]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/arch/i386/mm/init.c | |
3 | * | |
4 | * Copyright (C) 1995 Linus Torvalds | |
5 | * | |
6 | * Support of BIGMEM added by Gerhard Wichert, Siemens AG, July 1999 | |
7 | */ | |
8 | ||
1da177e4 LT |
9 | #include <linux/module.h> |
10 | #include <linux/signal.h> | |
11 | #include <linux/sched.h> | |
12 | #include <linux/kernel.h> | |
13 | #include <linux/errno.h> | |
14 | #include <linux/string.h> | |
15 | #include <linux/types.h> | |
16 | #include <linux/ptrace.h> | |
17 | #include <linux/mman.h> | |
18 | #include <linux/mm.h> | |
19 | #include <linux/hugetlb.h> | |
20 | #include <linux/swap.h> | |
21 | #include <linux/smp.h> | |
22 | #include <linux/init.h> | |
23 | #include <linux/highmem.h> | |
24 | #include <linux/pagemap.h> | |
6fb14755 | 25 | #include <linux/pfn.h> |
c9cf5528 | 26 | #include <linux/poison.h> |
1da177e4 LT |
27 | #include <linux/bootmem.h> |
28 | #include <linux/slab.h> | |
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> |
1da177e4 LT |
33 | |
34 | #include <asm/processor.h> | |
35 | #include <asm/system.h> | |
36 | #include <asm/uaccess.h> | |
37 | #include <asm/pgtable.h> | |
38 | #include <asm/dma.h> | |
39 | #include <asm/fixmap.h> | |
40 | #include <asm/e820.h> | |
41 | #include <asm/apic.h> | |
8550eb99 | 42 | #include <asm/bugs.h> |
1da177e4 LT |
43 | #include <asm/tlb.h> |
44 | #include <asm/tlbflush.h> | |
a5a19c63 | 45 | #include <asm/pgalloc.h> |
1da177e4 | 46 | #include <asm/sections.h> |
b239fb25 | 47 | #include <asm/paravirt.h> |
1da177e4 LT |
48 | |
49 | unsigned int __VMALLOC_RESERVE = 128 << 20; | |
50 | ||
51 | DEFINE_PER_CPU(struct mmu_gather, mmu_gathers); | |
52 | unsigned long highstart_pfn, highend_pfn; | |
53 | ||
8550eb99 | 54 | static noinline int do_test_wp_bit(void); |
1da177e4 LT |
55 | |
56 | /* | |
57 | * Creates a middle page table and puts a pointer to it in the | |
58 | * given global directory entry. This only returns the gd entry | |
59 | * in non-PAE compilation mode, since the middle layer is folded. | |
60 | */ | |
61 | static pmd_t * __init one_md_table_init(pgd_t *pgd) | |
62 | { | |
63 | pud_t *pud; | |
64 | pmd_t *pmd_table; | |
8550eb99 | 65 | |
1da177e4 | 66 | #ifdef CONFIG_X86_PAE |
b239fb25 JF |
67 | if (!(pgd_val(*pgd) & _PAGE_PRESENT)) { |
68 | pmd_table = (pmd_t *) alloc_bootmem_low_pages(PAGE_SIZE); | |
69 | ||
6c435456 | 70 | paravirt_alloc_pd(&init_mm, __pa(pmd_table) >> PAGE_SHIFT); |
b239fb25 JF |
71 | set_pgd(pgd, __pgd(__pa(pmd_table) | _PAGE_PRESENT)); |
72 | pud = pud_offset(pgd, 0); | |
8550eb99 | 73 | BUG_ON(pmd_table != pmd_offset(pud, 0)); |
b239fb25 JF |
74 | } |
75 | #endif | |
1da177e4 LT |
76 | pud = pud_offset(pgd, 0); |
77 | pmd_table = pmd_offset(pud, 0); | |
8550eb99 | 78 | |
1da177e4 LT |
79 | return pmd_table; |
80 | } | |
81 | ||
82 | /* | |
83 | * Create a page table and place a pointer to it in a middle page | |
8550eb99 | 84 | * directory entry: |
1da177e4 LT |
85 | */ |
86 | static pte_t * __init one_page_table_init(pmd_t *pmd) | |
87 | { | |
b239fb25 | 88 | if (!(pmd_val(*pmd) & _PAGE_PRESENT)) { |
509a80c4 IM |
89 | pte_t *page_table = NULL; |
90 | ||
91 | #ifdef CONFIG_DEBUG_PAGEALLOC | |
92 | page_table = (pte_t *) alloc_bootmem_pages(PAGE_SIZE); | |
93 | #endif | |
8550eb99 | 94 | if (!page_table) { |
509a80c4 IM |
95 | page_table = |
96 | (pte_t *)alloc_bootmem_low_pages(PAGE_SIZE); | |
8550eb99 | 97 | } |
b239fb25 | 98 | |
fdb4c338 | 99 | paravirt_alloc_pt(&init_mm, __pa(page_table) >> PAGE_SHIFT); |
1da177e4 | 100 | set_pmd(pmd, __pmd(__pa(page_table) | _PAGE_TABLE)); |
b239fb25 | 101 | BUG_ON(page_table != pte_offset_kernel(pmd, 0)); |
1da177e4 | 102 | } |
509a80c4 | 103 | |
1da177e4 LT |
104 | return pte_offset_kernel(pmd, 0); |
105 | } | |
106 | ||
107 | /* | |
8550eb99 | 108 | * This function initializes a certain range of kernel virtual memory |
