]>
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> | |
cfb80c9e | 24 | #include <linux/pci.h> |
6fb14755 | 25 | #include <linux/pfn.h> |
c9cf5528 | 26 | #include <linux/poison.h> |
1da177e4 | 27 | #include <linux/bootmem.h> |
a9ce6bc1 | 28 | #include <linux/memblock.h> |
1da177e4 | 29 | #include <linux/proc_fs.h> |
05039b92 | 30 | #include <linux/memory_hotplug.h> |
27d99f7e | 31 | #include <linux/initrd.h> |
55b2355e | 32 | #include <linux/cpumask.h> |
5a0e3ad6 | 33 | #include <linux/gfp.h> |
1da177e4 | 34 | |
f832ff18 | 35 | #include <asm/asm.h> |
46eaa670 | 36 | #include <asm/bios_ebda.h> |
1da177e4 | 37 | #include <asm/processor.h> |
1da177e4 LT |
38 | #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> | |
8550eb99 | 44 | #include <asm/bugs.h> |
1da177e4 LT |
45 | #include <asm/tlb.h> |
46 | #include <asm/tlbflush.h> | |
c10d1e26 | 47 | #include <asm/olpc_ofw.h> |
a5a19c63 | 48 | #include <asm/pgalloc.h> |
1da177e4 | 49 | #include <asm/sections.h> |
b239fb25 | 50 | #include <asm/paravirt.h> |
551889a6 | 51 | #include <asm/setup.h> |
7bfeab9a | 52 | #include <asm/cacheflush.h> |
2b72394e | 53 | #include <asm/page_types.h> |
4fcb2083 | 54 | #include <asm/init.h> |
1da177e4 | 55 | |
5c51bdbe YL |
56 | #include "mm_internal.h" |
57 | ||
1da177e4 LT |
58 | unsigned long highstart_pfn, highend_pfn; |
59 | ||
8550eb99 | 60 | static noinline int do_test_wp_bit(void); |
1da177e4 | 61 | |
dc16ecf7 | 62 | bool __read_mostly __vmalloc_start_set = false; |
4e29684c | 63 | |
1da177e4 LT |
64 | /* |
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 | */ | |
69 | static pmd_t * __init one_md_table_init(pgd_t *pgd) | |
70 | { | |
71 | pud_t *pud; | |
72 | pmd_t *pmd_table; | |
8550eb99 | 73 | |
1da177e4 | 74 | #ifdef CONFIG_X86_PAE |
b239fb25 | 75 | if (!(pgd_val(*pgd) & _PAGE_PRESENT)) { |
4e37a890 | 76 | pmd_table = (pmd_t *)alloc_low_page(); |
6944a9c8 | 77 | paravirt_alloc_pmd(&init_mm, __pa(pmd_table) >> PAGE_SHIFT); |
b239fb25 JF |
78 | set_pgd(pgd, __pgd(__pa(pmd_table) | _PAGE_PRESENT)); |
79 | pud = pud_offset(pgd, 0); | |
8550eb99 | 80 | BUG_ON(pmd_table != pmd_offset(pud, 0)); |
a376f30a Z |
81 | |
82 | return pmd_table; | |
b239fb25 JF |
83 | } |
84 | #endif | |
1da177e4 LT |
85 | pud = pud_offset(pgd, 0); |
86 | pmd_table = pmd_offset(pud, 0); | |
8550eb99 | 87 | |
1da177e4 LT |
88 | return pmd_table; |
89 | } | |
90 | ||
91 | /* | |
92 | * Create a page table and place a pointer to it in a middle page | |
8550eb99 | 93 | * directory entry: |
1da177e4 LT |
94 | */ |
95 | static pte_t * __init one_page_table_init(pmd_t *pmd) | |
96 | { | |
b239fb25 | 97 | if (!(pmd_val(*pmd) & _PAGE_PRESENT)) { |
4e37a890 | 98 | pte_t *page_table = (pte_t *)alloc_low_page(); |
b239fb25 | 99 | |
6944a9c8 | 100 | paravirt_alloc_pte(&init_mm, __pa(page_table) >> PAGE_SHIFT); |
1da177e4 | 101 | set_pmd(pmd, __pmd(__pa(page_table) | _PAGE_TABLE)); |
b239fb25 | 102 | BUG_ON(page_table != pte_offset_kernel(pmd, 0)); |
1da177e4 | 103 | } |
509a80c4 | 104 | |
1da177e4 LT |
105 | return pte_offset_kernel(pmd, 0); |
106 | } | |
107 | ||
458a3e64 | 108 | pmd_t * __init populate_extra_pmd(unsigned long vaddr) |
11124411 TH |
109 | { |
110 | int pgd_idx = pgd_index(vaddr); | |
111 | int pmd_idx = pmd_index(vaddr); | |
458a3e64 TH |
112 | |
113 | return one_md_table_init(swapper_pg_dir + pgd_idx) + pmd_idx; | |
114 | } | |
115 | ||
116 | pte_t * __init populate_extra_pte(unsigned long vaddr) | |
117 | { | |
118 | int pte_idx = pte_index(vaddr); | |
11124411 TH |
119 | pmd_t *pmd; |
120 | ||
458a3e64 TH |
121 | pmd = populate_extra_pmd(vaddr); |
122 | return one_page_table_init(pmd) + pte_idx; | |
11124411 TH |
123 | } |
124 | ||
719272c4 YL |
125 | static unsigned long __init |
126 | page_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 | ||
a3c6018e | 155 | static pte_t *__init page_table_kmap_check(pte_t *pte, pmd_t *pmd, |
719272c4 YL |
156 | unsigned long vaddr, pte_t *lastpte, |
157 | void **adr) | |
a3c6018e JB |
158 | { |
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 | |
