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/signal.h> |
9 | #include <linux/sched.h> | |
10 | #include <linux/kernel.h> | |
11 | #include <linux/errno.h> | |
12 | #include <linux/string.h> | |
13 | #include <linux/types.h> | |
14 | #include <linux/ptrace.h> | |
15 | #include <linux/mman.h> | |
16 | #include <linux/mm.h> | |
17 | #include <linux/hugetlb.h> | |
18 | #include <linux/swap.h> | |
19 | #include <linux/smp.h> | |
20 | #include <linux/init.h> | |
21 | #include <linux/highmem.h> | |
22 | #include <linux/pagemap.h> | |
cfb80c9e | 23 | #include <linux/pci.h> |
6fb14755 | 24 | #include <linux/pfn.h> |
c9cf5528 | 25 | #include <linux/poison.h> |
a9ce6bc1 | 26 | #include <linux/memblock.h> |
1da177e4 | 27 | #include <linux/proc_fs.h> |
05039b92 | 28 | #include <linux/memory_hotplug.h> |
27d99f7e | 29 | #include <linux/initrd.h> |
55b2355e | 30 | #include <linux/cpumask.h> |
5a0e3ad6 | 31 | #include <linux/gfp.h> |
1da177e4 | 32 | |
f832ff18 | 33 | #include <asm/asm.h> |
46eaa670 | 34 | #include <asm/bios_ebda.h> |
1da177e4 | 35 | #include <asm/processor.h> |
7c0f6ba6 | 36 | #include <linux/uaccess.h> |
1da177e4 LT |
37 | #include <asm/pgtable.h> |
38 | #include <asm/dma.h> | |
39 | #include <asm/fixmap.h> | |
66441bd3 | 40 | #include <asm/e820/api.h> |
1da177e4 | 41 | #include <asm/apic.h> |
8550eb99 | 42 | #include <asm/bugs.h> |
1da177e4 LT |
43 | #include <asm/tlb.h> |
44 | #include <asm/tlbflush.h> | |
c10d1e26 | 45 | #include <asm/olpc_ofw.h> |
a5a19c63 | 46 | #include <asm/pgalloc.h> |
1da177e4 | 47 | #include <asm/sections.h> |
b239fb25 | 48 | #include <asm/paravirt.h> |
551889a6 | 49 | #include <asm/setup.h> |
d1163651 | 50 | #include <asm/set_memory.h> |
2b72394e | 51 | #include <asm/page_types.h> |
92a0f81d | 52 | #include <asm/cpu_entry_area.h> |
4fcb2083 | 53 | #include <asm/init.h> |
1da177e4 | 54 | |
5c51bdbe YL |
55 | #include "mm_internal.h" |
56 | ||
1da177e4 LT |
57 | unsigned long highstart_pfn, highend_pfn; |
58 | ||
dc16ecf7 | 59 | bool __read_mostly __vmalloc_start_set = false; |
4e29684c | 60 | |
1da177e4 LT |
61 | /* |
62 | * Creates a middle page table and puts a pointer to it in the | |
63 | * given global directory entry. This only returns the gd entry | |
64 | * in non-PAE compilation mode, since the middle layer is folded. | |
65 | */ | |
66 | static pmd_t * __init one_md_table_init(pgd_t *pgd) | |
67 | { | |
e0c4f675 | 68 | p4d_t *p4d; |
1da177e4 LT |
69 | pud_t *pud; |
70 | pmd_t *pmd_table; | |
8550eb99 | 71 | |
1da177e4 | 72 | #ifdef CONFIG_X86_PAE |
b239fb25 | 73 | if (!(pgd_val(*pgd) & _PAGE_PRESENT)) { |
4e37a890 | 74 | pmd_table = (pmd_t *)alloc_low_page(); |
6944a9c8 | 75 | paravirt_alloc_pmd(&init_mm, __pa(pmd_table) >> PAGE_SHIFT); |
b239fb25 | 76 | set_pgd(pgd, __pgd(__pa(pmd_table) | _PAGE_PRESENT)); |
e0c4f675 KS |
77 | p4d = p4d_offset(pgd, 0); |
78 | pud = pud_offset(p4d, 0); | |
8550eb99 | 79 | BUG_ON(pmd_table != pmd_offset(pud, 0)); |
a376f30a Z |
80 | |
81 | return pmd_table; | |
b239fb25 JF |
82 | } |
83 | #endif | |
e0c4f675 KS |
84 | p4d = p4d_offset(pgd, 0); |
85 | pud = pud_offset(p4d, 0); | |
1da177e4 | 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); | |
9962eea9 | 140 | pmd_idx = pmd_index(vaddr); |
719272c4 YL |
141 | |
142 | for ( ; (pgd_idx < PTRS_PER_PGD) && (vaddr != end); pgd_idx++) { | |
143 | for (; (pmd_idx < PTRS_PER_PMD) && (vaddr != end); | |
144 | pmd_idx++) { | |
145 | if ((vaddr >> PMD_SHIFT) >= pmd_idx_kmap_begin && | |
146 | (vaddr >> PMD_SHIFT) <= pmd_idx_kmap_end) | |
147 | count++; | |
148 | vaddr += PMD_SIZE; | |
149 | } | |
150 | pmd_idx = 0; | |
151 | } | |
152 | #endif | |
153 | return count; | |
154 | } | |
155 | ||
a3c6018e | 156 | static pte_t *__init page_table_kmap_check(pte_t *pte, pmd_t *pmd, |
719272c4 YL |
157 | unsigned long vaddr, pte_t *lastpte, |
158 | void **adr) | |
a3c6018e JB |
159 | { |
160 | #ifdef CONFIG_HIGHMEM | |
161 | /* | |
162 | * Something (early fixmap) may already have put a pte | |
163 | * page here, which causes the page table allocation | |
164 | * to become nonlinear. Attempt to fix it, and if it | |
165 | * is still nonlinear then we have to bug. | |
166 | */ | |
167 | int pmd_idx_kmap_begin = fix_to_virt(FIX_KMAP_END) >> PMD_SHIFT; | |
168 | int pmd_idx_kmap_end = fix_to_virt(FIX_KMAP_BEGIN) >> PMD_SHIFT; | |
169 | ||
170 | if (pmd_idx_kmap_begin != pmd_idx_kmap_end | |
