Date: Wed, 3 Jan 1996 15:53:51 -0700 (MST) From: Terry Lambert <terry@lambert.org> To: scott@statsci.com Cc: chuckr@glue.umd.edu, questions@FreeBSD.org Subject: Re: Can FreeBSD be visible from Win95? Message-ID: <199601032253.PAA15991@phaeton.artisoft.com> In-Reply-To: <199601031906.LAA02229@block.statsci.com> from "Scott Blachowicz" at Jan 3, 96 11:06:19 am
next in thread | previous in thread | raw e-mail | index | archive | help
> > heard lots of folks saying that the msdos filesystem code in FreeBSD is > > not correct, and the reason I heard was because it tried to do buffering > > of the msdos stuff, but the locking mechanism was incorrect, causing > > errors to propagate. This is supposed to be most obvious when doing mv > > commands. I think you could get either Justin Gibbs or Terry (I forget > > his lastname, but you know who I mean) to explain it more authoritatively. > > I've brought this up a couple times before and the result of the > discussions was that there seemed to be a connection between using FIPS to > shrink the DOS slice and the BSD corruptions I was seeing. I believe the > FIPS author was involved in some of the discussion (though possibly not on > the FreeBSD lists??). I think it had something to do with changing the > size of the used portion of the FAT without changing the size of the table > itself (there's a check for this sort of condition in the mtools package > that the FIPS author said shouldn't be made because the condition is > legitimate). That's something that appears to be "legal" but confuses the > FreeBSD code (and, as I noted, MS-stuff and Linux have no problems with > it). I'm no expert in any of this (DOS, file systems, ...). > > And, I think it was Terry Lambert who was suggesting a possible connection > between having the FIPS-shrunk DOS slice before my BSD slice. I was repeating someone else. I believe (from looking at the code) that the expected cluster size is used with the expected cluster count for the disk size. When the cluster size is higher than expected, you get "extra" clusters written out, since it uses the calculated rather than the empirical cluster size/count to write out empirically sized clusters. When this happens (if it happens), you will get clusters of the expected cluster size written out to the expected disk locations, which will be off the end of the disk for the calculated count vs. the empirical size. This will have the effect of corrupting the partition following the DOS partition in the expected cluster locations as if it had been written by DOS on a larger partition. Terry Lambert terry@lambert.org --- Any opinions in this posting are my own and not those of my present or previous employers.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199601032253.PAA15991>