Hi Brad,
Brad Gass wrote:
> I just looked at the link Pete gave me, and that doesn't really match the
> symptoms of my oops, unless I mis-read his post.
I would guess that if ov511 were opened before it finished initializing you
would see some bad effects, but they would probably show up pretty quickly. I
am grateful to Pete for finding this bug though.
> I had the camera grabbing frames every 15 seconds or so for hours,
> specifically from Apr 1 15:47 to Apr 2 08:30, or about 16.5 hours. That's
> when the oops happened, and usb-ohci interrupts stopped coming in.
This sort of problem was common a long time ago, until I implemented delayed
deallocation of the frame buffer. It seems that continuously calling
rvmalloc() and rvfree() corrupts kernel memory after a while. Newer drivers
don't call rvfree() until 5 seconds after close() is called, and you can set
this timeout to whatever you want with the module parameter buf_timeout=(time
in seconds).
> So it's doesn't seem to be a hotplug issue, or hardware initialization
> issue. The camera is plugged into the root hub at the time the USB
> subsystem is loaded. I'm not fully versed in the internals of the USB or
> OV511 systems by any means, but it does work for quite awhile, just not
> continually over longer periods of time. As a point of reference, the
> number of interrupts fired up to the time of the oops was 4355919.
I would not be surprised if ov511's interrupt code, which does tons of
processing, is responsible for this oops. You might want to try driver version
1.35 with the ttpp=1 module parameter, which moves most processing from
interrupt time to task time.
--
Mark McClelland
[EMAIL PROTECTED]
_______________________________________________
[EMAIL PROTECTED]
To unsubscribe, use the last form field at:
http://lists.sourceforge.net/lists/listinfo/linux-usb-devel
