Skip to content

Commit

Permalink
avoid using aligned_alloc; posix_memalign is better-behaved
Browse files Browse the repository at this point in the history
  • Loading branch information
RalfJung committed May 14, 2024
1 parent ef15976 commit 5cc020d
Showing 1 changed file with 8 additions and 11 deletions.
19 changes: 8 additions & 11 deletions library/std/src/sys/pal/unix/alloc.rs
Original file line number Diff line number Diff line change
Expand Up @@ -89,21 +89,18 @@ cfg_if::cfg_if! {
// /memory/aligned_memory.cc
libc::memalign(layout.align(), layout.size()) as *mut u8
}
} else if #[cfg(target_os = "wasi")] {
#[inline]
unsafe fn aligned_malloc(layout: &Layout) -> *mut u8 {
// C11 aligned_alloc requires that the size be a multiple of the alignment.
// Layout already checks that the size rounded up doesn't overflow isize::MAX.
let align = layout.align();
let size = layout.size().next_multiple_of(align);
libc::aligned_alloc(align, size) as *mut u8
}
} else {
#[inline]
unsafe fn aligned_malloc(layout: &Layout) -> *mut u8 {
let mut out = ptr::null_mut();
// posix_memalign requires that the alignment be a multiple of `sizeof(void*)`.
// Since these are all powers of 2, we can just use max.
// We prefer posix_memalign over aligned_malloc since with aligned_malloc,
// implementations are making almost arbitrary choices for which alignments are
// "supported", making it hard to use. For instance, some implementations require the
// size to be a multiple of the alignment (wasi emmalloc), while others require the
// alignment to be at least the pointer size (Illumos, macOS) -- which may or may not be
// standards-compliant, but that does not help us.
// posix_memalign only has one, clear requirement: that the alignment be a multiple of
// `sizeof(void*)`. Since these are all powers of 2, we can just use max.
let align = layout.align().max(crate::mem::size_of::<usize>());
let ret = libc::posix_memalign(&mut out, align, layout.size());
if ret != 0 { ptr::null_mut() } else { out as *mut u8 }
Expand Down

0 comments on commit 5cc020d

Please sign in to comment.