Hello Alan!
Ah, excellent. Looks like we can bump the version already as version incompatibilities have arisen in the past and code is there to already support this. In sys/xfs/ext2/main.c I can see MSG_OLDMINT which we can bump for the XFS module. minixfs has something similar. But it's what to do in the kernel when loading older XFS modules ? Frank ?
This is not supported (e.g. the kernel always load the xfs/xdd; only the modules have a version check).
Yes, I took a deeper look into the code. It looks good. A question about this (I don't checked it): it is possible to catch read/write request at a igher level instead inside the xfs? This make much more sense (and avoid the problem if a newer kernel load an older xfs).
That's what I meant. I like to move the EISDIR error check into the filesystem layer. And, if this is possible (not checked), the open() on the directory shall be handled at the filesystem layer too.
Oh, and as for loading older XFS modules. That's not going to work really, as we're going to be allowing a read-only directory through to the XFS. Older XFS's don't understand that.
This avoid any modification on the xfs, we don't have any compatibility problems and the directories are handled generically inside the kernel.
Regards, Frank -- ATARI FALCON 060 // MILAN 060 ----------------------------- http://sparemint.org/ e-Mail: fnaumann@boerde.de