MythTV and FireWire on Fedora 9? Plugreport reports nothing?

[jarod-ajLrJawYSntWk0Htik3J/w at public.gmane.org (Jarod Wilson)]

On Thu, 2009-01-29 at 16:18 -0500, Dan Ritter wrote:
> On Thu, Jan 29, 2009 at 04:06:24PM -0500, Derek Atkins wrote:
> > machine, but I expected the kernel probes to find it.   I'm happy
> > to find out that I'm just missing a driver, but the firewire
> > drivers do appear to be loaded:
> > 
> > firewire_ohci          30468  0 
> > firewire_core          45632  1 firewire_ohci
> > crc_itu_t              10240  1 firewire_core
> 
> I don't see:
> 
> ieee1394
> dv1394
> raw1394
> eth1394
> ohci1394

As well you shouldn't. Those *1394 modules are the old 'ieee1394' stack.
Fedora (and now RHEL as well) ship the new 'firewire' stack. (ieee1394
and firewire are the subdir names containing the drivers underneath the
drivers/ dir in the kernel source). All the expected modules are loaded
on Derek's system.

However, the lack of stuff in dmesg is slightly disturbing... Shortly
after the first two lines, there should be (at least) a third, related
to device creation, like so:

firewire_ohci 0000:0e:00.0: PCI INT A -> GSI 16 (level, low) -> IRQ 16
firewire_ohci: Added fw-ohci device 0000:0e:00.0, OHCI version 1.10
firewire_core: created device fw0: GUID 00133b010001000f, S800

Derek Atkins wrote:
> Well, I'm hoping Jarod sees this soon..  He's claimed it works.

I still swear it does! ;)

> But I see no /dev/fw* device, and none of the old firewire tools
> seem to work.

So yeah, something is falling down. The card is seen, but no device
nodes are being created, so there's nothing for the tools to work with.
This sounds like a driver bug we've yet to address (at least in the
stack that is in the latest F9 kernel).

> But the card is seen on the PCI Bus:
> 
> 01:09.0 FireWire (IEEE 1394): Texas Instruments TSB12LV26 IEEE-1394
> Controller (Link)

Hrm. Noticed elsewhere that this is reported an ohci 1.0 controller... I
had thought pretty much all TI controllers were ohci 1.1, so it would
seem this is a fairly rare chipset... Ugh. I'd happily trade you a
known-working TI-based ohci 1.1 controller for that one, in order to 1)
get you working sooner and 2) investigate why the heck that one isn't
working. :)

There's actually a chance this card works w/current rawhide, there's
been some additional improvements to locking in recent weeks that might
help, but well, that would require running a rawhide kernel right now to
test, and I don't know how much dependency fallout there might be if you
were to try to add one to an F9 system... Another thing that could be
attempted is to enable some of the debug logging in the firewire-ohci
driver, to see if anything jumps out in the AT/AR packet dumps...


-- 
Jarod Wilson
jarod-ajLrJawYSntWk0Htik3J/w at public.gmane.org