[mythtv-users] A post about no problem, but still wondering ...

[Mike Thomas]

On Fri, 11 Apr 2014 21:53:28 -0500
"Saul A. Peebsen" <jaglover at gmail.com> wrote:
> On Fri, 11 Apr 2014 22:26:18 -0400
> "R. G. Newbury" <newbury at mandamus.org> wrote:
> 
> > On 11/04/14 09:30 PM, Saul A. Peebsen wrote:
> > > On Fri, 11 Apr 2014 16:28:46 +1200
> > > Stephen Worthington <stephen_agent at jsw.gen.nz> wrote:
> > >
> > >> On Thu, 10 Apr 2014 14:32:44 -0500, you wrote:
> > >>
> > >>> I had a crashing BE which crashed almost every time HVR2250 was
> > >>> accessed first time. Once it was working it kept working. It was
> > >>> so bad I enabled hardware watchdog support in my kernel to have
> > >>> this box reset automatically (headless box in a closet).
> > >>> I was thinking the motherboard is bad, so I replaced it. No
> > >>> cigar. Until I added 2 GiB of RAM to make it 4 GiB total ...
> > >>> not a single crash after that.
> > >>> Is there an explanation to this?
> > >>
> > >> Is it by chance a 32-bit system?
> > >
> > > Nope, 64. And it is not the memory, when I replaced the
> > > motherboard it came with new RAM.
> > 
> > Your old motherboard may have been the inadvertent source of the 
> > problem. I also have an HVR2250. It uses 3 modules, which have to
> > be loaded in the correct order. Otherwise the card will not work.
> > But I have never had the BE crash! The problem appears to be that
> > there can be race condition where one module starts to load, and
> > another starts to load before the first has finished loading and
> > stabilized. Hilarity ensues...</sarc>
> > 
> > 
> > I ended up adding the following lines to /etc/rc.d/rc.local
> > 
> > systemctl stop mythbackend.service
> > modprobe -r tda18271
> > modprobe -r s5h1411
> > modprobe -r saa7164
> > modprobe -r s5h1411
> > modprobe -r tda18271
> > modprobe -r s5h1411
> > echo "  Ignore 'FATAL error' messages above if next line is 'No
> > output...' " lsmod | grep tad18271
> > lsmod | grep s5h1411
> > lsmod | grep saa7164
> > echo "  No output means modules were properly unloaded"
> > 
> > modprobe  tda18271
> > modprobe  s5h1411
> > modprobe  saa7164
> > echo "  Modules re-loaded in proper order"
> > systemctl start mythbackend.service
> 
> Hmmm ... I have all HVR2250 modules built in, no modules. Firmware,
> too. Maybe I didn't describe the issue clearly enough.
> Every time I came home from work and tried to watch live TV the BE
> crashed when the channel was provided by HVR2250. After reboot it
> usually worked, I could change channels, no problem. But only until I
> switched to another card, switching back to HVR2250 resulted in
> another crash. And waiting until watchdog kicks in and reboots the
> box. All my scheduled recordings via HVR2250 did not record.
>  
> > Sounds like you are lucky. Your new motherboard +/- card combo +/-
> > your distro don't have that problem!
> 
> My new motherboard/RAM combo had exactly the same problem ... until I
> added some RAM. 
> 

Dear Saul,

Did you perchance have any other cards installed in the machine? I had a
similar problem when I used HVR2200/2250 cards in a box with with 2
Radeon cards.

The symptoms that computer experienced included HARDWARE ERROR
messages, tuner lock-ups, tuner driver's spewing error messages,
graphics stalls and so on, but none of the symptoms really pointed to
the presence of the graphics cards. Changing kernel, userspace,
physical RAM, motherboard, CPU and PSU chased the problem around but
never nailed it but when I removed one of the Radeon cards the
mysterious happenings went away.

I'm not saying it is down to your graphics cards or their driver but
don't be surprised if it is caused by something else on your computer.
Hardware or driver problems can show up in the most mysterious ways.

If you think it is a memory problem, you'll never be sure unless you
fit ECC memory and enable EDAC. Also as one of the other posters
recommended, you really must use a serial console to capture any
emergency output.

Yours,

Mike.