>
>
> So we noticed that EWRONGFS got remapped to ENOSYS, but that is purely
> a mount error, not an unmount. But, could this be to do with the errno
> remapping changes in xfs_linux.h on Mar 13th?
>
There is a code path in xfs_unmount_flush() which has problems not
being able to destinguish between EFSCORRUPTED and EIO. I'm in the
process of moving these two, and re-testing.
Mark
> Steve
>
> > Since I malformed the address this didn't make it to the list:
> >
> > >
> > >
> > > Danny writes:
> > > => All,
> > > => Mar 29 03:19:15 dsc_proto_1 kernel: XFS unmount got error 38
> > > => Mar 29 03:19:15 dsc_proto_1 kernel: linvfs_put_super: vfsp/0xc3bc2ae0
> > > => left dangling!
> > > => Mar 29 03:19:15 dsc_proto_1 kernel: VFS: Busy inodes after unmount.
> > > => Self-destruct in 5 seconds. Have a nice day...
> > > => Mar 29 03:19:56 dsc_proto_1 kernel: XFS: Filesystem has duplicate
> > > UUID
> > -
> > > => can't mount
> > >
> > > => Now: I'm pretty sure that a reboot would allow the mount to
> > > succeed;
> > in
> > > => other words, /dev/md1 is still "mounted" in some way in the kernel;
> > > => hence the dup UUID.
> > >
> > > If xfs_unmountfs doesn't get called, the uuid for the mounted xfs fs won't
> > > get removed from the uuid table, and further mounts of that filesystem
> > > will be rejected. There are a couple of paths in which the unmount can
> > > fail to call xfs_unmountfs, but they shouldn't happen unless something
> > > is really wrong anyway.
> > >
> > > What I'm really curious about is why the unmount failed in the first
> > > place - that definitely shouldn't be happening.
> > >
> > Curious, I just hit this running a forced shutdown test. Different
> > error, same effect. In my case, it definitely won't follow the code
> > path to call xfs_unmountfs(). In either event, it doesn't look like
> > the filesystem is mounted in any useful sense.
> >
> > I am curious to know how it got here. Any information would be
> > appreciated.
> >
> > > Can you give us some vitals on your setup? Are you using highmem, quota,
> > > acls, dmapi etc? Running any specific tests?
> > >
> > > => This machine is a test machine only. Is there anything you'd like
> > > me
> > > => to try before I reboot?
> > >
> > > I think it's too late now. If you can work out how to repeat this problem,
> > > we should be able to suggest some ways to debug it.
> > >
> > If I read the code right, there isn't much else which can be done besides
> > rebooting.....
> >
> > Mark
> >
>
>
|