719272c4 | 171 | && (vaddr >> PMD_SHIFT) <= pmd_idx_kmap_end) { |
a3c6018e JB |
172 | pte_t *newpte; |
173 | int i; | |
174 | ||
c464573c | 175 | BUG_ON(after_bootmem); |
719272c4 | 176 | newpte = *adr; |
a3c6018e JB |
177 | for (i = 0; i < PTRS_PER_PTE; i++) |
178 | set_pte(newpte + i, pte[i]); | |
719272c4 | 179 | *adr = (void *)(((unsigned long)(*adr)) + PAGE_SIZE); |
a3c6018e JB |
180 | |
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 | ||
1da177e4 | 196 | /* |
8550eb99 | 197 | * This function initializes a certain range of kernel virtual memory |
1da177e4 LT |
198 | * with new bootmem page tables, everywhere page tables are missing in |
199 | * the given range. | |
8550eb99 IM |
200 | * |
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 | |
1da177e4 LT |
203 | * checking the pgd every time. |
204 | */ | |
8550eb99 IM |
205 | static void __init |
206 | page_table_range_init(unsigned long start, unsigned long end, pgd_t *pgd_base) | |
1da177e4 | 207 | { |
1da177e4 LT |
208 | int pgd_idx, pmd_idx; |
209 | unsigned long vaddr; | |
8550eb99 IM |
210 | pgd_t *pgd; |
211 | pmd_t *pmd; | |
a3c6018e | 212 | pte_t *pte = NULL; |
719272c4 YL |
213 | unsigned long count = page_table_range_init_count(start, end); |
214 | void *adr = NULL; | |
215 | ||
216 | if (count) | |
217 | adr = alloc_low_pages(count); | |
1da177e4 LT |
218 | |
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++) { | |
b239fb25 JF |
225 | pmd = one_md_table_init(pgd); |
226 | pmd = pmd + pmd_index(vaddr); | |
8550eb99 IM |
227 | for (; (pmd_idx < PTRS_PER_PMD) && (vaddr != end); |
228 | pmd++, pmd_idx++) { | |
a3c6018e | 229 | pte = page_table_kmap_check(one_page_table_init(pmd), |
719272c4 | 230 | pmd, vaddr, pte, &adr); |
1da177e4 LT |
231 | |
232 | vaddr += PMD_SIZE; | |
233 | } | |
234 | pmd_idx = 0; | |
235 | } | |
236 | } | |
237 | ||
238 | static inline int is_kernel_text(unsigned long addr) | |
239 | { | |
5bd5a452 | 240 | if (addr >= (unsigned long)_text && addr <= (unsigned long)__init_end) |
1da177e4 LT |
241 | return 1; |
242 | return 0; | |
243 | } | |
244 | ||
245 | /* | |
8550eb99 IM |
246 | * 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: | |
1da177e4 | 249 | */ |
e53fb04f PE |
250 | unsigned long __init |
251 | kernel_physical_mapping_init(unsigned long start, | |
252 | unsigned long end, | |
253 | unsigned long page_size_mask) | |
1da177e4 | 254 | { |
e53fb04f | 255 | int use_pse = page_size_mask == (1<<PG_LEVEL_2M); |
c1fd1b43 | 256 | unsigned long last_map_addr = end; |
e53fb04f | 257 | unsigned long start_pfn, end_pfn; |
e7179853 | 258 | pgd_t *pgd_base = swapper_pg_dir; |
8550eb99 | 259 | int pgd_idx, pmd_idx, pte_ofs; |
1da177e4 LT |
260 | unsigned long pfn; |
261 | pgd_t *pgd; | |
262 | pmd_t *pmd; | |
263 | pte_t *pte; | |
a2699e47 SS |
264 | unsigned pages_2m, pages_4k; |
265 | int mapping_iter; | |
266 | ||
e53fb04f PE |
267 | start_pfn = start >> PAGE_SHIFT; |
268 | end_pfn = end >> PAGE_SHIFT; | |
269 | ||
a2699e47 SS |
270 | /* |
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; | |
1da177e4 | 285 | |
a04ad82d YL |
286 | if (!cpu_has_pse) |
287 | use_pse = 0; | |
1da177e4 | 288 | |
a2699e47 SS |
289 | repeat: |
290 | pages_2m = pages_4k = 0; | |
a04ad82d YL |
291 | pfn = start_pfn; |
292 | pgd_idx = pgd_index((pfn<<PAGE_SHIFT) + PAGE_OFFSET); | |
293 | pgd = pgd_base + pgd_idx; | |
1da177e4 LT |
294 | for (; pgd_idx < PTRS_PER_PGD; pgd++, pgd_idx++) { |
295 | pmd = one_md_table_init(pgd); | |
8550eb99 | 296 | |
a04ad82d YL |
297 | 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; | |
f3f20de8 | 306 | pmd++, pmd_idx++) { |
8550eb99 | 307 | unsigned int addr = pfn * PAGE_SIZE + PAGE_OFFSET; |
1da177e4 | 308 | |
8550eb99 IM |
309 | /* |
310 | * Map with big pages if possible, otherwise | |
311 | * create normal page tables: | |
312 | */ | |
a04ad82d | 313 | if (use_pse) { |
8550eb99 | 314 | unsigned int addr2; |
f3f20de8 | 315 | pgprot_t prot = PAGE_KERNEL_LARGE; |
a2699e47 SS |
316 | /* |
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); | |
