[Open-graphics] DRM on Linux, BSD, OpenSolaris, etc.

[James Richard Tyrer]

Is it a reasonable idea to try to figure out a way to play HD-DVDs on an 
Open Source Software OS?  I think that this is possible and it would be 
much more secure than M$ Vista.

Further research indicates that the license for AACS requires that the 
decompressed signal sent over the connecting bus (AGP, PCI, PCIe, HTX, 
etc.) be encrypted.  These people are seriously paranoid -- it doesn't 
do any good to be paranoid because hackers have already found out how to 
get the disk key for HD-DVDs.

IAC, this actually does help solve the problem.

So, my design so far for the graphics board:

We have two PCI addresses (secondary addresses) one to communicate with 
the system and one for input of streaming video.

We have an AES decoder for the streaming video input that can be turned 
on and off.

We have a MCU on the board (soldered to the board) with built in ROM and 
I2C to communicate with the HDCP enabled HDMI transmitter and an 
external connection.

There are two things that the OS driver can do:

1.	Turn HDCP on and off BUT not override the MCU turning it on.

2.	Turn the AES decoder on and off.

The MCU will know if the AES decoder is on and the video resolution 
being used, not by being told by the OS but by reading the information 
from the registers in the video controller and or being told what to do 
over the I2C bus.  If the decoder is on, it is presumed that video is 
being received -- the OS needs to turn it off if no video is being 
received.  When the decoder is on, the MCU will turn HDCP on or off and 
other outputs on or off based on the requirements of the video disk's DRM.

Now, we can't stop somebody from making a software decoder for AACS that 
produces 'clear' (unencrypted) video, but you can't stop hackers from 
making one for Windows (which has already happened).

Now we also make a video decoder board that has an AACS decoder for 
input and a AES encoder for output.  If we use something like the NXP 
chip the board would have firmware on board.  If this needs to be 
loadable or flashable, then you would also need an MCU with ROM 
(soldered to the board) to handle the issues of encryption, DRM and 
communicate with the graphics board.

So, this video decoder board accepts the unchanged data from the HD-DVD 
or BR decodes the AACS or CSS (if needed), decompresses the video, and 
encrypts the output with AES (if needed).

The board can communicate with the graphics board over the I2C bus 
regarding resolution, HD rules, and encryption.

The I2C bus is not considered a consumer accessible bus so this is OK.

This is totally secure to the extent that appears to be required by the 
DHCP & AACS licenses -- you can't get unencrypted HD out of the HDMI 
port, or the computer's internal bus, without serious hacking of the 
hardware.  We can even publish the source code for the MCU firmware on 
the boards since it is in a ROM soldered to the boards.

Does anyone see something that I missed?  Can this be broken without 
using a soldering iron and a ROM programmer, or tapping into the I2C bus?

I have some questions about this.

AES requires a common key -- the encoder and the decoder need the same 
key.  Do we give all of our boards the same key, or would each graphics 
board get a random key which the video decoder board would need to 
obtain over the I2C bus?

This does not address the issue of copying.  Obviously, allowing no 
copying is simple.  If the disk says no copies, then the video decoder 
board would refuse to provide the needed output.  Remember, it won't 
work to try to record the AES encrypted video data.

I still have not figured out how making a copy is prevented.  I would 
think that you could just directly copy the contents of the disk.  I 
presume that there is something in the drive that prevents it. 
Otherwise, you could just install 2 drives in a PC an make copies  IAC, 
it isn't our problem.

What I don't know about is what to do when a disk allows copies.  How 
can you determine if a copy has been made since you can't burn a pressed 
disk.  Perhaps 1 copy means one level of copying which can be 
implemented in the video decoder board if it also has an AACS encoder. 
You just output a header with 1 less copy allowed.

-- 
JRT