On Sat Oct 27 2007 at 00:21:32 -0400, Blair Sadewitz wrote:
> Recently I tried running a -current installation on an LFS root
> partition, and unfortunately I've been getting deadlocks/freezes again
> (no panics thusfar).  Each time in ddb, the backtrace showed a call to
> ltsleep() somewhere in the bowels of lfs.

"somewhere"?  Is it different place each time?

I have a r/w version of lfs kinda working under rump now, but it
needs some patches which need to be architected better that they can
be committed.  Hopefully it will help in exercising the file system in
ways which are more helpful for fixing bugs.

> I was wondering: if there's anything preventing us from converting
> lfs_interlock to a mutex and perhaps getting rid of some of the spl()
> wrapping.
> And does anyone have any other ideas on how we might improve
> synchronization in this beast?

I'm all for getting rid of ltsleeps, but it's likely to be a fairly long
and arduous journey with lfs.

-- 
Antti Kantee <pooka@iki.fi>                     Of course he runs NetBSD
http://www.iki.fi/pooka/                          http://www.NetBSD.org/
    "la qualité la plus indispensable du cuisinier est l'exactitude"