blob: b299724f6e34e9ace068a4c7f39d504a7cc5c7b4 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07002 *
3 * Copyright (C) 1995 Linus Torvalds
4 *
5 * Support of BIGMEM added by Gerhard Wichert, Siemens AG, July 1999
6 */
7
Linus Torvalds1da177e2005-04-16 15:20:36 -07008#include <linux/module.h>
9#include <linux/signal.h>
10#include <linux/sched.h>
11#include <linux/kernel.h>
12#include <linux/errno.h>
13#include <linux/string.h>
14#include <linux/types.h>
15#include <linux/ptrace.h>
16#include <linux/mman.h>
17#include <linux/mm.h>
18#include <linux/hugetlb.h>
19#include <linux/swap.h>
20#include <linux/smp.h>
21#include <linux/init.h>
22#include <linux/highmem.h>
23#include <linux/pagemap.h>
Jeremy Fitzhardingecfb80c92008-12-16 12:17:36 -080024#include <linux/pci.h>
Jan Beulich6fb14752007-05-02 19:27:10 +020025#include <linux/pfn.h>
Randy Dunlapc9cf5522006-06-27 02:53:52 -070026#include <linux/poison.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include <linux/bootmem.h>
Yinghai Lua9ce6bc2010-08-25 13:39:17 -070028#include <linux/memblock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070029#include <linux/proc_fs.h>
Dave Hansen05039b92005-10-29 18:16:57 -070030#include <linux/memory_hotplug.h>
Adrian Bunk27d99f72005-11-13 16:06:51 -080031#include <linux/initrd.h>
Shaohua Li55b23552006-06-23 02:04:49 -070032#include <linux/cpumask.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090033#include <linux/gfp.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070034
H. Peter Anvinf832ff12008-02-04 16:47:58 +010035#include <asm/asm.h>
Ingo Molnar46eaa672008-10-12 15:06:29 +020036#include <asm/bios_ebda.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070037#include <asm/processor.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070038#include <asm/uaccess.h>
39#include <asm/pgtable.h>
40#include <asm/dma.h>
41#include <asm/fixmap.h>
42#include <asm/e820.h>
43#include <asm/apic.h>
Ingo Molnar8550eb92008-01-30 13:34:10 +010044#include <asm/bugs.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070045#include <asm/tlb.h>
46#include <asm/tlbflush.h>
Andres Salomonc10d1e22010-11-17 06:09:52 +000047#include <asm/olpc_ofw.h>
Jeremy Fitzhardingea5a19c62008-01-30 13:33:39 +010048#include <asm/pgalloc.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070049#include <asm/sections.h>
Jeremy Fitzhardingeb239fb22007-05-02 19:27:13 +020050#include <asm/paravirt.h>
Ian Campbell551889a2008-02-09 23:24:09 +010051#include <asm/setup.h>
Harvey Harrison7bfeab92008-02-12 12:12:01 -080052#include <asm/cacheflush.h>
Pekka Enberg2b723942009-04-28 16:00:49 +030053#include <asm/page_types.h>
Pekka Enberg4fcb2082009-03-05 14:55:08 +020054#include <asm/init.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070055
Yinghai Lu5c51bdb2012-11-16 19:39:01 -080056#include "mm_internal.h"
57
Linus Torvalds1da177e2005-04-16 15:20:36 -070058unsigned long highstart_pfn, highend_pfn;
59
Ingo Molnar8550eb92008-01-30 13:34:10 +010060static noinline int do_test_wp_bit(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -070061
Jeremy Fitzhardingedc16ecf2009-03-04 16:10:44 -080062bool __read_mostly __vmalloc_start_set = false;
Yinghai Lu4e296842008-06-24 12:18:14 -070063
Linus Torvalds1da177e2005-04-16 15:20:36 -070064/*
65 * Creates a middle page table and puts a pointer to it in the
66 * given global directory entry. This only returns the gd entry
67 * in non-PAE compilation mode, since the middle layer is folded.
68 */
69static pmd_t * __init one_md_table_init(pgd_t *pgd)
70{
71 pud_t *pud;
72 pmd_t *pmd_table;
Ingo Molnar8550eb92008-01-30 13:34:10 +010073
Linus Torvalds1da177e2005-04-16 15:20:36 -070074#ifdef CONFIG_X86_PAE
Jeremy Fitzhardingeb239fb22007-05-02 19:27:13 +020075 if (!(pgd_val(*pgd) & _PAGE_PRESENT)) {
Yinghai Lu4e37a892012-11-16 19:39:16 -080076 pmd_table = (pmd_t *)alloc_low_page();
Jeremy Fitzhardinge6944a9c2008-03-17 16:37:01 -070077 paravirt_alloc_pmd(&init_mm, __pa(pmd_table) >> PAGE_SHIFT);
Jeremy Fitzhardingeb239fb22007-05-02 19:27:13 +020078 set_pgd(pgd, __pgd(__pa(pmd_table) | _PAGE_PRESENT));
79 pud = pud_offset(pgd, 0);
Ingo Molnar8550eb92008-01-30 13:34:10 +010080 BUG_ON(pmd_table != pmd_offset(pud, 0));
Zhaoleia376f302008-10-31 17:43:04 +080081
82 return pmd_table;
Jeremy Fitzhardingeb239fb22007-05-02 19:27:13 +020083 }
84#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -070085 pud = pud_offset(pgd, 0);
86 pmd_table = pmd_offset(pud, 0);
Ingo Molnar8550eb92008-01-30 13:34:10 +010087
Linus Torvalds1da177e2005-04-16 15:20:36 -070088 return pmd_table;
89}
90
91/*
92 * Create a page table and place a pointer to it in a middle page
Ingo Molnar8550eb92008-01-30 13:34:10 +010093 * directory entry:
Linus Torvalds1da177e2005-04-16 15:20:36 -070094 */
95static pte_t * __init one_page_table_init(pmd_t *pmd)
96{
Jeremy Fitzhardingeb239fb22007-05-02 19:27:13 +020097 if (!(pmd_val(*pmd) & _PAGE_PRESENT)) {
Yinghai Lu4e37a892012-11-16 19:39:16 -080098 pte_t *page_table = (pte_t *)alloc_low_page();
Jeremy Fitzhardingeb239fb22007-05-02 19:27:13 +020099
Jeremy Fitzhardinge6944a9c2008-03-17 16:37:01 -0700100 paravirt_alloc_pte(&init_mm, __pa(page_table) >> PAGE_SHIFT);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101 set_pmd(pmd, __pmd(__pa(page_table) | _PAGE_TABLE));
Jeremy Fitzhardingeb239fb22007-05-02 19:27:13 +0200102 BUG_ON(page_table != pte_offset_kernel(pmd, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103 }
Ingo Molnar509a80c2007-10-17 18:04:34 +0200104
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105 return pte_offset_kernel(pmd, 0);
106}
107
Tejun Heo458a3e62009-02-24 11:57:21 +0900108pmd_t * __init populate_extra_pmd(unsigned long vaddr)
Tejun Heo11124412009-02-20 16:29:09 +0900109{
110 int pgd_idx = pgd_index(vaddr);
111 int pmd_idx = pmd_index(vaddr);
Tejun Heo458a3e62009-02-24 11:57:21 +0900112
113 return one_md_table_init(swapper_pg_dir + pgd_idx) + pmd_idx;
114}
115
116pte_t * __init populate_extra_pte(unsigned long vaddr)
117{
118 int pte_idx = pte_index(vaddr);
Tejun Heo11124412009-02-20 16:29:09 +0900119 pmd_t *pmd;
120
Tejun Heo458a3e62009-02-24 11:57:21 +0900121 pmd = populate_extra_pmd(vaddr);
122 return one_page_table_init(pmd) + pte_idx;
Tejun Heo11124412009-02-20 16:29:09 +0900123}
124
Yinghai Lu719272c2012-11-16 19:39:06 -0800125static unsigned long __init
126page_table_range_init_count(unsigned long start, unsigned long end)
