On Tue, Nov 4, 2025 at 3:49 PM Maciej Wieczor-Retman m.wieczorretman@pm.me wrote:

From: Maciej Wieczor-Retman maciej.wieczor-retman@intel.com

A KASAN tag mismatch, possibly causing a kernel panic, can be observed on systems with a tag-based KASAN enabled and with multiple NUMA nodes. It was reported on arm64 and reproduced on x86. It can be explained in the following points:

    1. There can be more than one virtual memory chunk.
    2. Chunk's base address has a tag.
    3. The base address points at the first chunk and thus inherits
       the tag of the first chunk.
    4. The subsequent chunks will be accessed with the tag from the
       first chunk.
    5. Thus, the subsequent chunks need to have their tag set to
       match that of the first chunk.

Refactor code by moving it into a helper in preparation for the actual fix.

Fixes: 1d96320f8d53 ("kasan, vmalloc: add vmalloc tagging for SW_TAGS") Cc: stable@vger.kernel.org # 6.1+ Signed-off-by: Maciej Wieczor-Retman maciej.wieczor-retman@intel.com Tested-by: Baoquan He bhe@redhat.com

---

Changelog v1 (after splitting of from the KASAN series):

• Rewrite first paragraph of the patch message to point at the user impact of the issue.
• Move helper to common.c so it can be compiled in all KASAN modes.

include/linux/kasan.h | 10 ++++++++++ mm/kasan/common.c | 11 +++++++++++ mm/vmalloc.c | 4 +--- 3 files changed, 22 insertions(+), 3 deletions(-)

diff --git a/include/linux/kasan.h b/include/linux/kasan.h index d12e1a5f5a9a..b00849ea8ffd 100644 --- a/include/linux/kasan.h +++ b/include/linux/kasan.h @@ -614,6 +614,13 @@ static __always_inline void kasan_poison_vmalloc(const void *start, __kasan_poison_vmalloc(start, size); }

+void __kasan_unpoison_vmap_areas(struct vm_struct **vms, int nr_vms); +static __always_inline void kasan_unpoison_vmap_areas(struct vm_struct **vms, int nr_vms) +{

•   if (kasan_enabled())
  

•           __kasan_unpoison_vmap_areas(vms, nr_vms);
  

+}

#else /* CONFIG_KASAN_VMALLOC */

static inline void kasan_populate_early_vm_area_shadow(void *start, @@ -638,6 +645,9 @@ static inline void *kasan_unpoison_vmalloc(const void *start, static inline void kasan_poison_vmalloc(const void *start, unsigned long size) { }

+static inline void kasan_unpoison_vmap_areas(struct vm_struct **vms, int nr_vms) +{ }

#endif /* CONFIG_KASAN_VMALLOC */

#if (defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)) && \ diff --git a/mm/kasan/common.c b/mm/kasan/common.c index d4c14359feaf..c63544a98c24 100644 --- a/mm/kasan/common.c +++ b/mm/kasan/common.c @@ -28,6 +28,7 @@ #include <linux/string.h> #include <linux/types.h> #include <linux/bug.h> +#include <linux/vmalloc.h>

#include "kasan.h" #include "../slab.h" @@ -582,3 +583,13 @@ bool __kasan_check_byte(const void *address, unsigned long ip) } return true; }

+void __kasan_unpoison_vmap_areas(struct vm_struct **vms, int nr_vms) +{

•   int area;
  

•   for (area = 0 ; area < nr_vms ; area++) {
  

•           kasan_poison(vms[area]->addr, vms[area]->size,
  

•                        arch_kasan_get_tag(vms[area]->addr), false);
  

The patch description says this patch is a refactoring, but the patch changes the logic of the code.

We don't call __kasan_unpoison_vmalloc() anymore and don't perform all the related checks. This might be OK, assuming the checks always succeed/fail, but this needs to be explained (note that there two versions of __kasan_unpoison_vmalloc() with different checks).

And also we don't assign a random tag anymore - we should.

Also, you can just use get/set_tag(), no need to use the arch_ version (and in the following patch too).

•   }
  

+} diff --git a/mm/vmalloc.c b/mm/vmalloc.c index 798b2ed21e46..934c8bfbcebf 100644 --- a/mm/vmalloc.c +++ b/mm/vmalloc.c @@ -4870,9 +4870,7 @@ struct vm_struct **pcpu_get_vm_areas(const unsigned long *offsets, * With hardware tag-based KASAN, marking is skipped for * non-VM_ALLOC mappings, see __kasan_unpoison_vmalloc(). */

•   for (area = 0; area < nr_vms; area++)
  

•           vms[area]->addr = kasan_unpoison_vmalloc(vms[area]->addr,
  

•                           vms[area]->size, KASAN_VMALLOC_PROT_NORMAL);
  

•   kasan_unpoison_vmap_areas(vms, nr_vms);
  
    kfree(vas);
    return vms;
  

-- 2.51.0