Don't worry, I haven't dropped off the face of the earth, I've been
spending time being pulled in many directions :)
Here's what I have so far. It's the interface polling thread method
discussed earlier. Right now I'm looking at where the AFS code
currently searches the ILL structure to get its interface info (in
osi_SetServerPrefs in afs_server.c and the MTU probing in rx/SOLARIS/
knet.c) and figuring out the the best way to ingrate code into there
without making things messy.
http://elektronkind.org/outbox/afs/osi_net.c
So, this thread (osi_NetIFPoller) is kicked off by a taskq (see taskq
(9f)) in afs_call.c:afs_InitSetup.
It sets itself up and enters a while(1) loop, gathers the interface
info that other parts of AFS desires, aquires a RW lock on the global
array, updates the global array, releases the RW lock, and sleeps.
The thread sleeps for a tunable number of seconds, the default being
30. This can be adjusted via putting "set afs:afs_if_poll_rate =
<seconds>" in /etc/system.
The code currently doesn't care about pruning interfaces from the
global array that were up, but then down'd. It also assumes that the
max interfaces it'll ever see is 1024. The maximum interfaces that
solaris supports is, in the words of one Solaris network driver guy,
"arbitrarily huge"... so in the interest of keeping memory
requirements low and the code time, it's capped at 1024.
So, as I move on to finish this and hopefully provide a patch, could
someone please take a look at my logic/code and nit pick or suggest
improvements? One on my mind is whether afs_InitSetup is the right
place to spawn this thread.
Mucho thanks.
/dale
--
Dale Ghent
Specialist, Storage and UNIX Systems
UMBC - Office of Information Technology
ECS 201 - x51705
_______________________________________________
OpenAFS-devel mailing list
[email protected]
https://lists.openafs.org/mailman/listinfo/openafs-devel
