serious issue caused by usb device, stalling almost all operations

[Alexander Best]

On Mon Oct 25 10, Alexander Motin wrote:
> Hans Petter Selasky wrote:
> > On Wednesday 20 October 2010 17:30:40 Alexander Best wrote:
> >> hi there,
> >>
> >> i'm running HEAD (r213495; amd64). i stumbled upon this severe problem:
> >>
> >> after attaching my mobile phone, it simply resets without doing mount or
> >> anything. however after letting the device come up again it won't show up
> >> in the console. after detaching it the usb subsystem seemed to have
> >> completely crashed. but that's not all. the following programs now simply
> >> hang without producing any output C-C won't do anything:
> >>
> >> - dmesg
> >> - top
> >> - ps
> >> - killall
> >> - camcontrol devlist
> >> - usbconfig
> > 
> > That's most likely because USB's umass driver is waiting for cam to drain. 
> > Possibly some refcounting is not correct. I suspect this is not a USB problem. 
> > Try to enter into the debugger, and look for backtrace for function stuck in 
> > umass_detach.

i set debug.kdb.panic=1, but didn't work, because writing the core dump stalled
and watchdog came up.

any advice?

cheers.
alex

> 
> It is a bit suspicious that problem happens only when device dies during
> request. Are you sure that running command was properly aborted when
> device got detached? Every running command has own set of references,
> denying detach.
> 
> -- 
> Alexander Motin

-- 
a13x