> I'm having problems with my Philips webcam and an OHCI controller (buffer underflow).
> The machine freezes rock-solid if I then try to kill -9 the application which is 
>using the
> usb/webcam. No OOPS, nothing.

How about when you unplug the camera?

In either case, OHCI is unlikely to be directly involved -- that's more
suggestive of a bug in the PWC driver, though of course one can't rule
out OHCI with just these facts.

You might try enabling slab poisoning:  mm/slab.c is set up for a
CONFIG_DEBUG_SLAB option which you can set through
kernel config options (under kernel debugging).  Sometimes drivers
don't refcount their devices correctly, or handle synchronization
between disconnect() and close() correctly.  Sometimes nastier failure
modes show up with OHCI than with the UHCIs.


> - PCI OHCI controller, 2 ports: "i/o CARD-USB", with OPTi 82C861 chip 

The OPTi chips may be less reliable as OHCI goes.  Evidently MacOS X
has some really nasty workarounds for them.  (I've also noticed that they
can't enable the "UE" exception.)


> I've compiled in as many debugging features as I possibly could (in pwc and ohci) 
>and attached
> the messages file, zipped. Things go wrong around when the driver reports "pwc Frame
> buffer underflow", some way down the file.

Curious:  the "USB-error/status: 8" bit means OHCI saw a data overrun (camera
sent too much data), but that doesn't seem to have been reported in any "RET URB"
messages.  And those overruns were reported before the "Frame buffer underflow"
messages.

- Dave



_______________________________________________
[EMAIL PROTECTED]
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel

Reply via email to