Kernel-mode IRQs and Kernel heap demand paging can't be enabled simultaneously #16
Description
In __kcore/__classes/heap.cpp, if we allow fakemapping/demand paging the heap, then:
inside of malloc(), while the heap's WaitLock is still held, we'll attempt to initialize a header at the top of new allocations. When initializing this header, we'll encounter a #PF because the new memory will be fakemapped -- and this will trigger the #PF.
If CONFIG_RT_KERNEL_IRQS is enabled, then the kernel will enable local #INTs during exceptions and SWIs. If the kernel enables local #INTs during exceptions, this will include #PFs, of course. Our #PF which was triggered by the fakemapped pages in the heap, will also of course, enable local #INTs.
If an IRQ occurs during this #PF, and that IRQ allocates an object on the heap, then we will get a deadlock because the #PF occured while we still held the heap's WaitLock.
We should probably just disable kernel heap demand paging -- or perhaps all demand paging in the kernel altogether. It was really just a vanity feature and the fact that #PFs can occur while we hold a lock is extremely concerning.