127{
128 unsigned long count = 0;
129#ifdef CONFIG_HIGHMEM
130 int pmd_idx_kmap_begin = fix_to_virt(FIX_KMAP_END) >> PMD_SHIFT;
131 int pmd_idx_kmap_end = fix_to_virt(FIX_KMAP_BEGIN) >> PMD_SHIFT;
132 int pgd_idx, pmd_idx;
133 unsigned long vaddr;
134
135 if (pmd_idx_kmap_begin == pmd_idx_kmap_end)
136 return 0;
137
138 vaddr = start;
139 pgd_idx = pgd_index(vaddr);
140
141 for ( ; (pgd_idx < PTRS_PER_PGD) && (vaddr != end); pgd_idx++) {
142 for (; (pmd_idx < PTRS_PER_PMD) && (vaddr != end);
143 pmd_idx++) {
144 if ((vaddr >> PMD_SHIFT) >= pmd_idx_kmap_begin &&
145 (vaddr >> PMD_SHIFT) <= pmd_idx_kmap_end)
146 count++;
147 vaddr += PMD_SIZE;
148 }
149 pmd_idx = 0;
150 }
151#endif
152 return count;
153}
154
Jan Beulicha3c60182009-01-16 11:59:33 +0000155static pte_t *__init page_table_kmap_check(pte_t *pte, pmd_t *pmd,
Yinghai Lu719272c2012-11-16 19:39:06 -0800156 unsigned long vaddr, pte_t *lastpte,
157 void **adr)
Jan Beulicha3c60182009-01-16 11:59:33 +0000158{
159#ifdef CONFIG_HIGHMEM
160 /*
161 * Something (early fixmap) may already have put a pte
162 * page here, which causes the page table allocation
163 * to become nonlinear. Attempt to fix it, and if it
164 * is still nonlinear then we have to bug.
165 */
166 int pmd_idx_kmap_begin = fix_to_virt(FIX_KMAP_END) >> PMD_SHIFT;
167 int pmd_idx_kmap_end = fix_to_virt(FIX_KMAP_BEGIN) >> PMD_SHIFT;
168
169 if (pmd_idx_kmap_begin != pmd_idx_kmap_end
170 && (vaddr >> PMD_SHIFT) >= pmd_idx_kmap_begin
Yinghai Lu719272c2012-11-16 19:39:06 -0800171 && (vaddr >> PMD_SHIFT) <= pmd_idx_kmap_end) {
Jan Beulicha3c60182009-01-16 11:59:33 +0000172 pte_t *newpte;
173 int i;
174
Pekka Enbergc4645732009-03-05 14:54:58 +0200175 BUG_ON(after_bootmem);
Yinghai Lu719272c2012-11-16 19:39:06 -0800176 newpte = *adr;
Jan Beulicha3c60182009-01-16 11:59:33 +0000177 for (i = 0; i < PTRS_PER_PTE; i++)
178 set_pte(newpte + i, pte[i]);
Yinghai Lu719272c2012-11-16 19:39:06 -0800179 *adr = (void *)(((unsigned long)(*adr)) + PAGE_SIZE);
Jan Beulicha3c60182009-01-16 11:59:33 +0000180
181 paravirt_alloc_pte(&init_mm, __pa(newpte) >> PAGE_SHIFT);
182 set_pmd(pmd, __pmd(__pa(newpte)|_PAGE_TABLE));
183 BUG_ON(newpte != pte_offset_kernel(pmd, 0));
184 __flush_tlb_all();
185
186 paravirt_release_pte(__pa(pte) >> PAGE_SHIFT);
187 pte = newpte;
188 }
189 BUG_ON(vaddr < fix_to_virt(FIX_KMAP_BEGIN - 1)
190 && vaddr > fix_to_virt(FIX_KMAP_END)
191 && lastpte && lastpte + PTRS_PER_PTE != pte);
192#endif
193 return pte;
194}
195
Linus Torvalds1da177e2005-04-16 15:20:36 -0700196/*
Ingo Molnar8550eb92008-01-30 13:34:10 +0100197 * This function initializes a certain range of kernel virtual memory
Linus Torvalds1da177e2005-04-16 15:20:36 -0700198 * with new bootmem page tables, everywhere page tables are missing in
199 * the given range.
Ingo Molnar8550eb92008-01-30 13:34:10 +0100200 *
201 * NOTE: The pagetables are allocated contiguous on the physical space
202 * so we can cache the place of the first one and move around without
Linus Torvalds1da177e2005-04-16 15:20:36 -0700203 * checking the pgd every time.
204 */
Ingo Molnar8550eb92008-01-30 13:34:10 +0100205static void __init
206page_table_range_init(unsigned long start, unsigned long end, pgd_t *pgd_base)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700207{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208 int pgd_idx, pmd_idx;
209 unsigned long vaddr;
Ingo Molnar8550eb92008-01-30 13:34:10 +0100210 pgd_t *pgd;
211 pmd_t *pmd;
Jan Beulicha3c60182009-01-16 11:59:33 +0000212 pte_t *pte = NULL;
Yinghai Lu719272c2012-11-16 19:39:06 -0800213 unsigned long count = page_table_range_init_count(start, end);
214 void *adr = NULL;
215
216 if (count)
217 adr = alloc_low_pages(count);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700218
219 vaddr = start;
220 pgd_idx = pgd_index(vaddr);
221 pmd_idx = pmd_index(vaddr);
222 pgd = pgd_base + pgd_idx;
223
224 for ( ; (pgd_idx < PTRS_PER_PGD) && (vaddr != end); pgd++, pgd_idx++) {
Jeremy Fitzhardingeb239fb22007-05-02 19:27:13 +0200225 pmd = one_md_table_init(pgd);
226 pmd = pmd + pmd_index(vaddr);
Ingo Molnar8550eb92008-01-30 13:34:10 +0100227 for (; (pmd_idx < PTRS_PER_PMD) && (vaddr != end);
228 pmd++, pmd_idx++) {
Jan Beulicha3c60182009-01-16 11:59:33 +0000229 pte = page_table_kmap_check(one_page_table_init(pmd),
Yinghai Lu719272c2012-11-16 19:39:06 -0800230 pmd, vaddr, pte, &adr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231
232 vaddr += PMD_SIZE;
233 }
234 pmd_idx = 0;
235 }
236}
237
238static inline int is_kernel_text(unsigned long addr)
239{
Matthieu Castet5bd5a452010-11-16 22:31:26 +0100240 if (addr >= (unsigned long)_text && addr <= (unsigned long)__init_end)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241 return 1;
242 return 0;
243}
244
245/*
Ingo Molnar8550eb92008-01-30 13:34:10 +0100246 * This maps the physical memory to kernel virtual address space, a total
247 * of max_low_pfn pages, by creating page tables starting from address
248 * PAGE_OFFSET:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249 */
Pekka Enberge53fb042009-03-05 14:55:07 +0200250unsigned long __init
251kernel_physical_mapping_init(unsigned long start,
252 unsigned long end,
253 unsigned long page_size_mask)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254{
Pekka Enberge53fb042009-03-05 14:55:07 +0200255 int use_pse = page_size_mask == (1<<PG_LEVEL_2M);
Pekka Enbergc1fd1b42010-02-24 17:04:47 +0200256 unsigned long last_map_addr = end;
Pekka Enberge53fb042009-03-05 14:55:07 +0200257 unsigned long start_pfn, end_pfn;
Pekka Enberge7179852009-03-05 14:54:55 +0200258 pgd_t *pgd_base = swapper_pg_dir;
Ingo Molnar8550eb92008-01-30 13:34:10 +0100259 int pgd_idx, pmd_idx, pte_ofs;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260 unsigned long pfn;
261 pgd_t *pgd;
262 pmd_t *pmd;
263 pte_t *pte;
Suresh Siddhaa2699e42008-09-23 14:00:38 -0700264 unsigned pages_2m, pages_4k;
265 int mapping_iter;
266
Pekka Enberge53fb042009-03-05 14:55:07 +0200267 start_pfn = start >> PAGE_SHIFT;
268 end_pfn = end >> PAGE_SHIFT;
269
Suresh Siddhaa2699e42008-09-23 14:00:38 -0700270 /*
271 * First iteration will setup identity mapping using large/small pages
272 * based on use_pse, with other attributes same as set by
273 * the early code in head_32.S
274 *
275 * Second iteration will setup the appropriate attributes (NX, GLOBAL..)
