diff options
Diffstat (limited to 'arch/x86/mm/init_32.c')
-rw-r--r-- | arch/x86/mm/init_32.c | 460 |
1 files changed, 154 insertions, 306 deletions
diff --git a/arch/x86/mm/init_32.c b/arch/x86/mm/init_32.c index 3eb2ed188a4c..d7f5060ab21c 100644 --- a/arch/x86/mm/init_32.c +++ b/arch/x86/mm/init_32.c | |||
@@ -49,9 +49,7 @@ | |||
49 | #include <asm/paravirt.h> | 49 | #include <asm/paravirt.h> |
50 | #include <asm/setup.h> | 50 | #include <asm/setup.h> |
51 | #include <asm/cacheflush.h> | 51 | #include <asm/cacheflush.h> |
52 | #include <asm/smp.h> | 52 | #include <asm/init.h> |
53 | |||
54 | unsigned int __VMALLOC_RESERVE = 128 << 20; | ||
55 | 53 | ||
56 | unsigned long max_low_pfn_mapped; | 54 | unsigned long max_low_pfn_mapped; |
57 | unsigned long max_pfn_mapped; | 55 | unsigned long max_pfn_mapped; |
@@ -61,19 +59,14 @@ unsigned long highstart_pfn, highend_pfn; | |||
61 | 59 | ||
62 | static noinline int do_test_wp_bit(void); | 60 | static noinline int do_test_wp_bit(void); |
63 | 61 | ||
64 | 62 | bool __read_mostly __vmalloc_start_set = false; | |
65 | static unsigned long __initdata table_start; | ||
66 | static unsigned long __meminitdata table_end; | ||
67 | static unsigned long __meminitdata table_top; | ||
68 | |||
69 | static int __initdata after_init_bootmem; | ||
70 | 63 | ||
71 | static __init void *alloc_low_page(void) | 64 | static __init void *alloc_low_page(void) |
72 | { | 65 | { |
73 | unsigned long pfn = table_end++; | 66 | unsigned long pfn = e820_table_end++; |
74 | void *adr; | 67 | void *adr; |
75 | 68 | ||
76 | if (pfn >= table_top) | 69 | if (pfn >= e820_table_top) |
77 | panic("alloc_low_page: ran out of memory"); | 70 | panic("alloc_low_page: ran out of memory"); |
78 | 71 | ||
79 | adr = __va(pfn * PAGE_SIZE); | 72 | adr = __va(pfn * PAGE_SIZE); |
@@ -93,7 +86,7 @@ static pmd_t * __init one_md_table_init(pgd_t *pgd) | |||
93 | 86 | ||
94 | #ifdef CONFIG_X86_PAE | 87 | #ifdef CONFIG_X86_PAE |
95 | if (!(pgd_val(*pgd) & _PAGE_PRESENT)) { | 88 | if (!(pgd_val(*pgd) & _PAGE_PRESENT)) { |
96 | if (after_init_bootmem) | 89 | if (after_bootmem) |
97 | pmd_table = (pmd_t *)alloc_bootmem_low_pages(PAGE_SIZE); | 90 | pmd_table = (pmd_t *)alloc_bootmem_low_pages(PAGE_SIZE); |
98 | else | 91 | else |
99 | pmd_table = (pmd_t *)alloc_low_page(); | 92 | pmd_table = (pmd_t *)alloc_low_page(); |
@@ -120,7 +113,7 @@ static pte_t * __init one_page_table_init(pmd_t *pmd) | |||
120 | if (!(pmd_val(*pmd) & _PAGE_PRESENT)) { | 113 | if (!(pmd_val(*pmd) & _PAGE_PRESENT)) { |
121 | pte_t *page_table = NULL; | 114 | pte_t *page_table = NULL; |
122 | 115 | ||
123 | if (after_init_bootmem) { | 116 | if (after_bootmem) { |
124 | #ifdef CONFIG_DEBUG_PAGEALLOC | 117 | #ifdef CONFIG_DEBUG_PAGEALLOC |
125 | page_table = (pte_t *) alloc_bootmem_pages(PAGE_SIZE); | 118 | page_table = (pte_t *) alloc_bootmem_pages(PAGE_SIZE); |
126 | #endif | 119 | #endif |
@@ -138,6 +131,23 @@ static pte_t * __init one_page_table_init(pmd_t *pmd) | |||
138 | return pte_offset_kernel(pmd, 0); | 131 | return pte_offset_kernel(pmd, 0); |
