Re: [OpenAFS] Re: OpenAFS kernel panic
On Wed, Oct 31, 2012 at 11:15:43AM -0500, Andrew Deason wrote: > On Mon, 29 Oct 2012 14:44:15 -0400 > Jack Neely wrote: > > > > You can save us a little time by providing the disassembly of > > > afs_Conn. You can get this by running > > > > > > objdump -d -r /path/to/libafs.ko > /some/file > > > > Attached. > > This looks like 'objdump -d', not 'objdump -d -r', but okay. > > This is issue 130714. Were you moving fileservers around at the time, or > trying to change their addresses or something? Yes, this was in the middle of lots of volume moves so that we could upgrade the storage space allocated to each AFS server. Jack > A fix is already in the tree and will be in the next 1.6 release. > > -- > Andrew Deason > adea...@sinenomine.net > > ___ > OpenAFS-info mailing list > OpenAFS-info@openafs.org > https://lists.openafs.org/mailman/listinfo/openafs-info -- Jack Neely Linux Czar, OIT Campus Linux Services Office of Information Technology, NC State University GPG Fingerprint: 1917 5AC1 E828 9337 7AA4 EA6B 213B 765F 3B6A 5B89 ___ OpenAFS-info mailing list OpenAFS-info@openafs.org https://lists.openafs.org/mailman/listinfo/openafs-info
Re: [OpenAFS] Re: NFS export oddness
On Fri, Nov 2, 2012 at 10:54 AM, Andrew Deason wrote: > On Thu, 1 Nov 2012 15:59:13 -0700 > Timothy Balcer wrote: > >> unfortunately, when I try to export the /afs volume, I get this: >> >> exportfs: /afs does not support NFS export >> >> and when I do the explicit fs exportafs nfs command, I get this: >> >> Sorry, the nfs-exporter type is currently not supported on this AFS >> client > > The NFS xlator is not built by default on Linux. The interfaces we need > to use to enable that functionality are no longer available to us as of > kernel 2.6.29 (according to Marc :). If you want to try to build it on a > kernel before then, you can edit src/config/param.linux26.h, remove the > lines that define AFS_NONFSTRANS, and try to build it again. It may or > may not work. (If you really want the kernel xlator to work, you can > also try older OpenAFS client versions if that doesn't work.) Some of the GPL issues have changed since (d_obtain_alias is no longer GPL only), but last time I checked a few years ago there were still sunrpc and maybe other functions that were still off limits. For newer kernels, it's also very likely that this code has bit-rotted and needs some TLC before it will compile and run, since it's been excluded from the build for a long time. Marc ___ OpenAFS-info mailing list OpenAFS-info@openafs.org https://lists.openafs.org/mailman/listinfo/openafs-info
Re: [OpenAFS] Re: NFS export oddness
On Fri, Nov 2, 2012 at 10:54 AM, Andrew Deason wrote: > On Thu, 1 Nov 2012 15:59:13 -0700 > Timothy Balcer wrote: > >> unfortunately, when I try to export the /afs volume, I get this: >> >> exportfs: /afs does not support NFS export >> >> and when I do the explicit fs exportafs nfs command, I get this: >> >> Sorry, the nfs-exporter type is currently not supported on this AFS >> client > > The NFS xlator is not built by default on Linux. The interfaces we need > to use to enable that functionality are no longer available to us as of > kernel 2.6.29 (according to Marc :). If you want to try to build it on a > kernel before then, you can edit src/config/param.linux26.h, remove the > lines that define AFS_NONFSTRANS, and try to build it again. It may or > may not work. (If you really want the kernel xlator to work, you can > also try older OpenAFS client versions if that doesn't work.) and if you're building your own kernel you may be able to make whatever interfaces you want available to yourself, of course, but this is an expert-level thing. > Your other options are to use the translator on a commercial unix > instead (it's known to work on Solaris), Solaris is actually the only other place it's likely to work. -- Derrick ___ OpenAFS-info mailing list OpenAFS-info@openafs.org https://lists.openafs.org/mailman/listinfo/openafs-info
[OpenAFS] Happy Birthday OpenAFS!
Folks, It went unremarked here, so I'm going to remark. :) Twelve years ago yesterday, on November 1 2000, IBM released AFS into the open source community, giving birth to OpenAFS. I believe Derrick was the first (or among the first) to download it from IBM shortly after midnight that morning. Happy 12th birthday OpenAFS! Evan -- Evan D. Macbeth – Director of OpenAFS Sine Nomine Associates emacb...@sinenomine.net - 703–723-6673 x1223
[OpenAFS] Re: NFS export oddness
On Thu, 1 Nov 2012 15:59:13 -0700 Timothy Balcer wrote: > unfortunately, when I try to export the /afs volume, I get this: > > exportfs: /afs does not support NFS export > > and when I do the explicit fs exportafs nfs command, I get this: > > Sorry, the nfs-exporter type is currently not supported on this AFS > client The NFS xlator is not built by default on Linux. The interfaces we need to use to enable that functionality are no longer available to us as of kernel 2.6.29 (according to Marc :). If you want to try to build it on a kernel before then, you can edit src/config/param.linux26.h, remove the lines that define AFS_NONFSTRANS, and try to build it again. It may or may not work. (If you really want the kernel xlator to work, you can also try older OpenAFS client versions if that doesn't work.) Your other options are to use the translator on a commercial unix instead (it's known to work on Solaris), or try a userspace NFS server, as Christof mentioned. We would like to sometime get a userspace NFS translator that has the afs-specific bits, but nobody seems to want it enough for that to actually happen. -- Andrew Deason adea...@sinenomine.net ___ OpenAFS-info mailing list OpenAFS-info@openafs.org https://lists.openafs.org/mailman/listinfo/openafs-info
Re: [OpenAFS] [workaround] NFS export oddness
Instead of this unionfs-workaround, you could also use a user-space nfs-server, like unfs http://unfs3.sourceforge.net/ but of course, you'll loose the kerberos-security as well. Christof Am Thu, 1 Nov 2012 17:31:31 -0700 schrieb Timothy Balcer : > I found a workaround for now.. I hope it finds some use amongst the > community, assuming the linux NFS translator really isn't working > properly (as seems to be intimated in the emails I recently found) > > You can use unionfs with an afs volume and a normal volume, and then > NFS export THAT unionfs directory, at which point the files will be > accessible. So if you just want to export afs via nfs, you can > unionfs it with an empty directory, and then NFS export that one. > > so for example > > unionfs-fuse /afs/foo.com/home/bar:/empty /test > > and in /etc/exports > > /test *(options) > > This works for read.. I haven't tested for write. > > So what I am going to do is have my application write to the local > volume, and have a process flush that directory to the afs volume > from time to time. The unionfs volume won't change at all, and will > present a uniform appearance. > > I'm still interested in NFS translation though. This would be a > securoty problem if the IPs were not entirely sequestered. ___ OpenAFS-info mailing list OpenAFS-info@openafs.org https://lists.openafs.org/mailman/listinfo/openafs-info
