Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | /* pgalloc.c: page directory & page table allocation |
| 2 | * |
| 3 | * Copyright (C) 2004 Red Hat, Inc. All Rights Reserved. |
| 4 | * Written by David Howells (dhowells@redhat.com) |
| 5 | * |
| 6 | * This program is free software; you can redistribute it and/or |
| 7 | * modify it under the terms of the GNU General Public License |
| 8 | * as published by the Free Software Foundation; either version |
| 9 | * 2 of the License, or (at your option) any later version. |
| 10 | */ |
| 11 | |
| 12 | #include <linux/sched.h> |
| 13 | #include <linux/slab.h> |
| 14 | #include <linux/mm.h> |
| 15 | #include <linux/highmem.h> |
| 16 | #include <asm/pgalloc.h> |
| 17 | #include <asm/page.h> |
| 18 | #include <asm/cacheflush.h> |
| 19 | |
| 20 | pgd_t swapper_pg_dir[PTRS_PER_PGD] __attribute__((aligned(PAGE_SIZE))); |
| 21 | kmem_cache_t *pgd_cache; |
| 22 | |
| 23 | pte_t *pte_alloc_one_kernel(struct mm_struct *mm, unsigned long address) |
| 24 | { |
| 25 | pte_t *pte = (pte_t *)__get_free_page(GFP_KERNEL|__GFP_REPEAT); |
| 26 | if (pte) |
| 27 | clear_page(pte); |
| 28 | return pte; |
| 29 | } |
| 30 | |
| 31 | struct page *pte_alloc_one(struct mm_struct *mm, unsigned long address) |
| 32 | { |
| 33 | struct page *page; |
| 34 | |
| 35 | #ifdef CONFIG_HIGHPTE |
| 36 | page = alloc_pages(GFP_KERNEL|__GFP_HIGHMEM|__GFP_REPEAT, 0); |
| 37 | #else |
| 38 | page = alloc_pages(GFP_KERNEL|__GFP_REPEAT, 0); |
| 39 | #endif |
| 40 | if (page) |
| 41 | clear_highpage(page); |
| 42 | flush_dcache_page(page); |
| 43 | return page; |
| 44 | } |
| 45 | |
| 46 | void __set_pmd(pmd_t *pmdptr, unsigned long pmd) |
| 47 | { |
| 48 | unsigned long *__ste_p = pmdptr->ste; |
| 49 | int loop; |
| 50 | |
| 51 | if (!pmd) { |
| 52 | memset(__ste_p, 0, PME_SIZE); |
| 53 | } |
| 54 | else { |
| 55 | BUG_ON(pmd & (0x3f00 | xAMPRx_SS | 0xe)); |
| 56 | |
| 57 | for (loop = PME_SIZE; loop > 0; loop -= 4) { |
| 58 | *__ste_p++ = pmd; |
| 59 | pmd += __frv_PT_SIZE; |
| 60 | } |
| 61 | } |
| 62 | |
| 63 | frv_dcache_writeback((unsigned long) pmdptr, (unsigned long) (pmdptr + 1)); |
| 64 | } |
| 65 | |
| 66 | /* |
| 67 | * List of all pgd's needed for non-PAE so it can invalidate entries |
| 68 | * in both cached and uncached pgd's; not needed for PAE since the |
| 69 | * kernel pmd is shared. If PAE were not to share the pmd a similar |
| 70 | * tactic would be needed. This is essentially codepath-based locking |
| 71 | * against pageattr.c; it is the unique case in which a valid change |
| 72 | * of kernel pagetables can't be lazily synchronized by vmalloc faults. |
| 73 | * vmalloc faults work because attached pagetables are never freed. |
| 74 | * If the locking proves to be non-performant, a ticketing scheme with |
