This is an initial working version of getting huge pages support on parisc. Can be pulled from my "hugetlb_hppa3" branch of my git tree as well: http://git.kernel.org/cgit/linux/kernel/git/deller/parisc-linux.git/log/?h=hugetlb_hppa3 Signed-off-by: Helge Deller <deller@xxxxxx> diff --git a/arch/parisc/Kconfig b/arch/parisc/Kconfig index c365469..2278497 100644 --- a/arch/parisc/Kconfig +++ b/arch/parisc/Kconfig @@ -108,6 +108,9 @@ config PGTABLE_LEVELS default 3 if 64BIT && PARISC_PAGE_SIZE_4KB default 2 +config SYS_SUPPORTS_HUGETLBFS + def_bool y if 64BIT + source "init/Kconfig" source "kernel/Kconfig.freezer" diff --git a/arch/parisc/include/asm/hugetlb.h b/arch/parisc/include/asm/hugetlb.h new file mode 100644 index 0000000..7d56a9c --- /dev/null +++ b/arch/parisc/include/asm/hugetlb.h @@ -0,0 +1,85 @@ +#ifndef _ASM_PARISC64_HUGETLB_H +#define _ASM_PARISC64_HUGETLB_H + +#include <asm/page.h> +#include <asm-generic/hugetlb.h> + + +void set_huge_pte_at(struct mm_struct *mm, unsigned long addr, + pte_t *ptep, pte_t pte); + +pte_t huge_ptep_get_and_clear(struct mm_struct *mm, unsigned long addr, + pte_t *ptep); + +static inline int is_hugepage_only_range(struct mm_struct *mm, + unsigned long addr, + unsigned long len) { + return 0; +} + +/* + * If the arch doesn't supply something else, assume that hugepage + * size aligned regions are ok without further preparation. + */ +static inline int prepare_hugepage_range(struct file *file, + unsigned long addr, unsigned long len) +{ + if (len & ~HPAGE_MASK) + return -EINVAL; + if (addr & ~HPAGE_MASK) + return -EINVAL; + return 0; +} + +static inline void hugetlb_free_pgd_range(struct mmu_gather *tlb, + unsigned long addr, unsigned long end, + unsigned long floor, + unsigned long ceiling) +{ + free_pgd_range(tlb, addr, end, floor, ceiling); +} + +static inline void huge_ptep_clear_flush(struct vm_area_struct *vma, + unsigned long addr, pte_t *ptep) +{ +} + +static inline int huge_pte_none(pte_t pte) +{ + return pte_none(pte); +} + +static inline pte_t huge_pte_wrprotect(pte_t pte) +{ + return pte_wrprotect(pte); +} + +static inline void huge_ptep_set_wrprotect(struct mm_struct *mm, + unsigned long addr, pte_t *ptep) +{ + pte_t old_pte = *ptep; + set_huge_pte_at(mm, addr, ptep, pte_wrprotect(old_pte)); +} + +static inline int huge_ptep_set_access_flags(struct vm_area_struct *vma, + unsigned long addr, pte_t *ptep, + pte_t pte, int dirty) +{ + int changed = !pte_same(*ptep, pte); + if (changed) { + set_huge_pte_at(vma->vm_mm, addr, ptep, pte); + flush_tlb_page(vma, addr); + } + return changed; +} + +static inline pte_t huge_ptep_get(pte_t *ptep) +{ + return *ptep; +} + +static inline void arch_clear_hugepage_flags(struct page *page) +{ +} + +#endif /* _ASM_PARISC64_HUGETLB_H */ diff --git a/arch/parisc/include/asm/page.h b/arch/parisc/include/asm/page.h index 60d5d17..ca2dff4 100644 --- a/arch/parisc/include/asm/page.h +++ b/arch/parisc/include/asm/page.h @@ -145,10 +145,13 @@ extern int npmem_ranges; #endif /* CONFIG_DISCONTIGMEM */ #ifdef CONFIG_HUGETLB_PAGE -#define HPAGE_SHIFT 22 /* 4MB (is this fixed?) */ +#define HPAGE_SHIFT PMD_SHIFT /* fixed for