Subject: Re: LFS on multi-processor systems
To: None <current-users@NetBSD.org>
From: Brian de Alwis <firstname.lastname@example.org>
Date: 09/20/2007 16:33:23
So I've been running a full build.sh in parallel with building firefox,
while also copying several copies of pkgsrc in the background, and haven't
encounted any hard hangs. The LFS file system is about 3GB in size,
configured to use a 16k block size, and mounted on /mnt. I've tried
a GENERIC.MPDEBUG kernel (with DEBUG enabled for the LFS files), and a
GENERIC.MP kernel. All while I was doing work in X. And with
a default kern.maxvnodes.
This is a bit of a different test setup than what I was running when I
did experience the hangs. Previously I had a 30GB file system, and it
was mounted on /home -- my home directory was there.
I had wondered if it might it be special files that cause a problem
(e.g., sockets)? All the files currently on /mnt are files, directories
or symlinks. Except my /home has nothing except for a single socket
file from gqmpeg, which I haven't used in years; copying it onto /mnt
hasn't made a difference.
Might these hard hangs have something to do with the size of the
BTW: Is there a way to cause writes to be transparently duplicated to
multiple filesystems? Something like mount_union, but that sends all
writes to both file systems? Then I could try running LFS in parallel
with FFS with little(r) risk of data loss.
Brian de Alwis | Software Practices Lab | UBC | http://www.cs.ubc.ca/~bsd/
"Amusement to an observing mind is study." - Benjamin Disraeli