Date: Fri, 25 Mar 2005 14:29:01 +1100 (EST) From: Bruce Evans <bde@zeta.org.au> To: Xin LI <delphij@frontfree.net> Cc: freebsd-audit@FreeBSD.org Subject: Re: [PATCH] nfsserver patch for review Message-ID: <20050325140634.N360@delplex.bde.org> In-Reply-To: <20050324180845.GA26524@frontfree.net> References: <20050324180845.GA26524@frontfree.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On Fri, 25 Mar 2005, Xin LI wrote: > From NetBSD: > > For nfs3 requests, don't gather writes since it's clients' responsibility. > > Will anyone object if I will commit this against -CURRENT? Not me. Is it actually faster and/or more efficient (hopefully both)? Write gathering may wrong even in the nfsv2 case, since the underlying file system should do write clustering, and at least ffs does it, and write clustering at the lowest level works well. BTW, there is lots of vaguely related bogusness involving fsyncing writes: - in the nfsv3 case, the server should just fsync what the client requests, but doesn't always. - the async mount flag is bogus for nfs, and the vfs.nfs.async sysctl is even more bogus. Interaction of these bogusnesses with not always honoring the client's fsync requests give many sub-bugs. Bruce
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20050325140634.N360>