171 | && (vaddr >> PMD_SHIFT) >= pmd_idx_kmap_begin | |
719272c4 | 172 | && (vaddr >> PMD_SHIFT) <= pmd_idx_kmap_end) { |
a3c6018e JB |
173 | pte_t *newpte; |
174 | int i; | |
175 | ||
c464573c | 176 | BUG_ON(after_bootmem); |
719272c4 | 177 | newpte = *adr; |
a3c6018e JB |
178 | for (i = 0; i < PTRS_PER_PTE; i++) |
179 | set_pte(newpte + i, pte[i]); | |
719272c4 | 180 | *adr = (void *)(((unsigned long)(*adr)) + PAGE_SIZE); |
a3c6018e JB |
181 | |
182 | paravirt_alloc_pte(&init_mm, __pa(newpte) >> PAGE_SHIFT); | |
183 | set_pmd(pmd, __pmd(__pa(newpte)|_PAGE_TABLE)); | |
184 | BUG_ON(newpte != pte_offset_kernel(pmd, 0)); | |
185 | __flush_tlb_all(); | |
186 | ||
187 | paravirt_release_pte(__pa(pte) >> PAGE_SHIFT); | |
188 | pte = newpte; | |
189 | } | |
190 | BUG_ON(vaddr < fix_to_virt(FIX_KMAP_BEGIN - 1) | |
191 | && vaddr > fix_to_virt(FIX_KMAP_END) | |
192 | && lastpte && lastpte + PTRS_PER_PTE != pte); | |
193 | #endif | |
194 | return pte; | |
195 | } | |
196 | ||
1da177e4 | 197 | /* |
8550eb99 | 198 | * This function initializes a certain range of kernel virtual memory |
1da177e4 LT |
199 | * with new bootmem page tables, everywhere page tables are missing in |
200 | * the given range. | |
8550eb99 IM |
201 | * |
202 | * NOTE: The pagetables are allocated contiguous on the physical space | |
203 | * so we can cache the place of the first one and move around without | |
1da177e4 LT |
204 | * checking the pgd every time. |
205 | */ | |
8550eb99 IM |
206 | static void __init |
207 | page_table_range_init(unsigned long start, unsigned long end, pgd_t *pgd_base) | |
1da177e4 | 208 | { |
1da177e4 LT |
209 | int pgd_idx, pmd_idx; |
210 | unsigned long vaddr; | |
8550eb99 IM |
211 | pgd_t *pgd; |
212 | pmd_t *pmd; | |
a3c6018e | 213 | pte_t *pte = NULL; |
719272c4 YL |
214 | unsigned long count = page_table_range_init_count(start, end); |
215 | void *adr = NULL; | |
216 | ||
217 | if (count) | |
218 | adr = alloc_low_pages(count); | |
1da177e4 LT |
219 | |
220 | vaddr = start; | |
221 | pgd_idx = pgd_index(vaddr); | |
222 | pmd_idx = pmd_index(vaddr); | |
223 | pgd = pgd_base + pgd_idx; | |
224 | ||
225 | for ( ; (pgd_idx < PTRS_PER_PGD) && (vaddr != end); pgd++, pgd_idx++) { | |
b239fb25 JF |
226 | pmd = one_md_table_init(pgd); |
227 | pmd = pmd + pmd_index(vaddr); | |
8550eb99 IM |
228 | for (; (pmd_idx < PTRS_PER_PMD) && (vaddr != end); |
229 | pmd++, pmd_idx++) { | |
a3c6018e | 230 | pte = page_table_kmap_check(one_page_table_init(pmd), |
719272c4 | 231 | pmd, vaddr, pte, &adr); |
1da177e4 LT |
232 | |
233 | vaddr += PMD_SIZE; | |
234 | } | |
235 | pmd_idx = 0; | |
236 | } | |
237 | } | |
238 | ||
239 | static inline int is_kernel_text(unsigned long addr) | |
240 | { | |
5bd5a452 | 241 | if (addr >= (unsigned long)_text && addr <= (unsigned long)__init_end) |
1da177e4 LT |
242 | return 1; |
243 | return 0; | |
244 | } | |
245 | ||
246 | /* | |
8550eb99 IM |
247 | * This maps the physical memory to kernel virtual address space, a total |
248 | * of max_low_pfn pages, by creating page tables starting from address | |
249 | * PAGE_OFFSET: | |
1da177e4 | 250 | */ |
e53fb04f PE |
251 | unsigned long __init |
252 | kernel_physical_mapping_init(unsigned long start, | |
253 | unsigned long end, | |
254 | unsigned long page_size_mask) | |
1da177e4 | 255 | { |
e53fb04f | 256 | int use_pse = page_size_mask == (1<<PG_LEVEL_2M); |
c1fd1b43 | 257 | unsigned long last_map_addr = end; |
e53fb04f | 258 | unsigned long start_pfn, end_pfn; |
e7179853 | 259 | pgd_t *pgd_base = swapper_pg_dir; |
8550eb99 | 260 | int pgd_idx, pmd_idx, pte_ofs; |
1da177e4 LT |
261 | unsigned long pfn; |
262 | pgd_t *pgd; | |
263 | pmd_t *pmd; | |
264 | pte_t *pte; | |
a2699e47 SS |
265 | unsigned pages_2m, pages_4k; |
266 | int mapping_iter; | |
267 | ||
e53fb04f PE |
268 | start_pfn = start >> PAGE_SHIFT; |
269 | end_pfn = end >> PAGE_SHIFT; | |
270 | ||
a2699e47 SS |
271 | /* |
272 | * First iteration will setup identity mapping using large/small pages | |
273 | * based on use_pse, with other attributes same as set by | |
274 | * the early code in head_32.S | |
275 | * | |
276 | * Second iteration will setup the appropriate attributes (NX, GLOBAL..) | |
277 | * as desired for the kernel identity mapping. | |
278 | * | |
279 | * This two pass mechanism conforms to the TLB app note which says: | |
280 | * | |
281 | * "Software should not write to a paging-structure entry in a way | |
