[PATCH] RDP compositor take 5

[Hardening]

On 30/03/2013 21:34, Jason Ekstrand wrote:
> Hardening,
> This is looking really good.  I only have a couple comments on the
> code below that should take all of 5 minutes to address.
> 
> That said, It's segfaulting on me a lot.  Most of them seem to have to
> do with wl_seat.  I really don't know why.  Also, it works with
The same here, i think seat releasing does not release every associated
resources (BTW this is mark as a TODO in the code). The RDP backend
triggers the bug easily as it creates one seat per incoming RDP peer. I
guess the seat hotplugging is hard to test on other backends. In
valgrind backtraces it looks like it is related to the pointer and/or
the keyboard.

> freerdp, but doesn't work with rdesktop; why is this?
The short response would be: there's bugs in both products, so they
don't mix well.

Full explanation: in RDP there's 3 kind of security: RDP security which
come with RDP4 and has many security flaws, TLS which uses a certificate
and key (RDP packets over a TLS transport), and finally NLA which is a
windows thing (and which is desactivated in the weston compositor).
rdesktop knows only RDP4 security. When the remote peer is 127.0.0.1
FreeRDP will announce a none encyption level and method. According to
the spec, rdesktop should not start a secret exchange, but it does. And
on the other hand FreeRDP is too strict at this step of the negociation.
I didn't inquiry more but it's also possible that rdesktop don't support
surface commands used to send images.

Hardening.