f3f20de8 | 323 | |
960ddb4f | 324 | pfn &= PMD_MASK >> PAGE_SHIFT; |
8550eb99 | 325 | addr2 = (pfn + PTRS_PER_PTE-1) * PAGE_SIZE + |
f3f20de8 JF |
326 | PAGE_OFFSET + PAGE_SIZE-1; |
327 | ||
8550eb99 IM |
328 | if (is_kernel_text(addr) || |
329 | is_kernel_text(addr2)) | |
f3f20de8 JF |
330 | prot = PAGE_KERNEL_LARGE_EXEC; |
331 | ||
ce0c0e50 | 332 | pages_2m++; |
a2699e47 SS |
333 | if (mapping_iter == 1) |
334 | set_pmd(pmd, pfn_pmd(pfn, init_prot)); | |
335 | else | |
336 | set_pmd(pmd, pfn_pmd(pfn, prot)); | |
b239fb25 | 337 | |
1da177e4 | 338 | pfn += PTRS_PER_PTE; |
8550eb99 IM |
339 | continue; |
340 | } | |
341 | pte = one_page_table_init(pmd); | |
1da177e4 | 342 | |
a04ad82d YL |
343 | pte_ofs = pte_index((pfn<<PAGE_SHIFT) + PAGE_OFFSET); |
344 | pte += pte_ofs; | |
345 | for (; pte_ofs < PTRS_PER_PTE && pfn < end_pfn; | |
8550eb99 IM |
346 | pte++, pfn++, pte_ofs++, addr += PAGE_SIZE) { |
347 | pgprot_t prot = PAGE_KERNEL; | |
a2699e47 SS |
348 | /* |
349 | * first pass will use the same initial | |
350 | * identity mapping attribute. | |
351 | */ | |
352 | pgprot_t init_prot = __pgprot(PTE_IDENT_ATTR); | |
f3f20de8 | 353 | |
8550eb99 IM |
354 | if (is_kernel_text(addr)) |
355 | prot = PAGE_KERNEL_EXEC; | |
f3f20de8 | 356 | |
ce0c0e50 | 357 | pages_4k++; |
c1fd1b43 | 358 | if (mapping_iter == 1) { |
a2699e47 | 359 | set_pte(pte, pfn_pte(pfn, init_prot)); |
c1fd1b43 PE |
360 | last_map_addr = (pfn << PAGE_SHIFT) + PAGE_SIZE; |
361 | } else | |
a2699e47 | 362 | set_pte(pte, pfn_pte(pfn, prot)); |
1da177e4 LT |
363 | } |
364 | } | |
365 | } | |
a2699e47 SS |
366 | 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 | } | |
c1fd1b43 | 386 | return last_map_addr; |
ae531c26 AV |
387 | } |
388 | ||
1da177e4 LT |
389 | pte_t *kmap_pte; |
390 | pgprot_t kmap_prot; | |
391 | ||
8550eb99 IM |
392 | static 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 | } | |
1da177e4 LT |
397 | |
398 | static void __init kmap_init(void) | |
399 | { | |
400 | unsigned long kmap_vstart; | |
401 | ||
8550eb99 IM |
402 | /* |
403 | * Cache the first kmap pte: | |
404 | */ | |
1da177e4 LT |
405 | 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 | ||
fd940934 | 411 | #ifdef CONFIG_HIGHMEM |
1da177e4 LT |
412 | static void __init permanent_kmaps_init(pgd_t *pgd_base) |
413 | { | |
8550eb99 | 414 | unsigned long vaddr; |
1da177e4 LT |
415 | pgd_t *pgd; |
416 | pud_t *pud; | |
417 | pmd_t *pmd; | |
418 | pte_t *pte; | |
1da177e4 LT |
419 | |
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); | |
8550eb99 | 427 | pkmap_page_table = pte; |
1da177e4 LT |
428 | } |
429 | ||
1d931264 YL |
430 | void __init add_highpages_with_active_regions(int nid, |
431 | unsigned long start_pfn, unsigned long end_pfn) | |
1da177e4 | 432 | { |
8a9ca34c TH |
433 | phys_addr_t start, end; |
434 | u64 i; | |
435 | ||
436 | for_each_free_mem_range(i, nid, &start, &end, NULL) { | |
437 | unsigned long pfn = clamp_t(unsigned long, PFN_UP(start), | |
438 | start_pfn, end_pfn); | |
439 | unsigned long e_pfn = clamp_t(unsigned long, PFN_DOWN(end), | |
440 | start_pfn, end_pfn); | |
441 | for ( ; pfn < e_pfn; pfn++) | |
442 | if (pfn_valid(pfn)) | |
5e7ccf86 | 443 | free_highmem_page(pfn_to_page(pfn)); |
23be8c7d | 444 | } |
b5bc6c0e | 445 | } |
1da177e4 | 446 | #else |
e8e32326 IB |
447 | static inline void permanent_kmaps_init(pgd_t *pgd_base) |
448 | { | |
449 | } | |
1da177e4 LT |
450 | #endif /* CONFIG_HIGHMEM */ |
451 | ||
7737b215 | 452 | void __init native_pagetable_init(void) |
1da177e4 | 453 | { |
551889a6 | 454 | unsigned long pfn, va; |
73090f89 | 455 | pgd_t *pgd, *base = swapper_pg_dir; |
551889a6 IC |
456 | pud_t *pud; |
457 | pmd_t *pmd; | |
458 | pte_t *pte; | |
b239fb25 JF |
459 | |
460 | /* | |
551889a6 | 461 | * Remove any mappings which extend past the end of physical |
11ed9e92 YL |
462 | * memory from the boot time page table. |
463 | * In virtual address space, we should have at least two pages | |
464 | * from VMALLOC_END to pkmap or fixmap according to VMALLOC_END | |
