* [PATCH 00/12] Make riscv use THP contpte support for arm64
@ 2024-05-08 19:19 Alexandre Ghiti
2024-05-08 19:19 ` [PATCH 01/12] mm, arm64: Rename ARM64_CONTPTE to THP_CONTPTE Alexandre Ghiti
` (11 more replies)
0 siblings, 12 replies; 16+ messages in thread
From: Alexandre Ghiti @ 2024-05-08 19:19 UTC (permalink / raw
To: Ryan Roberts, Catalin Marinas, Will Deacon, Alexander Potapenko,
Marco Elver, Dmitry Vyukov, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Ard Biesheuvel, Anup Patel, Atish Patra,
Andrey Ryabinin, Andrey Konovalov, Vincenzo Frascino,
Andrew Morton, linux-arm-kernel, linux-kernel, kasan-dev,
linux-riscv, linux-efi, kvm, kvm-riscv, linux-mm
Cc: Alexandre Ghiti
This allows riscv to support napot (riscv equivalent to contpte) THPs by
moving arm64 contpte support into mm, the previous series [1] only merging
riscv and arm64 implementations of hugetlbfs contpte.
riscv contpte specification allows for different contpte sizes, although
only 64KB is supported for now. So in this patchset is implemented the
support of multiple contpte sizes, which introduces a few arch specific
helpers to determine what sizes are supported. Even though only one size
is supported on riscv, the implementation of the multi size support is to
show what it will look like when we support other sizes, and make sure
it does not regress arm64.
I tested arm64 using the cow kselftest and a kernel build with 4KB base
page size and 64KB contpte. riscv was tested with the same tests on *all*
contpte sizes that fit in the last page table level (support for PMD sizes
is not present here). Both arch were only tested on qemu.
Alexandre Ghiti (12):
mm, arm64: Rename ARM64_CONTPTE to THP_CONTPTE
mm, riscv, arm64: Use common ptep_get() function
mm, riscv, arm64: Use common set_ptes() function
mm, riscv, arm64: Use common ptep_get_lockless() function
mm, riscv, arm64: Use common set_pte() function
mm, riscv, arm64: Use common pte_clear() function
mm, riscv, arm64: Use common ptep_get_and_clear() function
mm, riscv, arm64: Use common ptep_test_and_clear_young() function
mm, riscv, arm64: Use common ptep_clear_flush_young() function
mm, riscv, arm64: Use common ptep_set_access_flags() function
mm, riscv, arm64: Use common ptep_set_wrprotect()/wrprotect_ptes()
functions
mm, riscv, arm64: Use common
get_and_clear_full_ptes()/clear_full_ptes() functions
arch/arm64/Kconfig | 9 -
arch/arm64/include/asm/pgtable.h | 318 +++++---------
arch/arm64/mm/Makefile | 1 -
arch/arm64/mm/contpte.c | 408 ------------------
arch/arm64/mm/hugetlbpage.c | 6 +-
arch/arm64/mm/mmu.c | 2 +-
arch/riscv/include/asm/kfence.h | 4 +-
arch/riscv/include/asm/pgtable.h | 206 +++++++++-
arch/riscv/kernel/efi.c | 4 +-
arch/riscv/kernel/hibernate.c | 2 +-
arch/riscv/kvm/mmu.c | 26 +-
arch/riscv/mm/fault.c | 2 +-
arch/riscv/mm/init.c | 4 +-
arch/riscv/mm/kasan_init.c | 16 +-
arch/riscv/mm/pageattr.c | 8 +-
arch/riscv/mm/pgtable.c | 6 +-
include/linux/contpte.h | 37 ++
mm/Kconfig | 9 +
mm/contpte.c | 685 ++++++++++++++++++++++++++++++-
19 files changed, 1056 insertions(+), 697 deletions(-)
delete mode 100644 arch/arm64/mm/contpte.c
create mode 100644 include/linux/contpte.h
--
2.39.2
^ permalink raw reply [flat|nested] 16+ messages in thread
* [PATCH 01/12] mm, arm64: Rename ARM64_CONTPTE to THP_CONTPTE
2024-05-08 19:19 [PATCH 00/12] Make riscv use THP contpte support for arm64 Alexandre Ghiti
@ 2024-05-08 19:19 ` Alexandre Ghiti
2024-05-09 0:46 ` Barry Song
2024-05-08 19:19 ` [PATCH 02/12] mm, riscv, arm64: Use common ptep_get() function Alexandre Ghiti
` (10 subsequent siblings)
11 siblings, 1 reply; 16+ messages in thread
From: Alexandre Ghiti @ 2024-05-08 19:19 UTC (permalink / raw
To: Ryan Roberts, Catalin Marinas, Will Deacon, Alexander Potapenko,
Marco Elver, Dmitry Vyukov, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Ard Biesheuvel, Anup Patel, Atish Patra,
Andrey Ryabinin, Andrey Konovalov, Vincenzo Frascino,
Andrew Morton, linux-arm-kernel, linux-kernel, kasan-dev,
linux-riscv, linux-efi, kvm, kvm-riscv, linux-mm
Cc: Alexandre Ghiti
The ARM64_CONTPTE config represents the capability to transparently use
contpte mappings for THP userspace mappings, which will be implemented
in the next commits for riscv, so make this config more generic and move
it to mm.
Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com>
---
arch/arm64/Kconfig | 9 ---------
arch/arm64/include/asm/pgtable.h | 6 +++---
arch/arm64/mm/Makefile | 2 +-
mm/Kconfig | 9 +++++++++
4 files changed, 13 insertions(+), 13 deletions(-)
diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
index ac2f6d906cc3..9d823015b4e5 100644
--- a/arch/arm64/Kconfig
+++ b/arch/arm64/Kconfig
@@ -2227,15 +2227,6 @@ config UNWIND_PATCH_PAC_INTO_SCS
select UNWIND_TABLES
select DYNAMIC_SCS
-config ARM64_CONTPTE
- bool "Contiguous PTE mappings for user memory" if EXPERT
- depends on TRANSPARENT_HUGEPAGE
- default y
- help
- When enabled, user mappings are configured using the PTE contiguous
- bit, for any mappings that meet the size and alignment requirements.
- This reduces TLB pressure and improves performance.
-
endmenu # "Kernel Features"
menu "Boot options"
diff --git a/arch/arm64/include/asm/pgtable.h b/arch/arm64/include/asm/pgtable.h
index 7c2938cb70b9..1758ce71fae9 100644
--- a/arch/arm64/include/asm/pgtable.h
+++ b/arch/arm64/include/asm/pgtable.h
@@ -1369,7 +1369,7 @@ extern void ptep_modify_prot_commit(struct vm_area_struct *vma,
unsigned long addr, pte_t *ptep,
pte_t old_pte, pte_t new_pte);
-#ifdef CONFIG_ARM64_CONTPTE
+#ifdef CONFIG_THP_CONTPTE
/*
* The contpte APIs are used to transparently manage the contiguous bit in ptes
@@ -1622,7 +1622,7 @@ static inline int ptep_set_access_flags(struct vm_area_struct *vma,
return contpte_ptep_set_access_flags(vma, addr, ptep, entry, dirty);
}
-#else /* CONFIG_ARM64_CONTPTE */
+#else /* CONFIG_THP_CONTPTE */
#define ptep_get __ptep_get
#define set_pte __set_pte
@@ -1642,7 +1642,7 @@ static inline int ptep_set_access_flags(struct vm_area_struct *vma,
#define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
#define ptep_set_access_flags __ptep_set_access_flags
-#endif /* CONFIG_ARM64_CONTPTE */
+#endif /* CONFIG_THP_CONTPTE */
int find_num_contig(struct mm_struct *mm, unsigned long addr,
pte_t *ptep, size_t *pgsize);
diff --git a/arch/arm64/mm/Makefile b/arch/arm64/mm/Makefile
index 60454256945b..52a1b2082627 100644
--- a/arch/arm64/mm/Makefile
+++ b/arch/arm64/mm/Makefile
@@ -3,7 +3,7 @@ obj-y := dma-mapping.o extable.o fault.o init.o \
cache.o copypage.o flush.o \
ioremap.o mmap.o pgd.o mmu.o \
context.o proc.o pageattr.o fixmap.o
-obj-$(CONFIG_ARM64_CONTPTE) += contpte.o
+obj-$(CONFIG_THP_CONTPTE) += contpte.o
obj-$(CONFIG_HUGETLB_PAGE) += hugetlbpage.o
obj-$(CONFIG_PTDUMP_CORE) += ptdump.o
obj-$(CONFIG_PTDUMP_DEBUGFS) += ptdump_debugfs.o
diff --git a/mm/Kconfig b/mm/Kconfig
index c325003d6552..fd4de221a1c6 100644
--- a/mm/Kconfig
+++ b/mm/Kconfig
@@ -984,6 +984,15 @@ config ARCH_HAS_CACHE_LINE_SIZE
config ARCH_HAS_CONTPTE
bool
+config THP_CONTPTE
+ bool "Contiguous PTE mappings for user memory" if EXPERT
+ depends on ARCH_HAS_CONTPTE && TRANSPARENT_HUGEPAGE
+ default y
+ help
+ When enabled, user mappings are configured using the PTE contiguous
+ bit, for any mappings that meet the size and alignment requirements.
+ This reduces TLB pressure and improves performance.
+
config ARCH_HAS_CURRENT_STACK_POINTER
bool
help
--
2.39.2
^ permalink raw reply related [flat|nested] 16+ messages in thread
* [PATCH 02/12] mm, riscv, arm64: Use common ptep_get() function
2024-05-08 19:19 [PATCH 00/12] Make riscv use THP contpte support for arm64 Alexandre Ghiti
2024-05-08 19:19 ` [PATCH 01/12] mm, arm64: Rename ARM64_CONTPTE to THP_CONTPTE Alexandre Ghiti
@ 2024-05-08 19:19 ` Alexandre Ghiti
2024-05-08 19:19 ` [PATCH 03/12] mm, riscv, arm64: Use common set_ptes() function Alexandre Ghiti
` (9 subsequent siblings)
11 siblings, 0 replies; 16+ messages in thread
From: Alexandre Ghiti @ 2024-05-08 19:19 UTC (permalink / raw
To: Ryan Roberts, Catalin Marinas, Will Deacon, Alexander Potapenko,
Marco Elver, Dmitry Vyukov, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Ard Biesheuvel, Anup Patel, Atish Patra,
Andrey Ryabinin, Andrey Konovalov, Vincenzo Frascino,
Andrew Morton, linux-arm-kernel, linux-kernel, kasan-dev,
linux-riscv, linux-efi, kvm, kvm-riscv, linux-mm
Cc: Alexandre Ghiti
Make riscv use the contpte aware ptep_get() function from arm64.
Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com>
---
arch/arm64/include/asm/pgtable.h | 30 ++++++++++----------
arch/arm64/mm/contpte.c | 47 +++++---------------------------
arch/arm64/mm/hugetlbpage.c | 6 ++--
arch/riscv/include/asm/kfence.h | 4 +--
arch/riscv/include/asm/pgtable.h | 22 +++++++++++++++
arch/riscv/kernel/efi.c | 2 +-
arch/riscv/kvm/mmu.c | 16 +++++------
arch/riscv/mm/fault.c | 2 +-
arch/riscv/mm/kasan_init.c | 2 +-
arch/riscv/mm/pageattr.c | 4 +--
arch/riscv/mm/pgtable.c | 4 +--
include/linux/contpte.h | 12 ++++++++
mm/contpte.c | 45 ++++++++++++++++++++++++++++++
13 files changed, 122 insertions(+), 74 deletions(-)
create mode 100644 include/linux/contpte.h
diff --git a/arch/arm64/include/asm/pgtable.h b/arch/arm64/include/asm/pgtable.h
index 1758ce71fae9..a878735deb9f 100644
--- a/arch/arm64/include/asm/pgtable.h
+++ b/arch/arm64/include/asm/pgtable.h
@@ -38,6 +38,7 @@
#include <linux/mm_types.h>
#include <linux/sched.h>
#include <linux/page_table_check.h>
+#include <linux/contpte.h>
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
#define __HAVE_ARCH_FLUSH_PMD_TLB_RANGE
@@ -1379,8 +1380,7 @@ extern void ptep_modify_prot_commit(struct vm_area_struct *vma,
extern void __contpte_try_fold(struct mm_struct *mm, unsigned long addr,
pte_t *ptep, pte_t pte);
extern void __contpte_try_unfold(struct mm_struct *mm, unsigned long addr,
- pte_t *ptep, pte_t pte);
-extern pte_t contpte_ptep_get(pte_t *ptep, pte_t orig_pte);
+ pte_t *ptep, pte_t pte);
extern pte_t contpte_ptep_get_lockless(pte_t *orig_ptep);
extern void contpte_set_ptes(struct mm_struct *mm, unsigned long addr,
pte_t *ptep, pte_t pte, unsigned int nr);
@@ -1456,16 +1456,8 @@ static inline unsigned int pte_batch_hint(pte_t *ptep, pte_t pte)
* setting it in the pgtable.
*/
+extern pte_t ptep_get(pte_t *ptep);
#define ptep_get ptep_get
-static inline pte_t ptep_get(pte_t *ptep)
-{
- pte_t pte = __ptep_get(ptep);
-
- if (likely(!pte_valid_cont(pte)))
- return pte;
-
- return contpte_ptep_get(ptep, pte);
-}
#define ptep_get_lockless ptep_get_lockless
static inline pte_t ptep_get_lockless(pte_t *ptep)
@@ -1659,9 +1651,10 @@ static inline int arch_contpte_get_num_contig(struct mm_struct *mm,
* find out the number of contiguous ptes.
*/
if (size == 0)
- return find_num_contig(mm, addr, ptep, pgsize);
+ return mm ? find_num_contig(mm, addr, ptep, pgsize) : CONT_PTES;
- *pgsize = size;
+ if (pgsize)
+ *pgsize = size;
switch (size) {
#ifndef __PAGETABLE_PMD_FOLDED
@@ -1674,11 +1667,13 @@ static inline int arch_contpte_get_num_contig(struct mm_struct *mm,
contig_ptes = 1;
break;
case CONT_PMD_SIZE:
- *pgsize = PMD_SIZE;
+ if (pgsize)
+ *pgsize = PMD_SIZE;
contig_ptes = CONT_PMDS;
break;
case CONT_PTE_SIZE:
- *pgsize = PAGE_SIZE;
+ if (pgsize)
+ *pgsize = PAGE_SIZE;
contig_ptes = CONT_PTES;
break;
}
@@ -1686,6 +1681,11 @@ static inline int arch_contpte_get_num_contig(struct mm_struct *mm,
return contig_ptes;
}
+static inline pte_t *arch_contpte_align_down(pte_t *ptep)
+{
+ return PTR_ALIGN_DOWN(ptep, sizeof(*ptep) * CONT_PTES);
+}
+
#endif /* !__ASSEMBLY__ */
#endif /* __ASM_PGTABLE_H */
diff --git a/arch/arm64/mm/contpte.c b/arch/arm64/mm/contpte.c
index 1b64b4c3f8bf..d5512ebb26e9 100644
--- a/arch/arm64/mm/contpte.c
+++ b/arch/arm64/mm/contpte.c
@@ -21,11 +21,6 @@ static inline bool mm_is_user(struct mm_struct *mm)
return mm != &init_mm;
}
-static inline pte_t *contpte_align_down(pte_t *ptep)
-{
- return PTR_ALIGN_DOWN(ptep, sizeof(*ptep) * CONT_PTES);
-}
-
static void contpte_try_unfold_partial(struct mm_struct *mm, unsigned long addr,
pte_t *ptep, unsigned int nr)
{
@@ -34,10 +29,10 @@ static void contpte_try_unfold_partial(struct mm_struct *mm, unsigned long addr,
* of the range.
*/
- if (ptep != contpte_align_down(ptep) || nr < CONT_PTES)
+ if (ptep != arch_contpte_align_down(ptep) || nr < CONT_PTES)
contpte_try_unfold(mm, addr, ptep, __ptep_get(ptep));
- if (ptep + nr != contpte_align_down(ptep + nr)) {
+ if (ptep + nr != arch_contpte_align_down(ptep + nr)) {
unsigned long last_addr = addr + PAGE_SIZE * (nr - 1);
pte_t *last_ptep = ptep + nr - 1;
@@ -54,7 +49,7 @@ static void contpte_convert(struct mm_struct *mm, unsigned long addr,
pte_t *start_ptep;
int i;
- start_ptep = ptep = contpte_align_down(ptep);
+ start_ptep = ptep = arch_contpte_align_down(ptep);
start_addr = addr = ALIGN_DOWN(addr, CONT_PTE_SIZE);
pte = pfn_pte(ALIGN_DOWN(pte_pfn(pte), CONT_PTES), pte_pgprot(pte));
@@ -122,7 +117,7 @@ void __contpte_try_fold(struct mm_struct *mm, unsigned long addr,
prot = pte_pgprot(pte_mkold(pte_mkclean(pte)));
expected_pte = pfn_pte(pfn, prot);
orig_ptep = ptep;
- ptep = contpte_align_down(ptep);
+ ptep = arch_contpte_align_down(ptep);
for (i = 0; i < CONT_PTES; i++) {
subpte = pte_mkold(pte_mkclean(__ptep_get(ptep)));
@@ -152,34 +147,6 @@ void __contpte_try_unfold(struct mm_struct *mm, unsigned long addr,
}
EXPORT_SYMBOL_GPL(__contpte_try_unfold);
-pte_t contpte_ptep_get(pte_t *ptep, pte_t orig_pte)
-{
- /*
- * Gather access/dirty bits, which may be populated in any of the ptes
- * of the contig range. We are guaranteed to be holding the PTL, so any
- * contiguous range cannot be unfolded or otherwise modified under our
- * feet.
- */
-
- pte_t pte;
- int i;
-
- ptep = contpte_align_down(ptep);
-
- for (i = 0; i < CONT_PTES; i++, ptep++) {
- pte = __ptep_get(ptep);
-
- if (pte_dirty(pte))
- orig_pte = pte_mkdirty(orig_pte);
-
- if (pte_young(pte))
- orig_pte = pte_mkyoung(orig_pte);
- }
-
- return orig_pte;
-}
-EXPORT_SYMBOL_GPL(contpte_ptep_get);
-
pte_t contpte_ptep_get_lockless(pte_t *orig_ptep)
{
/*
@@ -214,7 +181,7 @@ pte_t contpte_ptep_get_lockless(pte_t *orig_ptep)
return orig_pte;
orig_prot = pte_pgprot(pte_mkold(pte_mkclean(orig_pte)));
- ptep = contpte_align_down(orig_ptep);
+ ptep = arch_contpte_align_down(orig_ptep);
pfn = pte_pfn(orig_pte) - (orig_ptep - ptep);
for (i = 0; i < CONT_PTES; i++, ptep++, pfn++) {
@@ -312,7 +279,7 @@ int contpte_ptep_test_and_clear_young(struct vm_area_struct *vma,
int young = 0;
int i;
- ptep = contpte_align_down(ptep);
+ ptep = arch_contpte_align_down(ptep);
addr = ALIGN_DOWN(addr, CONT_PTE_SIZE);
for (i = 0; i < CONT_PTES; i++, ptep++, addr += PAGE_SIZE)
@@ -389,7 +356,7 @@ int contpte_ptep_set_access_flags(struct vm_area_struct *vma,
* faults. Avoid per-page tlb flush in __ptep_set_access_flags()
* and instead flush the whole range at the end.
