[mythtv-users] Firewire on a combo backend/frontend

[John P Poet]

On 7/31/06, Steven Adeff <adeffs.mythtv at gmail.com> wrote:
> On 7/31/06, John P Poet <jppoet at gmail.com> wrote:
> > Since capturing firewire video data on my backend machine resulted in
> > a constant stream of discontinuity errors, I turned my frontend
> > machine into a combo backend/frontend, and hooked the firewire up
> > there.
> >
> > On this frontend machine, I am able to capture pretty clean video via
> > firewire.  However, if I try to watch anything while it is recording
> > via firewire, the video is very jumpy, and I get a lot of "video is 30
> > frames behind audio" type messages.  This is on a Athlon X2 4200,
> > runing 2.6.17.7.
> >
> > I figured that it had been long enough since Jim Westfall submitted
> > his ieee1394 dma patch that it would be a part of 2.6.17, but I
> > discovered it was not.  After applying his patch:
> >
> > http://www.gossamer-threads.com/lists/mythtv/dev/187783#187783
> >
> > and building a new kernel, the video playback while recording problem
> > is now solved.  I did have to "hand" apply the patch because the
> > formatting of the dma.c file in 2.6.17 has changed slightly.  Applying
> > this patch caused the CPU usage of mythbackend to drop from 10-11% to
> > 3-4% while recording via firewire.
> >
> > Unfortunately, I now had an audio problem.  The sound would drop-out
> > for a fraction of a second, every couple of seconds.  This only
> > happened if I was recording via firewire at the same time.  I figured
> > my firewire port and my sound card must be sharing an interrupt, but a
> > quick look at the motherboard manual disproved that theory.  I then
> > check the latencies of those two devices (lspci) and discovered that
> > both were at 32.  After adjusting both up to 64 (setpci), my audio
> > problem is also solved.
> >
> > I still have a little problem with channel changes.  The channel
> > change program I am using works 100% of the time, when I am testing it
> > via the command line.  With Myth in control it only seems to be 75%
> > reliable.  This program changes channels on my SA3250HD via firewire.
> >
> > I figured that to get 100% reliable channel changes, I would switch to
> > lirc.  I have an ir-blaster and decided to hook it up along side my
> > ir-reciever.  Unfortunately, both are serial and therefore use the
> > lirc_serial driver.  It turns out the lirc_serial driver cannot handle
> > two serial ports at the same time.  Furthermore, you cannot run two
> > lirc_serial drivers at the same time.  To get around this, some people
> > have hacked the lirc source tree to create a lirc_serial1.  This is
> > actually much harder than it sounds, and is generally a royal PITA.  I
> > may try a little harder to get channel changes to work via firewire...
>
> Did you try Jim's patch on your dedicated backend to see if that
> solved the dropout issue? I'm still hesitant to roll my own kernel
> just for this patch, though now I wonder why its been so long with the
> linux1394 people ignoring it.
>
> --
> Steve

Yes, but it did not solve the discontinuity problems.  Pretty sure it
is a cabling issue to that room.

John