Would like to see the GFS Patches included in the main kernel so that GFS is an easy add-on. ftp://ftp.globalfilesystem.org/pub/GFS/4.0/releases/ http://www.sistina.com/gfs/software/ [mebrown@carthage linux_2_4]$ ls -la total 52 drwx--S--- 2 mebrown mebrown 4096 Jan 1 12:50 . drwx--S--- 4 mebrown mebrown 4096 Jan 1 12:50 .. -rw------- 1 mebrown mebrown 1877 Dec 31 16:12 README -rw------- 1 mebrown mebrown 3904 Dec 31 16:12 bufcache.patch -rw------- 1 mebrown mebrown 2950 May 26 2000 cdb16.patch -rw------- 1 mebrown mebrown 679 May 26 2000 exportscsi.patch -rw------- 1 mebrown mebrown 17817 Dec 31 16:12 inode.patch -rw------- 1 mebrown mebrown 2062 Dec 31 16:12 ksyms.patch -rw------- 1 mebrown mebrown 342 Dec 31 16:12 task.patch From the GFS kernel patches README: Patches to the Linux Kernel 2.4 ================================= These patches are against 2.4.0-prerelease Patches: ---------------- [These are the essential ones to get GFS working under 2.4.xx kernels] bufcache.patch - Adds callbacks to buffer cache for GFS. Once GFS has migrated to using the page cache entirely, this will not be required. ksyms.patch - Exports symbols required by GFS. This patch contains the GFS specific changes that will be required to the kernel. From time to time this patch will be submitted to the main kernel tree, so if bits of it don't apply to your kernel source, it may be because its made its way into the main tree already. task.patch - Adds a hook to the task_struct to hang the GFS transaction structure from. inode.patch - Adds an extra argument to read_inode() to handle the 64bit inode numbers cleanly. N.B. This isn't the "right" way to do this, so this will change. Old Patches: ---------------- [These are not uptodate compared with the latest kernel version and may not still be required/working] exportscsi.patch - exports "rscsi_disks" symbol for Dlock. cdb16.patch - A version of cdb16.patch for 2.3.xx kernels
Date: Tue, 9 Jan 2001 12:04:30 -0600 From: Ken Preslan <kpreslan> To: michael_e_brown.com Cc: sass, gfs Subject: Re: Status of 2.4 patches > 1) How stable are the 2.4 patches, and how close do you think they are to > what the final patchset will need to look like. In particular, what is the > status of the exportscsi.patch and cdb16.patch (listed under 'Old > Patches') The old patches, exportscsi.patch and cdb16.patch, are out of date and need to be redone when we port Pool to 2.4. (But they aren't needed if you're not doing Pool.) Bufcache.patch, ksyms.patch, and inode.patch are close, but both need a bit more work. task.patch is pretty close to its final form. nfs.patch is just a bug fix. It has been submitted and will hopefully be included shortly. > 2) What is the status of pushing the GFS kernel patchset to Linus for > inclusion in the mainstream kernel. (Hint: It would be _really_ helpful to > me if you were to push the GFS patches to Alan Cox for inclusion in the > 2.4ac series, and in the 2.2 official kernel.) task.patch and bufcache.patch will both probably go in (or an equivelant will go in) when Stephen Tweedie get ext3 into the kernel. inode.patch needs to be synced up with what Al Viro wants to do with the new iget4() interface. ksyms.patch doesn't really do anything controversial, so we may submit that soon. -- Ken Preslan <kpreslan> Code Monkey Sistina Software, Inc.
GFS is now non-free software. The chances of us doing this therefor went from "good when the code quality improves sufficiently" to "no chance whatsoever, sorry".
http://www.opengfs.org ?? I see Alan and Arjan are active on the lists.
GFS is not in a state where it is supportable. Sistina GFS 4.1 has dozens and dozens of serious (some exploitable) bugs, OpenGFS is cleaning this up a lot now, but lack of equipment slows it down a bit.