282 | * that would change, for any linear address, both the page size | |
283 | * and either the page frame or attributes." | |
284 | */ | |
285 | mapping_iter = 1; | |
1da177e4 | 286 | |
16bf9226 | 287 | if (!boot_cpu_has(X86_FEATURE_PSE)) |
a04ad82d | 288 | use_pse = 0; |
1da177e4 | 289 | |
a2699e47 SS |
290 | repeat: |
291 | pages_2m = pages_4k = 0; | |
a04ad82d YL |
292 | pfn = start_pfn; |
293 | pgd_idx = pgd_index((pfn<<PAGE_SHIFT) + PAGE_OFFSET); | |
294 | pgd = pgd_base + pgd_idx; | |
1da177e4 LT |
295 | for (; pgd_idx < PTRS_PER_PGD; pgd++, pgd_idx++) { |
296 | pmd = one_md_table_init(pgd); | |
8550eb99 | 297 | |
a04ad82d YL |
298 | if (pfn >= end_pfn) |
299 | continue; | |
300 | #ifdef CONFIG_X86_PAE | |
301 | pmd_idx = pmd_index((pfn<<PAGE_SHIFT) + PAGE_OFFSET); | |
302 | pmd += pmd_idx; | |
303 | #else | |
304 | pmd_idx = 0; | |
305 | #endif | |
306 | for (; pmd_idx < PTRS_PER_PMD && pfn < end_pfn; | |
f3f20de8 | 307 | pmd++, pmd_idx++) { |
8550eb99 | 308 | unsigned int addr = pfn * PAGE_SIZE + PAGE_OFFSET; |
1da177e4 | 309 | |
8550eb99 IM |
310 | /* |
311 | * Map with big pages if possible, otherwise | |
312 | * create normal page tables: | |
313 | */ | |
a04ad82d | 314 | if (use_pse) { |
8550eb99 | 315 | unsigned int addr2; |
f3f20de8 | 316 | pgprot_t prot = PAGE_KERNEL_LARGE; |
a2699e47 SS |
317 | /* |
318 | * first pass will use the same initial | |
319 | * identity mapping attribute + _PAGE_PSE. | |
320 | */ | |
321 | pgprot_t init_prot = | |
322 | __pgprot(PTE_IDENT_ATTR | | |
323 | _PAGE_PSE); | |
f3f20de8 | 324 | |
960ddb4f | 325 | pfn &= PMD_MASK >> PAGE_SHIFT; |
8550eb99 | 326 | addr2 = (pfn + PTRS_PER_PTE-1) * PAGE_SIZE + |
f3f20de8 JF |
327 | PAGE_OFFSET + PAGE_SIZE-1; |
328 | ||
8550eb99 IM |
329 | if (is_kernel_text(addr) || |
330 | is_kernel_text(addr2)) | |
f3f20de8 JF |
331 | prot = PAGE_KERNEL_LARGE_EXEC; |
332 | ||
ce0c0e50 | 333 | pages_2m++; |
a2699e47 SS |
334 | if (mapping_iter == 1) |
335 | set_pmd(pmd, pfn_pmd(pfn, init_prot)); | |
336 | else | |
337 | set_pmd(pmd, pfn_pmd(pfn, prot)); | |
b239fb25 | 338 | |
1da177e4 | 339 | pfn += PTRS_PER_PTE; |
8550eb99 IM |
340 | continue; |
341 | } | |
342 | pte = one_page_table_init(pmd); | |
1da177e4 | 343 | |
a04ad82d YL |
344 | pte_ofs = pte_index((pfn<<PAGE_SHIFT) + PAGE_OFFSET); |
345 | pte += pte_ofs; | |
346 | for (; pte_ofs < PTRS_PER_PTE && pfn < end_pfn; | |
8550eb99 IM |
347 | pte++, pfn++, pte_ofs++, addr += PAGE_SIZE) { |
348 | pgprot_t prot = PAGE_KERNEL; | |
a2699e47 SS |
349 | /* |
350 | * first pass will use the same initial | |
351 | * identity mapping attribute. | |
352 | */ | |
353 | pgprot_t init_prot = __pgprot(PTE_IDENT_ATTR); | |
f3f20de8 | 354 | |
8550eb99 IM |
355 | if (is_kernel_text(addr)) |
356 | prot = PAGE_KERNEL_EXEC; | |
f3f20de8 | 357 | |
ce0c0e50 | 358 | pages_4k++; |
c1fd1b43 | 359 | if (mapping_iter == 1) { |
a2699e47 | 360 | set_pte(pte, pfn_pte(pfn, init_prot)); |
c1fd1b43 PE |
361 | last_map_addr = (pfn << PAGE_SHIFT) + PAGE_SIZE; |
362 | } else | |
a2699e47 | 363 | set_pte(pte, pfn_pte(pfn, prot)); |
1da177e4 LT |
364 | } |
365 | } | |
366 | } | |
a2699e47 SS |
367 | if (mapping_iter == 1) { |
368 | /* | |
369 | * update direct mapping page count only in the first | |
370 | * iteration. | |
371 | */ | |
372 | update_page_count(PG_LEVEL_2M, pages_2m); | |
373 | update_page_count(PG_LEVEL_4K, pages_4k); | |
374 | ||
375 | /* | |
376 | * local global flush tlb, which will flush the previous | |
377 | * mappings present in both small and large page TLB's. | |
378 | */ | |
379 | __flush_tlb_all(); | |
380 | ||
381 | /* | |
382 | * Second iteration will set the actual desired PTE attributes. | |
383 | */ | |
384 | mapping_iter = 2; | |
385 | goto repeat; | |
386 | } | |
c1fd1b43 | 387 | return last_map_addr; |
ae531c26 AV |
388 | } |
389 | ||
1da177e4 | 390 | pte_t *kmap_pte; |
1da177e4 | 391 | |
8550eb99 IM |
392 | static inline pte_t *kmap_get_fixmap_pte(unsigned long vaddr) |
393 | { | |
e0c4f675 KS |
394 | pgd_t *pgd = pgd_offset_k(vaddr); |
395 | p4d_t *p4d = p4d_offset(pgd, vaddr); | |
396 | pud_t *pud = pud_offset(p4d, vaddr); | |
397 | pmd_t *pmd = pmd_offset(pud, vaddr); | |
398 | return pte_offset_kernel(pmd, vaddr); | |
8550eb99 | 399 | } |
1da177e4 LT |
400 | |
401 | static void __init kmap_init(void) | |
402 | { | |
403 | unsigned long kmap_vstart; | |
404 | ||
8550eb99 IM |
405 | /* |
406 | * Cache the first kmap pte: | |
