From owner-freebsd-hackers Wed Aug 7 05:51:14 1996 Return-Path: owner-hackers Received: (from root@localhost) by freefall.freebsd.org (8.7.5/8.7.3) id FAA14556 for hackers-outgoing; Wed, 7 Aug 1996 05:51:14 -0700 (PDT) Received: from eac.iafrica.com ([196.7.192.149]) by freefall.freebsd.org (8.7.5/8.7.3) with SMTP id FAA14178 for ; Wed, 7 Aug 1996 05:48:15 -0700 (PDT) Received: (from rnordier@localhost) by eac.iafrica.com (8.6.12/8.6.12) id OAA01558; Wed, 7 Aug 1996 14:45:49 +0200 From: Robert Nordier Message-Id: <199608071245.OAA01558@eac.iafrica.com> Subject: Re: SAMBA and DOS file systems To: rpt@miles.sso.wdl.lmco.comm (Richard Toren) Date: Wed, 7 Aug 1996 14:45:47 +0200 (SAT) Cc: hackers@FreeBSD.org In-Reply-To: from "Richard Toren" at Aug 6, 96 07:04:09 pm X-Mailer: ELM [version 2.4 PL24 ME8a] Content-Type: text Sender: owner-hackers@FreeBSD.org X-Loop: FreeBSD.org Precedence: bulk Richard Toren wrote: > SAMBA seems to have really picked up as a topic these last couple of weeks. > > My question concerns the file systems that SAMBA uses to support the > remote requests. This machine is used for both FBSD and WFWG 3.11. I am > planning to use SAMBA as a file server for the WFWG files stored in the > DOS partitions, and used on another WFWG machine. > > I have been following (over the last year) the saga of the fitness of > using the DOS FS. Some people seem to have a disaster, and others say it > works just fine. In my case I have 3 DOS partitions (1 primary and 2 > extended) of 250 Mb each. > > Is SAMBA safe with DOS partitions? Should the FreeBSD partitions be the > only ones used? > > Any experiences are guidelines would be appreciated.... I can't comment about SAMBA, but the following is reasonably definitive regarding the state of the msdosfs itself: * Use of the msdosfs to access certain DOS partitions may cause corruption to a non-DOS partition not necessarily on the same drive. * The msdosfs code permits numerous race conditions and deadlocks, and also has major bugs at a more fundamental level (messing up when moving/renaming directories, as one example). Corruption to non-DOS partitions typically occurs in the case of 64-head IDE drives accessing large DOS partitions with 64-sector cluster sizes. From data I have gathered, use of FIPS is _not_ a necessary or sufficient condition for this. Apparently /dev/wd is. A while ago, I spent a couple of days investigating the corruption problem. However I have only four different non-SCSI drives here, and was unable to corrupt any of them. As I was lacking an EIDE/translating controller at the time, I didn't pursue this. The non-DOS partition corruption problem is a system configuration issue and is avoidable, with planning. Avoid it, and the msdosfs is perfectly usable for simple data transfers (one user copying individual files to and from a DOS filesystem). If you value your data, I'd say the msdosfs is pretty much _unusable_ for anything more complex. -- Robert Nordier