[mythtv-users] analog side of hvr-1600 and hvr-2250

[Devin Heitmueller]

On Wed, Dec 4, 2013 at 4:56 PM, R. G. Newbury <newbury at mandamus.org> wrote:
> Mike, I would be willing to try any udev rules you can come up with. I
> pinged Devin over a year ago about this problem but no fix has yet been
> found. We discussed this and when I sent him a copy of dmesg, I got back:
>
> On 07/19/2012 10:30 PM, Devin Heitmueller wrote:
>> Wow, ok.  That is *not* what I was expecting to see in your dmesg output.

Yeah, I remember this.

> Best guess: modules are being loaded before the device has stabilized.
> Should NOT happen, but it does for me,
> Most intriguing of course, is that just about no-one else sees this.
> Does Ubuntu have a udev script for these modules? Fedora does not.
> Wild ass-guess: Devin, is it possible that somehow systemd is behind this by
> starting everything in parallel?  I did not think that systemd had anything
> to do with the first passes of hardware initialization, but I know that it
> does impinge on some things (network mainlu AIUI).

It could be something such as udev or systemd trying to access the
tuner after the device nodes are created but before it's fully
initialized.  There is a history of issues where "v4l-id" gets invoked
by udev would run "v4l-id", which would access the analog side of the
tuner while the digital side was still coming up, causing all sorts of
crazy inconsistent state.  I had to fix it in a number of drivers
(cx88, em28xx, au0828), so perhaps such an issue exists with saa7164
as well.

It feels like that sort of race condition.

Devin

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