lots of different things....

Chaskiel Grundman cgrundman at gmail.com
Thu Oct 4 23:39:12 EDT 2012


So, for reasons that even I am unclear about, I've been hacking on
kafs/rxrpc things the last couple of days. Some things I have as a
result:

- an rxkad/krb5 aklog (produces "2b" tokens and hands them to the
kernel) (/afs/andrew.cmu.edu/usr18/cg2v/src/kafs/aklog-kafs.c)
- A significant portion of a library providing the client side of
openafs's RX API (/afs/andrew.cmu.edu/usr18/cg2v/src/rxrxrpc-poc)

- some bug reports. These were experienced on 3.0.0 and 3.2.0 (ubuntu)
but a quick look at master makes me think they are still present
there.

   oops (ASSERT) if you sendmsg with msg_name set on an unbound socket
(rx->local is not set in rxrpc_name_to_transport)

   you can't setsockopt after bind(), and I can't figure out why that is needed.

   oops (ASSERT) in rxrpc_key_read (ENCODE_DATA should not be used for
the key, since the keylength should not be copied to the output)

   kafs does not recover from dead servers (as far as I can tell,
nothing looks at server->fs_dead_jif and resets server->state)



More information about the linux-afs mailing list