]>
Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
1da177e4 LT |
2 | * |
3 | * Copyright (C) 1995 Linus Torvalds | |
4 | * | |
5 | * Support of BIGMEM added by Gerhard Wichert, Siemens AG, July 1999 | |
6 | */ | |
7 | ||
1da177e4 LT |
8 | #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> | |
6fb14755 | 24 | #include <linux/pfn.h> |
c9cf5528 | 25 | #include <linux/poison.h> |
1da177e4 LT |
26 | #include <linux/bootmem.h> |
27 | #include <linux/slab.h> | |
28 | #include <linux/proc_fs.h> | |
05039b92 | 29 | #include <linux/memory_hotplug.h> |
27d99f7e | 30 | #include <linux/initrd.h> |
55b2355e | 31 | #include <linux/cpumask.h> |
1da177e4 | 32 | |
f832ff18 | 33 | #include <asm/asm.h> |
1da177e4 LT |
34 | #include <asm/processor.h> |
35 | #include <asm/system.h> | |
36 | #include <asm/uaccess.h> | |
37 | #include <asm/pgtable.h> | |
38 | #include <asm/dma.h> | |
39 | #include <asm/fixmap.h> | |
40 | #include <asm/e820.h> | |
41 | #include <asm/apic.h> | |
8550eb99 | 42 | #include <asm/bugs.h> |
1da177e4 LT |
43 | #include <asm/tlb.h> |
44 | #include <asm/tlbflush.h> | |
a5a19c63 | 45 | #include <asm/pgalloc.h> |
1da177e4 | 46 | #include <asm/sections.h> |
b239fb25 | 47 | #include <asm/paravirt.h> |
551889a6 | 48 | #include <asm/setup.h> |
7bfeab9a | 49 | #include <asm/cacheflush.h> |
1da177e4 LT |
50 | |
51 | unsigned int __VMALLOC_RESERVE = 128 << 20; | |
52 | ||
f361a450 | 53 | unsigned long max_low_pfn_mapped; |
67794292 | 54 | unsigned long max_pfn_mapped; |
7d1116a9 | 55 | |
1da177e4 LT |
56 | DEFINE_PER_CPU(struct mmu_gather, mmu_gathers); |
57 | unsigned long highstart_pfn, highend_pfn; | |
58 | ||
8550eb99 | 59 | static noinline int do_test_wp_bit(void); |
1da177e4 | 60 | |
4e29684c YL |
61 | |
62 | static unsigned long __initdata table_start; | |
63 | static unsigned long __meminitdata table_end; | |
64 | static unsigned long __meminitdata table_top; | |
65 | ||
66 | static int __initdata after_init_bootmem; | |
67 | ||
68 | static __init void *alloc_low_page(unsigned long *phys) | |
69 | { | |
70 | unsigned long pfn = table_end++; | |
71 | void *adr; | |
72 | ||
73 | if (pfn >= table_top) | |
74 | panic("alloc_low_page: ran out of memory"); | |
75 | ||
76 | adr = __va(pfn * PAGE_SIZE); | |
77 | memset(adr, 0, PAGE_SIZE); | |
78 | *phys = pfn * PAGE_SIZE; | |
79 | return adr; | |
80 | } | |
81 | ||
1da177e4 LT |
82 | /* |
83 | * Creates a middle page table and puts a pointer to it in the | |
84 | * given global directory entry. This only returns the gd entry | |
85 | * in non-PAE compilation mode, since the middle layer is folded. | |
86 | */ | |
87 | static pmd_t * __init one_md_table_init(pgd_t *pgd) | |
88 | { | |
89 | pud_t *pud; | |
90 | pmd_t *pmd_table; | |
8550eb99 | 91 | |
1da177e4 | 92 | #ifdef CONFIG_X86_PAE |
4e29684c | 93 | unsigned long phys; |
b239fb25 | 94 | if (!(pgd_val(*pgd) & _PAGE_PRESENT)) { |
4e29684c YL |
95 | if (after_init_bootmem) |
96 | pmd_table = (pmd_t *)alloc_bootmem_low_pages(PAGE_SIZE); | |
97 | else | |
98 | pmd_table = (pmd_t *)alloc_low_page(&phys); | |
6944a9c8 | 99 | paravirt_alloc_pmd(&init_mm, __pa(pmd_table) >> PAGE_SHIFT); |
b239fb25 JF |
100 | set_pgd(pgd, __pgd(__pa(pmd_table) | _PAGE_PRESENT)); |
101 | pud = pud_offset(pgd, 0); | |
8550eb99 | 102 | BUG_ON(pmd_table != pmd_offset(pud, 0)); |
b239fb25 JF |
103 | } |
104 | #endif | |
1da177e4 LT |
105 | pud = pud_offset(pgd, 0); |
106 | pmd_table = pmd_offset(pud, 0); | |
8550eb99 | 107 | |
1da177e4 LT |
108 | return pmd_table; |
109 | } | |
110 | ||
111 | /* | |
112 | * Create a page table and place a pointer to it in a middle page | |
8550eb99 | 113 | * directory entry: |
1da177e4 LT |
114 | */ |
115 | static pte_t * __init one_page_table_init(pmd_t *pmd) | |
116 | { | |
b239fb25 | 117 | if (!(pmd_val(*pmd) & _PAGE_PRESENT)) { |
509a80c4 IM |
118 | pte_t *page_table = NULL; |
119 | ||
4e29684c | 120 | if (after_init_bootmem) { |
509a80c4 | 121 | #ifdef CONFIG_DEBUG_PAGEALLOC |
4e29684c | 122 | page_table = (pte_t *) alloc_bootmem_pages(PAGE_SIZE); |
509a80c4 | 123 | #endif |
4e29684c YL |
124 | if (!page_table) |
125 | page_table = | |
509a80c4 | 126 | (pte_t *)alloc_bootmem_low_pages(PAGE_SIZE); |
4e29684c YL |
127 | } else { |
128 | unsigned long phys; | |
129 | page_table = (pte_t *)alloc_low_page(&phys); | |
8550eb99 | 130 | } |
b239fb25 | 131 | |
6944a9c8 | 132 | paravirt_alloc_pte(&init_mm, __pa(page_table) >> PAGE_SHIFT); |
1da177e4 | 133 | set_pmd(pmd, __pmd(__pa(page_table) | _PAGE_TABLE)); |
b239fb25 | 134 | BUG_ON(page_table != pte_offset_kernel(pmd, 0)); |
1da177e4 | 135 | } |
509a80c4 | 136 | |
1da177e4 LT |
137 | return pte_offset_kernel(pmd, 0); |
138 | } | |
139 | ||
140 | /* | |
8550eb99 | 141 | * This function initializes a certain range of kernel virtual memory |
1da177e4 LT |
142 | * with new bootmem page tables, everywhere page tables are missing in |
143 | * the given range. | |
8550eb99 IM |
144 | * |
145 | * NOTE: The pagetables are allocated contiguous on the physical space | |
146 | * so we can cache the place of the first one and move around without | |
1da177e4 LT |
147 | * checking the pgd every time. |
148 | */ | |
8550eb99 IM |
149 | static void __init |
150 | page_table_range_init(unsigned long start, unsigned long end, pgd_t *pgd_base) | |
1da177e4 | 151 | { |
1da177e4 LT |
152 | int pgd_idx, pmd_idx; |
153 | unsigned long vaddr; | |
8550eb99 IM |
154 | pgd_t *pgd; |
155 | pmd_t *pmd; | |
1da177e4 LT |
156 | |
157 | vaddr = start; | |
158 | pgd_idx = pgd_index(vaddr); | |
