Philips saa7134 IR remote problem with linux kernel v2.6.35

Philips saa7134 IR remote problem with linux kernel v2.6.35

[Partha Guha Roy]

Hi,

I have a saa7134 analog tv card (Avermedia PCI pure m135a) with an IR
remote. The IR remote is recognized by a standard keyboard and lirc
used to work fine with this. However, from kernel v2.6.35, the IR
remote does not work properly. The major problem is that every
keystroke is registered after the next keystroke. So, if I press the
sequence "123" on the remote, it actually comes up with only "12". If
I then if I wait for 5 seconds, the "3" gets lost.

Now I tried to bisect the kernel and it lead to the following commit:

commit e40b1127f994a427568319d1be9b9e5ab1f58dd1
Author: David Härdeman <david@hardeman.nu>
Date:   Thu Apr 15 18:46:00 2010 -0300

    V4L/DVB: ir-core: change duration to be coded as a u32 integer

    This patch implements the agreed upon 1:31 integer encoded pulse/duration
    struct for ir-core raw decoders. All decoders have been tested after the
    change. Comments are welcome.

    Signed-off-by: David Härdeman <david@hardeman.nu>
    Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>

I am willing to test patches if needed.

Thanks and regards.

/Partha Roy

Re: Philips saa7134 IR remote problem with linux kernel v2.6.35

[Sean Young]

On Wed, Aug 08, 2012 at 12:23:46AM +0600, Partha Guha Roy wrote:
> I have a saa7134 analog tv card (Avermedia PCI pure m135a) with an IR
> remote. The IR remote is recognized by a standard keyboard and lirc
> used to work fine with this. However, from kernel v2.6.35, the IR
> remote does not work properly. The major problem is that every
> keystroke is registered after the next keystroke. So, if I press the
> sequence "123" on the remote, it actually comes up with only "12". If
> I then if I wait for 5 seconds, the "3" gets lost.
> 
> Now I tried to bisect the kernel and it lead to the following commit:
> 
> commit e40b1127f994a427568319d1be9b9e5ab1f58dd1
> Author: David Härdeman <david@hardeman.nu>
> Date:   Thu Apr 15 18:46:00 2010 -0300
> 
>     V4L/DVB: ir-core: change duration to be coded as a u32 integer
> 
>     This patch implements the agreed upon 1:31 integer encoded pulse/duration
>     struct for ir-core raw decoders. All decoders have been tested after the
>     change. Comments are welcome.
> 
>     Signed-off-by: David Härdeman <david@hardeman.nu>
>     Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
> 
> I am willing to test patches if needed.

Are you runnning the lircd user space process for input or relying on
the in-kernel decoders? Also what remote are you using (or more 
specifically, what IR protocol does it use)?

Can you reproduce the issue on a more contemporary kernel?

Note that the commit only affects kernel space IR decoders so it should
not affect lircd.

I wouldn't be surprised if the 15ms delay for processing in 
saa7134_raw_decode_irq (bottom of saa7134-input.c) needs increasing.


Sean

Re: Philips saa7134 IR remote problem with linux kernel v2.6.35

[Partha Guha Roy]

Hi Sean,

I am not an expert on the kernel. So please excuse me if I give the
wrong answer somewhere.

On Fri, Aug 10, 2012 at 3:47 PM, Sean Young <sean@mess.org> wrote:
>
> Are you runnning the lircd user space process for input or relying on
> the in-kernel decoders?

For my testing, I booted the vanilla kernel into ubuntu recovery mode
and just pressed a few keys on the remote. No lircd process was
running at that point. So, I am guessing that I used the in-kernel
decoders.

> Also what remote are you using (or more
> specifically, what IR protocol does it use)?
>

The remove came with the analog TV card (avermedia pci pure m135a). I
am not sure what protocol the remote uses. I'd really appreciate it if
you could let me know how I can find that out.

> Can you reproduce the issue on a more contemporary kernel?
>

Yes. The buggy behavior is present in Ubuntu 12.04 (IIRC, kernel
v3.2.*). I also know that the buggy behavior is present at v3.4.x of
the kernel. I haven't tested more recent kernels.

> Note that the commit only affects kernel space IR decoders so it should
> not affect lircd.
>

Ok. But as I mentioned, I think I am using kernel space decoders.

> I wouldn't be surprised if the 15ms delay for processing in
> saa7134_raw_decode_irq (bottom of saa7134-input.c) needs increasing.
>

Thank you very much for your feedback.

Regards.

/Partha Roy

Re: Philips saa7134 IR remote problem with linux kernel v2.6.35

[Sean Young]

On Fri, Aug 10, 2012 at 11:41:23PM +0600, Partha Guha Roy wrote:
> On Fri, Aug 10, 2012 at 3:47 PM, Sean Young <sean@mess.org> wrote:
> >
> > Are you runnning the lircd user space process for input or relying on
> > the in-kernel decoders?
> 
> For my testing, I booted the vanilla kernel into ubuntu recovery mode
> and just pressed a few keys on the remote. No lircd process was
> running at that point. So, I am guessing that I used the in-kernel
> decoders.

That sounds right.

> > Also what remote are you using (or more
> > specifically, what IR protocol does it use)?
> >
> 
> The remove came with the analog TV card (avermedia pci pure m135a). I
> am not sure what protocol the remote uses. I'd really appreciate it if
> you could let me know how I can find that out.

That's a nec remote. Using ir-keytable (in the ir-keyable package on
Ubuntu) you change the protocol and see what's being sent.

However I guess the problem is that IR edges aren't being reported until
triggered by more IR edges.

> > Can you reproduce the issue on a more contemporary kernel?
> >
> 
> Yes. The buggy behavior is present in Ubuntu 12.04 (IIRC, kernel
> v3.2.*). I also know that the buggy behavior is present at v3.4.x of
> the kernel. I haven't tested more recent kernels.

There is a lot code churn so it would be very helpful to have it 
reproduced on a recent kernel.

Unfortunately your bisect wasn't entirely useful. You ended up at
e40b1127f994a427568319d1be9b9e5ab1f58dd1. Unfortunately that commit
introduced a bug which was not entirely resolved until
9800b5b619cd9a013a6f0c7d5da0dbbc17a5af30. If you do a bisect 
again it would useful to skip anything between those commits.

Also is the issue intermittent or is your last key press *always*
missed? 


Sean

Philips saa7134 IR remote problem with linux kernel v2.6.35

[Partha Guha Roy]

Hi,

I have a saa7134 analog tv card (Avermedia PCI pure m135a) with an IR
remote. The IR remote is recognized by a standard keyboard and lirc
used to work fine with this. However, from kernel v2.6.35, the IR
remote does not work properly. The major problem is that every
keystroke is registered after the next keystroke. So, if I press the
sequence "123" on the remote, it actually comes up with only "12". If
I then if I wait for 5 seconds, the "3" gets lost.

Now I tried to bisect the kernel and it lead to the following commit:

commit e40b1127f994a427568319d1be9b9e5ab1f58dd1
Author: David Härdeman <david@hardeman.nu>
Date:   Thu Apr 15 18:46:00 2010 -0300

    V4L/DVB: ir-core: change duration to be coded as a u32 integer

    This patch implements the agreed upon 1:31 integer encoded pulse/duration
    struct for ir-core raw decoders. All decoders have been tested after the
    change. Comments are welcome.

    Signed-off-by: David Härdeman <david@hardeman.nu>
    Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>

I am willing to test patches if needed.

Thanks and regards.

/Partha Roy