465 | * definition. And max_low_pfn is set to VMALLOC_END physical | |
466 | * address. If initial memory mapping is doing right job, we | |
467 | * should have pte used near max_low_pfn or one pmd is not present. | |
b239fb25 | 468 | */ |
11ed9e92 | 469 | for (pfn = max_low_pfn; pfn < 1<<(32-PAGE_SHIFT); pfn++) { |
551889a6 IC |
470 | va = PAGE_OFFSET + (pfn<<PAGE_SHIFT); |
471 | pgd = base + pgd_index(va); | |
472 | if (!pgd_present(*pgd)) | |
473 | break; | |
474 | ||
475 | pud = pud_offset(pgd, va); | |
476 | pmd = pmd_offset(pud, va); | |
477 | if (!pmd_present(*pmd)) | |
478 | break; | |
479 | ||
11ed9e92 YL |
480 | /* should not be large page here */ |
481 | if (pmd_large(*pmd)) { | |
482 | 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", | |
483 | pfn, pmd, __pa(pmd)); | |
484 | BUG_ON(1); | |
485 | } | |
486 | ||
551889a6 IC |
487 | pte = pte_offset_kernel(pmd, va); |
488 | if (!pte_present(*pte)) | |
489 | break; | |
490 | ||
11ed9e92 YL |
491 | printk(KERN_DEBUG "clearing pte for ram above max_low_pfn: pfn: %lx pmd: %p pmd phys: %lx pte: %p pte phys: %lx\n", |
492 | pfn, pmd, __pa(pmd), pte, __pa(pte)); | |
551889a6 IC |
493 | pte_clear(NULL, va, pte); |
494 | } | |
6944a9c8 | 495 | paravirt_alloc_pmd(&init_mm, __pa(base) >> PAGE_SHIFT); |
843b8ed2 | 496 | paging_init(); |
b239fb25 JF |
497 | } |
498 | ||
b239fb25 JF |
499 | /* |
500 | * Build a proper pagetable for the kernel mappings. Up until this | |
501 | * point, we've been running on some set of pagetables constructed by | |
502 | * the boot process. | |
503 | * | |
504 | * If we're booting on native hardware, this will be a pagetable | |
551889a6 IC |
505 | * constructed in arch/x86/kernel/head_32.S. The root of the |
506 | * pagetable will be swapper_pg_dir. | |
b239fb25 JF |
507 | * |
508 | * If we're booting paravirtualized under a hypervisor, then there are | |
509 | * more options: we may already be running PAE, and the pagetable may | |
510 | * or may not be based in swapper_pg_dir. In any case, | |
7737b215 | 511 | * paravirt_pagetable_init() will set up swapper_pg_dir |
b239fb25 JF |
512 | * appropriately for the rest of the initialization to work. |
513 | * | |
514 | * In general, pagetable_init() assumes that the pagetable may already | |
515 | * be partially populated, and so it avoids stomping on any existing | |
516 | * mappings. | |
517 | */ | |
f765090a | 518 | void __init early_ioremap_page_table_range_init(void) |
b239fb25 | 519 | { |
e7179853 | 520 | pgd_t *pgd_base = swapper_pg_dir; |
8550eb99 | 521 | unsigned long vaddr, end; |
b239fb25 | 522 | |
1da177e4 LT |
523 | /* |
524 | * Fixed mappings, only the page table structure has to be | |
525 | * created - mappings will be set by set_fixmap(): | |
526 | */ | |
527 | vaddr = __fix_to_virt(__end_of_fixed_addresses - 1) & PMD_MASK; | |
b239fb25 JF |
528 | end = (FIXADDR_TOP + PMD_SIZE - 1) & PMD_MASK; |
529 | page_table_range_init(vaddr, end, pgd_base); | |
beacfaac | 530 | early_ioremap_reset(); |
e7b37895 YL |
531 | } |
532 | ||
533 | static void __init pagetable_init(void) | |
534 | { | |
535 | pgd_t *pgd_base = swapper_pg_dir; | |
536 | ||
1da177e4 | 537 | permanent_kmaps_init(pgd_base); |
1da177e4 LT |
538 | } |
539 | ||
be43d728 | 540 | pteval_t __supported_pte_mask __read_mostly = ~(_PAGE_NX | _PAGE_GLOBAL | _PAGE_IOMAP); |
6fdc05d4 JF |
541 | EXPORT_SYMBOL_GPL(__supported_pte_mask); |
542 | ||
90d967e0 YL |
543 | /* user-defined highmem size */ |
544 | static unsigned int highmem_pages = -1; | |
545 | ||
546 | /* | |
547 | * highmem=size forces highmem to be exactly 'size' bytes. | |
548 | * This works even on boxes that have no highmem otherwise. | |
549 | * This also works to reduce highmem size on bigger boxes. | |
550 | */ | |
551 | static int __init parse_highmem(char *arg) | |
552 | { | |
553 | if (!arg) | |
554 | return -EINVAL; | |
555 | ||
556 | highmem_pages = memparse(arg, &arg) >> PAGE_SHIFT; | |
557 | return 0; | |
558 | } | |
559 | early_param("highmem", parse_highmem); | |
560 | ||
4769843b IM |