159 | pmd_idx = pmd_index(vaddr); | |
160 | pgd = pgd_base + pgd_idx; | |
161 | ||
162 | for ( ; (pgd_idx < PTRS_PER_PGD) && (vaddr != end); pgd++, pgd_idx++) { | |
b239fb25 JF |
163 | pmd = one_md_table_init(pgd); |
164 | pmd = pmd + pmd_index(vaddr); | |
8550eb99 IM |
165 | for (; (pmd_idx < PTRS_PER_PMD) && (vaddr != end); |
166 | pmd++, pmd_idx++) { | |
b239fb25 | 167 | one_page_table_init(pmd); |
1da177e4 LT |
168 | |
169 | vaddr += PMD_SIZE; | |
170 | } | |
171 | pmd_idx = 0; | |
172 | } | |
173 | } | |
174 | ||
175 | static inline int is_kernel_text(unsigned long addr) | |
176 | { | |
177 | if (addr >= PAGE_OFFSET && addr <= (unsigned long)__init_end) | |
178 | return 1; | |
179 | return 0; | |
180 | } | |
181 | ||
182 | /* | |
8550eb99 IM |
183 | * This maps the physical memory to kernel virtual address space, a total |
184 | * of max_low_pfn pages, by creating page tables starting from address | |
185 | * PAGE_OFFSET: | |
1da177e4 | 186 | */ |
4e29684c | 187 | static void __init kernel_physical_mapping_init(pgd_t *pgd_base, |
a04ad82d YL |
188 | unsigned long start_pfn, |
189 | unsigned long end_pfn, | |
190 | int use_pse) | |
1da177e4 | 191 | { |
8550eb99 | 192 | int pgd_idx, pmd_idx, pte_ofs; |
1da177e4 LT |
193 | unsigned long pfn; |
194 | pgd_t *pgd; | |
195 | pmd_t *pmd; | |
196 | pte_t *pte; | |
a2699e47 SS |
197 | unsigned pages_2m, pages_4k; |
198 | int mapping_iter; | |
199 | ||
200 | /* | |
201 | * First iteration will setup identity mapping using large/small pages | |
202 | * based on use_pse, with other attributes same as set by | |
203 | * the early code in head_32.S | |
204 | * | |
205 | * Second iteration will setup the appropriate attributes (NX, GLOBAL..) | |
206 | * as desired for the kernel identity mapping. | |
207 | * | |
208 | * This two pass mechanism conforms to the TLB app note which says: | |
209 | * | |
210 | * "Software should not write to a paging-structure entry in a way | |
211 | * that would change, for any linear address, both the page size | |
212 | * and either the page frame or attributes." | |
213 | */ | |
214 | mapping_iter = 1; | |
1da177e4 | 215 | |
a04ad82d YL |
216 | if (!cpu_has_pse) |
217 | use_pse = 0; | |
1da177e4 | 218 | |
a2699e47 SS |
219 | repeat: |
220 | pages_2m = pages_4k = 0; | |
a04ad82d YL |
221 | pfn = start_pfn; |
222 | pgd_idx = pgd_index((pfn<<PAGE_SHIFT) + PAGE_OFFSET); | |
223 | pgd = pgd_base + pgd_idx; | |
1da177e4 LT |
224 | for (; pgd_idx < PTRS_PER_PGD; pgd++, pgd_idx++) { |
225 | pmd = one_md_table_init(pgd); | |
8550eb99 | 226 | |
a04ad82d YL |
227 | if (pfn >= end_pfn) |
228 | continue; | |
229 | #ifdef CONFIG_X86_PAE | |
230 | pmd_idx = pmd_index((pfn<<PAGE_SHIFT) + PAGE_OFFSET); | |
231 | pmd += pmd_idx; | |
232 | #else | |
233 | pmd_idx = 0; | |
234 | #endif | |
235 | for (; pmd_idx < PTRS_PER_PMD && pfn < end_pfn; | |
f3f20de8 | 236 | pmd++, pmd_idx++) { |
8550eb99 | 237 | unsigned int addr = pfn * PAGE_SIZE + PAGE_OFFSET; |
1da177e4 | 238 | |
8550eb99 IM |
239 | /* |
240 | * Map with big pages if possible, otherwise | |
241 | * create normal page tables: | |
242 | */ | |
a04ad82d | 243 | if (use_pse) { |
8550eb99 | 244 | unsigned int addr2; |
f3f20de8 | 245 | pgprot_t prot = PAGE_KERNEL_LARGE; |
a2699e47 SS |
246 | /* |
247 | * first pass will use the same initial | |
248 | * identity mapping attribute + _PAGE_PSE. | |
249 | */ | |
250 | pgprot_t init_prot = | |
251 | __pgprot(PTE_IDENT_ATTR | | |
252 | _PAGE_PSE); | |
f3f20de8 | 253 | |
8550eb99 | 254 | addr2 = (pfn + PTRS_PER_PTE-1) * PAGE_SIZE + |
f3f20de8 JF |
255 | PAGE_OFFSET + PAGE_SIZE-1; |
256 | ||
8550eb99 IM |
257 | if (is_kernel_text(addr) || |
258 | is_kernel_text(addr2)) | |
f3f20de8 JF |
259 | prot = PAGE_KERNEL_LARGE_EXEC; |
260 | ||
ce0c0e50 | 261 | pages_2m++; |
a2699e47 SS |
262 | if (mapping_iter == 1) |
263 | set_pmd(pmd, pfn_pmd(pfn, init_prot)); | |
264 | else | |
265 | set_pmd(pmd, pfn_pmd(pfn, prot)); | |
b239fb25 | 266 | |
1da177e4 | 267 | pfn += PTRS_PER_PTE; |
8550eb99 IM |
268 | continue; |
269 | } | |
270 | pte = one_page_table_init(pmd); | |
1da177e4 | 271 | |
a04ad82d YL |
272 | pte_ofs = pte_index((pfn<<PAGE_SHIFT) + PAGE_OFFSET); |
273 | pte += pte_ofs; | |
274 | for (; pte_ofs < PTRS_PER_PTE && pfn < end_pfn; | |
8550eb99 IM |
275 | pte++, pfn++, pte_ofs++, addr += PAGE_SIZE) { |
276 | pgprot_t prot = PAGE_KERNEL; | |
a2699e47 SS |
277 | /* |
278 | * first pass will use the same initial | |
279 | * identity mapping attribute. | |
280 | */ | |
281 | pgprot_t init_prot = __pgprot(PTE_IDENT_ATTR); | |
f3f20de8 | 282 | |
8550eb99 IM |
283 | if (is_kernel_text(addr)) |
284 | prot = PAGE_KERNEL_EXEC; | |
f3f20de8 | 285 | |
ce0c0e50 | 286 | pages_4k++; |
a2699e47 SS |
287 | if (mapping_iter == 1) |
288 | set_pte(pte, pfn_pte(pfn, init_prot)); | |
289 | else | |
290 | set_pte(pte, pfn_pte(pfn, prot)); | |
1da177e4 LT |
291 | } |
292 | } | |
293 | } | |
a2699e47 SS |
294 | if (mapping_iter == 1) { |
295 | /* | |
296 | * update direct mapping page count only in the first | |
297 | * iteration. | |
298 | */ | |
299 | update_page_count(PG_LEVEL_2M, pages_2m); | |
300 | update_page_count(PG_LEVEL_4K, pages_4k); | |
301 | ||
302 | /* | |
303 | * local global flush tlb, which will flush the previous | |
304 | * mappings present in both small and large page TLB's. | |
305 | */ | |
306 | __flush_tlb_all(); | |
307 | ||
308 | /* | |
309 | * Second iteration will set the actual desired PTE attributes. | |
310 | */ | |
311 | mapping_iter = 2; | |
312 | goto repeat; | |
