kernel oplock and broken pipe messages

[dss at cs.york.ac.uk]

Hi,

I have inherited responsibility for a group of Samba servers
and I am in the process of upgrading them all to run 2.0.6.

I have come across a couple of things I am not sure about and would
appreciate some advice.  I have checked in the FAQ lists and the mailing
list archives, but there seems to be no mention of these things in there.

First Question
==============

I have just upgraded a server which is running under IRIX 6.5.4,
fired up the smbd process and observed the message:

	check_kernel_oplocks: Kernel oplocks available and set to True.

in log.smb.  However, I am now noticing a lot of messages of the
following type appearing in the client log files:

	set_file_oplock: Unable to get kernel oplock on file {filename},
	dev = xxxxxx, inode = xxxxxx. Error was Invalid argument

Looking at the fcntl man page for IRIX it says that EINVAL is returned when
the file is in a non-XFS filesystem, and in some cases this is does appear
to be the problem, but in other cases it is not.

Has anyone else observed problems with this for files which _are_ on
an XFS filesystem?  Is it something to worry about?  Should I disable the
kernel oplocks feature?  Is it not a good idea to have it enabled on
a server which has some non-XFS filesystems?

Second question
===============

The clients now seem to be logging a lot of messages of the form:

	lib/util_sock.c:(537)
	write_socket_data: write failure. Error = Broken pipe

This also seems to be the case with 2.0.5a running under Solaris 2.5.
Again, is this anything to worry about, or is it just normal behaviour?

Thanks,

Dave