[pvrusb2] My tuners occasionally get stuck

[Mike Isely]

Jan:

I think Gary has a good point here.

Another angle you should consider is something I've seen here repeatedly 
over time, having nothing to do with the software: power.  External USB 
devices that are not USB-powered of course rely on their own power 
supply.  Cheap wall-wart power adapters simply do NOT have the 
resilience of a PC power supply when it comes to dirty power.  A nearby 
A/C compressor cycling on/off can glitch the power enough to crash an 
externally-wall-wart-powered USB device while PC internal power supplies 
usually have enough internal energy storage (i.e. capacitors) to ride 
out an event like that.  I've *personally witnessed* this sort of thing 
in the past and there have been repeated reports of people with randomly 
crashing USB devices that magically fix themselves when their power 
supply is moved behind, say, a UPS or sometimes just a good surge 
protector (depending on the severity of the power issue).

Your report mentions all 4 tuners crashing at once, which admittedly 
might be more likely a kernel USB stack bug.  Note that the pvrusb2 
driver instances itself once for every device you have connected.  
Since the driver instances generally operate independantly of one 
another, a driver misbehavior should not be able to cause simultaneous 
failure of all tuners at once.  The fact that a system reboot does not 
clear the problem is another factor that suggests the foul-up is in the 
tuners themselves, though the trigger could still be the kernel USB 
stack.

Since you have 4 tuners, you have a unique opportunity to try an 
experiment: Put just 2 of the tuners behind a UPS.  Wait for the next 
failure.  Do all four tuners fail at once again or just two?  If only 2 
fail now then that's almost a smoking gun clue indicating bad power.  
If you still have all 4 failing at once then I'd put my money on Gary's 
theory: a kernel USB stack bug.

Another idea - if you have a portable USB hard drive, plug it in as well 
and run a repeated massive tar command to generate large amounts of I/O.  
When you next get a failure, does the hard drive survive while the 
tuners still fail?  If the hard drive fails too, then you've just 
exonerated the entire V4L subsystem and I'd then be looking really hard 
at the USB stack.

Hope that helps.

  -Mike


On Wed, 21 Mar 2012, Jan Ceuleers wrote:

> On 21/03/12 17:09, Gary Buhrmaster wrote:
> > <conjecture>
> 
> Understood, and thanks for your insights.
> 
> > Almost certainly a kernel usb issue.  The 1900s and the
> > APC device are likely victims, not the cause.  These types
> > of bug reports (the 'clear tt' ones) with usb devices have a
> > tendency to reoccur on some irregular basis as the usb
> > subsystem in the kernel gets improved.  Usually the
> > problem is patched in the next kernel rev (because others
> > have reported it), although sometimes you have to do the
> > git bisect to find the patch that was "special" to get it fixed.
> 
> Indeed my thoughts align with yours: the tuners are significant USB traffic
> generators which are probably more likely than other devices to expose
> weaknesses in the kernel USB subsystem. But I did google the kernel log
> messages without success (probably because they are so terse).
> 
> I recently upgraded from Ubuntu 10.10 to 11.04, and I had this problem under
> 10.10 as well. So the problem is not restricted to the 2.6.38 kernel in 11.04.
> 
> But I will try a more recent kernel as well before taking it to linux-usb if I
> still need to.
> 
> Thanks, Jan
> _______________________________________________
> pvrusb2 mailing list
> pvrusb2 at isely.net
> http://www.isely.net/cgi-bin/mailman/listinfo/pvrusb2
> 

-- 

Mike Isely
isely @ isely (dot) net
PGP: 03 54 43 4D 75 E5 CC 92 71 16 01 E2 B5 F5 C1 E8