313 | } | |
1da177e4 LT |
314 | } |
315 | ||
ae531c26 AV |
316 | /* |
317 | * devmem_is_allowed() checks to see if /dev/mem access to a certain address | |
318 | * is valid. The argument is a physical page number. | |
319 | * | |
320 | * | |
321 | * On x86, access has to be given to the first megabyte of ram because that area | |
322 | * contains bios code and data regions used by X and dosemu and similar apps. | |
323 | * Access has to be given to non-kernel-ram areas as well, these contain the PCI | |
324 | * mmio resources as well as potential bios/acpi data regions. | |
325 | */ | |
326 | int devmem_is_allowed(unsigned long pagenr) | |
327 | { | |
328 | if (pagenr <= 256) | |
329 | return 1; | |
330 | if (!page_is_ram(pagenr)) | |
331 | return 1; | |
332 | return 0; | |
333 | } | |
334 | ||
1da177e4 LT |
335 | #ifdef CONFIG_HIGHMEM |
336 | pte_t *kmap_pte; | |
337 | pgprot_t kmap_prot; | |
338 | ||
8550eb99 IM |
339 | static inline pte_t *kmap_get_fixmap_pte(unsigned long vaddr) |
340 | { | |
341 | return pte_offset_kernel(pmd_offset(pud_offset(pgd_offset_k(vaddr), | |
342 | vaddr), vaddr), vaddr); | |
343 | } | |
1da177e4 LT |
344 | |
345 | static void __init kmap_init(void) | |
346 | { | |
347 | unsigned long kmap_vstart; | |
348 | ||
8550eb99 IM |
349 | /* |
350 | * Cache the first kmap pte: | |
351 | */ | |
1da177e4 LT |
352 | kmap_vstart = __fix_to_virt(FIX_KMAP_BEGIN); |
353 | kmap_pte = kmap_get_fixmap_pte(kmap_vstart); | |
354 | ||
355 | kmap_prot = PAGE_KERNEL; | |
356 | } | |
357 | ||
358 | static void __init permanent_kmaps_init(pgd_t *pgd_base) | |
359 | { | |
8550eb99 | 360 | unsigned long vaddr; |
1da177e4 LT |
361 | pgd_t *pgd; |
362 | pud_t *pud; | |
363 | pmd_t *pmd; | |
364 | pte_t *pte; | |
1da177e4 LT |
365 | |
366 | vaddr = PKMAP_BASE; | |
367 | page_table_range_init(vaddr, vaddr + PAGE_SIZE*LAST_PKMAP, pgd_base); | |
368 | ||
369 | pgd = swapper_pg_dir + pgd_index(vaddr); | |
370 | pud = pud_offset(pgd, vaddr); | |
371 | pmd = pmd_offset(pud, vaddr); | |
372 | pte = pte_offset_kernel(pmd, vaddr); | |
8550eb99 | 373 | pkmap_page_table = pte; |
1da177e4 LT |
374 | } |
375 | ||
cc9f7a0c | 376 | static void __init add_one_highpage_init(struct page *page, int pfn) |
1da177e4 | 377 | { |
cc9f7a0c YL |
378 | ClearPageReserved(page); |
379 | init_page_count(page); | |
380 | __free_page(page); | |
381 | totalhigh_pages++; | |
1da177e4 LT |
382 | } |
383 | ||
b5bc6c0e YL |
384 | struct add_highpages_data { |
385 | unsigned long start_pfn; | |
386 | unsigned long end_pfn; | |
b5bc6c0e YL |
387 | }; |
388 | ||
d52d53b8 | 389 | static int __init add_highpages_work_fn(unsigned long start_pfn, |
b5bc6c0e | 390 | unsigned long end_pfn, void *datax) |
1da177e4 | 391 | { |
b5bc6c0e YL |
392 | int node_pfn; |
393 | struct page *page; | |
394 | unsigned long final_start_pfn, final_end_pfn; | |
395 | struct add_highpages_data *data; | |
8550eb99 | 396 | |
b5bc6c0e | 397 | data = (struct add_highpages_data *)datax; |
b5bc6c0e YL |
398 | |
399 | final_start_pfn = max(start_pfn, data->start_pfn); | |
400 | final_end_pfn = min(end_pfn, data->end_pfn); | |
401 | if (final_start_pfn >= final_end_pfn) | |
d52d53b8 | 402 | return 0; |
b5bc6c0e YL |
403 | |
404 | for (node_pfn = final_start_pfn; node_pfn < final_end_pfn; | |
405 | node_pfn++) { | |
406 | if (!pfn_valid(node_pfn)) | |
407 | continue; | |
408 | page = pfn_to_page(node_pfn); | |
cc9f7a0c | 409 | add_one_highpage_init(page, node_pfn); |
23be8c7d | 410 | } |
b5bc6c0e | 411 | |
d52d53b8 YL |
412 | return 0; |
413 | ||
b5bc6c0e YL |
414 | } |
415 | ||
416 | void __init add_highpages_with_active_regions(int nid, unsigned long start_pfn, | |
cc9f7a0c | 417 | unsigned long end_pfn) |
b5bc6c0e YL |
418 | { |
419 | struct add_highpages_data data; | |
420 | ||
421 | data.start_pfn = start_pfn; | |
422 | data.end_pfn = end_pfn; | |
b5bc6c0e YL |
423 | |
424 | work_with_active_regions(nid, add_highpages_work_fn, &data); | |
425 | } | |
426 | ||
8550eb99 | 427 | #ifndef CONFIG_NUMA |
cc9f7a0c | 428 | static void __init set_highmem_pages_init(void) |
1da177e4 | 429 | { |
cc9f7a0c | 430 | add_highpages_with_active_regions(0, highstart_pfn, highend_pfn); |
b5bc6c0e | 431 | |
1da177e4 LT |
432 | totalram_pages += totalhigh_pages; |
433 | } | |
8550eb99 | 434 | #endif /* !CONFIG_NUMA */ |
1da177e4 LT |
435 | |
436 | #else | |
8550eb99 IM |
437 | # define kmap_init() do { } while (0) |
438 | # define permanent_kmaps_init(pgd_base) do { } while (0) | |
cc9f7a0c | 439 | # define set_highmem_pages_init() do { } while (0) |
1da177e4 LT |
440 | #endif /* CONFIG_HIGHMEM */ |
441 | ||
b239fb25 | 442 | void __init native_pagetable_setup_start(pgd_t *base) |
1da177e4 | 443 | { |
551889a6 IC |
444 | unsigned long pfn, va; |
445 | pgd_t *pgd; | |
446 | pud_t *pud; | |
447 | pmd_t *pmd; | |
448 | pte_t *pte; | |
b239fb25 JF |
449 | |
450 | /* | |
551889a6 IC |
451 | * Remove any mappings which extend past the end of physical |
452 | * memory from the boot time page table: | |
b239fb25 | 453 | */ |
551889a6 IC |
454 | for (pfn = max_low_pfn + 1; pfn < 1<<(32-PAGE_SHIFT); pfn++) { |
455 | va = PAGE_OFFSET + (pfn<<PAGE_SHIFT); | |
456 | pgd = base + pgd_index(va); | |
457 | if (!pgd_present(*pgd)) | |
458 | break; | |
459 | ||
460 | pud = pud_offset(pgd, va); | |
461 | pmd = pmd_offset(pud, va); | |
462 | if (!pmd_present(*pmd)) | |
463 | break; | |
464 | ||
465 | pte = pte_offset_kernel(pmd, va); | |
466 | if (!pte_present(*pte)) | |
467 | break; | |
468 | ||
469 | pte_clear(NULL, va, pte); | |
