Recent Samba (4.18.0+) winbind configuration ?

[Rowland Penny]

On 13/04/2023 08:08, Andreas Schneider wrote:
> On Tuesday, 11 April 2023 15:49:15 CEST Rowland Penny via samba-technical
> wrote:
>> On 11/04/2023 13:32, Jiří Šašek - Solaris Prague via samba-technical wrote:
>>> Story begins when Samba 4.14.0 changed its VFS so we decided to skip
>>> 4.14.x and also 4.15.x releases to bundle it on Solaris but various
>>> problems were also with later releases and we were only fixing issues in
>>> 4.13.0 . Currently (4.18.0) the situation can called be critical so I
>>> focused on recent release.
>>>
>>> winbindd is giving up and I think the key could be in log message:
>>>
>>> [2023/04/11 11:28:53.090661,  5, pid=1525, effective(0, 0), real(0, 0),
>>> class=winbind] ../../source3/winbindd/wb_lookupname.c:
>>> 52(wb_lookupname_send)
>>>
>>>     WB command lookupname start.
>>>     Search namespace 'smbsetup' and domain 'smbsetup' for name ''.
>>>
>>> ...where "smbsetup" is my domain  ...and I have added core-dump here to
>>> freeze the stack:
>>>
>>> [2023/04/11 11:28:53.090780,  0, pid=1525, effective(0, 0), real(0, 0)]
>>> ../../source3/lib/dumpcore.c:315(dump_core)
>>>
>>>     dumping core in /var/samba/log/cores/winbindd
>>>
>>> I went over such many frames of "tevent's" but I do not know where to
>>> start evaluation. I only catch the (PIDL) generated code was
>>> optimized-out by gcc (11.3.0) so I built the code with -O0 and now the
>>> code is available.
>>>
>>> In the mood of total madness I also tried ChatGPT which suspects my
>>> smb.conf but answer on my question:
>>>
>>>
>>> Q:
>>> what was changed in idmap configuration in smb.conf between 4.13 and 4.18
>>> ?
>>>
>>> A:
>>> There have been some changes to the idmap configuration in Samba between
>>> version 4.13 and 4.18, although they may not necessarily be related to
>>> the issue you're experiencing.
>>>
>>> One notable change is that in Samba 4.18, the idmap config parameter has
>>> been deprecated in favor of the idmap_ldb parameter. This new parameter
>>> uses an LDB backend for id mapping, which allows for better performance
>>> and scalability than the previous tdb backend.
>>>
>>> Another change is that the idmap config parameter in Samba 4.18 now
>>> accepts a backend option, which allows you to specify the backend to use
>>> for id mapping. The options are rid, ad, autorid, and autorid_compat.
>>>
>>> The rid backend continues to be the default, which means that if you
>>> don't specify a backend explicitly, Samba will use the RID algorithm to
>>> generate UIDs and GIDs.
>>>
>>> In summary, if you're upgrading from Samba 4.13 to 4.18, you may need to
>>> update your smb.conf to use the new idmap_ldb parameter and specify a
>>> backend if you're not using the default rid backend.
>>>
>>> ....looks suspectably because I can not find "idmap_ldb" in man smb.conf
>>> so I asked again:
>>>
>>> Q:
>>> backend option of idmap config was available even in samba 4.0.0 smb.con
>>> so what exactly has been changed in 4.18.0 ?
>>> A:
>>> Something went wrong. If this issue persists please contact us through
>>> our help center at help.openai.com.
>>>
>>> ...I do not want to push here the way how to shot-down the f'g ChatGPT
>>> but I would like to know if something was really changed. I went through
>>> the RELEASE NOTES and there is nothing directly concerns to idmap config
>>> AFAIK
>>>
>>> constraints:
>>>    - net ads join works as usual include adding of "A" record to DNS
>>>
>>>    - despite of successful join it looks DC can not be located:
>>> [2023/04/11 11:28:53.083220,  1, pid=1525, effective(0, 0), real(0, 0),
>>> class=rpc_parse] ../../librpc/ndr/ndr.c:490(ndr_print_
>>> function_debug)
>>>
>>>          wbint_PingDc: struct wbint_PingDc
>>>             in: struct wbint_PingDc
>>>
>>> [2023/04/11 11:28:53.083291,  1, pid=1525, effective(0, 0), real(0, 0),
>>> class=rpc_parse] ../../librpc/ndr/ndr.c:490(ndr_print_
>>> function_debug)
>>>
>>>          wbint_InitConnection: struct wbint_InitConnection
>>>             in: struct wbint_InitConnection
>>>                 dcname                   : *
>>>                     dcname                   : ''
>>>
>>> ...so it really looks like a stupid mis-configuration issue of mine.
>>>
>>> Google is silent on all my attempts to ask "him".
>>>
>>> If someone have some good idea please share it with me.
>>>
>>> Many thanks,
>>> Yura
>>
>> Yes, stop asking chatGTP, you have made my day by supplying me with
>> proof that chatGPT is STUPID!!!
> 
> It isn't really stupid, it just learns patterns it finds in a data set and
> then give you the results what it learned. This means that the data set is at
> fault here and the data set is "the Internet" :-)
>   
> 
> 

That doesn't make it stupid ?????

The internet is, in many ways, stupid

Rowland