1da177e4 LT |
109 | * with new bootmem page tables, everywhere page tables are missing in |
110 | * the given range. | |
8550eb99 IM |
111 | * |
112 | * NOTE: The pagetables are allocated contiguous on the physical space | |
113 | * so we can cache the place of the first one and move around without | |
1da177e4 LT |
114 | * checking the pgd every time. |
115 | */ | |
8550eb99 IM |
116 | static void __init |
117 | page_table_range_init(unsigned long start, unsigned long end, pgd_t *pgd_base) | |
1da177e4 | 118 | { |
1da177e4 LT |
119 | int pgd_idx, pmd_idx; |
120 | unsigned long vaddr; | |
8550eb99 IM |
121 | pgd_t *pgd; |
122 | pmd_t *pmd; | |
1da177e4 LT |
123 | |
124 | vaddr = start; | |
125 | pgd_idx = pgd_index(vaddr); | |
126 | pmd_idx = pmd_index(vaddr); | |
127 | pgd = pgd_base + pgd_idx; | |
128 | ||
129 | for ( ; (pgd_idx < PTRS_PER_PGD) && (vaddr != end); pgd++, pgd_idx++) { | |
b239fb25 JF |
130 | pmd = one_md_table_init(pgd); |
131 | pmd = pmd + pmd_index(vaddr); | |
8550eb99 IM |
132 | for (; (pmd_idx < PTRS_PER_PMD) && (vaddr != end); |
133 | pmd++, pmd_idx++) { | |
b239fb25 | 134 | one_page_table_init(pmd); |
1da177e4 LT |
135 | |
136 | vaddr += PMD_SIZE; | |
137 | } | |
138 | pmd_idx = 0; | |
139 | } | |
140 | } | |
141 | ||
142 | static inline int is_kernel_text(unsigned long addr) | |
143 | { | |
144 | if (addr >= PAGE_OFFSET && addr <= (unsigned long)__init_end) | |
145 | return 1; | |
146 | return 0; | |
147 | } | |
148 | ||
149 | /* | |
8550eb99 IM |
150 | * This maps the physical memory to kernel virtual address space, a total |
151 | * of max_low_pfn pages, by creating page tables starting from address | |
152 | * PAGE_OFFSET: | |
1da177e4 LT |
153 | */ |
154 | static void __init kernel_physical_mapping_init(pgd_t *pgd_base) | |
155 | { | |
8550eb99 | 156 | int pgd_idx, pmd_idx, pte_ofs; |
1da177e4 LT |
157 | unsigned long pfn; |
158 | pgd_t *pgd; | |
159 | pmd_t *pmd; | |
160 | pte_t *pte; | |
1da177e4 LT |
161 | |
162 | pgd_idx = pgd_index(PAGE_OFFSET); | |
163 | pgd = pgd_base + pgd_idx; | |
164 | pfn = 0; | |
165 | ||
166 | for (; pgd_idx < PTRS_PER_PGD; pgd++, pgd_idx++) { | |
167 | pmd = one_md_table_init(pgd); | |
168 | if (pfn >= max_low_pfn) | |
169 | continue; | |
8550eb99 | 170 | |
f3f20de8 JF |
171 | for (pmd_idx = 0; |
172 | pmd_idx < PTRS_PER_PMD && pfn < max_low_pfn; | |
173 | pmd++, pmd_idx++) { | |
8550eb99 | 174 | unsigned int addr = pfn * PAGE_SIZE + PAGE_OFFSET; |
1da177e4 | 175 | |
8550eb99 IM |
176 | /* |
177 | * Map with big pages if possible, otherwise | |
178 | * create normal page tables: | |
179 | */ | |
1da177e4 | 180 | if (cpu_has_pse) { |
8550eb99 | 181 | unsigned int addr2; |
f3f20de8 JF |
182 | pgprot_t prot = PAGE_KERNEL_LARGE; |
183 | ||
8550eb99 | 184 | addr2 = (pfn + PTRS_PER_PTE-1) * PAGE_SIZE + |
f3f20de8 JF |
185 | PAGE_OFFSET + PAGE_SIZE-1; |
186 | ||
8550eb99 IM |
187 | if (is_kernel_text(addr) || |
188 | is_kernel_text(addr2)) | |
f3f20de8 JF |
189 | prot = PAGE_KERNEL_LARGE_EXEC; |
190 | ||
191 | set_pmd(pmd, pfn_pmd(pfn, prot)); | |
b239fb25 | 192 | |
1da177e4 | 193 | pfn += PTRS_PER_PTE; |
8550eb99 IM |
194 | continue; |
195 | } | |
196 | pte = one_page_table_init(pmd); | |
1da177e4 | 197 | |
8550eb99 IM |
198 | for (pte_ofs = 0; |
199 | pte_ofs < PTRS_PER_PTE && pfn < max_low_pfn; | |
200 | pte++, pfn++, pte_ofs++, addr += PAGE_SIZE) { | |
201 | pgprot_t prot = PAGE_KERNEL; | |
f3f20de8 | 202 | |
8550eb99 IM |
203 | if (is_kernel_text(addr)) |
204 | prot = PAGE_KERNEL_EXEC; | |
f3f20de8 | 205 | |
8550eb99 | 206 | set_pte(pte, pfn_pte(pfn, prot)); |
1da177e4 LT |
207 | } |
208 | } | |
209 | } | |
210 | } | |
211 | ||
212 | static inline int page_kills_ppro(unsigned long pagenr) | |
213 | { | |
214 | if (pagenr >= 0x70000 && pagenr <= 0x7003F) | |
215 | return 1; | |
216 | return 0; | |
217 | } | |
218 | ||
1da177e4 LT |
