Talk:Evolution of Operating Systems Designs/State of current research

DRM
For those that don't know:

What DRM is trying to do is not at all impossible or futile from a design point of view. Sure, there is an analog hole, but that degrades quality. High-quality digital copies can be very well controlled by DRM.

Consider the next-generation AMD processor in a motherboard with a TPM chip. Here is how it works.


 * 1) A regular OS load happens. This may be compromised of course, but that gets detected later.
 * 2) The OS sets aside 64 kB of physically linear RAM, then loads some code into it.
 * 3) The OS executes a special new CPU instruction (opcode) to enter trusted (evil) mode.
 * 4) The CPU performs a special bus cycle to inform other CPUs and restrict DMA access to the 64 kB chunk.
 * 5) A well-protected chip on the motherboard is granted access to this memory.
 * 6) The well-protected chip verifies a digital signature, then grants access to itself.
 * 7) The CPU begins executing the trusted and protected 64 kB.
 * 8) The CPU is placed into a new hypervisor-like mode, with the 64 kB of code in control. Unlike normal CPUs, these CPUs have 8 protection rings (not just 4) and an enhanced ability to virtualize themselves. The 64 kB of code can do something a bit like VMWare, but with less performance loss.
 * 9) The 64 kB of code checks a digital signature on the regular OS.
 * 10) The regular OS, now somewhat trusted because of the digital signature, can continue execution. The 64 kB of special code remains protected by hardware though, and retains the ability to monitor execution of the regular OS. The regular OS can ask the 64 kB of special code to perform various services in cooperation with the well-protected chip.

Now remember that encryption is happening on the monitor cable, the FireWire cable, etc.

Oh, you want to stop this with a custom driver? Sorry. The obvious problem is that an unsigned driver could be refused, but remember that the 64 kB of special trusted code is running the rest of the OS in a sort of virtual machine. In particular, memory writes can be intercepted via the permission bits in the page tables. There are two sets of page tables. Stuff first gets remapped by the ones the regular OS is handling, then gets remapped by the page tables that the 64 kB of trusted code is handling.

By this mechanism and by control of DMA, the 64 kB can be safely expanded. Whole videos could be decrypted into protected memory, though this isn't really required since the regular OS can be verified and protected.

AlbertCahalan 05:22, 12 November 2005 (UTC)

Yeah, and exactly who the hell is researching DRM hardware? Excepting Intel maybe, nobody. The fact is, you're blatantly misusing language. Everyone OTHER than you understands DRM research to mean DRM software *development*. You mean it in a very narrow non-standard sense, and you never bother to explain what sense it is you do mean it. 24.200.176.92 02:35, 13 November 2005 (UTC)