*/
- ptep = contpte_align_down(ptep);
+ ptep = arch_contpte_align_down(ptep);
start_addr = addr = ALIGN_DOWN(addr, CONT_PTE_SIZE);
for (i = 0; i < CONT_PTES; i++, ptep++, addr += PAGE_SIZE)
diff --git a/arch/arm64/mm/hugetlbpage.c b/arch/arm64/mm/hugetlbpage.c
index 5869f20ca28e..083e80ac5790 100644
--- a/arch/arm64/mm/hugetlbpage.c
+++ b/arch/arm64/mm/hugetlbpage.c
@@ -101,12 +101,14 @@ int find_num_contig(struct mm_struct *mm, unsigned long addr,
pud_t *pudp;
pmd_t *pmdp;
- *pgsize = PAGE_SIZE;
+ if (pgsize)
+ *pgsize = PAGE_SIZE;
p4dp = p4d_offset(pgdp, addr);
pudp = pud_offset(p4dp, addr);
pmdp = pmd_offset(pudp, addr);
if ((pte_t *)pmdp == ptep) {
- *pgsize = PMD_SIZE;
+ if (pgsize)
+ *pgsize = PMD_SIZE;
return CONT_PMDS;
}
return CONT_PTES;
diff --git a/arch/riscv/include/asm/kfence.h b/arch/riscv/include/asm/kfence.h
index 7388edd88986..f303fef8591c 100644
--- a/arch/riscv/include/asm/kfence.h
+++ b/arch/riscv/include/asm/kfence.h
@@ -18,9 +18,9 @@ static inline bool kfence_protect_page(unsigned long addr, bool protect)
pte_t *pte = virt_to_kpte(addr);
if (protect)
- set_pte(pte, __pte(pte_val(ptep_get(pte)) & ~_PAGE_PRESENT));
+ set_pte(pte, __pte(pte_val(__ptep_get(pte)) & ~_PAGE_PRESENT));
else
- set_pte(pte, __pte(pte_val(ptep_get(pte)) | _PAGE_PRESENT));
+ set_pte(pte, __pte(pte_val(__ptep_get(pte)) | _PAGE_PRESENT));
flush_tlb_kernel_range(addr, addr + PAGE_SIZE);
diff --git a/arch/riscv/include/asm/pgtable.h b/arch/riscv/include/asm/pgtable.h
index 9e397935536e..8d05179f6bbe 100644
--- a/arch/riscv/include/asm/pgtable.h
+++ b/arch/riscv/include/asm/pgtable.h
@@ -299,6 +299,7 @@ static inline unsigned long pte_napot(pte_t pte)
#define pte_cont pte_napot
#define pte_valid_napot(pte) (pte_present(pte) && pte_napot(pte))
+#define pte_valid_cont pte_valid_napot
static inline pte_t pte_mknapot(pte_t pte, unsigned int order)
{
@@ -571,6 +572,17 @@ static inline int arch_contpte_get_num_contig(struct mm_struct *mm,
return size >> hugepage_shift;
}
+
+static inline pte_t *arch_contpte_align_down(pte_t *ptep)
+{
+ pte_t __pte = READ_ONCE(*ptep);
+ int ncontig;
+
+ ncontig = napot_pte_num(napot_cont_order(__pte));
+
+ return PTR_ALIGN_DOWN(ptep, sizeof(*ptep) * ncontig);
+}
+
#endif
static inline pte_t __ptep_get(pte_t *ptep)
@@ -696,8 +708,18 @@ static inline int ptep_clear_flush_young(struct vm_area_struct *vma,
return ptep_test_and_clear_young(vma, address, ptep);
}
+#ifdef CONFIG_THP_CONTPTE
+
+extern pte_t ptep_get(pte_t *ptep);
+#define ptep_get ptep_get
+
+#else /* CONFIG_THP_CONTPTE */
+
#define ptep_get __ptep_get
#define set_ptes __set_ptes
+
+#endif /* CONFIG_THP_CONTPTE */
+
#define __HAVE_ARCH_PTEP_GET_AND_CLEAR
#define ptep_get_and_clear __ptep_get_and_clear
#define pte_clear __pte_clear
diff --git a/arch/riscv/kernel/efi.c b/arch/riscv/kernel/efi.c
index b64bf1624a05..3d2a635c69ac 100644
--- a/arch/riscv/kernel/efi.c
+++ b/arch/riscv/kernel/efi.c
@@ -60,7 +60,7 @@ int __init efi_create_mapping(struct mm_struct *mm, efi_memory_desc_t *md)
static int __init set_permissions(pte_t *ptep, unsigned long addr, void *data)
{
efi_memory_desc_t *md = data;
- pte_t pte = ptep_get(ptep);
+ pte_t pte = __ptep_get(ptep);
unsigned long val;
if (md->attribute & EFI_MEMORY_RO) {
diff --git a/arch/riscv/kvm/mmu.c b/arch/riscv/kvm/mmu.c
index a9e2fd7245e1..70c6cb3864d6 100644
--- a/arch/riscv/kvm/mmu.c
+++ b/arch/riscv/kvm/mmu.c
@@ -103,7 +103,7 @@ static bool gstage_get_leaf_entry(struct kvm *kvm, gpa_t addr,
*ptep_level = current_level;
ptep = (pte_t *)kvm->arch.pgd;
ptep = &ptep[gstage_pte_index(addr, current_level)];
- while (ptep && pte_val(ptep_get(ptep))) {
+ while (ptep && pte_val(__ptep_get(ptep))) {
if (gstage_pte_leaf(ptep)) {
*ptep_level = current_level;
*ptepp = ptep;
@@ -113,7 +113,7 @@ static bool gstage_get_leaf_entry(struct kvm *kvm, gpa_t addr,
if (current_level) {
current_level--;
*ptep_level = current_level;
- ptep = (pte_t *)gstage_pte_page_vaddr(ptep_get(ptep));
+ ptep = (pte_t *)gstage_pte_page_vaddr(__ptep_get(ptep));
ptep = &ptep[gstage_pte_index(addr, current_level)];
} else {
ptep = NULL;
@@ -149,7 +149,7 @@ static int gstage_set_pte(struct kvm *kvm, u32 level,
if (gstage_pte_leaf(ptep))
return -EEXIST;
- if (!pte_val(ptep_get(ptep))) {
+ if (!pte_val(__ptep_get(ptep))) {
if (!pcache)
return -ENOMEM;
next_ptep = kvm_mmu_memory_cache_alloc(pcache);
@@ -160,7 +160,7 @@ static int gstage_set_pte(struct kvm *kvm, u32 level,
} else {
if (gstage_pte_leaf(ptep))
return -EEXIST;
- next_ptep = (pte_t *)gstage_pte_page_vaddr(ptep_get(ptep));
+ next_ptep = (pte_t *)gstage_pte_page_vaddr(__ptep_get(ptep));
}
current_level--;
@@ -239,11 +239,11 @@ static void gstage_op_pte(struct kvm *kvm, gpa_t addr,
BUG_ON(addr & (page_size - 1));
- if (!pte_val(ptep_get(ptep)))
+ if (!pte_val(__ptep_get(ptep)))
return;
if (ptep_level && !gstage_pte_leaf(ptep)) {
- next_ptep = (pte_t *)gstage_pte_page_vaddr(ptep_get(ptep));
+ next_ptep = (pte_t *)gstage_pte_page_vaddr(__ptep_get(ptep));
next_ptep_level = ptep_level - 1;
ret = gstage_level_to_page_size(next_ptep_level,
&next_page_size);
@@ -261,7 +261,7 @@ static void gstage_op_pte(struct kvm *kvm, gpa_t addr,
if (op == GSTAGE_OP_CLEAR)
set_pte(ptep, __pte(0));
else if (op == GSTAGE_OP_WP)
- set_pte(ptep, __pte(pte_val(ptep_get(ptep)) & ~_PAGE_WRITE));
+ set_pte(ptep, __pte(pte_val(__ptep_get(ptep)) & ~_PAGE_WRITE));
gstage_remote_tlb_flush(kvm, ptep_level, addr);
}
}
@@ -603,7 +603,7 @@ bool kvm_test_age_gfn(struct kvm *kvm, struct kvm_gfn_range *range)
&ptep, &ptep_level))
return false;
- return pte_young(ptep_get(ptep));
+ return pte_young(__ptep_get(ptep));
}
int kvm_riscv_gstage_map(struct kvm_vcpu *vcpu,
diff --git a/arch/riscv/mm/fault.c b/arch/riscv/mm/fault.c
index 3ba1d4dde5dd..0e08afc1fc6a 100644
--- a/arch/riscv/mm/fault.c
+++ b/arch/riscv/mm/fault.c
@@ -175,7 +175,7 @@ static inline void vmalloc_fault(struct pt_regs *regs, int code, unsigned long a
* silently loop forever.
*/
pte_k = pte_offset_kernel(pmd_k, addr);
- if (!pte_present(ptep_get(pte_k))) {
+ if (!pte_present(__ptep_get(pte_k))) {
no_context(regs, addr);
return;
}
diff --git a/arch/riscv/mm/kasan_init.c b/arch/riscv/mm/kasan_init.c
index c301c8d291d2..381d61f42ab8 100644
--- a/arch/riscv/mm/kasan_init.c
+++ b/arch/riscv/mm/kasan_init.c
@@ -39,7 +39,7 @@ static void __init kasan_populate_pte(pmd_t *pmd, unsigned long vaddr, unsigned
ptep = pte_offset_kernel(pmd, vaddr);
do {
- if (pte_none(ptep_get(ptep))) {
+ if (pte_none(__ptep_get(ptep))) {
phys_addr = memblock_phys_alloc(PAGE_SIZE, PAGE_SIZE);
set_pte(ptep, pfn_pte(PFN_DOWN(phys_addr), PAGE_KERNEL));
memset(__va(phys_addr), KASAN_SHADOW_INIT, PAGE_SIZE);
diff --git a/arch/riscv/mm/pageattr.c b/arch/riscv/mm/pageattr.c
index 410056a50aa9..98c9dc4b983c 100644
--- a/arch/riscv/mm/pageattr.c
+++ b/arch/riscv/mm/pageattr.c
@@ -68,7 +68,7 @@ static int pageattr_pmd_entry(pmd_t *pmd, unsigned long addr,
static int pageattr_pte_entry(pte_t *pte, unsigned long addr,
unsigned long next, struct mm_walk *walk)
{
- pte_t val = ptep_get(pte);
+ pte_t val = __ptep_get(pte);
val = __pte(set_pageattr_masks(pte_val(val), walk));
set_pte(pte, val);
@@ -435,5 +435,5 @@ bool kernel_page_present(struct page *page)
return true;
pte = pte_offset_kernel(pmd, addr);
- return pte_present(ptep_get(pte));
+ return pte_present(__ptep_get(pte));
}
diff --git a/arch/riscv/mm/pgtable.c b/arch/riscv/mm/pgtable.c
index e86df7ef193c..5756bde9eb42 100644
--- a/arch/riscv/mm/pgtable.c
+++ b/arch/riscv/mm/pgtable.c
@@ -9,7 +9,7 @@ int __ptep_set_access_flags(struct vm_area_struct *vma,
unsigned long address, pte_t *ptep,
pte_t entry, int dirty)
{
- if (!pte_same(ptep_get(ptep), entry))
+ if (!pte_same(__ptep_get(ptep), entry))
__set_pte_at(vma->vm_mm, ptep, entry);
/*
* update_mmu_cache will unconditionally execute, handling both
@@ -22,7 +22,7 @@ int ptep_test_and_clear_young(struct vm_area_struct *vma,
unsigned long address,
pte_t *ptep)
{
- if (!pte_young(ptep_get(ptep)))
+ if (!pte_young(__ptep_get(ptep)))
return 0;
return test_and_clear_bit(_PAGE_ACCESSED_OFFSET, &pte_val(*ptep));
}
diff --git a/include/linux/contpte.h b/include/linux/contpte.h
new file mode 100644
index 000000000000..46acac7222ca
--- /dev/null
+++ b/include/linux/contpte.h
@@ -0,0 +1,12 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _LINUX_CONTPTE_H
+#define _LINUX_CONTPTE_H
+
+/*
+ * The contpte APIs are used to transparently manage the contiguous bit in ptes
+ * where it is possible and makes sense to do so. The PTE_CONT bit is considered
+ * a private implementation detail of the public ptep API (see below).
+ */
+pte_t contpte_ptep_get(pte_t *ptep, pte_t orig_pte);
+
+#endif /* _LINUX_CONTPTE_H */
diff --git a/mm/contpte.c b/mm/contpte.c
index 15791f6d9c41..d365356bbf92 100644
--- a/mm/contpte.c
+++ b/mm/contpte.c
@@ -6,6 +6,7 @@
#include <linux/mm.h>
#include <linux/pgtable.h>
#include <linux/hugetlb.h>
+#include <linux/contpte.h>
/*
* Any arch that wants to use that needs to define:
@@ -17,6 +18,8 @@
* - __ptep_set_wrprotect()
* - pte_cont()
* - arch_contpte_get_num_contig()
+ * - pte_valid_cont()
+ * - arch_contpte_align_down()
*/
/*
@@ -28,6 +31,7 @@
* - huge_ptep_set_access_flags()
* - huge_ptep_set_wrprotect()
* - huge_ptep_clear_flush()
+ * - ptep_get()
*/
pte_t huge_ptep_get(pte_t *ptep)
@@ -270,3 +274,44 @@ pte_t huge_ptep_clear_flush(struct vm_area_struct *vma,
ncontig = arch_contpte_get_num_contig(mm, addr, ptep, 0, &pgsize);
return get_clear_contig_flush(mm, addr, ptep, pgsize, ncontig);
}
+
+#ifdef CONFIG_THP_CONTPTE
+pte_t contpte_ptep_get(pte_t *ptep, pte_t orig_pte)
+{
+ /*
+ * Gather access/dirty bits, which may be populated in any of the ptes
+ * of the contig range. We are guaranteed to be holding the PTL, so any
+ * contiguous range cannot be unfolded or otherwise modified under our
+ * feet.
+ */
+
+ pte_t pte;
+ int i, ncontig;
+
+ ptep = arch_contpte_align_down(ptep);
+ ncontig = arch_contpte_get_num_contig(NULL, 0, ptep, 0, NULL);
+
+ for (i = 0; i < ncontig; i++, ptep++) {
+ pte = __ptep_get(ptep);
+
+ if (pte_dirty(pte))
+ orig_pte = pte_mkdirty(orig_pte);
+
+ if (pte_young(pte))
+ orig_pte = pte_mkyoung(orig_pte);
+ }
+
+ return orig_pte;
+}
+EXPORT_SYMBOL_GPL(contpte_ptep_get);
+
+__always_inline pte_t ptep_get(pte_t *ptep)
+{
+ pte_t pte = __ptep_get(ptep);
+
+ if (likely(!pte_valid_cont(pte)))
+ return pte;
+
+ return contpte_ptep_get(ptep, pte);
+}
+#endif /* CONTPTE_THP_CONTPTE */
--
2.39.2
^ permalink raw reply related [flat|nested] 16+ messages in thread
* [PATCH 03/12] mm, riscv, arm64: Use common set_ptes() function
2024-05-08 19:19 [PATCH 00/12] Make riscv use THP contpte support for arm64 Alexandre Ghiti
2024-05-08 19:19 ` [PATCH 01/12] mm, arm64: Rename ARM64_CONTPTE to THP_CONTPTE Alexandre Ghiti
2024-05-08 19:19 ` [PATCH 02/12] mm, riscv, arm64: Use common ptep_get() function Alexandre Ghiti
@ 2024-05-08 19:19 ` Alexandre Ghiti
2024-05-08 19:19 ` [PATCH 04/12] mm, riscv, arm64: Use common ptep_get_lockless() function Alexandre Ghiti
` (8 subsequent siblings)
11 siblings, 0 replies; 16+ messages in thread
From: Alexandre Ghiti @ 2024-05-08 19:19 UTC (permalink / raw
To: Ryan Roberts, Catalin Marinas, Will Deacon, Alexander Potapenko,
Marco Elver, Dmitry Vyukov, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Ard Biesheuvel, Anup Patel, Atish Patra,
Andrey Ryabinin, Andrey Konovalov, Vincenzo Frascino,
Andrew Morton, linux-arm-kernel, linux-kernel, kasan-dev,
linux-riscv, linux-efi, kvm, kvm-riscv, linux-mm
Cc: Alexandre Ghiti
Make riscv use the contpte aware set_ptes() function from arm64.
Note that riscv can support multiple contpte sizes so the arm64 code was
modified to take into account this possibility.
Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com>
---
arch/arm64/include/asm/pgtable.h | 88 ++++------
arch/arm64/mm/contpte.c | 162 ------------------
arch/arm64/mm/mmu.c | 2 +-
arch/riscv/include/asm/pgtable.h | 76 +++++++++
include/linux/contpte.h | 10 ++
mm/contpte.c | 277 ++++++++++++++++++++++++++++++-
6 files changed, 398 insertions(+), 217 deletions(-)
diff --git a/arch/arm64/include/asm/pgtable.h b/arch/arm64/include/asm/pgtable.h
index a878735deb9f..e85b3a052a02 100644
--- a/arch/arm64/include/asm/pgtable.h
+++ b/arch/arm64/include/asm/pgtable.h
@@ -118,10 +118,17 @@ static inline pteval_t __phys_to_pte_val(phys_addr_t phys)
#define pte_tagged(pte) ((pte_val(pte) & PTE_ATTRINDX_MASK) == \
PTE_ATTRINDX(MT_NORMAL_TAGGED))
-#define pte_cont_addr_end(addr, end) \
-({ unsigned long __boundary = ((addr) + CONT_PTE_SIZE) & CONT_PTE_MASK; \
- (__boundary - 1 < (end) - 1) ? __boundary : (end); \
-})
+static inline unsigned long arch_contpte_addr_end(unsigned long addr,
+ unsigned long end,
+ int *ncontig)
+{
+ unsigned long __boundary = (addr + CONT_PTE_SIZE) & CONT_PTE_MASK;
+
+ if (ncontig)
+ *ncontig = CONT_PTES;
+
+ return (__boundary - 1 < end - 1) ? __boundary : end;
+}
#define pmd_cont_addr_end(addr, end) \
({ unsigned long __boundary = ((addr) + CONT_PMD_SIZE) & CONT_PMD_MASK; \
@@ -1377,13 +1384,7 @@ extern void ptep_modify_prot_commit(struct vm_area_struct *vma,
* where it is possible and makes sense to do so. The PTE_CONT bit is considered
* a private implementation detail of the public ptep API (see below).
*/
-extern void __contpte_try_fold(struct mm_struct *mm, unsigned long addr,
- pte_t *ptep, pte_t pte);
-extern void __contpte_try_unfold(struct mm_struct *mm, unsigned long addr,
- pte_t *ptep, pte_t pte);
extern pte_t contpte_ptep_get_lockless(pte_t *orig_ptep);
-extern void contpte_set_ptes(struct mm_struct *mm, unsigned long addr,
- pte_t *ptep, pte_t pte, unsigned int nr);
extern void contpte_clear_full_ptes(struct mm_struct *mm, unsigned long addr,
pte_t *ptep, unsigned int nr, int full);
extern pte_t contpte_get_and_clear_full_ptes(struct mm_struct *mm,
@@ -1399,36 +1400,6 @@ extern int contpte_ptep_set_access_flags(struct vm_area_struct *vma,
unsigned long addr, pte_t *ptep,
pte_t entry, int dirty);
-static __always_inline void contpte_try_fold(struct mm_struct *mm,
- unsigned long addr, pte_t *ptep, pte_t pte)
-{
- /*
- * Only bother trying if both the virtual and physical addresses are
- * aligned and correspond to the last entry in a contig range. The core
- * code mostly modifies ranges from low to high, so this is the likely
- * the last modification in the contig range, so a good time to fold.
- * We can't fold special mappings, because there is no associated folio.
- */
-
- const unsigned long contmask = CONT_PTES - 1;
- bool valign = ((addr >> PAGE_SHIFT) & contmask) == contmask;
-
- if (unlikely(valign)) {
- bool palign = (pte_pfn(pte) & contmask) == contmask;
-
- if (unlikely(palign &&
- pte_valid(pte) && !pte_cont(pte) && !pte_special(pte)))
- __contpte_try_fold(mm, addr, ptep, pte);
- }
-}
-
-static __always_inline void contpte_try_unfold(struct mm_struct *mm,
- unsigned long addr, pte_t *ptep, pte_t pte)
-{
- if (unlikely(pte_valid_cont(pte)))
- __contpte_try_unfold(mm, addr, ptep, pte);
-}
-
#define pte_batch_hint pte_batch_hint
static inline unsigned int pte_batch_hint(pte_t *ptep, pte_t pte)
{
@@ -1485,20 +1456,9 @@ static inline void set_pte(pte_t *ptep, pte_t pte)
__set_pte(ptep, pte_mknoncont(pte));
}
+extern void set_ptes(struct mm_struct *mm, unsigned long addr,
+ pte_t *ptep, pte_t pte, unsigned int nr);
#define set_ptes set_ptes
-static __always_inline void set_ptes(struct mm_struct *mm, unsigned long addr,
- pte_t *ptep, pte_t pte, unsigned int nr)
-{
- pte = pte_mknoncont(pte);
-
- if (likely(nr == 1)) {
- contpte_try_unfold(mm, addr, ptep, __ptep_get(ptep));
- __set_ptes(mm, addr, ptep, pte, 1);
- contpte_try_fold(mm, addr, ptep, pte);
- } else {
- contpte_set_ptes(mm, addr, ptep, pte, nr);
- }
-}
static inline void pte_clear(struct mm_struct *mm,
unsigned long addr, pte_t *ptep)
@@ -1686,6 +1646,28 @@ static inline pte_t *arch_contpte_align_down(pte_t *ptep)
return PTR_ALIGN_DOWN(ptep, sizeof(*ptep) * CONT_PTES);
}
+static inline void arch_contpte_flush_tlb_range(struct vm_area_struct *vma,
+ unsigned long start,
+ unsigned long end,
+ unsigned long stride)
+{
+ __flush_tlb_range(vma, start, end, stride, true, 3);
+}
+
+static inline int arch_contpte_get_first_ncontig(size_t *pgsize)
+{
+ if (pgsize)
+ *pgsize = PAGE_SIZE;
+
+ return CONT_PTES;
+}
+
+/* Must return 0 when ncontig does not have any next. */
+static inline int arch_contpte_get_next_ncontig(int ncontig, size_t *pgsize)
+{
+ return 0;
+}
+
#endif /* !__ASSEMBLY__ */
#endif /* __ASM_PGTABLE_H */
diff --git a/arch/arm64/mm/contpte.c b/arch/arm64/mm/contpte.c
index d5512ebb26e9..e225e458856e 100644
--- a/arch/arm64/mm/contpte.c
+++ b/arch/arm64/mm/contpte.c
@@ -8,19 +8,6 @@
#include <linux/export.h>
#include <asm/tlbflush.h>
-static inline bool mm_is_user(struct mm_struct *mm)
-{
- /*
- * Don't attempt to apply the contig bit to kernel mappings, because
- * dynamically adding/removing the contig bit can cause page faults.
- * These racing faults are ok for user space, since they get serialized
- * on the PTL. But kernel mappings can't tolerate faults.
- */
- if (unlikely(mm_is_efi(mm)))
- return false;
- return mm != &init_mm;
-}
-
static void contpte_try_unfold_partial(struct mm_struct *mm, unsigned long addr,
pte_t *ptep, unsigned int nr)
{
@@ -41,112 +28,6 @@ static void contpte_try_unfold_partial(struct mm_struct *mm, unsigned long addr,
}
}
-static void contpte_convert(struct mm_struct *mm, unsigned long addr,
- pte_t *ptep, pte_t pte)
-{
- struct vm_area_struct vma = TLB_FLUSH_VMA(mm, 0);
- unsigned long start_addr;
- pte_t *start_ptep;
- int i;
-
- start_ptep = ptep = arch_contpte_align_down(ptep);
- start_addr = addr = ALIGN_DOWN(addr, CONT_PTE_SIZE);
- pte = pfn_pte(ALIGN_DOWN(pte_pfn(pte), CONT_PTES), pte_pgprot(pte));
-
- for (i = 0; i < CONT_PTES; i++, ptep++, addr += PAGE_SIZE) {
- pte_t ptent = __ptep_get_and_clear(mm, addr, ptep);
-
- if (pte_dirty(ptent))
- pte = pte_mkdirty(pte);
-
- if (pte_young(ptent))
- pte = pte_mkyoung(pte);
- }
-
- __flush_tlb_range(&vma, start_addr, addr, PAGE_SIZE, true, 3);
-
- __set_ptes(mm, start_addr, start_ptep, pte, CONT_PTES);
-}
-
-void __contpte_try_fold(struct mm_struct *mm, unsigned long addr,
- pte_t *ptep, pte_t pte)
-{
- /*
- * We have already checked that the virtual and pysical addresses are
- * correctly aligned for a contpte mapping in contpte_try_fold() so the
- * remaining checks are to ensure that the contpte range is fully
- * covered by a single folio, and ensure that all the ptes are valid
- * with contiguous PFNs and matching prots. We ignore the state of the
- * access and dirty bits for the purpose of deciding if its a contiguous
- * range; the folding process will generate a single contpte entry which
- * has a single access and dirty bit. Those 2 bits are the logical OR of
- * their respective bits in the constituent pte entries. In order to
- * ensure the contpte range is covered by a single folio, we must
- * recover the folio from the pfn, but special mappings don't have a
- * folio backing them. Fortunately contpte_try_fold() already checked
- * that the pte is not special - we never try to fold special mappings.
- * Note we can't use vm_normal_page() for this since we don't have the
- * vma.
- */
-
- unsigned long folio_start, folio_end;
- unsigned long cont_start, cont_end;
- pte_t expected_pte, subpte;
- struct folio *folio;
- struct page *page;
- unsigned long pfn;
- pte_t *orig_ptep;
- pgprot_t prot;
-
- int i;
-
- if (!mm_is_user(mm))
- return;
-
- page = pte_page(pte);
- folio = page_folio(page);
- folio_start = addr - (page - &folio->page) * PAGE_SIZE;
- folio_end = folio_start + folio_nr_pages(folio) * PAGE_SIZE;
- cont_start = ALIGN_DOWN(addr, CONT_PTE_SIZE);
- cont_end = cont_start + CONT_PTE_SIZE;
-
- if (folio_start > cont_start || folio_end < cont_end)
- return;
-
- pfn = ALIGN_DOWN(pte_pfn(pte), CONT_PTES);
- prot = pte_pgprot(pte_mkold(pte_mkclean(pte)));
- expected_pte = pfn_pte(pfn, prot);
- orig_ptep = ptep;
- ptep = arch_contpte_align_down(ptep);
-
- for (i = 0; i < CONT_PTES; i++) {
- subpte = pte_mkold(pte_mkclean(__ptep_get(ptep)));
- if (!pte_same(subpte, expected_pte))
- return;
- expected_pte = pte_advance_pfn(expected_pte, 1);
- ptep++;
- }
-
- pte = pte_mkcont(pte);
- contpte_convert(mm, addr, orig_ptep, pte);
-}
-EXPORT_SYMBOL_GPL(__contpte_try_fold);
-
-void __contpte_try_unfold(struct mm_struct *mm, unsigned long addr,
- pte_t *ptep, pte_t pte)
-{
- /*
- * We have already checked that the ptes are contiguous in
- * contpte_try_unfold(), so just check that the mm is user space.