470 | } | |
6944a9c8 | 471 | paravirt_alloc_pmd(&init_mm, __pa(base) >> PAGE_SHIFT); |
b239fb25 JF |
472 | } |
473 | ||
474 | void __init native_pagetable_setup_done(pgd_t *base) | |
475 | { | |
b239fb25 JF |
476 | } |
477 | ||
478 | /* | |
479 | * Build a proper pagetable for the kernel mappings. Up until this | |
480 | * point, we've been running on some set of pagetables constructed by | |
481 | * the boot process. | |
482 | * | |
483 | * If we're booting on native hardware, this will be a pagetable | |
551889a6 IC |
484 | * constructed in arch/x86/kernel/head_32.S. The root of the |
485 | * pagetable will be swapper_pg_dir. | |
b239fb25 JF |
486 | * |
487 | * If we're booting paravirtualized under a hypervisor, then there are | |
488 | * more options: we may already be running PAE, and the pagetable may | |
489 | * or may not be based in swapper_pg_dir. In any case, | |
490 | * paravirt_pagetable_setup_start() will set up swapper_pg_dir | |
491 | * appropriately for the rest of the initialization to work. | |
492 | * | |
493 | * In general, pagetable_init() assumes that the pagetable may already | |
494 | * be partially populated, and so it avoids stomping on any existing | |
495 | * mappings. | |
496 | */ | |
e7b37895 | 497 | static void __init early_ioremap_page_table_range_init(pgd_t *pgd_base) |
b239fb25 | 498 | { |
8550eb99 | 499 | unsigned long vaddr, end; |
b239fb25 | 500 | |
1da177e4 LT |
501 | /* |
502 | * Fixed mappings, only the page table structure has to be | |
503 | * created - mappings will be set by set_fixmap(): | |
504 | */ | |
beacfaac | 505 | early_ioremap_clear(); |
1da177e4 | 506 | vaddr = __fix_to_virt(__end_of_fixed_addresses - 1) & PMD_MASK; |
b239fb25 JF |
507 | end = (FIXADDR_TOP + PMD_SIZE - 1) & PMD_MASK; |
508 | page_table_range_init(vaddr, end, pgd_base); | |
beacfaac | 509 | early_ioremap_reset(); |
e7b37895 YL |
510 | } |
511 | ||
512 | static void __init pagetable_init(void) | |
513 | { | |
514 | pgd_t *pgd_base = swapper_pg_dir; | |
515 | ||
516 | paravirt_pagetable_setup_start(pgd_base); | |
1da177e4 LT |
517 | |
518 | permanent_kmaps_init(pgd_base); | |
519 | ||
b239fb25 | 520 | paravirt_pagetable_setup_done(pgd_base); |
1da177e4 LT |
521 | } |
522 | ||
a6eb84bc | 523 | #ifdef CONFIG_ACPI_SLEEP |
1da177e4 | 524 | /* |
a6eb84bc | 525 | * ACPI suspend needs this for resume, because things like the intel-agp |
1da177e4 LT |
526 | * driver might have split up a kernel 4MB mapping. |
527 | */ | |
a6eb84bc | 528 | char swsusp_pg_dir[PAGE_SIZE] |
8550eb99 | 529 | __attribute__ ((aligned(PAGE_SIZE))); |
1da177e4 LT |
530 | |
531 | static inline void save_pg_dir(void) | |
532 | { | |
533 | memcpy(swsusp_pg_dir, swapper_pg_dir, PAGE_SIZE); | |
534 | } | |
a6eb84bc | 535 | #else /* !CONFIG_ACPI_SLEEP */ |
1da177e4 LT |
536 | static inline void save_pg_dir(void) |
537 | { | |
538 | } | |
a6eb84bc | 539 | #endif /* !CONFIG_ACPI_SLEEP */ |
1da177e4 | 540 | |
8550eb99 | 541 | void zap_low_mappings(void) |
1da177e4 LT |
542 | { |
543 | int i; | |
544 | ||
1da177e4 LT |
545 | /* |
546 | * Zap initial low-memory mappings. | |
547 | * | |
548 | * Note that "pgd_clear()" doesn't do it for | |
549 | * us, because pgd_clear() is a no-op on i386. | |
550 | */ | |
68db065c | 551 | for (i = 0; i < KERNEL_PGD_BOUNDARY; i++) { |
1da177e4 LT |
552 | #ifdef CONFIG_X86_PAE |
553 | set_pgd(swapper_pg_dir+i, __pgd(1 + __pa(empty_zero_page))); | |
554 | #else | |
555 | set_pgd(swapper_pg_dir+i, __pgd(0)); | |
556 | #endif | |
8550eb99 | 557 | } |
1da177e4 LT |
558 | flush_tlb_all(); |
559 | } | |
560 | ||
8550eb99 | 561 | int nx_enabled; |
d5321abe | 562 | |
ef5e94af | 563 | pteval_t __supported_pte_mask __read_mostly = ~(_PAGE_NX | _PAGE_GLOBAL); |
6fdc05d4 JF |
564 | EXPORT_SYMBOL_GPL(__supported_pte_mask); |
565 | ||
d5321abe JB |
566 | #ifdef CONFIG_X86_PAE |
567 | ||
8550eb99 | 568 | static int disable_nx __initdata; |
1da177e4 LT |
569 | |
570 | /* | |
571 | * noexec = on|off | |
572 | * | |
573 | * Control non executable mappings. | |
574 | * | |
575 | * on Enable | |
576 | * off Disable | |
577 | */ | |
1a3f239d | 578 | static int __init noexec_setup(char *str) |
1da177e4 | 579 | { |
1a3f239d RR |
580 | if (!str || !strcmp(str, "on")) { |
581 | if (cpu_has_nx) { | |
582 | __supported_pte_mask |= _PAGE_NX; | |
583 | disable_nx = 0; | |
584 | } | |
8550eb99 IM |
585 | } else { |
586 | if (!strcmp(str, "off")) { | |
587 | disable_nx = 1; | |
588 | __supported_pte_mask &= ~_PAGE_NX; | |
589 | } else { | |
590 | return -EINVAL; | |
591 | } | |
592 | } | |
1a3f239d RR |
593 | |
594 | return 0; | |
1da177e4 | 595 | } |
1a3f239d | 596 | early_param("noexec", noexec_setup); |
1da177e4 | 597 | |
1da177e4 LT |
598 | static void __init set_nx(void) |
599 | { | |
600 | unsigned int v[4], l, h; | |
601 | ||
602 | if (cpu_has_pae && (cpuid_eax(0x80000000) > 0x80000001)) { | |
603 | cpuid(0x80000001, &v[0], &v[1], &v[2], &v[3]); | |
8550eb99 | 604 | |
1da177e4 LT |
605 | if ((v[3] & (1 << 20)) && !disable_nx) { |
606 | rdmsr(MSR_EFER, l, h); | |
607 | l |= EFER_NX; | |
608 | wrmsr(MSR_EFER, l, h); | |
609 | nx_enabled = 1; | |
610 | __supported_pte_mask |= _PAGE_NX; | |
611 | } | |
612 | } | |
613 | } | |
1da177e4 LT |
614 | #endif |
615 | ||
90d967e0 YL |
616 | /* user-defined highmem size */ |
617 | static unsigned int highmem_pages = -1; | |
618 | ||
619 | /* | |
620 | * highmem=size forces highmem to be exactly 'size' bytes. | |
621 | * This works even on boxes that have no highmem otherwise. | |