276 * as desired for the kernel identity mapping.
277 *
278 * This two pass mechanism conforms to the TLB app note which says:
279 *
280 * "Software should not write to a paging-structure entry in a way
281 * that would change, for any linear address, both the page size
282 * and either the page frame or attributes."
283 */
284 mapping_iter = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285
Yinghai Lua04ad822008-06-29 00:39:06 -0700286 if (!cpu_has_pse)
287 use_pse = 0;
288
Suresh Siddhaa2699e42008-09-23 14:00:38 -0700289repeat:
290 pages_2m = pages_4k = 0;
Yinghai Lua04ad822008-06-29 00:39:06 -0700291 pfn = start_pfn;
292 pgd_idx = pgd_index((pfn<<PAGE_SHIFT) + PAGE_OFFSET);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293 pgd = pgd_base + pgd_idx;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294 for (; pgd_idx < PTRS_PER_PGD; pgd++, pgd_idx++) {
295 pmd = one_md_table_init(pgd);
Ingo Molnar8550eb92008-01-30 13:34:10 +0100296
Yinghai Lua04ad822008-06-29 00:39:06 -0700297 if (pfn >= end_pfn)
298 continue;
299#ifdef CONFIG_X86_PAE
300 pmd_idx = pmd_index((pfn<<PAGE_SHIFT) + PAGE_OFFSET);
301 pmd += pmd_idx;
302#else
303 pmd_idx = 0;
304#endif
305 for (; pmd_idx < PTRS_PER_PMD && pfn < end_pfn;
Jeremy Fitzhardingef3f20de2008-01-30 13:31:09 +0100306 pmd++, pmd_idx++) {
Ingo Molnar8550eb92008-01-30 13:34:10 +0100307 unsigned int addr = pfn * PAGE_SIZE + PAGE_OFFSET;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308
Ingo Molnar8550eb92008-01-30 13:34:10 +0100309 /*
310 * Map with big pages if possible, otherwise
311 * create normal page tables:
312 */
Yinghai Lua04ad822008-06-29 00:39:06 -0700313 if (use_pse) {
Ingo Molnar8550eb92008-01-30 13:34:10 +0100314 unsigned int addr2;
Jeremy Fitzhardingef3f20de2008-01-30 13:31:09 +0100315 pgprot_t prot = PAGE_KERNEL_LARGE;
Suresh Siddhaa2699e42008-09-23 14:00:38 -0700316 /*
317 * first pass will use the same initial
318 * identity mapping attribute + _PAGE_PSE.
319 */
320 pgprot_t init_prot =
321 __pgprot(PTE_IDENT_ATTR |
322 _PAGE_PSE);
Jeremy Fitzhardingef3f20de2008-01-30 13:31:09 +0100323
Yinghai Lu960ddb4f2012-11-16 19:38:54 -0800324 pfn &= PMD_MASK >> PAGE_SHIFT;
Ingo Molnar8550eb92008-01-30 13:34:10 +0100325 addr2 = (pfn + PTRS_PER_PTE-1) * PAGE_SIZE +
Jeremy Fitzhardingef3f20de2008-01-30 13:31:09 +0100326 PAGE_OFFSET + PAGE_SIZE-1;
327
Ingo Molnar8550eb92008-01-30 13:34:10 +0100328 if (is_kernel_text(addr) ||
329 is_kernel_text(addr2))
Jeremy Fitzhardingef3f20de2008-01-30 13:31:09 +0100330 prot = PAGE_KERNEL_LARGE_EXEC;
331
Andi Kleence0c0e52008-05-02 11:46:49 +0200332 pages_2m++;
Suresh Siddhaa2699e42008-09-23 14:00:38 -0700333 if (mapping_iter == 1)
334 set_pmd(pmd, pfn_pmd(pfn, init_prot));
335 else
336 set_pmd(pmd, pfn_pmd(pfn, prot));
Jeremy Fitzhardingeb239fb22007-05-02 19:27:13 +0200337
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338 pfn += PTRS_PER_PTE;
Ingo Molnar8550eb92008-01-30 13:34:10 +0100339 continue;
340 }
341 pte = one_page_table_init(pmd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342
Yinghai Lua04ad822008-06-29 00:39:06 -0700343 pte_ofs = pte_index((pfn<<PAGE_SHIFT) + PAGE_OFFSET);
344 pte += pte_ofs;
345 for (; pte_ofs < PTRS_PER_PTE && pfn < end_pfn;
Ingo Molnar8550eb92008-01-30 13:34:10 +0100346 pte++, pfn++, pte_ofs++, addr += PAGE_SIZE) {
347 pgprot_t prot = PAGE_KERNEL;
Suresh Siddhaa2699e42008-09-23 14:00:38 -0700348 /*
349 * first pass will use the same initial
350 * identity mapping attribute.
351 */
352 pgprot_t init_prot = __pgprot(PTE_IDENT_ATTR);
Jeremy Fitzhardingef3f20de2008-01-30 13:31:09 +0100353
Ingo Molnar8550eb92008-01-30 13:34:10 +0100354 if (is_kernel_text(addr))
355 prot = PAGE_KERNEL_EXEC;
Jeremy Fitzhardingef3f20de2008-01-30 13:31:09 +0100356
Andi Kleence0c0e52008-05-02 11:46:49 +0200357 pages_4k++;
Pekka Enbergc1fd1b42010-02-24 17:04:47 +0200358 if (mapping_iter == 1) {
Suresh Siddhaa2699e42008-09-23 14:00:38 -0700359 set_pte(pte, pfn_pte(pfn, init_prot));
Pekka Enbergc1fd1b42010-02-24 17:04:47 +0200360 last_map_addr = (pfn << PAGE_SHIFT) + PAGE_SIZE;
361 } else
Suresh Siddhaa2699e42008-09-23 14:00:38 -0700362 set_pte(pte, pfn_pte(pfn, prot));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363 }
364 }
365 }
Suresh Siddhaa2699e42008-09-23 14:00:38 -0700366 if (mapping_iter == 1) {
367 /*
368 * update direct mapping page count only in the first
369 * iteration.
