port-arm/38950: shark, hpcarm kernels prone to hard hangs on process exit

[rafal%pobox.com@localhost]

>Number:         38950
>Category:       port-arm
>Synopsis:       shark, hpcarm kernels prone to hard hangs on process exit
>Confidential:   no
>Severity:       critical
>Priority:       high
>Responsible:    port-arm-maintainer
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Fri Jun 13 14:00:01 +0000 2008
>Originator:     rafal%pobox.com@localhost
>Release:        NetBSD 4.99.64
>Organization:
        lacking
>Environment:
System: NetBSD j720 4.99.64 NetBSD NetBSD 4.99.64 (J720) #11: Fri Jun 13 
08:53:48 EDT 2008 
rafal@fvk:/extra/netbsd-src/current/sys/arch/hpcarm/compile/J720 hpcarm
Architecture: arm
Machine: hpcarm
>Description:
        Both the shark and hpcarm kernels cause hand lockups of the machine
        in what appears to be the process exit path.  Sometimes simple stuff
        like the on-boot fsck cause the hang, sometimes it requires something
        more 'vigorous' like a pkgsrc build, but when the machine does hang
        it doesn't respond to DDB or any interrupts.

        Both machines built from recent (< 2 days old) sources.  The hpcarm
        (a Jornada 728) seems more succeptible to the hang, but that machine
        has root at wd0 @ pcmcia vs. root on NFS for the shark, and I've also
        played with the hpcarm a bunch more, whereas the shark was sitting
        idle until a few days ago.

>How-To-Repeat:
        Have enough processes start/exit and odds are the machine will hang.
        On the Jornada often running the rc.d scripts going to multi-user is
        enough... on the Shark I tickled it with a pkgsrc build (over NFS),
        though I have not set up a complete NFS environment so I have only
        booted it in single-user mode (root == arm destdir on the build
        host ;)).

>Fix:
        Unknown

>Unformatted:
                sources from 6/11/08