407 | */ | |
1da177e4 LT |
408 | kmap_vstart = __fix_to_virt(FIX_KMAP_BEGIN); |
409 | kmap_pte = kmap_get_fixmap_pte(kmap_vstart); | |
1da177e4 LT |
410 | } |
411 | ||
fd940934 | 412 | #ifdef CONFIG_HIGHMEM |
1da177e4 LT |
413 | static void __init permanent_kmaps_init(pgd_t *pgd_base) |
414 | { | |
8550eb99 | 415 | unsigned long vaddr; |
1da177e4 | 416 | pgd_t *pgd; |
e0c4f675 | 417 | p4d_t *p4d; |
1da177e4 LT |
418 | pud_t *pud; |
419 | pmd_t *pmd; | |
420 | pte_t *pte; | |
1da177e4 LT |
421 | |
422 | vaddr = PKMAP_BASE; | |
423 | page_table_range_init(vaddr, vaddr + PAGE_SIZE*LAST_PKMAP, pgd_base); | |
424 | ||
425 | pgd = swapper_pg_dir + pgd_index(vaddr); | |
e0c4f675 KS |
426 | p4d = p4d_offset(pgd, vaddr); |
427 | pud = pud_offset(p4d, vaddr); | |
1da177e4 LT |
428 | pmd = pmd_offset(pud, vaddr); |
429 | pte = pte_offset_kernel(pmd, vaddr); | |
8550eb99 | 430 | pkmap_page_table = pte; |
1da177e4 LT |
431 | } |
432 | ||
1d931264 YL |
433 | void __init add_highpages_with_active_regions(int nid, |
434 | unsigned long start_pfn, unsigned long end_pfn) | |
1da177e4 | 435 | { |
8a9ca34c TH |
436 | phys_addr_t start, end; |
437 | u64 i; | |
438 | ||
fc6daaf9 | 439 | for_each_free_mem_range(i, nid, MEMBLOCK_NONE, &start, &end, NULL) { |
8a9ca34c TH |
440 | unsigned long pfn = clamp_t(unsigned long, PFN_UP(start), |
441 | start_pfn, end_pfn); | |
442 | unsigned long e_pfn = clamp_t(unsigned long, PFN_DOWN(end), | |
443 | start_pfn, end_pfn); | |
444 | for ( ; pfn < e_pfn; pfn++) | |
445 | if (pfn_valid(pfn)) | |
5e7ccf86 | 446 | free_highmem_page(pfn_to_page(pfn)); |
23be8c7d | 447 | } |
b5bc6c0e | 448 | } |
1da177e4 | 449 | #else |
e8e32326 IB |
450 | static inline void permanent_kmaps_init(pgd_t *pgd_base) |
451 | { | |
452 | } | |
1da177e4 LT |
453 | #endif /* CONFIG_HIGHMEM */ |
454 | ||
945fd17a TG |
455 | void __init sync_initial_page_table(void) |
456 | { | |
457 | clone_pgd_range(initial_page_table + KERNEL_PGD_BOUNDARY, | |
458 | swapper_pg_dir + KERNEL_PGD_BOUNDARY, | |
459 | KERNEL_PGD_PTRS); | |
460 | ||
461 | /* | |
462 | * sync back low identity map too. It is used for example | |
463 | * in the 32-bit EFI stub. | |
464 | */ | |
465 | clone_pgd_range(initial_page_table, | |
466 | swapper_pg_dir + KERNEL_PGD_BOUNDARY, | |
467 | min(KERNEL_PGD_PTRS, KERNEL_PGD_BOUNDARY)); | |
468 | } | |
469 | ||
7737b215 | 470 | void __init native_pagetable_init(void) |
1da177e4 | 471 | { |
551889a6 | 472 | unsigned long pfn, va; |
73090f89 | 473 | pgd_t *pgd, *base = swapper_pg_dir; |
e0c4f675 | 474 | p4d_t *p4d; |
551889a6 IC |
475 | pud_t *pud; |
476 | pmd_t *pmd; | |
477 | pte_t *pte; | |
b239fb25 JF |
478 | |
479 | /* | |
551889a6 | 480 | * Remove any mappings which extend past the end of physical |
11ed9e92 YL |
481 | * memory from the boot time page table. |
482 | * In virtual address space, we should have at least two pages | |
483 | * from VMALLOC_END to pkmap or fixmap according to VMALLOC_END | |
484 | * definition. And max_low_pfn is set to VMALLOC_END physical | |
485 | * address. If initial memory mapping is doing right job, we | |
486 | * should have pte used near max_low_pfn or one pmd is not present. | |
b239fb25 | 487 | */ |
11ed9e92 | 488 | for (pfn = max_low_pfn; pfn < 1<<(32-PAGE_SHIFT); pfn++) { |
551889a6 IC |
489 | va = PAGE_OFFSET + (pfn<<PAGE_SHIFT); |
490 | pgd = base + pgd_index(va); | |
491 | if (!pgd_present(*pgd)) | |
492 | break; | |
493 | ||
e0c4f675 KS |
494 | p4d = p4d_offset(pgd, va); |
495 | pud = pud_offset(p4d, va); | |
551889a6 IC |
496 | pmd = pmd_offset(pud, va); |
497 | if (!pmd_present(*pmd)) | |
498 | break; | |
499 | ||
11ed9e92 YL |
500 | /* should not be large page here */ |
501 | if (pmd_large(*pmd)) { | |
502 | 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", | |
503 | pfn, pmd, __pa(pmd)); | |
504 | BUG_ON(1); | |
505 | } | |
506 | ||
551889a6 IC |
507 | pte = pte_offset_kernel(pmd, va); |
508 | if (!pte_present(*pte)) | |
509 | break; | |
510 | ||
11ed9e92 YL |
511 | printk(KERN_DEBUG "clearing pte for ram above max_low_pfn: pfn: %lx pmd: %p pmd phys: %lx pte: %p pte phys: %lx\n", |
512 | pfn, pmd, __pa(pmd), pte, __pa(pte)); | |
551889a6 IC |
513 | pte_clear(NULL, va, pte); |
514 | } | |
6944a9c8 | 515 | paravirt_alloc_pmd(&init_mm, __pa(base) >> PAGE_SHIFT); |
843b8ed2 | 516 | paging_init(); |
b239fb25 JF |
517 | } |
518 | ||
b239fb25 JF |
519 | /* |
