Dear all,

Our team is developing a hardware security module (HSM) running QNX 6.4 on a Freescale p1022 CPU with 512 MB of RAM. For
 security reasons, our HSM must guarantee that sensitive information is always contained within the first 64MB of the 
machine’s physical address space. We define sensitive information at the process level: all data (heap + stack 
segments) of certain processes are considered sensitive, all other processes are not sensitive. Therefore, we would like
 to partition our RAM into the following zones:

0-64 : “secure memory” -> system page + kernel memory + sensitive processes memory (heap + stack)
64-512 : “non-secure memory” -> non-sensitive process memory (heap + stack)

Ideally, the code segments of all processes should be loaded in non-secure memory, but this is optional for us.

The immediate naïve solution to this is to use the ‘ –r’ startup option to “remove” the excess memory and simply 
use 64MB for everything. Of course, we would like to find a better solution ;-)

We have looked at this problem for several weeks but have not since devised a satisfactory solution. 
We did look at the QNX feature called “Memory Partitioning” (see http://community.qnx.com/sf/wiki/do/viewPage/projects
.core_os/wiki/Memory_Partitioning_Technology) but it is apparently experimental and unfortunately unsupported by QNX.

Did any of you encounter a similar use case of QNX? Did you find a satisfactory solution ?

Any input is gladly welcome !

Best Regards,

Baudouin Nyssen.