370 */
371 update_page_count(PG_LEVEL_2M, pages_2m);
372 update_page_count(PG_LEVEL_4K, pages_4k);
373
374 /*
375 * local global flush tlb, which will flush the previous
376 * mappings present in both small and large page TLB's.
377 */
378 __flush_tlb_all();
379
380 /*
381 * Second iteration will set the actual desired PTE attributes.
382 */
383 mapping_iter = 2;
384 goto repeat;
385 }
Pekka Enbergc1fd1b42010-02-24 17:04:47 +0200386 return last_map_addr;
Arjan van de Venae531c22008-04-24 23:40:47 +0200387}
388
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389pte_t *kmap_pte;
390pgprot_t kmap_prot;
391
Ingo Molnar8550eb92008-01-30 13:34:10 +0100392static inline pte_t *kmap_get_fixmap_pte(unsigned long vaddr)
393{
394 return pte_offset_kernel(pmd_offset(pud_offset(pgd_offset_k(vaddr),
395 vaddr), vaddr), vaddr);
396}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397
398static void __init kmap_init(void)
399{
400 unsigned long kmap_vstart;
401
Ingo Molnar8550eb92008-01-30 13:34:10 +0100402 /*
403 * Cache the first kmap pte:
404 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405 kmap_vstart = __fix_to_virt(FIX_KMAP_BEGIN);
406 kmap_pte = kmap_get_fixmap_pte(kmap_vstart);
407
408 kmap_prot = PAGE_KERNEL;
409}
410
Keith Packardfd940932008-10-30 19:37:09 -0700411#ifdef CONFIG_HIGHMEM
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412static void __init permanent_kmaps_init(pgd_t *pgd_base)
413{
Ingo Molnar8550eb92008-01-30 13:34:10 +0100414 unsigned long vaddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415 pgd_t *pgd;
416 pud_t *pud;
417 pmd_t *pmd;
418 pte_t *pte;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419
420 vaddr = PKMAP_BASE;
421 page_table_range_init(vaddr, vaddr + PAGE_SIZE*LAST_PKMAP, pgd_base);
422
423 pgd = swapper_pg_dir + pgd_index(vaddr);
424 pud = pud_offset(pgd, vaddr);
425 pmd = pmd_offset(pud, vaddr);
426 pte = pte_offset_kernel(pmd, vaddr);
Ingo Molnar8550eb92008-01-30 13:34:10 +0100427 pkmap_page_table = pte;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428}
429
Minchan Kimb1258ac2009-10-22 11:27:22 +0900430static void __init add_one_highpage_init(struct page *page)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431{
Yinghai Lucc9f7a02008-06-16 16:11:08 -0700432 ClearPageReserved(page);
433 init_page_count(page);
434 __free_page(page);
435 totalhigh_pages++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436}
437
Yinghai Lu1d931262010-10-05 16:15:15 -0700438void __init add_highpages_with_active_regions(int nid,
439 unsigned long start_pfn, unsigned long end_pfn)
Yinghai Lub5bc6c02008-06-14 18:32:52 -0700440{
Tejun Heo8a9ca342011-07-12 11:16:02 +0200441 phys_addr_t start, end;
442 u64 i;
Yinghai Lub5bc6c02008-06-14 18:32:52 -0700443
Tejun Heo8a9ca342011-07-12 11:16:02 +0200444 for_each_free_mem_range(i, nid, &start, &end, NULL) {
445 unsigned long pfn = clamp_t(unsigned long, PFN_UP(start),
446 start_pfn, end_pfn);
447 unsigned long e_pfn = clamp_t(unsigned long, PFN_DOWN(end),
448 start_pfn, end_pfn);
449 for ( ; pfn < e_pfn; pfn++)
450 if (pfn_valid(pfn))
451 add_one_highpage_init(pfn_to_page(pfn));
Yinghai Lub5bc6c02008-06-14 18:32:52 -0700452 }
Yinghai Lub5bc6c02008-06-14 18:32:52 -0700453}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454#else
Ingo Brueckle8e32322009-01-02 14:42:00 +0100455static inline void permanent_kmaps_init(pgd_t *pgd_base)
456{
457}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458#endif /* CONFIG_HIGHMEM */
459
Attilio Rao7737b212012-08-21 21:22:38 +0100460void __init native_pagetable_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461{
Ian Campbell551889a2008-02-09 23:24:09 +0100462 unsigned long pfn, va;
Attilio Rao73090f82012-08-21 21:22:37 +0100463 pgd_t *pgd, *base = swapper_pg_dir;
Ian Campbell551889a2008-02-09 23:24:09 +0100464 pud_t *pud;
465 pmd_t *pmd;
466 pte_t *pte;
Jeremy Fitzhardingeb239fb22007-05-02 19:27:13 +0200467
468 /*
Ian Campbell551889a2008-02-09 23:24:09 +0100469 * Remove any mappings which extend past the end of physical
Yinghai Lu11ed9e92012-11-16 19:39:11 -0800470 * memory from the boot time page table.
471 * In virtual address space, we should have at least two pages
472 * from VMALLOC_END to pkmap or fixmap according to VMALLOC_END
473 * definition. And max_low_pfn is set to VMALLOC_END physical
474 * address. If initial memory mapping is doing right job, we
475 * should have pte used near max_low_pfn or one pmd is not present.
Jeremy Fitzhardingeb239fb22007-05-02 19:27:13 +0200476 */
Yinghai Lu11ed9e92012-11-16 19:39:11 -0800477 for (pfn = max_low_pfn; pfn < 1<<(32-PAGE_SHIFT); pfn++) {
Ian Campbell551889a2008-02-09 23:24:09 +0100478 va = PAGE_OFFSET + (pfn<<PAGE_SHIFT);
479 pgd = base + pgd_index(va);
480 if (!pgd_present(*pgd))
481 break;
Jeremy Fitzhardingeb239fb22007-05-02 19:27:13 +0200482
Ian Campbell551889a2008-02-09 23:24:09 +0100483 pud = pud_offset(pgd, va);
484 pmd = pmd_offset(pud, va);
485 if (!pmd_present(*pmd))
486 break;
487
Yinghai Lu11ed9e92012-11-16 19:39:11 -0800488 /* should not be large page here */
489 if (pmd_large(*pmd)) {
490 pr_warn("try to clear pte for ram above max_low_pfn: pfn: %lx pmd: %p pmd phys: %lx, but pmd is big page and is not using pte !\n",
491 pfn, pmd, __pa(pmd));
492 BUG_ON(1);
493 }
494
Ian Campbell551889a2008-02-09 23:24:09 +0100495 pte = pte_offset_kernel(pmd, va);
496 if (!pte_present(*pte))
497 break;
498
Yinghai Lu11ed9e92012-11-16 19:39:11 -0800499 printk(KERN_DEBUG "clearing pte for ram above max_low_pfn: pfn: %lx pmd: %p pmd phys: %lx pte: %p pte phys: %lx\n",
500 pfn, pmd, __pa(pmd), pte, __pa(pte));
Ian Campbell551889a2008-02-09 23:24:09 +0100501 pte_clear(NULL, va, pte);
502 }
Jeremy Fitzhardinge6944a9c2008-03-17 16:37:01 -0700503 paravirt_alloc_pmd(&init_mm, __pa(base) >> PAGE_SHIFT);
Attilio Rao843b8ed2012-08-21 21:22:39 +0100504 paging_init();
Jeremy Fitzhardingeb239fb22007-05-02 19:27:13 +0200505}
506
Jeremy Fitzhardingeb239fb22007-05-02 19:27:13 +0200507/*
508 * Build a proper pagetable for the kernel mappings. Up until this
509 * point, we've been running on some set of pagetables constructed by
510 * the boot process.
