s4: fsmo role transfer

Anatoliy Atanasov anatoliy.atanasov at postpath.com
Thu Sep 9 04:13:15 MDT 2010

Hi Tridge,
> > Hi Nadya,
> Andrew and I looked over your FSMO role transfer patches today. Very
> nice work! You are producing some great code.
> Andrew spotted one thing, which is that the rootdse role transfers
> don't seem to be protected by any authentication or ACL. Maybe we
> could just have a simple security level check on these changes in the
> rootdse code?  Alternatively you could send the token in the irpc call
> to the drepl server and have it do authentication or ACL checks.
> The simplest is probably to say that rootdse modify is denied unless
> you are at least SECURITY_DOMAIN_CONTROLLER level.
> We also fixed up some build breakages caused by some recent changes in
> the drepl server, and the conversion of the IRPC code to use binding
> handles. We've put a fixed and rebased copy of your fsmo-fixed branch
> in the nadya-fsmo-fixed in my git tree.
Thanks for making this work i had troubles yesterday porting the IRPC stuff. I made a test for this implementation and i pushed it to http://git.samba.org/?p=anatoliy/anatoliy.git;a=shortlog;h=refs/heads/nadya-fsmo-fixed-test

There is the following problem with that test, it works with windows though:
1. schedule a partition pull with extended op
2. receive getncchanges with extended op, process it and return to sender
3. receive another getncchanges with extended op, process it and return to sender
4. callback is called for the 2nd but not for the first call
5. second call fails because there on step 3 we have already changed the role.

The transfer goes well but the second extended op is totally unexpected and shows error.
To reproduce this you can pull and run make test TESTS=fsmo


More information about the samba-technical mailing list