[PATCH v5 0/3] kasan: vmalloc: Fixes for the percpu allocator and vrealloc

[Maciej Wieczor-Retman]

Patches fix two issues related to KASAN and vmalloc.

The first one, a KASAN tag mismatch, possibly resulting in a kernel
panic, can be observed on systems with a tag-based KASAN enabled and
with multiple NUMA nodes. Initially it was only noticed on x86 [1] but
later a similar issue was also reported on arm64 [2].

Specifically the problem is related to how vm_structs interact with
pcpu_chunks - both when they are allocated, assigned and when pcpu_chunk
addresses are derived.

When vm_structs are allocated they are unpoisoned, each with a different
random tag, if vmalloc support is enabled along the KASAN mode. Later
when first pcpu chunk is allocated it gets its 'base_addr' field set to
the first allocated vm_struct. With that it inherits that vm_struct's
tag.

When pcpu_chunk addresses are later derived (by pcpu_chunk_addr(), for
example in pcpu_alloc_noprof()) the base_addr field is used and offsets
are added to it. If the initial conditions are satisfied then some of
the offsets will point into memory allocated with a different vm_struct.
So while the lower bits will get accurately derived the tag bits in the
top of the pointer won't match the shadow memory contents.

The solution (proposed at v2 of the x86 KASAN series [3]) is to unpoison
the vm_structs with the same tag when allocating them for the per cpu
allocator (in pcpu_get_vm_areas()).

The second one reported by syzkaller [4] is related to vrealloc and
happens because of random tag generation when unpoisoning memory without
allocating new pages. This breaks shadow memory tracking and needs to
reuse the existing tag instead of generating a new one. At the same time
an inconsistency in used flags is corrected.

The series is based on 6.19-rc1.

[1] https://lore.kernel.org/all/e7e04692866d02e6d3b32bb43b998e5d17092b...@intel.com/
[2] https://lore.kernel.org/all/aMUrW1Znp1GEj7St@MiWiFi-R3L-srv/
[3] https://lore.kernel.org/all/CAPAsAGxDRv_uFeMYu9TwhBVW...@mail.gmail.com/
[4] https://syzkaller.appspot.com/bug?extid=997752115a851cb0cf36

Changes v5:
- Rebased onto 6.19-rc1.

Changes v4:
- Added WARN_ON_ONCE() and removed pr_warn() from last patch.
- Added missing cc stable to the first patch.
- Fixed stray 'Changelog v1' in the patch messages.

Changes v3:
- Reworded the 4th and 5th paragraphs after finding the vms[] pointers
were untagged.
- Redo the patches by using a flag instead of a new
__kasan_vmalloc_unpoison() argument.
- Added Jiayuan's patch to the series.

Changes v2:
- Redid the patches since last version wasn't an actual refactor as the
patch promised.
- Also fixed multiple mistakes and retested everything.

Jiayuan Chen (1):
mm/kasan: Fix incorrect unpoisoning in vrealloc for KASAN

Maciej Wieczor-Retman (2):
kasan: Refactor pcpu kasan vmalloc unpoison
kasan: Unpoison vms[area] addresses with a common tag

include/linux/kasan.h | 16 ++++++++++++++++
mm/kasan/common.c | 32 ++++++++++++++++++++++++++++++++
mm/kasan/hw_tags.c | 2 +-
mm/kasan/shadow.c | 4 +++-
mm/vmalloc.c | 8 ++++----
5 files changed, 56 insertions(+), 6 deletions(-)

--
2.52.0

[Maciej Wieczor-Retman]

From: Jiayuan Chen <jiayua...@linux.dev>

Syzkaller reported a memory out-of-bounds bug [1]. This patch fixes two
issues:

1. In vrealloc the KASAN_VMALLOC_VM_ALLOC flag is missing when
unpoisoning the extended region. This flag is required to correctly
associate the allocation with KASAN's vmalloc tracking.

Note: In contrast, vzalloc (via __vmalloc_node_range_noprof) explicitly
sets KASAN_VMALLOC_VM_ALLOC and calls kasan_unpoison_vmalloc() with it.
vrealloc must behave consistently — especially when reusing existing
vmalloc regions — to ensure KASAN can track allocations correctly.

2. When vrealloc reuses an existing vmalloc region (without allocating
new pages) KASAN generates a new tag, which breaks tag-based memory
access tracking.

Introduce KASAN_VMALLOC_KEEP_TAG, a new KASAN flag that allows reusing
the tag already attached to the pointer, ensuring consistent tag
behavior during reallocation.

Pass KASAN_VMALLOC_KEEP_TAG and KASAN_VMALLOC_VM_ALLOC to the
kasan_unpoison_vmalloc inside vrealloc_node_align_noprof().