511 *
512 * If we're booting on native hardware, this will be a pagetable
Ian Campbell551889a2008-02-09 23:24:09 +0100513 * constructed in arch/x86/kernel/head_32.S. The root of the
514 * pagetable will be swapper_pg_dir.
Jeremy Fitzhardingeb239fb22007-05-02 19:27:13 +0200515 *
516 * If we're booting paravirtualized under a hypervisor, then there are
517 * more options: we may already be running PAE, and the pagetable may
518 * or may not be based in swapper_pg_dir. In any case,
Attilio Rao7737b212012-08-21 21:22:38 +0100519 * paravirt_pagetable_init() will set up swapper_pg_dir
Jeremy Fitzhardingeb239fb22007-05-02 19:27:13 +0200520 * appropriately for the rest of the initialization to work.
521 *
522 * In general, pagetable_init() assumes that the pagetable may already
523 * be partially populated, and so it avoids stomping on any existing
524 * mappings.
525 */
Pekka Enbergf7650902009-03-05 14:55:05 +0200526void __init early_ioremap_page_table_range_init(void)
Jeremy Fitzhardingeb239fb22007-05-02 19:27:13 +0200527{
Pekka Enberge7179852009-03-05 14:54:55 +0200528 pgd_t *pgd_base = swapper_pg_dir;
Ingo Molnar8550eb92008-01-30 13:34:10 +0100529 unsigned long vaddr, end;
Jeremy Fitzhardingeb239fb22007-05-02 19:27:13 +0200530
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531 /*
532 * Fixed mappings, only the page table structure has to be
533 * created - mappings will be set by set_fixmap():
534 */
535 vaddr = __fix_to_virt(__end_of_fixed_addresses - 1) & PMD_MASK;
Jeremy Fitzhardingeb239fb22007-05-02 19:27:13 +0200536 end = (FIXADDR_TOP + PMD_SIZE - 1) & PMD_MASK;
537 page_table_range_init(vaddr, end, pgd_base);
Huang, Yingbeacfaa2008-01-30 13:33:44 +0100538 early_ioremap_reset();
Yinghai Lue7b37892008-06-25 21:51:28 -0700539}
540
541static void __init pagetable_init(void)
542{
543 pgd_t *pgd_base = swapper_pg_dir;
544
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545 permanent_kmaps_init(pgd_base);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546}
547
Jeremy Fitzhardingebe43d722008-09-07 15:21:13 -0700548pteval_t __supported_pte_mask __read_mostly = ~(_PAGE_NX | _PAGE_GLOBAL | _PAGE_IOMAP);
Jeremy Fitzhardinge6fdc05d2008-01-30 13:32:57 +0100549EXPORT_SYMBOL_GPL(__supported_pte_mask);
550
Yinghai Lu90d967e2008-06-23 21:00:45 +0200551/* user-defined highmem size */
552static unsigned int highmem_pages = -1;
553
554/*
555 * highmem=size forces highmem to be exactly 'size' bytes.
556 * This works even on boxes that have no highmem otherwise.
557 * This also works to reduce highmem size on bigger boxes.
558 */
559static int __init parse_highmem(char *arg)
560{
561 if (!arg)
562 return -EINVAL;
563
564 highmem_pages = memparse(arg, &arg) >> PAGE_SHIFT;
565 return 0;
566}
567early_param("highmem", parse_highmem);
568
Ingo Molnar47698432009-02-12 13:31:41 +0100569#define MSG_HIGHMEM_TOO_BIG \
570 "highmem size (%luMB) is bigger than pages available (%luMB)!\n"
571
572#define MSG_LOWMEM_TOO_SMALL \
573 "highmem size (%luMB) results in <64MB lowmem, ignoring it!\n"
574/*
575 * All of RAM fits into lowmem - but if user wants highmem
576 * artificially via the highmem=x boot parameter then create
577 * it:
578 */
Yinghai Luf836e352012-11-16 19:39:09 -0800579static void __init lowmem_pfn_init(void)
Ingo Molnar47698432009-02-12 13:31:41 +0100580{
Ingo Molnard88316c22009-02-12 15:16:03 +0100581 /* max_low_pfn is 0, we already have early_res support */
582 max_low_pfn = max_pfn;
583
Ingo Molnar47698432009-02-12 13:31:41 +0100584 if (highmem_pages == -1)
585 highmem_pages = 0;
586#ifdef CONFIG_HIGHMEM
587 if (highmem_pages >= max_pfn) {
588 printk(KERN_ERR MSG_HIGHMEM_TOO_BIG,
589 pages_to_mb(highmem_pages), pages_to_mb(max_pfn));
590 highmem_pages = 0;
591 }
592 if (highmem_pages) {
593 if (max_low_pfn - highmem_pages < 64*1024*1024/PAGE_SIZE) {
594 printk(KERN_ERR MSG_LOWMEM_TOO_SMALL,
595 pages_to_mb(highmem_pages));
596 highmem_pages = 0;
597 }
598 max_low_pfn -= highmem_pages;
599 }
600#else
601 if (highmem_pages)
602 printk(KERN_ERR "ignoring highmem size on non-highmem kernel!\n");
603#endif
604}
605
606#define MSG_HIGHMEM_TOO_SMALL \
607 "only %luMB highmem pages available, ignoring highmem size of %luMB!\n"
608
609#define MSG_HIGHMEM_TRIMMED \
610 "Warning: only 4GB will be used. Use a HIGHMEM64G enabled kernel!\n"
611/*
612 * We have more RAM than fits into lowmem - we try to put it into
613 * highmem, also taking the highmem=x boot parameter into account:
614 */
Yinghai Luf836e352012-11-16 19:39:09 -0800615static void __init highmem_pfn_init(void)
Ingo Molnar47698432009-02-12 13:31:41 +0100616{
Ingo Molnard88316c22009-02-12 15:16:03 +0100617 max_low_pfn = MAXMEM_PFN;
618
Ingo Molnar47698432009-02-12 13:31:41 +0100619 if (highmem_pages == -1)
620 highmem_pages = max_pfn - MAXMEM_PFN;
621
622 if (highmem_pages + MAXMEM_PFN < max_pfn)
623 max_pfn = MAXMEM_PFN + highmem_pages;
624
625 if (highmem_pages + MAXMEM_PFN > max_pfn) {
626 printk(KERN_WARNING MSG_HIGHMEM_TOO_SMALL,
627 pages_to_mb(max_pfn - MAXMEM_PFN),
628 pages_to_mb(highmem_pages));
629 highmem_pages = 0;
630 }
Ingo Molnar47698432009-02-12 13:31:41 +0100631#ifndef CONFIG_HIGHMEM
632 /* Maximum memory usable is what is directly addressable */
633 printk(KERN_WARNING "Warning only %ldMB will be used.\n", MAXMEM>>20);
634 if (max_pfn > MAX_NONPAE_PFN)
635 printk(KERN_WARNING "Use a HIGHMEM64G enabled kernel.\n");
636 else
637 printk(KERN_WARNING "Use a HIGHMEM enabled kernel.\n");
638 max_pfn = MAXMEM_PFN;
639#else /* !CONFIG_HIGHMEM */
640#ifndef CONFIG_HIGHMEM64G
641 if (max_pfn > MAX_NONPAE_PFN) {
642 max_pfn = MAX_NONPAE_PFN;
643 printk(KERN_WARNING MSG_HIGHMEM_TRIMMED);
644 }
645#endif /* !CONFIG_HIGHMEM64G */
646#endif /* !CONFIG_HIGHMEM */
647}
648
Yinghai Lu90d967e2008-06-23 21:00:45 +0200649/*
650 * Determine low and high memory ranges:
