On tor, 2006-02-23 at 18:58 +0100, Jörg Dreßler wrote:
> Hello,
>
> i have met a really annoying problem with my tv-card and the new version
> of mythtv. I've read about this in the archive but the solutions did not
> work for me. Prior to trying some of the proposals in the archive the
> kernel-log looked like this:
> Jan 17 10:07:00 fortress kernel: bt878(0): irq FDSR FBUS risc_pc=2f74e008
> Jan 17 10:07:00 fortress kernel: bt878(0): irq FDSR FBUS risc_pc=2f74e008
> Jan 17 10:07:00 fortress kernel: bt878(0): irq FDSR risc_pc=2f74e008
> Jan 17 10:07:00 fortress kernel: bt878(0): irq FDSR risc_pc=2f74e008
> {20 - 30 pages kernel-log}
> Jan 17 10:07:25 fortress kernel: bt878(0): irq FDSR risc_pc=2f74e008
> Jan 17 10:07:25 fortress kernel: bt878(0): irq FDSR risc_pc=2f74e008
> Jan 17 10:07:25 fortress kernel: irq 217: nobody cared (try booting with
> the "irqpoll" option)
> Jan 17 10:07:25 fortress kernel: [<c014ed04>] __report_bad_irq+0x24/0x90
> Jan 17 10:07:25 fortress kernel: [<c014ee12>] note_interrupt+0x72/0xc0
> Jan 17 10:07:25 fortress kernel: [<c014e6e2>] __do_IRQ+0x132/0x150
> Jan 17 10:07:25 fortress kernel: [<c0107036>] do_IRQ+0x36/0x70
> Jan 17 10:07:25 fortress kernel: [<c010537a>] common_interrupt+0x1a/0x20
> Jan 17 10:07:25 fortress kernel: [<c01500ca>] find_get_pages_tag+0xaa/0xd0
> Jan 17 10:07:25 fortress kernel: [<c015a5da>] pagevec_lookup_tag+0x2a/0x40
> Jan 17 10:07:25 fortress kernel: [<c0196f0c>] mpage_writepages+0x15c/0x3f0
> Jan 17 10:07:25 fortress kernel: [<f95b4b50>]
> reiserfs_writepage+0x0/0x40 [reiserfs]
> Jan 17 10:07:25 fortress kernel: [<c027bace>] con_put_char+0x2e/0x40
> Jan 17 10:07:25 fortress kernel: [<c0120d88>] __wake_up+0x38/0x50
> Jan 17 10:07:25 fortress kernel: [<c014f5c2>]
> __filemap_fdatawrite_range+0x72/0x90
> Jan 17 10:07:25 fortress kernel: [<c014f602>] filemap_fdatawrite+0x22/0x30
> Jan 17 10:07:25 fortress kernel: [<c0172b81>] do_fsync+0x61/0xd0
> Jan 17 10:07:25 fortress kernel: [<c01042ab>] sysenter_past_esp+0x54/0x79
> Jan 17 10:07:25 fortress kernel: handlers:
> Jan 17 10:07:25 fortress kernel: [<f94ae7f0>] (bttv_irq+0x0/0x3a0 [bttv])
> Jan 17 10:07:25 fortress kernel: [<f92dc490>] (bt878_irq+0x0/0x240 [bt878])
> Jan 17 10:07:25 fortress kernel: Disabling IRQ #217
>
> Till now i changed the PCI slot, which changed the IRQ in the message.
> Next i raised the PCI latency which results in 40 - 50 pages of failure
> messages and finally in bt878.c:
> controlreg |= 0x13; // Nebula DigiTV - was: 0x1b;
>
> Now the kernel log looks like that:
> {40 - 50 pages kernel-log}
> Feb 22 21:04:51 fortress kernel: bt878(0): irq FDSR risc_pc=31231008
> Feb 22 21:04:51 fortress kernel: bt878(0): irq FDSR risc_pc=31231008
> Feb 22 21:04:51 fortress kernel: bt878(0): irq FDSR risc_pc=31231008
> Feb 22 21:04:51 fortress kernel: bt878(0): IRQ lockup, cleared int mask
>
> Needless to say, watching TV afterwards fails, even when the last line
> is missing sometimes.
> I suspect the failure is somehow triggered by the mythtv
> EPG/EIT-scanner, because of time correlation and it didn't occur when
> using the 0.18-version. And i've once noticed that tv_grab_dvb is
> causing the same failure, except for the IRQ-lock and dvb is still
> working afterwards. This is the really annoying thing about it as mythtv
> is scanning for EIT-data almost all the time.
> I'm using SUSE 10 and the 2.6.13-15.8-smp kernel. The DigiTV-card is the
> old one with the nxt6000 frontend.
> Anyone who knows how to get rid of this?
The FBUS error you see in your first case means the card had to dump
data because it didn't get access to the pci bus before it's internal
buffer overflowed. The controlreg change you did should reduce the
chances of getting the FBUS error, it should however (as I understand
it) not have anything to do with your other problems. What are the last
few messages before the mentioned 40-50 pages? (tip: install a sysloger
with "last message repeated NN times" feature). I have on some occations
gotten a SCERR error which usually gets followed by log patterns such as
what you show in the second example.
Also, please try latest v4l-dvb from hg, to make sure you are not
suffering from an allready fixed bug.
Sigmund
>
> regards
> Jörg Dreßler
>
>
> _______________________________________________
>
> linux-dvb@xxxxxxxxxxx
> http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb
_______________________________________________
linux-dvb@xxxxxxxxxxx
http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb
