]>
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 | ||
9 | #include <linux/config.h> | |
10 | #include <linux/module.h> | |
11 | #include <linux/signal.h> | |
12 | #include <linux/sched.h> | |
13 | #include <linux/kernel.h> | |
14 | #include <linux/errno.h> | |
15 | #include <linux/string.h> | |
16 | #include <linux/types.h> | |
17 | #include <linux/ptrace.h> | |
18 | #include <linux/mman.h> | |
19 | #include <linux/mm.h> | |
20 | #include <linux/hugetlb.h> | |
21 | #include <linux/swap.h> | |
22 | #include <linux/smp.h> | |
23 | #include <linux/init.h> | |
24 | #include <linux/highmem.h> | |
25 | #include <linux/pagemap.h> | |
26 | #include <linux/bootmem.h> | |
27 | #include <linux/slab.h> | |
28 | #include <linux/proc_fs.h> | |
29 | #include <linux/efi.h> | |
30 | ||
31 | #include <asm/processor.h> | |
32 | #include <asm/system.h> | |
33 | #include <asm/uaccess.h> | |
34 | #include <asm/pgtable.h> | |
35 | #include <asm/dma.h> | |
36 | #include <asm/fixmap.h> | |
37 | #include <asm/e820.h> | |
38 | #include <asm/apic.h> | |
39 | #include <asm/tlb.h> | |
40 | #include <asm/tlbflush.h> | |
41 | #include <asm/sections.h> | |
42 | ||
43 | unsigned int __VMALLOC_RESERVE = 128 << 20; | |
44 | ||
45 | DEFINE_PER_CPU(struct mmu_gather, mmu_gathers); | |
46 | unsigned long highstart_pfn, highend_pfn; | |
47 | ||
48 | static int noinline do_test_wp_bit(void); | |
49 | ||
50 | /* | |
51 | * Creates a middle page table and puts a pointer to it in the | |
52 | * given global directory entry. This only returns the gd entry | |
53 | * in non-PAE compilation mode, since the middle layer is folded. | |
54 | */ | |
55 | static pmd_t * __init one_md_table_init(pgd_t *pgd) | |
56 | { | |
57 | pud_t *pud; | |
58 | pmd_t *pmd_table; | |
59 | ||
60 | #ifdef CONFIG_X86_PAE | |
61 | pmd_table = (pmd_t *) alloc_bootmem_low_pages(PAGE_SIZE); | |
62 | set_pgd(pgd, __pgd(__pa(pmd_table) | _PAGE_PRESENT)); | |
63 | pud = pud_offset(pgd, 0); | |
64 | if (pmd_table != pmd_offset(pud, 0)) | |
65 | BUG(); | |
66 | #else | |
67 | pud = pud_offset(pgd, 0); | |
68 | pmd_table = pmd_offset(pud, 0); | |
69 | #endif | |
70 | ||
71 | return pmd_table; | |
72 | } | |
73 | ||
74 | /* | |
75 | * Create a page table and place a pointer to it in a middle page | |
76 | * directory entry. | |
77 | */ | |
78 | static pte_t * __init one_page_table_init(pmd_t *pmd) | |
79 | { | |
80 | if (pmd_none(*pmd)) { | |
81 | pte_t *page_table = (pte_t *) alloc_bootmem_low_pages(PAGE_SIZE); | |
82 | set_pmd(pmd, __pmd(__pa(page_table) | _PAGE_TABLE)); | |
83 | if (page_table != pte_offset_kernel(pmd, 0)) | |
84 | BUG(); | |
85 | ||
86 | return page_table; | |
87 | } | |
88 | ||
89 | return pte_offset_kernel(pmd, 0); | |
90 | } | |
91 | ||
92 | /* | |
93 | * This function initializes a certain range of kernel virtual memory | |
94 | * with new bootmem page tables, everywhere page tables are missing in | |
95 | * the given range. | |
96 | */ | |
97 | ||
98 | /* | |
99 | * NOTE: The pagetables are allocated contiguous on the physical space | |
100 | * so we can cache the place of the first one and move around without | |
101 | * checking the pgd every time. | |
102 | */ | |
103 | static void __init page_table_range_init (unsigned long start, unsigned long end, pgd_t *pgd_base) | |
104 | { | |
105 | pgd_t *pgd; | |
106 | pud_t *pud; | |
107 | pmd_t *pmd; | |
108 | int pgd_idx, pmd_idx; | |
109 | unsigned long vaddr; | |
110 | ||
111 | vaddr = start; | |
112 | pgd_idx = pgd_index(vaddr); | |
113 | pmd_idx = pmd_index(vaddr); | |
114 | pgd = pgd_base + pgd_idx; | |
115 | ||
116 | for ( ; (pgd_idx < PTRS_PER_PGD) && (vaddr != end); pgd++, pgd_idx++) { | |
117 | if (pgd_none(*pgd)) | |
118 | one_md_table_init(pgd); | |
119 | pud = pud_offset(pgd, vaddr); | |
120 | pmd = pmd_offset(pud, vaddr); | |
121 | for (; (pmd_idx < PTRS_PER_PMD) && (vaddr != end); pmd++, pmd_idx++) { | |
122 | if (pmd_none(*pmd)) | |
123 | one_page_table_init(pmd); | |
124 | ||
125 | vaddr += PMD_SIZE; | |
126 | } | |
127 | pmd_idx = 0; | |
128 | } | |
129 | } | |
130 | ||
131 | static inline int is_kernel_text(unsigned long addr) | |
132 | { | |
133 | if (addr >= PAGE_OFFSET && addr <= (unsigned long)__init_end) | |
134 | return 1; | |
135 | return 0; | |
136 | } | |
137 | ||
138 | /* | |
139 | * This maps the physical memory to kernel virtual address space, a total | |
140 | * of max_low_pfn pages, by creating page tables starting from address | |
141 | * PAGE_OFFSET. | |
142 | */ | |
143 | static void __init kernel_physical_mapping_init(pgd_t *pgd_base) | |
144 | { | |
145 | unsigned long pfn; | |
146 | pgd_t *pgd; | |
147 | pmd_t *pmd; | |
148 | pte_t *pte; | |
149 | int pgd_idx, pmd_idx, pte_ofs; | |
150 | ||
151 | pgd_idx = pgd_index(PAGE_OFFSET); | |
152 | pgd = pgd_base + pgd_idx; | |
153 | pfn = 0; | |
154 | ||
155 | for (; pgd_idx < PTRS_PER_PGD; pgd++, pgd_idx++) { | |
156 | pmd = one_md_table_init(pgd); | |
157 | if (pfn >= max_low_pfn) | |
158 | continue; | |
159 | for (pmd_idx = 0; pmd_idx < PTRS_PER_PMD && pfn < max_low_pfn; pmd++, pmd_idx++) { | |
160 | unsigned int address = pfn * PAGE_SIZE + PAGE_OFFSET; | |
161 | ||
162 | /* Map with big pages if possible, otherwise create normal page tables. */ | |
163 | if (cpu_has_pse) { | |
164 | unsigned int address2 = (pfn + PTRS_PER_PTE - 1) * PAGE_SIZE + PAGE_OFFSET + PAGE_SIZE-1; | |
165 | ||
166 | if (is_kernel_text(address) || is_kernel_text(address2)) | |
167 | set_pmd(pmd, pfn_pmd(pfn, PAGE_KERNEL_LARGE_EXEC)); | |
168 | else | |
169 | set_pmd(pmd, pfn_pmd(pfn, PAGE_KERNEL_LARGE)); | |
170 | pfn += PTRS_PER_PTE; | |
171 | } else { | |
172 | pte = one_page_table_init(pmd); | |
173 | ||
174 | for (pte_ofs = 0; pte_ofs < PTRS_PER_PTE && pfn < max_low_pfn; pte++, pfn++, pte_ofs++) { | |