622 | * This also works to reduce highmem size on bigger boxes. | |
623 | */ | |
624 | static int __init parse_highmem(char *arg) | |
625 | { | |
626 | if (!arg) | |
627 | return -EINVAL; | |
628 | ||
629 | highmem_pages = memparse(arg, &arg) >> PAGE_SHIFT; | |
630 | return 0; | |
631 | } | |
632 | early_param("highmem", parse_highmem); | |
633 | ||
634 | /* | |
635 | * Determine low and high memory ranges: | |
636 | */ | |
2ec65f8b | 637 | void __init find_low_pfn_range(void) |
90d967e0 | 638 | { |
2ec65f8b YL |
639 | /* it could update max_pfn */ |
640 | ||
346cafec | 641 | /* max_low_pfn is 0, we already have early_res support */ |
90d967e0 YL |
642 | |
643 | max_low_pfn = max_pfn; | |
644 | if (max_low_pfn > MAXMEM_PFN) { | |
645 | if (highmem_pages == -1) | |
646 | highmem_pages = max_pfn - MAXMEM_PFN; | |
647 | if (highmem_pages + MAXMEM_PFN < max_pfn) | |
648 | max_pfn = MAXMEM_PFN + highmem_pages; | |
649 | if (highmem_pages + MAXMEM_PFN > max_pfn) { | |
650 | printk(KERN_WARNING "only %luMB highmem pages " | |
651 | "available, ignoring highmem size of %uMB.\n", | |
652 | pages_to_mb(max_pfn - MAXMEM_PFN), | |
653 | pages_to_mb(highmem_pages)); | |
654 | highmem_pages = 0; | |
655 | } | |
656 | max_low_pfn = MAXMEM_PFN; | |
657 | #ifndef CONFIG_HIGHMEM | |
658 | /* Maximum memory usable is what is directly addressable */ | |
659 | printk(KERN_WARNING "Warning only %ldMB will be used.\n", | |
660 | MAXMEM>>20); | |
661 | if (max_pfn > MAX_NONPAE_PFN) | |
662 | printk(KERN_WARNING | |
663 | "Use a HIGHMEM64G enabled kernel.\n"); | |
664 | else | |
665 | printk(KERN_WARNING "Use a HIGHMEM enabled kernel.\n"); | |
666 | max_pfn = MAXMEM_PFN; | |
667 | #else /* !CONFIG_HIGHMEM */ | |
668 | #ifndef CONFIG_HIGHMEM64G | |
669 | if (max_pfn > MAX_NONPAE_PFN) { | |
670 | max_pfn = MAX_NONPAE_PFN; | |
671 | printk(KERN_WARNING "Warning only 4GB will be used." | |
672 | "Use a HIGHMEM64G enabled kernel.\n"); | |
673 | } | |
674 | #endif /* !CONFIG_HIGHMEM64G */ | |
675 | #endif /* !CONFIG_HIGHMEM */ | |
676 | } else { | |
677 | if (highmem_pages == -1) | |
678 | highmem_pages = 0; | |
679 | #ifdef CONFIG_HIGHMEM | |
680 | if (highmem_pages >= max_pfn) { | |
681 | printk(KERN_ERR "highmem size specified (%uMB) is " | |
682 | "bigger than pages available (%luMB)!.\n", | |
683 | pages_to_mb(highmem_pages), | |
684 | pages_to_mb(max_pfn)); | |
685 | highmem_pages = 0; | |
686 | } | |
687 | if (highmem_pages) { | |
688 | if (max_low_pfn - highmem_pages < | |
689 | 64*1024*1024/PAGE_SIZE){ | |
690 | printk(KERN_ERR "highmem size %uMB results in " | |
691 | "smaller than 64MB lowmem, ignoring it.\n" | |
692 | , pages_to_mb(highmem_pages)); | |
693 | highmem_pages = 0; | |
694 | } | |
695 | max_low_pfn -= highmem_pages; | |
696 | } | |
697 | #else | |
698 | if (highmem_pages) | |
699 | printk(KERN_ERR "ignoring highmem size on non-highmem" | |
700 | " kernel!\n"); | |
701 | #endif | |
702 | } | |
90d967e0 YL |
703 | } |
704 | ||
b2ac82a0 | 705 | #ifndef CONFIG_NEED_MULTIPLE_NODES |
2ec65f8b | 706 | void __init initmem_init(unsigned long start_pfn, |
b2ac82a0 YL |
707 | unsigned long end_pfn) |
708 | { | |
b2ac82a0 YL |
709 | #ifdef CONFIG_HIGHMEM |
710 | highstart_pfn = highend_pfn = max_pfn; | |
711 | if (max_pfn > max_low_pfn) | |
712 | highstart_pfn = max_low_pfn; | |
713 | memory_present(0, 0, highend_pfn); | |
cb95a13a | 714 | e820_register_active_regions(0, 0, highend_pfn); |
b2ac82a0 YL |
715 | printk(KERN_NOTICE "%ldMB HIGHMEM available.\n", |
716 | pages_to_mb(highend_pfn - highstart_pfn)); | |
717 | num_physpages = highend_pfn; | |
718 | high_memory = (void *) __va(highstart_pfn * PAGE_SIZE - 1) + 1; | |
719 | #else | |
720 | memory_present(0, 0, max_low_pfn); | |
cb95a13a | 721 | e820_register_active_regions(0, 0, max_low_pfn); |
b2ac82a0 YL |
722 | num_physpages = max_low_pfn; |
723 | high_memory = (void *) __va(max_low_pfn * PAGE_SIZE - 1) + 1; | |
724 | #endif | |
725 | #ifdef CONFIG_FLATMEM | |
726 | max_mapnr = num_physpages; | |
727 | #endif | |
728 | printk(KERN_NOTICE "%ldMB LOWMEM available.\n", | |
729 | pages_to_mb(max_low_pfn)); | |
730 | ||
731 | setup_bootmem_allocator(); | |
b2ac82a0 | 732 | } |
cb95a13a | 733 | #endif /* !CONFIG_NEED_MULTIPLE_NODES */ |
b2ac82a0 | 734 | |
cb95a13a | 735 | static void __init zone_sizes_init(void) |
b2ac82a0 YL |
736 | { |
737 | unsigned long max_zone_pfns[MAX_NR_ZONES]; | |
738 | memset(max_zone_pfns, 0, sizeof(max_zone_pfns)); | |
739 | max_zone_pfns[ZONE_DMA] = | |
740 | virt_to_phys((char *)MAX_DMA_ADDRESS) >> PAGE_SHIFT; | |
741 | max_zone_pfns[ZONE_NORMAL] = max_low_pfn; | |
b2ac82a0 YL |
742 | #ifdef CONFIG_HIGHMEM |
743 | max_zone_pfns[ZONE_HIGHMEM] = highend_pfn; | |
b2ac82a0 YL |
744 | #endif |
745 | ||
746 | free_area_init_nodes(max_zone_pfns); | |
747 | } | |
b2ac82a0 | 748 | |
b2ac82a0 YL |
749 | void __init setup_bootmem_allocator(void) |
750 | { | |
751 | int i; | |
752 | unsigned long bootmap_size, bootmap; | |
753 | /* | |
754 | * Initialize the boot-time allocator (with low memory only): | |
755 | */ | |
756 | bootmap_size = bootmem_bootmap_pages(max_low_pfn)<<PAGE_SHIFT; | |
757 | bootmap = find_e820_area(min_low_pfn<<PAGE_SHIFT, | |
758 | max_pfn_mapped<<PAGE_SHIFT, bootmap_size, | |
759 | PAGE_SIZE); | |
760 | if (bootmap == -1L) | |
761 | panic("Cannot find bootmem map of size %ld\n", bootmap_size); | |
762 | reserve_early(bootmap, bootmap + bootmap_size, "BOOTMAP"); | |
225c37d7 | 763 | |
346cafec YL |
764 | /* don't touch min_low_pfn */ |
