[PATCH 0/3] riscv: mm: clean up and optimize some page table get-and-clear functions

[PATCH 0/3] riscv: mm: clean up and optimize some page table get-and-clear functions

[Paul Walmsley]

This series implements some minor cleanups on the RISC-V page table
get-and-clear functions.  We continue the work from commit
546e42c8c6d94 ("riscv: Use an atomic xchg in
pudp_huge_get_and_clear()").  The two themes are:

- avoid atomic operations on non-SMP kernels

- use xchg() rather than atomic_long_xchg() on variables that aren't
  atomic_long_t

Tested on QEMU 10.0.6 on RV32 and RV64, SMP and !SMP, running kselftests.


- Paul

Paul Walmsley (3):
  riscv: mm: pmdp_huge_get_and_clear(): avoid atomic ops when
    !CONFIG_SMP
  riscv: mm: ptep_get_and_clear(): avoid atomic ops when !CONFIG_SMP
  riscv: mm: use xchg() on non-atomic_long_t variables, not
    atomic_long_xchg()

 arch/riscv/include/asm/pgtable.h | 16 ++++++++++++++--
 1 file changed, 14 insertions(+), 2 deletions(-)


base-commit: dcb6fa37fd7bc9c3d2b066329b0d27dedf8becaa
-- 
2.48.1


_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv

[PATCH 1/3] riscv: mm: pmdp_huge_get_and_clear(): avoid atomic ops when !CONFIG_SMP

[Paul Walmsley]

When !CONFIG_SMP, there's no need for atomic operations in
pmdp_huge_get_and_clear(), so, similar to what x86 does, let's not use
atomics in this case.  See also commit 546e42c8c6d94 ("riscv: Use an
atomic xchg in pudp_huge_get_and_clear()").

Cc: Alexandre Ghiti <alex@ghiti.fr>
Signed-off-by: Paul Walmsley <pjw@kernel.org>
---
 arch/riscv/include/asm/pgtable.h | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/arch/riscv/include/asm/pgtable.h b/arch/riscv/include/asm/pgtable.h
index 5a08eb5fe99f..64a2197e6d97 100644
--- a/arch/riscv/include/asm/pgtable.h
+++ b/arch/riscv/include/asm/pgtable.h
@@ -861,7 +861,13 @@ static inline int pmdp_test_and_clear_young(struct vm_area_struct *vma,
 static inline pmd_t pmdp_huge_get_and_clear(struct mm_struct *mm,
 					unsigned long address, pmd_t *pmdp)
 {
+#ifdef CONFIG_SMP
 	pmd_t pmd = __pmd(atomic_long_xchg((atomic_long_t *)pmdp, 0));
+#else
+	pmd_t pmd = *pmdp;
+
+	pmd_clear(pmdp);
+#endif
 
 	page_table_check_pmd_clear(mm, pmd);
 
-- 
2.48.1


_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv

[PATCH 2/3] riscv: mm: ptep_get_and_clear(): avoid atomic ops when !CONFIG_SMP

[Paul Walmsley]

When !CONFIG_SMP, there's no need for atomic operations in
ptep_get_and_clear(), so, similar to x86, let's not use atomics in
this case.

Cc: Alexandre Ghiti <alex@ghiti.fr>
Signed-off-by: Paul Walmsley <pjw@kernel.org>
---
 arch/riscv/include/asm/pgtable.h | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/arch/riscv/include/asm/pgtable.h b/arch/riscv/include/asm/pgtable.h
index 64a2197e6d97..e1610d031199 100644
--- a/arch/riscv/include/asm/pgtable.h
+++ b/arch/riscv/include/asm/pgtable.h
@@ -590,7 +590,13 @@ extern int ptep_test_and_clear_young(struct vm_area_struct *vma, unsigned long a
 static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
 				       unsigned long address, pte_t *ptep)
 {
+#ifdef CONFIG_SMP
 	pte_t pte = __pte(atomic_long_xchg((atomic_long_t *)ptep, 0));
+#else
+	pte_t pte = *ptep;
+
+	set_pte(ptep, __pte(0));
+#endif
 
 	page_table_check_pte_clear(mm, pte);
 
-- 
2.48.1


_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv

[PATCH 3/3] riscv: mm: use xchg() on non-atomic_long_t variables, not atomic_long_xchg()

[Paul Walmsley]

Let's not call atomic_long_xchg() on something that's not an
atomic_long_t, and just use xchg() instead.  Continues the cleanup
from commit 546e42c8c6d94 ("riscv: Use an atomic xchg in
pudp_huge_get_and_clear()"),

Cc: Alexandre Ghiti <alex@ghiti.fr>
Signed-off-by: Paul Walmsley <pjw@kernel.org>
---
 arch/riscv/include/asm/pgtable.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/arch/riscv/include/asm/pgtable.h b/arch/riscv/include/asm/pgtable.h
index e1610d031199..a1ffa502739b 100644
--- a/arch/riscv/include/asm/pgtable.h
+++ b/arch/riscv/include/asm/pgtable.h
@@ -591,7 +591,7 @@ static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
 				       unsigned long address, pte_t *ptep)
 {
 #ifdef CONFIG_SMP
-	pte_t pte = __pte(atomic_long_xchg((atomic_long_t *)ptep, 0));
+	pte_t pte = __pte(xchg(&ptep->pte, 0));
 #else
 	pte_t pte = *ptep;
 
@@ -868,7 +868,7 @@ static inline pmd_t pmdp_huge_get_and_clear(struct mm_struct *mm,
 					unsigned long address, pmd_t *pmdp)
 {
 #ifdef CONFIG_SMP
-	pmd_t pmd = __pmd(atomic_long_xchg((atomic_long_t *)pmdp, 0));
+	pmd_t pmd = __pmd(xchg(&pmdp->pmd, 0));
 #else
 	pmd_t pmd = *pmdp;
 
-- 
2.48.1


_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv

Re: [PATCH 0/3] riscv: mm: clean up and optimize some page table get-and-clear functions

[patchwork-bot+linux-riscv]

Hello:

This series was applied to riscv/linux.git (for-next)
by Paul Walmsley <pjw@kernel.org>:

On Wed, 19 Nov 2025 21:18:27 -0700 you wrote:
> This series implements some minor cleanups on the RISC-V page table
> get-and-clear functions.  We continue the work from commit
> 546e42c8c6d94 ("riscv: Use an atomic xchg in
> pudp_huge_get_and_clear()").  The two themes are:
> 
> - avoid atomic operations on non-SMP kernels
> 
> [...]

Here is the summary with links:
  - [1/3] riscv: mm: pmdp_huge_get_and_clear(): avoid atomic ops when !CONFIG_SMP
    https://git.kernel.org/riscv/c/44d98d7575d0
  - [2/3] riscv: mm: ptep_get_and_clear(): avoid atomic ops when !CONFIG_SMP
    https://git.kernel.org/riscv/c/ad6a7d4bcee7
  - [3/3] riscv: mm: use xchg() on non-atomic_long_t variables, not atomic_long_xchg()
    https://git.kernel.org/riscv/c/36313cea1273

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html



_______________________________________________
linux-riscv mailing list
linux-riscv@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-riscv