From: Matthias Andree (madt.e-technik.uni-dortmund.de)
Date: Fri Aug 01 2003 - 16:26:56 CDT

• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

On Fri, 01 Aug 2003, Wietse Venema wrote:

> > > This is not robust. Postfix will change and there is no guarantee
> > > that today's cscope result will be valid tomorrow.
> >
> > Of course not. There is no guarantee that tomorrow's Postfix doesn't
> > contain sprintf rather than vsprintf either.
>
> I will not reply to such a silly suggestion.

I don't see much of a difference there. These functions are supposed to
suppress errors. One error is the non-idempotence of NFS method, the
other is non-synchronicity of some local file system method.

At the moment, it appears Linux or BSD + SOFTDEP users are better off
mounting their Postfix file systems via NFSv3 knfsd onto a different
directory on the same machine and stuff "sync" into their
/etc/exports... yes, this looks silly, but might actually provide the
necessary semantics...

> By the way, mkdir needs to be synchronous too, because Postfix will
> create missing parent and parent-parent directories if it has to.
>
> So your sync_mkdir() would have to invoke sane_mkdir() etc.

A bit harder, but still no problem, open and fsync() the freshly created
dir (Linux 2.4 XFS 2.2/2.4 ext3fs 2.4 reiserfs BSD SOFTDEP) or its
respective parent (Linux 2.2 or ext2fs).

It IS ugly, but POSIX doesn't prescribe synchronous directory
modification semantics as much as we'd like to have these. Looking at
past experience, fsync() is the method that's the most likely to be
implemented by $RANDOM_LINUX_FILESYSTEM - and it's the magic that works
with most recent file systems including SOFTDEP/softupdates.

It used to work with ext2fs, this system had "fsync() == sync()"
semantics in the past, not sure if that's still true.

I've started a poll on linux-kernel again, we'll see how many people
answer and for what file systems we'll receive answers. I'll resend that
stuff to the respective maintainers in a couple of days when answers are
missing.

Check the mail with these headers for details:

From: Matthias Andree <matthias.andreegmx.de>
Subject: File system synchronous completion poll.
To: Linux-Kernel mailing list <linux-kernelvger.kernel.org>
Date: Fri, 1 Aug 2003 23:23:58 +0200
Message-ID: <20030801212358.GA18944merlin.emma.line.org>

--
Matthias Andree

• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]