[PATCH AUTOSEL 6.19-6.12] arm64: hugetlbpage: avoid unused-but-set-parameter warning (gcc-16)

[Sasha Levin]

From: Arnd Bergmann <arnd@xxxxxxxx>

[ Upstream commit 729a2e8e9ac47099a967567389cc9d73ef4194ca ]

gcc-16 warns about an instance that older compilers did not:

arch/arm64/mm/hugetlbpage.c: In function 'huge_pte_clear':
arch/arm64/mm/hugetlbpage.c:369:57: error: parameter 'addr' set but not used [-Werror=unused-but-set-parameter=]

The issue here is that __pte_clear() does not actually use its second
argument, but when CONFIG_ARM64_CONTPTE is enabled it still gets
updated.

Replace the macro with an inline function to let the compiler see
the argument getting passed down.

Suggested-by: Catalin Marinas <catalin.marinas@xxxxxxx>
Signed-off-by: Arnd Bergmann <arnd@xxxxxxxx>
Reviewed-by: Dev Jain <dev.jain@xxxxxxx>
Signed-off-by: Will Deacon <will@xxxxxxxxxx>
Signed-off-by: Sasha Levin <sashal@xxxxxxxxxx>
---

LLM Generated explanations, may be completely bogus:

## Analysis

### What the commit does

This commit replaces a `__pte_clear` macro with an equivalent inline
function to fix a gcc-16 build warning/error:

**Old (macro):**
```c
#define __pte_clear(mm, addr, ptep) \
    __set_pte(ptep, __pte(0))
```

**New (inline function):**
```c
static inline void __pte_clear(struct mm_struct *mm,
                               unsigned long addr, pte_t *ptep)
{
    __set_pte(ptep, __pte(0));
}
```

The two are functionally identical. The only difference is that the
inline function form allows the compiler to understand that `mm` and
`addr` are intentionally unused parameters passed through to satisfy the
interface, suppressing the `-Werror=unused-but-set-parameter` warning
that gcc-16 newly introduces.

### Classification: Build Fix

This is a **build fix** — it prevents compilation failure with gcc-16
when `-Werror` is enabled. Build fixes are explicitly listed as
acceptable stable material in the stable kernel rules.

### Risk Assessment

**Risk: Extremely low.** The macro-to-inline-function conversion
produces identical compiled output. The function body is identical
(`__set_pte(ptep, __pte(0))`), and the compiler will inline it, making
the generated code exactly the same. There is zero behavioral change at
runtime.

### Scope

- **1 file changed** (`arch/arm64/include/asm/pgtable.h`)
- Small, surgical change — macro removed, inline function added in
  appropriate location
- No logic changes whatsoever

### Considerations Against Backporting

- **gcc-16 is very new** — most stable kernel users and enterprise
  distributions won't use it for some time
- This is not a runtime bug — it only affects compilation
- The warning only becomes an error with `-Werror=unused-but-set-
  parameter` (which gcc-16 enables by default or through `-Werror`)

### Considerations For Backporting

- Build fixes are explicitly listed as stable-worthy in stable kernel
  rules
- The change is **zero risk** — functionally identical
- As distributions and users adopt gcc-16, they will hit this on older
  kernels
- Already reviewed and accepted by arm64 maintainers (Reviewed-by: Dev
  Jain, Signed-off-by: Will Deacon)

### Verification

- Verified the old macro and new inline function are functionally
  identical by reading the diff — both call `__set_pte(ptep, __pte(0))`
  and discard `mm` and `addr`
- Verified this is purely a build fix with no runtime behavioral change
- The commit message clearly describes the gcc-16 warning/error and the
  solution
- The commit has proper review chain (Suggested-by: Catalin Marinas,
  Reviewed-by: Dev Jain, Signed-off-by: Will Deacon — all arm64
  maintainers/reviewers)
- Could NOT verify whether stable trees already have other gcc-16 build
  fixes (unverified, but irrelevant to the merits of this specific fix)

### Conclusion

This is a zero-risk build fix that prevents compilation failure with
gcc-16. While gcc-16 is new and most stable users won't encounter this
immediately, the change is so low-risk (functionally identical macro-to-
inline conversion) that the benefit of proactively fixing compilation
clearly outweighs the negligible risk. Build fixes are explicitly called
out as stable material.

**YES**

 arch/arm64/include/asm/pgtable.h | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/arch/arm64/include/asm/pgtable.h b/arch/arm64/include/asm/pgtable.h
index 64d5f1d9cce96..5ab5fe3bef25e 100644
--- a/arch/arm64/include/asm/pgtable.h
+++ b/arch/arm64/include/asm/pgtable.h
@@ -179,8 +179,6 @@ static inline pteval_t __phys_to_pte_val(phys_addr_t phys)
 	__pte(__phys_to_pte_val((phys_addr_t)(pfn) << PAGE_SHIFT) | pgprot_val(prot))
 
 #define pte_none(pte)		(!pte_val(pte))
-#define __pte_clear(mm, addr, ptep) \
-				__set_pte(ptep, __pte(0))
 #define pte_page(pte)		(pfn_to_page(pte_pfn(pte)))
 
 /*
@@ -1320,6 +1318,13 @@ static inline bool pud_user_accessible_page(pud_t pud)
 /*
  * Atomic pte/pmd modifications.
  */
+
+static inline void __pte_clear(struct mm_struct *mm,
+			       unsigned long addr, pte_t *ptep)
+{
+	__set_pte(ptep, __pte(0));
+}
+
 static inline int __ptep_test_and_clear_young(struct vm_area_struct *vma,
 					      unsigned long address,
 					      pte_t *ptep)
-- 
2.51.0