Add try_alloc and try_alloc_pinned to kernel module.

[wedsonaf]

This allows us to fail gracefully when out of memory.

Signed-off-by: Wedson Almeida Filho wedsonaf@google.com

[alex]

Should we put these in alloc.sr

[alex]

Based on how Box works, I think we want something a little smarter. Looking at:

• https://github.com/rust-lang/rust/blob/master/library/alloc/src/boxed.rs#L262
• https://github.com/rust-lang/rust/blob/master/library/alloc/src/boxed.rs#L292

We want something like:

        let layout = Layout::new::<mem::MaybeUninit<T>>();
        let ptr = alloc::alloc::alloc(layout);
        // Handle nul ptr here
        let mut b = unsafe { Box::from_raw(ptr.as_ptr()) };
        b.as_mut_ptr().write(value);
        b.assume_init()

[wedsonaf]

For my own benefit: do you see any UB in the code above?

I've updated the code as you suggested. In process of adding SAFETY comments, I noticed that we cannot call alloc with a zero-sized layout, so I added some code to take care of that.

I also don't call assume_init() because it requires some nightly feature. I just do the into_raw/from_raw dance.

PTAL.

[wedsonaf]

Should we put these in alloc.sr

Do you mean allocator.rs or to create its own file? If the former, I felt that they didn't belong there because these new functions are not tied to this particular allocator, nor any other; they call whatever the global allocator happens to be.

WDYT?

[alex]

Yes, I meant allocator.rs. I don't feel strongly though, lib.rs is also ok.

…

On Mon, Dec 7, 2020 at 5:12 PM Wedson Almeida Filho < ***@***.***> wrote: Should we put these in alloc.sr Do you mean allocator.rs or to create its own file? If the former, I felt that they didn't belong there because these new functions are *not* tied to this particular allocator, nor any other; they call whatever the global allocator happens to be. WDYT? — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#45 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAAAGBHUQO4IRPRJLHSAGXDSTVHMLANCNFSM4URCCBYQ> .

-- All that is necessary for evil to succeed is for good people to do nothing.

[alex]

Thanks for working through this with me! Looks great!