No subject

[malcolm melville +44 171 542 8472]

Luke

Not sure what you mean by this:

....
the problem is that the connection to each of the password servers takes
too long to time out: long enough for the client to think that the server
is dead.

what is needed in clientgen.c is some code that opens a socket on port 139
and then checks that the socket has been connected: at present it goes
ahead and attempts to send an smb session request without validating that
the socket is really connected.  this is probably confused by my
understanding of sockets, by the way. 
....

Did clientgen.c change between 1.9.17p2 and 1.9.18p1? I haven't had time to
check. It looks to me like in the past a password check failure caused the
checking of the next server and now no longer does. Timeouts do not seem to be
the issue. In fact in 1.9.17p2 it was pretty fast at scanning the list to find
a server that could validate the user/password combination. I will try to check
with snoop/tcpdump and see what packets go on the wire, and then look at the
source code module. 

I guess the big question is: which behaviour is expected and wanted. Was it a
fluke in 1.9.17p2 that it worked and 1.9.18p1 is behaving correctly or vice
versa. If someone can answer that question (the smb.conf man page is not
explicit) then maybe a new smb.conf parameter could be put in place to set
which ever behaviour is non-default. Or maybe its only me that wnats the old
behaviour and I will have my own patched up version!

regards
malcolm

------------------------------------------------------------------------
Any views expressed in this message are those of the individual  sender,
except  where  the  sender  specifically  states them to be the views of
Reuters Ltd.