520 | * Build a proper pagetable for the kernel mappings. Up until this | |
521 | * point, we've been running on some set of pagetables constructed by | |
522 | * the boot process. | |
523 | * | |
524 | * If we're booting on native hardware, this will be a pagetable | |
551889a6 IC |
525 | * constructed in arch/x86/kernel/head_32.S. The root of the |
526 | * pagetable will be swapper_pg_dir. | |
b239fb25 JF |
527 | * |
528 | * If we're booting paravirtualized under a hypervisor, then there are | |
529 | * more options: we may already be running PAE, and the pagetable may | |
530 | * or may not be based in swapper_pg_dir. In any case, | |
7737b215 | 531 | * paravirt_pagetable_init() will set up swapper_pg_dir |
b239fb25 JF |
532 | * appropriately for the rest of the initialization to work. |
533 | * | |
534 | * In general, pagetable_init() assumes that the pagetable may already | |
535 | * be partially populated, and so it avoids stomping on any existing | |
536 | * mappings. | |
537 | */ | |
f765090a | 538 | void __init early_ioremap_page_table_range_init(void) |
b239fb25 | 539 | { |
e7179853 | 540 | pgd_t *pgd_base = swapper_pg_dir; |
8550eb99 | 541 | unsigned long vaddr, end; |
b239fb25 | 542 | |
1da177e4 LT |
543 | /* |
544 | * Fixed mappings, only the page table structure has to be | |
545 | * created - mappings will be set by set_fixmap(): | |
546 | */ | |
547 | vaddr = __fix_to_virt(__end_of_fixed_addresses - 1) & PMD_MASK; | |
b239fb25 JF |
548 | end = (FIXADDR_TOP + PMD_SIZE - 1) & PMD_MASK; |
549 | page_table_range_init(vaddr, end, pgd_base); | |
beacfaac | 550 | early_ioremap_reset(); |
e7b37895 YL |
551 | } |
552 | ||
553 | static void __init pagetable_init(void) | |
554 | { | |
555 | pgd_t *pgd_base = swapper_pg_dir; | |
556 | ||
1da177e4 | 557 | permanent_kmaps_init(pgd_base); |
1da177e4 LT |
558 | } |
559 | ||
8a57f484 DH |
560 | #define DEFAULT_PTE_MASK ~(_PAGE_NX | _PAGE_GLOBAL) |
561 | /* Bits supported by the hardware: */ | |
562 | pteval_t __supported_pte_mask __read_mostly = DEFAULT_PTE_MASK; | |
563 | /* Bits allowed in normal kernel mappings: */ | |
564 | pteval_t __default_kernel_pte_mask __read_mostly = DEFAULT_PTE_MASK; | |
6fdc05d4 | 565 | EXPORT_SYMBOL_GPL(__supported_pte_mask); |
8a57f484 DH |
566 | /* Used in PAGE_KERNEL_* macros which are reasonably used out-of-tree: */ |
567 | EXPORT_SYMBOL(__default_kernel_pte_mask); | |
6fdc05d4 | 568 | |
90d967e0 YL |
569 | /* user-defined highmem size */ |
570 | static unsigned int highmem_pages = -1; | |
571 | ||
572 | /* | |
573 | * highmem=size forces highmem to be exactly 'size' bytes. | |
574 | * This works even on boxes that have no highmem otherwise. | |
575 | * This also works to reduce highmem size on bigger boxes. | |
576 | */ | |
577 | static int __init parse_highmem(char *arg) | |
578 | { | |
579 | if (!arg) | |
580 | return -EINVAL; | |
581 | ||
582 | highmem_pages = memparse(arg, &arg) >> PAGE_SHIFT; | |
583 | return 0; | |
584 | } | |
585 | early_param("highmem", parse_highmem); | |
586 | ||
4769843b IM |
587 | #define MSG_HIGHMEM_TOO_BIG \ |
588 | "highmem size (%luMB) is bigger than pages available (%luMB)!\n" | |
589 | ||
590 | #define MSG_LOWMEM_TOO_SMALL \ | |
591 | "highmem size (%luMB) results in <64MB lowmem, ignoring it!\n" | |
90d967e0 | 592 | /* |
4769843b IM |
593 | * All of RAM fits into lowmem - but if user wants highmem |
594 | * artificially via the highmem=x boot parameter then create | |
595 | * it: | |
90d967e0 | 596 | */ |
f836e35a | 597 | static void __init lowmem_pfn_init(void) |
90d967e0 | 598 | { |
346cafec | 599 | /* max_low_pfn is 0, we already have early_res support */ |
90d967e0 | 600 | max_low_pfn = max_pfn; |
d88316c2 | 601 | |
4769843b IM |
602 | if (highmem_pages == -1) |
603 | highmem_pages = 0; | |
604 | #ifdef CONFIG_HIGHMEM | |
605 | if (highmem_pages >= max_pfn) { | |
606 | printk(KERN_ERR MSG_HIGHMEM_TOO_BIG, | |
607 | pages_to_mb(highmem_pages), pages_to_mb(max_pfn)); | |
608 | highmem_pages = 0; | |
609 | } | |
610 | if (highmem_pages) { | |
611 | if (max_low_pfn - highmem_pages < 64*1024*1024/PAGE_SIZE) { | |
612 | printk(KERN_ERR MSG_LOWMEM_TOO_SMALL, | |
90d967e0 YL |
613 | pages_to_mb(highmem_pages)); |
614 | highmem_pages = 0; | |
615 | } | |
4769843b IM |
616 | max_low_pfn -= highmem_pages; |
617 | } | |
618 | #else | |
619 | if (highmem_pages) | |
620 | printk(KERN_ERR "ignoring highmem size on non-highmem kernel!\n"); | |
