Align up the Hole initialization address

[tomaka]

ptr::write requires the address to be properly aligned. This PR makes sure that this is the case.

[phil-opp]

Thanks a lot for the pull request!

Just to ensure that I understand everything correctly: As far as I know, the HoleList::new function is only called once with the heap bounds when the allocator is initialized. So this issue should only occur when heap_bottom is not properly aligned. I just had a short look at your kernel to find out why heap_bottom could be misaligned in your case. I think the reason is that you're directly using the free address ranges reported by multiboot in combination with identity mapping. Is that right?

Assuming my understanding is correct, this seems like a use case where rounding up the address and leaving the first few bytes of the heap unused is indeed the best solution.

Could you add a short sentence about this to the documentation of HoleList::new? We should specify that the caller does not need to ensure any alignment of hole_addr since the function performs the required alignment itself. Otherwise the changes look good to me!

[tomaka]

I think the reason is that you're directly using the free address ranges reported by multiboot in combination with identity mapping. Is that right?

Indeed!

Could you add a short sentence about this to the documentation of HoleList::new? We should specify that the caller does not need to ensure any alignment of hole_addr since the function performs the required alignment itself. Otherwise the changes look good to me!

I added a small line. I must admit that I don't really know what to say in the comment.

[phil-opp]

Thanks! The comment is fine. To me, hole_addr sounds like something that needs to be capable of storing a Hole, including correct alignment. The function, however, now performs the alignment itself, so this isn't required anymore. I think it's worth documenting such facts to ensure that no confusion occurs.

[phil-opp]

Published as version 0.6.5