usb/78208: ulpt page fault

[Ian Dowse]

The following reply was made to PR usb/78208; it has been noted by GNATS.

From: Ian Dowse <iedowse at maths.tcd.ie>
To: Jan-Espen Pettersen <sigsegv at leakingmemory.org>
Cc: FreeBSD-gnats-submit at FreeBSD.org
Subject: Re: usb/78208: ulpt page fault 
Date: Sun, 13 Mar 2005 21:14:58 +0000

 In message <200502281829.j1SIT6sb003645 at endeavour.localnet.radiotube.org>, Jan-
 Espen Pettersen writes:
 >>Description:
 >I got this page fault trap just after printing via ulpt. The pages came out ju
 >st fine, which is why I think this problem is about handling EOF or close of t
 >ransaction with /dev/ulpt0. To me this looks like a timer (as the name *_tick)
 > which probably was not stopped in time, and therefore didn't have a valid xfe
 >r pointer to pass to the setup routine. These crashes has been going on for a 
 >while, but I haven't been able to get crashdumps before recently when I starte
 >d to press Ctrl+Alt+F1 shortly after starting the print jobs.
 
 This looks like it could be caused by a known race condition in
 kernel timers. Is the problem easy to reproduce for you? If so,
 could you try the patch at
 
 	http://people.freebsd.org/~iedowse/releng_5_callout.diff
 
 and see if it helps? Remove your patch to ulpt.c first. The above
 patch is a partial backport of some changes that went in to -CURRENT
 that fix some race conditions with Giant-locked kernel timers.
 
 Ian