transparent huge pages */ #define HPAGE_SIZE ((1UL) << HPAGE_SHIFT) #define HPAGE_MASK (~(HPAGE_SIZE - 1)) #define HUGETLB_PAGE_ORDER (HPAGE_SHIFT - PAGE_SHIFT) + +#define REAL_HPAGE_SHIFT 20 /* 20=1MB, 22=4MB */ +#define _HUGE_PAGE_SIZE_ENCODING_DEFAULT _PAGE_SIZE_ENCODING_1M #endif #define virt_addr_valid(kaddr) pfn_valid(__pa(kaddr) >> PAGE_SHIFT) diff --git a/arch/parisc/include/asm/pgtable.h b/arch/parisc/include/asm/pgtable.h index f93c4a4..148a243 100644 --- a/arch/parisc/include/asm/pgtable.h +++ b/arch/parisc/include/asm/pgtable.h @@ -167,7 +167,7 @@ static inline void purge_tlb_entries(struct mm_struct *mm, unsigned long addr) #define _PAGE_NO_CACHE_BIT 24 /* (0x080) Uncached Page (U bit) */ #define _PAGE_ACCESSED_BIT 23 /* (0x100) Software: Page Accessed */ #define _PAGE_PRESENT_BIT 22 /* (0x200) Software: translation valid */ -/* bit 21 was formerly the FLUSH bit but is now unused */ +#define _PAGE_HPAGE_BIT 21 /* (0x400) Huge Page */ #define _PAGE_USER_BIT 20 /* (0x800) Software: User accessible page */ /* N.B. The bits are defined in terms of a 32 bit word above, so the */ @@ -194,6 +194,7 @@ static inline void purge_tlb_entries(struct mm_struct *mm, unsigned long addr) #define _PAGE_NO_CACHE (1 << xlate_pabit(_PAGE_NO_CACHE_BIT)) #define _PAGE_ACCESSED (1 << xlate_pabit(_PAGE_ACCESSED_BIT)) #define _PAGE_PRESENT (1 << xlate_pabit(_PAGE_PRESENT_BIT)) +#define _PAGE_HUGE (1 << xlate_pabit(_PAGE_HPAGE_BIT)) #define _PAGE_USER (1 << xlate_pabit(_PAGE_USER_BIT)) #define _PAGE_TABLE (_PAGE_PRESENT | _PAGE_READ | _PAGE_WRITE | _PAGE_DIRTY | _PAGE_ACCESSED) @@ -217,7 +218,7 @@ static inline void purge_tlb_entries(struct mm_struct *mm, unsigned long addr) #define PxD_FLAG_VALID (1 << xlate_pabit(_PxD_VALID_BIT)) #define PxD_FLAG_MASK (0xf) #define PxD_FLAG_SHIFT (4) -#define PxD_VALUE_SHIFT (8) /* (PAGE_SHIFT-PxD_FLAG_SHIFT) */ +#define PxD_VALUE_SHIFT (PFN_PTE_SHIFT-PxD_FLAG_SHIFT) #ifndef __ASSEMBLY__ @@ -363,6 +364,13 @@ static inline pte_t pte_mkwrite(pte_t pte) { pte_val(pte) |= _PAGE_WRITE; return static inline pte_t pte_mkspecial(pte_t pte) { return pte; } /* + * Huge pte definitions. + */ +#define pte_huge(pte) (pte_val(pte) & _PAGE_HUGE) +#define pte_mkhuge(pte) (__pte(pte_val(pte) | _PAGE_HUGE)) + + +/* * Conversion functions: convert a page and protection to a page entry, * and a page entry and page directory to the page they refer to. */ @@ -410,8 +418,9 @@ static inline pte_t pte_modify(pte_t pte, pgprot_t newprot) /* Find an entry in the second-level page table.. */ #if CONFIG_PGTABLE_LEVELS == 3 +#define pmd_index(addr) (((addr) >> PMD_SHIFT) & (PTRS_PER_PMD - 1)) #define pmd_offset(dir,address) \ -((pmd_t *) pgd_page_vaddr(*(dir)) + (((address)>>PMD_SHIFT) & (PTRS_PER_PMD-1))) +((pmd_t *) pgd_page_vaddr(*(dir)) + pmd_index(address)) #else #define pmd_offset(dir,addr) ((pmd_t *) dir) #endif @@ -504,6 +513,8 @@ static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr, #define _PAGE_SIZE_ENCODING_4M 5 #define _PAGE_SIZE_ENCODING_16M 6 #define _PAGE_SIZE_ENCODING_64M 7 +#define _PAGE_SIZE_ENCODING_MASK 0x0f +#define PHYS_PAGE_SIZE_ENCODING(x) ((x) & _PAGE_SIZE_ENCODING_MASK) #if