219 | #ifdef CONFIG_HIGHMEM |
220 | pte_t *kmap_pte; | |
221 | pgprot_t kmap_prot; | |
222 | ||
8550eb99 IM |
223 | static inline pte_t *kmap_get_fixmap_pte(unsigned long vaddr) |
224 | { | |
225 | return pte_offset_kernel(pmd_offset(pud_offset(pgd_offset_k(vaddr), | |
226 | vaddr), vaddr), vaddr); | |
227 | } | |
1da177e4 LT |
228 | |
229 | static void __init kmap_init(void) | |
230 | { | |
231 | unsigned long kmap_vstart; | |
232 | ||
8550eb99 IM |
233 | /* |
234 | * Cache the first kmap pte: | |
235 | */ | |
1da177e4 LT |
236 | kmap_vstart = __fix_to_virt(FIX_KMAP_BEGIN); |
237 | kmap_pte = kmap_get_fixmap_pte(kmap_vstart); | |
238 | ||
239 | kmap_prot = PAGE_KERNEL; | |
240 | } | |
241 | ||
242 | static void __init permanent_kmaps_init(pgd_t *pgd_base) | |
243 | { | |
8550eb99 | 244 | unsigned long vaddr; |
1da177e4 LT |
245 | pgd_t *pgd; |
246 | pud_t *pud; | |
247 | pmd_t *pmd; | |
248 | pte_t *pte; | |
1da177e4 LT |
249 | |
250 | vaddr = PKMAP_BASE; | |
251 | page_table_range_init(vaddr, vaddr + PAGE_SIZE*LAST_PKMAP, pgd_base); | |
252 | ||
253 | pgd = swapper_pg_dir + pgd_index(vaddr); | |
254 | pud = pud_offset(pgd, vaddr); | |
255 | pmd = pmd_offset(pud, vaddr); | |
256 | pte = pte_offset_kernel(pmd, vaddr); | |
8550eb99 | 257 | pkmap_page_table = pte; |
1da177e4 LT |
258 | } |
259 | ||
c09b4240 | 260 | static void __meminit free_new_highpage(struct page *page) |
05039b92 | 261 | { |
7835e98b | 262 | init_page_count(page); |
05039b92 DH |
263 | __free_page(page); |
264 | totalhigh_pages++; | |
265 | } | |
266 | ||
267 | void __init add_one_highpage_init(struct page *page, int pfn, int bad_ppro) | |
1da177e4 LT |
268 | { |
269 | if (page_is_ram(pfn) && !(bad_ppro && page_kills_ppro(pfn))) { | |
270 | ClearPageReserved(page); | |
05039b92 | 271 | free_new_highpage(page); |
1da177e4 LT |
272 | } else |
273 | SetPageReserved(page); | |
274 | } | |
275 | ||
8550eb99 IM |
276 | static int __meminit |
277 | add_one_highpage_hotplug(struct page *page, unsigned long pfn) | |
05039b92 DH |
278 | { |
279 | free_new_highpage(page); | |
280 | totalram_pages++; | |
281 | #ifdef CONFIG_FLATMEM | |
282 | max_mapnr = max(pfn, max_mapnr); | |
283 | #endif | |
284 | num_physpages++; | |
8550eb99 | 285 | |
05039b92 DH |
286 | return 0; |
287 | } | |
288 | ||
289 | /* | |
290 | * Not currently handling the NUMA case. | |
291 | * Assuming single node and all memory that | |
292 | * has been added dynamically that would be | |
8550eb99 | 293 | * onlined here is in HIGHMEM. |
05039b92 | 294 | */ |
0e0be25d | 295 | void __meminit online_page(struct page *page) |
05039b92 DH |
296 | { |
297 | ClearPageReserved(page); | |
298 | add_one_highpage_hotplug(page, page_to_pfn(page)); | |
299 | } | |
300 | ||
8550eb99 | 301 | #ifndef CONFIG_NUMA |
1da177e4 LT |
302 | static void __init set_highmem_pages_init(int bad_ppro) |
303 | { | |
304 | int pfn; | |
8550eb99 | 305 | |
23be8c7d IM |
306 | for (pfn = highstart_pfn; pfn < highend_pfn; pfn++) { |
307 | /* | |
308 | * Holes under sparsemem might not have no mem_map[]: | |
309 | */ | |
310 | if (pfn_valid(pfn)) | |
311 | add_one_highpage_init(pfn_to_page(pfn), pfn, bad_ppro); | |
312 | } | |
1da177e4 LT |
313 | totalram_pages += totalhigh_pages; |
314 | } | |
8550eb99 | 315 | #endif /* !CONFIG_NUMA */ |
1da177e4 LT |
316 | |
317 | #else | |
8550eb99 IM |
318 | # define kmap_init() do { } while (0) |
319 | # define permanent_kmaps_init(pgd_base) do { } while (0) | |
320 | # define set_highmem_pages_init(bad_ppro) do { } while (0) | |
1da177e4 LT |
321 | #endif /* CONFIG_HIGHMEM */ |
322 | ||
c93c82bb | 323 | pteval_t __PAGE_KERNEL = _PAGE_KERNEL; |
129f6946 | 324 | EXPORT_SYMBOL(__PAGE_KERNEL); |
1da177e4 | 325 | |
8550eb99 | 326 | pteval_t __PAGE_KERNEL_EXEC = _PAGE_KERNEL_EXEC; |
1da177e4 | 327 | |
b239fb25 | 328 | void __init native_pagetable_setup_start(pgd_t *base) |