765 | bootmap_size = init_bootmem_node(NODE_DATA(0), bootmap >> PAGE_SHIFT, | |
766 | min_low_pfn, max_low_pfn); | |
b2ac82a0 YL |
767 | printk(KERN_INFO " mapped low ram: 0 - %08lx\n", |
768 | max_pfn_mapped<<PAGE_SHIFT); | |
769 | printk(KERN_INFO " low ram: %08lx - %08lx\n", | |
770 | min_low_pfn<<PAGE_SHIFT, max_low_pfn<<PAGE_SHIFT); | |
771 | printk(KERN_INFO " bootmap %08lx - %08lx\n", | |
772 | bootmap, bootmap + bootmap_size); | |
773 | for_each_online_node(i) | |
774 | free_bootmem_with_active_regions(i, max_low_pfn); | |
775 | early_res_to_bootmem(0, max_low_pfn<<PAGE_SHIFT); | |
776 | ||
4e29684c | 777 | after_init_bootmem = 1; |
b2ac82a0 YL |
778 | } |
779 | ||
4e29684c YL |
780 | static void __init find_early_table_space(unsigned long end) |
781 | { | |
7482b0e9 | 782 | unsigned long puds, pmds, ptes, tables, start; |
4e29684c YL |
783 | |
784 | puds = (end + PUD_SIZE - 1) >> PUD_SHIFT; | |
785 | tables = PAGE_ALIGN(puds * sizeof(pud_t)); | |
786 | ||
787 | pmds = (end + PMD_SIZE - 1) >> PMD_SHIFT; | |
788 | tables += PAGE_ALIGN(pmds * sizeof(pmd_t)); | |
789 | ||
7482b0e9 YL |
790 | if (cpu_has_pse) { |
791 | unsigned long extra; | |
a04ad82d YL |
792 | |
793 | extra = end - ((end>>PMD_SHIFT) << PMD_SHIFT); | |
794 | extra += PMD_SIZE; | |
7482b0e9 YL |
795 | ptes = (extra + PAGE_SIZE - 1) >> PAGE_SHIFT; |
796 | } else | |
797 | ptes = (end + PAGE_SIZE - 1) >> PAGE_SHIFT; | |
798 | ||
799 | tables += PAGE_ALIGN(ptes * sizeof(pte_t)); | |
8207c257 | 800 | |
a04ad82d YL |
801 | /* for fixmap */ |
802 | tables += PAGE_SIZE * 2; | |
803 | ||
4e29684c YL |
804 | /* |
805 | * RED-PEN putting page tables only on node 0 could | |
806 | * cause a hotspot and fill up ZONE_DMA. The page tables | |
807 | * need roughly 0.5KB per GB. | |
808 | */ | |
809 | start = 0x7000; | |
810 | table_start = find_e820_area(start, max_pfn_mapped<<PAGE_SHIFT, | |
811 | tables, PAGE_SIZE); | |
812 | if (table_start == -1UL) | |
813 | panic("Cannot find space for the kernel page tables"); | |
814 | ||
815 | table_start >>= PAGE_SHIFT; | |
816 | table_end = table_start; | |
817 | table_top = table_start + (tables>>PAGE_SHIFT); | |
818 | ||
819 | printk(KERN_DEBUG "kernel direct mapping tables up to %lx @ %lx-%lx\n", | |
820 | end, table_start << PAGE_SHIFT, | |
821 | (table_start << PAGE_SHIFT) + tables); | |
822 | } | |
823 | ||
824 | unsigned long __init_refok init_memory_mapping(unsigned long start, | |
825 | unsigned long end) | |
826 | { | |
827 | pgd_t *pgd_base = swapper_pg_dir; | |
a04ad82d YL |
828 | unsigned long start_pfn, end_pfn; |
829 | unsigned long big_page_start; | |
4e29684c YL |
830 | |
831 | /* | |
832 | * Find space for the kernel direct mapping tables. | |
833 | */ | |
834 | if (!after_init_bootmem) | |
835 | find_early_table_space(end); | |
836 | ||
837 | #ifdef CONFIG_X86_PAE | |
838 | set_nx(); | |
839 | if (nx_enabled) | |
840 | printk(KERN_INFO "NX (Execute Disable) protection: active\n"); | |
841 | #endif | |
842 | ||
843 | /* Enable PSE if available */ | |
844 | if (cpu_has_pse) | |
845 | set_in_cr4(X86_CR4_PSE); | |
846 | ||
847 | /* Enable PGE if available */ | |
848 | if (cpu_has_pge) { | |
849 | set_in_cr4(X86_CR4_PGE); | |
ef5e94af | 850 | __supported_pte_mask |= _PAGE_GLOBAL; |
4e29684c YL |
851 | } |
852 | ||
a04ad82d YL |
853 | /* |
854 | * Don't use a large page for the first 2/4MB of memory | |
855 | * because there are often fixed size MTRRs in there | |
856 | * and overlapping MTRRs into large pages can cause | |
857 | * slowdowns. | |
858 | */ | |
859 | big_page_start = PMD_SIZE; | |
860 | ||
861 | if (start < big_page_start) { | |
862 | start_pfn = start >> PAGE_SHIFT; | |
863 | end_pfn = min(big_page_start>>PAGE_SHIFT, end>>PAGE_SHIFT); | |
864 | } else { | |
865 | /* head is not big page alignment ? */ | |
866 | start_pfn = start >> PAGE_SHIFT; | |
867 | end_pfn = ((start + (PMD_SIZE - 1))>>PMD_SHIFT) | |
868 | << (PMD_SHIFT - PAGE_SHIFT); | |
869 | } | |
870 | if (start_pfn < end_pfn) | |
871 | kernel_physical_mapping_init(pgd_base, start_pfn, end_pfn, 0); | |
872 | ||
873 | /* big page range */ | |
874 | start_pfn = ((start + (PMD_SIZE - 1))>>PMD_SHIFT) | |
875 | << (PMD_SHIFT - PAGE_SHIFT); | |
876 | if (start_pfn < (big_page_start >> PAGE_SHIFT)) | |
877 | start_pfn = big_page_start >> PAGE_SHIFT; | |
878 | end_pfn = (end>>PMD_SHIFT) << (PMD_SHIFT - PAGE_SHIFT); | |
879 | if (start_pfn < end_pfn) | |
880 | kernel_physical_mapping_init(pgd_base, start_pfn, end_pfn, | |
881 | cpu_has_pse); | |
882 | ||
883 | /* tail is not big page alignment ? */ | |
884 | start_pfn = end_pfn; | |
885 | if (start_pfn > (big_page_start>>PAGE_SHIFT)) { | |
886 | end_pfn = end >> PAGE_SHIFT; | |
887 | if (start_pfn < end_pfn) | |
888 | kernel_physical_mapping_init(pgd_base, start_pfn, | |
889 | end_pfn, 0); | |
890 | } | |
4e29684c | 891 | |
e7b37895 YL |
892 | early_ioremap_page_table_range_init(pgd_base); |
893 | ||
4e29684c YL |
894 | load_cr3(swapper_pg_dir); |
895 | ||
896 | __flush_tlb_all(); | |
897 | ||
898 | if (!after_init_bootmem) | |
899 | reserve_early(table_start << PAGE_SHIFT, | |
900 | table_end << PAGE_SHIFT, "PGTABLE"); | |
901 | ||
caadbdce YL |
902 | if (!after_init_bootmem) |
903 | early_memtest(start, end); | |
904 | ||
4e29684c YL |
905 | return end >> PAGE_SHIFT; |
906 | } | |
907 | ||
e7b37895 | 908 | |
1da177e4 LT |
909 | /* |
910 | * paging_init() sets up the page tables - note that the first 8MB are | |
911 | * already mapped by head.S. | |
912 | * | |
913 | * This routines also unmaps the page at virtual kernel address 0, so | |