defined(CONFIG_PARISC_PAGE_SIZE_4KB) # define _PAGE_SIZE_ENCODING_DEFAULT _PAGE_SIZE_ENCODING_4K diff --git a/arch/parisc/include/uapi/asm/mman.h b/arch/parisc/include/uapi/asm/mman.h index 294d251..587005a 100644 --- a/arch/parisc/include/uapi/asm/mman.h +++ b/arch/parisc/include/uapi/asm/mman.h @@ -46,25 +46,15 @@ #define MADV_DONTFORK 10 /* don't inherit across fork */ #define MADV_DOFORK 11 /* do inherit across fork */ -/* The range 12-64 is reserved for page size specification. */ -#define MADV_4K_PAGES 12 /* Use 4K pages */ -#define MADV_16K_PAGES 14 /* Use 16K pages */ -#define MADV_64K_PAGES 16 /* Use 64K pages */ -#define MADV_256K_PAGES 18 /* Use 256K pages */ -#define MADV_1M_PAGES 20 /* Use 1 Megabyte pages */ -#define MADV_4M_PAGES 22 /* Use 4 Megabyte pages */ -#define MADV_16M_PAGES 24 /* Use 16 Megabyte pages */ -#define MADV_64M_PAGES 26 /* Use 64 Megabyte pages */ +#define MADV_MERGEABLE 12 /* KSM may merge identical pages */ +#define MADV_UNMERGEABLE 13 /* KSM may not merge identical pages */ -#define MADV_MERGEABLE 65 /* KSM may merge identical pages */ -#define MADV_UNMERGEABLE 66 /* KSM may not merge identical pages */ +#define MADV_HUGEPAGE 14 /* Worth backing with hugepages */ +#define MADV_NOHUGEPAGE 15 /* Not worth backing with hugepages */ -#define MADV_HUGEPAGE 67 /* Worth backing with hugepages */ -#define MADV_NOHUGEPAGE 68 /* Not worth backing with hugepages */ - -#define MADV_DONTDUMP 69 /* Explicity exclude from the core dump, +#define MADV_DONTDUMP 16 /* Explicity exclude from the core dump, overrides the coredump filter bits */ -#define MADV_DODUMP 70 /* Clear the MADV_NODUMP flag */ +#define MADV_DODUMP 17 /* Clear the MADV_NODUMP flag */ /* compatibility flags */ #define MAP_FILE 0 diff --git a/arch/parisc/kernel/entry.S b/arch/parisc/kernel/entry.S index c5ef408..f0e5157 100644 --- a/arch/parisc/kernel/entry.S +++ b/arch/parisc/kernel/entry.S @@ -504,19 +504,32 @@ /* bitshift difference between a PFN (based on kernel's PAGE_SIZE) * to a CPU TLB 4k PFN (4k => 12 bits to shift) */ - #define PAGE_ADD_SHIFT (PAGE_SHIFT-12) + #define PAGE_ADD_SHIFT (PAGE_SHIFT-12) + #define PAGE_ADD_HUGE_SHIFT (REAL_HPAGE_SHIFT-12) /* Drop prot bits and convert to page addr for iitlbt and idtlbt */ - .macro convert_for_tlb_insert20 pte + .macro convert_for_tlb_insert20 pte,tmp +#ifdef CONFIG_HUGETLB_PAGE + copy \pte,\tmp + extrd,u \tmp,(63-ASM_PFN_PTE_SHIFT)+(63-58)+PAGE_ADD_SHIFT,\ + 64-PAGE_SHIFT-PAGE_ADD_SHIFT,\pte + + depdi _PAGE_SIZE_ENCODING_DEFAULT,63,\ + (63-58)+PAGE_ADD_SHIFT,\pte + extrd,u,*= \tmp,_PAGE_HPAGE_BIT+32,1,%r0 + depdi _HUGE_PAGE_SIZE_ENCODING_DEFAULT,63,\ + (63-58)+PAGE_ADD_SHIFT,\pte +#else /* Huge pages disabled */ extrd,u \pte,(63-ASM_PFN_PTE_SHIFT)+(63-58)+PAGE_ADD_SHIFT,\ 64-PAGE_SHIFT-PAGE_ADD_SHIFT,\pte depdi _PAGE_SIZE_ENCODING_DEFAULT,63,\ (63-58)+PAGE_ADD_SHIFT,\pte +#endif .endm /* Convert the pte and prot to tlb insertion values. How * this happens is quite subtle, read below */ - .macro make_insert_tlb spc,pte,prot + .macro make_insert_tlb spc,pte,prot,tmp space_to_prot \spc \prot /* create prot id from space */ /* The following is the