1da177e4 | 329 | { |
1da177e4 LT |
330 | #ifdef CONFIG_X86_PAE |
331 | int i; | |
b239fb25 JF |
332 | |
333 | /* | |
334 | * Init entries of the first-level page table to the | |
335 | * zero page, if they haven't already been set up. | |
336 | * | |
337 | * In a normal native boot, we'll be running on a | |
338 | * pagetable rooted in swapper_pg_dir, but not in PAE | |
339 | * mode, so this will end up clobbering the mappings | |
340 | * for the lower 24Mbytes of the address space, | |
341 | * without affecting the kernel address space. | |
342 | */ | |
343 | for (i = 0; i < USER_PTRS_PER_PGD; i++) | |
344 | set_pgd(&base[i], | |
345 | __pgd(__pa(empty_zero_page) | _PAGE_PRESENT)); | |
346 | ||
347 | /* Make sure kernel address space is empty so that a pagetable | |
348 | will be allocated for it. */ | |
349 | memset(&base[USER_PTRS_PER_PGD], 0, | |
350 | KERNEL_PGD_PTRS * sizeof(pgd_t)); | |
c119ecce | 351 | #else |
6c435456 | 352 | paravirt_alloc_pd(&init_mm, __pa(base) >> PAGE_SHIFT); |
1da177e4 | 353 | #endif |
b239fb25 JF |
354 | } |
355 | ||
356 | void __init native_pagetable_setup_done(pgd_t *base) | |
357 | { | |
358 | #ifdef CONFIG_X86_PAE | |
359 | /* | |
360 | * Add low memory identity-mappings - SMP needs it when | |
361 | * starting up on an AP from real-mode. In the non-PAE | |
362 | * case we already have these mappings through head.S. | |
363 | * All user-space mappings are explicitly cleared after | |
364 | * SMP startup. | |
365 | */ | |
366 | set_pgd(&base[0], base[USER_PTRS_PER_PGD]); | |
367 | #endif | |
368 | } | |
369 | ||
370 | /* | |
371 | * Build a proper pagetable for the kernel mappings. Up until this | |
372 | * point, we've been running on some set of pagetables constructed by | |
373 | * the boot process. | |
374 | * | |
375 | * If we're booting on native hardware, this will be a pagetable | |
376 | * constructed in arch/i386/kernel/head.S, and not running in PAE mode | |
377 | * (even if we'll end up running in PAE). The root of the pagetable | |
378 | * will be swapper_pg_dir. | |
379 | * | |
380 | * If we're booting paravirtualized under a hypervisor, then there are | |
381 | * more options: we may already be running PAE, and the pagetable may | |
382 | * or may not be based in swapper_pg_dir. In any case, | |
383 | * paravirt_pagetable_setup_start() will set up swapper_pg_dir | |
384 | * appropriately for the rest of the initialization to work. | |
385 | * | |
386 | * In general, pagetable_init() assumes that the pagetable may already | |
387 | * be partially populated, and so it avoids stomping on any existing | |
388 | * mappings. | |
389 | */ | |
8550eb99 | 390 | static void __init pagetable_init(void) |
b239fb25 | 391 | { |
b239fb25 | 392 | pgd_t *pgd_base = swapper_pg_dir; |
8550eb99 | 393 | unsigned long vaddr, end; |
b239fb25 JF |
394 | |
395 | paravirt_pagetable_setup_start(pgd_base); | |
1da177e4 LT |
396 | |
397 | /* Enable PSE if available */ | |
b239fb25 | 398 | if (cpu_has_pse) |
1da177e4 | 399 | set_in_cr4(X86_CR4_PSE); |
1da177e4 LT |
400 | |
401 | /* Enable PGE if available */ | |
402 | if (cpu_has_pge) { | |
403 | set_in_cr4(X86_CR4_PGE); | |
404 | __PAGE_KERNEL |= _PAGE_GLOBAL; | |
405 | __PAGE_KERNEL_EXEC |= _PAGE_GLOBAL; | |
406 | } | |
407 | ||
408 | kernel_physical_mapping_init(pgd_base); | |
409 | remap_numa_kva(); | |
410 | ||
411 | /* | |
412 | * Fixed mappings, only the page table structure has to be | |
413 | * created - mappings will be set by set_fixmap(): | |
414 | */ | |
beacfaac | 415 | early_ioremap_clear(); |
1da177e4 | 416 | vaddr = __fix_to_virt(__end_of_fixed_addresses - 1) & PMD_MASK; |
b239fb25 JF |
417 | end = (FIXADDR_TOP + PMD_SIZE - 1) & PMD_MASK; |
418 | page_table_range_init(vaddr, end, pgd_base); | |
beacfaac | 419 | early_ioremap_reset(); |
1da177e4 LT |
420 | |
421 | permanent_kmaps_init(pgd_base); | |
422 | ||