- */
- if (!mm_is_user(mm))
- return;
-
- pte = pte_mknoncont(pte);
- contpte_convert(mm, addr, ptep, pte);
-}
-EXPORT_SYMBOL_GPL(__contpte_try_unfold);
-
pte_t contpte_ptep_get_lockless(pte_t *orig_ptep)
{
/*
@@ -204,49 +85,6 @@ pte_t contpte_ptep_get_lockless(pte_t *orig_ptep)
}
EXPORT_SYMBOL_GPL(contpte_ptep_get_lockless);
-void contpte_set_ptes(struct mm_struct *mm, unsigned long addr,
- pte_t *ptep, pte_t pte, unsigned int nr)
-{
- unsigned long next;
- unsigned long end;
- unsigned long pfn;
- pgprot_t prot;
-
- /*
- * The set_ptes() spec guarantees that when nr > 1, the initial state of
- * all ptes is not-present. Therefore we never need to unfold or
- * otherwise invalidate a range before we set the new ptes.
- * contpte_set_ptes() should never be called for nr < 2.
- */
- VM_WARN_ON(nr == 1);
-
- if (!mm_is_user(mm))
- return __set_ptes(mm, addr, ptep, pte, nr);
-
- end = addr + (nr << PAGE_SHIFT);
- pfn = pte_pfn(pte);
- prot = pte_pgprot(pte);
-
- do {
- next = pte_cont_addr_end(addr, end);
- nr = (next - addr) >> PAGE_SHIFT;
- pte = pfn_pte(pfn, prot);
-
- if (((addr | next | (pfn << PAGE_SHIFT)) & ~CONT_PTE_MASK) == 0)
- pte = pte_mkcont(pte);
- else
- pte = pte_mknoncont(pte);
-
- __set_ptes(mm, addr, ptep, pte, nr);
-
- addr = next;
- ptep += nr;
- pfn += nr;
-
- } while (addr != end);
-}
-EXPORT_SYMBOL_GPL(contpte_set_ptes);
-
void contpte_clear_full_ptes(struct mm_struct *mm, unsigned long addr,
pte_t *ptep, unsigned int nr, int full)
{
diff --git a/arch/arm64/mm/mmu.c b/arch/arm64/mm/mmu.c
index 495b732d5af3..b7ad732660aa 100644
--- a/arch/arm64/mm/mmu.c
+++ b/arch/arm64/mm/mmu.c
@@ -222,7 +222,7 @@ static void alloc_init_cont_pte(pmd_t *pmdp, unsigned long addr,
do {
pgprot_t __prot = prot;
- next = pte_cont_addr_end(addr, end);
+ next = arch_contpte_addr_end(addr, end, NULL);
/* use a contiguous mapping if the range is suitably aligned */
if ((((addr | next | phys) & ~CONT_PTE_MASK) == 0) &&
diff --git a/arch/riscv/include/asm/pgtable.h b/arch/riscv/include/asm/pgtable.h
index 8d05179f6bbe..ebfe6b16529e 100644
--- a/arch/riscv/include/asm/pgtable.h
+++ b/arch/riscv/include/asm/pgtable.h
@@ -301,6 +301,20 @@ static inline unsigned long pte_napot(pte_t pte)
#define pte_valid_napot(pte) (pte_present(pte) && pte_napot(pte))
#define pte_valid_cont pte_valid_napot
+/*
+ * contpte is what we expose to the core mm code, this is not exactly a napot
+ * mapping since the size is not encoded in the pfn yet.
+ */
+static inline pte_t pte_mknoncont(pte_t pte)
+{
+ return __pte(pte_val(pte) & ~_PAGE_NAPOT);
+}
+
+static inline pte_t pte_mkcont(pte_t pte)
+{
+ return __pte(pte_val(pte) | _PAGE_NAPOT);
+}
+
static inline pte_t pte_mknapot(pte_t pte, unsigned int order)
{
int pos = order - 1 + _PAGE_PFN_SHIFT;
@@ -329,6 +343,11 @@ static inline unsigned long pte_napot(pte_t pte)
#endif /* CONFIG_RISCV_ISA_SVNAPOT */
+static inline pgprot_t pte_pgprot(pte_t pte)
+{
+ return __pgprot(pte_val(pte) & ~_PAGE_PFN_MASK);
+}
+
/* Yields the page frame number (PFN) of a page table entry */
static inline unsigned long pte_pfn(pte_t pte)
{
@@ -354,6 +373,11 @@ static inline int pte_present(pte_t pte)
return (pte_val(pte) & (_PAGE_PRESENT | _PAGE_PROT_NONE));
}
+static inline int pte_valid(pte_t pte)
+{
+ return (pte_val(pte) & _PAGE_PRESENT);
+}
+
static inline int pte_none(pte_t pte)
{
return (pte_val(pte) == 0);
@@ -583,6 +607,55 @@ static inline pte_t *arch_contpte_align_down(pte_t *ptep)
return PTR_ALIGN_DOWN(ptep, sizeof(*ptep) * ncontig);
}
+static inline void arch_contpte_flush_tlb_range(struct vm_area_struct *vma,
+ unsigned long start,
+ unsigned long end,
+ unsigned long stride)
+{
+ flush_tlb_mm_range(vma->vm_mm, start, end, stride);
+}
+
+static inline int arch_contpte_get_first_ncontig(size_t *pgsize)
+{
+ if (pgsize)
+ *pgsize = PAGE_SIZE;
+
+ return 1 << NAPOT_CONT64KB_ORDER;
+}
+
+/* Must return 0 when ncontig does not have any next. */
+static inline int arch_contpte_get_next_ncontig(int ncontig, size_t *pgsize)
+{
+ return 0;
+}
+
+#define for_each_contpte_order_rev(ncontig, order, pgsize) \
+ for (pgsize = PAGE_SIZE, order = NAPOT_ORDER_MAX - 1, ncontig = BIT(order); \
+ ncontig >= BIT(NAPOT_CONT_ORDER_BASE); \
+ order--, ncontig = BIT(order))
+
+static inline unsigned long arch_contpte_addr_end(unsigned long addr,
+ unsigned long end,
+ int *ncontig)
+{
+ unsigned long contpte_saddr, contpte_eaddr, contpte_size;
+ size_t pgsize;
+ int contig, order;
+
+ for_each_contpte_order_rev(contig, order, pgsize) {
+ contpte_size = contig * pgsize;
+ contpte_saddr = ALIGN_DOWN(addr, contpte_size);
+ contpte_eaddr = contpte_saddr + contpte_size;
+
+ if (contpte_saddr >= addr && contpte_eaddr <= end) {
+ *ncontig = contig;
+ return contpte_eaddr;
+ }
+ }
+
+ *ncontig = 0;
+ return end;
+}
#endif
static inline pte_t __ptep_get(pte_t *ptep)
@@ -712,6 +785,9 @@ static inline int ptep_clear_flush_young(struct vm_area_struct *vma,
extern pte_t ptep_get(pte_t *ptep);
#define ptep_get ptep_get
+extern void set_ptes(struct mm_struct *mm, unsigned long addr,
+ pte_t *ptep, pte_t pteval, unsigned int nr);
+#define set_ptes set_ptes
#else /* CONFIG_THP_CONTPTE */
diff --git a/include/linux/contpte.h b/include/linux/contpte.h
index 46acac7222ca..54d10204e9af 100644
--- a/include/linux/contpte.h
+++ b/include/linux/contpte.h
@@ -8,5 +8,15 @@
* a private implementation detail of the public ptep API (see below).
*/
pte_t contpte_ptep_get(pte_t *ptep, pte_t orig_pte);
+void __contpte_try_fold(struct mm_struct *mm, unsigned long addr,
+ pte_t *ptep, pte_t pte);
+void contpte_try_fold(struct mm_struct *mm, unsigned long addr,
+ pte_t *ptep, pte_t pte);
+void __contpte_try_unfold(struct mm_struct *mm, unsigned long addr,
+ pte_t *ptep, pte_t pte);
+void contpte_try_unfold(struct mm_struct *mm, unsigned long addr,
+ pte_t *ptep, pte_t pte);
+void contpte_set_ptes(struct mm_struct *mm, unsigned long addr,
+ pte_t *ptep, pte_t pte, unsigned int nr);
#endif /* _LINUX_CONTPTE_H */
diff --git a/mm/contpte.c b/mm/contpte.c
index d365356bbf92..566745d7842f 100644
--- a/mm/contpte.c
+++ b/mm/contpte.c
@@ -7,6 +7,7 @@
#include <linux/pgtable.h>
#include <linux/hugetlb.h>
#include <linux/contpte.h>
+#include <linux/efi.h>
/*
* Any arch that wants to use that needs to define:
@@ -20,6 +21,14 @@
* - arch_contpte_get_num_contig()
* - pte_valid_cont()
* - arch_contpte_align_down()
+ * - arch_contpte_flush_tlb_range()
+ * - arch_contpte_get_first_ncontig()
+ * - arch_contpte_get_next_ncontig()
+ * - arch_contpte_addr_end()
+ * - pte_mkcont()
+ * - pte_mknoncont()
+ * - pte_valid()
+ * - pte_pgprot()
*/
/*
@@ -32,6 +41,7 @@
* - huge_ptep_set_wrprotect()
* - huge_ptep_clear_flush()
* - ptep_get()
+ * - set_ptes()
*/
pte_t huge_ptep_get(pte_t *ptep)
@@ -314,4 +324,269 @@ __always_inline pte_t ptep_get(pte_t *ptep)
return contpte_ptep_get(ptep, pte);
}
-#endif /* CONTPTE_THP_CONTPTE */
+EXPORT_SYMBOL_GPL(ptep_get);
+
+static inline bool mm_is_user(struct mm_struct *mm)
+{
+ /*
+ * Don't attempt to apply the contig bit to kernel mappings, because
+ * dynamically adding/removing the contig bit can cause page faults.
+ * These racing faults are ok for user space, since they get serialized
+ * on the PTL. But kernel mappings can't tolerate faults.
+ */
+ if (unlikely(mm_is_efi(mm)))
+ return false;
+ return mm != &init_mm;
+}
+
+static void contpte_convert(struct mm_struct *mm, unsigned long addr,
+ pte_t *ptep, pte_t pte,
+ int ncontig, size_t pgsize)
+{
+ struct vm_area_struct vma = TLB_FLUSH_VMA(mm, 0);
+ unsigned long start_addr;
+ pte_t *start_ptep;
+ int i;
+
+ start_addr = addr;
+ start_ptep = ptep;
+
+ for (i = 0; i < ncontig; i++, ptep++, addr += pgsize) {
+ pte_t ptent = __ptep_get_and_clear(mm, addr, ptep);
+
+ if (pte_dirty(ptent))
+ pte = pte_mkdirty(pte);
+
+ if (pte_young(ptent))
+ pte = pte_mkyoung(pte);
+ }
+
+ arch_contpte_flush_tlb_range(&vma, start_addr, addr, pgsize);
+
+ __set_ptes(mm, start_addr, start_ptep, pte, ncontig);
+}
+
+void __contpte_try_unfold(struct mm_struct *mm, unsigned long addr,
+ pte_t *ptep, pte_t pte)
+{
+ unsigned long start_addr;
+ pte_t *start_ptep;
+ size_t pgsize;
+ int ncontig;
+
+ /*
+ * We have already checked that the ptes are contiguous in
+ * contpte_try_unfold(), so just check that the mm is user space.
+ */
+ if (!mm_is_user(mm))
+ return;
+
+ pte = pte_mknoncont(pte);
+ start_ptep = arch_contpte_align_down(ptep);
+ ncontig = arch_contpte_get_num_contig(mm, addr, start_ptep, 0, &pgsize);
+ start_addr = ALIGN_DOWN(addr, ncontig * pgsize);
+ pte = pfn_pte(ALIGN_DOWN(pte_pfn(pte), ncontig), pte_pgprot(pte));
+
+ contpte_convert(mm, start_addr, start_ptep, pte, ncontig, pgsize);
+}
+
+__always_inline void contpte_try_unfold(struct mm_struct *mm,
+ unsigned long addr, pte_t *ptep,
+ pte_t pte)
+{
+ if (unlikely(pte_valid_cont(pte)))
+ __contpte_try_unfold(mm, addr, ptep, pte);
+}
+EXPORT_SYMBOL_GPL(contpte_try_unfold);
+
+static bool contpte_is_last_pte(unsigned long addr, pte_t pte, int ncontig)
+{
+ const unsigned long contmask = ncontig - 1;
+ bool valign = ((addr >> PAGE_SHIFT) & contmask) == contmask;
+
+ if (unlikely(valign)) {
+ bool palign = (pte_pfn(pte) & contmask) == contmask;
+
+ if (unlikely(palign &&
+ pte_valid(pte) && !pte_cont(pte) && !pte_special(pte)))
+ return true;
+ }
+
+ return false;
+}
+
+void __contpte_try_fold(struct mm_struct *mm, unsigned long addr,
+ pte_t *ptep, pte_t pte)
+{
+ /*
+ * We have already checked that the virtual and pysical addresses are
+ * correctly aligned for a contpte mapping in contpte_try_fold() so the
+ * remaining checks are to ensure that the contpte range is fully
+ * covered by a single folio, and ensure that all the ptes are valid
+ * with contiguous PFNs and matching prots. We ignore the state of the
+ * access and dirty bits for the purpose of deciding if its a contiguous
+ * range; the folding process will generate a single contpte entry which
+ * has a single access and dirty bit. Those 2 bits are the logical OR of
+ * their respective bits in the constituent pte entries. In order to
+ * ensure the contpte range is covered by a single folio, we must
+ * recover the folio from the pfn, but special mappings don't have a
+ * folio backing them. Fortunately contpte_try_fold() already checked
+ * that the pte is not special - we never try to fold special mappings.
+ * Note we can't use vm_normal_page() for this since we don't have the
+ * vma.
+ */
+
+ unsigned long folio_start, folio_end;
+ unsigned long cont_start, cont_end;
+ pte_t expected_pte, subpte;
+ struct folio *folio;
+ struct page *page;
+ unsigned long pfn;
+ pte_t *cur_ptep;
+ pgprot_t prot;
+ size_t pgsize;
+ int i, ncontig, ncontig_prev;
+
+ if (!mm_is_user(mm))
+ return;
+
+ page = pte_page(pte);
+ folio = page_folio(page);
+ folio_start = addr - (page - &folio->page) * PAGE_SIZE;
+ folio_end = folio_start + folio_nr_pages(folio) * PAGE_SIZE;
+
+ prot = pte_pgprot(pte_mkold(pte_mkclean(pte)));
+ ncontig_prev = 0;
+ ncontig = arch_contpte_get_first_ncontig(&pgsize);
+
+ do {
+ /* Make sure we still belong to the same folio */
+ cont_start = ALIGN_DOWN(addr, ncontig * pgsize);
+ cont_end = cont_start + ncontig * pgsize;
+
+ if (folio_start > cont_start || folio_end < cont_end)
+ break;
+
+ pfn = ALIGN_DOWN(pte_pfn(pte), ncontig);
+ expected_pte = pfn_pte(pfn, prot);
+ cur_ptep = PTR_ALIGN_DOWN(ptep, sizeof(*ptep) * ncontig);
+
+ /*
+ * Either the region is already a contpte mapping or make sure
+ * the ptes belong to a contpte region.
+ */
+ if (!pte_valid_cont(__ptep_get(cur_ptep))) {
+ for (i = 0; i < ncontig - ncontig_prev; i++) {
+ subpte = pte_mkold(pte_mkclean(__ptep_get(cur_ptep)));
+ if (!pte_same(subpte, expected_pte))
+ goto out;
+ expected_pte = pte_advance_pfn(expected_pte, 1);
+ cur_ptep++;
+ }
+ }
+
+ /*
+ * Compute the next contpte region to check: avoid checking
+ * the same region again by only checking the "second-half"
+ * (the "region buddy") of the current region, which keeps the
+ * whole thing in O(n).
+ */
+ ncontig_prev = ncontig;
+ ncontig = arch_contpte_get_next_ncontig(ncontig, &pgsize);
+ /* set_ptes spec states that "The PTEs are all in the same PMD" */
+ if (!ncontig || pgsize > PAGE_SIZE)
+ break;
+ } while (contpte_is_last_pte(addr, pte, ncontig));
+
+out:
+ if (!ncontig_prev)
+ return;
+
+ ptep = PTR_ALIGN_DOWN(ptep, sizeof(*ptep) * ncontig_prev);
+ cont_start = ALIGN_DOWN(addr, ncontig_prev * PAGE_SIZE);
+ pte = pte_mkcont(pte);
+ pte = pfn_pte(ALIGN_DOWN(pte_pfn(pte), ncontig_prev), pte_pgprot(pte));
+
+ contpte_convert(mm, cont_start, ptep, pte, ncontig_prev, PAGE_SIZE);
+}
+EXPORT_SYMBOL_GPL(__contpte_try_fold);
+
+__always_inline void contpte_try_fold(struct mm_struct *mm,
+ unsigned long addr, pte_t *ptep, pte_t pte)
+{
+ /*
+ * Only bother trying if both the virtual and physical addresses are
+ * aligned and correspond to the last entry in a contig range. The core
+ * code mostly modifies ranges from low to high, so this is the likely
+ * the last modification in the contig range, so a good time to fold.
+ * We can't fold special mappings, because there is no associated folio.
+ *
+ * Only test if the first ncontig is aligned, because it can't be the
+ * last pte of a size larger than the first ncontig and not the last
+ * of the first ncontig.
+ */
+ int ncontig = arch_contpte_get_first_ncontig(NULL);
+
+ if (contpte_is_last_pte(addr, pte, ncontig))
+ __contpte_try_fold(mm, addr, ptep, pte);
+}
+
+void contpte_set_ptes(struct mm_struct *mm, unsigned long addr,
+ pte_t *ptep, pte_t pte, unsigned int nr)
+{
+ unsigned long contmask;
+ unsigned long next;
+ unsigned long end;
+ unsigned long pfn;
+ pgprot_t prot;
+ int ncontig;
+
+ /*
+ * The set_ptes() spec guarantees that when nr > 1, the initial state of
+ * all ptes is not-present. Therefore we never need to unfold or
+ * otherwise invalidate a range before we set the new ptes.
+ * contpte_set_ptes() should never be called for nr < 2.
+ */
+ VM_WARN_ON(nr == 1);
+
+ if (!mm_is_user(mm))
+ return __set_ptes(mm, addr, ptep, pte, nr);
+
+ end = addr + (nr << PAGE_SHIFT);
+ pfn = pte_pfn(pte);
+ prot = pte_pgprot(pte);
+
+ do {
+ next = arch_contpte_addr_end(addr, end, &ncontig);
+ nr = (next - addr) >> PAGE_SHIFT;
+ pte = pfn_pte(pfn, prot);
+ contmask = (ncontig << PAGE_SHIFT) - 1;
+
+ if (ncontig && ((addr | next | (pfn << PAGE_SHIFT)) & contmask) == 0)
+ pte = pte_mkcont(pte);
+ else
+ pte = pte_mknoncont(pte);
+
+ __set_ptes(mm, addr, ptep, pte, nr);
+
+ addr = next;
+ ptep += nr;
+ pfn += nr;
+ } while (addr != end);
+}
+EXPORT_SYMBOL_GPL(contpte_set_ptes);
+
+__always_inline void set_ptes(struct mm_struct *mm, unsigned long addr,
+ pte_t *ptep, pte_t pte, unsigned int nr)
+{
+ pte = pte_mknoncont(pte);
+
+ if (likely(nr == 1)) {
+ contpte_try_unfold(mm, addr, ptep, __ptep_get(ptep));
+ __set_ptes(mm, addr, ptep, pte, 1);
+ contpte_try_fold(mm, addr, ptep, pte);
+ } else {
+ contpte_set_ptes(mm, addr, ptep, pte, nr);
+ }
+}
+#endif /* CONFIG_THP_CONTPTE */
--
2.39.2
^ permalink raw reply related [flat|nested] 16+ messages in thread
* [PATCH 04/12] mm, riscv, arm64: Use common ptep_get_lockless() function
2024-05-08 19:19 [PATCH 00/12] Make riscv use THP contpte support for arm64 Alexandre Ghiti
` (2 preceding siblings ...)
2024-05-08 19:19 ` [PATCH 03/12] mm, riscv, arm64: Use common set_ptes() function Alexandre Ghiti
@ 2024-05-08 19:19 ` Alexandre Ghiti
2024-05-08 19:19 ` [PATCH 05/12] mm, riscv, arm64: Use common set_pte() function Alexandre Ghiti
` (7 subsequent siblings)
11 siblings, 0 replies; 16+ messages in thread
From: Alexandre Ghiti @ 2024-05-08 19:19 UTC (permalink / raw
To: Ryan Roberts, Catalin Marinas, Will Deacon, Alexander Potapenko,
Marco Elver, Dmitry Vyukov, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Ard Biesheuvel, Anup Patel, Atish Patra,
Andrey Ryabinin, Andrey Konovalov, Vincenzo Frascino,
Andrew Morton, linux-arm-kernel, linux-kernel, kasan-dev,
linux-riscv, linux-efi, kvm, kvm-riscv, linux-mm
Cc: Alexandre Ghiti
Make riscv use the contpte aware ptep_get_lockless() function from arm64.
Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com>
---
arch/arm64/include/asm/pgtable.h | 11 +----
arch/arm64/mm/contpte.c | 57 --------------------------
arch/riscv/include/asm/pgtable.h | 2 +
include/linux/contpte.h | 1 +
mm/contpte.c | 69 ++++++++++++++++++++++++++++++++
5 files changed, 73 insertions(+), 67 deletions(-)
diff --git a/arch/arm64/include/asm/pgtable.h b/arch/arm64/include/asm/pgtable.h
index e85b3a052a02..8a0603257436 100644
--- a/arch/arm64/include/asm/pgtable.h
+++ b/arch/arm64/include/asm/pgtable.h
@@ -1384,7 +1384,6 @@ extern void ptep_modify_prot_commit(struct vm_area_struct *vma,
* where it is possible and makes sense to do so. The PTE_CONT bit is considered
* a private implementation detail of the public ptep API (see below).
*/
-extern pte_t contpte_ptep_get_lockless(pte_t *orig_ptep);
extern void contpte_clear_full_ptes(struct mm_struct *mm, unsigned long addr,
pte_t *ptep, unsigned int nr, int full);
extern pte_t contpte_get_and_clear_full_ptes(struct mm_struct *mm,
@@ -1430,16 +1429,8 @@ static inline unsigned int pte_batch_hint(pte_t *ptep, pte_t pte)
extern pte_t ptep_get(pte_t *ptep);
#define ptep_get ptep_get
+extern pte_t ptep_get_lockless(pte_t *ptep);
#define ptep_get_lockless ptep_get_lockless
-static inline pte_t ptep_get_lockless(pte_t *ptep)
-{
- pte_t pte = __ptep_get(ptep);
-
- if (likely(!pte_valid_cont(pte)))
- return pte;
-
- return contpte_ptep_get_lockless(ptep);
-}
static inline void set_pte(pte_t *ptep, pte_t pte)
{
diff --git a/arch/arm64/mm/contpte.c b/arch/arm64/mm/contpte.c
index e225e458856e..5e9e40145085 100644
--- a/arch/arm64/mm/contpte.c
+++ b/arch/arm64/mm/contpte.c
@@ -28,63 +28,6 @@ static void contpte_try_unfold_partial(struct mm_struct *mm, unsigned long addr,
}
}
-pte_t contpte_ptep_get_lockless(pte_t *orig_ptep)
-{
- /*
- * The ptep_get_lockless() API requires us to read and return *orig_ptep
- * so that it is self-consistent, without the PTL held, so we may be
- * racing with other threads modifying the pte. Usually a READ_ONCE()
- * would suffice, but for the contpte case, we also need to gather the
- * access and dirty bits from across all ptes in the contiguous block,
- * and we can't read all of those neighbouring ptes atomically, so any
- * contiguous range may be unfolded/modified/refolded under our feet.
