illumos gate

From Matt: This appears to be an ndrgen bug. So far i've isolated it to the difference between Windows providing an all-0 LM challenge buffer, and (linux) mount.cifs providing a NULL LM challenge pointer. The former works, the later adds an extraneous WORD between the NT challenge and the Validation Level, which makes it look like we're setting Validation Level to 0. (the correct '3' WORD appears as a 2-byte 'long frame' in wireshark.)

Also from Matt:
Currently, when a client gives us a NULL LM challenge response pointer, when we pass on this information to the DC via Netr SamLogon, two extra bytes get added after the NT challenge response and before the validation level. This screws up the server's decoding of our packet, and it reads these two zero bytes as the validation level. These bytes get added by ndrgen's padding routine, which defaults to 4-byte alignment. The solution, then, is to use 2-byte alignment for this interface. While I can't find a reference to alignment in the spec(s), several other functions in NETR use 2-byte alignment, and that solution works here, so that's what I'm going with.