b239fb25 | 423 | paravirt_pagetable_setup_done(pgd_base); |
1da177e4 LT |
424 | } |
425 | ||
b0cb1a19 | 426 | #if defined(CONFIG_HIBERNATION) || defined(CONFIG_ACPI) |
1da177e4 LT |
427 | /* |
428 | * Swap suspend & friends need this for resume because things like the intel-agp | |
429 | * driver might have split up a kernel 4MB mapping. | |
430 | */ | |
431 | char __nosavedata swsusp_pg_dir[PAGE_SIZE] | |
8550eb99 | 432 | __attribute__ ((aligned(PAGE_SIZE))); |
1da177e4 LT |
433 | |
434 | static inline void save_pg_dir(void) | |
435 | { | |
436 | memcpy(swsusp_pg_dir, swapper_pg_dir, PAGE_SIZE); | |
437 | } | |
438 | #else | |
439 | static inline void save_pg_dir(void) | |
440 | { | |
441 | } | |
442 | #endif | |
443 | ||
8550eb99 | 444 | void zap_low_mappings(void) |
1da177e4 LT |
445 | { |
446 | int i; | |
447 | ||
448 | save_pg_dir(); | |
449 | ||
450 | /* | |
451 | * Zap initial low-memory mappings. | |
452 | * | |
453 | * Note that "pgd_clear()" doesn't do it for | |
454 | * us, because pgd_clear() is a no-op on i386. | |
455 | */ | |
8550eb99 | 456 | for (i = 0; i < USER_PTRS_PER_PGD; i++) { |
1da177e4 LT |
457 | #ifdef CONFIG_X86_PAE |
458 | set_pgd(swapper_pg_dir+i, __pgd(1 + __pa(empty_zero_page))); | |
459 | #else | |
460 | set_pgd(swapper_pg_dir+i, __pgd(0)); | |
461 | #endif | |
8550eb99 | 462 | } |
1da177e4 LT |
463 | flush_tlb_all(); |
464 | } | |
465 | ||
8550eb99 | 466 | int nx_enabled; |
d5321abe | 467 | |
6fdc05d4 JF |
468 | pteval_t __supported_pte_mask __read_mostly = ~_PAGE_NX; |
469 | EXPORT_SYMBOL_GPL(__supported_pte_mask); | |
470 | ||
d5321abe JB |
471 | #ifdef CONFIG_X86_PAE |
472 | ||
8550eb99 | 473 | static int disable_nx __initdata; |
1da177e4 LT |
474 | |
475 | /* | |
476 | * noexec = on|off | |
477 | * | |
478 | * Control non executable mappings. | |
479 | * | |
480 | * on Enable | |
481 | * off Disable | |
482 | */ | |
1a3f239d | 483 | static int __init noexec_setup(char *str) |
1da177e4 | 484 | { |
1a3f239d RR |
485 | if (!str || !strcmp(str, "on")) { |
486 | if (cpu_has_nx) { | |
487 | __supported_pte_mask |= _PAGE_NX; | |
488 | disable_nx = 0; | |
489 | } | |
8550eb99 IM |
490 | } else { |
491 | if (!strcmp(str, "off")) { | |
492 | disable_nx = 1; | |
493 | __supported_pte_mask &= ~_PAGE_NX; | |
494 | } else { | |
495 | return -EINVAL; | |
496 | } | |
497 | } | |
1a3f239d RR |
498 | |
499 | return 0; | |
1da177e4 | 500 | } |
1a3f239d | 501 | early_param("noexec", noexec_setup); |
1da177e4 | 502 | |
1da177e4 LT |
503 | static void __init set_nx(void) |
504 | { | |
505 | unsigned int v[4], l, h; | |
506 | ||
507 | if (cpu_has_pae && (cpuid_eax(0x80000000) > 0x80000001)) { | |
508 | cpuid(0x80000001, &v[0], &v[1], &v[2], &v[3]); | |
8550eb99 | 509 | |
1da177e4 LT |
510 | if ((v[3] & (1 << 20)) && !disable_nx) { |
511 | rdmsr(MSR_EFER, l, h); | |
512 | l |= EFER_NX; | |
513 | wrmsr(MSR_EFER, l, h); | |
514 | nx_enabled = 1; | |
515 | __supported_pte_mask |= _PAGE_NX; | |
516 | } | |
517 | } | |
518 | } | |
1da177e4 LT |
519 | #endif |
520 | ||
521 | /* | |
522 | * paging_init() sets up the page tables - note that the first 8MB are | |
523 | * already mapped by head.S. | |
524 | * | |
525 | * This routines also unmaps the page at virtual kernel address 0, so | |
526 | * that we can trap those pesky NULL-reference errors in the kernel. | |
527 | */ | |
528 | void __init paging_init(void) | |
529 | { | |
530 | #ifdef CONFIG_X86_PAE | |
531 | set_nx(); | |
532 | if (nx_enabled) | |
533 | printk("NX (Execute Disable) protection: active\n"); | |
534 | #endif | |
1da177e4 LT |
535 | pagetable_init(); |
536 | ||
537 | load_cr3(swapper_pg_dir); | |
538 | ||
539 | #ifdef CONFIG_X86_PAE | |
540 | /* | |
541 | * We will bail out later - printk doesn't work right now so | |
542 | * the user would just see a hanging kernel. | |
543 | */ | |
544 | if (cpu_has_pae) | |
