[PATCH 2/2] s4:dsdb/rootdse: Support netlogon request

[Benjamin Franzke]

Hi Andrew,

I've read a bit of the nbt code, to see whether I did some wrong guesses or
so,
but then i came to the conclusion maybe the data is invalidated (e.g.
free'd) too early,
before it is actually send out to the client.
That would explain this occasional, hard-to-reproduce error.

Then I found something that *probably* caused that.
I used ldb_msg_add_value to add the netlogon blob to the message:
https://git.bnfr.net/samba/tree/source4/dsdb/samdb/ldb_modules/rootdse.c?h=netlogon-5#n810
That means it stays part of requests talloc tree, which may be freed before
sending the message(?).
ldb_msg_add_steal_value should be used instead.

I've pushed a patch to my netlogon-8 branch,
could you please run another autobuild to see whether that fixed it?

Thanks, Ben


2013/11/6 Andrew Bartlett <abartlet at samba.org>

> On Wed, 2013-11-06 at 10:26 +1300, Andrew Bartlett wrote:
> > On Tue, 2013-11-05 at 21:19 +0100, Benjamin Franzke wrote:
> > > Hi Andrew,
> > >
> > >
> > > I still could not find the cause of the test failure (Reminder: please
> > > send me the pcap file).
> >
> > Because it's for the whole run, it's 289MB even after XZ, and 1.7GB
> > uncompressed for loading into wireshark. Are you sure you still want
> > it?
> >
> > Andrew Bartlett
>
> As a clue, I think it's either related to an NDR parsing error:
>         case NDR_ERR_UNREAD_BYTES:
>                 return NT_STATUS_PORT_MESSAGE_TOO_LONG;
>
> or due to pushing the 'udp' message via socket_wrapper over our internal
> datagram sockets:
>
> {NT_STATUS_PORT_MESSAGE_TOO_LONG, EMSGSIZE},
>
> Andrew Bartlett
>
> --
> Andrew Bartlett
> http://samba.org/~abartlet/
> Authentication Developer, Samba Team           http://samba.org
> Samba Developer, Catalyst IT                   http://catalyst.net.nz
>
>
>