[mythtv-users] HVR-950Q analog
dheitmueller at kernellabs.com
Thu Mar 31 15:38:30 UTC 2011
On Wed, Mar 30, 2011 at 10:58 AM, Mark Lord <mythtv at rtr.ca> wrote:
> On 11-03-30 09:25 AM, Devin Heitmueller wrote:
>> I did the original driver work using 0.22 as my testbed (and was
>> comfortable that it was working pretty well). You aren't the first
>> person to say "I upgraded to 0.24 and now everything is unreliable".
>> There haven't been any driver changes to the 950q since I did the
>> original 0.22 work, so I suspect strongly that there was a regression
>> in MythTV - although I haven't had any time to install and debug it.
> I wonder if this could have anything to with V4L vs V4L2 interfaces
> or something like that? I have another analog tuner here (ConvertX),
> and I believe that Mythtv will use only V4L (not V4L2) interfaces
> to control it. Which is a problem, since the in-kernel-staging driver
> doesn't implement the old interfaces.
Feels doubtful. I think what is more likely is a race condition
switching between analog and digital modes. The 950Q is a rather
extreme case in that it has a single chip which provides both the
analog video decoder and the digital demodulator (this is unusual
compared to most other tuner designs). Hence, if the power management
isn't *exactly* correct (in both the driver and the application) in
terms of shutting down analog and starting up digital (or vice versa),
then you're entirely likely to put the chip into an unknown state.
Now if you're also seeing analog audio issues with a completely
unrelated design like the PCTV 800i, then that suggests it may have
*nothing* to do with the 950q's power management at all and is more
likely some general MythTV problem reading from an ALSA device
associated with a framegrabber.
I was pretty comfortable that this stuff was working in 0.22, since I
did take the time to debug the known race conditions and had the 950q
working properly. If something changed in 0.24, either it's a bug in
MythTV, or the changes to MythTV exposed some *different* race
condition in the 950q driver that we never encountered before.
Somebody's got to find the time to get an environment setup (building
both MythTV and the 950q driver from source), and better understand
the actual failure. Unfortunately, I don't have the cycles to do it
as a result of my day job.
Devin J. Heitmueller - Kernel Labs
More information about the mythtv-users