545 | set_in_cr4(X86_CR4_PAE); | |
546 | #endif | |
547 | __flush_tlb_all(); | |
548 | ||
549 | kmap_init(); | |
550 | } | |
551 | ||
552 | /* | |
553 | * Test if the WP bit works in supervisor mode. It isn't supported on 386's | |
554 | * and also on some strange 486's (NexGen etc.). All 586+'s are OK. This | |
555 | * used to involve black magic jumps to work around some nasty CPU bugs, | |
556 | * but fortunately the switch to using exceptions got rid of all that. | |
557 | */ | |
1da177e4 LT |
558 | static void __init test_wp_bit(void) |
559 | { | |
560 | printk("Checking if this processor honours the WP bit even in supervisor mode... "); | |
561 | ||
562 | /* Any page-aligned address will do, the test is non-destructive */ | |
563 | __set_fixmap(FIX_WP_TEST, __pa(&swapper_pg_dir), PAGE_READONLY); | |
564 | boot_cpu_data.wp_works_ok = do_test_wp_bit(); | |
565 | clear_fixmap(FIX_WP_TEST); | |
566 | ||
567 | if (!boot_cpu_data.wp_works_ok) { | |
568 | printk("No.\n"); | |
569 | #ifdef CONFIG_X86_WP_WORKS_OK | |
570 | panic("This kernel doesn't support CPU's with broken WP. Recompile it for a 386!"); | |
571 | #endif | |
572 | } else { | |
573 | printk("Ok.\n"); | |
574 | } | |
575 | } | |
576 | ||
8550eb99 | 577 | static struct kcore_list kcore_mem, kcore_vmalloc; |
1da177e4 LT |
578 | |
579 | void __init mem_init(void) | |
580 | { | |
1da177e4 | 581 | int codesize, reservedpages, datasize, initsize; |
8550eb99 | 582 | int tmp, bad_ppro; |
1da177e4 | 583 | |
05b79bdc | 584 | #ifdef CONFIG_FLATMEM |
8d8f3cbe | 585 | BUG_ON(!mem_map); |
1da177e4 | 586 | #endif |
1da177e4 LT |
587 | bad_ppro = ppro_with_ram_bug(); |
588 | ||
589 | #ifdef CONFIG_HIGHMEM | |
590 | /* check that fixmap and pkmap do not overlap */ | |
591 | if (PKMAP_BASE+LAST_PKMAP*PAGE_SIZE >= FIXADDR_START) { | |
592 | printk(KERN_ERR "fixmap and kmap areas overlap - this will crash\n"); | |
593 | printk(KERN_ERR "pkstart: %lxh pkend: %lxh fixstart %lxh\n", | |
594 | PKMAP_BASE, PKMAP_BASE+LAST_PKMAP*PAGE_SIZE, FIXADDR_START); | |
595 | BUG(); | |
596 | } | |
597 | #endif | |
1da177e4 LT |
598 | /* this will put all low memory onto the freelists */ |
599 | totalram_pages += free_all_bootmem(); | |
600 | ||
601 | reservedpages = 0; | |
602 | for (tmp = 0; tmp < max_low_pfn; tmp++) | |
603 | /* | |
8550eb99 | 604 | * Only count reserved RAM pages: |
1da177e4 LT |
605 | */ |
606 | if (page_is_ram(tmp) && PageReserved(pfn_to_page(tmp))) | |
607 | reservedpages++; | |
608 | ||
609 | set_highmem_pages_init(bad_ppro); | |
610 | ||
611 | codesize = (unsigned long) &_etext - (unsigned long) &_text; | |
612 | datasize = (unsigned long) &_edata - (unsigned long) &_etext; | |
613 | initsize = (unsigned long) &__init_end - (unsigned long) &__init_begin; | |
614 | ||
8550eb99 IM |
615 | kclist_add(&kcore_mem, __va(0), max_low_pfn << PAGE_SHIFT); |
616 | kclist_add(&kcore_vmalloc, (void *)VMALLOC_START, | |
1da177e4 LT |
617 | VMALLOC_END-VMALLOC_START); |
618 | ||
8550eb99 IM |
619 | printk(KERN_INFO "Memory: %luk/%luk available (%dk kernel code, " |
620 | "%dk reserved, %dk data, %dk init, %ldk highmem)\n", | |
1da177e4 LT |
621 | (unsigned long) nr_free_pages() << (PAGE_SHIFT-10), |
622 | num_physpages << (PAGE_SHIFT-10), | |
623 | codesize >> 10, | |
624 | reservedpages << (PAGE_SHIFT-10), | |
625 | datasize >> 10, | |
626 | initsize >> 10, | |
627 | (unsigned long) (totalhigh_pages << (PAGE_SHIFT-10)) | |
628 | ); | |
629 | ||
052e7994 JF |
630 | #if 1 /* double-sanity-check paranoia */ |
631 | printk("virtual kernel memory layout:\n" | |
8550eb99 | 632 | " fixmap : 0x%08lx - 0x%08lx (%4ld kB)\n" |
052e7994 | 633 | #ifdef CONFIG_HIGHMEM |
8550eb99 | 634 | " pkmap : 0x%08lx - 0x%08lx (%4ld kB)\n" |
052e7994 | 635 | #endif |
8550eb99 IM |
636 | " vmalloc : 0x%08lx - 0x%08lx (%4ld MB)\n" |
637 | " lowmem : 0x%08lx - 0x%08lx (%4ld MB)\n" | |
638 | " .init : 0x%08lx - 0x%08lx (%4ld kB)\n" | |
639 | " .data : 0x%08lx - 0x%08lx (%4ld kB)\n" | |
640 | " .text : 0x%08lx - 0x%08lx (%4ld kB)\n", | |
641 | FIXADDR_START, FIXADDR_TOP, | |
642 | (FIXADDR_TOP - FIXADDR_START) >> 10, | |
052e7994 JF |
643 | |
644 | #ifdef CONFIG_HIGHMEM | |
8550eb99 IM |
645 | PKMAP_BASE, PKMAP_BASE+LAST_PKMAP*PAGE_SIZE, |
646 | (LAST_PKMAP*PAGE_SIZE) >> 10, | |
052e7994 JF |
647 | #endif |
648 | ||
8550eb99 IM |
649 | VMALLOC_START, VMALLOC_END, |
650 | (VMALLOC_END - VMALLOC_START) >> 20, | |
052e7994 | 651 | |
8550eb99 IM |
652 | (unsigned long)__va(0), (unsigned long)high_memory, |
653 | ((unsigned long)high_memory - (unsigned long)__va(0)) >> 20, | |
052e7994 | 654 | |
8550eb99 IM |
655 | (unsigned long)&__init_begin, (unsigned long)&__init_end, |
656 | ((unsigned long)&__init_end - | |
657 | (unsigned long)&__init_begin) >> 10, | |
052e7994 | 658 | |
8550eb99 IM |
659 | (unsigned long)&_etext, (unsigned long)&_edata, |
660 | ((unsigned long)&_edata - (unsigned long)&_etext) >> 10, | |
052e7994 | 661 | |
8550eb99 IM |
662 | (unsigned long)&_text, (unsigned long)&_etext, |
663 | ((unsigned long)&_etext - (unsigned long)&_text) >> 10); | |
052e7994 JF |
664 | |
665 | #ifdef CONFIG_HIGHMEM | |
8550eb99 IM |
666 | BUG_ON(PKMAP_BASE + LAST_PKMAP*PAGE_SIZE > FIXADDR_START); |
667 | BUG_ON(VMALLOC_END > PKMAP_BASE); | |
052e7994 | 668 | #endif |
8550eb99 IM |
669 | BUG_ON(VMALLOC_START > VMALLOC_END); |
670 | BUG_ON((unsigned long)high_memory > VMALLOC_START); | |
052e7994 JF |
671 | #endif /* double-sanity-check paranoia */ |
672 | ||
1da177e4 LT |
673 | #ifdef CONFIG_X86_PAE |
674 | if (!cpu_has_pae) | |
675 | panic("cannot execute a PAE-enabled kernel on a PAE-less CPU!"); | |
676 | #endif | |
677 | if (boot_cpu_data.wp_works_ok < 0) | |
678 | test_wp_bit(); | |
679 | ||
680 | /* | |
681 | * Subtle. SMP is doing it's boot stuff late (because it has to | |
682 | * fork idle threads) - but it also needs low mappings for the | |
683 | * protected-mode entry to work. We zap these entries only after | |
684 | * the WP-bit has been tested. | |
685 | */ | |
686 | #ifndef CONFIG_SMP | |
687 | zap_low_mappings(); | |
688 | #endif | |
689 | } | |
690 | ||
ad8f5797 | 691 | #ifdef CONFIG_MEMORY_HOTPLUG |
bc02af93 | 692 | int arch_add_memory(int nid, u64 start, u64 size) |
05039b92 | 693 | { |
7c7e9425 | 694 | struct pglist_data *pgdata = NODE_DATA(nid); |
776ed98b | 695 | struct zone *zone = pgdata->node_zones + ZONE_HIGHMEM; |
05039b92 DH |
696 | unsigned long start_pfn = start >> PAGE_SHIFT; |
697 | unsigned long nr_pages = size >> PAGE_SHIFT; | |
698 | ||
699 | return __add_pages(zone, start_pfn, nr_pages); | |
700 | } | |
9d99aaa3 | 701 | #endif |
05039b92 | 702 | |
e18b890b | 703 | struct kmem_cache *pmd_cache; |
1da177e4 LT |
704 | |
705 | void __init pgtable_cache_init(void) | |
706 | { | |
8550eb99 | 707 | if (PTRS_PER_PMD > 1) { |
1da177e4 | 708 | pmd_cache = kmem_cache_create("pmd", |
4f817847 JF |
709 | PTRS_PER_PMD*sizeof(pmd_t), |
710 | PTRS_PER_PMD*sizeof(pmd_t), | |
711 | SLAB_PANIC, | |
712 | pmd_ctor); | |
8550eb99 | 713 | } |
1da177e4 LT |
714 | } |
715 | ||
716 | /* | |
717 | * This function cannot be __init, since exceptions don't work in that | |
718 | * section. Put this after the callers, so that it cannot be inlined. | |
719 | */ | |
8550eb99 | 720 | static noinline int do_test_wp_bit(void) |
1da177e4 LT |
721 | { |
722 | char tmp_reg; | |
723 | int flag; | |
724 | ||
725 | __asm__ __volatile__( | |
8550eb99 IM |
726 | " movb %0, %1 \n" |
727 | "1: movb %1, %0 \n" | |
728 | " xorl %2, %2 \n" | |
1da177e4 | 729 | "2: \n" |
8550eb99 | 730 | ".section __ex_table, \"a\"\n" |
1da177e4 | 731 | " .align 4 \n" |
8550eb99 | 732 | " .long 1b, 2b \n" |
1da177e4 LT |
733 | ".previous \n" |
734 | :"=m" (*(char *)fix_to_virt(FIX_WP_TEST)), | |
735 | "=q" (tmp_reg), | |
