During page table walking, ensure the range being processed matches
the expected size of a leaf entry (P4D, PUD, or PMD). While pxd_addr_end() 
functions handle boundary checks, they do not inherently validate whether 
the entry is a leaf node of the expected size.

Add WARN_ON_ONCE() to detect misalignments or partial updates of huge 
pages. If a mismatch is detected, return -EINVAL to prevent potential
corruption of page table entries.

Signed-off-by: Austin Kim <austindh.kim@gmail.com>
---
 arch/riscv/mm/pageattr.c | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/arch/riscv/mm/pageattr.c b/arch/riscv/mm/pageattr.c
index 3f76db3d2..1461bf77f 100644
--- a/arch/riscv/mm/pageattr.c
+++ b/arch/riscv/mm/pageattr.c
@@ -32,6 +32,8 @@ static int pageattr_p4d_entry(p4d_t *p4d, unsigned long addr,
 	p4d_t val = p4dp_get(p4d);
 
 	if (p4d_leaf(val)) {
+		if (WARN_ON_ONCE((next - addr) != P4D_SIZE))
+			return -EINVAL;
 		val = __p4d(set_pageattr_masks(p4d_val(val), walk));
 		set_p4d(p4d, val);
 	}
@@ -45,6 +47,8 @@ static int pageattr_pud_entry(pud_t *pud, unsigned long addr,
 	pud_t val = pudp_get(pud);
 
 	if (pud_leaf(val)) {
+		if (WARN_ON_ONCE((next - addr) != PUD_SIZE))
+			return -EINVAL;
 		val = __pud(set_pageattr_masks(pud_val(val), walk));
 		set_pud(pud, val);
 	}
@@ -58,6 +62,8 @@ static int pageattr_pmd_entry(pmd_t *pmd, unsigned long addr,
 	pmd_t val = pmdp_get(pmd);
 
 	if (pmd_leaf(val)) {
+		if (WARN_ON_ONCE((next - addr) != PMD_SIZE))
+			return -EINVAL;
 		val = __pmd(set_pageattr_masks(pmd_val(val), walk));
 		set_pmd(pmd, val);
 	}
-- 
2.34.1

Hi Austin,

On Mon, 20 Apr 2026, Austin Kim wrote:

> During page table walking, ensure the range being processed matches
> the expected size of a leaf entry (P4D, PUD, or PMD). While pxd_addr_end() 
> functions handle boundary checks, they do not inherently validate whether 
> the entry is a leaf node of the expected size.
> 
> Add WARN_ON_ONCE() to detect misalignments or partial updates of huge 
> pages. If a mismatch is detected, return -EINVAL to prevent potential
> corruption of page table entries.
> 
> Signed-off-by: Austin Kim <austindh.kim@gmail.com>

Is this in response to a failure mode that you've seen?  Or is it simply 
prophylactic?

Seems like a better place for this might be in common code, so other 
architectures could benefit as well?


- Paul

Hello Paul,

2026년 5월 23일 (토) 오전 8:34, Paul Walmsley <pjw@kernel.org>님이 작성:
>
> Hi Austin,
>
> On Mon, 20 Apr 2026, Austin Kim wrote:
>
> > During page table walking, ensure the range being processed matches
> > the expected size of a leaf entry (P4D, PUD, or PMD). While pxd_addr_end()
> > functions handle boundary checks, they do not inherently validate whether
> > the entry is a leaf node of the expected size.
> >
> > Add WARN_ON_ONCE() to detect misalignments or partial updates of huge
> > pages. If a mismatch is detected, return -EINVAL to prevent potential
> > corruption of page table entries.
> >
> > Signed-off-by: Austin Kim <austindh.kim@gmail.com>
>
> Is this in response to a failure mode that you've seen?  Or is it simply
> prophylactic?
>
> Seems like a better place for this might be in common code, so other
> architectures could benefit as well?

The answer is that this change is mainly prophylactic,
although it was tested on a RISC-V board to make sure
it does not introduce any unexpected side effects.

The pageattr_p4d_entry() function is registered as a callback
and is called from the page walker code:

mm/pagewalk.c
static int walk_p4d_range(pgd_t *pgd, unsigned long addr,
                          unsigned long end,
                          struct mm_walk *walk)
{
        ...
        if (ops->p4d_entry) {
                err = ops->p4d_entry(p4d, addr, next, walk);

With the current code flow, this condition should not be reachable in
normal operation.
The WARN_ON() is intended to catch unexpected situations if the page
walker behavior changes
in the future or if related code is modified.

I will also look for a more common place where a similar check
could be implemented so that other architectures may benefit as well.

Thank you for your feedback.

BR,
Austin Kim

>
>
> - Paul