From owner-freebsd-hackers Tue Jul 7 07:46:11 1998 Return-Path: Received: (from majordom@localhost) by hub.freebsd.org (8.8.8/8.8.8) id HAA07699 for freebsd-hackers-outgoing; Tue, 7 Jul 1998 07:46:11 -0700 (PDT) (envelope-from owner-freebsd-hackers@FreeBSD.ORG) Received: from dark.uwaterloo.ca (dark.uwaterloo.ca [129.97.50.72]) by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id HAA07675; Tue, 7 Jul 1998 07:46:05 -0700 (PDT) (envelope-from erick@dark.uwaterloo.ca) Received: from localhost (erick@localhost) by dark.uwaterloo.ca (8.8.7/8.8.7) with SMTP id KAA16379; Tue, 7 Jul 1998 10:38:59 -0400 (EDT) (envelope-from erick@dark.uwaterloo.ca) Date: Tue, 7 Jul 1998 10:38:58 -0400 (EDT) From: Erick Engelke To: "Alton, Matthew" cc: "'hackers@freebsd.org'" , "'freebsd-fs@freebsd.org'" , "Smallie, Scott" Subject: RE: lfs In-Reply-To: <31B3F0BF1C40D11192A700805FD48BF901776600@STLABCEXG011> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-freebsd-hackers@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG On Mon, 6 Jul 1998, Alton, Matthew wrote: > erick wrote: > > I see that a newlfs'd partition shows up nicely with the ifile log > > file and lost+found. But as soon as sync happens, whether timed or > > due to a umount or file i/o, we have a panic. > [Alton, Matthew] > Are you getting a core dump for kgdb? I don't have a scratch > box running this code right now. I've got a nasty old 486/33 > almost up for testing purposes... Now that I'm more up to speed on kdb (thanks for the tips), I got a bit farther. My scratch pc is a 486/50... not much better, eh? The actual error is in the lfs_writeseg() of lfs_segment.c where lfs is copying individual bufs into contiguous memory for a big long write. (called from lfs_segwrite, from lfs_sync) It looks like I'm dying at ++cbp->b_vp->v_numoutput... and a quick look suggests it is possible the vnode points to garbage. I'll have to verify that with kdb and experiment some more. While on the topic, LFS has a lot of hacks, many specifically about bufferring. Some other PC OS's simply manipulate the page tables to map buffers into contiguous RAM... (and then instruct 3rd parties to never do the same). I don't imagine that pagetable fiddling practice is ever used with FreeBSD, is it? > > Hmmm, where do printfs send their output... nowhere I happen to be > > looking? > [Alton, Matthew] > I thought they went to /dev/console They do. My mistake. Thanks for the hints. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message