621 | #endif | |
622 | } | |
623 | ||
624 | #define MSG_HIGHMEM_TOO_SMALL \ | |
625 | "only %luMB highmem pages available, ignoring highmem size of %luMB!\n" | |
626 | ||
627 | #define MSG_HIGHMEM_TRIMMED \ | |
628 | "Warning: only 4GB will be used. Use a HIGHMEM64G enabled kernel!\n" | |
629 | /* | |
630 | * We have more RAM than fits into lowmem - we try to put it into | |
631 | * highmem, also taking the highmem=x boot parameter into account: | |
632 | */ | |
f836e35a | 633 | static void __init highmem_pfn_init(void) |
4769843b | 634 | { |
d88316c2 IM |
635 | max_low_pfn = MAXMEM_PFN; |
636 | ||
4769843b IM |
637 | if (highmem_pages == -1) |
638 | highmem_pages = max_pfn - MAXMEM_PFN; | |
639 | ||
640 | if (highmem_pages + MAXMEM_PFN < max_pfn) | |
641 | max_pfn = MAXMEM_PFN + highmem_pages; | |
642 | ||
643 | if (highmem_pages + MAXMEM_PFN > max_pfn) { | |
644 | printk(KERN_WARNING MSG_HIGHMEM_TOO_SMALL, | |
645 | pages_to_mb(max_pfn - MAXMEM_PFN), | |
646 | pages_to_mb(highmem_pages)); | |
647 | highmem_pages = 0; | |
648 | } | |
90d967e0 | 649 | #ifndef CONFIG_HIGHMEM |
4769843b IM |
650 | /* Maximum memory usable is what is directly addressable */ |
651 | printk(KERN_WARNING "Warning only %ldMB will be used.\n", MAXMEM>>20); | |
652 | if (max_pfn > MAX_NONPAE_PFN) | |
653 | printk(KERN_WARNING "Use a HIGHMEM64G enabled kernel.\n"); | |
654 | else | |
655 | printk(KERN_WARNING "Use a HIGHMEM enabled kernel.\n"); | |
656 | max_pfn = MAXMEM_PFN; | |
90d967e0 YL |
657 | #else /* !CONFIG_HIGHMEM */ |
658 | #ifndef CONFIG_HIGHMEM64G | |
4769843b IM |
659 | if (max_pfn > MAX_NONPAE_PFN) { |
660 | max_pfn = MAX_NONPAE_PFN; | |
661 | printk(KERN_WARNING MSG_HIGHMEM_TRIMMED); | |
662 | } | |
90d967e0 YL |
663 | #endif /* !CONFIG_HIGHMEM64G */ |
664 | #endif /* !CONFIG_HIGHMEM */ | |
4769843b IM |
665 | } |
666 | ||
667 | /* | |
668 | * Determine low and high memory ranges: | |
669 | */ | |
670 | void __init find_low_pfn_range(void) | |
671 | { | |
672 | /* it could update max_pfn */ | |
673 | ||
d88316c2 | 674 | if (max_pfn <= MAXMEM_PFN) |
4769843b | 675 | lowmem_pfn_init(); |
d88316c2 IM |
676 | else |
677 | highmem_pfn_init(); | |
90d967e0 YL |
678 | } |
679 | ||
b2ac82a0 | 680 | #ifndef CONFIG_NEED_MULTIPLE_NODES |
d8fc3afc | 681 | void __init initmem_init(void) |
b2ac82a0 | 682 | { |
b2ac82a0 YL |
683 | #ifdef CONFIG_HIGHMEM |
684 | highstart_pfn = highend_pfn = max_pfn; | |
685 | if (max_pfn > max_low_pfn) | |
686 | highstart_pfn = max_low_pfn; | |
b2ac82a0 YL |
687 | printk(KERN_NOTICE "%ldMB HIGHMEM available.\n", |
688 | pages_to_mb(highend_pfn - highstart_pfn)); | |
b2ac82a0 YL |
689 | high_memory = (void *) __va(highstart_pfn * PAGE_SIZE - 1) + 1; |
690 | #else | |
b2ac82a0 YL |
691 | high_memory = (void *) __va(max_low_pfn * PAGE_SIZE - 1) + 1; |
692 | #endif | |
0608f70c | 693 | |
d7dc899a | 694 | memblock_set_node(0, PHYS_ADDR_MAX, &memblock.memory, 0); |
0608f70c TH |
695 | sparse_memory_present_with_active_regions(0); |
696 | ||
b2ac82a0 | 697 | #ifdef CONFIG_FLATMEM |
46a84132 | 698 | max_mapnr = IS_ENABLED(CONFIG_HIGHMEM) ? highend_pfn : max_low_pfn; |
b2ac82a0 | 699 | #endif |
dc16ecf7 JF |
700 | __vmalloc_start_set = true; |
701 | ||
b2ac82a0 YL |
702 | printk(KERN_NOTICE "%ldMB LOWMEM available.\n", |
703 | pages_to_mb(max_low_pfn)); | |
704 | ||
705 | setup_bootmem_allocator(); | |
b2ac82a0 | 706 | } |
cb95a13a | 707 | #endif /* !CONFIG_NEED_MULTIPLE_NODES */ |
b2ac82a0 | 708 | |
b2ac82a0 YL |
709 | void __init setup_bootmem_allocator(void) |
710 | { | |
b2ac82a0 YL |
711 | printk(KERN_INFO " mapped low ram: 0 - %08lx\n", |
712 | max_pfn_mapped<<PAGE_SHIFT); | |
fc5efe39 | 713 | printk(KERN_INFO " low ram: 0 - %08lx\n", max_low_pfn<<PAGE_SHIFT); |
4e29684c YL |
714 | } |
715 | ||
1da177e4 LT |
716 | /* |
717 | * paging_init() sets up the page tables - note that the first 8MB are | |
718 | * already mapped by head.S. | |
719 | * | |
720 | * This routines also unmaps the page at virtual kernel address 0, so | |
721 | * that we can trap those pesky NULL-reference errors in the kernel. | |
722 | */ | |
723 | void __init paging_init(void) | |
724 | { | |
1da177e4 LT |
725 | pagetable_init(); |
726 | ||
1da177e4 LT |
727 | __flush_tlb_all(); |
728 | ||
729 | kmap_init(); | |
11cd0bc1 YL |
730 | |
731 | /* | |
732 | * NOTE: at this point the bootmem allocator is fully available. | |
733 | */ | |
