On Tue, 28 Oct 2008, Ralf Baechle wrote:
> I see your point. If we dynamically allocate memory for exception handlers
> at run-time and point ebase to it multi-kernel systems should till work
> unless maybe the firmware gets disturbed by such a change of ebase.
A run-time sanity check would do I suppose. I.e. that ebase points to
somewhere within installed RAM and does not clash with the kernel in some
way. We can mark that page as reserved in the memory map then; something
we ought to regardless and which I gather we do not at the moment.
Maciej
|