561 | #define MSG_HIGHMEM_TOO_BIG \ |
562 | "highmem size (%luMB) is bigger than pages available (%luMB)!\n" | |
563 | ||
564 | #define MSG_LOWMEM_TOO_SMALL \ | |
565 | "highmem size (%luMB) results in <64MB lowmem, ignoring it!\n" | |
90d967e0 | 566 | /* |
4769843b IM |
567 | * All of RAM fits into lowmem - but if user wants highmem |
568 | * artificially via the highmem=x boot parameter then create | |
569 | * it: | |
90d967e0 | 570 | */ |
f836e35a | 571 | static void __init lowmem_pfn_init(void) |
90d967e0 | 572 | { |
346cafec | 573 | /* max_low_pfn is 0, we already have early_res support */ |
90d967e0 | 574 | max_low_pfn = max_pfn; |
d88316c2 | 575 | |
4769843b IM |
576 | if (highmem_pages == -1) |
577 | highmem_pages = 0; | |
578 | #ifdef CONFIG_HIGHMEM | |
579 | if (highmem_pages >= max_pfn) { | |
580 | printk(KERN_ERR MSG_HIGHMEM_TOO_BIG, | |
581 | pages_to_mb(highmem_pages), pages_to_mb(max_pfn)); | |
582 | highmem_pages = 0; | |
583 | } | |
584 | if (highmem_pages) { | |
585 | if (max_low_pfn - highmem_pages < 64*1024*1024/PAGE_SIZE) { | |
586 | printk(KERN_ERR MSG_LOWMEM_TOO_SMALL, | |
90d967e0 YL |
587 | pages_to_mb(highmem_pages)); |
588 | highmem_pages = 0; | |
589 | } | |
4769843b IM |
590 | max_low_pfn -= highmem_pages; |
591 | } | |
592 | #else | |
593 | if (highmem_pages) | |
594 | printk(KERN_ERR "ignoring highmem size on non-highmem kernel!\n"); | |
595 | #endif | |
596 | } | |
597 | ||
598 | #define MSG_HIGHMEM_TOO_SMALL \ | |
599 | "only %luMB highmem pages available, ignoring highmem size of %luMB!\n" | |
600 | ||
601 | #define MSG_HIGHMEM_TRIMMED \ | |
602 | "Warning: only 4GB will be used. Use a HIGHMEM64G enabled kernel!\n" | |
603 | /* | |
604 | * We have more RAM than fits into lowmem - we try to put it into | |
605 | * highmem, also taking the highmem=x boot parameter into account: | |
606 | */ | |
f836e35a | 607 | static void __init highmem_pfn_init(void) |
4769843b | 608 | { |
d88316c2 IM |
609 | max_low_pfn = MAXMEM_PFN; |
610 | ||
4769843b IM |
611 | if (highmem_pages == -1) |
612 | highmem_pages = max_pfn - MAXMEM_PFN; | |
613 | ||
614 | if (highmem_pages + MAXMEM_PFN < max_pfn) | |
615 | max_pfn = MAXMEM_PFN + highmem_pages; | |
616 | ||
617 | if (highmem_pages + MAXMEM_PFN > max_pfn) { | |
618 | printk(KERN_WARNING MSG_HIGHMEM_TOO_SMALL, | |
619 | pages_to_mb(max_pfn - MAXMEM_PFN), | |
620 | pages_to_mb(highmem_pages)); | |
621 | highmem_pages = 0; | |
622 | } | |
90d967e0 | 623 | #ifndef CONFIG_HIGHMEM |
4769843b IM |
624 | /* Maximum memory usable is what is directly addressable */ |
625 | printk(KERN_WARNING "Warning only %ldMB will be used.\n", MAXMEM>>20); | |
626 | if (max_pfn > MAX_NONPAE_PFN) | |
627 | printk(KERN_WARNING "Use a HIGHMEM64G enabled kernel.\n"); | |
628 | else | |
629 | printk(KERN_WARNING "Use a HIGHMEM enabled kernel.\n"); | |
630 | max_pfn = MAXMEM_PFN; | |
90d967e0 YL |
631 | #else /* !CONFIG_HIGHMEM */ |
632 | #ifndef CONFIG_HIGHMEM64G | |
4769843b IM |
633 | if (max_pfn > MAX_NONPAE_PFN) { |
634 | max_pfn = MAX_NONPAE_PFN; | |
635 | printk(KERN_WARNING MSG_HIGHMEM_TRIMMED); | |
636 | } | |
90d967e0 YL |
637 | #endif /* !CONFIG_HIGHMEM64G */ |
638 | #endif /* !CONFIG_HIGHMEM */ | |
4769843b IM |
639 | } |
640 | ||
641 | /* | |
642 | * Determine low and high memory ranges: | |
643 | */ | |
644 | void __init find_low_pfn_range(void) | |
645 | { | |
646 | /* it could update max_pfn */ | |
647 | ||
d88316c2 | 648 | if (max_pfn <= MAXMEM_PFN) |
4769843b | 649 | lowmem_pfn_init(); |
d88316c2 IM |
650 | else |
651 | highmem_pfn_init(); | |
90d967e0 YL |
652 | } |
653 | ||
b2ac82a0 | 654 | #ifndef CONFIG_NEED_MULTIPLE_NODES |
d8fc3afc | 655 | void __init initmem_init(void) |
b2ac82a0 | 656 | { |
b2ac82a0 YL |
657 | #ifdef CONFIG_HIGHMEM |
658 | highstart_pfn = highend_pfn = max_pfn; | |
659 | if (max_pfn > max_low_pfn) | |
660 | highstart_pfn = max_low_pfn; | |
b2ac82a0 YL |
661 | printk(KERN_NOTICE "%ldMB HIGHMEM available.\n", |
662 | pages_to_mb(highend_pfn - highstart_pfn)); | |
b2ac82a0 YL |
