Stefan (metze) Metzmacher metze at samba.org
Mon Oct 17 02:11:20 MDT 2011

Hi Andrew,

>>>>> For now, the module does not use any event context, so I've made no
>>>>> change here yet.  
>>>> If only the gensec_update() routines use any event context
>>>> stuff at all, wouldn't it be better to pass the event
>>>> context explicitly there instead of putting an event context
>>>> into the gensec_security struct? This way the risk of using
>>>> gensec wrongly leading to nested event loops is greatly
>>>> reduced.
>>>> If we absolutely have to accept the whole gensec thing into
>>>> the Samba3 code (something which needs much broader
>>>> discussion I think), then we should only do it if we can
>>>> agree on handling the event stuff. Looking at 'struct
>>>> gensec_security' right now I don't think we are there yet.
>>>> The risk of accidentially getting nested event loops into
>>>> the main Samba3 code, leading to very hard to debug
>>>> situations is much too high for my taste with the tevent
>>>> context being part of a central structure.
>>> I've updated my branch at and addressed the suggestions that you and
>>> metze made, using the approach indicated in the commits you referenced
>>> earlier.  gensec_update() now takes a tevent context, and is the only
>>> gensec function that needs one. 
>>> http://git.samba.org/?p=abartlet/samba.git/.git;a=shortlog;h=refs/heads/s3-auth-gensec-module-2
>> Thanks! I'll have a closer look later today.
>> But
>> http://git.samba.org/?p=abartlet/samba.git/.git;a=commitdiff;h=ffc9033d0a7ed07
>> looks good, except for py_gensec_update() which seems to lack a
>> TALLOC_FREE(ev); after gensec_update().
> The mem_ctx here is only valid for the life of the call, so there is no
> need. 

Ah, ok.

>> In the long run we should change the callers and backends to
>> gensec_update_send/recv(),
>> but for now having the event context just on gensec_update() is fine for me.
> Great, thanks!

I've fixed some formatting and fixed the autoconf build
(auth/ntlmssp/gensec_ntlmssp.c was missing)

Can you take a look at the TODO's?
Why do we downgrade to the next spnego mech on LOGON_FAILURE?

I think the backend has to decide that and convert LOGON_FAILURE into
NT_STATUS_INVALID_PARAMETER if it's started as a subcontext.

Not related to your code, but can you have a look at
I think we should add the unix id of the user to the group array if
it's mapped with WBC_ID_TYPE_BOTH.

> I'll move on to the client code shortly.



-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 262 bytes
Desc: OpenPGP digital signature
URL: <http://lists.samba.org/pipermail/samba-technical/attachments/20111017/e93fb182/attachment.pgp>

More information about the samba-technical mailing list