914 | * that we can trap those pesky NULL-reference errors in the kernel. | |
915 | */ | |
916 | void __init paging_init(void) | |
917 | { | |
1da177e4 LT |
918 | pagetable_init(); |
919 | ||
1da177e4 LT |
920 | __flush_tlb_all(); |
921 | ||
922 | kmap_init(); | |
11cd0bc1 YL |
923 | |
924 | /* | |
925 | * NOTE: at this point the bootmem allocator is fully available. | |
926 | */ | |
11cd0bc1 YL |
927 | sparse_init(); |
928 | zone_sizes_init(); | |
1da177e4 LT |
929 | } |
930 | ||
931 | /* | |
932 | * Test if the WP bit works in supervisor mode. It isn't supported on 386's | |
f7f17a67 DV |
933 | * and also on some strange 486's. All 586+'s are OK. This used to involve |
934 | * black magic jumps to work around some nasty CPU bugs, but fortunately the | |
935 | * switch to using exceptions got rid of all that. | |
1da177e4 | 936 | */ |
1da177e4 LT |
937 | static void __init test_wp_bit(void) |
938 | { | |
d7d119d7 IM |
939 | printk(KERN_INFO |
940 | "Checking if this processor honours the WP bit even in supervisor mode..."); | |
1da177e4 LT |
941 | |
942 | /* Any page-aligned address will do, the test is non-destructive */ | |
943 | __set_fixmap(FIX_WP_TEST, __pa(&swapper_pg_dir), PAGE_READONLY); | |
944 | boot_cpu_data.wp_works_ok = do_test_wp_bit(); | |
945 | clear_fixmap(FIX_WP_TEST); | |
946 | ||
947 | if (!boot_cpu_data.wp_works_ok) { | |
d7d119d7 | 948 | printk(KERN_CONT "No.\n"); |
1da177e4 | 949 | #ifdef CONFIG_X86_WP_WORKS_OK |
d7d119d7 IM |
950 | panic( |
951 | "This kernel doesn't support CPU's with broken WP. Recompile it for a 386!"); | |
1da177e4 LT |
952 | #endif |
953 | } else { | |
d7d119d7 | 954 | printk(KERN_CONT "Ok.\n"); |
1da177e4 LT |
955 | } |
956 | } | |
957 | ||
8550eb99 | 958 | static struct kcore_list kcore_mem, kcore_vmalloc; |
1da177e4 LT |
959 | |
960 | void __init mem_init(void) | |
961 | { | |
1da177e4 | 962 | int codesize, reservedpages, datasize, initsize; |
cc9f7a0c | 963 | int tmp; |
1da177e4 | 964 | |
05b79bdc | 965 | #ifdef CONFIG_FLATMEM |
8d8f3cbe | 966 | BUG_ON(!mem_map); |
1da177e4 | 967 | #endif |
1da177e4 LT |
968 | /* this will put all low memory onto the freelists */ |
969 | totalram_pages += free_all_bootmem(); | |
970 | ||
971 | reservedpages = 0; | |
972 | for (tmp = 0; tmp < max_low_pfn; tmp++) | |
973 | /* | |
8550eb99 | 974 | * Only count reserved RAM pages: |
1da177e4 LT |
975 | */ |
976 | if (page_is_ram(tmp) && PageReserved(pfn_to_page(tmp))) | |
977 | reservedpages++; | |
978 | ||
cc9f7a0c | 979 | set_highmem_pages_init(); |
1da177e4 LT |
980 | |
981 | codesize = (unsigned long) &_etext - (unsigned long) &_text; | |
982 | datasize = (unsigned long) &_edata - (unsigned long) &_etext; | |
983 | initsize = (unsigned long) &__init_end - (unsigned long) &__init_begin; | |
984 | ||
8550eb99 IM |
985 | kclist_add(&kcore_mem, __va(0), max_low_pfn << PAGE_SHIFT); |
986 | kclist_add(&kcore_vmalloc, (void *)VMALLOC_START, | |
1da177e4 LT |
987 | VMALLOC_END-VMALLOC_START); |
988 | ||
8550eb99 IM |
989 | printk(KERN_INFO "Memory: %luk/%luk available (%dk kernel code, " |
990 | "%dk reserved, %dk data, %dk init, %ldk highmem)\n", | |
1da177e4 LT |
991 | (unsigned long) nr_free_pages() << (PAGE_SHIFT-10), |
992 | num_physpages << (PAGE_SHIFT-10), | |
993 | codesize >> 10, | |
994 | reservedpages << (PAGE_SHIFT-10), | |
995 | datasize >> 10, | |
996 | initsize >> 10, | |
997 | (unsigned long) (totalhigh_pages << (PAGE_SHIFT-10)) | |
998 | ); | |
999 | ||
d7d119d7 | 1000 | printk(KERN_INFO "virtual kernel memory layout:\n" |
8550eb99 | 1001 | " fixmap : 0x%08lx - 0x%08lx (%4ld kB)\n" |
052e7994 | 1002 | #ifdef CONFIG_HIGHMEM |
8550eb99 | 1003 | " pkmap : 0x%08lx - 0x%08lx (%4ld kB)\n" |
052e7994 | 1004 | #endif |
8550eb99 IM |
1005 | " vmalloc : 0x%08lx - 0x%08lx (%4ld MB)\n" |
1006 | " lowmem : 0x%08lx - 0x%08lx (%4ld MB)\n" | |
1007 | " .init : 0x%08lx - 0x%08lx (%4ld kB)\n" | |
1008 | " .data : 0x%08lx - 0x%08lx (%4ld kB)\n" | |
1009 | " .text : 0x%08lx - 0x%08lx (%4ld kB)\n", | |
1010 | FIXADDR_START, FIXADDR_TOP, | |
1011 | (FIXADDR_TOP - FIXADDR_START) >> 10, | |
052e7994 JF |
1012 | |
1013 | #ifdef CONFIG_HIGHMEM | |
8550eb99 IM |
1014 | PKMAP_BASE, PKMAP_BASE+LAST_PKMAP*PAGE_SIZE, |
1015 | (LAST_PKMAP*PAGE_SIZE) >> 10, | |
052e7994 JF |
1016 | #endif |
1017 | ||
8550eb99 IM |
1018 | VMALLOC_START, VMALLOC_END, |
1019 | (VMALLOC_END - VMALLOC_START) >> 20, | |
052e7994 | 1020 | |
8550eb99 IM |
1021 | (unsigned long)__va(0), (unsigned long)high_memory, |
1022 | ((unsigned long)high_memory - (unsigned long)__va(0)) >> 20, | |
052e7994 | 1023 | |
8550eb99 IM |
1024 | (unsigned long)&__init_begin, (unsigned long)&__init_end, |
1025 | ((unsigned long)&__init_end - | |
1026 | (unsigned long)&__init_begin) >> 10, | |
052e7994 | 1027 | |
8550eb99 IM |
1028 | (unsigned long)&_etext, (unsigned long)&_edata, |
1029 | ((unsigned long)&_edata - (unsigned long)&_etext) >> 10, | |
052e7994 | 1030 | |
8550eb99 IM |
1031 | (unsigned long)&_text, (unsigned long)&_etext, |
1032 | ((unsigned long)&_etext - (unsigned long)&_text) >> 10); | |
052e7994 JF |
1033 | |
1034 | #ifdef CONFIG_HIGHMEM | |
8550eb99 IM |
1035 | BUG_ON(PKMAP_BASE + LAST_PKMAP*PAGE_SIZE > FIXADDR_START); |
1036 | BUG_ON(VMALLOC_END > PKMAP_BASE); | |
052e7994 | 1037 | #endif |
8550eb99 IM |
1038 | BUG_ON(VMALLOC_START > VMALLOC_END); |
1039 | BUG_ON((unsigned long)high_memory > VMALLOC_START); | |
052e7994 | 1040 | |
1da177e4 LT |
1041 | if (boot_cpu_data.wp_works_ok < 0) |
1042 | test_wp_bit(); | |