real subtlety. This is depositing * T <-> _PAGE_REFTRAP @@ -553,7 +566,7 @@ depdi 1,12,1,\prot /* Drop prot bits and convert to page addr for iitlbt and idtlbt */ - convert_for_tlb_insert20 \pte + convert_for_tlb_insert20 \pte \tmp .endm /* Identical macro to make_insert_tlb above, except it @@ -1147,7 +1160,7 @@ dtlb_miss_20w: tlb_lock spc,ptp,pte,t0,t1,dtlb_check_alias_20w update_accessed ptp,pte,t0,t1 - make_insert_tlb spc,pte,prot + make_insert_tlb spc,pte,prot,t0 idtlbt pte,prot @@ -1173,7 +1186,7 @@ nadtlb_miss_20w: tlb_lock spc,ptp,pte,t0,t1,nadtlb_check_alias_20w update_accessed ptp,pte,t0,t1 - make_insert_tlb spc,pte,prot + make_insert_tlb spc,pte,prot,t0 idtlbt pte,prot @@ -1267,7 +1280,7 @@ dtlb_miss_20: tlb_lock spc,ptp,pte,t0,t1,dtlb_check_alias_20 update_accessed ptp,pte,t0,t1 - make_insert_tlb spc,pte,prot + make_insert_tlb spc,pte,prot,t0 f_extend pte,t1 @@ -1295,7 +1308,7 @@ nadtlb_miss_20: tlb_lock spc,ptp,pte,t0,t1,nadtlb_check_alias_20 update_accessed ptp,pte,t0,t1 - make_insert_tlb spc,pte,prot + make_insert_tlb spc,pte,prot,t0 f_extend pte,t1 @@ -1404,7 +1417,7 @@ itlb_miss_20w: tlb_lock spc,ptp,pte,t0,t1,itlb_fault update_accessed ptp,pte,t0,t1 - make_insert_tlb spc,pte,prot + make_insert_tlb spc,pte,prot,t0 iitlbt pte,prot @@ -1428,7 +1441,7 @@ naitlb_miss_20w: tlb_lock spc,ptp,pte,t0,t1,naitlb_check_alias_20w update_accessed ptp,pte,t0,t1 - make_insert_tlb spc,pte,prot + make_insert_tlb spc,pte,prot,t0 iitlbt pte,prot @@ -1514,7 +1527,7 @@ itlb_miss_20: tlb_lock spc,ptp,pte,t0,t1,itlb_fault update_accessed ptp,pte,t0,t1 - make_insert_tlb spc,pte,prot + make_insert_tlb spc,pte,prot,t0 f_extend pte,t1 @@ -1534,7 +1547,7 @@ naitlb_miss_20: tlb_lock spc,ptp,pte,t0,t1,naitlb_check_alias_20 update_accessed ptp,pte,t0,t1 - make_insert_tlb spc,pte,prot + make_insert_tlb spc,pte,prot,t0 f_extend pte,t1 @@ -1566,7 +1579,7 @@ dbit_trap_20w: tlb_lock spc,ptp,pte,t0,t1,dbit_fault update_dirty ptp,pte,t1 - make_insert_tlb spc,pte,prot + make_insert_tlb spc,pte,prot,t0 idtlbt pte,prot @@ -1610,7 +1623,7 @@ dbit_trap_20: tlb_lock spc,ptp,pte,t0,t1,dbit_fault update_dirty ptp,pte,t1 - make_insert_tlb spc,pte,prot + make_insert_tlb spc,pte,prot,t0 f_extend pte,t1 diff --git a/arch/parisc/mm/Makefile b/arch/parisc/mm/Makefile index 758ceef..134393d 100644 --- a/arch/parisc/mm/Makefile +++ b/arch/parisc/mm/Makefile @@ -3,3 +3,4 @@ # obj-y := init.o fault.o ioremap.o +obj-$(CONFIG_HUGETLB_PAGE) += hugetlbpage.o diff --git a/arch/parisc/mm/hugetlbpage.c b/arch/parisc/mm/hugetlbpage.c new file mode 100644 index 0000000..3e82f86 --- /dev/null +++ b/arch/parisc/mm/hugetlbpage.c @@ -0,0 +1,160 @@ +/* + * PARISC64 Huge TLB page support. + * + * Copyright (C) 2002, 2003, 2006 David S. Miller (davem@xxxxxxxxxxxxx) + */ + +#include <linux/fs.h> +#include <linux/mm.h> +#include <linux/hugetlb.h> +#include <linux/pagemap.h> +#include <linux/sysctl.h> + +#include <asm/mman.h> +#include <asm/pgalloc.h> +#include <asm/tlb.h> +#include <asm/tlbflush.h> +#include <asm/cacheflush.h> +#include <asm/mmu_context.h> + + +unsigned long +hugetlb_get_unmapped_area(struct file *file, unsigned long addr, + unsigned