Disagree - the author intended for the macro to be safe to use as they didn't expect new unsafe code to be insertable - they wouldn't have marked the macro as unsafe even if that was a supported feature, because this isn't their intent for it to be used that way. If the default of the language was that a new unsafe block was required here as I'm proposing, then the macro would have been safe:
alloc_pages!(unsafe { *(0x41414141 as *mut usize) } );
Agree with the general sentiment here. I've previously raised that using unsafe in a macro is difficult to reason about because of things like overloading basic traits like arithmetic and comparisons. Generics is an interesting point to consider too.