29:04 2006 -0400. the
download with
Fri Apr 28 10:14:45 2006 -0300 in the changelog it had stopped working -
it might have happened before that, this was just the next version I
downloaded.

The symptom is that the radio applications cannot tune into a signal,
they just get static. If I just use the version in the 2.6.16 kernel,
the radio applications tune into the signal and give good reception. So
its not a hardware/aerial issue - just the drivers.

I see the following displayed if I run gnomeradio (or radio) from the
command line:

VIDICGAUDI: Invalid argument
VIDICSAUDI: Invalid argument

I've included the dmesg loading the latest drivers below.
I have had a look at some of the code in the cx88 modules, but I can't
work out what might have changed to cause this.
If someone could point me at some specific stuff to try I'm happy to
experiment to fix it.

My thought so far is that this could be a v4l -v4l2 compatibility problem?
Thanks
Regards
Ian

I can confirm your observations. All radio apps querying v4l1
capabilities fail, only kradio still works in v4l2 and issues only a
warning for v4l1.

Should be related to the removing/changing of v4l1 ioctls.

Cheers,
Hermann

Hi Hermann,

hermann pitton wrote:

I can confirm your observations. All radio apps querying v4l1
capabilities fail, only kradio still works in v4l2 and issues only a
warning for v4l1.

Should be related to the removing/changing of v4l1 ioctls.

Thanks for the tip - kradio is masked in Gentoo - I unmasked it and
installed, and I can listen to radio again :-).

So, xawtv (radio) and gnomeradio need updating to use v4l2. I think
xawtv is sort of related to this list, would one of its maintainers be
listening?

For gnomeradio - how does one go about getting a change request into the
Gnome project? Will they care?

Thanks
Regards
Ian

Am Samstag, den 06.05.2006, 22:45 +0100 schrieb Ian Pickworth:
Hi Hermann,

hermann pitton wrote:

I can confirm your observations. All radio apps querying v4l1
capabilities fail, only kradio still works in v4l2 and issues only a
warning for v4l1.

Should be related to the removing/changing of v4l1 ioctls.

Thanks for the tip - kradio is masked in Gentoo - I unmasked it and
installed, and I can listen to radio again :-).

So, xawtv (radio) and gnomeradio need updating to use v4l2. I think
xawtv is sort of related to this list, would one of its maintainers be
listening?

For gnomeradio - how does one go about getting a change request into the
Gnome project? Will they care?

Guess most likely we'll get "what about you?" for all this currently.

Greetings,
Hermann

Hi Ian,

Am Sonntag, den 07.05.2006, 00:57 +0200 schrieb hermann pitton:
Am Samstag, den 06.05.2006, 22:45 +0100 schrieb Ian Pickworth:
Hi Hermann,

hermann pitton wrote:

I can confirm your observations. All radio apps querying v4l1
capabilities fail, only kradio still works in v4l2 and issues only a
warning for v4l1.

Should be related to the removing/changing of v4l1 ioctls.

Thanks for the tip - kradio is masked in Gentoo - I unmasked it and
installed, and I can listen to radio again :-).

So, xawtv (radio) and gnomeradio need updating to use v4l2. I think
xawtv is sort of related to this list, would one of its maintainers be
listening?

For gnomeradio - how does one go about getting a change request into the
Gnome project? Will they care?

Guess most likely we'll get "what about you?" for all this currently.

after finding some time, as Mauro announced the v4l1-compat layer is
still usable, it just doesn't get configured and compiled with default
make and make install.

