FW: RPC restructuring finished (mostly).

[Luke Kenneth Casson Leighton]

jeremy,

remember that this work was performed over two years ago, and it wasn't
you that carried out the split of interface and implementation parts.
i'll do a follow-up, hope you don't mind.

jf was doing spoolss, so i was unable to get the spoolss up to the same
consistent level, ready for doing replacement with IDL.

btw IDL is going to be a bitch for spoolss.  it's a two-level affair, with
the data "wrapped" and typecast to a void*, length, info_level.  the
top-level doesn't even know what the void* _is_.

which is a stupid, horrible and wasteful way to do dce/rpc programming.

then, at the second-level, they use internal marshalling/unmarshalling,
just like is currently done with security descriptors and SERVICE_INFO.

this explains why these structures, on spoolss, sec_desc and SERVICE_INFO
are all self-relative.

so, the code that does the self-relative modes (hard-coded) structures i
advise you to keep, for now, and deal with as a separate sub-project.

btw it is also worthwhile investigating whether receiving data from
non-intel-byte-order systems affects these second-level-marshalled
structures or not.

i doubt it.

which means that sun cascade will fail to work against samba, because
samba will incorrectly unmarshall the sec-desc (etc) - due to
receiver-makes-right rules.

hope this helps.


also, the reason that we didn't use talloc is because there are possible
circumstances in which the destruction of memory at the end of the rpc
marshalling will cause memory corruption.

remember, dce/rpc can pass memory structures over-the-wire that can be
used beyond the scope of one single dce/rpc function call, at both client
end and server end.

so, fragile or not, memory leaks are better than freeing memory and still
using it!

luke

> -----Original Message-----
> From: samba-technical-admin at us5.samba.org
> [mailto:samba-technical-admin at us5.samba.org]On Behalf Of Jeremy Allison
> Sent: 27 February 2001 01:48
> To: samba-technical at samba.org; jeremy at valinux.com; samba-team at samba.org
> Subject: RPC restructuring finished (mostly).
> 
> 
> Hi all,
> 
> 	I've been very busy lately splitting the RPC pipe
> serving code in smbd into interface and implementation parts. The
> ultimate goal (for Samba 3.0) is that the interface parts
> will be auto-generated from IDL files, and only the implementation
> backend part will need to be written by hand.
> 
> If you take a look at the new code, you'll find that the
> files in rpc_server/ have been split in two. A srv_XXX.c
> file, which is the interface part, and a srv_XXX_nt.c
> file, which is the implementation part.
> 
> The srv_XXX.c files are now very simple (other than srv_spoolss.c
> which still needs a little work), and have a completely
> regular format.
> 
> For example, if the RPC call is sent a SERVER_Q_XXX structure
> (Q is for query) and returns a SERVER_R_XXX structure (R is
> for reply) then the code to set up the interface and call the
> implementation looks like (where the "server" prefix is replaced
> by the RPC pipe name in the real code) :
> 
> static BOOL api_server_XXX(pipes_struct *p)
> {
>     SERVER_Q_XXX q_u;
>     SERVER_R_XXX r_u;
>     prs_struct *data = &p->in_data.data;
>     prs_struct *rdata = &p->out_data.rdata;
> 
>     ZERO_STRUCT(q_u);
>     ZERO_STRUCT(r_u);
> 
>     if(!server_io_q_xxx("", &q_u, data, 0)) {
>         DEBUG(0,("api_server_XXX: unable to unmarshall SERVER_Q_XXX.\n"));
>         return False;
>     }
> 
>     r_u.status = _server_XXX(p, &q_u, &r_u);
> 
>     /* store the response in the SMB stream */
>     if(!server_io_r_XXX("", &r_u, rdata, 0)) {
>         DEBUG(0,("api_server_XXX: unable to marshall SERVER_R_XXX.\n"));
>         return False;
>     }
> 
>     return True;
> } 
> 
> Given the RPC name server_XXX then everything in this
> function can be auto-generated from IDL files. In addition,
> the implementation interface is standardised to be (as mentioned,
> the spoolss interface is currently the only exception to this) :
> 
> uint32 _server_XXX(pipes_struct *p, SERVER_Q_XXX *q_u, SERVER_R_XXX *r_u)
> 
> All the current srv_XXX_nt.c files (except srv_spoolss_nt.c)
> follow this format, and so it should be much easier to add
> more RPC functions as they are decoded. To add a new RPC call
> you write the structure definition in the include/rpc_XXX.h
> file, add the interface function to the rpc_server/srv_XXX.c
> file in the above format and then do the real work of fleshing
> out the implementation in the rpc_server/srv_XXX_nt.c file.
> 
> In addition, the memory allocation has been greatly improved.
> A talloc context is created when an RPC call is begun, and is
> guarenteed to stay alive until the last packet in the reply
> RPC has been read by the client. This means that all malloc()/free()
> code that has to back out mallocs on error can be replaced with
> either prs_alloc_mem() or talloc() calls - they both allocate
> from the same RPC talloc context. I've used this to throw out a
> lot of fragile error processing code, and make the flow easier
> to understand.
> 
> I know this is a lot of changes, but it should give us a 
> very robust framework for adding more services/rpc's and
> make Andrew's job of moving us to IDL generated RPC code
> (which will be more correct and make it easier to add NT
> RPCs) much more simple.
> 
> HEAD is currently lagging behind slightly in this work in
> that the SAMR pipe has not been converted in HEAD, due to
> the passdb database interface changes there. I intend to
> work on that as soon as 2.2 has stabilized.
> 
> Just letting everyone know what the flurry of CVS commits
> on 2.2 have been about - let me know any comments/questions.
> 
> Cheers,
> 
> 	Jeremy Allison,
> 	Samba Team.
> 
> -- 
> --------------------------------------------------------
> Buying an operating system without source is like buying
> a self-assembly Space Shuttle with no instructions.
> --------------------------------------------------------
> 
> 
> 

 ----- Luke Kenneth Casson Leighton <lkcl at samba-tng.org> -----

"i want a world of dreams, run by near-sighted visionaries"
"good.  that's them sorted out.  now, on _this_ world..."