From owner-freebsd-fs Fri Jul 17 11:59:33 1998 Return-Path: Received: (from majordom@localhost) by hub.freebsd.org (8.8.8/8.8.8) id LAA00371 for freebsd-fs-outgoing; Fri, 17 Jul 1998 11:59:33 -0700 (PDT) (envelope-from owner-freebsd-fs@FreeBSD.ORG) Received: from parkplace.cet.co.jp (parkplace.cet.co.jp [202.32.64.1]) by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id LAA00366 for ; Fri, 17 Jul 1998 11:59:32 -0700 (PDT) (envelope-from michaelh@cet.co.jp) Received: from localhost (michaelh@localhost) by parkplace.cet.co.jp (8.8.8/CET-v2.2) with SMTP id SAA11011; Fri, 17 Jul 1998 18:57:49 GMT Date: Sat, 18 Jul 1998 03:57:49 +0900 (JST) From: Michael Hancock To: "Alton, Matthew" cc: "'FreeBSD-fs@FreeBSD.ORG'" Subject: RE: LFS Hacking In-Reply-To: <31B3F0BF1C40D11192A700805FD48BF901776614@STLABCEXG011> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-freebsd-fs@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.org On Fri, 17 Jul 1998, Alton, Matthew wrote: > Is Heideman's (or anyone else's) FS stacking being worked on > right now? I was working on fixing layering violations with ref counting and locks on the VOP name ops. Heidemann's framework requires a symmetry such that the ops are the same above and below. These layering violations break the symmetry indirectly by side-effect. Terry uses the term "making it reflexive" to describe coding such that where the caller is responsible for releasing any references it takes. This is important in the case of stacking. He gets a lot of heat for using these odd terms but I find it convenient to use a single word to describe something that would normally take a sentence or two. I got most of it done before all these projects came along at work and I'm just too buried right now to finish the final 3 cases. One of these will be the toughest, VOP_RENAME. I understand the ufs_rename code now enough to do it correctly. I actually had to print the source out and highlight all the gotos to different exit points among other things before I had a clue as what was happening in there. Union_rename might be a good challenge, though. I expected NFS to be hard but now I expect it to be one of the easier ones. It's just bigger code, I mean the guy just didn't believe in subroutines or something. If I can just get these done then I can get rid of all the WILLRELE infrastructure. After this is done, work will be need to done to deal with vnode and cached object coherency, where cached objects are data pages or attributes such as the file size. John Heidemann has a paper on an approach that incorporates a central cache manager. It's a good paper but I'm not entirely convinced that his hand-off approach to locking and distributed protocols is really workable. > I tell ya, I spent a good 10 minutes last night in lamentation > mode over the > egregious and unforgivable termination of the ARPA BSD funding. > I'm > acclimating to the whole cathedral/bazaar thing but I just jones > away > something awful for the academic rigor and coordination that the > old-timers > had. I'm going to think up a way to do something about this. I'm still impressed by cathedrals and these days you can see Michelangelos in a bazaar now and then. Regards, Mike To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-fs" in the body of the message