From owner-freebsd-hackers Thu Mar 2 16:33:26 2000 Delivered-To: freebsd-hackers@freebsd.org Received: from cube.gelatinous.com (cube.gelatinous.com [207.82.194.150]) by hub.freebsd.org (Postfix) with SMTP id 7F8F837B5F1 for ; Thu, 2 Mar 2000 16:33:22 -0800 (PST) (envelope-from aaron@gelatinous.com) Received: (qmail 50245 invoked by uid 1000); 3 Mar 2000 00:33:18 -0000 Date: Thu, 2 Mar 2000 16:33:18 -0800 From: Aaron Smith To: freebsd-hackers@freebsd.org Subject: user-space filesystems Message-ID: <20000302163318.F7995@gelatinous.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii X-Mailer: Mutt 1.0.1i Sender: owner-freebsd-hackers@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG hello, i've done some searching and i've seen discussion of userland fs before. has there been any progress in the user-space filesystem area? i have a nifty project and i would like to avoid using loopback NFS; have we got anything akin to linux's userfs yet? if freebsd doesn't have this capability, where would a good place to start be on loopback NFS? maybe somebody has a loopback NFS skeleton i can start from? any pointers/discussion would be helpful. aaron here's one of the messages that made me say "yeah, like that!": > Date: Sat, 17 Jul 1999 14:57:45 -0400 > From: "David E. Cross" > To: freebsd-hackers@freebsd.org > Subject: USFS (User Space File System) > Message-ID: <199907171857.OAA81681@cs.rpi.edu> > > I am looking at a project that will require a user based process to > interact with the system as if it were a filesystem. The traditional way I > have seen this done is as the system NFS mounting itself (ala AMD). I > would really like a more clean approach to this. What I am interested in > is a 'User Space File System' that would interact with a user process in a > similiar manor to how nfsd's do. A process would issue a mount (ok, this > is different than NFSDs), then it would make a special system call with a > structure, that call would return whenever a request was pending with the > structure filled in with the appropriate information. The user process > would fulfill the request, pack the return data into the structure and call > kernel again. > > I have a number of questions on more specific ideas (like caching, > inode/vnode interaction, etc). But I am just feeling arround for what > people think about this. Any ideas/comments? To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message