736 | "=r" (flag) | |
737 | :"2" (1) | |
738 | :"memory"); | |
8550eb99 | 739 | |
1da177e4 LT |
740 | return flag; |
741 | } | |
742 | ||
63aaf308 | 743 | #ifdef CONFIG_DEBUG_RODATA |
edeed305 AV |
744 | const int rodata_test_data = 0xC3; |
745 | EXPORT_SYMBOL_GPL(rodata_test_data); | |
63aaf308 | 746 | |
63aaf308 AV |
747 | void mark_rodata_ro(void) |
748 | { | |
6fb14755 JB |
749 | unsigned long start = PFN_ALIGN(_text); |
750 | unsigned long size = PFN_ALIGN(_etext) - start; | |
63aaf308 | 751 | |
602033ed LT |
752 | #ifndef CONFIG_KPROBES |
753 | #ifdef CONFIG_HOTPLUG_CPU | |
754 | /* It must still be possible to apply SMP alternatives. */ | |
755 | if (num_possible_cpus() <= 1) | |
756 | #endif | |
757 | { | |
6d238cc4 | 758 | set_pages_ro(virt_to_page(start), size >> PAGE_SHIFT); |
602033ed | 759 | printk("Write protecting the kernel text: %luk\n", size >> 10); |
0c42f392 AK |
760 | |
761 | #ifdef CONFIG_CPA_DEBUG | |
0c42f392 | 762 | printk("Testing CPA: Reverting %lx-%lx\n", start, start+size); |
6d238cc4 | 763 | set_pages_rw(virt_to_page(start), size>>PAGE_SHIFT); |
0c42f392 AK |
764 | |
765 | printk("Testing CPA: write protecting again\n"); | |
6d238cc4 | 766 | set_pages_ro(virt_to_page(start), size>>PAGE_SHIFT); |
0c42f392 | 767 | #endif |
602033ed LT |
768 | } |
769 | #endif | |
6fb14755 JB |
770 | start += size; |
771 | size = (unsigned long)__end_rodata - start; | |
6d238cc4 | 772 | set_pages_ro(virt_to_page(start), size >> PAGE_SHIFT); |
6fb14755 JB |
773 | printk("Write protecting the kernel read-only data: %luk\n", |
774 | size >> 10); | |
edeed305 | 775 | rodata_test(); |
63aaf308 | 776 | |
0c42f392 AK |
777 | #ifdef CONFIG_CPA_DEBUG |
778 | printk("Testing CPA: undo %lx-%lx\n", start, start + size); | |
6d238cc4 | 779 | set_pages_rw(virt_to_page(start), size >> PAGE_SHIFT); |
0c42f392 AK |
780 | |
781 | printk("Testing CPA: write protecting again\n"); | |
6d238cc4 | 782 | set_pages_ro(virt_to_page(start), size >> PAGE_SHIFT); |
0c42f392 | 783 | #endif |
63aaf308 AV |
784 | } |
785 | #endif | |
786 | ||
9a0b5817 GH |
787 | void free_init_pages(char *what, unsigned long begin, unsigned long end) |
788 | { | |
ee01f112 IM |
789 | #ifdef CONFIG_DEBUG_PAGEALLOC |
790 | /* | |
791 | * If debugging page accesses then do not free this memory but | |
792 | * mark them not present - any buggy init-section access will | |
793 | * create a kernel page fault: | |
794 | */ | |
795 | printk(KERN_INFO "debug: unmapping init memory %08lx..%08lx\n", | |
796 | begin, PAGE_ALIGN(end)); | |
797 | set_memory_np(begin, (end - begin) >> PAGE_SHIFT); | |
798 | #else | |
86f03989 IM |
799 | unsigned long addr; |
800 | ||
3c1df68b AV |
801 | /* |
802 | * We just marked the kernel text read only above, now that | |
803 | * we are going to free part of that, we need to make that | |
804 | * writeable first. | |
805 | */ | |
806 | set_memory_rw(begin, (end - begin) >> PAGE_SHIFT); | |
807 | ||
9a0b5817 | 808 | for (addr = begin; addr < end; addr += PAGE_SIZE) { |
e3ebadd9 LT |
809 | ClearPageReserved(virt_to_page(addr)); |
810 | init_page_count(virt_to_page(addr)); | |
811 | memset((void *)addr, POISON_FREE_INITMEM, PAGE_SIZE); | |
812 | free_page(addr); | |
9a0b5817 GH |
813 | totalram_pages++; |
814 | } | |
6fb14755 | 815 | printk(KERN_INFO "Freeing %s: %luk freed\n", what, (end - begin) >> 10); |
ee01f112 | 816 | #endif |
9a0b5817 GH |
817 | } |
818 | ||
819 | void free_initmem(void) | |
820 | { | |
821 | free_init_pages("unused kernel memory", | |
e3ebadd9 LT |
822 | (unsigned long)(&__init_begin), |
823 | (unsigned long)(&__init_end)); | |
9a0b5817 | 824 | } |
63aaf308 | 825 | |
1da177e4 LT |
826 | #ifdef CONFIG_BLK_DEV_INITRD |
827 | void free_initrd_mem(unsigned long start, unsigned long end) | |
828 | { | |
e3ebadd9 | 829 | free_init_pages("initrd memory", start, end); |
1da177e4 LT |
830 | } |
831 | #endif |