[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [MiNT] Security again



On Tue, Dec 07, 1999 at 12:27:59PM +0100, Konrad M. Kokoszkiewicz wrote:
> 
> Speaking of address space virtualization, what is the position of GEM AES
> in all that? As I see it - though I must say I don't have a clue about how
> specific things are solved in an AES - the concept of the AES' ability to
> read/write application's address space (with F_OS_SPECIAL privileges)
> relies on the fact that the memory is physically mapped. Or I am errant?

Since the AES does not do DMA or similar, virtual mapping should pose no
problems, as long as the AES and the applications use the same (common)
address space.

BTW: are there any places besides USERDEF-callbacks and the parameters
passed in the in/out arrays where the AES needs to access an application's
memory?

With appropriate support in AES and kernel, the AES might even perform such
callbacks when different virtual memory spaces are active, and the
parameters can easily be copied to/from the AES's memory space.

cu
Michael
-- 
Michael Schwingen, Ahornstrasse 36, 52074 Aachen