Re: wapbl panic in NetBSD6

[David Holland <dholland-current%netbsd.org@localhost>]

On Wed, Oct 24, 2012 at 01:16:00PM -0400, Greg Kerr wrote:
 > After deleting 100K+ rogue files in /tmp on NetBSD6 ("stable",
 > cvs'd from netbsd.org yesterday), machine froze with a panic in
 > wapbl_register_deallocation.

This one? :(

   /* XXX should eventually instead tie this into resource estimation */
   /*
    * XXX this panic needs locking/mutex analysis and the
    * ability to cope with the failure.
    */
   /* XXX this XXX doesn't have enough XXX */
   if (__predict_false(wl->wl_dealloccnt >= wl->wl_dealloclim))
        panic("wapbl_register_deallocation: out of resources");


 > Rebooted into single-user and fsck -Pfy / ; mv /tmp /tmp.o mkdir
 > /tmp; chmod 1777 /tmp ? was the only way I could recover. Can't get
 > rid of tmp.o though.

If that's the panic, probably you can delete the files in batches,
something like

   cd /tmp.o
   while :; do
        VICTIMS="$(ls | head -1000)"
        if [ x"$VICTIMS" = x ]; then
                break;
        fi
        rm $VICTIMS
        sync; sync
        sleep 1
   done

although if there are subdirs in there it may take somewhat more
complex logic.

-- 
David A. Holland
dholland%netbsd.org@localhost