> If Red Hat is a requirement, then what are the perceived obstacles to > making the in-kernel kafs module do all the filesystem stuff, and all > OpenAFS packages need to do is provide userspace tools?
FYI, I have a qemu image working now that boots a kernel and initrd, and then mounts the root filesystem from AFS. It boots with both the OpenAFS.ko module, and with the kafs.ko module. Last time I tried this (several years ago) kafs would kernel panic, so I think there's been improvement. What do scientific linux users need to make this usable? Just a way to get tokens and set ACLs? The only other thing I can think of is the full 'fs' command, and this could be solved by either adding a magic-file based pioctl interface, or some sort of dirty hack involving linking libuafs and making the 'fs' command a full-fledged AFS client that doesn't need anything from the host system other than a path and a UDP connection to the AFS server(s) One of the features of the kafs client is every volume shows up in /proc/mounts, along with the quota of the volume. _______________________________________________ OpenAFS-devel mailing list OpenAFS-devel@openafs.org https://lists.openafs.org/mailman/listinfo/openafs-devel
