Date: Wed, 20 Mar 2002 21:38:04 +0100 From: Poul-Henning Kamp <phk@critter.freebsd.dk> To: undisclosed-recipients:; Subject: UFS2, GEOM & DARPA - don't get all excited, OK ? Message-ID: <13286.1016656684@critter.freebsd.dk>
next in thread | raw e-mail | index | archive | help
------- =_aaaaaaaaaa Subject: UFS2, GEOM & DARPA - don't get all excited, OK ? From: Poul-Henning Kamp <phk@freebsd.org> Date: Wed, 20 Mar 2002 21:38:04 +0100 Message-ID: <13286.1016656684@critter.freebsd.dk> Sender: phk@critter.freebsd.dk Bcc: Blind Distribution List: ; MIME-Version: 1.0 Ok, Kirk and I thought it would stirr a buzz once we even mentioned "UFS2" so let me set the record straight, (or at least firmly crooked): UFS2 is UFS Extended Attributes in the inodes. That's it. No more, no less. In particular that means: No journaling. No Btree-directories, no in-directory inodes. Because we need to increase the size of the inode we chose to "run a revision" and call it UFS2, and obviously, we will make sure all relevant fields get the space they need, or at reserve space for their growth as part of this. Specifically inode numbers, block numbers and time_t will have 64 bit available. We may attempt a couple of neat tricks at the same time: per inode blocksize and lazy inode initializtion. The former might improve I/O performance down the road, the latter speed up newfs and fsck. We will try to avoid forking sys/ufs/ufs if we can, we may have to push some stuff from ufs to ffs to make that happen. In practice this also means a sweep through all the userland stuff to make it work with UFS2: newfs, fsck, tunefs, quotactl, sysinstall etc (that's largely going to be my problem). Incidently there is a not insignificant crossover between UFS and disklabels and from there into GEOM, so some of the things I will be doing in that corner will be hard to attribute only to one or the other of these two DARPA funded projects. The one thing we RSN will cry to have is endian/wordsize agnostic UFS. That is not part of the DARPA project description and it is unlikely to "sneak" into it. We have it in mind though. (The NETBSD work is not uninteresting, but both Kirk and I feel that there might be a better and even more general solution.) And this is yet a problem with significant crossover to the issues I have in GEOM with reading labeling data structures of disks in alien formats so some of the technology I plan for GEOM might be applicable to UFS as well. Finally, If you are in the US and you think it is a good thing that DARPA sponsors stuff like this, don't forget to say so to people who might, directly or indirectly, provide feedback to DARPA. I hope this answers the FAQ on UFS2, GEOM and all that. -- Poul-Henning Kamp | UNIX since Zilog Zeus 3.20 phk@FreeBSD.ORG | TCP/IP since RFC 956 FreeBSD committer | BSD since 4.3-tahoe Never attribute to malice what can adequately be explained by incompetence. ------- =_aaaaaaaaaa-- To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?13286.1016656684>