- * Therefore we ensure we read a _consistent_ contpte range by checking
- * that all ptes in the range are valid and have CONT_PTE set, that all
- * pfns are contiguous and that all pgprots are the same (ignoring
- * access/dirty). If we find a pte that is not consistent, then we must
- * be racing with an update so start again. If the target pte does not
- * have CONT_PTE set then that is considered consistent on its own
- * because it is not part of a contpte range.
- */
-
- pgprot_t orig_prot;
- unsigned long pfn;
- pte_t orig_pte;
- pgprot_t prot;
- pte_t *ptep;
- pte_t pte;
- int i;
-
-retry:
- orig_pte = __ptep_get(orig_ptep);
-
- if (!pte_valid_cont(orig_pte))
- return orig_pte;
-
- orig_prot = pte_pgprot(pte_mkold(pte_mkclean(orig_pte)));
- ptep = arch_contpte_align_down(orig_ptep);
- pfn = pte_pfn(orig_pte) - (orig_ptep - ptep);
-
- for (i = 0; i < CONT_PTES; i++, ptep++, pfn++) {
- pte = __ptep_get(ptep);
- prot = pte_pgprot(pte_mkold(pte_mkclean(pte)));
-
- if (!pte_valid_cont(pte) ||
- pte_pfn(pte) != pfn ||
- pgprot_val(prot) != pgprot_val(orig_prot))
- goto retry;
-
- if (pte_dirty(pte))
- orig_pte = pte_mkdirty(orig_pte);
-
- if (pte_young(pte))
- orig_pte = pte_mkyoung(orig_pte);
- }
-
- return orig_pte;
-}
-EXPORT_SYMBOL_GPL(contpte_ptep_get_lockless);
-
void contpte_clear_full_ptes(struct mm_struct *mm, unsigned long addr,
pte_t *ptep, unsigned int nr, int full)
{
diff --git a/arch/riscv/include/asm/pgtable.h b/arch/riscv/include/asm/pgtable.h
index ebfe6b16529e..62cad1b974f1 100644
--- a/arch/riscv/include/asm/pgtable.h
+++ b/arch/riscv/include/asm/pgtable.h
@@ -785,6 +785,8 @@ static inline int ptep_clear_flush_young(struct vm_area_struct *vma,
extern pte_t ptep_get(pte_t *ptep);
#define ptep_get ptep_get
+extern pte_t ptep_get_lockless(pte_t *ptep);
+#define ptep_get_lockless ptep_get_lockless
extern void set_ptes(struct mm_struct *mm, unsigned long addr,
pte_t *ptep, pte_t pteval, unsigned int nr);
#define set_ptes set_ptes
diff --git a/include/linux/contpte.h b/include/linux/contpte.h
index 54d10204e9af..01da4bfc3af6 100644
--- a/include/linux/contpte.h
+++ b/include/linux/contpte.h
@@ -8,6 +8,7 @@
* a private implementation detail of the public ptep API (see below).
*/
pte_t contpte_ptep_get(pte_t *ptep, pte_t orig_pte);
+pte_t contpte_ptep_get_lockless(pte_t *orig_ptep);
void __contpte_try_fold(struct mm_struct *mm, unsigned long addr,
pte_t *ptep, pte_t pte);
void contpte_try_fold(struct mm_struct *mm, unsigned long addr,
diff --git a/mm/contpte.c b/mm/contpte.c
index 566745d7842f..060e0bc1a2a3 100644
--- a/mm/contpte.c
+++ b/mm/contpte.c
@@ -42,6 +42,7 @@
* - huge_ptep_clear_flush()
* - ptep_get()
* - set_ptes()
+ * - ptep_get_lockless()
*/
pte_t huge_ptep_get(pte_t *ptep)
@@ -589,4 +590,72 @@ __always_inline void set_ptes(struct mm_struct *mm, unsigned long addr,
contpte_set_ptes(mm, addr, ptep, pte, nr);
}
}
+
+pte_t contpte_ptep_get_lockless(pte_t *orig_ptep)
+{
+ /*
+ * The ptep_get_lockless() API requires us to read and return *orig_ptep
+ * so that it is self-consistent, without the PTL held, so we may be
+ * racing with other threads modifying the pte. Usually a READ_ONCE()
+ * would suffice, but for the contpte case, we also need to gather the
+ * access and dirty bits from across all ptes in the contiguous block,
+ * and we can't read all of those neighbouring ptes atomically, so any
+ * contiguous range may be unfolded/modified/refolded under our feet.
+ * Therefore we ensure we read a _consistent_ contpte range by checking
+ * that all ptes in the range are valid and have CONT_PTE set, that all
+ * pfns are contiguous and that all pgprots are the same (ignoring
+ * access/dirty). If we find a pte that is not consistent, then we must
+ * be racing with an update so start again. If the target pte does not
+ * have CONT_PTE set then that is considered consistent on its own
+ * because it is not part of a contpte range.
+ */
+
+ pgprot_t orig_prot;
+ unsigned long pfn;
+ pte_t orig_pte;
+ pgprot_t prot;
+ pte_t *ptep;
+ pte_t pte;
+ int i, ncontig;
+
+retry:
+ orig_pte = __ptep_get(orig_ptep);
+
+ if (!pte_valid_cont(orig_pte))
+ return orig_pte;
+
+ orig_prot = pte_pgprot(pte_mkold(pte_mkclean(orig_pte)));
+ ptep = arch_contpte_align_down(orig_ptep);
+ ncontig = arch_contpte_get_num_contig(NULL, 0, ptep, 0, NULL);
+ pfn = pte_pfn(orig_pte) - (orig_ptep - ptep);
+
+ for (i = 0; i < ncontig; i++, ptep++, pfn++) {
+ pte = __ptep_get(ptep);
+ prot = pte_pgprot(pte_mkold(pte_mkclean(pte)));
+
+ if (!pte_valid_cont(pte) ||
+ pte_pfn(pte) != pfn ||
+ pgprot_val(prot) != pgprot_val(orig_prot))
+ goto retry;
+
+ if (pte_dirty(pte))
+ orig_pte = pte_mkdirty(orig_pte);
+
+ if (pte_young(pte))
+ orig_pte = pte_mkyoung(orig_pte);
+ }
+
+ return orig_pte;
+}
+EXPORT_SYMBOL_GPL(contpte_ptep_get_lockless);
+
+__always_inline pte_t ptep_get_lockless(pte_t *ptep)
+{
+ pte_t pte = __ptep_get(ptep);
+
+ if (likely(!pte_valid_cont(pte)))
+ return pte;
+
+ return contpte_ptep_get_lockless(ptep);
+}
#endif /* CONFIG_THP_CONTPTE */
--
2.39.2
^ permalink raw reply related [flat|nested] 16+ messages in thread
* [PATCH 05/12] mm, riscv, arm64: Use common set_pte() function
2024-05-08 19:19 [PATCH 00/12] Make riscv use THP contpte support for arm64 Alexandre Ghiti
` (3 preceding siblings ...)
2024-05-08 19:19 ` [PATCH 04/12] mm, riscv, arm64: Use common ptep_get_lockless() function Alexandre Ghiti
@ 2024-05-08 19:19 ` Alexandre Ghiti
2024-05-08 19:19 ` [PATCH 06/12] mm, riscv, arm64: Use common pte_clear() function Alexandre Ghiti
` (6 subsequent siblings)
11 siblings, 0 replies; 16+ messages in thread
From: Alexandre Ghiti @ 2024-05-08 19:19 UTC (permalink / raw
To: Ryan Roberts, Catalin Marinas, Will Deacon, Alexander Potapenko,
Marco Elver, Dmitry Vyukov, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Ard Biesheuvel, Anup Patel, Atish Patra,
Andrey Ryabinin, Andrey Konovalov, Vincenzo Frascino,
Andrew Morton, linux-arm-kernel, linux-kernel, kasan-dev,
linux-riscv, linux-efi, kvm, kvm-riscv, linux-mm
Cc: Alexandre Ghiti
Make riscv use the contpte aware set_pte() function from arm64.
Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com>
---
arch/arm64/include/asm/pgtable.h | 16 ++--------------
arch/riscv/include/asm/kfence.h | 4 ++--
arch/riscv/include/asm/pgtable.h | 7 +++++--
arch/riscv/kernel/efi.c | 2 +-
arch/riscv/kernel/hibernate.c | 2 +-
arch/riscv/kvm/mmu.c | 10 +++++-----
arch/riscv/mm/init.c | 2 +-
arch/riscv/mm/kasan_init.c | 14 +++++++-------
arch/riscv/mm/pageattr.c | 4 ++--
mm/contpte.c | 18 ++++++++++++++++++
10 files changed, 44 insertions(+), 35 deletions(-)
diff --git a/arch/arm64/include/asm/pgtable.h b/arch/arm64/include/asm/pgtable.h
index 8a0603257436..bb6210fb72c8 100644
--- a/arch/arm64/include/asm/pgtable.h
+++ b/arch/arm64/include/asm/pgtable.h
@@ -1432,20 +1432,8 @@ extern pte_t ptep_get(pte_t *ptep);
extern pte_t ptep_get_lockless(pte_t *ptep);
#define ptep_get_lockless ptep_get_lockless
-static inline void set_pte(pte_t *ptep, pte_t pte)
-{
- /*
- * We don't have the mm or vaddr so cannot unfold contig entries (since
- * it requires tlb maintenance). set_pte() is not used in core code, so
- * this should never even be called. Regardless do our best to service
- * any call and emit a warning if there is any attempt to set a pte on
- * top of an existing contig range.
- */
- pte_t orig_pte = __ptep_get(ptep);
-
- WARN_ON_ONCE(pte_valid_cont(orig_pte));
- __set_pte(ptep, pte_mknoncont(pte));
-}
+extern void set_pte(pte_t *ptep, pte_t pte);
+#define set_pte set_pte
extern void set_ptes(struct mm_struct *mm, unsigned long addr,
pte_t *ptep, pte_t pte, unsigned int nr);
diff --git a/arch/riscv/include/asm/kfence.h b/arch/riscv/include/asm/kfence.h
index f303fef8591c..36e9f638abf6 100644
--- a/arch/riscv/include/asm/kfence.h
+++ b/arch/riscv/include/asm/kfence.h
@@ -18,9 +18,9 @@ static inline bool kfence_protect_page(unsigned long addr, bool protect)
pte_t *pte = virt_to_kpte(addr);
if (protect)
- set_pte(pte, __pte(pte_val(__ptep_get(pte)) & ~_PAGE_PRESENT));
+ __set_pte(pte, __pte(pte_val(__ptep_get(pte)) & ~_PAGE_PRESENT));
else
- set_pte(pte, __pte(pte_val(__ptep_get(pte)) | _PAGE_PRESENT));
+ __set_pte(pte, __pte(pte_val(__ptep_get(pte)) | _PAGE_PRESENT));
flush_tlb_kernel_range(addr, addr + PAGE_SIZE);
diff --git a/arch/riscv/include/asm/pgtable.h b/arch/riscv/include/asm/pgtable.h
index 62cad1b974f1..4f8f673787e7 100644
--- a/arch/riscv/include/asm/pgtable.h
+++ b/arch/riscv/include/asm/pgtable.h
@@ -539,7 +539,7 @@ static inline int pte_same(pte_t pte_a, pte_t pte_b)
* a page table are directly modified. Thus, the following hook is
* made available.
*/
-static inline void set_pte(pte_t *ptep, pte_t pteval)
+static inline void __set_pte(pte_t *ptep, pte_t pteval)
{
WRITE_ONCE(*ptep, pteval);
}
@@ -551,7 +551,7 @@ static inline void __set_pte_at(struct mm_struct *mm, pte_t *ptep, pte_t pteval)
if (pte_present(pteval) && pte_exec(pteval))
flush_icache_pte(mm, pteval);
- set_pte(ptep, pteval);
+ __set_pte(ptep, pteval);
}
#define PFN_PTE_SHIFT _PAGE_PFN_SHIFT
@@ -790,11 +790,14 @@ extern pte_t ptep_get_lockless(pte_t *ptep);
extern void set_ptes(struct mm_struct *mm, unsigned long addr,
pte_t *ptep, pte_t pteval, unsigned int nr);
#define set_ptes set_ptes
+extern void set_pte(pte_t *ptep, pte_t pte);
+#define set_pte set_pte
#else /* CONFIG_THP_CONTPTE */
#define ptep_get __ptep_get
#define set_ptes __set_ptes
+#define set_pte __set_pte
#endif /* CONFIG_THP_CONTPTE */
diff --git a/arch/riscv/kernel/efi.c b/arch/riscv/kernel/efi.c
index 3d2a635c69ac..673eca7705ba 100644
--- a/arch/riscv/kernel/efi.c
+++ b/arch/riscv/kernel/efi.c
@@ -72,7 +72,7 @@ static int __init set_permissions(pte_t *ptep, unsigned long addr, void *data)
val = pte_val(pte) & ~_PAGE_EXEC;
pte = __pte(val);
}
- set_pte(ptep, pte);
+ __set_pte(ptep, pte);
return 0;
}
diff --git a/arch/riscv/kernel/hibernate.c b/arch/riscv/kernel/hibernate.c
index 671b686c0158..97ed3df7a308 100644
--- a/arch/riscv/kernel/hibernate.c
+++ b/arch/riscv/kernel/hibernate.c
@@ -186,7 +186,7 @@ static int temp_pgtable_map_pte(pmd_t *dst_pmdp, pmd_t *src_pmdp, unsigned long
pte_t pte = READ_ONCE(*src_ptep);
if (pte_present(pte))
- set_pte(dst_ptep, __pte(pte_val(pte) | pgprot_val(prot)));
+ __set_pte(dst_ptep, __pte(pte_val(pte) | pgprot_val(prot)));
} while (dst_ptep++, src_ptep++, start += PAGE_SIZE, start < end);
return 0;
diff --git a/arch/riscv/kvm/mmu.c b/arch/riscv/kvm/mmu.c
index 70c6cb3864d6..1ee6139d495f 100644
--- a/arch/riscv/kvm/mmu.c
+++ b/arch/riscv/kvm/mmu.c
@@ -155,7 +155,7 @@ static int gstage_set_pte(struct kvm *kvm, u32 level,
next_ptep = kvm_mmu_memory_cache_alloc(pcache);
if (!next_ptep)
return -ENOMEM;
- set_pte(ptep, pfn_pte(PFN_DOWN(__pa(next_ptep)),
+ __set_pte(ptep, pfn_pte(PFN_DOWN(__pa(next_ptep)),
__pgprot(_PAGE_TABLE)));
} else {
if (gstage_pte_leaf(ptep))
@@ -167,7 +167,7 @@ static int gstage_set_pte(struct kvm *kvm, u32 level,
ptep = &next_ptep[gstage_pte_index(addr, current_level)];
}
- set_pte(ptep, *new_pte);
+ __set_pte(ptep, *new_pte);
if (gstage_pte_leaf(ptep))
gstage_remote_tlb_flush(kvm, current_level, addr);
@@ -251,7 +251,7 @@ static void gstage_op_pte(struct kvm *kvm, gpa_t addr,
return;
if (op == GSTAGE_OP_CLEAR)
- set_pte(ptep, __pte(0));
+ __set_pte(ptep, __pte(0));
for (i = 0; i < PTRS_PER_PTE; i++)
gstage_op_pte(kvm, addr + i * next_page_size,
&next_ptep[i], next_ptep_level, op);
@@ -259,9 +259,9 @@ static void gstage_op_pte(struct kvm *kvm, gpa_t addr,
put_page(virt_to_page(next_ptep));
} else {
if (op == GSTAGE_OP_CLEAR)
- set_pte(ptep, __pte(0));
+ __set_pte(ptep, __pte(0));
else if (op == GSTAGE_OP_WP)
- set_pte(ptep, __pte(pte_val(__ptep_get(ptep)) & ~_PAGE_WRITE));
+ __set_pte(ptep, __pte(pte_val(__ptep_get(ptep)) & ~_PAGE_WRITE));
gstage_remote_tlb_flush(kvm, ptep_level, addr);
}
}
diff --git a/arch/riscv/mm/init.c b/arch/riscv/mm/init.c
index fe8e159394d8..bb5c6578204c 100644
--- a/arch/riscv/mm/init.c
+++ b/arch/riscv/mm/init.c
@@ -325,7 +325,7 @@ void __set_fixmap(enum fixed_addresses idx, phys_addr_t phys, pgprot_t prot)
ptep = &fixmap_pte[pte_index(addr)];
if (pgprot_val(prot))
- set_pte(ptep, pfn_pte(phys >> PAGE_SHIFT, prot));
+ __set_pte(ptep, pfn_pte(phys >> PAGE_SHIFT, prot));
else
pte_clear(&init_mm, addr, ptep);
local_flush_tlb_page(addr);
diff --git a/arch/riscv/mm/kasan_init.c b/arch/riscv/mm/kasan_init.c
index 381d61f42ab8..b5061cb3ce4d 100644
--- a/arch/riscv/mm/kasan_init.c
+++ b/arch/riscv/mm/kasan_init.c
@@ -41,7 +41,7 @@ static void __init kasan_populate_pte(pmd_t *pmd, unsigned long vaddr, unsigned
do {
if (pte_none(__ptep_get(ptep))) {
phys_addr = memblock_phys_alloc(PAGE_SIZE, PAGE_SIZE);
- set_pte(ptep, pfn_pte(PFN_DOWN(phys_addr), PAGE_KERNEL));
+ __set_pte(ptep, pfn_pte(PFN_DOWN(phys_addr), PAGE_KERNEL));
memset(__va(phys_addr), KASAN_SHADOW_INIT, PAGE_SIZE);
}
} while (ptep++, vaddr += PAGE_SIZE, vaddr != end);
@@ -327,8 +327,8 @@ asmlinkage void __init kasan_early_init(void)
KASAN_SHADOW_END - (1UL << (64 - KASAN_SHADOW_SCALE_SHIFT)));
for (i = 0; i < PTRS_PER_PTE; ++i)
- set_pte(kasan_early_shadow_pte + i,
- pfn_pte(virt_to_pfn(kasan_early_shadow_page), PAGE_KERNEL));
+ __set_pte(kasan_early_shadow_pte + i,
+ pfn_pte(virt_to_pfn(kasan_early_shadow_page), PAGE_KERNEL));
for (i = 0; i < PTRS_PER_PMD; ++i)
set_pmd(kasan_early_shadow_pmd + i,
@@ -523,10 +523,10 @@ void __init kasan_init(void)
kasan_mem_to_shadow((const void *)MODULES_VADDR + SZ_2G));
for (i = 0; i < PTRS_PER_PTE; i++)
- set_pte(&kasan_early_shadow_pte[i],
- mk_pte(virt_to_page(kasan_early_shadow_page),
- __pgprot(_PAGE_PRESENT | _PAGE_READ |
- _PAGE_ACCESSED)));
+ __set_pte(&kasan_early_shadow_pte[i],
+ mk_pte(virt_to_page(kasan_early_shadow_page),
+ __pgprot(_PAGE_PRESENT | _PAGE_READ |
+ _PAGE_ACCESSED)));
memset(kasan_early_shadow_page, KASAN_SHADOW_INIT, PAGE_SIZE);
init_task.kasan_depth = 0;
diff --git a/arch/riscv/mm/pageattr.c b/arch/riscv/mm/pageattr.c
index 98c9dc4b983c..d623e4fc11fc 100644
--- a/arch/riscv/mm/pageattr.c
+++ b/arch/riscv/mm/pageattr.c
@@ -71,7 +71,7 @@ static int pageattr_pte_entry(pte_t *pte, unsigned long addr,
pte_t val = __ptep_get(pte);
val = __pte(set_pageattr_masks(pte_val(val), walk));
- set_pte(pte, val);
+ __set_pte(pte, val);
return 0;
}
@@ -121,7 +121,7 @@ static int __split_linear_mapping_pmd(pud_t *pudp,
ptep_new = (pte_t *)page_address(pte_page);
for (i = 0; i < PTRS_PER_PTE; ++i, ++ptep_new)
- set_pte(ptep_new, pfn_pte(pfn + i, prot));
+ __set_pte(ptep_new, pfn_pte(pfn + i, prot));
smp_wmb();
diff --git a/mm/contpte.c b/mm/contpte.c
index 060e0bc1a2a3..543ae5b5a863 100644
--- a/mm/contpte.c
+++ b/mm/contpte.c
@@ -17,6 +17,7 @@
* - __pte_clear()
* - __ptep_set_access_flags()
* - __ptep_set_wrprotect()
+ * - __set_pte()
* - pte_cont()
* - arch_contpte_get_num_contig()
* - pte_valid_cont()
@@ -43,6 +44,7 @@
* - ptep_get()
* - set_ptes()
* - ptep_get_lockless()
+ * - set_pte()
*/
pte_t huge_ptep_get(pte_t *ptep)
@@ -658,4 +660,20 @@ __always_inline pte_t ptep_get_lockless(pte_t *ptep)
return contpte_ptep_get_lockless(ptep);
}
+
+void set_pte(pte_t *ptep, pte_t pte)
+{
+ /*
+ * We don't have the mm or vaddr so cannot unfold contig entries (since
+ * it requires tlb maintenance). set_pte() is not used in core code, so
+ * this should never even be called. Regardless do our best to service
+ * any call and emit a warning if there is any attempt to set a pte on
+ * top of an existing contig range.
+ */
+ pte_t orig_pte = __ptep_get(ptep);
+
+ WARN_ON_ONCE(pte_valid_cont(orig_pte));
+ __set_pte(ptep, pte_mknoncont(pte));
+}
+
#endif /* CONFIG_THP_CONTPTE */
--
2.39.2
^ permalink raw reply related [flat|nested] 16+ messages in thread
* [PATCH 06/12] mm, riscv, arm64: Use common pte_clear() function
2024-05-08 19:19 [PATCH 00/12] Make riscv use THP contpte support for arm64 Alexandre Ghiti
` (4 preceding siblings ...)
2024-05-08 19:19 ` [PATCH 05/12] mm, riscv, arm64: Use common set_pte() function Alexandre Ghiti
@ 2024-05-08 19:19 ` Alexandre Ghiti
2024-05-08 19:19 ` [PATCH 07/12] mm, riscv, arm64: Use common ptep_get_and_clear() function Alexandre Ghiti
` (5 subsequent siblings)
11 siblings, 0 replies; 16+ messages in thread
From: Alexandre Ghiti @ 2024-05-08 19:19 UTC (permalink / raw
To: Ryan Roberts, Catalin Marinas, Will Deacon, Alexander Potapenko,
Marco Elver, Dmitry Vyukov, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Ard Biesheuvel, Anup Patel, Atish Patra,
Andrey Ryabinin, Andrey Konovalov, Vincenzo Frascino,
Andrew Morton, linux-arm-kernel, linux-kernel, kasan-dev,
linux-riscv, linux-efi, kvm, kvm-riscv, linux-mm
Cc: Alexandre Ghiti
Make riscv use the contpte aware pte_clear() function from arm64.
Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com>
---
arch/arm64/include/asm/pgtable.h | 9 +++------
arch/riscv/include/asm/pgtable.h | 4 +++-
arch/riscv/mm/init.c | 2 +-
mm/contpte.c | 6 ++++++
4 files changed, 13 insertions(+), 8 deletions(-)
diff --git a/arch/arm64/include/asm/pgtable.h b/arch/arm64/include/asm/pgtable.h
index bb6210fb72c8..74e582f2884f 100644
--- a/arch/arm64/include/asm/pgtable.h
+++ b/arch/arm64/include/asm/pgtable.h
@@ -1439,12 +1439,9 @@ extern void set_ptes(struct mm_struct *mm, unsigned long addr,
pte_t *ptep, pte_t pte, unsigned int nr);
#define set_ptes set_ptes
-static inline void pte_clear(struct mm_struct *mm,
- unsigned long addr, pte_t *ptep)
-{
- contpte_try_unfold(mm, addr, ptep, __ptep_get(ptep));
- __pte_clear(mm, addr, ptep);
-}
+extern void pte_clear(struct mm_struct *mm,
+ unsigned long addr, pte_t *ptep);
+#define pte_clear pte_clear
#define clear_full_ptes clear_full_ptes
static inline void clear_full_ptes(struct mm_struct *mm, unsigned long addr,
diff --git a/arch/riscv/include/asm/pgtable.h b/arch/riscv/include/asm/pgtable.h
index 4f8f673787e7..41534f4b8a6d 100644
--- a/arch/riscv/include/asm/pgtable.h
+++ b/arch/riscv/include/asm/pgtable.h
@@ -792,18 +792,20 @@ extern void set_ptes(struct mm_struct *mm, unsigned long addr,
#define set_ptes set_ptes
extern void set_pte(pte_t *ptep, pte_t pte);
#define set_pte set_pte
+extern void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep);
+#define pte_clear pte_clear
#else /* CONFIG_THP_CONTPTE */
#define ptep_get __ptep_get
#define set_ptes __set_ptes
#define set_pte __set_pte
+#define pte_clear __pte_clear
#endif /* CONFIG_THP_CONTPTE */
#define __HAVE_ARCH_PTEP_GET_AND_CLEAR
#define ptep_get_and_clear __ptep_get_and_clear
-#define pte_clear __pte_clear
#define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
#define ptep_set_access_flags __ptep_set_access_flags
#define __HAVE_ARCH_PTEP_SET_WRPROTECT
diff --git a/arch/riscv/mm/init.c b/arch/riscv/mm/init.c
index bb5c6578204c..c82f17b3060b 100644
--- a/arch/riscv/mm/init.c
+++ b/arch/riscv/mm/init.c
@@ -327,7 +327,7 @@ void __set_fixmap(enum fixed_addresses idx, phys_addr_t phys, pgprot_t prot)
if (pgprot_val(prot))
__set_pte(ptep, pfn_pte(phys >> PAGE_SHIFT, prot));
else
- pte_clear(&init_mm, addr, ptep);
+ __pte_clear(&init_mm, addr, ptep);
local_flush_tlb_page(addr);
}
diff --git a/mm/contpte.c b/mm/contpte.c
index 543ae5b5a863..c9eff6426ca0 100644
--- a/mm/contpte.c
+++ b/mm/contpte.c
@@ -45,6 +45,7 @@
* - set_ptes()
* - ptep_get_lockless()
* - set_pte()
+ * - pte_clear()
*/
pte_t huge_ptep_get(pte_t *ptep)
@@ -676,4 +677,9 @@ void set_pte(pte_t *ptep, pte_t pte)
__set_pte(ptep, pte_mknoncont(pte));
}
+void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
+{
+ contpte_try_unfold(mm, addr, ptep, __ptep_get(ptep));
+ __pte_clear(mm, addr, ptep);
+}
#endif /* CONFIG_THP_CONTPTE */
--
2.39.2
^ permalink raw reply related [flat|nested] 16+ messages in thread
* [PATCH 07/12] mm, riscv, arm64: Use common ptep_get_and_clear() function
2024-05-08 19:19 [PATCH 00/12] Make riscv use THP contpte support for arm64 Alexandre Ghiti
` (5 preceding siblings ...)
2024-05-08 19:19 ` [PATCH 06/12] mm, riscv, arm64: Use common pte_clear() function Alexandre Ghiti
@ 2024-05-08 19:19 ` Alexandre Ghiti
2024-05-08 19:19 ` [PATCH 08/12] mm, riscv, arm64: Use common ptep_test_and_clear_young() function Alexandre Ghiti
` (4 subsequent siblings)
11 siblings, 0 replies; 16+ messages in thread
From: Alexandre Ghiti @ 2024-05-08 19:19 UTC (permalink / raw
To: Ryan Roberts, Catalin Marinas, Will Deacon, Alexander Potapenko,
Marco Elver, Dmitry Vyukov, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Ard Biesheuvel, Anup Patel, Atish Patra,
Andrey Ryabinin, Andrey Konovalov, Vincenzo Frascino,
Andrew Morton, linux-arm-kernel, linux-kernel, kasan-dev,
linux-riscv, linux-efi, kvm, kvm-riscv, linux-mm
Cc: Alexandre Ghiti
Make riscv use the contpte aware ptep_get_and_clear() function from arm64.
Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com>
---
arch/arm64/include/asm/pgtable.h | 8 ++------
arch/riscv/include/asm/pgtable.h | 7 +++++--
mm/contpte.c | 8 ++++++++
3 files changed, 15 insertions(+), 8 deletions(-)
diff --git a/arch/arm64/include/asm/pgtable.h b/arch/arm64/include/asm/pgtable.h
index 74e582f2884f..ff7fe1d9cabe 100644
--- a/arch/arm64/include/asm/pgtable.h
+++ b/arch/arm64/include/asm/pgtable.h
@@ -1473,12 +1473,8 @@ static inline pte_t get_and_clear_full_ptes(struct mm_struct *mm,
}
#define __HAVE_ARCH_PTEP_GET_AND_CLEAR
-static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
- unsigned long addr, pte_t *ptep)
-{
- contpte_try_unfold(mm, addr, ptep, __ptep_get(ptep));
- return __ptep_get_and_clear(mm, addr, ptep);
-}
+extern pte_t ptep_get_and_clear(struct mm_struct *mm,
+ unsigned long addr, pte_t *ptep);
#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
static inline int ptep_test_and_clear_young(struct vm_area_struct *vma,
diff --git a/arch/riscv/include/asm/pgtable.h b/arch/riscv/include/asm/pgtable.h
index 41534f4b8a6d..03cd640137ed 100644
--- a/arch/riscv/include/asm/pgtable.h
+++ b/arch/riscv/include/asm/pgtable.h
@@ -794,6 +794,9 @@ extern void set_pte(pte_t *ptep, pte_t pte);
#define set_pte set_pte
extern void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep);
#define pte_clear pte_clear
+#define __HAVE_ARCH_PTEP_GET_AND_CLEAR
+extern pte_t ptep_get_and_clear(struct mm_struct *mm,
+ unsigned long addr, pte_t *ptep);
#else /* CONFIG_THP_CONTPTE */
@@ -801,11 +804,11 @@ extern void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep);
#define set_ptes __set_ptes
#define set_pte __set_pte
#define pte_clear __pte_clear
+#define __HAVE_ARCH_PTEP_GET_AND_CLEAR
+#define ptep_get_and_clear __ptep_get_and_clear
#endif /* CONFIG_THP_CONTPTE */
-#define __HAVE_ARCH_PTEP_GET_AND_CLEAR
-#define ptep_get_and_clear __ptep_get_and_clear
#define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
#define ptep_set_access_flags __ptep_set_access_flags
#define __HAVE_ARCH_PTEP_SET_WRPROTECT
diff --git a/mm/contpte.c b/mm/contpte.c
index c9eff6426ca0..5bf939639233 100644
--- a/mm/contpte.c
+++ b/mm/contpte.c
@@ -46,6 +46,7 @@
* - ptep_get_lockless()
* - set_pte()
* - pte_clear()
+ * - ptep_get_and_clear()
*/
pte_t huge_ptep_get(pte_t *ptep)
@@ -682,4 +683,11 @@ void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
contpte_try_unfold(mm, addr, ptep, __ptep_get(ptep));
__pte_clear(mm, addr, ptep);
}
+
+pte_t ptep_get_and_clear(struct mm_struct *mm,
+ unsigned long addr, pte_t *ptep)
+{
+ contpte_try_unfold(mm, addr, ptep, __ptep_get(ptep));
+ return __ptep_get_and_clear(mm, addr, ptep);
+}
#endif /* CONFIG_THP_CONTPTE */
--
2.39.2
^ permalink raw reply related [flat|nested] 16+ messages in thread
* [PATCH 08/12] mm, riscv, arm64: Use common ptep_test_and_clear_young() function
2024-05-08 19:19 [PATCH 00/12] Make riscv use THP contpte support for arm64 Alexandre Ghiti
` (6 preceding siblings ...)
2024-05-08 19:19 ` [PATCH 07/12] mm, riscv, arm64: Use common ptep_get_and_clear() function Alexandre Ghiti
@ 2024-05-08 19:19 ` Alexandre Ghiti
2024-05-08 19:19 ` [PATCH 09/12] mm, riscv, arm64: Use common ptep_clear_flush_young() function Alexandre Ghiti
` (3 subsequent siblings)
11 siblings, 0 replies; 16+ messages in thread
From: Alexandre Ghiti @ 2024-05-08 19:19 UTC (permalink / raw
To: Ryan Roberts, Catalin Marinas, Will Deacon, Alexander Potapenko,
Marco Elver, Dmitry Vyukov, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Ard Biesheuvel, Anup Patel, Atish Patra,
Andrey Ryabinin, Andrey Konovalov, Vincenzo Frascino,
Andrew Morton, linux-arm-kernel, linux-kernel, kasan-dev,
linux-riscv, linux-efi, kvm, kvm-riscv, linux-mm
Cc: Alexandre Ghiti
Make riscv use the contpte aware ptep_test_and_clear_young() function from
arm64.
Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com>
---
arch/arm64/include/asm/pgtable.h | 14 ++----------
arch/arm64/mm/contpte.c | 25 --------------------
arch/riscv/include/asm/pgtable.h | 12 ++++++----
arch/riscv/kvm/mmu.c | 2 +-
arch/riscv/mm/pgtable.c | 2 +-
include/linux/contpte.h | 2 ++
mm/contpte.c | 39 ++++++++++++++++++++++++++++++++
7 files changed, 53 insertions(+), 43 deletions(-)
diff --git a/arch/arm64/include/asm/pgtable.h b/arch/arm64/include/asm/pgtable.h
index ff7fe1d9cabe..9a8702d1ad00 100644
--- a/arch/arm64/include/asm/pgtable.h
+++ b/arch/arm64/include/asm/pgtable.h
@@ -1389,8 +1389,6 @@ extern void contpte_clear_full_ptes(struct mm_struct *mm, unsigned long addr,
extern pte_t contpte_get_and_clear_full_ptes(struct mm_struct *mm,
unsigned long addr, pte_t *ptep,
unsigned int nr, int full);
-extern int contpte_ptep_test_and_clear_young(struct vm_area_struct *vma,
- unsigned long addr, pte_t *ptep);
extern int contpte_ptep_clear_flush_young(struct vm_area_struct *vma,
unsigned long addr, pte_t *ptep);
extern void contpte_wrprotect_ptes(struct mm_struct *mm, unsigned long addr,
@@ -1477,16 +1475,8 @@ extern pte_t ptep_get_and_clear(struct mm_struct *mm,
unsigned long addr, pte_t *ptep);
#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
-static inline int ptep_test_and_clear_young(struct vm_area_struct *vma,
- unsigned long addr, pte_t *ptep)
-{
- pte_t orig_pte = __ptep_get(ptep);
-
- if (likely(!pte_valid_cont(orig_pte)))
- return __ptep_test_and_clear_young(vma, addr, ptep);
-
- return contpte_ptep_test_and_clear_young(vma, addr, ptep);
-}
+extern int ptep_test_and_clear_young(struct vm_area_struct *vma,
+ unsigned long addr, pte_t *ptep);
#define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
static inline int ptep_clear_flush_young(struct vm_area_struct *vma,
diff --git a/arch/arm64/mm/contpte.c b/arch/arm64/mm/contpte.c
index 5e9e40145085..9bf471633ca4 100644
--- a/arch/arm64/mm/contpte.c
+++ b/arch/arm64/mm/contpte.c
@@ -45,31 +45,6 @@ pte_t contpte_get_and_clear_full_ptes(struct mm_struct *mm,
}
EXPORT_SYMBOL_GPL(contpte_get_and_clear_full_ptes);
-int contpte_ptep_test_and_clear_young(struct vm_area_struct *vma,
- unsigned long addr, pte_t *ptep)
-{
- /*
- * ptep_clear_flush_young() technically requires us to clear the access
- * flag for a _single_ pte. However, the core-mm code actually tracks
- * access/dirty per folio, not per page. And since we only create a
- * contig range when the range is covered by a single folio, we can get
- * away with clearing young for the whole contig range here, so we avoid
- * having to unfold.
- */
-
- int young = 0;
- int i;
-
- ptep = arch_contpte_align_down(ptep);
- addr = ALIGN_DOWN(addr, CONT_PTE_SIZE);
-
- for (i = 0; i < CONT_PTES; i++, ptep++, addr += PAGE_SIZE)
- young |= __ptep_test_and_clear_young(vma, addr, ptep);
-
- return young;
-}
-EXPORT_SYMBOL_GPL(contpte_ptep_test_and_clear_young);
-
int contpte_ptep_clear_flush_young(struct vm_area_struct *vma,
unsigned long addr, pte_t *ptep)
{
diff --git a/arch/riscv/include/asm/pgtable.h b/arch/riscv/include/asm/pgtable.h
index 03cd640137ed..d39cb24c6c4a 100644
--- a/arch/riscv/include/asm/pgtable.h
+++ b/arch/riscv/include/asm/pgtable.h
@@ -739,8 +739,7 @@ static inline void __pte_clear(struct mm_struct *mm,
extern int __ptep_set_access_flags(struct vm_area_struct *vma, unsigned long address,
pte_t *ptep, pte_t entry, int dirty);
-#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG /* defined in mm/pgtable.c */
-extern int ptep_test_and_clear_young(struct vm_area_struct *vma, unsigned long address,
+extern int __ptep_test_and_clear_young(struct vm_area_struct *vma, unsigned long address,
pte_t *ptep);
static inline pte_t __ptep_get_and_clear(struct mm_struct *mm,
@@ -778,7 +777,7 @@ static inline int ptep_clear_flush_young(struct vm_area_struct *vma,
* shouldn't really matter because there's no real memory
* pressure for swapout to react to. ]
*/
- return ptep_test_and_clear_young(vma, address, ptep);
+ return __ptep_test_and_clear_young(vma, address, ptep);
}
#ifdef CONFIG_THP_CONTPTE
@@ -797,6 +796,9 @@ extern void pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep);
#define __HAVE_ARCH_PTEP_GET_AND_CLEAR
extern pte_t ptep_get_and_clear(struct mm_struct *mm,
unsigned long addr, pte_t *ptep);
+#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
+extern int ptep_test_and_clear_young(struct vm_area_struct *vma,
+ unsigned long addr, pte_t *ptep);
#else /* CONFIG_THP_CONTPTE */
@@ -806,6 +808,8 @@ extern pte_t ptep_get_and_clear(struct mm_struct *mm,
#define pte_clear __pte_clear
#define __HAVE_ARCH_PTEP_GET_AND_CLEAR
#define ptep_get_and_clear __ptep_get_and_clear
+#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
+#define ptep_test_and_clear_young __ptep_test_and_clear_young
#endif /* CONFIG_THP_CONTPTE */
@@ -987,7 +991,7 @@ static inline int pmdp_set_access_flags(struct vm_area_struct *vma,
static inline int pmdp_test_and_clear_young(struct vm_area_struct *vma,
unsigned long address, pmd_t *pmdp)
{
- return ptep_test_and_clear_young(vma, address, (pte_t *)pmdp);
+ return __ptep_test_and_clear_young(vma, address, (pte_t *)pmdp);
}
#define __HAVE_ARCH_PMDP_HUGE_GET_AND_CLEAR
diff --git a/arch/riscv/kvm/mmu.c b/arch/riscv/kvm/mmu.c
index 1ee6139d495f..554926e33760 100644
--- a/arch/riscv/kvm/mmu.c
+++ b/arch/riscv/kvm/mmu.c
@@ -585,7 +585,7 @@ bool kvm_age_gfn(struct kvm *kvm, struct kvm_gfn_range *range)
&ptep, &ptep_level))
return false;
- return ptep_test_and_clear_young(NULL, 0, ptep);
+ return __ptep_test_and_clear_young(NULL, 0, ptep);
}
bool kvm_test_age_gfn(struct kvm *kvm, struct kvm_gfn_range *range)
diff --git a/arch/riscv/mm/pgtable.c b/arch/riscv/mm/pgtable.c
index 5756bde9eb42..5f31d0594109 100644
--- a/arch/riscv/mm/pgtable.c
+++ b/arch/riscv/mm/pgtable.c
@@ -18,7 +18,7 @@ int __ptep_set_access_flags(struct vm_area_struct *vma,
return true;
}
-int ptep_test_and_clear_young(struct vm_area_struct *vma,
+int __ptep_test_and_clear_young(struct vm_area_struct *vma,
unsigned long address,
pte_t *ptep)
{
diff --git a/include/linux/contpte.h b/include/linux/contpte.h
index 01da4bfc3af6..38092adbe0d4 100644
--- a/include/linux/contpte.h
+++ b/include/linux/contpte.h
@@ -19,5 +19,7 @@ void contpte_try_unfold(struct mm_struct *mm, unsigned long addr,
pte_t *ptep, pte_t pte);
void contpte_set_ptes(struct mm_struct *mm, unsigned long addr,
pte_t *ptep, pte_t pte, unsigned int nr);
+int contpte_ptep_test_and_clear_young(struct vm_area_struct *vma,
+ unsigned long addr, pte_t *ptep);
#endif /* _LINUX_CONTPTE_H */
diff --git a/mm/contpte.c b/mm/contpte.c
index 5bf939639233..220e9d81f401 100644
--- a/mm/contpte.c
+++ b/mm/contpte.c
@@ -47,6 +47,7 @@
* - set_pte()
* - pte_clear()
* - ptep_get_and_clear()
+ * - ptep_test_and_clear_young()
*/
pte_t huge_ptep_get(pte_t *ptep)
@@ -690,4 +691,42 @@ pte_t ptep_get_and_clear(struct mm_struct *mm,
contpte_try_unfold(mm, addr, ptep, __ptep_get(ptep));
return __ptep_get_and_clear(mm, addr, ptep);
}
+
+int contpte_ptep_test_and_clear_young(struct vm_area_struct *vma,
+ unsigned long addr, pte_t *ptep)
+{
+ /*
+ * ptep_clear_flush_young() technically requires us to clear the access
+ * flag for a _single_ pte. However, the core-mm code actually tracks
+ * access/dirty per folio, not per page. And since we only create a
+ * contig range when the range is covered by a single folio, we can get
+ * away with clearing young for the whole contig range here, so we avoid
+ * having to unfold.
+ */
+
+ size_t pgsize;
+ int young = 0;
+ int i, ncontig;
+
+ ptep = arch_contpte_align_down(ptep);
+ ncontig = arch_contpte_get_num_contig(vma->vm_mm, addr, ptep, 0, &pgsize);
+ addr = ALIGN_DOWN(addr, ncontig * pgsize);
+
+ for (i = 0; i < ncontig; i++, ptep++, addr += pgsize)
+ young |= __ptep_test_and_clear_young(vma, addr, ptep);
+
+ return young;
+}
+EXPORT_SYMBOL_GPL(contpte_ptep_test_and_clear_young);
+
+__always_inline int ptep_test_and_clear_young(struct vm_area_struct *vma,
+ unsigned long addr, pte_t *ptep)
+{
+ pte_t orig_pte = __ptep_get(ptep);
+
+ if (likely(!pte_valid_cont(orig_pte)))
+ return __ptep_test_and_clear_young(vma, addr, ptep);
+
+ return contpte_ptep_test_and_clear_young(vma, addr, ptep);
+}
#endif /* CONFIG_THP_CONTPTE */
--
2.39.2
^ permalink raw reply related [flat|nested] 16+ messages in thread
* [PATCH 09/12] mm, riscv, arm64: Use common ptep_clear_flush_young() function
2024-05-08 19:19 [PATCH 00/12] Make riscv use THP contpte support for arm64 Alexandre Ghiti
` (7 preceding siblings ...)
2024-05-08 19:19 ` [PATCH 08/12] mm, riscv, arm64: Use common ptep_test_and_clear_young() function Alexandre Ghiti
@ 2024-05-08 19:19 ` Alexandre Ghiti
2024-05-08 19:19 ` [PATCH 10/12] mm, riscv, arm64: Use common ptep_set_access_flags() function Alexandre Ghiti
` (2 subsequent siblings)
11 siblings, 0 replies; 16+ messages in thread
From: Alexandre Ghiti @ 2024-05-08 19:19 UTC (permalink / raw
To: Ryan Roberts, Catalin Marinas, Will Deacon, Alexander Potapenko,
Marco Elver, Dmitry Vyukov, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Ard Biesheuvel, Anup Patel, Atish Patra,
Andrey Ryabinin, Andrey Konovalov, Vincenzo Frascino,
Andrew Morton, linux-arm-kernel, linux-kernel, kasan-dev,
linux-riscv, linux-efi, kvm, kvm-riscv, linux-mm
Cc: Alexandre Ghiti
Make riscv use the contpte aware ptep_clear_flush_young() function from
arm64.
Note that riscv used to not flush the tlb after clearing the accessed
bit, which it does now: this will be improved when we implement svinval
support.
Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com>
---
arch/arm64/include/asm/pgtable.h | 22 ++++++++----------
arch/arm64/mm/contpte.c | 21 -----------------
arch/riscv/include/asm/pgtable.h | 12 +++++++---
include/linux/contpte.h | 2 ++
mm/contpte.c | 40 ++++++++++++++++++++++++++++++++
5 files changed, 61 insertions(+), 36 deletions(-)
diff --git a/arch/arm64/include/asm/pgtable.h b/arch/arm64/include/asm/pgtable.h
index 9a8702d1ad00..92c12fb85cb4 100644
--- a/arch/arm64/include/asm/pgtable.h
+++ b/arch/arm64/include/asm/pgtable.h
@@ -1389,8 +1389,6 @@ extern void contpte_clear_full_ptes(struct mm_struct *mm, unsigned long addr,
extern pte_t contpte_get_and_clear_full_ptes(struct mm_struct *mm,
unsigned long addr, pte_t *ptep,
unsigned int nr, int full);
-extern int contpte_ptep_clear_flush_young(struct vm_area_struct *vma,
- unsigned long addr, pte_t *ptep);
extern void contpte_wrprotect_ptes(struct mm_struct *mm, unsigned long addr,
pte_t *ptep, unsigned int nr);
extern int contpte_ptep_set_access_flags(struct vm_area_struct *vma,
@@ -1479,16 +1477,8 @@ extern int ptep_test_and_clear_young(struct vm_area_struct *vma,
unsigned long addr, pte_t *ptep);
#define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
-static inline int ptep_clear_flush_young(struct vm_area_struct *vma,
- unsigned long addr, pte_t *ptep)
-{
- pte_t orig_pte = __ptep_get(ptep);
-
- if (likely(!pte_valid_cont(orig_pte)))
- return __ptep_clear_flush_young(vma, addr, ptep);
-
- return contpte_ptep_clear_flush_young(vma, addr, ptep);
-}
+extern int ptep_clear_flush_young(struct vm_area_struct *vma,
+ unsigned long addr, pte_t *ptep);
#define wrprotect_ptes wrprotect_ptes
static __always_inline void wrprotect_ptes(struct mm_struct *mm,
@@ -1616,6 +1606,14 @@ static inline void arch_contpte_flush_tlb_range(struct vm_area_struct *vma,
__flush_tlb_range(vma, start, end, stride, true, 3);
}
+static inline void arch_contpte_flush_tlb_range_nosync(struct vm_area_struct *vma,
+ unsigned long start,
+ unsigned long end,
+ unsigned long stride)
+{
+ __flush_tlb_range_nosync(vma, start, end, stride, true, 3);
+}
+
static inline int arch_contpte_get_first_ncontig(size_t *pgsize)
{
if (pgsize)
diff --git a/arch/arm64/mm/contpte.c b/arch/arm64/mm/contpte.c
index 9bf471633ca4..16940511943c 100644
--- a/arch/arm64/mm/contpte.c
+++ b/arch/arm64/mm/contpte.c
@@ -45,27 +45,6 @@ pte_t contpte_get_and_clear_full_ptes(struct mm_struct *mm,
}
EXPORT_SYMBOL_GPL(contpte_get_and_clear_full_ptes);
-int contpte_ptep_clear_flush_young(struct vm_area_struct *vma,
- unsigned long addr, pte_t *ptep)
-{
- int young;
-
- young = contpte_ptep_test_and_clear_young(vma, addr, ptep);
-
- if (young) {
- /*
- * See comment in __ptep_clear_flush_young(); same rationale for
- * eliding the trailing DSB applies here.
- */
- addr = ALIGN_DOWN(addr, CONT_PTE_SIZE);
- __flush_tlb_range_nosync(vma, addr, addr + CONT_PTE_SIZE,
- PAGE_SIZE, true, 3);
- }
-
- return young;
-}
-EXPORT_SYMBOL_GPL(contpte_ptep_clear_flush_young);
-
void contpte_wrprotect_ptes(struct mm_struct *mm, unsigned long addr,
pte_t *ptep, unsigned int nr)
{
diff --git a/arch/riscv/include/asm/pgtable.h b/arch/riscv/include/asm/pgtable.h
index d39cb24c6c4a..42c7884b8d2e 100644
--- a/arch/riscv/include/asm/pgtable.h
+++ b/arch/riscv/include/asm/pgtable.h
@@ -615,6 +615,8 @@ static inline void arch_contpte_flush_tlb_range(struct vm_area_struct *vma,
flush_tlb_mm_range(vma->vm_mm, start, end, stride);
}
+#define arch_contpte_flush_tlb_range_nosync arch_contpte_flush_tlb_range
+
static inline int arch_contpte_get_first_ncontig(size_t *pgsize)
{
if (pgsize)
@@ -758,9 +760,8 @@ static inline void __ptep_set_wrprotect(struct mm_struct *mm,
atomic_long_and(~(unsigned long)_PAGE_WRITE, (atomic_long_t *)ptep);
}
-#define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
-static inline int ptep_clear_flush_young(struct vm_area_struct *vma,
- unsigned long address, pte_t *ptep)
+static inline int __ptep_clear_flush_young(struct vm_area_struct *vma,
+ unsigned long address, pte_t *ptep)
{
/*
* This comment is borrowed from x86, but applies equally to RISC-V:
@@ -799,6 +800,9 @@ extern pte_t ptep_get_and_clear(struct mm_struct *mm,
#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
extern int ptep_test_and_clear_young(struct vm_area_struct *vma,
unsigned long addr, pte_t *ptep);
+#define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
+extern int ptep_clear_flush_young(struct vm_area_struct *vma,
+ unsigned long addr, pte_t *ptep);
#else /* CONFIG_THP_CONTPTE */
@@ -810,6 +814,8 @@ extern int ptep_test_and_clear_young(struct vm_area_struct *vma,
#define ptep_get_and_clear __ptep_get_and_clear
#define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
#define ptep_test_and_clear_young __ptep_test_and_clear_young
+#define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
+#define ptep_clear_flush_young __ptep_clear_flush_young
#endif /* CONFIG_THP_CONTPTE */
diff --git a/include/linux/contpte.h b/include/linux/contpte.h
index 38092adbe0d4..76a49ac8b6f5 100644
--- a/include/linux/contpte.h
+++ b/include/linux/contpte.h
@@ -21,5 +21,7 @@ void contpte_set_ptes(struct mm_struct *mm, unsigned long addr,
pte_t *ptep, pte_t pte, unsigned int nr);
int contpte_ptep_test_and_clear_young(struct vm_area_struct *vma,
unsigned long addr, pte_t *ptep);
+int contpte_ptep_clear_flush_young(struct vm_area_struct *vma,
+ unsigned long addr, pte_t *ptep);
#endif /* _LINUX_CONTPTE_H */
diff --git a/mm/contpte.c b/mm/contpte.c
index 220e9d81f401..600277b1196c 100644
--- a/mm/contpte.c
+++ b/mm/contpte.c
@@ -48,6 +48,7 @@
* - pte_clear()
* - ptep_get_and_clear()
* - ptep_test_and_clear_young()
+ * - ptep_clear_flush_young()
*/
pte_t huge_ptep_get(pte_t *ptep)
@@ -729,4 +730,43 @@ __always_inline int ptep_test_and_clear_young(struct vm_area_struct *vma,
return contpte_ptep_test_and_clear_young(vma, addr, ptep);
}
+
+int contpte_ptep_clear_flush_young(struct vm_area_struct *vma,
+ unsigned long addr, pte_t *ptep)
+{
+ int young;
+
+ young = contpte_ptep_test_and_clear_young(vma, addr, ptep);
+
+ if (young) {
+ /*
+ * See comment in __ptep_clear_flush_young(); same rationale for
+ * eliding the trailing DSB applies here.
+ */
+ size_t pgsize;
+ int ncontig;
+
+ ncontig = arch_contpte_get_num_contig(vma->vm_mm, addr, ptep,
+ 0, &pgsize);
+
+ addr = ALIGN_DOWN(addr, ncontig * pgsize);
+ arch_contpte_flush_tlb_range_nosync(vma, addr,
+ addr + ncontig * pgsize,
+ pgsize);
+ }
+
+ return young;
+}
+EXPORT_SYMBOL_GPL(contpte_ptep_clear_flush_young);
+
+__always_inline int ptep_clear_flush_young(struct vm_area_struct *vma,
+ unsigned long addr, pte_t *ptep)
+{
+ pte_t orig_pte = __ptep_get(ptep);
+
+ if (likely(!pte_valid_cont(orig_pte)))
+ return __ptep_clear_flush_young(vma, addr, ptep);
+
+ return contpte_ptep_clear_flush_young(vma, addr, ptep);
+}
#endif /* CONFIG_THP_CONTPTE */
--
2.39.2
^ permalink raw reply related [flat|nested] 16+ messages in thread
* [PATCH 10/12] mm, riscv, arm64: Use common ptep_set_access_flags() function
2024-05-08 19:19 [PATCH 00/12] Make riscv use THP contpte support for arm64 Alexandre Ghiti
` (8 preceding siblings ...)
2024-05-08 19:19 ` [PATCH 09/12] mm, riscv, arm64: Use common ptep_clear_flush_young() function Alexandre Ghiti
@ 2024-05-08 19:19 ` Alexandre Ghiti
2024-05-08 19:19 ` [PATCH 11/12] mm, riscv, arm64: Use common ptep_set_wrprotect()/wrprotect_ptes() functions Alexandre Ghiti
2024-05-08 19:19 ` [PATCH 12/12] mm, riscv, arm64: Use common get_and_clear_full_ptes()/clear_full_ptes() functions Alexandre Ghiti
11 siblings, 0 replies; 16+ messages in thread
From: Alexandre Ghiti @ 2024-05-08 19:19 UTC (permalink / raw
To: Ryan Roberts, Catalin Marinas, Will Deacon, Alexander Potapenko,
Marco Elver, Dmitry Vyukov, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Ard Biesheuvel, Anup Patel, Atish Patra,
Andrey Ryabinin, Andrey Konovalov, Vincenzo Frascino,
Andrew Morton, linux-arm-kernel, linux-kernel, kasan-dev,
linux-riscv, linux-efi, kvm, kvm-riscv, linux-mm
Cc: Alexandre Ghiti
Make riscv use the contpte aware ptep_set_access_flags() function from
arm64.
Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com>
---
arch/arm64/include/asm/pgtable.h | 19 ++--------
arch/arm64/mm/contpte.c | 46 -----------------------
arch/riscv/include/asm/pgtable.h | 10 +++--
include/linux/contpte.h | 3 ++
mm/contpte.c | 63 ++++++++++++++++++++++++++++++++
5 files changed, 76 insertions(+), 65 deletions(-)
diff --git a/arch/arm64/include/asm/pgtable.h b/arch/arm64/include/asm/pgtable.h
index 92c12fb85cb4..6591aab11c67 100644
--- a/arch/arm64/include/asm/pgtable.h
+++ b/arch/arm64/include/asm/pgtable.h
@@ -1391,9 +1391,6 @@ extern pte_t contpte_get_and_clear_full_ptes(struct mm_struct *mm,
unsigned int nr, int full);
extern void contpte_wrprotect_ptes(struct mm_struct *mm, unsigned long addr,
pte_t *ptep, unsigned int nr);
-extern int contpte_ptep_set_access_flags(struct vm_area_struct *vma,
- unsigned long addr, pte_t *ptep,
- pte_t entry, int dirty);
#define pte_batch_hint pte_batch_hint
static inline unsigned int pte_batch_hint(pte_t *ptep, pte_t pte)
@@ -1512,19 +1509,9 @@ static inline void ptep_set_wrprotect(struct mm_struct *mm,
}
#define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
-static inline int ptep_set_access_flags(struct vm_area_struct *vma,
- unsigned long addr, pte_t *ptep,
- pte_t entry, int dirty)
-{
- pte_t orig_pte = __ptep_get(ptep);
-
- entry = pte_mknoncont(entry);
-
- if (likely(!pte_valid_cont(orig_pte)))
- return __ptep_set_access_flags(vma, addr, ptep, entry, dirty);
-
- return contpte_ptep_set_access_flags(vma, addr, ptep, entry, dirty);
-}
+extern int ptep_set_access_flags(struct vm_area_struct *vma,
+ unsigned long addr, pte_t *ptep,
+ pte_t entry, int dirty);
#else /* CONFIG_THP_CONTPTE */
diff --git a/arch/arm64/mm/contpte.c b/arch/arm64/mm/contpte.c
index 16940511943c..5675a61452ac 100644
--- a/arch/arm64/mm/contpte.c
+++ b/arch/arm64/mm/contpte.c
@@ -62,49 +62,3 @@ void contpte_wrprotect_ptes(struct mm_struct *mm, unsigned long addr,
__wrprotect_ptes(mm, addr, ptep, nr);
}
EXPORT_SYMBOL_GPL(contpte_wrprotect_ptes);
-
-int contpte_ptep_set_access_flags(struct vm_area_struct *vma,
- unsigned long addr, pte_t *ptep,
- pte_t entry, int dirty)
-{
- unsigned long start_addr;
- pte_t orig_pte;
- int i;
-
- /*
- * Gather the access/dirty bits for the contiguous range. If nothing has
- * changed, its a noop.
- */
- orig_pte = pte_mknoncont(ptep_get(ptep));
- if (pte_val(orig_pte) == pte_val(entry))
- return 0;
-
- /*
- * We can fix up access/dirty bits without having to unfold the contig
- * range. But if the write bit is changing, we must unfold.
- */
- if (pte_write(orig_pte) == pte_write(entry)) {
- /*
- * For HW access management, we technically only need to update
- * the flag on a single pte in the range. But for SW access
- * management, we need to update all the ptes to prevent extra
- * faults. Avoid per-page tlb flush in __ptep_set_access_flags()
- * and instead flush the whole range at the end.
- */
- ptep = arch_contpte_align_down(ptep);
- start_addr = addr = ALIGN_DOWN(addr, CONT_PTE_SIZE);
-
- for (i = 0; i < CONT_PTES; i++, ptep++, addr += PAGE_SIZE)
- __ptep_set_access_flags(vma, addr, ptep, entry, 0);
-
- if (dirty)
- __flush_tlb_range(vma, start_addr, addr,
- PAGE_SIZE, true, 3);
- } else {
- __contpte_try_unfold(vma->vm_mm, addr, ptep, orig_pte);
- __ptep_set_access_flags(vma, addr, ptep, entry, dirty);
- }
-
- return 1;
-}
-EXPORT_SYMBOL_GPL(contpte_ptep_set_access_flags);
diff --git a/arch/riscv/include/asm/pgtable.h b/arch/riscv/include/asm/pgtable.h
index 42c7884b8d2e..b151a5aa4de8 100644
--- a/arch/riscv/include/asm/pgtable.h
+++ b/arch/riscv/include/asm/pgtable.h
@@ -803,6 +803,10 @@ extern int ptep_test_and_clear_young(struct vm_area_struct *vma,
#define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
extern int ptep_clear_flush_young(struct vm_area_struct *vma,
unsigned long addr, pte_t *ptep);
+#define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
+extern int ptep_set_access_flags(struct vm_area_struct *vma,
+ unsigned long address, pte_t *ptep,
+ pte_t entry, int dirty);
#else /* CONFIG_THP_CONTPTE */
@@ -816,11 +820,11 @@ extern int ptep_clear_flush_young(struct vm_area_struct *vma,
#define ptep_test_and_clear_young __ptep_test_and_clear_young
#define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
#define ptep_clear_flush_young __ptep_clear_flush_young
+#define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
+#define ptep_set_access_flags __ptep_set_access_flags
#endif /* CONFIG_THP_CONTPTE */
-#define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
-#define ptep_set_access_flags __ptep_set_access_flags
#define __HAVE_ARCH_PTEP_SET_WRPROTECT
#define ptep_set_wrprotect __ptep_set_wrprotect
@@ -990,7 +994,7 @@ static inline int pmdp_set_access_flags(struct vm_area_struct *vma,
unsigned long address, pmd_t *pmdp,
pmd_t entry, int dirty)
{
- return ptep_set_access_flags(vma, address, (pte_t *)pmdp, pmd_pte(entry), dirty);
+ return __ptep_set_access_flags(vma, address, (pte_t *)pmdp, pmd_pte(entry), dirty);
}
#define __HAVE_ARCH_PMDP_TEST_AND_CLEAR_YOUNG
diff --git a/include/linux/contpte.h b/include/linux/contpte.h
index 76a49ac8b6f5..76244b0c678a 100644
--- a/include/linux/contpte.h
+++ b/include/linux/contpte.h
@@ -23,5 +23,8 @@ int contpte_ptep_test_and_clear_young(struct vm_area_struct *vma,
unsigned long addr, pte_t *ptep);
int contpte_ptep_clear_flush_young(struct vm_area_struct *vma,
unsigned long addr, pte_t *ptep);
+int contpte_ptep_set_access_flags(struct vm_area_struct *vma,
+ unsigned long addr, pte_t *ptep,
+ pte_t entry, int dirty);
#endif /* _LINUX_CONTPTE_H */
diff --git a/mm/contpte.c b/mm/contpte.c
index 600277b1196c..9cbbff1f67ad 100644
--- a/mm/contpte.c
+++ b/mm/contpte.c
@@ -769,4 +769,67 @@ __always_inline int ptep_clear_flush_young(struct vm_area_struct *vma,
return contpte_ptep_clear_flush_young(vma, addr, ptep);
}
+
+int contpte_ptep_set_access_flags(struct vm_area_struct *vma,
+ unsigned long addr, pte_t *ptep,
+ pte_t entry, int dirty)
+{
+ unsigned long start_addr;
+ pte_t orig_pte;
+ int i;
+
+ /*
+ * Gather the access/dirty bits for the contiguous range. If nothing has
+ * changed, its a noop.
+ */
+ orig_pte = pte_mknoncont(ptep_get(ptep));
+ if (pte_val(orig_pte) == pte_val(entry))
+ return 0;
+
+ /*
+ * We can fix up access/dirty bits without having to unfold the contig
+ * range. But if the write bit is changing, we must unfold.
+ */
+ if (pte_write(orig_pte) == pte_write(entry)) {
+ /*
+ * For HW access management, we technically only need to update
+ * the flag on a single pte in the range. But for SW access
+ * management, we need to update all the ptes to prevent extra
+ * faults. Avoid per-page tlb flush in __ptep_set_access_flags()
+ * and instead flush the whole range at the end.
+ */
+ size_t pgsize;
+ int ncontig;
+
+ ptep = arch_contpte_align_down(ptep);
+ ncontig = arch_contpte_get_num_contig(vma->vm_mm, addr, ptep, 0, &pgsize);
+ start_addr = addr = ALIGN_DOWN(addr, ncontig * pgsize);
+
+ for (i = 0; i < ncontig; i++, ptep++, addr += pgsize)
+ __ptep_set_access_flags(vma, addr, ptep, entry, 0);
+
+ if (dirty)
+ arch_contpte_flush_tlb_range(vma, start_addr, addr, pgsize);
+ } else {
+ __contpte_try_unfold(vma->vm_mm, addr, ptep, orig_pte);
+ __ptep_set_access_flags(vma, addr, ptep, entry, dirty);
+ }
+
+ return 1;
+}
+EXPORT_SYMBOL_GPL(contpte_ptep_set_access_flags);
+
+__always_inline int ptep_set_access_flags(struct vm_area_struct *vma,
+ unsigned long addr, pte_t *ptep,
+ pte_t entry, int dirty)
+{
+ pte_t orig_pte = __ptep_get(ptep);
+
+ entry = pte_mknoncont(entry);
+
+ if (likely(!pte_valid_cont(orig_pte)))
+ return __ptep_set_access_flags(vma, addr, ptep, entry, dirty);
+
+ return contpte_ptep_set_access_flags(vma, addr, ptep, entry, dirty);
+}
#endif /* CONFIG_THP_CONTPTE */
--
2.39.2
^ permalink raw reply related [flat|nested] 16+ messages in thread
* [PATCH 11/12] mm, riscv, arm64: Use common ptep_set_wrprotect()/wrprotect_ptes() functions
2024-05-08 19:19 [PATCH 00/12] Make riscv use THP contpte support for arm64 Alexandre Ghiti
` (9 preceding siblings ...)
2024-05-08 19:19 ` [PATCH 10/12] mm, riscv, arm64: Use common ptep_set_access_flags() function Alexandre Ghiti
@ 2024-05-08 19:19 ` Alexandre Ghiti
2024-05-08 19:19 ` [PATCH 12/12] mm, riscv, arm64: Use common get_and_clear_full_ptes()/clear_full_ptes() functions Alexandre Ghiti
11 siblings, 0 replies; 16+ messages in thread
From: Alexandre Ghiti @ 2024-05-08 19:19 UTC (permalink / raw
To: Ryan Roberts, Catalin Marinas, Will Deacon, Alexander Potapenko,
Marco Elver, Dmitry Vyukov, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Ard Biesheuvel, Anup Patel, Atish Patra,
Andrey Ryabinin, Andrey Konovalov, Vincenzo Frascino,
Andrew Morton, linux-arm-kernel, linux-kernel, kasan-dev,
linux-riscv, linux-efi, kvm, kvm-riscv, linux-mm
Cc: Alexandre Ghiti
Make riscv use the contpte aware ptep_set_wrprotect()/wrprotect_ptes()
function from arm64.
Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com>
---
arch/arm64/include/asm/pgtable.h | 56 ++++++------------------
arch/arm64/mm/contpte.c | 18 --------
arch/riscv/include/asm/pgtable.h | 25 +++++++++--
include/linux/contpte.h | 2 +
mm/contpte.c | 75 +++++++++++++++++++++++++++++++-
5 files changed, 110 insertions(+), 66 deletions(-)
diff --git a/arch/arm64/include/asm/pgtable.h b/arch/arm64/include/asm/pgtable.h
index 6591aab11c67..162efd9647dd 100644
--- a/arch/arm64/include/asm/pgtable.h
+++ b/arch/arm64/include/asm/pgtable.h
@@ -1208,7 +1208,11 @@ static inline pmd_t pmdp_huge_get_and_clear(struct mm_struct *mm,
}
#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
-static inline void ___ptep_set_wrprotect(struct mm_struct *mm,
+/*
+ * __ptep_set_wrprotect - mark read-only while trasferring potential hardware
+ * dirty status (PTE_DBM && !PTE_RDONLY) to the software PTE_DIRTY bit.
+ */
+static inline void __ptep_set_wrprotect(struct mm_struct *mm,
unsigned long address, pte_t *ptep,
pte_t pte)
{
@@ -1222,23 +1226,13 @@ static inline void ___ptep_set_wrprotect(struct mm_struct *mm,
} while (pte_val(pte) != pte_val(old_pte));
}
-/*
- * __ptep_set_wrprotect - mark read-only while trasferring potential hardware
- * dirty status (PTE_DBM && !PTE_RDONLY) to the software PTE_DIRTY bit.
- */
-static inline void __ptep_set_wrprotect(struct mm_struct *mm,
- unsigned long address, pte_t *ptep)
-{
- ___ptep_set_wrprotect(mm, address, ptep, __ptep_get(ptep));
-}
-
static inline void __wrprotect_ptes(struct mm_struct *mm, unsigned long address,
pte_t *ptep, unsigned int nr)
{
unsigned int i;
for (i = 0; i < nr; i++, address += PAGE_SIZE, ptep++)
- __ptep_set_wrprotect(mm, address, ptep);
+ __ptep_set_wrprotect(mm, address, ptep, __ptep_get(ptep));
}
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
@@ -1246,7 +1240,7 @@ static inline void __wrprotect_ptes(struct mm_struct *mm, unsigned long address,
static inline void pmdp_set_wrprotect(struct mm_struct *mm,
unsigned long address, pmd_t *pmdp)
{
- __ptep_set_wrprotect(mm, address, (pte_t *)pmdp);
+ __ptep_set_wrprotect(mm, address, (pte_t *)pmdp, __ptep_get((pte_t *)pmdp));
}
#define pmdp_establish pmdp_establish
@@ -1389,8 +1383,6 @@ extern void contpte_clear_full_ptes(struct mm_struct *mm, unsigned long addr,
extern pte_t contpte_get_and_clear_full_ptes(struct mm_struct *mm,
unsigned long addr, pte_t *ptep,
unsigned int nr, int full);
-extern void contpte_wrprotect_ptes(struct mm_struct *mm, unsigned long addr,
- pte_t *ptep, unsigned int nr);
#define pte_batch_hint pte_batch_hint
static inline unsigned int pte_batch_hint(pte_t *ptep, pte_t pte)
@@ -1478,35 +1470,12 @@ extern int ptep_clear_flush_young(struct vm_area_struct *vma,
unsigned long addr, pte_t *ptep);
#define wrprotect_ptes wrprotect_ptes
-static __always_inline void wrprotect_ptes(struct mm_struct *mm,
- unsigned long addr, pte_t *ptep, unsigned int nr)
-{
- if (likely(nr == 1)) {
- /*
- * Optimization: wrprotect_ptes() can only be called for present
- * ptes so we only need to check contig bit as condition for
- * unfold, and we can remove the contig bit from the pte we read
- * to avoid re-reading. This speeds up fork() which is sensitive
- * for order-0 folios. Equivalent to contpte_try_unfold().
