Is something like `realloc_zeroed` and `grow_in_place_zeroed` useful?

[TimDiekmann]

When growing an allocation, is there a reason why there is no method that guarantees that the new memory is zeroed?

[petertodd]

+1

Implementation-wise for large reallocations there's definitely cases where allocators can efficiently do things like mmap additional (zeroed) pages to the end of an allocation. Also can have a default implementation that just forwards to realloc() so impls don't need to worry about them if they don't want too.

[gnzlbg]

I think these two can be useful. Whether these are worth having in an MVP, I don't know, we can always add them later.

[TimDiekmann]

AllocRef has alloc_zeroed, but no realloc_zeroed and grow_in_place_zeroed. Do we want to add those?

Shall we introduce a NonZeroLayout and use this in the alloc trait(s)?

Currently, zeroed allocation is pretty error prone and even experienced users may miss this: https://github.com/rust-lang/rust/issues/63291#issuecomment-538692745. On the other hand, we would restrict implementations which are fine to get zeroed layout like a bump allocator.

	Add them now	Not for an MVP	No need
@Amanieu	:+1:
@Ericson2314
@glandium
@gnzlbg
@Lokathor		:confused:
@scottjmaddox	:+1:
@TimDiekmann	:+1:
@Wodann	:+1:

Please vote with

• :+1: merge: Add them now
• :confused: postpone: I want to have them, but not for an MVP
• :-1: close: We don't need those (why?)

[Lokathor]

This seems like maybe it's about two things.

Is this about having newly allocated memory blocks already be zeroed? Or is this about being able to pass a zero sized allocation request into the allocator?

[TimDiekmann]

It's just like alloc_zeroed: The newly allocated memory is guaranteed to be zeroed. A default implementation would look like alloc_zeroed.

[Lokathor]

Ah, that's very good, but i agree that it's not really necessary for the minimum "get something working" stage. Voted.

[scottjmaddox]

It would be better to have these before stabilizing AllocRef.

[TimDiekmann]

I think stabilizing the whole thing will take at least one year

[Lokathor]

I also wouldn't actually stabilize AllocRef without this

But it's low priority to just get people using AllocRef and get people hammering on usage problems in Nightly and so on.

[TimDiekmann]

I like to mention though, that implementing those function is trivial and straight forward. But we shouldn't clutter the trait unless it's more or less fixed.

[Wodann]

I find it hard to make a call on this one for the following reasons:

• What is the design goal of the std's AllocRef API? Should it be a minimal API that can be used to build others APIs in external crates? If so, alloc_zeroed, realloc_zeroed, etc. could be in a non-std crate.
• Should the API focus on being general-purpose? If so this might be a better design:

pub fn alloc_with<F>(self, layout: Layout, f: F) -> Result<NonNull<u8>, AllocErr>
    where
        F: FnOnce(ptr: NonNull<u8>, size: usize) -> NonNull<u8>;

[Amanieu]

I think these should be added since they can be efficiently implemented with the mremap system call on Linux. mremap allows you to move/grow/shrink a memory mapping, and any new pages added for growth are guaranteed to be zeroed.

If AllocRef does not have these methods then the user will have to manually write zeroes to the added memory since the API makes no guarantees on their contents.