oplock and SCO 5.0.4

Alastair Broom abroom at valleyt.co.uk
Fri Sep 17 18:48:51 GMT 1999

Just thought I'd "help" everyone by stoking the argument about oplocks
having a timing problem with a vague set of observations  :-)  
I have a strange problem using samba 2.0.5a on SCO 5.0.4:
With the binary compiled versions kindly provided by Stephen Davies on
ftp.sdc.com.au I have no problems reading or writing files to the SCO
server at all.

However, when I compile my own binaries with either gcc or SCO's cc, I end
up with smbd's exiting for no apparent reason (and not producing a core)
when copying a directory of files from the SCO samba box to a W98 client.
The directory contains 30ish files totalling 6M.  The W98 client happily
shows the progress window as it copies the files, until it gives up on the
9th, 13th or 20somethingth file with the "Error Copying File" window which
says "Cannot copy <filename>: The specified network resource or device is
no longer available"
Sure enough, the smbd process which *was* handling the request
(according to log.smb) has exited without any sort of error or core dump.

The smb.conf contains  oplocks = True 
This error happens with debugging up to 4.
When you turn debugging up to 5, there are *no* problems with the copy and
everything runs perfectly, thus suggesting a timing problem that is
averted by all the extra logging going on.

Alternatively, setting oplocks = False means everything runs fine.

I have copies of log files, ps listings, and much much more testing 
information if anyone is interested.  I'm posting this in case it relates
to the other problems people have been seeing before they disabled oplocks.
Alastair Broom               Valley Technology, Edinburgh, Bonnie Scotland
abroom at valleyt.co.uk

More information about the samba mailing list