651 */
Yinghai Lu2ec65f82008-06-23 03:05:30 -0700652void __init find_low_pfn_range(void)
Yinghai Lu90d967e2008-06-23 21:00:45 +0200653{
Yinghai Lu2ec65f82008-06-23 03:05:30 -0700654 /* it could update max_pfn */
655
Ingo Molnard88316c22009-02-12 15:16:03 +0100656 if (max_pfn <= MAXMEM_PFN)
Ingo Molnar47698432009-02-12 13:31:41 +0100657 lowmem_pfn_init();
Ingo Molnard88316c22009-02-12 15:16:03 +0100658 else
659 highmem_pfn_init();
Yinghai Lu90d967e2008-06-23 21:00:45 +0200660}
661
Yinghai Lub2ac82a2008-06-22 02:45:39 -0700662#ifndef CONFIG_NEED_MULTIPLE_NODES
Tejun Heod8fc3af2011-02-16 12:13:06 +0100663void __init initmem_init(void)
Yinghai Lub2ac82a2008-06-22 02:45:39 -0700664{
Yinghai Lub2ac82a2008-06-22 02:45:39 -0700665#ifdef CONFIG_HIGHMEM
666 highstart_pfn = highend_pfn = max_pfn;
667 if (max_pfn > max_low_pfn)
668 highstart_pfn = max_low_pfn;
Yinghai Lub2ac82a2008-06-22 02:45:39 -0700669 printk(KERN_NOTICE "%ldMB HIGHMEM available.\n",
670 pages_to_mb(highend_pfn - highstart_pfn));
671 num_physpages = highend_pfn;
672 high_memory = (void *) __va(highstart_pfn * PAGE_SIZE - 1) + 1;
673#else
Yinghai Lub2ac82a2008-06-22 02:45:39 -0700674 num_physpages = max_low_pfn;
675 high_memory = (void *) __va(max_low_pfn * PAGE_SIZE - 1) + 1;
676#endif
Tejun Heo0608f702011-07-14 11:44:23 +0200677
678 memblock_set_node(0, (phys_addr_t)ULLONG_MAX, 0);
679 sparse_memory_present_with_active_regions(0);
680
Yinghai Lub2ac82a2008-06-22 02:45:39 -0700681#ifdef CONFIG_FLATMEM
682 max_mapnr = num_physpages;
683#endif
Jeremy Fitzhardingedc16ecf2009-03-04 16:10:44 -0800684 __vmalloc_start_set = true;
685
Yinghai Lub2ac82a2008-06-22 02:45:39 -0700686 printk(KERN_NOTICE "%ldMB LOWMEM available.\n",
687 pages_to_mb(max_low_pfn));
688
689 setup_bootmem_allocator();
Yinghai Lub2ac82a2008-06-22 02:45:39 -0700690}
Yinghai Lucb95a132008-07-02 00:31:02 -0700691#endif /* !CONFIG_NEED_MULTIPLE_NODES */
Yinghai Lub2ac82a2008-06-22 02:45:39 -0700692
Yinghai Lub2ac82a2008-06-22 02:45:39 -0700693void __init setup_bootmem_allocator(void)
694{
Yinghai Lub2ac82a2008-06-22 02:45:39 -0700695 printk(KERN_INFO " mapped low ram: 0 - %08lx\n",
696 max_pfn_mapped<<PAGE_SHIFT);
Yinghai Lufc5efe32009-03-04 12:21:24 -0800697 printk(KERN_INFO " low ram: 0 - %08lx\n", max_low_pfn<<PAGE_SHIFT);
Yinghai Lu4e296842008-06-24 12:18:14 -0700698}
699
Linus Torvalds1da177e2005-04-16 15:20:36 -0700700/*
701 * paging_init() sets up the page tables - note that the first 8MB are
702 * already mapped by head.S.
703 *
704 * This routines also unmaps the page at virtual kernel address 0, so
705 * that we can trap those pesky NULL-reference errors in the kernel.
706 */
707void __init paging_init(void)
708{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700709 pagetable_init();
710
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711 __flush_tlb_all();
712
713 kmap_init();
Yinghai Lu11cd0bc2008-06-23 19:51:10 -0700714
715 /*
716 * NOTE: at this point the bootmem allocator is fully available.
717 */
Andres Salomonc10d1e22010-11-17 06:09:52 +0000718 olpc_dt_build_devicetree();
Tejun Heo797390d2011-05-02 14:18:52 +0200719 sparse_memory_present_with_active_regions(MAX_NUMNODES);
Yinghai Lu11cd0bc2008-06-23 19:51:10 -0700720 sparse_init();
721 zone_sizes_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722}
723
724/*
725 * Test if the WP bit works in supervisor mode. It isn't supported on 386's
Dmitri Vorobievf7f17a62008-04-21 00:47:55 +0400726 * and also on some strange 486's. All 586+'s are OK. This used to involve
727 * black magic jumps to work around some nasty CPU bugs, but fortunately the
728 * switch to using exceptions got rid of all that.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730static void __init test_wp_bit(void)
731{
Ingo Molnard7d119d2008-01-30 13:34:10 +0100732 printk(KERN_INFO
733 "Checking if this processor honours the WP bit even in supervisor mode...");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734
735 /* Any page-aligned address will do, the test is non-destructive */
H. Peter Anvin8bd753b2012-09-21 12:43:06 -0700736 __set_fixmap(FIX_WP_TEST, __pa(&swapper_pg_dir), PAGE_KERNEL_RO);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700737 boot_cpu_data.wp_works_ok = do_test_wp_bit();
738 clear_fixmap(FIX_WP_TEST);
739
740 if (!boot_cpu_data.wp_works_ok) {
Ingo Molnard7d119d2008-01-30 13:34:10 +0100741 printk(KERN_CONT "No.\n");
H. Peter Anvina5c2a892012-11-28 11:50:28 -0800742 panic("Linux doesn't support CPUs with broken WP.");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743 } else {
Ingo Molnard7d119d2008-01-30 13:34:10 +0100744 printk(KERN_CONT "Ok.\n");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745 }
746}
747
Linus Torvalds1da177e2005-04-16 15:20:36 -0700748void __init mem_init(void)
749{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750 int codesize, reservedpages, datasize, initsize;
Yinghai Lucc9f7a02008-06-16 16:11:08 -0700751 int tmp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752
Jeremy Fitzhardingecfb80c92008-12-16 12:17:36 -0800753 pci_iommu_alloc();
754
Andy Whitcroft05b79bd2005-06-23 00:07:57 -0700755#ifdef CONFIG_FLATMEM
Eric Sesterhenn8d8f3cb2006-10-03 23:34:58 +0200756 BUG_ON(!mem_map);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757#endif
Stanislaw Gruszka855c7432011-12-06 09:08:34 +0100758 /*
759 * With CONFIG_DEBUG_PAGEALLOC initialization of highmem pages has to
760 * be done before free_all_bootmem(). Memblock use free low memory for
761 * temporary data (see find_range_array()) and for this purpose can use
762 * pages that was already passed to the buddy allocator, hence marked as
763 * not accessible in the page tables when compiled with
764 * CONFIG_DEBUG_PAGEALLOC. Otherwise order of initialization is not
765 * important here.