175 | if (is_kernel_text(address)) | |
176 | set_pte(pte, pfn_pte(pfn, PAGE_KERNEL_EXEC)); | |
177 | else | |
178 | set_pte(pte, pfn_pte(pfn, PAGE_KERNEL)); | |
179 | } | |
180 | } | |
181 | } | |
182 | } | |
183 | } | |
184 | ||
185 | static inline int page_kills_ppro(unsigned long pagenr) | |
186 | { | |
187 | if (pagenr >= 0x70000 && pagenr <= 0x7003F) | |
188 | return 1; | |
189 | return 0; | |
190 | } | |
191 | ||
192 | extern int is_available_memory(efi_memory_desc_t *); | |
193 | ||
5b505b90 | 194 | int page_is_ram(unsigned long pagenr) |
1da177e4 LT |
195 | { |
196 | int i; | |
197 | unsigned long addr, end; | |
198 | ||
199 | if (efi_enabled) { | |
200 | efi_memory_desc_t *md; | |
7ae65fd3 | 201 | void *p; |
1da177e4 | 202 | |
7ae65fd3 MT |
203 | for (p = memmap.map; p < memmap.map_end; p += memmap.desc_size) { |
204 | md = p; | |
1da177e4 LT |
205 | if (!is_available_memory(md)) |
206 | continue; | |
207 | addr = (md->phys_addr+PAGE_SIZE-1) >> PAGE_SHIFT; | |
208 | end = (md->phys_addr + (md->num_pages << EFI_PAGE_SHIFT)) >> PAGE_SHIFT; | |
209 | ||
210 | if ((pagenr >= addr) && (pagenr < end)) | |
211 | return 1; | |
212 | } | |
213 | return 0; | |
214 | } | |
215 | ||
216 | for (i = 0; i < e820.nr_map; i++) { | |
217 | ||
218 | if (e820.map[i].type != E820_RAM) /* not usable memory */ | |
219 | continue; | |
220 | /* | |
221 | * !!!FIXME!!! Some BIOSen report areas as RAM that | |
222 | * are not. Notably the 640->1Mb area. We need a sanity | |
223 | * check here. | |
224 | */ | |
225 | addr = (e820.map[i].addr+PAGE_SIZE-1) >> PAGE_SHIFT; | |
226 | end = (e820.map[i].addr+e820.map[i].size) >> PAGE_SHIFT; | |
227 | if ((pagenr >= addr) && (pagenr < end)) | |
228 | return 1; | |
229 | } | |
230 | return 0; | |
231 | } | |
232 | ||
233 | #ifdef CONFIG_HIGHMEM | |
234 | pte_t *kmap_pte; | |
235 | pgprot_t kmap_prot; | |
236 | ||
237 | #define kmap_get_fixmap_pte(vaddr) \ | |
238 | pte_offset_kernel(pmd_offset(pud_offset(pgd_offset_k(vaddr), vaddr), (vaddr)), (vaddr)) | |
239 | ||
240 | static void __init kmap_init(void) | |
241 | { | |
242 | unsigned long kmap_vstart; | |
243 | ||
244 | /* cache the first kmap pte */ | |
245 | kmap_vstart = __fix_to_virt(FIX_KMAP_BEGIN); | |
246 | kmap_pte = kmap_get_fixmap_pte(kmap_vstart); | |
247 | ||
248 | kmap_prot = PAGE_KERNEL; | |
249 | } | |
250 | ||
251 | static void __init permanent_kmaps_init(pgd_t *pgd_base) | |
252 | { | |
253 | pgd_t *pgd; | |
254 | pud_t *pud; | |
255 | pmd_t *pmd; | |
256 | pte_t *pte; | |
257 | unsigned long vaddr; | |
258 | ||
259 | vaddr = PKMAP_BASE; | |
260 | page_table_range_init(vaddr, vaddr + PAGE_SIZE*LAST_PKMAP, pgd_base); | |
261 | ||
262 | pgd = swapper_pg_dir + pgd_index(vaddr); | |
263 | pud = pud_offset(pgd, vaddr); | |
264 | pmd = pmd_offset(pud, vaddr); | |
