[RFC,3/4] mm: support Svnapot in hugetlb page

Commit Message

Qinglin Pan Oct. 18, 2021, 1:59 a.m. UTC

From: Qinglin Pan <[e-mail]>

Svnapot can be used to support 64KB hugetlb page, so it can become a new
option when using hugetlbfs. This patch adds a basic implementation of
hugetlb page, and support 64KB as a size in it by using Svnapot.

For test, boot kernel with command line contains "hugepagesz=64K
hugepages=20" and run a simple test like this:

int main() {
	void *addr;
	addr = mmap(NULL, 64 * 1024, PROT_WRITE | PROT_READ,
			MAP_PRIVATE | MAP_ANONYMOUS | MAP_HUGETLB | MAP_HUGE_64KB, -1, 0);
	printf("back from mmap \n");
	long *ptr = (long *)addr;
	unsigned int i = 0;
	for(; i < 8 * 1024;i += 512) {
		printf("%lp \n", ptr);
		*ptr = 0xdeafabcd12345678;
		ptr += 512;
	}
	ptr = (long *)addr;
	i = 0;
	for(; i < 8 * 1024;i += 512) {
		if (*ptr != 0xdeafabcd12345678) {
			printf("failed! 0x%lx \n", *ptr);
			break;
		}
		ptr += 512;
	}
	if(i == 8 * 1024)
		printf("simple test passed!\n");
}

And it should be passed.

Signed-off-by: Qinglin Pan <[e-mail]>
---
 arch/riscv/Kconfig               |   2 +-
 arch/riscv/include/asm/hugetlb.h |  29 +++-
 arch/riscv/include/asm/page.h    |   2 +-
 arch/riscv/mm/hugetlbpage.c      | 227 +++++++++++++++++++++++++++++++
 4 files changed, 257 insertions(+), 3 deletions(-)

Patch

diff --git a/arch/riscv/Kconfig b/arch/riscv/Kconfig
index 301a54233c7e..0ae025686faf 100644
--- a/arch/riscv/Kconfig
+++ b/arch/riscv/Kconfig
@@ -176,7 +176,7 @@  config ARCH_SELECT_MEMORY_MODEL
 	def_bool ARCH_SPARSEMEM_ENABLE
 
 config ARCH_WANT_GENERAL_HUGETLB
-	def_bool y
+	def_bool !HUGETLB_PAGE
 
 config ARCH_SUPPORTS_UPROBES
 	def_bool y
diff --git a/arch/riscv/include/asm/hugetlb.h b/arch/riscv/include/asm/hugetlb.h
index a5c2ca1d1cd8..fa99fb9226f7 100644
--- a/arch/riscv/include/asm/hugetlb.h
+++ b/arch/riscv/include/asm/hugetlb.h
@@ -2,7 +2,34 @@ 
 #ifndef _ASM_RISCV_HUGETLB_H
 #define _ASM_RISCV_HUGETLB_H
 
-#include <asm-generic/hugetlb.h>
 #include <asm/page.h>
 
+extern pte_t arch_make_huge_pte(pte_t entry, unsigned int shift,
+				       vm_flags_t flags);
+#define arch_make_huge_pte arch_make_huge_pte
+#define __HAVE_ARCH_HUGE_SET_HUGE_PTE_AT
+extern void set_huge_pte_at(struct mm_struct *mm, unsigned long addr,
+			    pte_t *ptep, pte_t pte);
+#define __HAVE_ARCH_HUGE_PTEP_GET_AND_CLEAR
+extern pte_t huge_ptep_get_and_clear(struct mm_struct *mm,
+				     unsigned long addr, pte_t *ptep);
+#define __HAVE_ARCH_HUGE_PTEP_CLEAR_FLUSH
+extern void huge_ptep_clear_flush(struct vm_area_struct *vma,
+				  unsigned long addr, pte_t *ptep);
+#define __HAVE_ARCH_HUGE_PTEP_SET_ACCESS_FLAGS
+extern int huge_ptep_set_access_flags(struct vm_area_struct *vma,
+				      unsigned long addr, pte_t *ptep,
+				      pte_t pte, int dirty);
+#define __HAVE_ARCH_HUGE_PTEP_SET_WRPROTECT
+extern void huge_ptep_set_wrprotect(struct mm_struct *mm,
+				    unsigned long addr, pte_t *ptep);
+#define __HAVE_ARCH_HUGE_PTE_CLEAR
+extern void huge_pte_clear(struct mm_struct *mm, unsigned long addr,
+			   pte_t *ptep, unsigned long sz);
+#define set_huge_swap_pte_at riscv_set_huge_swap_pte_at
+extern void riscv_set_huge_swap_pte_at(struct mm_struct *mm, unsigned long addr,
+				 pte_t *ptep, pte_t pte, unsigned long sz);
+
+#include <asm-generic/hugetlb.h>
+
 #endif /* _ASM_RISCV_HUGETLB_H */
diff --git a/arch/riscv/include/asm/page.h b/arch/riscv/include/asm/page.h
index 109c97e991a6..e67506dbcd53 100644
--- a/arch/riscv/include/asm/page.h
+++ b/arch/riscv/include/asm/page.h
@@ -17,7 +17,7 @@ 
 #define PAGE_MASK	(~(PAGE_SIZE - 1))
 
