[OpenAFS-devel] documentation on afs internals

[aeneous@speakeasy.org]

> I am currently curious about the format of read only volumes and read write
> volumes. More specifically, how they differ and what they look like on disk. Any
> pointers are appreciated.

Readonly and readwrite volumes do not differ significantly on disk.  There is a bit in the header that tells the fileserver that the volume is readonly, but that's it.

A volume consists of the header file (Vxxxx.vol) and a bunch of inodes.

There is one inode per file, one inode per directory, and one inode per unique ACL.  These are native filesystem inodes, not some AFS-specific structure.

The volume header file contents are defined in the vol/volume.h file.

When you read that file, you will note that the volume header also references several additional inodes.   The non-obvious ones are the "large vnode" and "small vnode" file.  These are lists of all the directory and file inodes in the volume, and are used for translating the AFS "vnode number" to a native inode number.