663 | high_memory = (void *) __va(highstart_pfn * PAGE_SIZE - 1) + 1; |
664 | #else | |
b2ac82a0 YL |
665 | high_memory = (void *) __va(max_low_pfn * PAGE_SIZE - 1) + 1; |
666 | #endif | |
0608f70c | 667 | |
e7e8de59 | 668 | memblock_set_node(0, (phys_addr_t)ULLONG_MAX, &memblock.memory, 0); |
0608f70c TH |
669 | sparse_memory_present_with_active_regions(0); |
670 | ||
b2ac82a0 | 671 | #ifdef CONFIG_FLATMEM |
46a84132 | 672 | max_mapnr = IS_ENABLED(CONFIG_HIGHMEM) ? highend_pfn : max_low_pfn; |
b2ac82a0 | 673 | #endif |
dc16ecf7 JF |
674 | __vmalloc_start_set = true; |
675 | ||
b2ac82a0 YL |
676 | printk(KERN_NOTICE "%ldMB LOWMEM available.\n", |
677 | pages_to_mb(max_low_pfn)); | |
678 | ||
679 | setup_bootmem_allocator(); | |
b2ac82a0 | 680 | } |
cb95a13a | 681 | #endif /* !CONFIG_NEED_MULTIPLE_NODES */ |
b2ac82a0 | 682 | |
b2ac82a0 YL |
683 | void __init setup_bootmem_allocator(void) |
684 | { | |
b2ac82a0 YL |
685 | printk(KERN_INFO " mapped low ram: 0 - %08lx\n", |
686 | max_pfn_mapped<<PAGE_SHIFT); | |
fc5efe39 | 687 | printk(KERN_INFO " low ram: 0 - %08lx\n", max_low_pfn<<PAGE_SHIFT); |
4e29684c YL |
688 | } |
689 | ||
1da177e4 LT |
690 | /* |
691 | * paging_init() sets up the page tables - note that the first 8MB are | |
692 | * already mapped by head.S. | |
693 | * | |
694 | * This routines also unmaps the page at virtual kernel address 0, so | |
695 | * that we can trap those pesky NULL-reference errors in the kernel. | |
696 | */ | |
697 | void __init paging_init(void) | |
698 | { | |
1da177e4 LT |
699 | pagetable_init(); |
700 | ||
1da177e4 LT |
701 | __flush_tlb_all(); |
702 | ||
703 | kmap_init(); | |
11cd0bc1 YL |
704 | |
705 | /* | |
706 | * NOTE: at this point the bootmem allocator is fully available. | |
707 | */ | |
c10d1e26 | 708 | olpc_dt_build_devicetree(); |
797390d8 | 709 | sparse_memory_present_with_active_regions(MAX_NUMNODES); |
11cd0bc1 YL |
710 | sparse_init(); |
711 | zone_sizes_init(); | |
1da177e4 LT |
712 | } |
713 | ||
714 | /* | |
715 | * Test if the WP bit works in supervisor mode. It isn't supported on 386's | |
f7f17a67 DV |
716 | * and also on some strange 486's. All 586+'s are OK. This used to involve |
717 | * black magic jumps to work around some nasty CPU bugs, but fortunately the | |
718 | * switch to using exceptions got rid of all that. | |
1da177e4 | 719 | */ |
1da177e4 LT |
720 | static void __init test_wp_bit(void) |
721 | { | |
d7d119d7 IM |
722 | printk(KERN_INFO |
723 | "Checking if this processor honours the WP bit even in supervisor mode..."); | |
1da177e4 LT |
724 | |
725 | /* Any page-aligned address will do, the test is non-destructive */ | |
8bd753be | 726 | __set_fixmap(FIX_WP_TEST, __pa(&swapper_pg_dir), PAGE_KERNEL_RO); |
1da177e4 LT |
727 | boot_cpu_data.wp_works_ok = do_test_wp_bit(); |
728 | clear_fixmap(FIX_WP_TEST); | |
729 | ||
730 | if (!boot_cpu_data.wp_works_ok) { | |
d7d119d7 | 731 | printk(KERN_CONT "No.\n"); |
a5c2a893 | 732 | panic("Linux doesn't support CPUs with broken WP."); |
1da177e4 | 733 | } else { |
d7d119d7 | 734 | printk(KERN_CONT "Ok.\n"); |
1da177e4 LT |
735 | } |
736 | } | |
737 | ||
1da177e4 LT |
738 | void __init mem_init(void) |
739 | { | |
cfb80c9e JF |
740 | pci_iommu_alloc(); |
741 | ||
05b79bdc | 742 | #ifdef CONFIG_FLATMEM |
8d8f3cbe | 743 | BUG_ON(!mem_map); |
1da177e4 | 744 | #endif |
855c743a SG |
745 | /* |
746 | * With CONFIG_DEBUG_PAGEALLOC initialization of highmem pages has to | |
747 | * be done before free_all_bootmem(). Memblock use free low memory for | |
748 | * temporary data (see find_range_array()) and for this purpose can use | |
749 | * pages that was already passed to the buddy allocator, hence marked as | |
750 | * not accessible in the page tables when compiled with | |
751 | * CONFIG_DEBUG_PAGEALLOC. Otherwise order of initialization is not | |
752 | * important here. | |
753 | */ | |
754 | set_highmem_pages_init(); | |
755 | ||
1da177e4 | 756 | /* this will put all low memory onto the freelists */ |
0c988534 | 757 | free_all_bootmem(); |
1da177e4 | 758 | |
4e37a890 YL |
759 | after_bootmem = 1; |
