[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: raw1394 problems galore



Stefan Richter wrote:
On 10/16/2006, Gene Heskett wrote to lkml and fedora-list:
Greetings all;

After about 4 days of snooping around trying to get kino, either 08.xx or 0.9.2, to run here, I'm getting noplace at a very high rate of speed.

System is an HP dv5120us lappy, with ddual boot of xp and FC5, with FC5 about as uptodate as can be managed when the kmod stuff is always 1 or 2 versions behind the kernel, so I'm still running 2.6.17-1.2174_FC5 for a kernel.

Specifically, kino-0.8.0-2.lvn5.i386.rpm can receive and record a/v from my camera, a Sony TRV-460, but cannot control it via the av/c controls as it cannot 'see' a device to do the controls in the prefs->ieee1394 screen.

Could be version problems between kernel--library--app. Maybe somebody
else knows more about it. But it's more probably the same problem as the
following:

Then kino-0.9.2-1.lvn5.i386.rpm cannot even see the raw1394 interface, reporting on its screen that the raw1394 kernel module isn't loaded, or that it cannot read/write /dev/raw1394.

HP dv5120us is based on Turion 64. Do you run a 64bit kernel on it? Then
the following bug may prevent access:
http://bugzilla.kernel.org/show_bug.cgi?id=4779
The bug has been fixed for read and write operations on /dev/raw1394 in
Linux 2.6.17 (actually 2.6.16-git11).

No 64 bit kernels ever, this has always been a 32 bit install.

The bug has _not_ been fixed yet for "ioctl" operations. The so-called
rawiso interface for isochronous transmission and reception uses ioctls.
I suppose the older Kino version could have slightly more success for
you simply because it uses a different interface for iso transmission
and reception.

My last comment in the bugzilla entry points to explanations on
64bit<-->32bit compatibility code for ioctls, in case anybody feels up
to hack the kernel and contribute the rest of the fix.

Gene, if you run a 64bit kernel, check if you could switch to a 32bit
kernel --- if only to confirm or deny that this is indeed the bug that
is biting your setup. Another thing to try is to test whether gscanbus
or 1394commander can work with raw1394. But as the _very first step_,
check the kernel log for messages from the 1394 drivers.
After last reboot, having reset udev.conf for err only, as debug is so verbose it prevents booting if wlan0 is enabled.

Oct 16 11:16:31 diablo kernel: ohci1394: fw-host0: OHCI-1394 1.1 (PCI): IRQ=[10] MMIO=[c0209000-c02097ff] Max Packet=[2048] IR/IT contexts=[
4/8]

Then, manually loaded via 'modprobe raw1394':

Oct 16 11:50:11 diablo kernel: ieee1394: raw1394: /dev/raw1394 device initialized Oct 16 11:50:11 diablo kernel: audit(1161013811.874:4): avc: denied { getattr } for pid=2753 comm="pam_console_app" name="raw1394" dev=tmpfs ino=10625 scontext=system_u:system_r:pam_console_t:s0-s0:c0.c255 tcontext=system_u:object_r:device_t:s0 tclass=chr_file Oct 16 11:50:11 diablo kernel: audit(1161013811.874:5): avc: denied { setattr } for pid=2753 comm="pam_console_app" name="raw1394" dev=tmpfs ino=10625 scontext=system_u:system_r:pam_console_t:s0-s0:c0.c255 tcontext=system_u:object_r:device_t:s0 tclass=chr_file

SELinux is in permissive mode, and /dev/raw1394 has perms of:
[root diablo ~]# ls -l /dev/raw1394
crw-rw-rw- 1 root root 171, 0 Oct 16 11:50 /dev/raw1394

As I had given up, the camera is packed away, but I'll get it out and connect it again for grins:

And no further messages were logged when I plugged it in and turned it on.

kino-0.8 receives video from it in real time and is doing so right now, and can capture it to file, and then play/edit that file, or could saturday when I last tried it. I ASSume that kino-0.9.2 could also play/edit that file, but have not verified that by reinstalling 0.9.2.

The module is loaded when the cable is plugged in, and the /etc/security/console.perms.d/50-defaults has been edited so that /dev/raw1394 as created by the hotplug event now has these perms:
crw-rw-rw- 1 root root 171, 0 Oct 15 20:20 /dev/raw1394

This has had no effect on the performance of either version of kino.

So when do we actually get a functioning firewire interface, which we DID have 18 months ago in FC2,

Did you run FC2 as 32bit environment on 32bit kernel?

Yes, and kino-0.7.5 died with kernel changes in the ieee1394 code someplace at about 2.6.9 IIRC. That FC2 box is my kernel test box, is 950 miles SE of me in West Virginia and powered down ATM, but its running 2.6.19-rc1 just fine except for the ieee1394 problems. That was why I thought I'd see if I could make it work here on my lappy with FC5 and release kernels. I upgraded from kernel-2.6.17-1.2174_FC5 to kernel-2.6.17-1.2187_FC5 this morning with no detectable difference.

before someone just had to rewrite the 1394 stuff again?

The 1394 kernel drivers are not being rewritten.

I was told it was a total rewrite of bad code when I complained about a year ago. My reply at the time was that it worked, and I don't often fix things that are working. I'm getting lazy in my dotage I guess.

 We just discover old
bugs, try to fix them on a rather slow rate, and sometimes introduce
regressions along with bug fixes. (We try not to do so, but there is
weird hardware out there.) I think similar things can be said about the
status quo of libraw1394's development; minus the regressions I hope. I
don't know about the other libraries and about the status of the
application software. The last big "rewrite" concerning the whole stack
from kernel to apps was the addition of the rawiso programming
interface, which was quite a long time ago, but migration of application
programs to it seems to be a still ongoing process, including
deployment. (Even Linux 2.4 has the rawiso interface.)

(I added linux1394-user to the Cc list. You don't need to subscribe
there; just keep all Ccs during the discussion.)

As for 1394commander or gscanbus, I have not managed to find rpms of those in any of the repos yumex or SPM shows me. They apparently are not part of the FC5 tree. I'd love to see what those 2 might have to say about the system. The only thing I do have is dvcont, which reports this:

[root diablo ~]# dvcont dev 0 play
Could not find any AV/C devices on the 1394 bus.

The camera is still plugged in and powered up.

Thanks Stefan. Silly Q though, are you any kin to Thomas, a math prof at math-tu? I've known him since back in my amiga days.

--
Cheers, Gene


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]