long len, unsigned long pgoff, unsigned long flags) +{ + struct hstate *h = hstate_file(file); + + if (len & ~huge_page_mask(h)) + return -EINVAL; + if (len > TASK_SIZE) + return -ENOMEM; + + if (flags & MAP_FIXED) + if (prepare_hugepage_range(file, addr, len)) + return -EINVAL; + + if (addr) + addr = ALIGN(addr, huge_page_size(h)); + + /* we need to make sure the colouring is OK */ + return arch_get_unmapped_area(file, addr, len, pgoff, flags); +} + + +pte_t *huge_pte_alloc(struct mm_struct *mm, + unsigned long addr, unsigned long sz) +{ + pgd_t *pgd; + pud_t *pud; + pmd_t *pmd; + pte_t *pte = NULL; + + /* We must align the address, because our caller will run + * set_huge_pte_at() on whatever we return, which writes out + * all of the sub-ptes for the hugepage range. So we have + * to give it the first such sub-pte. + */ + addr &= HPAGE_MASK; + + pgd = pgd_offset(mm, addr); + pud = pud_alloc(mm, pgd, addr); + if (pud) { + pmd = pmd_alloc(mm, pud, addr); + if (pmd) + pte = pte_alloc_map(mm, NULL, pmd, addr); + } + return pte; +} + +pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr) +{ + pgd_t *pgd; + pud_t *pud; + pmd_t *pmd; + pte_t *pte = NULL; + + addr &= HPAGE_MASK; + + pgd = pgd_offset(mm, addr); + if (!pgd_none(*pgd)) { + pud = pud_offset(pgd, addr); + if (!pud_none(*pud)) { + pmd = pmd_offset(pud, addr); + if (!pmd_none(*pmd)) + pte = pte_offset_map(pmd, addr); + } + } + return pte; +} + +/* Purge data and instruction TLB entries. Must be called holding + * the pa_tlb_lock. The TLB purge instructions are slow on SMP + * machines since the purge must be broadcast to all CPUs. + */ +static inline void purge_tlb_entries_huge(struct mm_struct *mm, unsigned long addr) +{ + // we use 2 physical huge pages to emulate Linux huge pages + BUILD_BUG_ON(REAL_HPAGE_SHIFT+1 != HPAGE_SHIFT); + + mtsp(mm->context, 1); + addr &= HPAGE_MASK; + addr |= _HUGE_PAGE_SIZE_ENCODING_DEFAULT; + pdtlb(addr); + pdtlb(addr + (1UL << REAL_HPAGE_SHIFT)); + if (unlikely(split_tlb)) { + pitlb(addr); + pitlb(addr + (1UL << REAL_HPAGE_SHIFT)); + } +} + +void set_huge_pte_at(struct mm_struct *mm, unsigned long addr, + pte_t *ptep, pte_t entry) +{ + unsigned long flags; + unsigned long addr_start; + int i; + +#if 0 // HELGE + if (!pte_present(*ptep) && pte_present(entry)) + mm->context.huge_pte_count++; +#endif + + addr &= HPAGE_MASK; + addr_start = addr; + + spin_lock_irqsave(&pa_tlb_lock, flags); + + for (i = 0; i < (1 << HUGETLB_PAGE_ORDER); i++) { + *ptep = entry; // set_pte_at(mm, addr, ptep, entry); + ptep++; + purge_tlb_entries(mm, addr); // non-huge + addr += PAGE_SIZE; + pte_val(entry) += PAGE_SIZE; + } + + purge_tlb_entries_huge(mm, addr_start); + + spin_unlock_irqrestore(&pa_tlb_lock, flags); +} + + +pte_t huge_ptep_get_and_clear(struct mm_struct *mm, unsigned long addr, + pte_t *ptep) +{ + pte_t entry; + + entry = *ptep; + set_huge_pte_at(mm, addr, ptep, __pte(0)); + + return entry; +} + +int pmd_huge(pmd_t pmd) +{ + return 0; +} + +int pud_huge(pud_t pud) +{ + return 0; +} -- To unsubscribe from this list: send the line "unsubscribe linux-parisc" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
