Created attachment 17989 [details]
WIP patches for master

I like the fact that you check for 
NT_STATUS_EQUAL(status, NT_STATUS_RPC_BAD_STUB_DATA) 
and set the result to NT_STATUS_RPC_PROCNUM_OUT_OF_RANGE (is that the same as INVALID?)

but as you noted later, the sequence will be out of sync since a patched server will not call dcesrv_netr_creds_server_step_check whereas a not patched will (but is unable to return return_authenticator).

Don't know if there is a way to make this compatible with both patched/windows and not patched samba..

(In reply to Michael Saxl from comment #3)

It is compatible with patched/unpatched Windows/Samba.

notice the state->tmp_creds = *state->creds and *state->creds = state->tmp_creds, which allows us rollback when step function
was not called on the server.

(In reply to Michael Saxl from comment #3)

NT_STATUS_RPC_PROCNUM_OUT_OF_RANGE is mapped from DCERPC_NCA_S_FAULT_INVALID_TAG

(In reply to Stefan Metzmacher from comment #5)

Oh, NT_STATUS_RPC_PROCNUM_OUT_OF_RANGE is wrong, it
should be NT_STATUS_RPC_ENUM_VALUE_OUT_OF_RANGE

Comment on attachment 17989 [details]
WIP patches for master

See https://gitlab.com/samba-team/samba/-/merge_requests/3189 for the latest version of patches

(In reply to Stefan Metzmacher from comment #4)

but how do you decide if you rollback?
The intended way is to rollback since a patched samba/windows will not step, but samba unpatched will.

I'm quite dumb at the specifics but as far as I read you make the step if all succeeds with
*state->creds = state->tmp_creds;

else a rollback is done.

What happens if the server side does the step (as far as I understood the unpatched server will)?

For reference: https://learn.microsoft.com/en-us/openspecs/windows_protocols/ms-winerrata/69ffd0ac-a0dd-49f2-96ad-6720441b0a93