Re: [PATCH v3 2/2] rust: support align and NUMA id in allocations

[Uladzislau Rezki <urezki@gmail.com>]

On Wed, Jun 25, 2025 at 09:12:21AM -0700, Boqun Feng wrote:
> On Wed, Jun 25, 2025 at 06:07:06PM +0200, Uladzislau Rezki wrote:
> > On Wed, Jun 25, 2025 at 06:12:52AM -0700, Boqun Feng wrote:
> > > On Wed, Jun 25, 2025 at 08:30:26AM +0200, Vitaly Wool wrote:
> > > [...]
> > > > @@ -151,16 +159,11 @@ unsafe fn realloc(
> > > >          layout: Layout,
> > > >          old_layout: Layout,
> > > >          flags: Flags,
> > > > +        nid: Option<i32>,
> > > >      ) -> Result<NonNull<[u8]>, AllocError> {
> > > > -        // TODO: Support alignments larger than PAGE_SIZE.
> > > 
> > > Thanks a lot for doing this! While you're at it, maybe we can add a few
> > > tests for various alignments of allocation? I'm thinking:
> > > 
> > > #[repr(align(65536)]
> > > pub struct Test64k {
> > >     a: i32
> > > }
> > > 
> > > #[kunit_tests(rust_vbox)]
> > > mod tests {
> > >     #[test]
> > >     fn large_allocation() -> Result {
> > >         // Better use `new_uninit()` to avoid allocation on the stack.
> > >         let x = VBox::<Test64k>::new_uninit(...)?;
> > > 
> > > 	assert!(x.as_ptr().addr() & (kernel::sizes::SZ_64K - 1) == 0);
> > >     }
> > > }
> > > 
> > > Thoughts?
> > > 
> > > Regards,
> > > Boqun
> > > 
> > > > -        if layout.align() > bindings::PAGE_SIZE {
> > > > -            pr_warn!("Vmalloc does not support alignments larger than PAGE_SIZE yet.\n");
> > > > -            return Err(AllocError);
> > > > -        }
> > > > -
> > > >          // SAFETY: If not `None`, `ptr` is guaranteed to point to valid memory, which was previously
> > > >          // allocated with this `Allocator`.
> > > > -        unsafe { ReallocFunc::VREALLOC.call(ptr, layout, old_layout, flags) }
> > > > +        unsafe { ReallocFunc::VREALLOC.call(ptr, layout, old_layout, flags, nid) }
> > > >      }
> > > >  }
> > > >  
> > > [...]
> > >
> > At least we are lacking of vrealloc() exercising in the vmalloc-test suite.
> > I am not sure it makes a lot of sense to add a kunit test on top of rust-wrapper
> > around vrealloc().
> > 
> > From my side, i will add the test case to the test_vmalloc.c test-suite.
> > 
> 
> Thanks! But we will need these tests from Rust side anyway, to test
> 1) whether the Rust wrapper does the right thing, and 2) whether any C
> change cause the behavior changes on the API that Rust wrapper rely on.
> 
Ah. Got it :)

--
Uladzislau Rezki