 #ifdef CONFIG_64BIT
-#define HUGE_MAX_HSTATE		2
+#define HUGE_MAX_HSTATE		3
 #else
 #define HUGE_MAX_HSTATE		1
 #endif
diff --git a/arch/riscv/mm/hugetlbpage.c b/arch/riscv/mm/hugetlbpage.c
index 932dadfdca54..b88a8dbfec3e 100644
--- a/arch/riscv/mm/hugetlbpage.c
+++ b/arch/riscv/mm/hugetlbpage.c
@@ -2,6 +2,224 @@ 
 #include <linux/hugetlb.h>
 #include <linux/err.h>
 
+pte_t *huge_pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
+			unsigned long addr, unsigned long sz)
+{
+	pgd_t *pgdp = pgd_offset(mm, addr);
+	p4d_t *p4dp = p4d_alloc(mm, pgdp, addr);
+	pud_t *pudp = pud_alloc(mm, p4dp, addr);
+	pmd_t *pmdp = pmd_alloc(mm, pudp, addr);
+
+	if (sz == NAPOT_CONT64KB_SIZE) {
+		if (!pmdp)
+			return NULL;
+		WARN_ON(addr & (sz - 1));
+		return pte_alloc_map(mm, pmdp, addr);
+	}
+
+	return NULL;
+}
+
+pte_t *huge_pte_offset(struct mm_struct *mm,
+		       unsigned long addr, unsigned long sz)
+{
+	pgd_t *pgdp;
+	p4d_t *p4dp;
+	pud_t *pudp;
+	pmd_t *pmdp;
+	pte_t *ptep = NULL;
+
+	pgdp = pgd_offset(mm, addr);
+	if (!pgd_present(READ_ONCE(*pgdp)))
+		return NULL;
+
+	p4dp = p4d_offset(pgdp, addr);
+	if (!p4d_present(READ_ONCE(*p4dp)))
+		return NULL;
+
+	pudp = pud_offset(p4dp, addr);
+	if (!pud_present(READ_ONCE(*pudp)))
+		return NULL;
+
+	pmdp = pmd_offset(pudp, addr);
+	if (!pmd_present(READ_ONCE(*pmdp)))
+		return NULL;
+
+	if (sz == NAPOT_CONT64KB_SIZE)
+		ptep = pte_offset_kernel(pmdp, (addr & ~NAPOT_CONT64KB_MASK));
+
+	return ptep;
+}
+
+int napot_pte_num(pte_t pte)
+{
+	if (!(pte_val(pte) & NAPOT_64KB_MASK))
+		return NAPOT_64KB_PTE_NUM;
+
+	pr_warn("%s: unrecognized napot pte size 0x%lx\n",
+		__func__, pte_val(pte));
+	return 1;
+}
+
+static pte_t get_clear_flush(struct mm_struct *mm,
+			     unsigned long addr,
+			     pte_t *ptep,
+			     unsigned long pte_num)
+{
+	pte_t orig_pte = huge_ptep_get(ptep);
+	bool valid = pte_val(orig_pte);
+	unsigned long i, saddr = addr;
+
+	for (i = 0; i < pte_num; i++, addr += PAGE_SIZE, ptep++) {
+		pte_t pte = ptep_get_and_clear(mm, addr, ptep);
+
+		if (pte_dirty(pte))
+			orig_pte = pte_mkdirty(orig_pte);
+
+		if (pte_young(pte))
+			orig_pte = pte_mkyoung(orig_pte);
+	}
+
+	if (valid) {
+		struct vm_area_struct vma = TLB_FLUSH_VMA(mm, 0);
+
+		flush_tlb_range(&vma, saddr, addr);
+	}
+	return orig_pte;
+}
+
+static void clear_flush(struct mm_struct *mm,
+			     unsigned long addr,
+			     pte_t *ptep,
+			     unsigned long pte_num)
+{
+	struct vm_area_struct vma = TLB_FLUSH_VMA(mm, 0);
+	unsigned long i, saddr = addr;
+
+	for (i = 0; i < pte_num; i++, addr += PAGE_SIZE, ptep++)
+		pte_clear(mm, addr, ptep);
+
+	flush_tlb_range(&vma, saddr, addr);
+}
+
+pte_t arch_make_huge_pte(pte_t entry, unsigned int shift,
+				       vm_flags_t flags)
+{
+	if (shift == NAPOT_CONT64KB_SHIFT)
+		entry = pte_mknapot(entry, NAPOT_CONT64KB_SHIFT - PAGE_SHIFT);
+
+	return entry;
+}
+
+void set_huge_pte_at(struct mm_struct *mm, unsigned long addr,
+			    pte_t *ptep, pte_t pte)
+{
+	int i;
+	int pte_num;
+