1043 | ||
76ebd054 | 1044 | cpa_init(); |
61165d7a | 1045 | save_pg_dir(); |
1da177e4 | 1046 | zap_low_mappings(); |
1da177e4 LT |
1047 | } |
1048 | ||
ad8f5797 | 1049 | #ifdef CONFIG_MEMORY_HOTPLUG |
bc02af93 | 1050 | int arch_add_memory(int nid, u64 start, u64 size) |
05039b92 | 1051 | { |
7c7e9425 | 1052 | struct pglist_data *pgdata = NODE_DATA(nid); |
776ed98b | 1053 | struct zone *zone = pgdata->node_zones + ZONE_HIGHMEM; |
05039b92 DH |
1054 | unsigned long start_pfn = start >> PAGE_SHIFT; |
1055 | unsigned long nr_pages = size >> PAGE_SHIFT; | |
1056 | ||
1057 | return __add_pages(zone, start_pfn, nr_pages); | |
1058 | } | |
9d99aaa3 | 1059 | #endif |
05039b92 | 1060 | |
1da177e4 LT |
1061 | /* |
1062 | * This function cannot be __init, since exceptions don't work in that | |
1063 | * section. Put this after the callers, so that it cannot be inlined. | |
1064 | */ | |
8550eb99 | 1065 | static noinline int do_test_wp_bit(void) |
1da177e4 LT |
1066 | { |
1067 | char tmp_reg; | |
1068 | int flag; | |
1069 | ||
1070 | __asm__ __volatile__( | |
8550eb99 IM |
1071 | " movb %0, %1 \n" |
1072 | "1: movb %1, %0 \n" | |
1073 | " xorl %2, %2 \n" | |
1da177e4 | 1074 | "2: \n" |
f832ff18 | 1075 | _ASM_EXTABLE(1b,2b) |
1da177e4 LT |
1076 | :"=m" (*(char *)fix_to_virt(FIX_WP_TEST)), |
1077 | "=q" (tmp_reg), | |
1078 | "=r" (flag) | |
1079 | :"2" (1) | |
1080 | :"memory"); | |
8550eb99 | 1081 | |
1da177e4 LT |
1082 | return flag; |
1083 | } | |
1084 | ||
63aaf308 | 1085 | #ifdef CONFIG_DEBUG_RODATA |
edeed305 AV |
1086 | const int rodata_test_data = 0xC3; |
1087 | EXPORT_SYMBOL_GPL(rodata_test_data); | |
63aaf308 | 1088 | |
63aaf308 AV |
1089 | void mark_rodata_ro(void) |
1090 | { | |
6fb14755 JB |
1091 | unsigned long start = PFN_ALIGN(_text); |
1092 | unsigned long size = PFN_ALIGN(_etext) - start; | |
63aaf308 | 1093 | |
8f0f996e SR |
1094 | #ifndef CONFIG_DYNAMIC_FTRACE |
1095 | /* Dynamic tracing modifies the kernel text section */ | |
4e4eee0e MD |
1096 | set_pages_ro(virt_to_page(start), size >> PAGE_SHIFT); |
1097 | printk(KERN_INFO "Write protecting the kernel text: %luk\n", | |
1098 | size >> 10); | |
0c42f392 AK |
1099 | |
1100 | #ifdef CONFIG_CPA_DEBUG | |
4e4eee0e MD |
1101 | printk(KERN_INFO "Testing CPA: Reverting %lx-%lx\n", |
1102 | start, start+size); | |
1103 | set_pages_rw(virt_to_page(start), size>>PAGE_SHIFT); | |
0c42f392 | 1104 | |
4e4eee0e MD |
1105 | printk(KERN_INFO "Testing CPA: write protecting again\n"); |
1106 | set_pages_ro(virt_to_page(start), size>>PAGE_SHIFT); | |
602033ed | 1107 | #endif |
8f0f996e SR |
1108 | #endif /* CONFIG_DYNAMIC_FTRACE */ |
1109 | ||
6fb14755 JB |
1110 | start += size; |
1111 | size = (unsigned long)__end_rodata - start; | |
6d238cc4 | 1112 | set_pages_ro(virt_to_page(start), size >> PAGE_SHIFT); |
d7d119d7 IM |
1113 | printk(KERN_INFO "Write protecting the kernel read-only data: %luk\n", |
1114 | size >> 10); | |
edeed305 | 1115 | rodata_test(); |
63aaf308 | 1116 | |
0c42f392 | 1117 | #ifdef CONFIG_CPA_DEBUG |
d7d119d7 | 1118 | printk(KERN_INFO "Testing CPA: undo %lx-%lx\n", start, start + size); |
6d238cc4 | 1119 | set_pages_rw(virt_to_page(start), size >> PAGE_SHIFT); |
0c42f392 | 1120 | |
d7d119d7 | 1121 | printk(KERN_INFO "Testing CPA: write protecting again\n"); |
6d238cc4 | 1122 | set_pages_ro(virt_to_page(start), size >> PAGE_SHIFT); |
0c42f392 | 1123 | #endif |
63aaf308 AV |
1124 | } |
1125 | #endif | |
1126 | ||
9a0b5817 GH |
1127 | void free_init_pages(char *what, unsigned long begin, unsigned long end) |
1128 | { | |
ee01f112 IM |
1129 | #ifdef CONFIG_DEBUG_PAGEALLOC |
1130 | /* | |
1131 | * If debugging page accesses then do not free this memory but | |
1132 | * mark them not present - any buggy init-section access will | |
1133 | * create a kernel page fault: | |
1134 | */ | |
1135 | printk(KERN_INFO "debug: unmapping init memory %08lx..%08lx\n", | |
1136 | begin, PAGE_ALIGN(end)); | |
1137 | set_memory_np(begin, (end - begin) >> PAGE_SHIFT); | |
1138 | #else | |
86f03989 IM |
1139 | unsigned long addr; |
1140 | ||
3c1df68b AV |
1141 | /* |
1142 | * We just marked the kernel text read only above, now that | |
1143 | * we are going to free part of that, we need to make that | |
1144 | * writeable first. | |
1145 | */ | |
1146 | set_memory_rw(begin, (end - begin) >> PAGE_SHIFT); | |
1147 | ||
9a0b5817 | 1148 | for (addr = begin; addr < end; addr += PAGE_SIZE) { |
e3ebadd9 LT |
1149 | ClearPageReserved(virt_to_page(addr)); |
1150 | init_page_count(virt_to_page(addr)); | |
1151 | memset((void *)addr, POISON_FREE_INITMEM, PAGE_SIZE); | |
1152 | free_page(addr); | |
9a0b5817 GH |
1153 | totalram_pages++; |
1154 | } | |
6fb14755 | 1155 | printk(KERN_INFO "Freeing %s: %luk freed\n", what, (end - begin) >> 10); |
ee01f112 | 1156 | #endif |
9a0b5817 GH |
1157 | } |
1158 | ||
1159 | void free_initmem(void) | |
1160 | { | |
1161 | free_init_pages("unused kernel memory", | |
e3ebadd9 LT |
1162 | (unsigned long)(&__init_begin), |
1163 | (unsigned long)(&__init_end)); | |
9a0b5817 | 1164 | } |
63aaf308 | 1165 | |
1da177e4 LT |
1166 | #ifdef CONFIG_BLK_DEV_INITRD |
1167 | void free_initrd_mem(unsigned long start, unsigned long end) | |
1168 | { | |
e3ebadd9 | 1169 | free_init_pages("initrd memory", start, end); |
1da177e4 LT |
1170 | } |
1171 | #endif | |
d2dbf343 YL |
1172 | |
1173 | int __init reserve_bootmem_generic(unsigned long phys, unsigned long len, | |
1174 | int flags) | |
1175 | { | |
1176 | return reserve_bootmem(phys, len, flags); | |
1177 | } |