265 | pte = pte_offset_kernel(pmd, vaddr); | |
266 | pkmap_page_table = pte; | |
267 | } | |
268 | ||
269 | void __init one_highpage_init(struct page *page, int pfn, int bad_ppro) | |
270 | { | |
271 | if (page_is_ram(pfn) && !(bad_ppro && page_kills_ppro(pfn))) { | |
272 | ClearPageReserved(page); | |
1da177e4 LT |
273 | set_page_count(page, 1); |
274 | __free_page(page); | |
275 | totalhigh_pages++; | |
276 | } else | |
277 | SetPageReserved(page); | |
278 | } | |
279 | ||
05b79bdc AW |
280 | #ifdef CONFIG_NUMA |
281 | extern void set_highmem_pages_init(int); | |
282 | #else | |
1da177e4 LT |
283 | static void __init set_highmem_pages_init(int bad_ppro) |
284 | { | |
285 | int pfn; | |
286 | for (pfn = highstart_pfn; pfn < highend_pfn; pfn++) | |
287 | one_highpage_init(pfn_to_page(pfn), pfn, bad_ppro); | |
288 | totalram_pages += totalhigh_pages; | |
289 | } | |
05b79bdc | 290 | #endif /* CONFIG_FLATMEM */ |
1da177e4 LT |
291 | |
292 | #else | |
293 | #define kmap_init() do { } while (0) | |
294 | #define permanent_kmaps_init(pgd_base) do { } while (0) | |
295 | #define set_highmem_pages_init(bad_ppro) do { } while (0) | |
296 | #endif /* CONFIG_HIGHMEM */ | |
297 | ||
298 | unsigned long long __PAGE_KERNEL = _PAGE_KERNEL; | |
129f6946 | 299 | EXPORT_SYMBOL(__PAGE_KERNEL); |
1da177e4 LT |
300 | unsigned long long __PAGE_KERNEL_EXEC = _PAGE_KERNEL_EXEC; |
301 | ||
05b79bdc | 302 | #ifdef CONFIG_NUMA |
1da177e4 | 303 | extern void __init remap_numa_kva(void); |
05b79bdc AW |
304 | #else |
305 | #define remap_numa_kva() do {} while (0) | |
1da177e4 LT |
306 | #endif |
307 | ||
308 | static void __init pagetable_init (void) | |
309 | { | |
310 | unsigned long vaddr; | |
311 | pgd_t *pgd_base = swapper_pg_dir; | |
312 | ||
313 | #ifdef CONFIG_X86_PAE | |
314 | int i; | |
315 | /* Init entries of the first-level page table to the zero page */ | |
316 | for (i = 0; i < PTRS_PER_PGD; i++) | |
317 | set_pgd(pgd_base + i, __pgd(__pa(empty_zero_page) | _PAGE_PRESENT)); | |
318 | #endif | |
319 | ||
320 | /* Enable PSE if available */ | |
321 | if (cpu_has_pse) { | |
322 | set_in_cr4(X86_CR4_PSE); | |
323 | } | |
324 | ||
325 | /* Enable PGE if available */ | |
326 | if (cpu_has_pge) { | |
327 | set_in_cr4(X86_CR4_PGE); | |
328 | __PAGE_KERNEL |= _PAGE_GLOBAL; | |
329 | __PAGE_KERNEL_EXEC |= _PAGE_GLOBAL; | |
330 | } | |
331 | ||
332 | kernel_physical_mapping_init(pgd_base); | |
333 | remap_numa_kva(); | |
334 | ||
335 | /* | |
336 | * Fixed mappings, only the page table structure has to be | |
337 | * created - mappings will be set by set_fixmap(): | |
338 | */ | |
339 | vaddr = __fix_to_virt(__end_of_fixed_addresses - 1) & PMD_MASK; | |
340 | page_table_range_init(vaddr, 0, pgd_base); | |
341 | ||
342 | permanent_kmaps_init(pgd_base); | |
343 | ||
344 | #ifdef CONFIG_X86_PAE | |
345 | /* | |
346 | * Add low memory identity-mappings - SMP needs it when | |