139 | } | 132 | } |
140 | 133 | ||
134 | pmd_t * __init populate_extra_pmd(unsigned long vaddr) | ||
135 | { | ||
136 | int pgd_idx = pgd_index(vaddr); | ||
137 | int pmd_idx = pmd_index(vaddr); | ||
138 | |||
139 | return one_md_table_init(swapper_pg_dir + pgd_idx) + pmd_idx; | ||
140 | } | ||
141 | |||
142 | pte_t * __init populate_extra_pte(unsigned long vaddr) | ||
143 | { | ||
144 | int pte_idx = pte_index(vaddr); | ||
145 | pmd_t *pmd; | ||
146 | |||
147 | pmd = populate_extra_pmd(vaddr); | ||
148 | return one_page_table_init(pmd) + pte_idx; | ||
149 | } | ||
150 | |||
141 | static pte_t *__init page_table_kmap_check(pte_t *pte, pmd_t *pmd, | 151 | static pte_t *__init page_table_kmap_check(pte_t *pte, pmd_t *pmd, |
142 | unsigned long vaddr, pte_t *lastpte) | 152 | unsigned long vaddr, pte_t *lastpte) |
143 | { | 153 | { |
@@ -154,12 +164,12 @@ static pte_t *__init page_table_kmap_check(pte_t *pte, pmd_t *pmd, | |||
154 | if (pmd_idx_kmap_begin != pmd_idx_kmap_end | 164 | if (pmd_idx_kmap_begin != pmd_idx_kmap_end |
155 | && (vaddr >> PMD_SHIFT) >= pmd_idx_kmap_begin | 165 | && (vaddr >> PMD_SHIFT) >= pmd_idx_kmap_begin |
156 | && (vaddr >> PMD_SHIFT) <= pmd_idx_kmap_end | 166 | && (vaddr >> PMD_SHIFT) <= pmd_idx_kmap_end |
157 | && ((__pa(pte) >> PAGE_SHIFT) < table_start | 167 | && ((__pa(pte) >> PAGE_SHIFT) < e820_table_start |
158 | || (__pa(pte) >> PAGE_SHIFT) >= table_end)) { | 168 | || (__pa(pte) >> PAGE_SHIFT) >= e820_table_end)) { |
159 | pte_t *newpte; | 169 | pte_t *newpte; |
160 | int i; | 170 | int i; |
161 | 171 | ||
162 | BUG_ON(after_init_bootmem); | 172 | BUG_ON(after_bootmem); |
163 | newpte = alloc_low_page(); | 173 | newpte = alloc_low_page(); |
164 | for (i = 0; i < PTRS_PER_PTE; i++) | 174 | for (i = 0; i < PTRS_PER_PTE; i++) |
165 | set_pte(newpte + i, pte[i]); | 175 | set_pte(newpte + i, pte[i]); |
@@ -228,11 +238,14 @@ static inline int is_kernel_text(unsigned long addr) | |||
228 | * of max_low_pfn pages, by creating page tables starting from address | 238 | * of max_low_pfn pages, by creating page tables starting from address |
229 | * PAGE_OFFSET: | 239 | * PAGE_OFFSET: |
230 | */ | 240 | */ |
231 | static void __init kernel_physical_mapping_init(pgd_t *pgd_base, | 241 | unsigned long __init |
232 | unsigned long start_pfn, | 242 | kernel_physical_mapping_init(unsigned long start, |
233 | unsigned long end_pfn, | 243 | unsigned long end, |
234 | int use_pse) | 244 | unsigned long page_size_mask) |
235 | { | 245 | { |
246 | int use_pse = page_size_mask == (1<<PG_LEVEL_2M); | ||
247 | unsigned long start_pfn, end_pfn; | ||
248 | pgd_t *pgd_base = swapper_pg_dir; | ||
236 | int pgd_idx, pmd_idx, pte_ofs; | 249 | int pgd_idx, pmd_idx, pte_ofs; |
237 | unsigned long pfn; | 250 | unsigned long pfn; |
238 | pgd_t *pgd; | 251 | pgd_t *pgd; |
@@ -241,6 +254,9 @@ static void __init kernel_physical_mapping_init(pgd_t *pgd_base, | |||
241 | unsigned pages_2m, pages_4k; | 254 | unsigned pages_2m, pages_4k; |
242 | int mapping_iter; | 255 | int mapping_iter; |
243 | 256 | ||