+	if (!pte_napot(pte)) {
+		set_pte_at(mm, addr, ptep, pte);
+		return;
+	}
+
+	pte_num = napot_pte_num(pte);
+	for (i = 0; i < pte_num; i++, ptep++, addr += PAGE_SIZE)
+		set_pte_at(mm, addr, ptep, pte);
+}
+
+int huge_ptep_set_access_flags(struct vm_area_struct *vma,
+				      unsigned long addr, pte_t *ptep,
+				      pte_t pte, int dirty)
+{
+	pte_t orig_pte;
+	int i;
+	int pte_num;
+
+	if (!pte_napot(pte))
+		return ptep_set_access_flags(vma, addr, ptep, pte, dirty);
+
+	pte_num = napot_pte_num(pte);
+	ptep = huge_pte_offset(vma->vm_mm, addr, NAPOT_CONT64KB_SIZE);
+	orig_pte = huge_ptep_get(ptep);
+
+	if (pte_dirty(orig_pte))
+		pte = pte_mkdirty(pte);
+
+	if (pte_young(orig_pte))
+		pte = pte_mkyoung(pte);
+
+	for (i = 0; i < pte_num; i++, addr += PAGE_SIZE, ptep++)
+		ptep_set_access_flags(vma, addr, ptep, pte, dirty);
+
+	return true;
+}
+
+pte_t huge_ptep_get_and_clear(struct mm_struct *mm,
+				     unsigned long addr, pte_t *ptep)
+{
+	int pte_num;
+	pte_t orig_pte = huge_ptep_get(ptep);
+
+	if (!pte_napot(orig_pte))
+		return ptep_get_and_clear(mm, addr, ptep);
+
+	pte_num = napot_pte_num(orig_pte);
+	return get_clear_flush(mm, addr, ptep, pte_num);
+}
+
+void huge_ptep_set_wrprotect(struct mm_struct *mm,
+				    unsigned long addr, pte_t *ptep)
+{
+	int i;
+	int pte_num;
+	pte_t pte = READ_ONCE(*ptep);
+
+	if (!pte_napot(pte))
+		return ptep_set_wrprotect(mm, addr, ptep);
+
+	pte_num = napot_pte_num(pte);
+	ptep = huge_pte_offset(mm, addr, NAPOT_CONT64KB_SIZE);
+
+	for (i = 0; i < pte_num; i++, addr += PAGE_SIZE, ptep++)
+		ptep_set_wrprotect(mm, addr, ptep);
+}
+
+void huge_ptep_clear_flush(struct vm_area_struct *vma,
+				  unsigned long addr, pte_t *ptep)
+{
+	int pte_num;
+	pte_t pte = READ_ONCE(*ptep);
+
+	if (!pte_napot(pte)) {
+		ptep_clear_flush(vma, addr, ptep);
+		return;
+	}
+
+	pte_num = napot_pte_num(pte);
+	clear_flush(vma->vm_mm, addr, ptep, pte_num);
+}
+
+void huge_pte_clear(struct mm_struct *mm, unsigned long addr,
+			   pte_t *ptep, unsigned long sz)
+{
+	int i, pte_num;
+
+	pte_num = napot_pte_num(READ_ONCE(*ptep));
+	for (i = 0; i < pte_num; i++, addr += PAGE_SIZE, ptep++)
+		pte_clear(mm, addr, ptep);
+}
+
+void riscv_set_huge_swap_pte_at(struct mm_struct *mm, unsigned long addr,
+				 pte_t *ptep, pte_t pte, unsigned long sz)
+{
+	int i, pte_num;
+
+	pte_num = napot_pte_num(READ_ONCE(*ptep));
+
+	for (i = 0; i < pte_num; i++, ptep++)
+		set_pte(ptep, pte);
+}
+
 int pud_huge(pud_t pud)
 {
 	return pud_leaf(pud);
@@ -18,6 +236,8 @@  bool __init arch_hugetlb_valid_size(unsigned long size)
 		return true;
 	else if (IS_ENABLED(CONFIG_64BIT) && size == PUD_SIZE)
 		return true;
+	else if (size == NAPOT_CONT64KB_SIZE)
+		return true;
 	else
 		return false;
 }
@@ -32,3 +252,10 @@  static __init int gigantic_pages_init(void)
 }
 arch_initcall(gigantic_pages_init);
 #endif
+
+static int __init hugetlbpage_init(void)
+{
+	hugetlb_add_hstate(NAPOT_CONT64KB_SHIFT - PAGE_SHIFT);
+	return 0;
+}
+arch_initcall(hugetlbpage_init);