| 75 | * checks at dup_mmap(), exec(), and other mmlist addition points |
| 76 | * could be used. The locking scheme was chosen on the basis of |
| 77 | * manfred's recommendations and having no core impact whatsoever. |
| 78 | * -- wli |
| 79 | */ |
| 80 | DEFINE_SPINLOCK(pgd_lock); |
| 81 | struct page *pgd_list; |
| 82 | |
| 83 | static inline void pgd_list_add(pgd_t *pgd) |
| 84 | { |
| 85 | struct page *page = virt_to_page(pgd); |
| 86 | page->index = (unsigned long) pgd_list; |
| 87 | if (pgd_list) |
| 88 | pgd_list->private = (unsigned long) &page->index; |
| 89 | pgd_list = page; |
| 90 | page->private = (unsigned long) &pgd_list; |
| 91 | } |
| 92 | |
| 93 | static inline void pgd_list_del(pgd_t *pgd) |
| 94 | { |
| 95 | struct page *next, **pprev, *page = virt_to_page(pgd); |
| 96 | next = (struct page *) page->index; |
| 97 | pprev = (struct page **) page->private; |
| 98 | *pprev = next; |
| 99 | if (next) |
| 100 | next->private = (unsigned long) pprev; |
| 101 | } |
| 102 | |
| 103 | void pgd_ctor(void *pgd, kmem_cache_t *cache, unsigned long unused) |
| 104 | { |
| 105 | unsigned long flags; |
| 106 | |
| 107 | if (PTRS_PER_PMD == 1) |
| 108 | spin_lock_irqsave(&pgd_lock, flags); |
| 109 | |
| 110 | memcpy((pgd_t *) pgd + USER_PGDS_IN_LAST_PML4, |
| 111 | swapper_pg_dir + USER_PGDS_IN_LAST_PML4, |
| 112 | (PTRS_PER_PGD - USER_PGDS_IN_LAST_PML4) * sizeof(pgd_t)); |
| 113 | |
| 114 | if (PTRS_PER_PMD > 1) |
| 115 | return; |
| 116 | |
| 117 | pgd_list_add(pgd); |
| 118 | spin_unlock_irqrestore(&pgd_lock, flags); |
| 119 | memset(pgd, 0, USER_PGDS_IN_LAST_PML4 * sizeof(pgd_t)); |
| 120 | } |
| 121 | |
| 122 | /* never called when PTRS_PER_PMD > 1 */ |
| 123 | void pgd_dtor(void *pgd, kmem_cache_t *cache, unsigned long unused) |
| 124 | { |
| 125 | unsigned long flags; /* can be called from interrupt context */ |
| 126 | |
| 127 | spin_lock_irqsave(&pgd_lock, flags); |
| 128 | pgd_list_del(pgd); |
| 129 | spin_unlock_irqrestore(&pgd_lock, flags); |
| 130 | } |
| 131 | |
| 132 | pgd_t *pgd_alloc(struct mm_struct *mm) |
| 133 | { |
| 134 | pgd_t *pgd; |
| 135 | |
| 136 | pgd = kmem_cache_alloc(pgd_cache, GFP_KERNEL); |
| 137 | if (!pgd) |
| 138 | return pgd; |
| 139 | |
| 140 | return pgd; |
| 141 | } |
| 142 | |
| 143 | void pgd_free(pgd_t *pgd) |
| 144 | { |
| 145 | /* in the non-PAE case, clear_page_tables() clears user pgd entries */ |
| 146 | kmem_cache_free(pgd_cache, pgd); |
| 147 | } |
| 148 | |
| 149 | void __init pgtable_cache_init(void) |
| 150 | { |
| 151 | pgd_cache = kmem_cache_create("pgd", |
| 152 | PTRS_PER_PGD * sizeof(pgd_t), |
| 153 | PTRS_PER_PGD * sizeof(pgd_t), |
| 154 | 0, |
| 155 | pgd_ctor, |
| 156 | pgd_dtor); |
| 157 | if (!pgd_cache) |
| 158 | panic("pgtable_cache_init(): Cannot create pgd cache"); |
| 159 | } |