dce/rpc thing

[jra at dp.samba.org]

On Tue, Apr 15, 2003 at 06:37:54PM +0100, Ronan Waide wrote:
> Hi folks,
> 
> I finally got a successful run with a large RPC buffer, but before I
> offer any patches I have a question about some of the stuff I'm
> modifying. In cli_pipe.c, function rpc_api_pipe_req(), there is this:
> 
> 	/*
> 	 * calc how much actual data we can send in a PDU fragment
> 	 */
> 	max_data = cli->max_xmit_frag - RPC_HEADER_LEN - RPC_HDR_REQ_LEN -
>             (auth_verify ? RPC_HDR_AUTH_LEN : 0) - auth_len - 8;
> 
> This results ultimately in PDUs that are 8 bytes shorter than the
> max_xmit_frag, which apparently breaks something (having tweaked just
> about every other field going through, I'm pretty sure about this!) on
> the Windows end of things. So, em, what's the -8 for?
> 
> Secondly, it appears that large RPC stuff with NT4 requires
> PIPE_RAW_MODE to be implemented. I've done this, but I'm not sure how
> it affects the signing/sealing hoo-hah. I'm also aware that this may
> be a moot point since perhaps Samba is more focussed on Win2K than
> NT4, and Win2K certainly seems happy enough without the raw mode
> stuff. A possible solution is to only switch on the raw mode code if
> we're dealing with a WinNT box (or anything else that proves to need
> it), but I don't know how to do that.
> 
> If anyone can shed a little light on either of these, I'd appreciate
> it. I will be testing the code against a Win2K box to see if it
> behaves sanely, plus there are still a few glitches to hammer out, but
> hopefully I should have a patch to offer before the end of the week.

This code and the schannel stuff is actively being worked on at the
SambaXP hack-fest in Germany.... Please keep cvs updating, some
things will change.

Jeremy.