347 | * starting up on an AP from real-mode. In the non-PAE | |
348 | * case we already have these mappings through head.S. | |
349 | * All user-space mappings are explicitly cleared after | |
350 | * SMP startup. | |
351 | */ | |
c9b02a24 | 352 | set_pgd(&pgd_base[0], pgd_base[USER_PTRS_PER_PGD]); |
1da177e4 LT |
353 | #endif |
354 | } | |
355 | ||
648be318 | 356 | #ifdef CONFIG_SOFTWARE_SUSPEND |
1da177e4 LT |
357 | /* |
358 | * Swap suspend & friends need this for resume because things like the intel-agp | |
359 | * driver might have split up a kernel 4MB mapping. | |
360 | */ | |
361 | char __nosavedata swsusp_pg_dir[PAGE_SIZE] | |
362 | __attribute__ ((aligned (PAGE_SIZE))); | |
363 | ||
364 | static inline void save_pg_dir(void) | |
365 | { | |
366 | memcpy(swsusp_pg_dir, swapper_pg_dir, PAGE_SIZE); | |
367 | } | |
368 | #else | |
369 | static inline void save_pg_dir(void) | |
370 | { | |
371 | } | |
372 | #endif | |
373 | ||
374 | void zap_low_mappings (void) | |
375 | { | |
376 | int i; | |
377 | ||
378 | save_pg_dir(); | |
379 | ||
380 | /* | |
381 | * Zap initial low-memory mappings. | |
382 | * | |
383 | * Note that "pgd_clear()" doesn't do it for | |
384 | * us, because pgd_clear() is a no-op on i386. | |
385 | */ | |
386 | for (i = 0; i < USER_PTRS_PER_PGD; i++) | |
387 | #ifdef CONFIG_X86_PAE | |
388 | set_pgd(swapper_pg_dir+i, __pgd(1 + __pa(empty_zero_page))); | |
389 | #else | |
390 | set_pgd(swapper_pg_dir+i, __pgd(0)); | |
391 | #endif | |
392 | flush_tlb_all(); | |
393 | } | |
394 | ||
395 | static int disable_nx __initdata = 0; | |
6c231b7b | 396 | u64 __supported_pte_mask __read_mostly = ~_PAGE_NX; |
1da177e4 LT |
397 | |
398 | /* | |
399 | * noexec = on|off | |
400 | * | |
401 | * Control non executable mappings. | |
402 | * | |
403 | * on Enable | |
404 | * off Disable | |
405 | */ | |
406 | void __init noexec_setup(const char *str) | |
407 | { | |
408 | if (!strncmp(str, "on",2) && cpu_has_nx) { | |
409 | __supported_pte_mask |= _PAGE_NX; | |
410 | disable_nx = 0; | |
411 | } else if (!strncmp(str,"off",3)) { | |
412 | disable_nx = 1; | |
413 | __supported_pte_mask &= ~_PAGE_NX; | |
414 | } | |
415 | } | |
416 | ||
417 | int nx_enabled = 0; | |
418 | #ifdef CONFIG_X86_PAE | |
419 | ||
420 | static void __init set_nx(void) | |
421 | { | |
422 | unsigned int v[4], l, h; | |
423 | ||
424 | if (cpu_has_pae && (cpuid_eax(0x80000000) > 0x80000001)) { | |
425 | cpuid(0x80000001, &v[0], &v[1], &v[2], &v[3]); | |
426 | if ((v[3] & (1 << 20)) && !disable_nx) { | |
427 | rdmsr(MSR_EFER, l, h); | |
428 | l |= EFER_NX; | |
429 | wrmsr(MSR_EFER, l, h); | |
430 | nx_enabled = 1; | |
431 | __supported_pte_mask |= _PAGE_NX; | |
432 | } | |
433 | } | |
434 | } | |
435 | ||
436 | /* | |
437 | * Enables/disables executability of a given kernel page and | |