c10d1e26 | 734 | olpc_dt_build_devicetree(); |
797390d8 | 735 | sparse_memory_present_with_active_regions(MAX_NUMNODES); |
11cd0bc1 YL |
736 | sparse_init(); |
737 | zone_sizes_init(); | |
1da177e4 LT |
738 | } |
739 | ||
740 | /* | |
741 | * Test if the WP bit works in supervisor mode. It isn't supported on 386's | |
f7f17a67 DV |
742 | * and also on some strange 486's. All 586+'s are OK. This used to involve |
743 | * black magic jumps to work around some nasty CPU bugs, but fortunately the | |
744 | * switch to using exceptions got rid of all that. | |
1da177e4 | 745 | */ |
1da177e4 LT |
746 | static void __init test_wp_bit(void) |
747 | { | |
4af17110 | 748 | char z = 0; |
6415813b | 749 | |
952a6c2c | 750 | printk(KERN_INFO "Checking if this processor honours the WP bit even in supervisor mode..."); |
1da177e4 | 751 | |
4af17110 | 752 | __set_fixmap(FIX_WP_TEST, __pa_symbol(empty_zero_page), PAGE_KERNEL_RO); |
1da177e4 | 753 | |
952a6c2c BP |
754 | if (probe_kernel_write((char *)fix_to_virt(FIX_WP_TEST), &z, 1)) { |
755 | clear_fixmap(FIX_WP_TEST); | |
756 | printk(KERN_CONT "Ok.\n"); | |
757 | return; | |
1da177e4 | 758 | } |
4af17110 | 759 | |
952a6c2c BP |
760 | printk(KERN_CONT "No.\n"); |
761 | panic("Linux doesn't support CPUs with broken WP."); | |
1da177e4 LT |
762 | } |
763 | ||
1da177e4 LT |
764 | void __init mem_init(void) |
765 | { | |
cfb80c9e JF |
766 | pci_iommu_alloc(); |
767 | ||
05b79bdc | 768 | #ifdef CONFIG_FLATMEM |
8d8f3cbe | 769 | BUG_ON(!mem_map); |
1da177e4 | 770 | #endif |
855c743a SG |
771 | /* |
772 | * With CONFIG_DEBUG_PAGEALLOC initialization of highmem pages has to | |
c6ffc5ca | 773 | * be done before memblock_free_all(). Memblock use free low memory for |
855c743a SG |
774 | * temporary data (see find_range_array()) and for this purpose can use |
775 | * pages that was already passed to the buddy allocator, hence marked as | |
776 | * not accessible in the page tables when compiled with | |
777 | * CONFIG_DEBUG_PAGEALLOC. Otherwise order of initialization is not | |
778 | * important here. | |
779 | */ | |
780 | set_highmem_pages_init(); | |
781 | ||
1da177e4 | 782 | /* this will put all low memory onto the freelists */ |
c6ffc5ca | 783 | memblock_free_all(); |
1da177e4 | 784 | |
4e37a890 | 785 | after_bootmem = 1; |
6f84f8d1 | 786 | x86_init.hyper.init_after_bootmem(); |
4e37a890 | 787 | |
46a84132 | 788 | mem_init_print_info(NULL); |
d7d119d7 | 789 | printk(KERN_INFO "virtual kernel memory layout:\n" |
8550eb99 | 790 | " fixmap : 0x%08lx - 0x%08lx (%4ld kB)\n" |
92a0f81d | 791 | " cpu_entry : 0x%08lx - 0x%08lx (%4ld kB)\n" |
052e7994 | 792 | #ifdef CONFIG_HIGHMEM |
8550eb99 | 793 | " pkmap : 0x%08lx - 0x%08lx (%4ld kB)\n" |
052e7994 | 794 | #endif |
8550eb99 IM |
795 | " vmalloc : 0x%08lx - 0x%08lx (%4ld MB)\n" |
796 | " lowmem : 0x%08lx - 0x%08lx (%4ld MB)\n" | |
797 | " .init : 0x%08lx - 0x%08lx (%4ld kB)\n" | |
798 | " .data : 0x%08lx - 0x%08lx (%4ld kB)\n" | |
799 | " .text : 0x%08lx - 0x%08lx (%4ld kB)\n", | |
800 | FIXADDR_START, FIXADDR_TOP, | |
801 | (FIXADDR_TOP - FIXADDR_START) >> 10, | |
052e7994 | 802 | |
92a0f81d TG |
803 | CPU_ENTRY_AREA_BASE, |
804 | CPU_ENTRY_AREA_BASE + CPU_ENTRY_AREA_MAP_SIZE, | |
805 | CPU_ENTRY_AREA_MAP_SIZE >> 10, | |
806 | ||
052e7994 | 807 | #ifdef CONFIG_HIGHMEM |
8550eb99 IM |
808 | PKMAP_BASE, PKMAP_BASE+LAST_PKMAP*PAGE_SIZE, |
809 | (LAST_PKMAP*PAGE_SIZE) >> 10, | |
052e7994 JF |
810 | #endif |
811 | ||
8550eb99 IM |
812 | VMALLOC_START, VMALLOC_END, |
813 | (VMALLOC_END - VMALLOC_START) >> 20, | |
052e7994 | 814 | |
8550eb99 IM |
815 | (unsigned long)__va(0), (unsigned long)high_memory, |
816 | ((unsigned long)high_memory - (unsigned long)__va(0)) >> 20, | |
052e7994 | 817 | |
8550eb99 IM |
818 | (unsigned long)&__init_begin, (unsigned long)&__init_end, |
819 | ((unsigned long)&__init_end - | |
820 | (unsigned long)&__init_begin) >> 10, | |
052e7994 | 821 | |
8550eb99 IM |
822 | (unsigned long)&_etext, (unsigned long)&_edata, |
823 | ((unsigned long)&_edata - (unsigned long)&_etext) >> 10, | |
052e7994 | 824 | |
8550eb99 IM |
825 | (unsigned long)&_text, (unsigned long)&_etext, |
826 | ((unsigned long)&_etext - (unsigned long)&_text) >> 10); | |
052e7994 | 827 | |
beeb4195 JB |
828 | /* |
829 | * Check boundaries twice: Some fundamental inconsistencies can | |
830 | * be detected at build time already. | |
831 | */ | |
832 | #define __FIXADDR_TOP (-PAGE_SIZE) | |
