*BSD News Article 22225


Return to BSD News archive

Path: sserve!newshost.anu.edu.au!munnari.oz.au!news.Hawaii.Edu!ames!saimiri.primate.wisc.edu!news.crd.ge.com!sunblossom!knight.vf.ge.com!news.ge.com!combs!steve
From: steve@@salem.ge.com (Stephen F. Combs)
Newsgroups: comp.os.386bsd.questions
Subject: Re: what is fs_clean for?
Date: 11 Oct 1993 16:49:13 GMT
Organization: GE Drive Systems, Salem, VA
Lines: 80
Distribution: world
Message-ID: <29c2q9$32v@alva.ge.com>
References: <BLYMN.93Oct3175636@mallee.awadi.com.au>
Reply-To: steve@@salem.ge.com
NNTP-Posting-Host: 3.29.5.200

In article 93Oct3175636@mallee.awadi.com.au, blymn@awadi.com.au (Brett Lymn) writes:
>>>>>> On 1 Oct 1993 17:10:28 GMT, mark@elea.umd.edu (Mark Sienkiewicz) said:
>Mark> NNTP-Posting-Host: elea.umd.edu
>
>Mark> In article <28fmis$12b9@acsc.com>, Jerry Chen <jerry@acsc.com> wrote:
>>During the mount time, what should be done if the file system is not clean?
>>Should the mount request be rejected or should the mount succeed?  Thanks
>>for the answer.
>
>Mark> There are two answers to this:
>
>Mark> WRONG:	Reject the mount of filesystems that are not clean.
>
>Mark> RIGHT:  Mount it anyway.
>
>Mark> You may detect a bit of bias in my statements. :)
>
>And IMHO you are wrong :-)
>
>Mark> The argument for rejecting it is that you don't know that the filesystem
>Mark> is clean - by mounting it, you can make it worse.  Also, you might have
>Mark> a detrimental effect on the rest of the system.
>
>Exactly.
>
>Mark> The argument for allowing the mount is that the filesystem probably is not
>Mark> in very bad shape.  You can *safely* run for *months* with blocks missing
>Mark> from the free list or unreferenced inodes allocated.  So I don't want my
>Mark> entire system failing (e.g. can't mount /usr) because I lost a temporary file.
>
>Similarly I do not want my system crashing down around my ears some
>months down the track because a block allocated to /usr/bin/login just
>happened to make it's way onto the free list.
>
>Mark> I also don't like systems that say "I won't do what you told me to do because
>Mark> somebody programmed me to be smarter than you".  Computers that say that
>Mark> are lying.
>
>True in some cases but you would have a difficult time convincing me
>that you could inspect a file system and tell me that there are no
>inconsistencies that are not going to cause a catastrophe later.
>
>Mark>  Maybe *I* know that the damage is not bad enough to prevent me
>Mark> from doing what I want to do.
>
>And how would you know this?
>
>Mark> Ideally, one of these two behaviours could be selected by an option in
>Mark> the config file.  e.g.
>Mark> 	options NO_MOUNT_DIRTY			# behaviour I thought was wrong
>Mark> 						# but some people might like
>
>Nope, better would be to make fsck (and other fs tools) pay attention
>to the fsclean flag, if the fs is clean then the fsck justs exits
>otherwise fsck does it's stuff.  The only dodgy is handling the /
>filesystem but that one is special anyway (just consider which
>filesystem the / mount point is in....).
>
>--
>Brett Lymn


Not to sound pedantic, or start a flame-war, BUT you should NEVER mount
a filesystem which is suspect UNTIL you have cleaned it up.  I've seen
too many times when a system will barf at the LEAST acceptable time if you
attempt to bypass stability checks.

---

=======================================================================
Stephen F. Combs
GE Industry Sales & Services		My Employer is in NO way
GE Drive Systems			Responsible or Liable for
Network Services			Any Opinions expressed by
1501 Roanoke Blvd.			Myself.
Salem, VA  24153
Internet E-Mail:	CombsSF@Salem.GE.COM
Novell via Internet:	Combs-SF@Salem.GE.COM
=======================================================================