Date: Mon, 11 May 1998 15:06:41 -0700 From: Kirk McKusick <mckusick@McKusick.COM> To: Ollivier Robert <roberto@keltia.freenix.fr> Cc: julian@whistle.com, luoqi@watermarkgroup.com (Luoqi Chen), current@FreeBSD.ORG Subject: Why Soft Updates are not a mount option Message-ID: <199805112206.PAA28767@flamingo.McKusick.COM> In-Reply-To: Your message of "Sat, 09 May 1998 23:36:02 EDT." <9805100336.AA06453@watermarkgroup.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Soft Updates are set in the superblock with tunefs for several reasons: 1) This is an interim measure during the testing phase of soft updates. In the long run, they will always be used as the normal course of events. I do not want legacy mount options lying around. 2) The soft update code is not prepared to be turned on when the filesystem is active. It currently has no code to find all files actively being written and building up the necessary dependency information for them. Writing such code is non-trivial and not a worthwhile exercise in my opinion, especially given goal #1. 3) Fsck uses different algorithms for cleaning up on filesystems run with soft updates. It is much more reliable to have it check for the bit in the superblock than it is to try and figure out whether the flag was set when it was mounted. Also, the algorithms can only be used if the filesystem was continuously run with soft updates throughout the time that it was mounted. If it could be updated, that would somehow have to be recorded. Kirk McKusick 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?199805112206.PAA28767>