438 | * returns the previous setting. | |
439 | */ | |
440 | int __init set_kernel_exec(unsigned long vaddr, int enable) | |
441 | { | |
442 | pte_t *pte; | |
443 | int ret = 1; | |
444 | ||
445 | if (!nx_enabled) | |
446 | goto out; | |
447 | ||
448 | pte = lookup_address(vaddr); | |
449 | BUG_ON(!pte); | |
450 | ||
451 | if (!pte_exec_kernel(*pte)) | |
452 | ret = 0; | |
453 | ||
454 | if (enable) | |
455 | pte->pte_high &= ~(1 << (_PAGE_BIT_NX - 32)); | |
456 | else | |
457 | pte->pte_high |= 1 << (_PAGE_BIT_NX - 32); | |
458 | __flush_tlb_all(); | |
459 | out: | |
460 | return ret; | |
461 | } | |
462 | ||
463 | #endif | |
464 | ||
465 | /* | |
466 | * paging_init() sets up the page tables - note that the first 8MB are | |
467 | * already mapped by head.S. | |
468 | * | |
469 | * This routines also unmaps the page at virtual kernel address 0, so | |
470 | * that we can trap those pesky NULL-reference errors in the kernel. | |
471 | */ | |
472 | void __init paging_init(void) | |
473 | { | |
474 | #ifdef CONFIG_X86_PAE | |
475 | set_nx(); | |
476 | if (nx_enabled) | |
477 | printk("NX (Execute Disable) protection: active\n"); | |
478 | #endif | |
479 | ||
480 | pagetable_init(); | |
481 | ||
482 | load_cr3(swapper_pg_dir); | |
483 | ||
484 | #ifdef CONFIG_X86_PAE | |
485 | /* | |
486 | * We will bail out later - printk doesn't work right now so | |
487 | * the user would just see a hanging kernel. | |
488 | */ | |
489 | if (cpu_has_pae) | |
490 | set_in_cr4(X86_CR4_PAE); | |
491 | #endif | |
492 | __flush_tlb_all(); | |
493 | ||
494 | kmap_init(); | |
495 | } | |
496 | ||
497 | /* | |
498 | * Test if the WP bit works in supervisor mode. It isn't supported on 386's | |
499 | * and also on some strange 486's (NexGen etc.). All 586+'s are OK. This | |
500 | * used to involve black magic jumps to work around some nasty CPU bugs, | |
501 | * but fortunately the switch to using exceptions got rid of all that. | |
502 | */ | |
503 | ||
504 | static void __init test_wp_bit(void) | |
505 | { | |
506 | printk("Checking if this processor honours the WP bit even in supervisor mode... "); | |
507 | ||
508 | /* Any page-aligned address will do, the test is non-destructive */ | |
509 | __set_fixmap(FIX_WP_TEST, __pa(&swapper_pg_dir), PAGE_READONLY); | |
510 | boot_cpu_data.wp_works_ok = do_test_wp_bit(); | |
511 | clear_fixmap(FIX_WP_TEST); | |
512 | ||
513 | if (!boot_cpu_data.wp_works_ok) { | |
514 | printk("No.\n"); | |
515 | #ifdef CONFIG_X86_WP_WORKS_OK | |
516 | panic("This kernel doesn't support CPU's with broken WP. Recompile it for a 386!"); | |
517 | #endif | |
518 | } else { | |
519 | printk("Ok.\n"); | |
520 | } | |
521 | } | |
522 | ||
523 | static void __init set_max_mapnr_init(void) | |
524 | { | |
525 | #ifdef CONFIG_HIGHMEM | |
526 | num_physpages = highend_pfn; | |
527 | #else | |
528 | num_physpages = max_low_pfn; | |
529 | #endif | |
05b79bdc | 530 | #ifdef CONFIG_FLATMEM |
1da177e4 LT |
531 | max_mapnr = num_physpages; |
532 | #endif | |
533 | } | |
534 | ||
535 | static struct kcore_list kcore_mem, kcore_vmalloc; | |
536 | ||
537 | void __init mem_init(void) | |
538 | { | |
539 | extern int ppro_with_ram_bug(void); | |
540 | int codesize, reservedpages, datasize, initsize; | |
541 | int tmp; | |
542 | int bad_ppro; | |
543 | ||
05b79bdc | 544 | #ifdef CONFIG_FLATMEM |
1da177e4 LT |
545 | if (!mem_map) |
546 | BUG(); | |
547 | #endif | |
548 | ||
549 | bad_ppro = ppro_with_ram_bug(); | |
550 | ||
551 | #ifdef CONFIG_HIGHMEM | |
552 | /* check that fixmap and pkmap do not overlap */ | |
553 | if (PKMAP_BASE+LAST_PKMAP*PAGE_SIZE >= FIXADDR_START) { | |
554 | printk(KERN_ERR "fixmap and kmap areas overlap - this will crash\n"); | |
555 | printk(KERN_ERR "pkstart: %lxh pkend: %lxh fixstart %lxh\n", | |
556 | PKMAP_BASE, PKMAP_BASE+LAST_PKMAP*PAGE_SIZE, FIXADDR_START); | |
557 | BUG(); | |
558 | } | |
559 | #endif | |
560 | ||
561 | set_max_mapnr_init(); | |
562 | ||
563 | #ifdef CONFIG_HIGHMEM | |
564 | high_memory = (void *) __va(highstart_pfn * PAGE_SIZE - 1) + 1; | |
565 | #else | |
566 | high_memory = (void *) __va(max_low_pfn * PAGE_SIZE - 1) + 1; | |
567 | #endif | |
568 | ||
569 | /* this will put all low memory onto the freelists */ | |
570 | totalram_pages += free_all_bootmem(); | |
571 | ||
572 | reservedpages = 0; | |
573 | for (tmp = 0; tmp < max_low_pfn; tmp++) | |
574 | /* | |
575 | * Only count reserved RAM pages | |
576 | */ | |
577 | if (page_is_ram(tmp) && PageReserved(pfn_to_page(tmp))) | |
578 | reservedpages++; | |
579 | ||
580 | set_highmem_pages_init(bad_ppro); | |
581 | ||
582 | codesize = (unsigned long) &_etext - (unsigned long) &_text; | |
583 | datasize = (unsigned long) &_edata - (unsigned long) &_etext; | |
584 | initsize = (unsigned long) &__init_end - (unsigned long) &__init_begin; | |
585 | ||
586 | kclist_add(&kcore_mem, __va(0), max_low_pfn << PAGE_SHIFT); | |
587 | kclist_add(&kcore_vmalloc, (void *)VMALLOC_START, | |
588 | VMALLOC_END-VMALLOC_START); | |
589 | ||
590 | printk(KERN_INFO "Memory: %luk/%luk available (%dk kernel code, %dk reserved, %dk data, %dk init, %ldk highmem)\n", | |
591 | (unsigned long) nr_free_pages() << (PAGE_SHIFT-10), | |
592 | num_physpages << (PAGE_SHIFT-10), | |
593 | codesize >> 10, | |
594 | reservedpages << (PAGE_SHIFT-10), | |
595 | datasize >> 10, | |
596 | initsize >> 10, | |
597 | (unsigned long) (totalhigh_pages << (PAGE_SHIFT-10)) | |
598 | ); | |
599 | ||
600 | #ifdef CONFIG_X86_PAE | |
601 | if (!cpu_has_pae) | |
602 | panic("cannot execute a PAE-enabled kernel on a PAE-less CPU!"); | |
603 | #endif | |
604 | if (boot_cpu_data.wp_works_ok < 0) | |
605 | test_wp_bit(); | |
606 | ||
607 | /* | |
608 | * Subtle. SMP is doing it's boot stuff late (because it has to | |
