Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 11 May 1998 00:16:39 +0000 (GMT)
From:      Terry Lambert <tlambert@primenet.com>
To:        roberto@keltia.freenix.fr (Ollivier Robert)
Cc:        current@FreeBSD.ORG
Subject:   Re: Soft Updates
Message-ID:  <199805110016.RAA29954@usr05.primenet.com>
In-Reply-To: <19980510004603.A16999@keltia.freenix.fr> from "Ollivier Robert" at May 10, 98 00:46:03 am

next in thread | previous in thread | raw e-mail | index | archive | help
> > you need to use the -n option of tunefs on an unmounted filesystem
> > it's not a mount option..
> 
> Why not BTW ? async/noasync are mount options so one could expect
> softupdates to be the same. Is there an reason for the current scheme (not
> that I mind but it would make switch it on/off easier, especially for /).

The argument is that when you change a mount option on the FS, the
in core buffers do not get flushed.

In this case, you could have some writes pending upon which a newly
scheduled soft updates write operation should depend, but since the
write was scheduled before the soft updates dependency tracking was
engaged, the dependency is unrecoverable.

There are potentially some issues with FS metadata cached off of the
device vnode, as well (this is more of a FreeBSD architectural issue
that causes problems unless it is carefully addressed; among other
things, maximum FS size is limited by this; otherwise the maximum FS
size would be limited vy the offset size, not the maximum VM object
size; you would still have a file size limited by VM object size,
however).


It's very possible to work around this.  Specifically, the conversion
of an FS from read-write to read-only effectively does an unmount
followed by a mount under the covers.

The reason Kirk gave for not doing this was that the "under the covers"
code was rather gross and convoluted, and it would have had to have
been (effectively) duplicated once again for soft updates being switched
on and off.


I tend to agree that the code is gross and convoluted.  See my post
in the "NFS exporting an MFS" thread on what could be done to the
VFS ops to make this easier to deal with, work for all FS's, and in
general, more robust.


					Terry Lambert
					terry@lambert.org
---
Any opinions in this posting are my own and not those of my present
or previous employers.

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-current" in the body of the message



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199805110016.RAA29954>