Changeset is here:
;style=gitweb

Default in videodev2.h is now:

/* Compatibility layer interface -- v4l1-compat module */
typedef int (*v4l2_kioctl)(struct inode *inode, struct file *file,
unsigned int cmd, void *arg);

#ifdef CNFIG_VIDEV4L1_CMPAT
int v4l_compat_translate_ioctl(struct inode *inode, struct file *file,
int cmd, void *arg, v4l2_kioctl driver_ioctl);
#else
#define v4l_compat_translate_ioctl(inode,file,cmd,arg,ioct l) -EINVAL
#endif

If someone needs the v4l1-compat, e.g. kradio didn't compile on all platforms,
simplest way to get it again is to call "make xconfig" or what else was used to build
the kernel from within v4l-dvb and keep it enabled, but the plan is to find out, which apps
have problems with pure v4l2. Radio seems clear, kradio is the only working app I know about ;)

Cheers,
Hermann

Am Mittwoch, den 24.05.2006, 18:05 -0300 schrieb Mauro Carvalho Chehab:
Hermann,
Em Ter, 2006-05-23 *s 20:02 +0200, hermann pitton escreveu:
Hi Ian,

after finding some time, as Mauro announced the v4l1-compat layer is
still usable, it just doesn't get configured and compiled with default
make and make install.

By default, v4l1-compat will be compiled, so older radio apps should
work as well. The advantage of having such options (for now) is to test
app compatibility with pure V4L2 devices.

Cheers,
Mauro.

Hi Mauro,

hmm, after revisiting it yesterday I noticed that it is not compiled and
installed with just make and make install on mercurial and Ian must have
had the same.

Calling "make xconfig" on FC3 and FC5 has it enabled as default then
in .config with the desired effect that the v4l1 ioctls don't return
EINVAL anymore.

It made me a little nervous for the upcoming kernels without looking
closer first, but now turns out it is/was only caused within mercurial
or if you say just make and make install should do, it must be something
I have locally. I'll try again tomorrow.

Cheers,
Hermann

Hermann,
Em Ter, 2006-05-23 20:02 +0200, hermann pitton escreveu:
Hi Ian,

after finding some time, as Mauro announced the v4l1-compat layer is
still usable, it just doesn't get configured and compiled with default
make and make install.

By default, v4l1-compat will be compiled, so older radio apps should
work as well. The advantage of having such options (for now) is to test
app compatibility with pure V4L2 devices.

Cheers,
Mauro.

Am Mittwoch, den 24.05.2006, 23:39 +0200 schrieb hermann pitton:
Am Mittwoch, den 24.05.2006, 18:05 -0300 schrieb Mauro Carvalho Chehab:
Hermann,
Em Ter, 2006-05-23 *s 20:02 +0200, hermann pitton escreveu:
Hi Ian,

after finding some time, as Mauro announced the v4l1-compat layer is
still usable, it just doesn't get configured and compiled with default
make and make install.

By default, v4l1-compat will be compiled, so older radio apps should
work as well. The advantage of having such options (for now) is to test
app compatibility with pure V4L2 devices.

Cheers,
Mauro.

Hi Mauro,

hmm, after revisiting it yesterday I noticed that it is not compiled and
installed with just make and make install on mercurial and Ian must have
had the same.

Calling "make xconfig" on FC3 and FC5 has it enabled as default then
in .config with the desired effect that the v4l1 ioctls don't return
EINVAL anymore.

It made me a little nervous for the upcoming kernels without looking
closer first, but now turns out it is/was only caused within mercurial
or if you say just make and make install should do, it must be something
I have locally. I'll try again tomorrow.

Mauro,

just a quick test. Without calling Kconfig at least once, v4l1-compat
stays disabled here on mercurial with just "make" and "make install".

Greetings,
Hermann

Am Donnerstag, den 25.05.2006, 00:48 +0200 schrieb hermann pitton:
Am Mittwoch, den 24.05.2006, 23:39 +0200 schrieb hermann pitton:
Am Mittwoch, den 24.05.2006, 18:05 -0300 schrieb Mauro Carvalho Chehab:
Hermann,
Em Ter, 2006-05-23 *s 20:02 +0200, hermann pitton escreveu:
[]

Mauro,

just a quick test. Without calling Kconfig at least once, v4l1-compat
stays disabled here on mercurial with just "make" and "make install".

Looks like this should be resolved on rc5.

Should some hint remain for those on prior stuff?
At least allmodconfig is very verbose about it

Sorry,
hermann in delay