milko question

[Love]

Jack Neely <jjneely at eos.ncsu.edu> writes:

> Hey all!
> 
> I few questions about milko.  Can milko read an AFS 3.4 server volume?

On disk, no.

> Is the format of the data files that make up an AFS volume known?

Not really, but by reading the old afs papers and the header-files I belive
it orginized something like this:

/vicepN/vol<vol# padded to 8 har> Contain inode-number of the SmallVnode
  inode (files and symlinks), LargeVnode inode (dirs, it has acls), and
  some summery infomation inode is some order (or just a pointer to a inode
  with summery infomation, doesn't really matter).

The SmallVnode contain the on-disk equvalent of a AFSFetchStatus + a inode
number. LargeVnode the same + AFSOpaque for acl. Don't know how store the
bits. It might be that the acl is stored now in a separete inode, but I
don't know.

All inodes created is taged with extra information (the the volume the
belong to and some other things I don't know) in the inode. This is a
source of pain for transarc (and their users). Resently Digital/Compaq
decided (with osf 4.0d) to use to more fields in the inode (32 bits support
for UID and GID). Then they use the one of the fields transarc stored some
info. The kernel "corrected" any "bad" inodes it would find. Fsck did the
same. Not fun at all. A usual it took transarc half a year to fix the
problem (we had the source to osf and backed out Digital\s changes).

I'm not really sure its that interesting to know in more detail how they
did it, but if some care I wouldn't mind to know more. (Please don't turn
the world upside down to find it the info, we will do diffrently anyway).

Why do you want to read the ondisk bits ? 

Just do a vos backup | vos restore. The dump format isn't known, but its
easier to figure out.

Love