766 */
767 set_highmem_pages_init();
768
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769 /* this will put all low memory onto the freelists */
770 totalram_pages += free_all_bootmem();
771
772 reservedpages = 0;
773 for (tmp = 0; tmp < max_low_pfn; tmp++)
774 /*
Ingo Molnar8550eb92008-01-30 13:34:10 +0100775 * Only count reserved RAM pages:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776 */
777 if (page_is_ram(tmp) && PageReserved(pfn_to_page(tmp)))
778 reservedpages++;
779
Yinghai Lu4e37a892012-11-16 19:39:16 -0800780 after_bootmem = 1;
781
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782 codesize = (unsigned long) &_etext - (unsigned long) &_text;
783 datasize = (unsigned long) &_edata - (unsigned long) &_etext;
784 initsize = (unsigned long) &__init_end - (unsigned long) &__init_begin;
785
Ingo Molnar8550eb92008-01-30 13:34:10 +0100786 printk(KERN_INFO "Memory: %luk/%luk available (%dk kernel code, "
787 "%dk reserved, %dk data, %dk init, %ldk highmem)\n",
Geert Uytterhoevencc013a82009-09-21 17:02:36 -0700788 nr_free_pages() << (PAGE_SHIFT-10),
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789 num_physpages << (PAGE_SHIFT-10),
790 codesize >> 10,
791 reservedpages << (PAGE_SHIFT-10),
792 datasize >> 10,
793 initsize >> 10,
Andreas Fenkart4b529402010-01-08 14:42:31 -0800794 totalhigh_pages << (PAGE_SHIFT-10));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795
Ingo Molnard7d119d2008-01-30 13:34:10 +0100796 printk(KERN_INFO "virtual kernel memory layout:\n"
Ingo Molnar8550eb92008-01-30 13:34:10 +0100797 " fixmap : 0x%08lx - 0x%08lx (%4ld kB)\n"
Jeremy Fitzhardinge052e7992006-09-25 23:32:25 -0700798#ifdef CONFIG_HIGHMEM
Ingo Molnar8550eb92008-01-30 13:34:10 +0100799 " pkmap : 0x%08lx - 0x%08lx (%4ld kB)\n"
Jeremy Fitzhardinge052e7992006-09-25 23:32:25 -0700800#endif
Ingo Molnar8550eb92008-01-30 13:34:10 +0100801 " vmalloc : 0x%08lx - 0x%08lx (%4ld MB)\n"
802 " lowmem : 0x%08lx - 0x%08lx (%4ld MB)\n"
803 " .init : 0x%08lx - 0x%08lx (%4ld kB)\n"
804 " .data : 0x%08lx - 0x%08lx (%4ld kB)\n"
805 " .text : 0x%08lx - 0x%08lx (%4ld kB)\n",
806 FIXADDR_START, FIXADDR_TOP,
807 (FIXADDR_TOP - FIXADDR_START) >> 10,
Jeremy Fitzhardinge052e7992006-09-25 23:32:25 -0700808
809#ifdef CONFIG_HIGHMEM
Ingo Molnar8550eb92008-01-30 13:34:10 +0100810 PKMAP_BASE, PKMAP_BASE+LAST_PKMAP*PAGE_SIZE,
811 (LAST_PKMAP*PAGE_SIZE) >> 10,
Jeremy Fitzhardinge052e7992006-09-25 23:32:25 -0700812#endif
813
Ingo Molnar8550eb92008-01-30 13:34:10 +0100814 VMALLOC_START, VMALLOC_END,
815 (VMALLOC_END - VMALLOC_START) >> 20,
Jeremy Fitzhardinge052e7992006-09-25 23:32:25 -0700816
Ingo Molnar8550eb92008-01-30 13:34:10 +0100817 (unsigned long)__va(0), (unsigned long)high_memory,
818 ((unsigned long)high_memory - (unsigned long)__va(0)) >> 20,
Jeremy Fitzhardinge052e7992006-09-25 23:32:25 -0700819
Ingo Molnar8550eb92008-01-30 13:34:10 +0100820 (unsigned long)&__init_begin, (unsigned long)&__init_end,
821 ((unsigned long)&__init_end -
822 (unsigned long)&__init_begin) >> 10,
Jeremy Fitzhardinge052e7992006-09-25 23:32:25 -0700823
Ingo Molnar8550eb92008-01-30 13:34:10 +0100824 (unsigned long)&_etext, (unsigned long)&_edata,
825 ((unsigned long)&_edata - (unsigned long)&_etext) >> 10,
Jeremy Fitzhardinge052e7992006-09-25 23:32:25 -0700826
Ingo Molnar8550eb92008-01-30 13:34:10 +0100827 (unsigned long)&_text, (unsigned long)&_etext,
828 ((unsigned long)&_etext - (unsigned long)&_text) >> 10);
Jeremy Fitzhardinge052e7992006-09-25 23:32:25 -0700829
Jan Beulichbeeb4192008-12-16 11:45:56 +0000830 /*
831 * Check boundaries twice: Some fundamental inconsistencies can
832 * be detected at build time already.
833 */
834#define __FIXADDR_TOP (-PAGE_SIZE)
835#ifdef CONFIG_HIGHMEM
836 BUILD_BUG_ON(PKMAP_BASE + LAST_PKMAP*PAGE_SIZE > FIXADDR_START);
837 BUILD_BUG_ON(VMALLOC_END > PKMAP_BASE);
838#endif
839#define high_memory (-128UL << 20)
840 BUILD_BUG_ON(VMALLOC_START >= VMALLOC_END);
841#undef high_memory
842#undef __FIXADDR_TOP
843
Jeremy Fitzhardinge052e7992006-09-25 23:32:25 -0700844#ifdef CONFIG_HIGHMEM
Ingo Molnar8550eb92008-01-30 13:34:10 +0100845 BUG_ON(PKMAP_BASE + LAST_PKMAP*PAGE_SIZE > FIXADDR_START);
846 BUG_ON(VMALLOC_END > PKMAP_BASE);
Jeremy Fitzhardinge052e7992006-09-25 23:32:25 -0700847#endif
Jan Beulichbeeb4192008-12-16 11:45:56 +0000848 BUG_ON(VMALLOC_START >= VMALLOC_END);
Ingo Molnar8550eb92008-01-30 13:34:10 +0100849 BUG_ON((unsigned long)high_memory > VMALLOC_START);
Jeremy Fitzhardinge052e7992006-09-25 23:32:25 -0700850
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851 if (boot_cpu_data.wp_works_ok < 0)
852 test_wp_bit();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853}
854
KAMEZAWA Hiroyukiad8f5792006-05-20 15:00:03 -0700855#ifdef CONFIG_MEMORY_HOTPLUG
Yasunori Gotobc02af92006-06-27 02:53:30 -0700856int arch_add_memory(int nid, u64 start, u64 size)
Dave Hansen05039b92005-10-29 18:16:57 -0700857{
Yasunori Goto7c7e9422006-12-22 01:11:13 -0800858 struct pglist_data *pgdata = NODE_DATA(nid);
Christoph Lameter776ed982006-09-25 23:31:09 -0700859 struct zone *zone = pgdata->node_zones + ZONE_HIGHMEM;
Dave Hansen05039b92005-10-29 18:16:57 -0700860 unsigned long start_pfn = start >> PAGE_SHIFT;
861 unsigned long nr_pages = size >> PAGE_SHIFT;
862
Gary Hadec04fc582009-01-06 14:39:14 -0800863 return __add_pages(nid, zone, start_pfn, nr_pages);
Dave Hansen05039b92005-10-29 18:16:57 -0700864}
Andi Kleen9d99aaa2006-04-07 19:49:15 +0200865#endif
Dave Hansen05039b92005-10-29 18:16:57 -0700866
Linus Torvalds1da177e2005-04-16 15:20:36 -0700867/*
868 * This function cannot be __init, since exceptions don't work in that
869 * section. Put this after the callers, so that it cannot be inlined.