760 | ||
46a84132 | 761 | mem_init_print_info(NULL); |
d7d119d7 | 762 | printk(KERN_INFO "virtual kernel memory layout:\n" |
8550eb99 | 763 | " fixmap : 0x%08lx - 0x%08lx (%4ld kB)\n" |
052e7994 | 764 | #ifdef CONFIG_HIGHMEM |
8550eb99 | 765 | " pkmap : 0x%08lx - 0x%08lx (%4ld kB)\n" |
052e7994 | 766 | #endif |
8550eb99 IM |
767 | " vmalloc : 0x%08lx - 0x%08lx (%4ld MB)\n" |
768 | " lowmem : 0x%08lx - 0x%08lx (%4ld MB)\n" | |
769 | " .init : 0x%08lx - 0x%08lx (%4ld kB)\n" | |
770 | " .data : 0x%08lx - 0x%08lx (%4ld kB)\n" | |
771 | " .text : 0x%08lx - 0x%08lx (%4ld kB)\n", | |
772 | FIXADDR_START, FIXADDR_TOP, | |
773 | (FIXADDR_TOP - FIXADDR_START) >> 10, | |
052e7994 JF |
774 | |
775 | #ifdef CONFIG_HIGHMEM | |
8550eb99 IM |
776 | PKMAP_BASE, PKMAP_BASE+LAST_PKMAP*PAGE_SIZE, |
777 | (LAST_PKMAP*PAGE_SIZE) >> 10, | |
052e7994 JF |
778 | #endif |
779 | ||
8550eb99 IM |
780 | VMALLOC_START, VMALLOC_END, |
781 | (VMALLOC_END - VMALLOC_START) >> 20, | |
052e7994 | 782 | |
8550eb99 IM |
783 | (unsigned long)__va(0), (unsigned long)high_memory, |
784 | ((unsigned long)high_memory - (unsigned long)__va(0)) >> 20, | |
052e7994 | 785 | |
8550eb99 IM |
786 | (unsigned long)&__init_begin, (unsigned long)&__init_end, |
787 | ((unsigned long)&__init_end - | |
788 | (unsigned long)&__init_begin) >> 10, | |
052e7994 | 789 | |
8550eb99 IM |
790 | (unsigned long)&_etext, (unsigned long)&_edata, |
791 | ((unsigned long)&_edata - (unsigned long)&_etext) >> 10, | |
052e7994 | 792 | |
8550eb99 IM |
793 | (unsigned long)&_text, (unsigned long)&_etext, |
794 | ((unsigned long)&_etext - (unsigned long)&_text) >> 10); | |
052e7994 | 795 | |
beeb4195 JB |
796 | /* |
797 | * Check boundaries twice: Some fundamental inconsistencies can | |
798 | * be detected at build time already. | |
799 | */ | |
800 | #define __FIXADDR_TOP (-PAGE_SIZE) | |
801 | #ifdef CONFIG_HIGHMEM | |
802 | BUILD_BUG_ON(PKMAP_BASE + LAST_PKMAP*PAGE_SIZE > FIXADDR_START); | |
803 | BUILD_BUG_ON(VMALLOC_END > PKMAP_BASE); | |
804 | #endif | |
805 | #define high_memory (-128UL << 20) | |
806 | BUILD_BUG_ON(VMALLOC_START >= VMALLOC_END); | |
807 | #undef high_memory | |
808 | #undef __FIXADDR_TOP | |
6145cfe3 KC |
809 | #ifdef CONFIG_RANDOMIZE_BASE |
810 | BUILD_BUG_ON(CONFIG_RANDOMIZE_BASE_MAX_OFFSET > KERNEL_IMAGE_SIZE); | |
811 | #endif | |
beeb4195 | 812 | |
052e7994 | 813 | #ifdef CONFIG_HIGHMEM |
8550eb99 IM |
814 | BUG_ON(PKMAP_BASE + LAST_PKMAP*PAGE_SIZE > FIXADDR_START); |
815 | BUG_ON(VMALLOC_END > PKMAP_BASE); | |
052e7994 | 816 | #endif |
beeb4195 | 817 | BUG_ON(VMALLOC_START >= VMALLOC_END); |
8550eb99 | 818 | BUG_ON((unsigned long)high_memory > VMALLOC_START); |
052e7994 | 819 | |
1da177e4 LT |
820 | if (boot_cpu_data.wp_works_ok < 0) |
821 | test_wp_bit(); | |
1da177e4 LT |
822 | } |
823 | ||
ad8f5797 | 824 | #ifdef CONFIG_MEMORY_HOTPLUG |
bc02af93 | 825 | int arch_add_memory(int nid, u64 start, u64 size) |
05039b92 | 826 | { |
7c7e9425 | 827 | struct pglist_data *pgdata = NODE_DATA(nid); |
03d4be64 WN |
828 | struct zone *zone = pgdata->node_zones + |
829 | zone_for_memory(nid, start, size, ZONE_HIGHMEM); | |
05039b92 DH |
830 | unsigned long start_pfn = start >> PAGE_SHIFT; |
831 | unsigned long nr_pages = size >> PAGE_SHIFT; | |
832 | ||
c04fc586 | 833 | return __add_pages(nid, zone, start_pfn, nr_pages); |
05039b92 | 834 | } |
24d335ca WC |
835 | |
836 | #ifdef CONFIG_MEMORY_HOTREMOVE | |
837 | int arch_remove_memory(u64 start, u64 size) | |
838 | { | |
839 | unsigned long start_pfn = start >> PAGE_SHIFT; | |
840 | unsigned long nr_pages = size >> PAGE_SHIFT; | |
841 | struct zone *zone; | |
842 | ||
843 | zone = page_zone(pfn_to_page(start_pfn)); | |
844 | return __remove_pages(zone, start_pfn, nr_pages); | |
845 | } | |
846 | #endif | |
9d99aaa3 | 847 | #endif |
05039b92 | 848 | |
1da177e4 LT |
849 | /* |
850 | * This function cannot be __init, since exceptions don't work in that | |
851 | * section. Put this after the callers, so that it cannot be inlined. | |
852 | */ | |
8550eb99 | 853 | static noinline int do_test_wp_bit(void) |
1da177e4 LT |
854 | { |
855 | char tmp_reg; | |
856 | int flag; | |
857 | ||
858 | __asm__ __volatile__( | |
8550eb99 IM |
859 | " movb %0, %1 \n" |
860 | "1: movb %1, %0 \n" | |
861 | " xorl %2, %2 \n" | |
1da177e4 | 862 | "2: \n" |
f832ff18 | 863 | _ASM_EXTABLE(1b,2b) |
1da177e4 LT |
864 | :"=m" (*(char *)fix_to_virt(FIX_WP_TEST)), |
865 | "=q" (tmp_reg), | |
866 | "=r" (flag) | |
867 | :"2" (1) | |
868 | :"memory"); | |
8550eb99 | 869 | |
1da177e4 LT |
870 | return flag; |
871 | } | |
872 | ||
63aaf308 | 873 | #ifdef CONFIG_DEBUG_RODATA |
edeed305 AV |
874 | const int rodata_test_data = 0xC3; |
875 | EXPORT_SYMBOL_GPL(rodata_test_data); | |
63aaf308 | 876 | |
502f6604 | 877 | int kernel_set_to_readonly __read_mostly; |
16239630 SR |
878 | |
879 | void set_kernel_text_rw(void) | |
880 | { | |
881 | unsigned long start = PFN_ALIGN(_text); | |
882 | unsigned long size = PFN_ALIGN(_etext) - start; | |
883 | ||
884 | if (!kernel_set_to_readonly) | |
885 | return; | |
886 | ||
887 | pr_debug("Set kernel text: %lx - %lx for read write\n", | |
888 | start, start+size); | |
889 | ||
890 | set_pages_rw(virt_to_page(start), size >> PAGE_SHIFT); | |
891 | } | |
892 | ||
893 | void set_kernel_text_ro(void) | |
894 | { | |
895 | unsigned long start = PFN_ALIGN(_text); | |
896 | unsigned long size = PFN_ALIGN(_etext) - start; | |
897 | ||
898 | if (!kernel_set_to_readonly) | |
899 | return; | |
900 | ||
901 | pr_debug("Set kernel text: %lx - %lx for read only\n", | |
902 | start, start+size); | |
903 | ||
904 | set_pages_ro(virt_to_page(start), size >> PAGE_SHIFT); | |
905 | } | |
906 | ||
5bd5a452 MC |
907 | static void mark_nxdata_nx(void) |
908 | { | |
909 | /* | |
910 | * When this called, init has already been executed and released, | |
0d2eb44f | 911 | * so everything past _etext should be NX. |
5bd5a452 MC |
912 | */ |
913 | unsigned long start = PFN_ALIGN(_etext); | |
914 | /* | |
915 | * This comes from is_kernel_text upper limit. Also HPAGE where used: | |
916 | */ | |
917 | unsigned long size = (((unsigned long)__init_end + HPAGE_SIZE) & HPAGE_MASK) - start; | |
918 | ||
919 | if (__supported_pte_mask & _PAGE_NX) | |
920 | printk(KERN_INFO "NX-protecting the kernel data: %luk\n", size >> 10); | |
921 | set_pages_nx(virt_to_page(start), size >> PAGE_SHIFT); | |
922 | } | |
923 | ||
63aaf308 AV |
924 | void mark_rodata_ro(void) |
925 | { | |
6fb14755 JB |
926 | unsigned long start = PFN_ALIGN(_text); |
927 | unsigned long size = PFN_ALIGN(_etext) - start; | |
63aaf308 | 928 | |
4e4eee0e MD |
929 | set_pages_ro(virt_to_page(start), size >> PAGE_SHIFT); |
930 | printk(KERN_INFO "Write protecting the kernel text: %luk\n", | |
931 | size >> 10); | |
0c42f392 | 932 | |
16239630 SR |
933 | kernel_set_to_readonly = 1; |
934 | ||
0c42f392 | 935 | #ifdef CONFIG_CPA_DEBUG |
4e4eee0e MD |
936 | printk(KERN_INFO "Testing CPA: Reverting %lx-%lx\n", |
937 | start, start+size); | |
938 | set_pages_rw(virt_to_page(start), size>>PAGE_SHIFT); | |
0c42f392 | 939 | |
4e4eee0e MD |
940 | printk(KERN_INFO "Testing CPA: write protecting again\n"); |
941 | set_pages_ro(virt_to_page(start), size>>PAGE_SHIFT); | |
602033ed | 942 | #endif |
8f0f996e | 943 | |
6fb14755 JB |
944 | start += size; |
945 | size = (unsigned long)__end_rodata - start; | |
6d238cc4 | 946 | set_pages_ro(virt_to_page(start), size >> PAGE_SHIFT); |
d7d119d7 IM |
947 | printk(KERN_INFO "Write protecting the kernel read-only data: %luk\n", |
948 | size >> 10); | |
edeed305 | 949 | rodata_test(); |
63aaf308 | 950 | |
0c42f392 | 951 | #ifdef CONFIG_CPA_DEBUG |
d7d119d7 | 952 | printk(KERN_INFO "Testing CPA: undo %lx-%lx\n", start, start + size); |
6d238cc4 | 953 | set_pages_rw(virt_to_page(start), size >> PAGE_SHIFT); |
0c42f392 | 954 | |
d7d119d7 | 955 | printk(KERN_INFO "Testing CPA: write protecting again\n"); |
6d238cc4 | 956 | set_pages_ro(virt_to_page(start), size >> PAGE_SHIFT); |
0c42f392 | 957 | #endif |
5bd5a452 | 958 | mark_nxdata_nx(); |
63aaf308 AV |
959 | } |
960 | #endif | |
961 |