- */
- pte_t orig_pte = __ptep_get(ptep);
-
- if (unlikely(pte_cont(orig_pte))) {
- __contpte_try_unfold(mm, addr, ptep, orig_pte);
- orig_pte = pte_mknoncont(orig_pte);
- }
- ___ptep_set_wrprotect(mm, addr, ptep, orig_pte);
- } else {
- contpte_wrprotect_ptes(mm, addr, ptep, nr);
- }
-}
+extern void wrprotect_ptes(struct mm_struct *mm,
+ unsigned long addr, pte_t *ptep, unsigned int nr);
#define __HAVE_ARCH_PTEP_SET_WRPROTECT
-static inline void ptep_set_wrprotect(struct mm_struct *mm,
- unsigned long addr, pte_t *ptep)
-{
- wrprotect_ptes(mm, addr, ptep, 1);
-}
+extern void ptep_set_wrprotect(struct mm_struct *mm,
+ unsigned long addr, pte_t *ptep);
#define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
extern int ptep_set_access_flags(struct vm_area_struct *vma,
@@ -1528,7 +1497,8 @@ extern int ptep_set_access_flags(struct vm_area_struct *vma,
#define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
#define ptep_clear_flush_young __ptep_clear_flush_young
#define __HAVE_ARCH_PTEP_SET_WRPROTECT
-#define ptep_set_wrprotect __ptep_set_wrprotect
+#define ptep_set_wrprotect(mm, addr, ptep) \
+ __ptep_set_wrprotect(mm, addr, ptep, __ptep_get(ptep))
#define wrprotect_ptes __wrprotect_ptes
#define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
#define ptep_set_access_flags __ptep_set_access_flags
diff --git a/arch/arm64/mm/contpte.c b/arch/arm64/mm/contpte.c
index 5675a61452ac..1cef93b15d6e 100644
--- a/arch/arm64/mm/contpte.c
+++ b/arch/arm64/mm/contpte.c
@@ -44,21 +44,3 @@ pte_t contpte_get_and_clear_full_ptes(struct mm_struct *mm,
return __get_and_clear_full_ptes(mm, addr, ptep, nr, full);
}
EXPORT_SYMBOL_GPL(contpte_get_and_clear_full_ptes);
-
-void contpte_wrprotect_ptes(struct mm_struct *mm, unsigned long addr,
- pte_t *ptep, unsigned int nr)
-{
- /*
- * If wrprotecting an entire contig range, we can avoid unfolding. Just
- * set wrprotect and wait for the later mmu_gather flush to invalidate
- * the tlb. Until the flush, the page may or may not be wrprotected.
- * After the flush, it is guaranteed wrprotected. If it's a partial
- * range though, we must unfold, because we can't have a case where
- * CONT_PTE is set but wrprotect applies to a subset of the PTEs; this
- * would cause it to continue to be unpredictable after the flush.
- */
-
- contpte_try_unfold_partial(mm, addr, ptep, nr);
- __wrprotect_ptes(mm, addr, ptep, nr);
-}
-EXPORT_SYMBOL_GPL(contpte_wrprotect_ptes);
diff --git a/arch/riscv/include/asm/pgtable.h b/arch/riscv/include/asm/pgtable.h
index b151a5aa4de8..728f31da5e6a 100644
--- a/arch/riscv/include/asm/pgtable.h
+++ b/arch/riscv/include/asm/pgtable.h
@@ -755,11 +755,21 @@ static inline pte_t __ptep_get_and_clear(struct mm_struct *mm,
}
static inline void __ptep_set_wrprotect(struct mm_struct *mm,
- unsigned long address, pte_t *ptep)
+ unsigned long address, pte_t *ptep,
+ pte_t pte)
{
atomic_long_and(~(unsigned long)_PAGE_WRITE, (atomic_long_t *)ptep);
}
+static inline void __wrprotect_ptes(struct mm_struct *mm, unsigned long address,
+ pte_t *ptep, unsigned int nr)
+{
+ unsigned int i;
+
+ for (i = 0; i < nr; i++, address += PAGE_SIZE, ptep++)
+ __ptep_set_wrprotect(mm, address, ptep, __ptep_get(ptep));
+}
+
static inline int __ptep_clear_flush_young(struct vm_area_struct *vma,
unsigned long address, pte_t *ptep)
{
@@ -807,6 +817,12 @@ extern int ptep_clear_flush_young(struct vm_area_struct *vma,
extern int ptep_set_access_flags(struct vm_area_struct *vma,
unsigned long address, pte_t *ptep,
pte_t entry, int dirty);
+#define __HAVE_ARCH_PTEP_SET_WRPROTECT
+extern void ptep_set_wrprotect(struct mm_struct *mm,
+ unsigned long addr, pte_t *ptep);
+extern void wrprotect_ptes(struct mm_struct *mm, unsigned long addr,
+ pte_t *ptep, unsigned int nr);
+#define wrprotect_ptes wrprotect_ptes
#else /* CONFIG_THP_CONTPTE */
@@ -822,12 +838,13 @@ extern int ptep_set_access_flags(struct vm_area_struct *vma,
#define ptep_clear_flush_young __ptep_clear_flush_young
#define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
#define ptep_set_access_flags __ptep_set_access_flags
+#define __HAVE_ARCH_PTEP_SET_WRPROTECT
+#define ptep_set_wrprotect(mm, addr, ptep) \
+ __ptep_set_wrprotect(mm, addr, ptep, __ptep_get(ptep))
+#define wrprotect_ptes __wrprotect_ptes
#endif /* CONFIG_THP_CONTPTE */
-#define __HAVE_ARCH_PTEP_SET_WRPROTECT
-#define ptep_set_wrprotect __ptep_set_wrprotect
-
#define pgprot_nx pgprot_nx
static inline pgprot_t pgprot_nx(pgprot_t _prot)
{
diff --git a/include/linux/contpte.h b/include/linux/contpte.h
index 76244b0c678a..d1439db1706c 100644
--- a/include/linux/contpte.h
+++ b/include/linux/contpte.h
@@ -26,5 +26,7 @@ int contpte_ptep_clear_flush_young(struct vm_area_struct *vma,
int contpte_ptep_set_access_flags(struct vm_area_struct *vma,
unsigned long addr, pte_t *ptep,
pte_t entry, int dirty);
+void contpte_wrprotect_ptes(struct mm_struct *mm, unsigned long addr,
+ pte_t *ptep, unsigned int nr);
#endif /* _LINUX_CONTPTE_H */
diff --git a/mm/contpte.c b/mm/contpte.c
index 9cbbff1f67ad..fe36b6b1d20a 100644
--- a/mm/contpte.c
+++ b/mm/contpte.c
@@ -49,6 +49,8 @@
* - ptep_get_and_clear()
* - ptep_test_and_clear_young()
* - ptep_clear_flush_young()
+ * - wrprotect_ptes()
+ * - ptep_set_wrprotect()
*/
pte_t huge_ptep_get(pte_t *ptep)
@@ -266,7 +268,7 @@ void huge_ptep_set_wrprotect(struct mm_struct *mm,
pte_t pte;
if (!pte_cont(__ptep_get(ptep))) {
- __ptep_set_wrprotect(mm, addr, ptep);
+ __ptep_set_wrprotect(mm, addr, ptep, __ptep_get(ptep));
return;
}
@@ -832,4 +834,75 @@ __always_inline int ptep_set_access_flags(struct vm_area_struct *vma,
return contpte_ptep_set_access_flags(vma, addr, ptep, entry, dirty);
}
+
+static void contpte_try_unfold_partial(struct mm_struct *mm, unsigned long addr,
+ pte_t *ptep, unsigned int nr)
+{
+ /*
+ * Unfold any partially covered contpte block at the beginning and end
+ * of the range.
+ */
+ size_t pgsize;
+ int ncontig;
+
+ ncontig = arch_contpte_get_num_contig(mm, addr, ptep, 0, &pgsize);
+
+ if (ptep != arch_contpte_align_down(ptep) || nr < ncontig)
+ contpte_try_unfold(mm, addr, ptep, __ptep_get(ptep));
+
+ if (ptep + nr != arch_contpte_align_down(ptep + nr)) {
+ unsigned long last_addr = addr + pgsize * (nr - 1);
+ pte_t *last_ptep = ptep + nr - 1;
+
+ contpte_try_unfold(mm, last_addr, last_ptep,
+ __ptep_get(last_ptep));
+ }
+}
+
+void contpte_wrprotect_ptes(struct mm_struct *mm, unsigned long addr,
+ pte_t *ptep, unsigned int nr)
+{
+ /*
+ * If wrprotecting an entire contig range, we can avoid unfolding. Just
+ * set wrprotect and wait for the later mmu_gather flush to invalidate
+ * the tlb. Until the flush, the page may or may not be wrprotected.
+ * After the flush, it is guaranteed wrprotected. If it's a partial
+ * range though, we must unfold, because we can't have a case where
+ * CONT_PTE is set but wrprotect applies to a subset of the PTEs; this
+ * would cause it to continue to be unpredictable after the flush.
+ */
+
+ contpte_try_unfold_partial(mm, addr, ptep, nr);
+ __wrprotect_ptes(mm, addr, ptep, nr);
+}
+EXPORT_SYMBOL_GPL(contpte_wrprotect_ptes);
+
+__always_inline void wrprotect_ptes(struct mm_struct *mm, unsigned long addr,
+ pte_t *ptep, unsigned int nr)
+{
+ if (likely(nr == 1)) {
+ /*
+ * Optimization: wrprotect_ptes() can only be called for present
+ * ptes so we only need to check contig bit as condition for
+ * unfold, and we can remove the contig bit from the pte we read
+ * to avoid re-reading. This speeds up fork() which is sensitive
+ * for order-0 folios. Equivalent to contpte_try_unfold().
+ */
+ pte_t orig_pte = __ptep_get(ptep);
+
+ if (unlikely(pte_cont(orig_pte))) {
+ __contpte_try_unfold(mm, addr, ptep, orig_pte);
+ orig_pte = pte_mknoncont(orig_pte);
+ }
+ __ptep_set_wrprotect(mm, addr, ptep, orig_pte);
+ } else {
+ contpte_wrprotect_ptes(mm, addr, ptep, nr);
+ }
+}
+
+__always_inline void ptep_set_wrprotect(struct mm_struct *mm,
+ unsigned long addr, pte_t *ptep)
+{
+ wrprotect_ptes(mm, addr, ptep, 1);
+}
#endif /* CONFIG_THP_CONTPTE */
--
2.39.2
^ permalink raw reply related [flat|nested] 16+ messages in thread
* [PATCH 12/12] mm, riscv, arm64: Use common get_and_clear_full_ptes()/clear_full_ptes() functions
2024-05-08 19:19 [PATCH 00/12] Make riscv use THP contpte support for arm64 Alexandre Ghiti
` (10 preceding siblings ...)
2024-05-08 19:19 ` [PATCH 11/12] mm, riscv, arm64: Use common ptep_set_wrprotect()/wrprotect_ptes() functions Alexandre Ghiti
@ 2024-05-08 19:19 ` Alexandre Ghiti
11 siblings, 0 replies; 16+ messages in thread
From: Alexandre Ghiti @ 2024-05-08 19:19 UTC (permalink / raw
To: Ryan Roberts, Catalin Marinas, Will Deacon, Alexander Potapenko,
Marco Elver, Dmitry Vyukov, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Ard Biesheuvel, Anup Patel, Atish Patra,
Andrey Ryabinin, Andrey Konovalov, Vincenzo Frascino,
Andrew Morton, linux-arm-kernel, linux-kernel, kasan-dev,
linux-riscv, linux-efi, kvm, kvm-riscv, linux-mm
Cc: Alexandre Ghiti
Make riscv use the contpte aware get_and_clear_full_ptes()/clear_full_ptes()
function from arm64.
Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com>
---
arch/arm64/include/asm/pgtable.h | 41 ++++------------------------
arch/arm64/mm/Makefile | 1 -
arch/arm64/mm/contpte.c | 46 -------------------------------
arch/riscv/include/asm/pgtable.h | 39 ++++++++++++++++++++++++++
include/linux/contpte.h | 5 ++++
mm/contpte.c | 47 ++++++++++++++++++++++++++++++++
6 files changed, 96 insertions(+), 83 deletions(-)
delete mode 100644 arch/arm64/mm/contpte.c
diff --git a/arch/arm64/include/asm/pgtable.h b/arch/arm64/include/asm/pgtable.h
index 162efd9647dd..f8a3159f9df0 100644
--- a/arch/arm64/include/asm/pgtable.h
+++ b/arch/arm64/include/asm/pgtable.h
@@ -1373,17 +1373,6 @@ extern void ptep_modify_prot_commit(struct vm_area_struct *vma,
#ifdef CONFIG_THP_CONTPTE
-/*
- * The contpte APIs are used to transparently manage the contiguous bit in ptes
- * where it is possible and makes sense to do so. The PTE_CONT bit is considered
- * a private implementation detail of the public ptep API (see below).
- */
-extern void contpte_clear_full_ptes(struct mm_struct *mm, unsigned long addr,
- pte_t *ptep, unsigned int nr, int full);
-extern pte_t contpte_get_and_clear_full_ptes(struct mm_struct *mm,
- unsigned long addr, pte_t *ptep,
- unsigned int nr, int full);
-
#define pte_batch_hint pte_batch_hint
static inline unsigned int pte_batch_hint(pte_t *ptep, pte_t pte)
{
@@ -1428,34 +1417,14 @@ extern void pte_clear(struct mm_struct *mm,
unsigned long addr, pte_t *ptep);
#define pte_clear pte_clear
+extern void clear_full_ptes(struct mm_struct *mm, unsigned long addr,
+ pte_t *ptep, unsigned int nr, int full);
#define clear_full_ptes clear_full_ptes
-static inline void clear_full_ptes(struct mm_struct *mm, unsigned long addr,
- pte_t *ptep, unsigned int nr, int full)
-{
- if (likely(nr == 1)) {
- contpte_try_unfold(mm, addr, ptep, __ptep_get(ptep));
- __clear_full_ptes(mm, addr, ptep, nr, full);
- } else {
- contpte_clear_full_ptes(mm, addr, ptep, nr, full);
- }
-}
+extern pte_t get_and_clear_full_ptes(struct mm_struct *mm,
+ unsigned long addr, pte_t *ptep,
+ unsigned int nr, int full);
#define get_and_clear_full_ptes get_and_clear_full_ptes
-static inline pte_t get_and_clear_full_ptes(struct mm_struct *mm,
- unsigned long addr, pte_t *ptep,
- unsigned int nr, int full)
-{
- pte_t pte;
-
- if (likely(nr == 1)) {
- contpte_try_unfold(mm, addr, ptep, __ptep_get(ptep));
- pte = __get_and_clear_full_ptes(mm, addr, ptep, nr, full);
- } else {
- pte = contpte_get_and_clear_full_ptes(mm, addr, ptep, nr, full);
- }
-
- return pte;
-}
#define __HAVE_ARCH_PTEP_GET_AND_CLEAR
extern pte_t ptep_get_and_clear(struct mm_struct *mm,
diff --git a/arch/arm64/mm/Makefile b/arch/arm64/mm/Makefile
index 52a1b2082627..dbd1bc95967d 100644
--- a/arch/arm64/mm/Makefile
+++ b/arch/arm64/mm/Makefile
@@ -3,7 +3,6 @@ obj-y := dma-mapping.o extable.o fault.o init.o \
cache.o copypage.o flush.o \
ioremap.o mmap.o pgd.o mmu.o \
context.o proc.o pageattr.o fixmap.o
-obj-$(CONFIG_THP_CONTPTE) += contpte.o
obj-$(CONFIG_HUGETLB_PAGE) += hugetlbpage.o
obj-$(CONFIG_PTDUMP_CORE) += ptdump.o
obj-$(CONFIG_PTDUMP_DEBUGFS) += ptdump_debugfs.o
diff --git a/arch/arm64/mm/contpte.c b/arch/arm64/mm/contpte.c
deleted file mode 100644
index 1cef93b15d6e..000000000000
--- a/arch/arm64/mm/contpte.c
+++ /dev/null
@@ -1,46 +0,0 @@
-// SPDX-License-Identifier: GPL-2.0-only
-/*
- * Copyright (C) 2023 ARM Ltd.
- */
-
-#include <linux/mm.h>
-#include <linux/efi.h>
-#include <linux/export.h>
-#include <asm/tlbflush.h>
-
-static void contpte_try_unfold_partial(struct mm_struct *mm, unsigned long addr,
- pte_t *ptep, unsigned int nr)
-{
- /*
- * Unfold any partially covered contpte block at the beginning and end
- * of the range.
- */
-
- if (ptep != arch_contpte_align_down(ptep) || nr < CONT_PTES)
- contpte_try_unfold(mm, addr, ptep, __ptep_get(ptep));
-
- if (ptep + nr != arch_contpte_align_down(ptep + nr)) {
- unsigned long last_addr = addr + PAGE_SIZE * (nr - 1);
- pte_t *last_ptep = ptep + nr - 1;
-
- contpte_try_unfold(mm, last_addr, last_ptep,
- __ptep_get(last_ptep));
- }
-}
-
-void contpte_clear_full_ptes(struct mm_struct *mm, unsigned long addr,
- pte_t *ptep, unsigned int nr, int full)
-{
- contpte_try_unfold_partial(mm, addr, ptep, nr);
- __clear_full_ptes(mm, addr, ptep, nr, full);
-}
-EXPORT_SYMBOL_GPL(contpte_clear_full_ptes);
-
-pte_t contpte_get_and_clear_full_ptes(struct mm_struct *mm,
- unsigned long addr, pte_t *ptep,
- unsigned int nr, int full)
-{
- contpte_try_unfold_partial(mm, addr, ptep, nr);
- return __get_and_clear_full_ptes(mm, addr, ptep, nr, full);
-}
-EXPORT_SYMBOL_GPL(contpte_get_and_clear_full_ptes);
diff --git a/arch/riscv/include/asm/pgtable.h b/arch/riscv/include/asm/pgtable.h
index 728f31da5e6a..a4843bdfdb37 100644
--- a/arch/riscv/include/asm/pgtable.h
+++ b/arch/riscv/include/asm/pgtable.h
@@ -754,6 +754,37 @@ static inline pte_t __ptep_get_and_clear(struct mm_struct *mm,
return pte;
}
+static inline void __clear_full_ptes(struct mm_struct *mm, unsigned long addr,
+ pte_t *ptep, unsigned int nr, int full)
+{
+ for (;;) {
+ __ptep_get_and_clear(mm, addr, ptep);
+ if (--nr == 0)
+ break;
+ ptep++;
+ addr += PAGE_SIZE;
+ }
+}
+
+static inline pte_t __get_and_clear_full_ptes(struct mm_struct *mm,
+ unsigned long addr, pte_t *ptep,
+ unsigned int nr, int full)
+{
+ pte_t pte, tmp_pte;
+
+ pte = __ptep_get_and_clear(mm, addr, ptep);
+ while (--nr) {
+ ptep++;
+ addr += PAGE_SIZE;
+ tmp_pte = __ptep_get_and_clear(mm, addr, ptep);
+ if (pte_dirty(tmp_pte))
+ pte = pte_mkdirty(pte);
+ if (pte_young(tmp_pte))
+ pte = pte_mkyoung(pte);
+ }
+ return pte;
+}
+
static inline void __ptep_set_wrprotect(struct mm_struct *mm,
unsigned long address, pte_t *ptep,
pte_t pte)
@@ -823,6 +854,13 @@ extern void ptep_set_wrprotect(struct mm_struct *mm,
extern void wrprotect_ptes(struct mm_struct *mm, unsigned long addr,
pte_t *ptep, unsigned int nr);
#define wrprotect_ptes wrprotect_ptes
+extern void clear_full_ptes(struct mm_struct *mm, unsigned long addr,
+ pte_t *ptep, unsigned int nr, int full);
+#define clear_full_ptes clear_full_ptes
+extern pte_t get_and_clear_full_ptes(struct mm_struct *mm,
+ unsigned long addr, pte_t *ptep,
+ unsigned int nr, int full);
+#define get_and_clear_full_ptes get_and_clear_full_ptes
#else /* CONFIG_THP_CONTPTE */
@@ -842,6 +880,7 @@ extern void wrprotect_ptes(struct mm_struct *mm, unsigned long addr,
#define ptep_set_wrprotect(mm, addr, ptep) \
__ptep_set_wrprotect(mm, addr, ptep, __ptep_get(ptep))
#define wrprotect_ptes __wrprotect_ptes
+#define clear_full_ptes __clear_full_ptes
#endif /* CONFIG_THP_CONTPTE */
diff --git a/include/linux/contpte.h b/include/linux/contpte.h
index d1439db1706c..b24554ebca41 100644
--- a/include/linux/contpte.h
+++ b/include/linux/contpte.h
@@ -28,5 +28,10 @@ int contpte_ptep_set_access_flags(struct vm_area_struct *vma,
pte_t entry, int dirty);
void contpte_wrprotect_ptes(struct mm_struct *mm, unsigned long addr,
pte_t *ptep, unsigned int nr);
+void contpte_clear_full_ptes(struct mm_struct *mm, unsigned long addr,
+ pte_t *ptep, unsigned int nr, int full);
+pte_t contpte_get_and_clear_full_ptes(struct mm_struct *mm,
+ unsigned long addr, pte_t *ptep,
+ unsigned int nr, int full);
#endif /* _LINUX_CONTPTE_H */
diff --git a/mm/contpte.c b/mm/contpte.c
index fe36b6b1d20a..677344e0e3c3 100644
--- a/mm/contpte.c
+++ b/mm/contpte.c
@@ -51,6 +51,8 @@
* - ptep_clear_flush_young()
* - wrprotect_ptes()
* - ptep_set_wrprotect()
+ * - clear_full_ptes()
+ * - get_and_clear_full_ptes()
*/
pte_t huge_ptep_get(pte_t *ptep)
@@ -905,4 +907,49 @@ __always_inline void ptep_set_wrprotect(struct mm_struct *mm,
{
wrprotect_ptes(mm, addr, ptep, 1);
}
+
+void contpte_clear_full_ptes(struct mm_struct *mm, unsigned long addr,
+ pte_t *ptep, unsigned int nr, int full)
+{
+ contpte_try_unfold_partial(mm, addr, ptep, nr);
+ __clear_full_ptes(mm, addr, ptep, nr, full);
+}
+EXPORT_SYMBOL_GPL(contpte_clear_full_ptes);
+
+pte_t contpte_get_and_clear_full_ptes(struct mm_struct *mm,
+ unsigned long addr, pte_t *ptep,
+ unsigned int nr, int full)
+{
+ contpte_try_unfold_partial(mm, addr, ptep, nr);
+ return __get_and_clear_full_ptes(mm, addr, ptep, nr, full);
+}
+EXPORT_SYMBOL_GPL(contpte_get_and_clear_full_ptes);
+
+__always_inline void clear_full_ptes(struct mm_struct *mm, unsigned long addr,
+ pte_t *ptep, unsigned int nr, int full)
+{
+ if (likely(nr == 1)) {
+ contpte_try_unfold(mm, addr, ptep, __ptep_get(ptep));
+ __clear_full_ptes(mm, addr, ptep, nr, full);
+ } else {
+ contpte_clear_full_ptes(mm, addr, ptep, nr, full);
+ }
+}
+
+__always_inline pte_t get_and_clear_full_ptes(struct mm_struct *mm,
+ unsigned long addr, pte_t *ptep,
+ unsigned int nr, int full)
+{
+ pte_t pte;
+
+ if (likely(nr == 1)) {
+ contpte_try_unfold(mm, addr, ptep, __ptep_get(ptep));
+ pte = __get_and_clear_full_ptes(mm, addr, ptep, nr, full);
+ } else {
+ pte = contpte_get_and_clear_full_ptes(mm, addr, ptep, nr, full);
+ }
+
+ return pte;
+}
+
#endif /* CONFIG_THP_CONTPTE */
--
2.39.2
^ permalink raw reply related [flat|nested] 16+ messages in thread
* Re: [PATCH 01/12] mm, arm64: Rename ARM64_CONTPTE to THP_CONTPTE
2024-05-08 19:19 ` [PATCH 01/12] mm, arm64: Rename ARM64_CONTPTE to THP_CONTPTE Alexandre Ghiti
@ 2024-05-09 0:46 ` Barry Song
2024-05-13 13:09 ` Alexandre Ghiti
0 siblings, 1 reply; 16+ messages in thread
From: Barry Song @ 2024-05-09 0:46 UTC (permalink / raw
To: Alexandre Ghiti
Cc: Ryan Roberts, Catalin Marinas, Will Deacon, Alexander Potapenko,
Marco Elver, Dmitry Vyukov, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Ard Biesheuvel, Anup Patel, Atish Patra,
Andrey Ryabinin, Andrey Konovalov, Vincenzo Frascino,
Andrew Morton, linux-arm-kernel, linux-kernel, kasan-dev,
linux-riscv, linux-efi, kvm, kvm-riscv, linux-mm
On Thu, May 9, 2024 at 7:20 AM Alexandre Ghiti <alexghiti@rivosinc.com> wrote:
>
> The ARM64_CONTPTE config represents the capability to transparently use
> contpte mappings for THP userspace mappings, which will be implemented
> in the next commits for riscv, so make this config more generic and move
> it to mm.
