new shutdown order, part 2

[David Young <dyoung%pobox.com@localhost>]

I'm making good progress on tearing down stacks of disks and filesystems
at shutdown, however, there are a couple of outstanding problems where I
need some help.

Today I am running these patches,
<ftp://cuw.ojctech.com/users/netbsd-780cd11e/new-shutdown-order.patch1>.

ccd(4) and cgd(4) need a major overhaul before they will attach to and
detach from the device tree. dk(4) and raid(4) require less involved
changes.  I've started the raid(4) changes; if you would like to tackle
dk(4), ccd(4), or cgd(4), let me know.

It seems that WAPBL is giving me some trouble.  The kernel cannot force
FFS on root to unmount if logging is enabled, but the unmount hangs
trying to re-acquire wl_mtx at line 578 in wapbl_discard():

573                              * The buffer will be unlocked and
574                              * removed from the transaction in brelse
575                              */
576                             mutex_exit(&wl->wl_mtx);
577                             brelsel(bp, 0);
578                             mutex_enter(&wl->wl_mtx);
579                     }
580             }
581             mutex_exit(&wl->wl_mtx);
582             mutex_exit(&bufcache_lock);

I put the DDB session log at
<ftp://cuw.ojctech.com/users/netbsd-129a1f62/wapbl.shutdown.hang>.

Dave

-- 
David Young             OJC Technologies
dyoung%ojctech.com@localhost      Urbana, IL * (217) 278-3933