833 | #ifdef CONFIG_HIGHMEM | |
834 | BUILD_BUG_ON(PKMAP_BASE + LAST_PKMAP*PAGE_SIZE > FIXADDR_START); | |
835 | BUILD_BUG_ON(VMALLOC_END > PKMAP_BASE); | |
836 | #endif | |
837 | #define high_memory (-128UL << 20) | |
838 | BUILD_BUG_ON(VMALLOC_START >= VMALLOC_END); | |
839 | #undef high_memory | |
840 | #undef __FIXADDR_TOP | |
841 | ||
052e7994 | 842 | #ifdef CONFIG_HIGHMEM |
8550eb99 IM |
843 | BUG_ON(PKMAP_BASE + LAST_PKMAP*PAGE_SIZE > FIXADDR_START); |
844 | BUG_ON(VMALLOC_END > PKMAP_BASE); | |
052e7994 | 845 | #endif |
beeb4195 | 846 | BUG_ON(VMALLOC_START >= VMALLOC_END); |
8550eb99 | 847 | BUG_ON((unsigned long)high_memory > VMALLOC_START); |
052e7994 | 848 | |
6415813b | 849 | test_wp_bit(); |
1da177e4 LT |
850 | } |
851 | ||
ad8f5797 | 852 | #ifdef CONFIG_MEMORY_HOTPLUG |
24e6d5a5 CH |
853 | int arch_add_memory(int nid, u64 start, u64 size, struct vmem_altmap *altmap, |
854 | bool want_memblock) | |
05039b92 | 855 | { |
05039b92 DH |
856 | unsigned long start_pfn = start >> PAGE_SHIFT; |
857 | unsigned long nr_pages = size >> PAGE_SHIFT; | |
858 | ||
24e6d5a5 | 859 | return __add_pages(nid, start_pfn, nr_pages, altmap, want_memblock); |
05039b92 | 860 | } |
24d335ca WC |
861 | |
862 | #ifdef CONFIG_MEMORY_HOTREMOVE | |
da024512 | 863 | int arch_remove_memory(u64 start, u64 size, struct vmem_altmap *altmap) |
24d335ca WC |
864 | { |
865 | unsigned long start_pfn = start >> PAGE_SHIFT; | |
866 | unsigned long nr_pages = size >> PAGE_SHIFT; | |
867 | struct zone *zone; | |
868 | ||
869 | zone = page_zone(pfn_to_page(start_pfn)); | |
da024512 | 870 | return __remove_pages(zone, start_pfn, nr_pages, altmap); |
24d335ca WC |
871 | } |
872 | #endif | |
9d99aaa3 | 873 | #endif |
05039b92 | 874 | |
502f6604 | 875 | int kernel_set_to_readonly __read_mostly; |
16239630 SR |
876 | |
877 | void set_kernel_text_rw(void) | |
878 | { | |
879 | unsigned long start = PFN_ALIGN(_text); | |
880 | unsigned long size = PFN_ALIGN(_etext) - start; | |
881 | ||
882 | if (!kernel_set_to_readonly) | |
883 | return; | |
884 | ||
885 | pr_debug("Set kernel text: %lx - %lx for read write\n", | |
886 | start, start+size); | |
887 | ||
888 | set_pages_rw(virt_to_page(start), size >> PAGE_SHIFT); | |
889 | } | |
890 | ||
891 | void set_kernel_text_ro(void) | |
892 | { | |
893 | unsigned long start = PFN_ALIGN(_text); | |
894 | unsigned long size = PFN_ALIGN(_etext) - start; | |
895 | ||
896 | if (!kernel_set_to_readonly) | |
897 | return; | |
898 | ||
899 | pr_debug("Set kernel text: %lx - %lx for read only\n", | |
900 | start, start+size); | |
901 | ||
902 | set_pages_ro(virt_to_page(start), size >> PAGE_SHIFT); | |
903 | } | |
904 | ||
5bd5a452 MC |
905 | static void mark_nxdata_nx(void) |
906 | { | |
907 | /* | |
908 | * When this called, init has already been executed and released, | |
0d2eb44f | 909 | * so everything past _etext should be NX. |
5bd5a452 MC |
910 | */ |
911 | unsigned long start = PFN_ALIGN(_etext); | |
912 | /* | |
913 | * This comes from is_kernel_text upper limit. Also HPAGE where used: | |
914 | */ | |
915 | unsigned long size = (((unsigned long)__init_end + HPAGE_SIZE) & HPAGE_MASK) - start; | |
916 | ||
917 | if (__supported_pte_mask & _PAGE_NX) | |
918 | printk(KERN_INFO "NX-protecting the kernel data: %luk\n", size >> 10); | |
919 | set_pages_nx(virt_to_page(start), size >> PAGE_SHIFT); | |
920 | } | |
921 | ||
63aaf308 AV |
922 | void mark_rodata_ro(void) |
923 | { | |
6fb14755 | 924 | unsigned long start = PFN_ALIGN(_text); |
2a25dc7c | 925 | unsigned long size = (unsigned long)__end_rodata - start; |
63aaf308 | 926 | |
4e4eee0e | 927 | set_pages_ro(virt_to_page(start), size >> PAGE_SHIFT); |
2a25dc7c | 928 | pr_info("Write protecting kernel text and read-only data: %luk\n", |
4e4eee0e | 929 | size >> 10); |
0c42f392 | 930 | |
16239630 SR |
931 | kernel_set_to_readonly = 1; |
932 | ||
0c42f392 | 933 | #ifdef CONFIG_CPA_DEBUG |
2a25dc7c | 934 | pr_info("Testing CPA: Reverting %lx-%lx\n", start, start + size); |
6d238cc4 | 935 | set_pages_rw(virt_to_page(start), size >> PAGE_SHIFT); |
0c42f392 | 936 | |
2a25dc7c | 937 | pr_info("Testing CPA: write protecting again\n"); |
6d238cc4 | 938 | set_pages_ro(virt_to_page(start), size >> PAGE_SHIFT); |
0c42f392 | 939 | #endif |
5bd5a452 | 940 | mark_nxdata_nx(); |
e1a58320 SS |
941 | if (__supported_pte_mask & _PAGE_NX) |
942 | debug_checkwx(); | |
63aaf308 | 943 | } |