We are looking at a 32-bit architecture implementation were we can have
distinct address spaces for user and kernel, thus allowing 4G's for
each. In doing this we have come across the use of TASK_SIZE to
determine if an address is user vs kernel (example mm/memory.c). I'm
wondering is it just sufficient to set TASK_SIZE to 0xffffffff? This
feels wrong to me, since it would imply that all the places that are
testing will never need access to the kernel memory space.
thanks
- kumar