[RFC] Advice on SMB client python bindings?

[Stefan Metzmacher]

Hi Tim,

> Currently certain samba-tool commands don't work if SMBv1 is disabled on
> the server [1]. I'm trying to fix this, and just wanted to check if
> anyone has any input before I go too far down the wrong path.
> 
> Basically the problem is the current SMB python bindings that we're
> using (pysmb.c) don't have proper SMBv2 support.
> 
> To start with, I looked at the changes Metze made to the source4 RPC
> code to support SMBv2 (i.e. commit 4422f7382aad309) and tried to use the
> same approach. I got a SMBv2 connection established, but quickly
> realized that although source4 libcli contains SMBv2 code, there's no
> common API to handle *either* a SMBv1 or SMBv2 connection. The SMBv1 and
> SMBv2 code use different data structures, and so an extra layer would be
> need to be added to the source4 libcli code to handle this.
> 
> It seems like the source3 libsmb code has already solved this problem
> (i.e. the smbXcli_conn_protocol(cli->conn) >= PROTOCOL_SMB2_02 checks).
> So now I'm looking at moving the python bindings into the source3 tree
> instead, so we can use the source3 SMB client library code underneath.
> Does this approach sound reasonable?
> 
> Finally, I noticed there's some existing source3 python bindings
> (pylibsmb.c, aka libsmb_samba_internal), however, they don't really
> appear to be used. Should I consolidate this code with what I'm
> changing, so we have a single set of SMB client python bindings? If so,
> does anything actually rely on libsmb_samba_internal? (Removing the
> .get_oplock_break API so we can use the synchronous SMB client APIs
> would make my life a lot easier).

I think it would be good to have just one set of python bindings.

I recently had the need for changes as well. See
https://git.samba.org/?p=metze/samba/wip.git;a=shortlog;h=refs/heads/master4-dcerpc-auth
https://git.samba.org/?p=metze/samba/wip.git;a=commitdiff;h=cb40716a117e512fa35897f8e19cf2e5029af4d5
https://git.samba.org/?p=metze/samba/wip.git;a=commitdiff;h=b1ebc6f68e067dba3288e000e723738bd1414a98
https://git.samba.org/?p=metze/samba/wip.git;a=commitdiff;h=0da84a6006f814c0df9760ea4aa4d44122d2a0c6

I'd also keep the async stuff, it doesn't look too hard at the python
layer.

I think we could disable the poll_mt stuff by default and let the caller
select between the two py_cli_state_setup_ev() implementations in
py_cli_state_init().

metze


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: OpenPGP digital signature
URL: <http://lists.samba.org/pipermail/samba-technical/attachments/20181126/cb5be676/signature.sig>