609 | * fork idle threads) - but it also needs low mappings for the | |
610 | * protected-mode entry to work. We zap these entries only after | |
611 | * the WP-bit has been tested. | |
612 | */ | |
613 | #ifndef CONFIG_SMP | |
614 | zap_low_mappings(); | |
615 | #endif | |
616 | } | |
617 | ||
618 | kmem_cache_t *pgd_cache; | |
619 | kmem_cache_t *pmd_cache; | |
620 | ||
621 | void __init pgtable_cache_init(void) | |
622 | { | |
623 | if (PTRS_PER_PMD > 1) { | |
624 | pmd_cache = kmem_cache_create("pmd", | |
625 | PTRS_PER_PMD*sizeof(pmd_t), | |
626 | PTRS_PER_PMD*sizeof(pmd_t), | |
627 | 0, | |
628 | pmd_ctor, | |
629 | NULL); | |
630 | if (!pmd_cache) | |
631 | panic("pgtable_cache_init(): cannot create pmd cache"); | |
632 | } | |
633 | pgd_cache = kmem_cache_create("pgd", | |
634 | PTRS_PER_PGD*sizeof(pgd_t), | |
635 | PTRS_PER_PGD*sizeof(pgd_t), | |
636 | 0, | |
637 | pgd_ctor, | |
638 | PTRS_PER_PMD == 1 ? pgd_dtor : NULL); | |
639 | if (!pgd_cache) | |
640 | panic("pgtable_cache_init(): Cannot create pgd cache"); | |
641 | } | |
642 | ||
643 | /* | |
644 | * This function cannot be __init, since exceptions don't work in that | |
645 | * section. Put this after the callers, so that it cannot be inlined. | |
646 | */ | |
647 | static int noinline do_test_wp_bit(void) | |
648 | { | |
649 | char tmp_reg; | |
650 | int flag; | |
651 | ||
652 | __asm__ __volatile__( | |
653 | " movb %0,%1 \n" | |
654 | "1: movb %1,%0 \n" | |
655 | " xorl %2,%2 \n" | |
656 | "2: \n" | |
657 | ".section __ex_table,\"a\"\n" | |
658 | " .align 4 \n" | |
659 | " .long 1b,2b \n" | |
660 | ".previous \n" | |
661 | :"=m" (*(char *)fix_to_virt(FIX_WP_TEST)), | |
662 | "=q" (tmp_reg), | |
663 | "=r" (flag) | |
664 | :"2" (1) | |
665 | :"memory"); | |
666 | ||
667 | return flag; | |
668 | } | |
669 | ||
670 | void free_initmem(void) | |
671 | { | |
672 | unsigned long addr; | |
673 | ||
674 | addr = (unsigned long)(&__init_begin); | |
675 | for (; addr < (unsigned long)(&__init_end); addr += PAGE_SIZE) { | |
676 | ClearPageReserved(virt_to_page(addr)); | |
677 | set_page_count(virt_to_page(addr), 1); | |
678 | memset((void *)addr, 0xcc, PAGE_SIZE); | |
679 | free_page(addr); | |
680 | totalram_pages++; | |
681 | } | |
682 | printk (KERN_INFO "Freeing unused kernel memory: %dk freed\n", (__init_end - __init_begin) >> 10); | |
683 | } | |
684 | ||
685 | #ifdef CONFIG_BLK_DEV_INITRD | |
686 | void free_initrd_mem(unsigned long start, unsigned long end) | |
687 | { | |
688 | if (start < end) | |
689 | printk (KERN_INFO "Freeing initrd memory: %ldk freed\n", (end - start) >> 10); | |
690 | for (; start < end; start += PAGE_SIZE) { | |
691 | ClearPageReserved(virt_to_page(start)); | |
692 | set_page_count(virt_to_page(start), 1); | |
693 | free_page(start); | |
694 | totalram_pages++; | |
695 | } | |
696 | } | |
697 | #endif |