Date: Thu, 27 Nov 2003 00:08:53 +0000 From: Jay Cornwall <jay@evilrealms.net> To: Martin <nakal@web.de> Cc: freebsd-current@freebsd.org Subject: Re: Panic with ugen Message-ID: <3FC54095.6030209@evilrealms.net> In-Reply-To: <1069888991.2521.7.camel@klotz.local> References: <1069874342.704.18.camel@klotz.local> <3FC502B5.5020700@evilrealms.net> <1069888991.2521.7.camel@klotz.local>
next in thread | previous in thread | raw e-mail | index | archive | help
Martin wrote: > It looks like this: > panic() > destroy_dev() > ugen_destroy_devnodes() > ugen_set_config() Yes, that's the one, and I think I can see why. The existing code fixed devfs problems for normal ugen_set_config calls, but doesn't account for what happens when an error occurs (which is presumably happening in your example program, as you said it gives an error the first time round) - the devfs stuff only half completes. (actually, looking at that error handling code, it doesn't look like it's been thought through well anyway - /* XXX should only do this after setting new altno has succeeded */ - maybe time to clean this code up?) After the device endpoints are destroyed (sys/dev/ugen.c:1038), the returns on lines 1055 and 1058 need to be covered by a devnode recovery procedure - particularly tricky given we just wiped the endpoint descriptors clean. I'll look at restructuring this code tomorrow, if Bernd doesn't beat me to it. -- Cheers, Jay http://www.evilrealms.net/ - Systems Administrator & Developer http://www.imperial.ac.uk/ - 3rd year CS student
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3FC54095.6030209>