870 */
Ingo Molnar8550eb92008-01-30 13:34:10 +0100871static noinline int do_test_wp_bit(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872{
873 char tmp_reg;
874 int flag;
875
876 __asm__ __volatile__(
Ingo Molnar8550eb92008-01-30 13:34:10 +0100877 " movb %0, %1 \n"
878 "1: movb %1, %0 \n"
879 " xorl %2, %2 \n"
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880 "2: \n"
H. Peter Anvinf832ff12008-02-04 16:47:58 +0100881 _ASM_EXTABLE(1b,2b)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700882 :"=m" (*(char *)fix_to_virt(FIX_WP_TEST)),
883 "=q" (tmp_reg),
884 "=r" (flag)
885 :"2" (1)
886 :"memory");
Ingo Molnar8550eb92008-01-30 13:34:10 +0100887
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888 return flag;
889}
890
Arjan van de Ven63aaf302006-01-06 00:12:02 -0800891#ifdef CONFIG_DEBUG_RODATA
Arjan van de Venedeed302008-01-30 13:34:08 +0100892const int rodata_test_data = 0xC3;
893EXPORT_SYMBOL_GPL(rodata_test_data);
Arjan van de Ven63aaf302006-01-06 00:12:02 -0800894
Suresh Siddha502f6602009-10-28 18:46:56 -0800895int kernel_set_to_readonly __read_mostly;
Steven Rostedt16239632009-02-17 17:57:30 -0500896
897void set_kernel_text_rw(void)
898{
899 unsigned long start = PFN_ALIGN(_text);
900 unsigned long size = PFN_ALIGN(_etext) - start;
901
902 if (!kernel_set_to_readonly)
903 return;
904
905 pr_debug("Set kernel text: %lx - %lx for read write\n",
906 start, start+size);
907
908 set_pages_rw(virt_to_page(start), size >> PAGE_SHIFT);
909}
910
911void set_kernel_text_ro(void)
912{
913 unsigned long start = PFN_ALIGN(_text);
914 unsigned long size = PFN_ALIGN(_etext) - start;
915
916 if (!kernel_set_to_readonly)
917 return;
918
919 pr_debug("Set kernel text: %lx - %lx for read only\n",
920 start, start+size);
921
922 set_pages_ro(virt_to_page(start), size >> PAGE_SHIFT);
923}
924
Matthieu Castet5bd5a452010-11-16 22:31:26 +0100925static void mark_nxdata_nx(void)
926{
927 /*
928 * When this called, init has already been executed and released,
Lucas De Marchi0d2eb442011-03-17 16:24:16 -0300929 * so everything past _etext should be NX.
Matthieu Castet5bd5a452010-11-16 22:31:26 +0100930 */
931 unsigned long start = PFN_ALIGN(_etext);
932 /*
933 * This comes from is_kernel_text upper limit. Also HPAGE where used:
934 */
935 unsigned long size = (((unsigned long)__init_end + HPAGE_SIZE) & HPAGE_MASK) - start;
936
937 if (__supported_pte_mask & _PAGE_NX)
938 printk(KERN_INFO "NX-protecting the kernel data: %luk\n", size >> 10);
939 set_pages_nx(virt_to_page(start), size >> PAGE_SHIFT);
940}
941
Arjan van de Ven63aaf302006-01-06 00:12:02 -0800942void mark_rodata_ro(void)
943{
Jan Beulich6fb14752007-05-02 19:27:10 +0200944 unsigned long start = PFN_ALIGN(_text);
945 unsigned long size = PFN_ALIGN(_etext) - start;
Arjan van de Ven63aaf302006-01-06 00:12:02 -0800946
Mathieu Desnoyers4e4eee02008-02-02 15:42:20 -0500947 set_pages_ro(virt_to_page(start), size >> PAGE_SHIFT);
948 printk(KERN_INFO "Write protecting the kernel text: %luk\n",
949 size >> 10);
Andi Kleen0c42f392008-01-30 13:33:42 +0100950
Steven Rostedt16239632009-02-17 17:57:30 -0500951 kernel_set_to_readonly = 1;
952
Andi Kleen0c42f392008-01-30 13:33:42 +0100953#ifdef CONFIG_CPA_DEBUG
Mathieu Desnoyers4e4eee02008-02-02 15:42:20 -0500954 printk(KERN_INFO "Testing CPA: Reverting %lx-%lx\n",
955 start, start+size);
956 set_pages_rw(virt_to_page(start), size>>PAGE_SHIFT);
Andi Kleen0c42f392008-01-30 13:33:42 +0100957
Mathieu Desnoyers4e4eee02008-02-02 15:42:20 -0500958 printk(KERN_INFO "Testing CPA: write protecting again\n");
959 set_pages_ro(virt_to_page(start), size>>PAGE_SHIFT);
Linus Torvalds602033e2007-07-26 12:07:21 -0700960#endif
Steven Rostedt8f0f9962008-05-12 21:20:56 +0200961
Jan Beulich6fb14752007-05-02 19:27:10 +0200962 start += size;
963 size = (unsigned long)__end_rodata - start;
Arjan van de Ven6d238cc2008-01-30 13:34:06 +0100964 set_pages_ro(virt_to_page(start), size >> PAGE_SHIFT);
Ingo Molnard7d119d2008-01-30 13:34:10 +0100965 printk(KERN_INFO "Write protecting the kernel read-only data: %luk\n",
966 size >> 10);
Arjan van de Venedeed302008-01-30 13:34:08 +0100967 rodata_test();
Arjan van de Ven63aaf302006-01-06 00:12:02 -0800968
Andi Kleen0c42f392008-01-30 13:33:42 +0100969#ifdef CONFIG_CPA_DEBUG
Ingo Molnard7d119d2008-01-30 13:34:10 +0100970 printk(KERN_INFO "Testing CPA: undo %lx-%lx\n", start, start + size);
Arjan van de Ven6d238cc2008-01-30 13:34:06 +0100971 set_pages_rw(virt_to_page(start), size >> PAGE_SHIFT);
Andi Kleen0c42f392008-01-30 13:33:42 +0100972
Ingo Molnard7d119d2008-01-30 13:34:10 +0100973 printk(KERN_INFO "Testing CPA: write protecting again\n");
Arjan van de Ven6d238cc2008-01-30 13:34:06 +0100974 set_pages_ro(virt_to_page(start), size >> PAGE_SHIFT);
Andi Kleen0c42f392008-01-30 13:33:42 +0100975#endif
Matthieu Castet5bd5a452010-11-16 22:31:26 +0100976 mark_nxdata_nx();
Arjan van de Ven63aaf302006-01-06 00:12:02 -0800977}
978#endif
979