blob: 4fb63a36bd52ff0333f9d849a0a75479008aaa5e [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/* 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>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090013#include <linux/gfp.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070014#include <linux/mm.h>
15#include <linux/highmem.h>
Christoph Lameter8defab32007-05-09 02:32:48 -070016#include <linux/quicklist.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070017#include <asm/pgalloc.h>
18#include <asm/page.h>
19#include <asm/cacheflush.h>
20
21pgd_t swapper_pg_dir[PTRS_PER_PGD] __attribute__((aligned(PAGE_SIZE)));
Linus Torvalds1da177e2005-04-16 15:20:36 -070022
23pte_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
Martin Schwidefsky2f569af2008-02-08 04:22:04 -080031pgtable_t pte_alloc_one(struct mm_struct *mm, unsigned long address)
Linus Torvalds1da177e2005-04-16 15:20:36 -070032{
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
Martin Schwidefsky2f569af2008-02-08 04:22:04 -080040 if (page) {
Linus Torvalds1da177e2005-04-16 15:20:36 -070041 clear_highpage(page);
Martin Schwidefsky2f569af2008-02-08 04:22:04 -080042 pgtable_page_ctor(page);
43 flush_dcache_page(page);
44 }
Linus Torvalds1da177e2005-04-16 15:20:36 -070045 return page;
46}
47
48void __set_pmd(pmd_t *pmdptr, unsigned long pmd)
49{
50 unsigned long *__ste_p = pmdptr->ste;
51 int loop;
52
53 if (!pmd) {
54 memset(__ste_p, 0, PME_SIZE);
55 }
56 else {
57 BUG_ON(pmd & (0x3f00 | xAMPRx_SS | 0xe));
58
59 for (loop = PME_SIZE; loop > 0; loop -= 4) {
60 *__ste_p++ = pmd;
61 pmd += __frv_PT_SIZE;
62 }
63 }
64
65 frv_dcache_writeback((unsigned long) pmdptr, (unsigned long) (pmdptr + 1));
66}
67
68/*
69 * List of all pgd's needed for non-PAE so it can invalidate entries
70 * in both cached and uncached pgd's; not needed for PAE since the
71 * kernel pmd is shared. If PAE were not to share the pmd a similar
72 * tactic would be needed. This is essentially codepath-based locking
73 * against pageattr.c; it is the unique case in which a valid change
74 * of kernel pagetables can't be lazily synchronized by vmalloc faults.
75 * vmalloc faults work because attached pagetables are never freed.
76 * If the locking proves to be non-performant, a ticketing scheme with
77 * checks at dup_mmap(), exec(), and other mmlist addition points
78 * could be used. The locking scheme was chosen on the basis of
79 * manfred's recommendations and having no core impact whatsoever.
80 * -- wli
81 */
82DEFINE_SPINLOCK(pgd_lock);
83struct page *pgd_list;
84
85static inline void pgd_list_add(pgd_t *pgd)
86{
87 struct page *page = virt_to_page(pgd);
88 page->index = (unsigned long) pgd_list;
89 if (pgd_list)
David Howells8080f232005-11-28 13:43:51 -080090 set_page_private(pgd_list, (unsigned long) &page->index);
Linus Torvalds1da177e2005-04-16 15:20:36 -070091 pgd_list = page;
Hugh Dickins4c21e2f2005-10-29 18:16:40 -070092 set_page_private(page, (unsigned long)&pgd_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -070093}
94
95static inline void pgd_list_del(pgd_t *pgd)
96{
97 struct page *next, **pprev, *page = virt_to_page(pgd);
98 next = (struct page *) page->index;
David Howells8080f232005-11-28 13:43:51 -080099 pprev = (struct page **) page_private(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100 *pprev = next;
101 if (next)
David Howells8080f232005-11-28 13:43:51 -0800102 set_page_private(next, (unsigned long) pprev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103}
104
Christoph Lameter8defab32007-05-09 02:32:48 -0700105void pgd_ctor(void *pgd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106{
107 unsigned long flags;
108
109 if (PTRS_PER_PMD == 1)
110 spin_lock_irqsave(&pgd_lock, flags);
111
112 memcpy((pgd_t *) pgd + USER_PGDS_IN_LAST_PML4,
113 swapper_pg_dir + USER_PGDS_IN_LAST_PML4,
114 (PTRS_PER_PGD - USER_PGDS_IN_LAST_PML4) * sizeof(pgd_t));
115
116 if (PTRS_PER_PMD > 1)
117 return;
118
119 pgd_list_add(pgd);
120 spin_unlock_irqrestore(&pgd_lock, flags);
121 memset(pgd, 0, USER_PGDS_IN_LAST_PML4 * sizeof(pgd_t));
122}
123
124/* never called when PTRS_PER_PMD > 1 */
Christoph Lameter8defab32007-05-09 02:32:48 -0700125void pgd_dtor(void *pgd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700126{
127 unsigned long flags; /* can be called from interrupt context */
128
129 spin_lock_irqsave(&pgd_lock, flags);
130 pgd_list_del(pgd);
131 spin_unlock_irqrestore(&pgd_lock, flags);
132}
133
134pgd_t *pgd_alloc(struct mm_struct *mm)
135{
Greg Dietschedbfe8982011-07-29 16:46:16 +0100136 return quicklist_alloc(0, GFP_KERNEL, pgd_ctor);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137}
138
Benjamin Herrenschmidt5e541972008-02-04 22:29:14 -0800139void pgd_free(struct mm_struct *mm, pgd_t *pgd)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700140{
141 /* in the non-PAE case, clear_page_tables() clears user pgd entries */
Christoph Lameter8defab32007-05-09 02:32:48 -0700142 quicklist_free(0, pgd_dtor, pgd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143}
144
145void __init pgtable_cache_init(void)
146{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147}
Christoph Lameter8defab32007-05-09 02:32:48 -0700148
149void check_pgt_cache(void)
150{
151 quicklist_trim(0, pgd_dtor, 25, 16);
152}
153