[1]: https://syzkaller.appspot.com/bug?extid=997752115a851cb0cf36

Fixes: a0309faf1cb0 ("mm: vmalloc: support more granular vrealloc() sizing")
Cc: <sta...@vger.kernel.org>
Reported-by: syzbot+997752...@syzkaller.appspotmail.com
Closes: https://lore.kernel.org/all/68e243a2.050a022...@google.com/T/
Signed-off-by: Jiayuan Chen <jiayua...@linux.dev>
Co-developed-by: Maciej Wieczor-Retman <maciej.wie...@intel.com>
Signed-off-by: Maciej Wieczor-Retman <maciej.wie...@intel.com>
Reviewed-by: Andrey Konovalov <andre...@gmail.com>
---
include/linux/kasan.h | 1 +

mm/kasan/hw_tags.c | 2 +-
mm/kasan/shadow.c | 4 +++-

mm/vmalloc.c | 4 +++-
4 files changed, 8 insertions(+), 3 deletions(-)

diff --git a/include/linux/kasan.h b/include/linux/kasan.h
index f335c1d7b61d..df3d8567dde9 100644
--- a/include/linux/kasan.h
+++ b/include/linux/kasan.h
@@ -28,6 +28,7 @@ typedef unsigned int __bitwise kasan_vmalloc_flags_t;
#define KASAN_VMALLOC_INIT ((__force kasan_vmalloc_flags_t)0x01u)
#define KASAN_VMALLOC_VM_ALLOC ((__force kasan_vmalloc_flags_t)0x02u)
#define KASAN_VMALLOC_PROT_NORMAL ((__force kasan_vmalloc_flags_t)0x04u)
+#define KASAN_VMALLOC_KEEP_TAG ((__force kasan_vmalloc_flags_t)0x08u)

#define KASAN_VMALLOC_PAGE_RANGE 0x1 /* Apply exsiting page range */
#define KASAN_VMALLOC_TLB_FLUSH 0x2 /* TLB flush */
diff --git a/mm/kasan/hw_tags.c b/mm/kasan/hw_tags.c
index 1c373cc4b3fa..cbef5e450954 100644
--- a/mm/kasan/hw_tags.c
+++ b/mm/kasan/hw_tags.c
@@ -361,7 +361,7 @@ void *__kasan_unpoison_vmalloc(const void *start, unsigned long size,
return (void *)start;
}

- tag = kasan_random_tag();
+ tag = (flags & KASAN_VMALLOC_KEEP_TAG) ? get_tag(start) : kasan_random_tag();
start = set_tag(start, tag);

/* Unpoison and initialize memory up to size. */
diff --git a/mm/kasan/shadow.c b/mm/kasan/shadow.c
index 29a751a8a08d..32fbdf759ea2 100644
--- a/mm/kasan/shadow.c
+++ b/mm/kasan/shadow.c
@@ -631,7 +631,9 @@ void *__kasan_unpoison_vmalloc(const void *start, unsigned long size,
!(flags & KASAN_VMALLOC_PROT_NORMAL))
return (void *)start;

- start = set_tag(start, kasan_random_tag());
+ if (unlikely(!(flags & KASAN_VMALLOC_KEEP_TAG)))
+ start = set_tag(start, kasan_random_tag());
+
kasan_unpoison(start, size, false);
return (void *)start;
}
diff --git a/mm/vmalloc.c b/mm/vmalloc.c
index ecbac900c35f..94c0a9262a46 100644
--- a/mm/vmalloc.c
+++ b/mm/vmalloc.c
@@ -4331,7 +4331,9 @@ void *vrealloc_node_align_noprof(const void *p, size_t size, unsigned long align
*/
if (size <= alloced_size) {
kasan_unpoison_vmalloc(p + old_size, size - old_size,
- KASAN_VMALLOC_PROT_NORMAL);
+ KASAN_VMALLOC_PROT_NORMAL |
+ KASAN_VMALLOC_VM_ALLOC |
+ KASAN_VMALLOC_KEEP_TAG);
/*
* No need to zero memory here, as unused memory will have
* already been zeroed at initial allocation time or during
--
2.52.0

[Maciej Wieczor-Retman]

From: Maciej Wieczor-Retman <maciej.wie...@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 reusing __kasan_unpoison_vmalloc in a new helper in
preparation for the actual fix.

Fixes: 1d96320f8d53 ("kasan, vmalloc: add vmalloc tagging for SW_TAGS")
Cc: sta...@vger.kernel.org # 6.1+
Reviewed-by: Andrey Konovalov <andre...@gmail.com>
Signed-off-by: Maciej Wieczor-Retman <maciej.wie...@intel.com>
---

Changelog v3:
- Redo the patch after applying Andrey's comments to align the code more
with what's already in include/linux/kasan.h

Changelog v2:
- Redo the whole patch so it's an actual refactor.