>
> Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com>
> ---
> arch/arm64/Kconfig | 9 ---------
> arch/arm64/include/asm/pgtable.h | 6 +++---
> arch/arm64/mm/Makefile | 2 +-
> mm/Kconfig | 9 +++++++++
> 4 files changed, 13 insertions(+), 13 deletions(-)
>
> diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
> index ac2f6d906cc3..9d823015b4e5 100644
> --- a/arch/arm64/Kconfig
> +++ b/arch/arm64/Kconfig
> @@ -2227,15 +2227,6 @@ config UNWIND_PATCH_PAC_INTO_SCS
> select UNWIND_TABLES
> select DYNAMIC_SCS
>
> -config ARM64_CONTPTE
> - bool "Contiguous PTE mappings for user memory" if EXPERT
> - depends on TRANSPARENT_HUGEPAGE
> - default y
> - help
> - When enabled, user mappings are configured using the PTE contiguous
> - bit, for any mappings that meet the size and alignment requirements.
> - This reduces TLB pressure and improves performance.
> -
> endmenu # "Kernel Features"
>
> menu "Boot options"
> diff --git a/arch/arm64/include/asm/pgtable.h b/arch/arm64/include/asm/pgtable.h
> index 7c2938cb70b9..1758ce71fae9 100644
> --- a/arch/arm64/include/asm/pgtable.h
> +++ b/arch/arm64/include/asm/pgtable.h
> @@ -1369,7 +1369,7 @@ extern void ptep_modify_prot_commit(struct vm_area_struct *vma,
> unsigned long addr, pte_t *ptep,
> pte_t old_pte, pte_t new_pte);
>
> -#ifdef CONFIG_ARM64_CONTPTE
> +#ifdef CONFIG_THP_CONTPTE
Is it necessarily THP? can't be hugetlb or others? I feel THP_CONTPTE
isn't a good name.
>
> /*
> * The contpte APIs are used to transparently manage the contiguous bit in ptes
> @@ -1622,7 +1622,7 @@ static inline int ptep_set_access_flags(struct vm_area_struct *vma,
> return contpte_ptep_set_access_flags(vma, addr, ptep, entry, dirty);
> }
>
> -#else /* CONFIG_ARM64_CONTPTE */
> +#else /* CONFIG_THP_CONTPTE */
>
> #define ptep_get __ptep_get
> #define set_pte __set_pte
> @@ -1642,7 +1642,7 @@ static inline int ptep_set_access_flags(struct vm_area_struct *vma,
> #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
> #define ptep_set_access_flags __ptep_set_access_flags
>
> -#endif /* CONFIG_ARM64_CONTPTE */
> +#endif /* CONFIG_THP_CONTPTE */
>
> int find_num_contig(struct mm_struct *mm, unsigned long addr,
> pte_t *ptep, size_t *pgsize);
> diff --git a/arch/arm64/mm/Makefile b/arch/arm64/mm/Makefile
> index 60454256945b..52a1b2082627 100644
> --- a/arch/arm64/mm/Makefile
> +++ b/arch/arm64/mm/Makefile
> @@ -3,7 +3,7 @@ obj-y := dma-mapping.o extable.o fault.o init.o \
> cache.o copypage.o flush.o \
> ioremap.o mmap.o pgd.o mmu.o \
> context.o proc.o pageattr.o fixmap.o
> -obj-$(CONFIG_ARM64_CONTPTE) += contpte.o
> +obj-$(CONFIG_THP_CONTPTE) += contpte.o
> obj-$(CONFIG_HUGETLB_PAGE) += hugetlbpage.o
> obj-$(CONFIG_PTDUMP_CORE) += ptdump.o
> obj-$(CONFIG_PTDUMP_DEBUGFS) += ptdump_debugfs.o
> diff --git a/mm/Kconfig b/mm/Kconfig
> index c325003d6552..fd4de221a1c6 100644
> --- a/mm/Kconfig
> +++ b/mm/Kconfig
> @@ -984,6 +984,15 @@ config ARCH_HAS_CACHE_LINE_SIZE
> config ARCH_HAS_CONTPTE
> bool
>
> +config THP_CONTPTE
> + bool "Contiguous PTE mappings for user memory" if EXPERT
> + depends on ARCH_HAS_CONTPTE && TRANSPARENT_HUGEPAGE
> + default y
> + help
> + When enabled, user mappings are configured using the PTE contiguous
> + bit, for any mappings that meet the size and alignment requirements.
> + This reduces TLB pressure and improves performance.
> +
> config ARCH_HAS_CURRENT_STACK_POINTER
> bool
> help
> --
> 2.39.2
Thanks
Barry
^ permalink raw reply [flat|nested] 16+ messages in thread
* Re: [PATCH 01/12] mm, arm64: Rename ARM64_CONTPTE to THP_CONTPTE
2024-05-09 0:46 ` Barry Song
@ 2024-05-13 13:09 ` Alexandre Ghiti
2024-05-14 9:30 ` Barry Song
0 siblings, 1 reply; 16+ messages in thread
From: Alexandre Ghiti @ 2024-05-13 13:09 UTC (permalink / raw
To: Barry Song
Cc: Ryan Roberts, Catalin Marinas, Will Deacon, Alexander Potapenko,
Marco Elver, Dmitry Vyukov, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Ard Biesheuvel, Anup Patel, Atish Patra,
Andrey Ryabinin, Andrey Konovalov, Vincenzo Frascino,
Andrew Morton, linux-arm-kernel, linux-kernel, kasan-dev,
linux-riscv, linux-efi, kvm, kvm-riscv, linux-mm
Hi Barry,
On Thu, May 9, 2024 at 2:46 AM Barry Song <21cnbao@gmail.com> wrote:
>
> On Thu, May 9, 2024 at 7:20 AM Alexandre Ghiti <alexghiti@rivosinc.com> wrote:
> >
> > The ARM64_CONTPTE config represents the capability to transparently use
> > contpte mappings for THP userspace mappings, which will be implemented
> > in the next commits for riscv, so make this config more generic and move
> > it to mm.
> >
> > Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com>
> > ---
> > arch/arm64/Kconfig | 9 ---------
> > arch/arm64/include/asm/pgtable.h | 6 +++---
> > arch/arm64/mm/Makefile | 2 +-
> > mm/Kconfig | 9 +++++++++
> > 4 files changed, 13 insertions(+), 13 deletions(-)
> >
> > diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
> > index ac2f6d906cc3..9d823015b4e5 100644
> > --- a/arch/arm64/Kconfig
> > +++ b/arch/arm64/Kconfig
> > @@ -2227,15 +2227,6 @@ config UNWIND_PATCH_PAC_INTO_SCS
> > select UNWIND_TABLES
> > select DYNAMIC_SCS
> >
> > -config ARM64_CONTPTE
> > - bool "Contiguous PTE mappings for user memory" if EXPERT
> > - depends on TRANSPARENT_HUGEPAGE
> > - default y
> > - help
> > - When enabled, user mappings are configured using the PTE contiguous
> > - bit, for any mappings that meet the size and alignment requirements.
> > - This reduces TLB pressure and improves performance.
> > -
> > endmenu # "Kernel Features"
> >
> > menu "Boot options"
> > diff --git a/arch/arm64/include/asm/pgtable.h b/arch/arm64/include/asm/pgtable.h
> > index 7c2938cb70b9..1758ce71fae9 100644
> > --- a/arch/arm64/include/asm/pgtable.h
> > +++ b/arch/arm64/include/asm/pgtable.h
> > @@ -1369,7 +1369,7 @@ extern void ptep_modify_prot_commit(struct vm_area_struct *vma,
> > unsigned long addr, pte_t *ptep,
> > pte_t old_pte, pte_t new_pte);
> >
> > -#ifdef CONFIG_ARM64_CONTPTE
> > +#ifdef CONFIG_THP_CONTPTE
>
> Is it necessarily THP? can't be hugetlb or others? I feel THP_CONTPTE
> isn't a good name.
This does not target hugetlbfs (see my other patchset for that here
https://lore.kernel.org/linux-riscv/7504a525-8211-48b3-becb-a6e838c1b42e@arm.com/T/#m57d273d680fc531b3aa1074e6f8558a52ba5badc).
What could be "others" here?
Thanks for your comment,
Alex
>
> >
> > /*
> > * The contpte APIs are used to transparently manage the contiguous bit in ptes
> > @@ -1622,7 +1622,7 @@ static inline int ptep_set_access_flags(struct vm_area_struct *vma,
> > return contpte_ptep_set_access_flags(vma, addr, ptep, entry, dirty);
> > }
> >
> > -#else /* CONFIG_ARM64_CONTPTE */
> > +#else /* CONFIG_THP_CONTPTE */
> >
> > #define ptep_get __ptep_get
> > #define set_pte __set_pte
> > @@ -1642,7 +1642,7 @@ static inline int ptep_set_access_flags(struct vm_area_struct *vma,
> > #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
> > #define ptep_set_access_flags __ptep_set_access_flags
> >
> > -#endif /* CONFIG_ARM64_CONTPTE */
> > +#endif /* CONFIG_THP_CONTPTE */
> >
> > int find_num_contig(struct mm_struct *mm, unsigned long addr,
> > pte_t *ptep, size_t *pgsize);
> > diff --git a/arch/arm64/mm/Makefile b/arch/arm64/mm/Makefile
> > index 60454256945b..52a1b2082627 100644
> > --- a/arch/arm64/mm/Makefile
> > +++ b/arch/arm64/mm/Makefile
> > @@ -3,7 +3,7 @@ obj-y := dma-mapping.o extable.o fault.o init.o \
> > cache.o copypage.o flush.o \
> > ioremap.o mmap.o pgd.o mmu.o \
> > context.o proc.o pageattr.o fixmap.o
> > -obj-$(CONFIG_ARM64_CONTPTE) += contpte.o
> > +obj-$(CONFIG_THP_CONTPTE) += contpte.o
> > obj-$(CONFIG_HUGETLB_PAGE) += hugetlbpage.o
> > obj-$(CONFIG_PTDUMP_CORE) += ptdump.o
> > obj-$(CONFIG_PTDUMP_DEBUGFS) += ptdump_debugfs.o
> > diff --git a/mm/Kconfig b/mm/Kconfig
> > index c325003d6552..fd4de221a1c6 100644
> > --- a/mm/Kconfig
> > +++ b/mm/Kconfig
> > @@ -984,6 +984,15 @@ config ARCH_HAS_CACHE_LINE_SIZE
> > config ARCH_HAS_CONTPTE
> > bool
> >
> > +config THP_CONTPTE
> > + bool "Contiguous PTE mappings for user memory" if EXPERT
> > + depends on ARCH_HAS_CONTPTE && TRANSPARENT_HUGEPAGE
> > + default y
> > + help
> > + When enabled, user mappings are configured using the PTE contiguous
> > + bit, for any mappings that meet the size and alignment requirements.
> > + This reduces TLB pressure and improves performance.
> > +
> > config ARCH_HAS_CURRENT_STACK_POINTER
> > bool
> > help
> > --
> > 2.39.2
>
> Thanks
> Barry
^ permalink raw reply [flat|nested] 16+ messages in thread
* Re: [PATCH 01/12] mm, arm64: Rename ARM64_CONTPTE to THP_CONTPTE
2024-05-13 13:09 ` Alexandre Ghiti
@ 2024-05-14 9:30 ` Barry Song
0 siblings, 0 replies; 16+ messages in thread
From: Barry Song @ 2024-05-14 9:30 UTC (permalink / raw
To: Alexandre Ghiti
Cc: Ryan Roberts, Catalin Marinas, Will Deacon, Alexander Potapenko,
Marco Elver, Dmitry Vyukov, Paul Walmsley, Palmer Dabbelt,
Albert Ou, Ard Biesheuvel, Anup Patel, Atish Patra,
Andrey Ryabinin, Andrey Konovalov, Vincenzo Frascino,
Andrew Morton, linux-arm-kernel, linux-kernel, kasan-dev,
linux-riscv, linux-efi, kvm, kvm-riscv, linux-mm
On Tue, May 14, 2024 at 1:09 AM Alexandre Ghiti <alexghiti@rivosinc.com> wrote:
>
> Hi Barry,
>
> On Thu, May 9, 2024 at 2:46 AM Barry Song <21cnbao@gmail.com> wrote:
> >
> > On Thu, May 9, 2024 at 7:20 AM Alexandre Ghiti <alexghiti@rivosinc.com> wrote:
> > >
> > > The ARM64_CONTPTE config represents the capability to transparently use
> > > contpte mappings for THP userspace mappings, which will be implemented
> > > in the next commits for riscv, so make this config more generic and move
> > > it to mm.
> > >
> > > Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com>
> > > ---
> > > arch/arm64/Kconfig | 9 ---------
> > > arch/arm64/include/asm/pgtable.h | 6 +++---
> > > arch/arm64/mm/Makefile | 2 +-
> > > mm/Kconfig | 9 +++++++++
> > > 4 files changed, 13 insertions(+), 13 deletions(-)
> > >
> > > diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
> > > index ac2f6d906cc3..9d823015b4e5 100644
> > > --- a/arch/arm64/Kconfig
> > > +++ b/arch/arm64/Kconfig
> > > @@ -2227,15 +2227,6 @@ config UNWIND_PATCH_PAC_INTO_SCS
> > > select UNWIND_TABLES
> > > select DYNAMIC_SCS
> > >
> > > -config ARM64_CONTPTE
> > > - bool "Contiguous PTE mappings for user memory" if EXPERT
> > > - depends on TRANSPARENT_HUGEPAGE
> > > - default y
> > > - help
> > > - When enabled, user mappings are configured using the PTE contiguous
> > > - bit, for any mappings that meet the size and alignment requirements.
> > > - This reduces TLB pressure and improves performance.
> > > -
> > > endmenu # "Kernel Features"
> > >
> > > menu "Boot options"
> > > diff --git a/arch/arm64/include/asm/pgtable.h b/arch/arm64/include/asm/pgtable.h
> > > index 7c2938cb70b9..1758ce71fae9 100644
> > > --- a/arch/arm64/include/asm/pgtable.h
> > > +++ b/arch/arm64/include/asm/pgtable.h
> > > @@ -1369,7 +1369,7 @@ extern void ptep_modify_prot_commit(struct vm_area_struct *vma,
> > > unsigned long addr, pte_t *ptep,
> > > pte_t old_pte, pte_t new_pte);
> > >
> > > -#ifdef CONFIG_ARM64_CONTPTE
> > > +#ifdef CONFIG_THP_CONTPTE
> >
> > Is it necessarily THP? can't be hugetlb or others? I feel THP_CONTPTE
> > isn't a good name.
>
> This does not target hugetlbfs (see my other patchset for that here
> https://lore.kernel.org/linux-riscv/7504a525-8211-48b3-becb-a6e838c1b42e@arm.com/T/#m57d273d680fc531b3aa1074e6f8558a52ba5badc).
>
> What could be "others" here?
I acknowledge that the current focus is on Transparent Huge Pages. However,
many aspects of CONT-PTE appear to be applicable to the mm-core in general.
For example,
/*
* The below functions constitute the public API that arm64 presents to the
* core-mm to manipulate PTE entries within their page tables (or at least this
* is the subset of the API that arm64 needs to implement). These public
* versions will automatically and transparently apply the contiguous bit where
* it makes sense to do so. Therefore any users that are contig-aware (e.g.
* hugetlb, kernel mapper) should NOT use these APIs, but instead use the
* private versions, which are prefixed with double underscore. All of these
* APIs except for ptep_get_lockless() are expected to be called with the PTL
* held. Although the contiguous bit is considered private to the
* implementation, it is deliberately allowed to leak through the getters (e.g.
* ptep_get()), back to core code. This is required so that pte_leaf_size() can
* provide an accurate size for perf_get_pgtable_size(). But this leakage means
* its possible a pte will be passed to a setter with the contiguous bit set, so
* we explicitly clear the contiguous bit in those cases to prevent accidentally
* setting it in the pgtable.
*/
#define ptep_get ptep_get
static inline pte_t ptep_get(pte_t *ptep)
{
pte_t pte = __ptep_get(ptep);
if (likely(!pte_valid_cont(pte)))
return pte;
return contpte_ptep_get(ptep, pte);
}
Could it possibly be given a more generic name such as "PGTABLE_CONTPTE"?
>
> Thanks for your comment,
>
> Alex
>
> >
> > >
> > > /*
> > > * The contpte APIs are used to transparently manage the contiguous bit in ptes
> > > @@ -1622,7 +1622,7 @@ static inline int ptep_set_access_flags(struct vm_area_struct *vma,
> > > return contpte_ptep_set_access_flags(vma, addr, ptep, entry, dirty);
> > > }
> > >
> > > -#else /* CONFIG_ARM64_CONTPTE */
> > > +#else /* CONFIG_THP_CONTPTE */
> > >
> > > #define ptep_get __ptep_get
> > > #define set_pte __set_pte
> > > @@ -1642,7 +1642,7 @@ static inline int ptep_set_access_flags(struct vm_area_struct *vma,
> > > #define __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
> > > #define ptep_set_access_flags __ptep_set_access_flags
> > >
> > > -#endif /* CONFIG_ARM64_CONTPTE */
> > > +#endif /* CONFIG_THP_CONTPTE */
> > >
> > > int find_num_contig(struct mm_struct *mm, unsigned long addr,
> > > pte_t *ptep, size_t *pgsize);
> > > diff --git a/arch/arm64/mm/Makefile b/arch/arm64/mm/Makefile
> > > index 60454256945b..52a1b2082627 100644
> > > --- a/arch/arm64/mm/Makefile
> > > +++ b/arch/arm64/mm/Makefile
> > > @@ -3,7 +3,7 @@ obj-y := dma-mapping.o extable.o fault.o init.o \
> > > cache.o copypage.o flush.o \
> > > ioremap.o mmap.o pgd.o mmu.o \
> > > context.o proc.o pageattr.o fixmap.o
> > > -obj-$(CONFIG_ARM64_CONTPTE) += contpte.o
> > > +obj-$(CONFIG_THP_CONTPTE) += contpte.o
> > > obj-$(CONFIG_HUGETLB_PAGE) += hugetlbpage.o
> > > obj-$(CONFIG_PTDUMP_CORE) += ptdump.o
> > > obj-$(CONFIG_PTDUMP_DEBUGFS) += ptdump_debugfs.o
> > > diff --git a/mm/Kconfig b/mm/Kconfig
> > > index c325003d6552..fd4de221a1c6 100644
> > > --- a/mm/Kconfig
> > > +++ b/mm/Kconfig
> > > @@ -984,6 +984,15 @@ config ARCH_HAS_CACHE_LINE_SIZE
> > > config ARCH_HAS_CONTPTE
> > > bool
> > >
> > > +config THP_CONTPTE
> > > + bool "Contiguous PTE mappings for user memory" if EXPERT
> > > + depends on ARCH_HAS_CONTPTE && TRANSPARENT_HUGEPAGE
> > > + default y
> > > + help
> > > + When enabled, user mappings are configured using the PTE contiguous
> > > + bit, for any mappings that meet the size and alignment requirements.
> > > + This reduces TLB pressure and improves performance.
> > > +
> > > config ARCH_HAS_CURRENT_STACK_POINTER
> > > bool
> > > help
> > > --
> > > 2.39.2
> >
Thanks
Barry
^ permalink raw reply [flat|nested] 16+ messages in thread
end of thread, other threads:[~2024-05-14 9:30 UTC | newest]
Thread overview: 16+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2024-05-08 19:19 [PATCH 00/12] Make riscv use THP contpte support for arm64 Alexandre Ghiti
2024-05-08 19:19 ` [PATCH 01/12] mm, arm64: Rename ARM64_CONTPTE to THP_CONTPTE Alexandre Ghiti
2024-05-09 0:46 ` Barry Song
2024-05-13 13:09 ` Alexandre Ghiti
2024-05-14 9:30 ` Barry Song
2024-05-08 19:19 ` [PATCH 02/12] mm, riscv, arm64: Use common ptep_get() function Alexandre Ghiti
2024-05-08 19:19 ` [PATCH 03/12] mm, riscv, arm64: Use common set_ptes() function Alexandre Ghiti
2024-05-08 19:19 ` [PATCH 04/12] mm, riscv, arm64: Use common ptep_get_lockless() function Alexandre Ghiti
2024-05-08 19:19 ` [PATCH 05/12] mm, riscv, arm64: Use common set_pte() function Alexandre Ghiti
2024-05-08 19:19 ` [PATCH 06/12] mm, riscv, arm64: Use common pte_clear() function Alexandre Ghiti
2024-05-08 19:19 ` [PATCH 07/12] mm, riscv, arm64: Use common ptep_get_and_clear() function Alexandre Ghiti
2024-05-08 19:19 ` [PATCH 08/12] mm, riscv, arm64: Use common ptep_test_and_clear_young() function Alexandre Ghiti
2024-05-08 19:19 ` [PATCH 09/12] mm, riscv, arm64: Use common ptep_clear_flush_young() function Alexandre Ghiti
2024-05-08 19:19 ` [PATCH 10/12] mm, riscv, arm64: Use common ptep_set_access_flags() function Alexandre Ghiti
2024-05-08 19:19 ` [PATCH 11/12] mm, riscv, arm64: Use common ptep_set_wrprotect()/wrprotect_ptes() functions Alexandre Ghiti
2024-05-08 19:19 ` [PATCH 12/12] mm, riscv, arm64: Use common get_and_clear_full_ptes()/clear_full_ptes() functions Alexandre Ghiti
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).