* [PATCH] rust: alloc: reject offline NUMA nodes in allocator
@ 2026-11-20 2:16 Franklin-Qi
0 siblings, 0 replies; only message in thread
From: Franklin-Qi @ 2026-11-20 2:16 UTC (permalink / raw)
To: Miguel Ojeda, Danilo Krummrich, Alice Ryhl, Vlastimil Babka
Cc: rust-for-linux, linux-mm, Boqun Feng, Gary Guo,
Björn Roy Baron, Benno Lossin, Andreas Hindborg,
Trevor Gross, Daniel Almeida, Tamir Duberstein, Alexandre Courbot,
Onur Özkan, Lorenzo Stoakes, Liam R . Howlett,
Uladzislau Rezki, Harry Yoo, Georgios Androutsopoulos,
Vitaly Wool, stable, Franklin-Qi
NumaNode::new() only checks that the id is in [0, MAX_NUMNODES). Safe
Allocator APIs can still forward an offline node into
k[v]realloc_node_align(), which may dereference NODE_DATA(nid) == NULL
and panic (GPF).
Reject offline nodes at the shared ReallocFunc boundary for nonzero
allocations. Leave NUMA_NO_NODE and zero-sized free/realloc paths
unchanged. A residual hotplug TOCTOU remains if a node goes offline
after the check; this matches practical C usage for rejecting
statically missing nodes from safe Rust.
Fixes: 7760b6421b6c ("rust: add support for NUMA ids in allocations")
Cc: stable@vger.kernel.org
Reported-by: Georgios Androutsopoulos <georgeandrout13@gmail.com>
Closes: https://github.com/Rust-for-Linux/linux/issues/1263
Signed-off-by: Franklin-Qi <yusq77@gmail.com>
---
rust/helpers/mm.c | 6 ++++++
rust/kernel/alloc/allocator.rs | 18 ++++++++++++++++++
2 files changed, 24 insertions(+)
diff --git a/rust/helpers/mm.c b/rust/helpers/mm.c
index b5540997b..b61e54e16 100644
--- a/rust/helpers/mm.c
+++ b/rust/helpers/mm.c
@@ -1,6 +1,7 @@
// SPDX-License-Identifier: GPL-2.0
#include <linux/mm.h>
+#include <linux/nodemask.h>
#include <linux/sched/mm.h>
__rust_helper void rust_helper_mmgrab(struct mm_struct *mm)
@@ -38,6 +39,11 @@ __rust_helper void rust_helper_mmap_read_unlock(struct mm_struct *mm)
mmap_read_unlock(mm);
}
+__rust_helper bool rust_helper_node_online(int nid)
+{
+ return node_online(nid);
+}
+
__rust_helper struct vm_area_struct *
rust_helper_vma_lookup(struct mm_struct *mm, unsigned long addr)
{
diff --git a/rust/kernel/alloc/allocator.rs b/rust/kernel/alloc/allocator.rs
index cd4203f27..15e17f598 100644
--- a/rust/kernel/alloc/allocator.rs
+++ b/rust/kernel/alloc/allocator.rs
@@ -101,6 +101,23 @@ unsafe fn call(
nid: NumaNode,
) -> Result<NonNull<[u8]>, AllocError> {
let size = layout.size();
+
+ // `NumaNode::new` validates the ID range, not whether the node is online.
+ // The C allocation paths require an online node for nonzero allocations.
+ // Keep `NUMA_NO_NODE` and zero-sized deallocations unchanged.
+ //
+ // Residual TOCTOU remains if a node goes offline after this check and
+ // before the allocator runs; that matches practical C usage for rejecting
+ // statically missing/offline nodes from safe Rust.
+ if size != 0 && nid != NumaNode::NO_NODE {
+ // SAFETY: The helper only reads the kernel's online-node mask. `nid`
+ // is either `NUMA_NO_NODE` (excluded above) or a value in
+ // `[0, MAX_NUMNODES)` from `NumaNode::new`.
+ if !unsafe { bindings::node_online(nid.0) } {
+ return Err(AllocError);
+ }
+ }
+
let ptr = match ptr {
Some(ptr) => {
if old_layout.size() == 0 {
@@ -116,6 +133,7 @@ unsafe fn call(
// - `self.0` is one of `krealloc`, `vrealloc`, `kvrealloc` and thus only requires that
// `ptr` is NULL or valid.
// - `ptr` is either NULL or valid by the safety requirements of this function.
+ // - For nonzero allocations with a specific node, `nid` was checked online above.
//
// GUARANTEE:
// - `self.0` is one of `krealloc`, `vrealloc`, `kvrealloc`.
--
2.55.0
^ permalink raw reply related [flat|nested] only message in thread
only message in thread, other threads:[~2026-11-20 2:17 UTC | newest]
Thread overview: (only message) (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-11-20 2:16 [PATCH] rust: alloc: reject offline NUMA nodes in allocator Franklin-Qi
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox