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)) { |
c464573c | 76 | if (after_bootmem) |
3c1596ef | 77 | pmd_table = (pmd_t *)alloc_bootmem_pages(PAGE_SIZE); |
4e29684c | 78 | else |
d6be89ad | 79 | pmd_table = (pmd_t *)alloc_low_page(); |
6944a9c8 | 80 | paravirt_alloc_pmd(&init_mm, __pa(pmd_table) >> PAGE_SHIFT); |
b239fb25 JF |
81 | set_pgd(pgd, __pgd(__pa(pmd_table) | _PAGE_PRESENT)); |
82 | pud = pud_offset(pgd, 0); | |
8550eb99 | 83 | BUG_ON(pmd_table != pmd_offset(pud, 0)); |
a376f30a Z |
84 | |
85 | return pmd_table; | |
b239fb25 JF |
86 | } |
87 | #endif | |
1da177e4 LT |
88 | pud = pud_offset(pgd, 0); |
89 | pmd_table = pmd_offset(pud, 0); | |
8550eb99 | 90 | |
1da177e4 LT |
91 | return pmd_table; |
92 | } | |
93 | ||
94 | /* | |
95 | * Create a page table and place a pointer to it in a middle page | |
8550eb99 | 96 | * directory entry: |
1da177e4 LT |
97 | */ |
98 | static pte_t * __init one_page_table_init(pmd_t *pmd) | |
99 | { | |
b239fb25 | 100 | if (!(pmd_val(*pmd) & _PAGE_PRESENT)) { |
509a80c4 IM |
101 | pte_t *page_table = NULL; |
102 | ||
c464573c | 103 | if (after_bootmem) { |
f8561296 | 104 | #if defined(CONFIG_DEBUG_PAGEALLOC) || defined(CONFIG_KMEMCHECK) |
4e29684c | 105 | page_table = (pte_t *) alloc_bootmem_pages(PAGE_SIZE); |
509a80c4 | 106 | #endif |
4e29684c YL |
107 | if (!page_table) |
108 | page_table = | |
3c1596ef | 109 | (pte_t *)alloc_bootmem_pages(PAGE_SIZE); |
d6be89ad JB |
110 | } else |
111 | page_table = (pte_t *)alloc_low_page(); | |
b239fb25 | 112 | |
6944a9c8 | 113 | paravirt_alloc_pte(&init_mm, __pa(page_table) >> PAGE_SHIFT); |
1da177e4 | 114 | set_pmd(pmd, __pmd(__pa(page_table) | _PAGE_TABLE)); |
b239fb25 | 115 | BUG_ON(page_table != pte_offset_kernel(pmd, 0)); |
1da177e4 | 116 | } |
509a80c4 | 117 | |
1da177e4 LT |
118 | return pte_offset_kernel(pmd, 0); |
119 | } | |
120 | ||
458a3e64 | 121 | pmd_t * __init populate_extra_pmd(unsigned long vaddr) |
11124411 TH |
122 | { |
123 | int pgd_idx = pgd_index(vaddr); | |
124 | int pmd_idx = pmd_index(vaddr); | |
458a3e64 TH |
125 | |
126 | return one_md_table_init(swapper_pg_dir + pgd_idx) + pmd_idx; | |
127 | } | |
128 | ||
129 | pte_t * __init populate_extra_pte(unsigned long vaddr) | |
130 | { | |
131 | int pte_idx = pte_index(vaddr); | |
11124411 TH |
132 | pmd_t *pmd; |
133 | ||
458a3e64 TH |
134 | pmd = populate_extra_pmd(vaddr); |
135 | return one_page_table_init(pmd) + pte_idx; | |
11124411 TH |
136 | } |
137 | ||
719272c4 YL |
138 | static unsigned long __init |
139 | page_table_range_init_count(unsigned long start, unsigned long end) | |
140 | { | |
141 | unsigned long count = 0; | |
142 | #ifdef CONFIG_HIGHMEM | |
143 | int pmd_idx_kmap_begin = fix_to_virt(FIX_KMAP_END) >> PMD_SHIFT; | |
144 | int pmd_idx_kmap_end = fix_to_virt(FIX_KMAP_BEGIN) >> PMD_SHIFT; | |
145 | int pgd_idx, pmd_idx; | |
146 | unsigned long vaddr; | |
147 | ||
148 | if (pmd_idx_kmap_begin == pmd_idx_kmap_end) | |
149 | return 0; | |
150 | ||
151 | vaddr = start; | |
152 | pgd_idx = pgd_index(vaddr); | |
153 | ||
154 | for ( ; (pgd_idx < PTRS_PER_PGD) && (vaddr != end); pgd_idx++) { | |
155 | for (; (pmd_idx < PTRS_PER_PMD) && (vaddr != end); | |
156 | pmd_idx++) { | |
157 | if ((vaddr >> PMD_SHIFT) >= pmd_idx_kmap_begin && | |
158 | (vaddr >> PMD_SHIFT) <= pmd_idx_kmap_end) | |
159 | count++; | |
160 | vaddr += PMD_SIZE; | |
161 | } | |
162 | pmd_idx = 0; | |
163 | } | |
164 | #endif | |
165 | return count; | |
166 | } | |
167 | ||
a3c6018e | 168 | static pte_t *__init page_table_kmap_check(pte_t *pte, pmd_t *pmd, |
719272c4 YL |
169 | unsigned long vaddr, pte_t *lastpte, |
170 | void **adr) | |
a3c6018e JB |
171 | { |
172 | #ifdef CONFIG_HIGHMEM | |
173 | /* | |
174 | * Something (early fixmap) may already have put a pte | |
175 | * page here, which causes the page table allocation | |
176 | * to become nonlinear. Attempt to fix it, and if it | |
177 | * is still nonlinear then we have to bug. | |
178 | */ | |
179 | int pmd_idx_kmap_begin = fix_to_virt(FIX_KMAP_END) >> PMD_SHIFT; | |
180 | int pmd_idx_kmap_end = fix_to_virt(FIX_KMAP_BEGIN) >> PMD_SHIFT; | |
181 | ||
182 | if (pmd_idx_kmap_begin != pmd_idx_kmap_end | |
183 | && (vaddr >> PMD_SHIFT) >= pmd_idx_kmap_begin | |
719272c4 | 184 | && (vaddr >> PMD_SHIFT) <= pmd_idx_kmap_end) { |
a3c6018e JB |
185 | pte_t *newpte; |
186 | int i; | |
187 | ||
c464573c | 188 | BUG_ON(after_bootmem); |
719272c4 | 189 | newpte = *adr; |
a3c6018e JB |
190 | for (i = 0; i < PTRS_PER_PTE; i++) |
191 | set_pte(newpte + i, pte[i]); | |
719272c4 | 192 | *adr = (void *)(((unsigned long)(*adr)) + PAGE_SIZE); |
a3c6018e JB |
193 | |
194 | paravirt_alloc_pte(&init_mm, __pa(newpte) >> PAGE_SHIFT); | |
195 | set_pmd(pmd, __pmd(__pa(newpte)|_PAGE_TABLE)); | |
196 | BUG_ON(newpte != pte_offset_kernel(pmd, 0)); | |
197 | __flush_tlb_all(); | |
198 | ||
199 | paravirt_release_pte(__pa(pte) >> PAGE_SHIFT); | |
200 | pte = newpte; | |
201 | } | |
202 | BUG_ON(vaddr < fix_to_virt(FIX_KMAP_BEGIN - 1) | |
203 | && vaddr > fix_to_virt(FIX_KMAP_END) | |
204 | && lastpte && lastpte + PTRS_PER_PTE != pte); | |
205 | #endif | |
206 | return pte; | |
207 | } | |
208 | ||
1da177e4 | 209 | /* |
8550eb99 | 210 | * This function initializes a certain range of kernel virtual memory |
1da177e4 LT |
211 | * with new bootmem page tables, everywhere page tables are missing in |
212 | * the given range. | |
8550eb99 IM |
213 | * |
214 | * NOTE: The pagetables are allocated contiguous on the physical space | |
215 | * so we can cache the place of the first one and move around without | |
1da177e4 LT |
216 | * checking the pgd every time. |
217 | */ | |
8550eb99 IM |
218 | static void __init |
219 | page_table_range_init(unsigned long start, unsigned long end, pgd_t *pgd_base) | |
1da177e4 | 220 | { |
1da177e4 LT |
221 | int pgd_idx, pmd_idx; |
222 | unsigned long vaddr; | |
8550eb99 IM |
223 | pgd_t *pgd; |
224 | pmd_t *pmd; | |
a3c6018e | 225 | pte_t *pte = NULL; |
719272c4 YL |
226 | unsigned long count = page_table_range_init_count(start, end); |
227 | void *adr = NULL; | |
228 | ||
229 | if (count) | |
230 | adr = alloc_low_pages(count); | |
1da177e4 LT |
231 | |
232 | vaddr = start; | |
233 | pgd_idx = pgd_index(vaddr); | |
234 | pmd_idx = pmd_index(vaddr); | |
235 | pgd = pgd_base + pgd_idx; | |
236 | ||
237 | for ( ; (pgd_idx < PTRS_PER_PGD) && (vaddr != end); pgd++, pgd_idx++) { | |
b239fb25 JF |
238 | pmd = one_md_table_init(pgd); |
239 | pmd = pmd + pmd_index(vaddr); | |
8550eb99 IM |
240 | for (; (pmd_idx < PTRS_PER_PMD) && (vaddr != end); |
241 | pmd++, pmd_idx++) { | |
a3c6018e | 242 | pte = page_table_kmap_check(one_page_table_init(pmd), |
719272c4 | 243 | pmd, vaddr, pte, &adr); |
1da177e4 LT |
244 | |
245 | vaddr += PMD_SIZE; | |
246 | } | |
247 | pmd_idx = 0; | |
248 | } | |
249 | } | |
250 | ||
251 | static inline int is_kernel_text(unsigned long addr) | |
252 | { | |
5bd5a452 | 253 | if (addr >= (unsigned long)_text && addr <= (unsigned long)__init_end) |
1da177e4 LT |
254 | return 1; |
255 | return 0; | |
256 | } | |
257 | ||
258 | /* | |
8550eb99 IM |
259 | * This maps the physical memory to kernel virtual address space, a total |
260 | * of max_low_pfn pages, by creating page tables starting from address | |
261 | * PAGE_OFFSET: | |
1da177e4 | 262 | */ |
e53fb04f PE |
263 | unsigned long __init |
264 | kernel_physical_mapping_init(unsigned long start, | |
265 | unsigned long end, | |
266 | unsigned long page_size_mask) | |
1da177e4 | 267 | { |
e53fb04f | 268 | int use_pse = page_size_mask == (1<<PG_LEVEL_2M); |
c1fd1b43 | 269 | unsigned long last_map_addr = end; |
e53fb04f | 270 | unsigned long start_pfn, end_pfn; |
e7179853 | 271 | pgd_t *pgd_base = swapper_pg_dir; |
8550eb99 | 272 | int pgd_idx, pmd_idx, pte_ofs; |
1da177e4 LT |
273 | unsigned long pfn; |
274 | pgd_t *pgd; | |
275 | pmd_t *pmd; | |
276 | pte_t *pte; | |
a2699e47 SS |
277 | unsigned pages_2m, pages_4k; |
278 | int mapping_iter; | |
279 | ||
e53fb04f PE |
280 | start_pfn = start >> PAGE_SHIFT; |
281 | end_pfn = end >> PAGE_SHIFT; | |
282 | ||
a2699e47 SS |
283 | /* |
284 | * First iteration will setup identity mapping using large/small pages | |
285 | * based on use_pse, with other attributes same as set by | |
286 | * the early code in head_32.S | |
287 | * | |
288 | * Second iteration will setup the appropriate attributes (NX, GLOBAL..) | |
289 | * as desired for the kernel identity mapping. | |
290 | * | |
291 | * This two pass mechanism conforms to the TLB app note which says: | |
292 | * | |
293 | * "Software should not write to a paging-structure entry in a way | |
294 | * that would change, for any linear address, both the page size | |
295 | * and either the page frame or attributes." | |
296 | */ | |
297 | mapping_iter = 1; | |
1da177e4 | 298 | |
a04ad82d YL |
299 | if (!cpu_has_pse) |
300 | use_pse = 0; | |
1da177e4 | 301 | |
a2699e47 SS |
302 | repeat: |
303 | pages_2m = pages_4k = 0; | |
a04ad82d YL |
304 | pfn = start_pfn; |
305 | pgd_idx = pgd_index((pfn<<PAGE_SHIFT) + PAGE_OFFSET); | |
306 | pgd = pgd_base + pgd_idx; | |
1da177e4 LT |
307 | for (; pgd_idx < PTRS_PER_PGD; pgd++, pgd_idx++) { |
308 | pmd = one_md_table_init(pgd); | |
8550eb99 | 309 | |
a04ad82d YL |
310 | if (pfn >= end_pfn) |
311 | continue; | |
312 | #ifdef CONFIG_X86_PAE | |
313 | pmd_idx = pmd_index((pfn<<PAGE_SHIFT) + PAGE_OFFSET); | |
314 | pmd += pmd_idx; | |
315 | #else | |
316 | pmd_idx = 0; | |
317 | #endif | |
318 | for (; pmd_idx < PTRS_PER_PMD && pfn < end_pfn; | |
f3f20de8 | 319 | pmd++, pmd_idx++) { |
8550eb99 | 320 | unsigned int addr = pfn * PAGE_SIZE + PAGE_OFFSET; |
1da177e4 | 321 | |
8550eb99 IM |
322 | /* |
323 | * Map with big pages if possible, otherwise | |
324 | * create normal page tables: | |
325 | */ | |
a04ad82d | 326 | if (use_pse) { |
8550eb99 | 327 | unsigned int addr2; |
f3f20de8 | 328 | pgprot_t prot = PAGE_KERNEL_LARGE; |
a2699e47 SS |
329 | /* |
330 | * first pass will use the same initial | |
331 | * identity mapping attribute + _PAGE_PSE. | |
332 | */ | |
333 | pgprot_t init_prot = | |
334 | __pgprot(PTE_IDENT_ATTR | | |
335 | _PAGE_PSE); | |
f3f20de8 | 336 | |
960ddb4f | 337 | pfn &= PMD_MASK >> PAGE_SHIFT; |
8550eb99 | 338 | addr2 = (pfn + PTRS_PER_PTE-1) * PAGE_SIZE + |
f3f20de8 JF |
339 | PAGE_OFFSET + PAGE_SIZE-1; |
340 | ||
8550eb99 IM |
341 | if (is_kernel_text(addr) || |
342 | is_kernel_text(addr2)) | |
f3f20de8 JF |
343 | prot = PAGE_KERNEL_LARGE_EXEC; |
344 | ||
ce0c0e50 | 345 | pages_2m++; |
a2699e47 SS |
346 | if (mapping_iter == 1) |
347 | set_pmd(pmd, pfn_pmd(pfn, init_prot)); | |
348 | else | |
349 | set_pmd(pmd, pfn_pmd(pfn, prot)); | |
b239fb25 | 350 | |
1da177e4 | 351 | pfn += PTRS_PER_PTE; |
8550eb99 IM |
352 | continue; |
353 | } | |
354 | pte = one_page_table_init(pmd); | |
1da177e4 | 355 | |
a04ad82d YL |
356 | pte_ofs = pte_index((pfn<<PAGE_SHIFT) + PAGE_OFFSET); |
357 | pte += pte_ofs; | |
358 | for (; pte_ofs < PTRS_PER_PTE && pfn < end_pfn; | |
8550eb99 IM |
359 | pte++, pfn++, pte_ofs++, addr += PAGE_SIZE) { |
360 | pgprot_t prot = PAGE_KERNEL; | |
a2699e47 SS |
361 | /* |
362 | * first pass will use the same initial | |
363 | * identity mapping attribute. | |
364 | */ | |
365 | pgprot_t init_prot = __pgprot(PTE_IDENT_ATTR); | |
f3f20de8 | 366 | |
8550eb99 IM |
367 | if (is_kernel_text(addr)) |
368 | prot = PAGE_KERNEL_EXEC; | |
f3f20de8 | 369 | |
ce0c0e50 | 370 | pages_4k++; |
c1fd1b43 | 371 | if (mapping_iter == 1) { |
a2699e47 | 372 | set_pte(pte, pfn_pte(pfn, init_prot)); |
c1fd1b43 PE |
373 | last_map_addr = (pfn << PAGE_SHIFT) + PAGE_SIZE; |
374 | } else | |
a2699e47 | 375 | set_pte(pte, pfn_pte(pfn, prot)); |
1da177e4 LT |
376 | } |
377 | } | |
378 | } | |
a2699e47 SS |
379 | if (mapping_iter == 1) { |
380 | /* | |
381 | * update direct mapping page count only in the first | |
382 | * iteration. | |
383 | */ | |
384 | update_page_count(PG_LEVEL_2M, pages_2m); | |
385 | update_page_count(PG_LEVEL_4K, pages_4k); | |
386 | ||
387 | /* | |
388 | * local global flush tlb, which will flush the previous | |
389 | * mappings present in both small and large page TLB's. | |
390 | */ | |
391 | __flush_tlb_all(); | |
392 | ||
393 | /* | |
394 | * Second iteration will set the actual desired PTE attributes. | |
395 | */ | |
396 | mapping_iter = 2; | |
397 | goto repeat; | |
398 | } | |
c1fd1b43 | 399 | return last_map_addr; |
ae531c26 AV |
400 | } |
401 | ||
1da177e4 LT |
402 | pte_t *kmap_pte; |
403 | pgprot_t kmap_prot; | |
404 | ||
8550eb99 IM |
405 | static inline pte_t *kmap_get_fixmap_pte(unsigned long vaddr) |
406 | { | |
407 | return pte_offset_kernel(pmd_offset(pud_offset(pgd_offset_k(vaddr), | |
408 | vaddr), vaddr), vaddr); | |
409 | } | |
1da177e4 LT |
410 | |
411 | static void __init kmap_init(void) | |
412 | { | |
413 | unsigned long kmap_vstart; | |
414 | ||
8550eb99 IM |
415 | /* |
416 | * Cache the first kmap pte: | |
417 | */ | |
1da177e4 LT |
418 | kmap_vstart = __fix_to_virt(FIX_KMAP_BEGIN); |
419 | kmap_pte = kmap_get_fixmap_pte(kmap_vstart); | |
420 | ||
421 | kmap_prot = PAGE_KERNEL; | |
422 | } | |
423 | ||
fd940934 | 424 | #ifdef CONFIG_HIGHMEM |
1da177e4 LT |
425 | static void __init permanent_kmaps_init(pgd_t *pgd_base) |
426 | { | |
8550eb99 | 427 | unsigned long vaddr; |
1da177e4 LT |
428 | pgd_t *pgd; |
429 | pud_t *pud; | |
430 | pmd_t *pmd; | |
431 | pte_t *pte; | |
1da177e4 LT |
432 | |
433 | vaddr = PKMAP_BASE; | |
434 | page_table_range_init(vaddr, vaddr + PAGE_SIZE*LAST_PKMAP, pgd_base); | |
435 | ||
436 | pgd = swapper_pg_dir + pgd_index(vaddr); | |
437 | pud = pud_offset(pgd, vaddr); | |
438 | pmd = pmd_offset(pud, vaddr); | |
439 | pte = pte_offset_kernel(pmd, vaddr); | |
8550eb99 | 440 | pkmap_page_table = pte; |
1da177e4 LT |
441 | } |
442 | ||
b1258ac2 | 443 | static void __init add_one_highpage_init(struct page *page) |
1da177e4 | 444 | { |
cc9f7a0c YL |
445 | ClearPageReserved(page); |
446 | init_page_count(page); | |
447 | __free_page(page); | |
448 | totalhigh_pages++; | |
1da177e4 LT |
449 | } |
450 | ||
1d931264 YL |
451 | void __init add_highpages_with_active_regions(int nid, |
452 | unsigned long start_pfn, unsigned long end_pfn) | |
1da177e4 | 453 | { |
8a9ca34c TH |
454 | phys_addr_t start, end; |
455 | u64 i; | |
456 | ||
457 | for_each_free_mem_range(i, nid, &start, &end, NULL) { | |
458 | unsigned long pfn = clamp_t(unsigned long, PFN_UP(start), | |
459 | start_pfn, end_pfn); | |
460 | unsigned long e_pfn = clamp_t(unsigned long, PFN_DOWN(end), | |
461 | start_pfn, end_pfn); | |
462 | for ( ; pfn < e_pfn; pfn++) | |
463 | if (pfn_valid(pfn)) | |
464 | add_one_highpage_init(pfn_to_page(pfn)); | |
23be8c7d | 465 | } |
b5bc6c0e | 466 | } |
1da177e4 | 467 | #else |
e8e32326 IB |
468 | static inline void permanent_kmaps_init(pgd_t *pgd_base) |
469 | { | |
470 | } | |
1da177e4 LT |
471 | #endif /* CONFIG_HIGHMEM */ |
472 | ||
7737b215 | 473 | void __init native_pagetable_init(void) |
1da177e4 | 474 | { |
551889a6 | 475 | unsigned long pfn, va; |
73090f89 | 476 | pgd_t *pgd, *base = swapper_pg_dir; |
551889a6 IC |
477 | pud_t *pud; |
478 | pmd_t *pmd; | |
479 | pte_t *pte; | |
b239fb25 JF |
480 | |
481 | /* | |
551889a6 | 482 | * Remove any mappings which extend past the end of physical |
11ed9e92 YL |
483 | * memory from the boot time page table. |
484 | * In virtual address space, we should have at least two pages | |
485 | * from VMALLOC_END to pkmap or fixmap according to VMALLOC_END | |
486 | * definition. And max_low_pfn is set to VMALLOC_END physical | |
487 | * address. If initial memory mapping is doing right job, we | |
488 | * should have pte used near max_low_pfn or one pmd is not present. | |
b239fb25 | 489 | */ |
11ed9e92 | 490 | for (pfn = max_low_pfn; pfn < 1<<(32-PAGE_SHIFT); pfn++) { |
551889a6 IC |
491 | va = PAGE_OFFSET + (pfn<<PAGE_SHIFT); |
492 | pgd = base + pgd_index(va); | |
493 | if (!pgd_present(*pgd)) | |
494 | break; | |
495 | ||
496 | pud = pud_offset(pgd, va); | |
497 | pmd = pmd_offset(pud, va); | |
498 | if (!pmd_present(*pmd)) | |
499 | break; | |
500 | ||
11ed9e92 YL |
501 | /* should not be large page here */ |
502 | if (pmd_large(*pmd)) { | |
503 | 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", | |
504 | pfn, pmd, __pa(pmd)); | |
505 | BUG_ON(1); | |
506 | } | |
507 | ||
551889a6 IC |
508 | pte = pte_offset_kernel(pmd, va); |
509 | if (!pte_present(*pte)) | |
510 | break; | |
511 | ||
11ed9e92 YL |
512 | printk(KERN_DEBUG "clearing pte for ram above max_low_pfn: pfn: %lx pmd: %p pmd phys: %lx pte: %p pte phys: %lx\n", |
513 | pfn, pmd, __pa(pmd), pte, __pa(pte)); | |
551889a6 IC |
514 | pte_clear(NULL, va, pte); |
515 | } | |
6944a9c8 | 516 | paravirt_alloc_pmd(&init_mm, __pa(base) >> PAGE_SHIFT); |
843b8ed2 | 517 | paging_init(); |
b239fb25 JF |
518 | } |
519 | ||
b239fb25 JF |
520 | /* |
521 | * Build a proper pagetable for the kernel mappings. Up until this | |
522 | * point, we've been running on some set of pagetables constructed by | |
523 | * the boot process. | |
524 | * | |
525 | * If we're booting on native hardware, this will be a pagetable | |
551889a6 IC |
526 | * constructed in arch/x86/kernel/head_32.S. The root of the |
527 | * pagetable will be swapper_pg_dir. | |
b239fb25 JF |
528 | * |
529 | * If we're booting paravirtualized under a hypervisor, then there are | |
530 | * more options: we may already be running PAE, and the pagetable may | |
531 | * or may not be based in swapper_pg_dir. In any case, | |
7737b215 | 532 | * paravirt_pagetable_init() will set up swapper_pg_dir |
b239fb25 JF |
533 | * appropriately for the rest of the initialization to work. |
534 | * | |
535 | * In general, pagetable_init() assumes that the pagetable may already | |
536 | * be partially populated, and so it avoids stomping on any existing | |
537 | * mappings. | |
538 | */ | |
f765090a | 539 | void __init early_ioremap_page_table_range_init(void) |
b239fb25 | 540 | { |
e7179853 | 541 | pgd_t *pgd_base = swapper_pg_dir; |
8550eb99 | 542 | unsigned long vaddr, end; |
b239fb25 | 543 | |
1da177e4 LT |
544 | /* |
545 | * Fixed mappings, only the page table structure has to be | |
546 | * created - mappings will be set by set_fixmap(): | |
547 | */ | |
548 | vaddr = __fix_to_virt(__end_of_fixed_addresses - 1) & PMD_MASK; | |
b239fb25 JF |
549 | end = (FIXADDR_TOP + PMD_SIZE - 1) & PMD_MASK; |
550 | page_table_range_init(vaddr, end, pgd_base); | |
beacfaac | 551 | early_ioremap_reset(); |
e7b37895 YL |
552 | } |
553 | ||
554 | static void __init pagetable_init(void) | |
555 | { | |
556 | pgd_t *pgd_base = swapper_pg_dir; | |
557 | ||
1da177e4 | 558 | permanent_kmaps_init(pgd_base); |
1da177e4 LT |
559 | } |
560 | ||
be43d728 | 561 | pteval_t __supported_pte_mask __read_mostly = ~(_PAGE_NX | _PAGE_GLOBAL | _PAGE_IOMAP); |
6fdc05d4 JF |
562 | EXPORT_SYMBOL_GPL(__supported_pte_mask); |
563 | ||
90d967e0 YL |
564 | /* user-defined highmem size */ |
565 | static unsigned int highmem_pages = -1; | |
566 | ||
567 | /* | |
568 | * highmem=size forces highmem to be exactly 'size' bytes. | |
569 | * This works even on boxes that have no highmem otherwise. | |
570 | * This also works to reduce highmem size on bigger boxes. | |
571 | */ | |
572 | static int __init parse_highmem(char *arg) | |
573 | { | |
574 | if (!arg) | |
575 | return -EINVAL; | |
576 | ||
577 | highmem_pages = memparse(arg, &arg) >> PAGE_SHIFT; | |
578 | return 0; | |
579 | } | |
580 | early_param("highmem", parse_highmem); | |
581 | ||
4769843b IM |
582 | #define MSG_HIGHMEM_TOO_BIG \ |
583 | "highmem size (%luMB) is bigger than pages available (%luMB)!\n" | |
584 | ||
585 | #define MSG_LOWMEM_TOO_SMALL \ | |
586 | "highmem size (%luMB) results in <64MB lowmem, ignoring it!\n" | |
90d967e0 | 587 | /* |
4769843b IM |
588 | * All of RAM fits into lowmem - but if user wants highmem |
589 | * artificially via the highmem=x boot parameter then create | |
590 | * it: | |
90d967e0 | 591 | */ |
f836e35a | 592 | static void __init lowmem_pfn_init(void) |
90d967e0 | 593 | { |
346cafec | 594 | /* max_low_pfn is 0, we already have early_res support */ |
90d967e0 | 595 | max_low_pfn = max_pfn; |
d88316c2 | 596 | |
4769843b IM |
597 | if (highmem_pages == -1) |
598 | highmem_pages = 0; | |
599 | #ifdef CONFIG_HIGHMEM | |
600 | if (highmem_pages >= max_pfn) { | |
601 | printk(KERN_ERR MSG_HIGHMEM_TOO_BIG, | |
602 | pages_to_mb(highmem_pages), pages_to_mb(max_pfn)); | |
603 | highmem_pages = 0; | |
604 | } | |
605 | if (highmem_pages) { | |
606 | if (max_low_pfn - highmem_pages < 64*1024*1024/PAGE_SIZE) { | |
607 | printk(KERN_ERR MSG_LOWMEM_TOO_SMALL, | |
90d967e0 YL |
608 | pages_to_mb(highmem_pages)); |
609 | highmem_pages = 0; | |
610 | } | |
4769843b IM |
611 | max_low_pfn -= highmem_pages; |
612 | } | |
613 | #else | |
614 | if (highmem_pages) | |
615 | printk(KERN_ERR "ignoring highmem size on non-highmem kernel!\n"); | |
616 | #endif | |
617 | } | |
618 | ||
619 | #define MSG_HIGHMEM_TOO_SMALL \ | |
620 | "only %luMB highmem pages available, ignoring highmem size of %luMB!\n" | |
621 | ||
622 | #define MSG_HIGHMEM_TRIMMED \ | |
623 | "Warning: only 4GB will be used. Use a HIGHMEM64G enabled kernel!\n" | |
624 | /* | |
625 | * We have more RAM than fits into lowmem - we try to put it into | |
626 | * highmem, also taking the highmem=x boot parameter into account: | |
627 | */ | |
f836e35a | 628 | static void __init highmem_pfn_init(void) |
4769843b | 629 | { |
d88316c2 IM |
630 | max_low_pfn = MAXMEM_PFN; |
631 | ||
4769843b IM |
632 | if (highmem_pages == -1) |
633 | highmem_pages = max_pfn - MAXMEM_PFN; | |
634 | ||
635 | if (highmem_pages + MAXMEM_PFN < max_pfn) | |
636 | max_pfn = MAXMEM_PFN + highmem_pages; | |
637 | ||
638 | if (highmem_pages + MAXMEM_PFN > max_pfn) { | |
639 | printk(KERN_WARNING MSG_HIGHMEM_TOO_SMALL, | |
640 | pages_to_mb(max_pfn - MAXMEM_PFN), | |
641 | pages_to_mb(highmem_pages)); | |
642 | highmem_pages = 0; | |
643 | } | |
90d967e0 | 644 | #ifndef CONFIG_HIGHMEM |
4769843b IM |
645 | /* Maximum memory usable is what is directly addressable */ |
646 | printk(KERN_WARNING "Warning only %ldMB will be used.\n", MAXMEM>>20); | |
647 | if (max_pfn > MAX_NONPAE_PFN) | |
648 | printk(KERN_WARNING "Use a HIGHMEM64G enabled kernel.\n"); | |
649 | else | |
650 | printk(KERN_WARNING "Use a HIGHMEM enabled kernel.\n"); | |
651 | max_pfn = MAXMEM_PFN; | |
90d967e0 YL |
652 | #else /* !CONFIG_HIGHMEM */ |
653 | #ifndef CONFIG_HIGHMEM64G | |
4769843b IM |
654 | if (max_pfn > MAX_NONPAE_PFN) { |
655 | max_pfn = MAX_NONPAE_PFN; | |
656 | printk(KERN_WARNING MSG_HIGHMEM_TRIMMED); | |
657 | } | |
90d967e0 YL |
658 | #endif /* !CONFIG_HIGHMEM64G */ |
659 | #endif /* !CONFIG_HIGHMEM */ | |
4769843b IM |
660 | } |
661 | ||
662 | /* | |
663 | * Determine low and high memory ranges: | |
664 | */ | |
665 | void __init find_low_pfn_range(void) | |
666 | { | |
667 | /* it could update max_pfn */ | |
668 | ||
d88316c2 | 669 | if (max_pfn <= MAXMEM_PFN) |
4769843b | 670 | lowmem_pfn_init(); |
d88316c2 IM |
671 | else |
672 | highmem_pfn_init(); | |
90d967e0 YL |
673 | } |
674 | ||
b2ac82a0 | 675 | #ifndef CONFIG_NEED_MULTIPLE_NODES |
d8fc3afc | 676 | void __init initmem_init(void) |
b2ac82a0 | 677 | { |
b2ac82a0 YL |
678 | #ifdef CONFIG_HIGHMEM |
679 | highstart_pfn = highend_pfn = max_pfn; | |
680 | if (max_pfn > max_low_pfn) | |
681 | highstart_pfn = max_low_pfn; | |
b2ac82a0 YL |
682 | printk(KERN_NOTICE "%ldMB HIGHMEM available.\n", |
683 | pages_to_mb(highend_pfn - highstart_pfn)); | |
684 | num_physpages = highend_pfn; | |
685 | high_memory = (void *) __va(highstart_pfn * PAGE_SIZE - 1) + 1; | |
686 | #else | |
b2ac82a0 YL |
687 | num_physpages = max_low_pfn; |
688 | high_memory = (void *) __va(max_low_pfn * PAGE_SIZE - 1) + 1; | |
689 | #endif | |
0608f70c TH |
690 | |
691 | memblock_set_node(0, (phys_addr_t)ULLONG_MAX, 0); | |
692 | sparse_memory_present_with_active_regions(0); | |
693 | ||
b2ac82a0 YL |
694 | #ifdef CONFIG_FLATMEM |
695 | max_mapnr = num_physpages; | |
696 | #endif | |
dc16ecf7 JF |
697 | __vmalloc_start_set = true; |
698 | ||
b2ac82a0 YL |
699 | printk(KERN_NOTICE "%ldMB LOWMEM available.\n", |
700 | pages_to_mb(max_low_pfn)); | |
701 | ||
702 | setup_bootmem_allocator(); | |
b2ac82a0 | 703 | } |
cb95a13a | 704 | #endif /* !CONFIG_NEED_MULTIPLE_NODES */ |
b2ac82a0 | 705 | |
b2ac82a0 YL |
706 | void __init setup_bootmem_allocator(void) |
707 | { | |
b2ac82a0 YL |
708 | printk(KERN_INFO " mapped low ram: 0 - %08lx\n", |
709 | max_pfn_mapped<<PAGE_SHIFT); | |
fc5efe39 | 710 | printk(KERN_INFO " low ram: 0 - %08lx\n", max_low_pfn<<PAGE_SHIFT); |
7482b0e9 | 711 | |
c464573c | 712 | after_bootmem = 1; |
4e29684c YL |
713 | } |
714 | ||
1da177e4 LT |
715 | /* |
716 | * paging_init() sets up the page tables - note that the first 8MB are | |
717 | * already mapped by head.S. | |
718 | * | |
719 | * This routines also unmaps the page at virtual kernel address 0, so | |
720 | * that we can trap those pesky NULL-reference errors in the kernel. | |
721 | */ | |
722 | void __init paging_init(void) | |
723 | { | |
1da177e4 LT |
724 | pagetable_init(); |
725 | ||
1da177e4 LT |
726 | __flush_tlb_all(); |
727 | ||
728 | kmap_init(); | |
11cd0bc1 YL |
729 | |
730 | /* | |
731 | * NOTE: at this point the bootmem allocator is fully available. | |
732 | */ | |
c10d1e26 | 733 | olpc_dt_build_devicetree(); |
797390d8 | 734 | sparse_memory_present_with_active_regions(MAX_NUMNODES); |
11cd0bc1 YL |
735 | sparse_init(); |
736 | zone_sizes_init(); | |
1da177e4 LT |
737 | } |
738 | ||
739 | /* | |
740 | * Test if the WP bit works in supervisor mode. It isn't supported on 386's | |
f7f17a67 DV |
741 | * and also on some strange 486's. All 586+'s are OK. This used to involve |
742 | * black magic jumps to work around some nasty CPU bugs, but fortunately the | |
743 | * switch to using exceptions got rid of all that. | |
1da177e4 | 744 | */ |
1da177e4 LT |
745 | static void __init test_wp_bit(void) |
746 | { | |
d7d119d7 IM |
747 | printk(KERN_INFO |
748 | "Checking if this processor honours the WP bit even in supervisor mode..."); | |
1da177e4 LT |
749 | |
750 | /* Any page-aligned address will do, the test is non-destructive */ | |
8bd753be | 751 | __set_fixmap(FIX_WP_TEST, __pa(&swapper_pg_dir), PAGE_KERNEL_RO); |
1da177e4 LT |
752 | boot_cpu_data.wp_works_ok = do_test_wp_bit(); |
753 | clear_fixmap(FIX_WP_TEST); | |
754 | ||
755 | if (!boot_cpu_data.wp_works_ok) { | |
d7d119d7 | 756 | printk(KERN_CONT "No.\n"); |
1da177e4 | 757 | #ifdef CONFIG_X86_WP_WORKS_OK |
d7d119d7 IM |
758 | panic( |
759 | "This kernel doesn't support CPU's with broken WP. Recompile it for a 386!"); | |
1da177e4 LT |
760 | #endif |
761 | } else { | |
d7d119d7 | 762 | printk(KERN_CONT "Ok.\n"); |
1da177e4 LT |
763 | } |
764 | } | |
765 | ||
1da177e4 LT |
766 | void __init mem_init(void) |
767 | { | |
1da177e4 | 768 | int codesize, reservedpages, datasize, initsize; |
cc9f7a0c | 769 | int tmp; |
1da177e4 | 770 | |
cfb80c9e JF |
771 | pci_iommu_alloc(); |
772 | ||
05b79bdc | 773 | #ifdef CONFIG_FLATMEM |
8d8f3cbe | 774 | BUG_ON(!mem_map); |
1da177e4 | 775 | #endif |
855c743a SG |
776 | /* |
777 | * With CONFIG_DEBUG_PAGEALLOC initialization of highmem pages has to | |
778 | * be done before free_all_bootmem(). Memblock use free low memory for | |
779 | * temporary data (see find_range_array()) and for this purpose can use | |
780 | * pages that was already passed to the buddy allocator, hence marked as | |
781 | * not accessible in the page tables when compiled with | |
782 | * CONFIG_DEBUG_PAGEALLOC. Otherwise order of initialization is not | |
783 | * important here. | |
784 | */ | |
785 | set_highmem_pages_init(); | |
786 | ||
1da177e4 LT |
787 | /* this will put all low memory onto the freelists */ |
788 | totalram_pages += free_all_bootmem(); | |
789 | ||
790 | reservedpages = 0; | |
791 | for (tmp = 0; tmp < max_low_pfn; tmp++) | |
792 | /* | |
8550eb99 | 793 | * Only count reserved RAM pages: |
1da177e4 LT |
794 | */ |
795 | if (page_is_ram(tmp) && PageReserved(pfn_to_page(tmp))) | |
796 | reservedpages++; | |
797 | ||
1da177e4 LT |
798 | codesize = (unsigned long) &_etext - (unsigned long) &_text; |
799 | datasize = (unsigned long) &_edata - (unsigned long) &_etext; | |
800 | initsize = (unsigned long) &__init_end - (unsigned long) &__init_begin; | |
801 | ||
8550eb99 IM |
802 | printk(KERN_INFO "Memory: %luk/%luk available (%dk kernel code, " |
803 | "%dk reserved, %dk data, %dk init, %ldk highmem)\n", | |
cc013a88 | 804 | nr_free_pages() << (PAGE_SHIFT-10), |
1da177e4 LT |
805 | num_physpages << (PAGE_SHIFT-10), |
806 | codesize >> 10, | |
807 | reservedpages << (PAGE_SHIFT-10), | |
808 | datasize >> 10, | |
809 | initsize >> 10, | |
4b529401 | 810 | totalhigh_pages << (PAGE_SHIFT-10)); |
1da177e4 | 811 | |
d7d119d7 | 812 | printk(KERN_INFO "virtual kernel memory layout:\n" |
8550eb99 | 813 | " fixmap : 0x%08lx - 0x%08lx (%4ld kB)\n" |
052e7994 | 814 | #ifdef CONFIG_HIGHMEM |
8550eb99 | 815 | " pkmap : 0x%08lx - 0x%08lx (%4ld kB)\n" |
052e7994 | 816 | #endif |
8550eb99 IM |
817 | " vmalloc : 0x%08lx - 0x%08lx (%4ld MB)\n" |
818 | " lowmem : 0x%08lx - 0x%08lx (%4ld MB)\n" | |
819 | " .init : 0x%08lx - 0x%08lx (%4ld kB)\n" | |
820 | " .data : 0x%08lx - 0x%08lx (%4ld kB)\n" | |
821 | " .text : 0x%08lx - 0x%08lx (%4ld kB)\n", | |
822 | FIXADDR_START, FIXADDR_TOP, | |
823 | (FIXADDR_TOP - FIXADDR_START) >> 10, | |
052e7994 JF |
824 | |
825 | #ifdef CONFIG_HIGHMEM | |
8550eb99 IM |
826 | PKMAP_BASE, PKMAP_BASE+LAST_PKMAP*PAGE_SIZE, |
827 | (LAST_PKMAP*PAGE_SIZE) >> 10, | |
052e7994 JF |
828 | #endif |
829 | ||
8550eb99 IM |
830 | VMALLOC_START, VMALLOC_END, |
831 | (VMALLOC_END - VMALLOC_START) >> 20, | |
052e7994 | 832 | |
8550eb99 IM |
833 | (unsigned long)__va(0), (unsigned long)high_memory, |
834 | ((unsigned long)high_memory - (unsigned long)__va(0)) >> 20, | |
052e7994 | 835 | |
8550eb99 IM |
836 | (unsigned long)&__init_begin, (unsigned long)&__init_end, |
837 | ((unsigned long)&__init_end - | |
838 | (unsigned long)&__init_begin) >> 10, | |
052e7994 | 839 | |
8550eb99 IM |
840 | (unsigned long)&_etext, (unsigned long)&_edata, |
841 | ((unsigned long)&_edata - (unsigned long)&_etext) >> 10, | |
052e7994 | 842 | |
8550eb99 IM |
843 | (unsigned long)&_text, (unsigned long)&_etext, |
844 | ((unsigned long)&_etext - (unsigned long)&_text) >> 10); | |
052e7994 | 845 | |
beeb4195 JB |
846 | /* |
847 | * Check boundaries twice: Some fundamental inconsistencies can | |
848 | * be detected at build time already. | |
849 | */ | |
850 | #define __FIXADDR_TOP (-PAGE_SIZE) | |
851 | #ifdef CONFIG_HIGHMEM | |
852 | BUILD_BUG_ON(PKMAP_BASE + LAST_PKMAP*PAGE_SIZE > FIXADDR_START); | |
853 | BUILD_BUG_ON(VMALLOC_END > PKMAP_BASE); | |
854 | #endif | |
855 | #define high_memory (-128UL << 20) | |
856 | BUILD_BUG_ON(VMALLOC_START >= VMALLOC_END); | |
857 | #undef high_memory | |
858 | #undef __FIXADDR_TOP | |
859 | ||
052e7994 | 860 | #ifdef CONFIG_HIGHMEM |
8550eb99 IM |
861 | BUG_ON(PKMAP_BASE + LAST_PKMAP*PAGE_SIZE > FIXADDR_START); |
862 | BUG_ON(VMALLOC_END > PKMAP_BASE); | |
052e7994 | 863 | #endif |
beeb4195 | 864 | BUG_ON(VMALLOC_START >= VMALLOC_END); |
8550eb99 | 865 | BUG_ON((unsigned long)high_memory > VMALLOC_START); |
052e7994 | 866 | |
1da177e4 LT |
867 | if (boot_cpu_data.wp_works_ok < 0) |
868 | test_wp_bit(); | |
1da177e4 LT |
869 | } |
870 | ||
ad8f5797 | 871 | #ifdef CONFIG_MEMORY_HOTPLUG |
bc02af93 | 872 | int arch_add_memory(int nid, u64 start, u64 size) |
05039b92 | 873 | { |
7c7e9425 | 874 | struct pglist_data *pgdata = NODE_DATA(nid); |
776ed98b | 875 | struct zone *zone = pgdata->node_zones + ZONE_HIGHMEM; |
05039b92 DH |
876 | unsigned long start_pfn = start >> PAGE_SHIFT; |
877 | unsigned long nr_pages = size >> PAGE_SHIFT; | |
878 | ||
c04fc586 | 879 | return __add_pages(nid, zone, start_pfn, nr_pages); |
05039b92 | 880 | } |
9d99aaa3 | 881 | #endif |
05039b92 | 882 | |
1da177e4 LT |
883 | /* |
884 | * This function cannot be __init, since exceptions don't work in that | |
885 | * section. Put this after the callers, so that it cannot be inlined. | |
886 | */ | |
8550eb99 | 887 | static noinline int do_test_wp_bit(void) |
1da177e4 LT |
888 | { |
889 | char tmp_reg; | |
890 | int flag; | |
891 | ||
892 | __asm__ __volatile__( | |
8550eb99 IM |
893 | " movb %0, %1 \n" |
894 | "1: movb %1, %0 \n" | |
895 | " xorl %2, %2 \n" | |
1da177e4 | 896 | "2: \n" |
f832ff18 | 897 | _ASM_EXTABLE(1b,2b) |
1da177e4 LT |
898 | :"=m" (*(char *)fix_to_virt(FIX_WP_TEST)), |
899 | "=q" (tmp_reg), | |
900 | "=r" (flag) | |
901 | :"2" (1) | |
902 | :"memory"); | |
8550eb99 | 903 | |
1da177e4 LT |
904 | return flag; |
905 | } | |
906 | ||
63aaf308 | 907 | #ifdef CONFIG_DEBUG_RODATA |
edeed305 AV |
908 | const int rodata_test_data = 0xC3; |
909 | EXPORT_SYMBOL_GPL(rodata_test_data); | |
63aaf308 | 910 | |
502f6604 | 911 | int kernel_set_to_readonly __read_mostly; |
16239630 SR |
912 | |
913 | void set_kernel_text_rw(void) | |
914 | { | |
915 | unsigned long start = PFN_ALIGN(_text); | |
916 | unsigned long size = PFN_ALIGN(_etext) - start; | |
917 | ||
918 | if (!kernel_set_to_readonly) | |
919 | return; | |
920 | ||
921 | pr_debug("Set kernel text: %lx - %lx for read write\n", | |
922 | start, start+size); | |
923 | ||
924 | set_pages_rw(virt_to_page(start), size >> PAGE_SHIFT); | |
925 | } | |
926 | ||
927 | void set_kernel_text_ro(void) | |
928 | { | |
929 | unsigned long start = PFN_ALIGN(_text); | |
930 | unsigned long size = PFN_ALIGN(_etext) - start; | |
931 | ||
932 | if (!kernel_set_to_readonly) | |
933 | return; | |
934 | ||
935 | pr_debug("Set kernel text: %lx - %lx for read only\n", | |
936 | start, start+size); | |
937 | ||
938 | set_pages_ro(virt_to_page(start), size >> PAGE_SHIFT); | |
939 | } | |
940 | ||
5bd5a452 MC |
941 | static void mark_nxdata_nx(void) |
942 | { | |
943 | /* | |
944 | * When this called, init has already been executed and released, | |
0d2eb44f | 945 | * so everything past _etext should be NX. |
5bd5a452 MC |
946 | */ |
947 | unsigned long start = PFN_ALIGN(_etext); | |
948 | /* | |
949 | * This comes from is_kernel_text upper limit. Also HPAGE where used: | |
950 | */ | |
951 | unsigned long size = (((unsigned long)__init_end + HPAGE_SIZE) & HPAGE_MASK) - start; | |
952 | ||
953 | if (__supported_pte_mask & _PAGE_NX) | |
954 | printk(KERN_INFO "NX-protecting the kernel data: %luk\n", size >> 10); | |
955 | set_pages_nx(virt_to_page(start), size >> PAGE_SHIFT); | |
956 | } | |
957 | ||
63aaf308 AV |
958 | void mark_rodata_ro(void) |
959 | { | |
6fb14755 JB |
960 | unsigned long start = PFN_ALIGN(_text); |
961 | unsigned long size = PFN_ALIGN(_etext) - start; | |
63aaf308 | 962 | |
4e4eee0e MD |
963 | set_pages_ro(virt_to_page(start), size >> PAGE_SHIFT); |
964 | printk(KERN_INFO "Write protecting the kernel text: %luk\n", | |
965 | size >> 10); | |
0c42f392 | 966 | |
16239630 SR |
967 | kernel_set_to_readonly = 1; |
968 | ||
0c42f392 | 969 | #ifdef CONFIG_CPA_DEBUG |
4e4eee0e MD |
970 | printk(KERN_INFO "Testing CPA: Reverting %lx-%lx\n", |
971 | start, start+size); | |
972 | set_pages_rw(virt_to_page(start), size>>PAGE_SHIFT); | |
0c42f392 | 973 | |
4e4eee0e MD |
974 | printk(KERN_INFO "Testing CPA: write protecting again\n"); |
975 | set_pages_ro(virt_to_page(start), size>>PAGE_SHIFT); | |
602033ed | 976 | #endif |
8f0f996e | 977 | |
6fb14755 JB |
978 | start += size; |
979 | size = (unsigned long)__end_rodata - start; | |
6d238cc4 | 980 | set_pages_ro(virt_to_page(start), size >> PAGE_SHIFT); |
d7d119d7 IM |
981 | printk(KERN_INFO "Write protecting the kernel read-only data: %luk\n", |
982 | size >> 10); | |
edeed305 | 983 | rodata_test(); |
63aaf308 | 984 | |
0c42f392 | 985 | #ifdef CONFIG_CPA_DEBUG |
d7d119d7 | 986 | printk(KERN_INFO "Testing CPA: undo %lx-%lx\n", start, start + size); |
6d238cc4 | 987 | set_pages_rw(virt_to_page(start), size >> PAGE_SHIFT); |
0c42f392 | 988 | |
d7d119d7 | 989 | printk(KERN_INFO "Testing CPA: write protecting again\n"); |
6d238cc4 | 990 | set_pages_ro(virt_to_page(start), size >> PAGE_SHIFT); |
0c42f392 | 991 | #endif |
5bd5a452 | 992 | mark_nxdata_nx(); |
63aaf308 AV |
993 | } |
994 | #endif | |
995 |