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 | 15 +++++++++++++++
mm/kasan/common.c | 17 +++++++++++++++++
mm/vmalloc.c | 4 +---
3 files changed, 33 insertions(+), 3 deletions(-)

diff --git a/include/linux/kasan.h b/include/linux/kasan.h
index df3d8567dde9..9c6ac4b62eb9 100644
--- a/include/linux/kasan.h
+++ b/include/linux/kasan.h
@@ -631,6 +631,16 @@ 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,
+ kasan_vmalloc_flags_t flags);
+static __always_inline void
+kasan_unpoison_vmap_areas(struct vm_struct **vms, int nr_vms,
+ kasan_vmalloc_flags_t flags)
+{
+ if (kasan_enabled())
+ __kasan_unpoison_vmap_areas(vms, nr_vms, flags);
+}
+
#else /* CONFIG_KASAN_VMALLOC */

static inline void kasan_populate_early_vm_area_shadow(void *start,
@@ -655,6 +665,11 @@ static inline void *kasan_unpoison_vmalloc(const void *start,
static inline void kasan_poison_vmalloc(const void *start, unsigned long size)
{ }

+static __always_inline void
+kasan_unpoison_vmap_areas(struct vm_struct **vms, int nr_vms,
+ kasan_vmalloc_flags_t flags)
+{ }
+
#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 1d27f1bd260b..b2b40c59ce18 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"
@@ -575,3 +576,19 @@ bool __kasan_check_byte(const void *address, unsigned long ip)
}
return true;
}
+
+#ifdef CONFIG_KASAN_VMALLOC
+void __kasan_unpoison_vmap_areas(struct vm_struct **vms, int nr_vms,
+ kasan_vmalloc_flags_t flags)
+{
+ unsigned long size;
+ void *addr;
+ int area;
+
+ for (area = 0 ; area < nr_vms ; area++) {
+ size = vms[area]->size;
+ addr = vms[area]->addr;
+ vms[area]->addr = __kasan_unpoison_vmalloc(addr, size, flags);
+ }
+}
+#endif
diff --git a/mm/vmalloc.c b/mm/vmalloc.c
index 94c0a9262a46..41dd01e8430c 100644
--- a/mm/vmalloc.c
+++ b/mm/vmalloc.c
@@ -5027,9 +5027,7 @@ 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, KASAN_VMALLOC_PROT_NORMAL);

kfree(vas);
return vms;
--
2.52.0

[Maciej Wieczor-Retman]

From: Maciej Wieczor-Retman <maciej.wie...@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.

Use the new vmalloc flag that disables random tag assignment in
__kasan_unpoison_vmalloc() - pass the same random tag to all the
vm_structs by tagging the pointers before they go inside
__kasan_unpoison_vmalloc(). Assigning a common tag resolves the pcpu
chunk address mismatch.

Fixes: 1d96320f8d53 ("kasan, vmalloc: add vmalloc tagging for SW_TAGS")
Cc: sta...@vger.kernel.org # 6.1+
Reviewed-by: Andrey Konovalov <andre...@gmail.com>
Signed-off-by: Maciej Wieczor-Retman <maciej.wie...@intel.com>
---

Changelog v4:
- Add WARN_ON_ONCE() if the new flag is already set in the helper.
(Andrey)
- Remove pr_warn() since the comment should be enough. (Andrey)

Changelog v3:
- Redo the patch by using a flag instead of a new argument in
__kasan_unpoison_vmalloc() (Andrey Konovalov)

Changelog v2:
- Revise the whole patch to match the fixed refactorization from the
first patch.

Changelog v1:
- Rewrite 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.

mm/kasan/common.c | 21 ++++++++++++++++++---
1 file changed, 18 insertions(+), 3 deletions(-)

diff --git a/mm/kasan/common.c b/mm/kasan/common.c
index b2b40c59ce18..ed489a14dddf 100644
--- a/mm/kasan/common.c
+++ b/mm/kasan/common.c
@@ -584,11 +584,26 @@ void __kasan_unpoison_vmap_areas(struct vm_struct **vms, int nr_vms,
unsigned long size;
void *addr;
int area;
+ u8 tag;
+
+ /*
+ * If KASAN_VMALLOC_KEEP_TAG was set at this point, all vms[] pointers
+ * would be unpoisoned with the KASAN_TAG_KERNEL which would disable
+ * KASAN checks down the line.
+ */
+ if (WARN_ON_ONCE(flags & KASAN_VMALLOC_KEEP_TAG))
+ return;
+
+ size = vms[0]->size;
+ addr = vms[0]->addr;
+ vms[0]->addr = __kasan_unpoison_vmalloc(addr, size, flags);
+ tag = get_tag(vms[0]->addr);

- for (area = 0 ; area < nr_vms ; area++) {
+ for (area = 1 ; area < nr_vms ; area++) {

size = vms[area]->size;

- addr = vms[area]->addr;
- vms[area]->addr = __kasan_unpoison_vmalloc(addr, size, flags);
+ addr = set_tag(vms[area]->addr, tag);

+ vms[area]->addr =

+ __kasan_unpoison_vmalloc(addr, size, flags | KASAN_VMALLOC_KEEP_TAG);
}
}
#endif
--
2.52.0

[Andrew Morton]

On Wed, 17 Dec 2025 13:48:44 +0000 Maciej Wieczor-Retman <m.wiecz...@pm.me> wrote:

> Patches fix two issues related to KASAN and vmalloc.

Thanks for the refresh. This is the same as we already have in
mm.git's mm-hotfixes-unstable branch. So I updated the changelog
metadata a bit and left the v4 series in place.