Date: Tue, 06 Feb 2001 20:44:11 +0100 From: Andre Oppermann <oppermann@monzoon.net> To: Matt Dillon <dillon@earth.backplane.com> Cc: Mike Silbersack <silby@silby.com>, Poul-Henning Kamp <phk@critter.freebsd.dk>, Charles Randall <crandall@matchlogic.com>, Dan Phoenix <dphoenix@bravenet.com>, Alfred Perlstein <bright@wintelcom.net>, Jos Backus <josb@cncdsl.com>, freebsd-hackers@FreeBSD.ORG Subject: Re: soft updates and qmail (RE: qmail IO problems) Message-ID: <3A80540B.CFF35C0F@monzoon.net> References: <Pine.BSF.4.31.0102061305430.15212-100000@achilles.silby.com> <200102061920.f16JKdN64467@earth.backplane.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Matt Dillon wrote: > > :On Tue, 6 Feb 2001, Matt Dillon wrote: > : > :> I did a quick search of the qmail site but couldn't find an email > :> address to report the FAQ issue to. If QMail calls fsync() in a > :> reasonable manner, then softupdates is perfectly safe and the QMail > :> FAQ needs to be updated to recommend softupdates rather then > :> disrecommend it. > :> > :> -Matt > : > :The question still reamins about link/unlink/rename. Is a fsync of the > :directory necessary to ensure that they completed properly? As they take > :filenames instead of fds, an fsync after the operation seems > :non-intuitive. The rename manpage seems to imply that the operation is > :synchronous - the other two are ambiguous. > : > :Mike "Silby" Silbersack > > If you rename a file and then fsync() the file's descriptor, you > will guarentee the directory structure. Same with link. If you > you unlink a file and still have an open descriptor you can fsync() > the descriptor to guarentee the directory unlink. It might be a good > idea to ftruncate() the file to 0-length in addition to unlinking it, > prior to calling fsync(), to avoid unnecessary data block writes. Qmail has no need for guaranteed unlink()s. It only needs atomically link()s. > I've never tried fsync()ing a directory descriptor so I don't know > if that would work or not. This is neccessary for ext2 even in sychronous mount. > No other filesystem makes guarentees for these operations either. > Not even normal UFS mounts. -- Andre To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3A80540B.CFF35C0F>