Subject: kern/31559: kernel panic uvm_unmap_remove()
To: None <kern-bug-people@netbsd.org, gnats-admin@netbsd.org,>
From: Wolfgang S.Rupprecht <wolfgang+gnus20051011T115600@dailyplanet.dontspam.wsrcc.com>
List: netbsd-bugs
Date: 10/11/2005 19:08:00
>Number: 31559
>Category: kern
>Synopsis: kernel panic uvm_unmap_remove()
>Confidential: no
>Severity: critical
>Priority: high
>Responsible: kern-bug-people
>State: open
>Class: sw-bug
>Submitter-Id: net
>Arrival-Date: Tue Oct 11 19:08:00 +0000 2005
>Originator: Wolfgang S. Rupprecht
>Release: NetBSD 3.99.9
>Organization:
W S Rupprecht Computer Consulting, Fremont CA
>Environment:
NetBSD capsicum.wsrcc.com 3.99.9 NetBSD 3.99.9 (WSRCC_MPACPI) #31: Thu Oct 6 23:18:29 PDT 2005 wolfgang@capsicum.wsrcc.com:/var/obj/netbsd/sys/arch/i386/compile/WSRCC_MPACPI i386
Architecture: i386
Machine: i386
>Description:
Sources were from the 6th. Panic was from gqmpeg playing an
ogg just as one song finished and the next was about to start.
Backtrace shows the panic was in gqmpeg closef().
uvm_unmap_remove()
uvm_unmap1()
pipe_free_kmem()
pipeclose()
pipe_close()
closef()
syscall_plane()
--- syscall (number 6) ---
Sorry, no serial console. I have a jpeg though. Folks might
need to cut and paste the url to get around a referrer-based
anti-leaching security hack. (sorry!)
http://www.wsrcc.com/wolfgang/private/netbsd-audio-panic/pic20051011_082410.jpg
>How-To-Repeat:
This is very repeatable and takes 5-10 tries.
1) I selected 4 1-hour long ogg files on the command line of gqmpeg.
gqmpeg 1.ogg 2.ogg 3.ogg 4.ogg
2) start playing
3) open the directory window by clicking on the button that
has up arrow with the line under it on the second row on the left.
4) in the directory window double click on the second song to
forcefully play it.
5) repeat until the last song is selected then go onto
selecting the first song.
6) at this point I got the panic.
This panic happened to me 5 or so times while playing music
with qqmpeg and it was always during some switching or within
5 seconds of doing so.
This may be related to re-selecting a song that was previously
deselected but still in the buffer cache. The machine is very
lightly loaded (few daemons) and has 768Meg dram in it. It
should have lots of very old stuff in the buffer cache to
deselect/reselect.
>Fix:
>Unformatted: