pomidorabelisima wrote:
> Hans Verkuil wrote:
>> On Thursday 08 March 2007 23:05, pomidorabelisima
wrote:
>>> hard lockup with 2.6.19-1.2911.6.5.fc6 kernel
>>> firmware crash with 2.6.19-1.2911.6.4.el5(fc6)
kernel
>>> firmware crash with 2.6.20 series kernels from
>>> FC7/rawhide
>>> and
>>> at least 2 days uptime with 2.6.18-1.2798.fc6
kernel
>>> with
>>> DEC DMA ERROR ffffffff
>>> DMA ERROR ffffffff ffffffff ffffffff -1
>>> DMA ERROR 00000003 00000000 00000013 -1
>>> ENC DMA ERROR b
>>>
>>> drawing to conclusion;
>>> something really weard going on with 2.6.19 and
2.6.20
>>> series kernels
>>> no matter how kernels are configured
>>> mama mia
>> Ah, so you DO get DMA errors. Possible causes:
>
> not me,
> THE card
> ;))
>
>> - you have a CPU frequency changer daemon running.
Disable it: the PVR
>> hardware doesn't like that at all.
>
> nope;
> /etc/init.d/cpuspeed status
> cpuspeed is stoped
>
>> - software or hardware RAID can also cause it
>
> partialy,
> usr tree IS on:
> /dev/md4 on /usr type ext3 (rw)
> BUT
> input/output for mythtv mpegs IS on:
> /dev/sdd5 on /mythtv type xfs (rw)
> NON RAID disk
> so,
> partialy excluded
>
>> - el cheapo graphics cards using main memory as the
framebuffer also
>> cause similar problems.
>
> nope;
> NV34 [GeForce FX 5200]
> DO NOT DO THAT
> there is NO onboard GPU
>
>> - older via motherboards are particularly prone to
this.
>
> nope,
> nForce2 ultra chipset
>
>> - in general, VIA MBs are more sensitive to this
issue than nVidia MBs,
>> and Intel MBs seem to be the best.
>
> IS nForce* chipsets prone to DMA issues,
> need more research
> so,
> partialy excluded
>
>> The root cause is a buggy DMA engine and/or
hardware of the cx23416. The
>
> yes yes you said that 100 times
> I DO NOT blame nither critisize YOU
> also, I understand that people from hauppauge
produced THE card
> for usage on OS like win
> you know, it must reboot once per day..
> :))
>
>> Hauppauge PVR500 is actually the only card that
seems to avoid this
>> problem (mostly) due to the PCI bridge that seems
to isolate the
>> cx23416 from most PCI DMA issues.
>
> before pvr-150 was installed
> with only pvr-350
> system hasn't got problem at all
> so cx23415 IS NOT problematic
> and yes
> cx23416 ON pvr-150 IS problematic
> cx23416 on PG600-2 need testing!
>
> found some stuff on pci latency
> pci_latency.sh from old trunk
> refered to
> Linux hardware stability guide, Part 2:
>
http://www-128.ibm.com/developerworks/library/l-hw2.html
>
> and
> setpci -v -s 01:07.0 latency_timer=ff
> setpci -v -s 01:09.0 latency_timer=ff
> now there are:
> setpci -v -s 01:07.0 latency_timer
> 01:07.0:0d = fc
> setpci -v -s 01:09.0 latency_timer
> 01:09.0:0d = fc
> fc==252
>
> I'll post what's going on
>
>
> rgds
> poma
same hw now on:
2.6.18-8.1.82.el5
CentOS release 5 (Final)
all that schmoozy&snoozy was indirect generated by
broken nvidia onboard gigabit nic
HAH!
sorry hauppauge!
not yours fault
EOS
poma
____________________________________________________________________________________
Looking for a deal? Find great prices on flights and hotels with Yahoo!
FareChase.
http://farechase.yahoo.com/
_______________________________________________
ivtv-devel mailing list
[email protected]
http://ivtvdriver.org/mailman/listinfo/ivtv-devel
