From owner-freebsd-stable Thu Aug 3 14:26:35 2000 Delivered-To: freebsd-stable@freebsd.org Received: from alcanet.com.au (mail.alcanet.com.au [203.62.196.10]) by hub.freebsd.org (Postfix) with SMTP id 0AA0D37B820 for ; Thu, 3 Aug 2000 14:26:21 -0700 (PDT) (envelope-from jeremyp@pc0640.alcatel.com.au) Received: by border.alcanet.com.au id <115835>; Fri, 4 Aug 2000 07:25:49 +1000 Date: Fri, 04 Aug 2000 05:41:04 +1000 From: Peter Jeremy Subject: Re: Cached versus non cached disk I/O In-reply-to: <396B8DC4.28CD9727@geodesic.com>; from pac@geodesic.com on Tue, Jul 11, 2000 at 04:12:36PM -0500 In-reply-to: ; from tom@uniserve.com on Tue, Jul 11, 2000 at 09:06:31PM -0700 To: Paul Coyne , Tom Cc: freebsd-stable@FreeBSD.ORG Message-Id: <00Aug4.072549est.115835@border.alcanet.com.au> MIME-version: 1.0 X-Mailer: Mutt 1.0i Content-type: text/plain; charset=us-ascii References: <396B8DC4.28CD9727@geodesic.com> Sender: owner-freebsd-stable@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG [catching up on some old mail] On Tue, Jul 11, 2000 at 04:12:36PM -0500, Paul Coyne wrote: >Is a mkdir cached? It depends... The following table lists the possible UFS mount modes, together with their caching behaviour, in descending order of FS integrity: mode data metadata softupdates async* async* sync sync sync `normal' async sync async async async Where sync means write-though caching and async means write-back caching. async* means that write-back caching is used, but the write ordering is controlled to ensure FS consistency on disk. mkdir(2) and symlink(2) entail metadata updates, so they are cached for both softupdates and async mounts. >Does a mkdir call from an NFS client wait for a positive response from >an NFS server prior to handling the next request? Again, this depends... By default, NFSv2 is fully synchronous - the client will wait until the server has committed the change to disk. NFSv3 allows some asynchronous operations, though I'm uncertain of the details. The defaults can be over-ridden by a sysctl. On Tue, Jul 11, 2000 at 09:06:31PM -0700, Tom wrote: > It >should be obvious that write-buffering metadata can cause problems, even >with softupdates, though softupdates is clearly better than async. Not quite. softupdates is actually more robust than a normal FS mount (and far more robust than async). The softupdates code controls the ordering of both data and metadata writes to ensure that the FS on disk is always internally consistent. With a normal mount, the metadata is mostly internally consistent, but is not necessarily consistent with the data. Peter To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message