Solaris 7

[assar@stacken.kth.se]

Anthony.Iano-Fletcher at nih.gov writes:
> 		/etc/name_to_major:	xfs 140
> 		/etc/name_to_sysnum:	xfs   40
> 
> 	'arlad -t' works nicely and I got the kernel module xfs loaded
> just fine and I did the drvconfig and the the mount_xfs, followed by
> starting the arlad daemon.
> 
> When I do 'ls /afs' then I start getting these messages on the
> console...
> 
> 	xfs_message_receive opcode = 1
> 	xfs_message_wakeup
> 	xfs_devwrite error = 0
> 	xfs_message_rpc after sleep
> 	xfs_message_rpc opcode = 12
> 	messageq = 7004ea84, next = 7004ea88first: 12:40
> 	xfs_message_rpc before sleep
> 	xfs_devselect dev = 36700160, events = 64, anyyet = 0
> 	xfs_devread dev = 36700160
> 	xfs_devwrite dev = 36700160
> 	xfs_message_receive opcode = 9
> 	xfs_message_installdata
> 	xfs_node_find
> 	cache_name = '@001'
> 	XFS PANIC Warning: xfs_message_installdata failed to lookup cache file = @001, error = 6
> 
> over and over again..... and nothing ever works. The file
> /usr/arla/cache/@001 does exist and does contain a list of cells just as
> one would expect under /afs...... but I never get a response from 'ls'.
> 
> When I kill arlad the system stops giving errors..... Ive tried running
> arlad with debugging but this give me nothing.
> 
> What have I missed? how do I find where the xfs module is looking for
> @001?

It sounds like arlad is unable to find and use the correct system call
(40 in this case).  You can probably see if that's what's happening by
running arlad under truss and see what system call it uses.

You need a libkafs from a krb4 of 19980701 or more recent to be able
to specify the system call at run-time and then set the environment
variable AFS_SYSCALL=xfs before starting arlad.

/assar