[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <DG8WJPYVA0H1.1PO95BAW0TK3Y@kernel.org>
Date: Sat, 07 Feb 2026 18:23:05 +0100
From: "Danilo Krummrich" <dakr@...nel.org>
To: "Shivam Kalra via B4 Relay" <devnull+shivamklr.cock.li@...nel.org>
Cc: <shivamklr@...k.li>, "Lorenzo Stoakes" <lorenzo.stoakes@...cle.com>,
"Vlastimil Babka" <vbabka@...e.cz>, "Liam R. Howlett"
<Liam.Howlett@...cle.com>, "Uladzislau Rezki" <urezki@...il.com>, "Miguel
Ojeda" <ojeda@...nel.org>, "Boqun Feng" <boqun.feng@...il.com>, "Gary Guo"
<gary@...yguo.net>, Björn Roy Baron
<bjorn3_gh@...tonmail.com>, "Benno Lossin" <lossin@...nel.org>, "Andreas
Hindborg" <a.hindborg@...nel.org>, "Alice Ryhl" <aliceryhl@...gle.com>,
"Trevor Gross" <tmgross@...ch.edu>, "Greg Kroah-Hartman"
<gregkh@...uxfoundation.org>, Arve Hjønnevåg
<arve@...roid.com>, "Todd Kjos" <tkjos@...roid.com>, "Christian Brauner"
<brauner@...nel.org>, "Carlos Llamas" <cmllamas@...gle.com>,
<rust-for-linux@...r.kernel.org>, <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH v3 2/4] rust: kvec: implement shrink_to and
shrink_to_fit for Vec
On Sat Feb 7, 2026 at 12:32 PM CET, Shivam Kalra via B4 Relay wrote:
> +impl<T, A: Shrinkable> Vec<T, A> {
I don't think we should have a Shrinkable trait with is_shrinkable(). This is a
decision taken by the backing Allocator's realloc() function already.
Instead, shrink_to() should be a normal method of Vec<A, T> and just call
A::realloc().
For the temporary workaround we can have a temporary ShrinkQuirk trait that has
methods that take the same arguments as shrink_to().
In Vec::shrink_to() we can then hook in before calling A::realloc() and apply
the quirk.
fn shrink_to() {
if self.shrink_needs_quirk() {
return self.shrink_quirk();
}
// Allocator backend decides.
A::realloc();
}
> + pub fn shrink_to(&mut self, min_capacity: usize, flags: Flags) -> Result<(), AllocError> {
> + let target_cap = core::cmp::max(self.len(), min_capacity);
> +
> + if self.capacity() <= target_cap {
> + return Ok(());
> + }
> +
> + if Self::is_zst() {
> + return Ok(());
> + }
> +
> + // SAFETY: `self.ptr` is valid by the type invariant.
> + if !unsafe { A::is_shrinkable(self.ptr.cast()) } {
> + return Ok(());
> + }
> +
> + // Only shrink if we would free at least one page.
> + let current_size = self.capacity() * core::mem::size_of::<T>();
> + let target_size = target_cap * core::mem::size_of::<T>();
> + let current_pages = current_size.div_ceil(PAGE_SIZE);
> + let target_pages = target_size.div_ceil(PAGE_SIZE);
This is the specific heuristic we use for the Vmalloc shrink workaround
(including when for KVmalloc is_vmalloc_addr() is true) and it doesn't belong
into the common code path.
But this goes away anyways with the above changes.
> + if current_pages <= target_pages {
> + return Ok(());
> + }
> +
> + if target_cap == 0 {
> + if !self.layout.is_empty() {
> + // SAFETY: `self.ptr` was allocated with `A`, layout matches.
> + unsafe { A::free(self.ptr.cast(), self.layout.into()) };
> + }
> + self.ptr = NonNull::dangling();
> + self.layout = ArrayLayout::empty();
> + return Ok(());
> + }
> +
> + // SAFETY: `target_cap <= self.capacity()` and original capacity was valid.
> + let new_layout = unsafe { ArrayLayout::<T>::new_unchecked(target_cap) };
> +
> + // TODO: Once vrealloc supports in-place shrinking (mm/vmalloc.c:4316), this
> + // explicit alloc+copy+free can potentially be replaced with realloc.
> + let new_ptr = A::alloc(new_layout.into(), flags, NumaNode::NO_NODE)?;
> +
> + // SAFETY: Both pointers are valid, non-overlapping, and properly aligned.
> + unsafe {
> + ptr::copy_nonoverlapping(self.as_ptr(), new_ptr.as_ptr().cast::<T>(), self.len);
> + }
> +
> + // SAFETY: `self.ptr` was allocated with `A`, layout matches.
> + unsafe { A::free(self.ptr.cast(), self.layout.into()) };
> +
> + // SAFETY: `new_ptr` is non-null because `A::alloc` succeeded.
> + self.ptr = unsafe { NonNull::new_unchecked(new_ptr.as_ptr().cast::<T>()) };
> + self.layout = new_layout;
> +
> + Ok(())
> + }
> +
> + /// Shrinks the capacity of the vector as much as possible.
> + ///
> + /// This is equivalent to calling `shrink_to(0, flags)`. See [`Vec::shrink_to`] for details.
> + ///
> + /// # Examples
> + ///
> + /// ```
> + /// use kernel::alloc::allocator::Vmalloc;
> + ///
> + /// let elements_per_page = kernel::page::PAGE_SIZE / core::mem::size_of::<u32>();
> + /// let mut v: Vec<u32, Vmalloc> = Vec::with_capacity(elements_per_page * 4, GFP_KERNEL)?;
You can just use VVec<u32>.
> + /// v.push(1, GFP_KERNEL)?;
> + /// v.push(2, GFP_KERNEL)?;
> + /// v.push(3, GFP_KERNEL)?;
> + ///
> + /// v.shrink_to_fit(GFP_KERNEL)?;
> + /// # Ok::<(), Error>(())
> + /// ```
> + pub fn shrink_to_fit(&mut self, flags: Flags) -> Result<(), AllocError> {
> + self.shrink_to(0, flags)
> + }
> +}
> +
> impl<T: Clone, A: Allocator> Vec<T, A> {
> /// Extend the vector by `n` clones of `value`.
> pub fn extend_with(&mut self, n: usize, value: T, flags: Flags) -> Result<(), AllocError> {
>
> --
> 2.43.0
Powered by blists - more mailing lists