kern/41161: FFSv2 w/ log and softdep: kernel panics on mount

[vs%nifelheim.info@localhost]

>Number:         41161
>Category:       kern
>Synopsis:       FFSv2 w/ log and softdep: kernel panics on mount
>Confidential:   no
>Severity:       serious
>Priority:       low
>Responsible:    kern-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Mon Apr 06 21:15:00 +0000 2009
>Originator:     Volkmar Seifert
>Release:        NetBSD 5.0 RC3 / Sparc64
>Organization:
>Environment:
NetBSD nidhoegg 5.0_RC3 NetBSD 5.0_RC3 (GENERIC) #0: Fri Mar 20 18:16:48 UTC 
2009  
builds%b4.netbsd.org@localhost:/home/builds/ab/netbsd-5-0-RC3/sparc64/200903200521Z-obj/home/builds/ab/netbsd-5-0-RC3/src/sys/arch/sparc64/compile/GENERIC
 sparc64
>Description:
I have freshly installed NetBSD 5.0RC3 on my SUN Ultra-5, a Sparc64.
All my partitions are using FFSv2 with the log-option (except the 
root-partition, which is of course using FFSv1), and because I was ignorant of 
the fact log and softdep are mutual exclusive, I used both options at the same 
time. Upon mount, the kernel told me that softdep would be ignored because log 
was specified, and then trapped and fell into the debugger. This was absolutely 
reproducable, and could be solved only by removing softdep from /etc/fstab. 

That was Saturday. While I just now attempted to reproduce the problem for this 
PR, the kernel directly -panicked-, dumped, and rebooted. No more traps and 
kernel-debuggers. If necessary, I can provide the compressed core-dump for 
download (it's about 32MB).

As you can see from the environment, it was a standard "NetBSD 5.0_RC3 
(GENERIC)"-Kernel, which is why I do not provide the compressed kernel, too. (I 
can do so, though, if really necessary)
>How-To-Repeat:
mount an FFSv2-drive with both options log and softdep using a NetBSD/sparc64 
5.0RC3 Generic kernel.
>Fix:
Do not specify both options at the same time. 
They are mutually exclusive anyway, as can be read in the according man-pages.

A cleaner approach would be to really ignore the softdep-option when log was 
specified, and not only to print a message claiming to ignore it.