257 | start_pfn = start >> PAGE_SHIFT; | ||
258 | end_pfn = end >> PAGE_SHIFT; | ||
259 | |||
244 | /* | 260 | /* |
245 | * First iteration will setup identity mapping using large/small pages | 261 | * First iteration will setup identity mapping using large/small pages |
246 | * based on use_pse, with other attributes same as set by | 262 | * based on use_pse, with other attributes same as set by |
@@ -355,26 +371,6 @@ repeat: | |||
355 | mapping_iter = 2; | 371 | mapping_iter = 2; |
356 | goto repeat; | 372 | goto repeat; |
357 | } | 373 | } |
358 | } | ||
359 | |||
360 | /* | ||
361 | * devmem_is_allowed() checks to see if /dev/mem access to a certain address | ||
362 | * is valid. The argument is a physical page number. | ||
363 | * | ||
364 | * | ||
365 | * On x86, access has to be given to the first megabyte of ram because that area | ||
366 | * contains bios code and data regions used by X and dosemu and similar apps. | ||
367 | * Access has to be given to non-kernel-ram areas as well, these contain the PCI | ||
368 | * mmio resources as well as potential bios/acpi data regions. | ||
369 | */ | ||
370 | int devmem_is_allowed(unsigned long pagenr) | ||
371 | { | ||
372 | if (pagenr <= 256) | ||
373 | return 1; | ||
374 | if (iomem_is_exclusive(pagenr << PAGE_SHIFT)) | ||
375 | return 0; | ||
376 | if (!page_is_ram(pagenr)) | ||
377 | return 1; | ||
378 | return 0; | 374 | return 0; |
379 | } | 375 | } |
380 | 376 | ||
@@ -470,22 +466,10 @@ void __init add_highpages_with_active_regions(int nid, unsigned long start_pfn, | |||
470 | work_with_active_regions(nid, add_highpages_work_fn, &data); | 466 | work_with_active_regions(nid, add_highpages_work_fn, &data); |
471 | } | 467 | } |
472 | 468 | ||
473 | #ifndef CONFIG_NUMA | ||
474 | static void __init set_highmem_pages_init(void) | ||
475 | { | ||
476 | add_highpages_with_active_regions(0, highstart_pfn, highend_pfn); | ||
477 | |||
478 | totalram_pages += totalhigh_pages; | ||
479 | } | ||
480 | #endif /* !CONFIG_NUMA */ | ||
481 | |||
482 | #else | 469 | #else |
483 | static inline void permanent_kmaps_init(pgd_t *pgd_base) | 470 | static inline void permanent_kmaps_init(pgd_t *pgd_base) |
484 | { | 471 | { |
485 | } | 472 | } |
486 | static inline void set_highmem_pages_init(void) | ||
487 | { | ||
488 | } | ||
489 | #endif /* CONFIG_HIGHMEM */ | 473 | #endif /* CONFIG_HIGHMEM */ |
490 | 474 | ||
491 | void __init native_pagetable_setup_start(pgd_t *base) | 475 | void __init native_pagetable_setup_start(pgd_t *base) |
@@ -543,8 +527,9 @@ void __init native_pagetable_setup_done(pgd_t *base) | |||
543 | * be partially populated, and so it avoids stomping on any existing | 527 | * be partially populated, and so it avoids stomping on any existing |
544 | * mappings. | 528 | * mappings. |
545 | */ | 529 | */ |
546 | static void __init early_ioremap_page_table_range_init(pgd_t *pgd_base) | 530 | void __init early_ioremap_page_table_range_init(void) |
547 | { | 531 | { |
532 | pgd_t *pgd_base = swapper_pg_dir; | ||
548 | unsigned long vaddr, end; | 533 | unsigned long vaddr, end; |
549 | 534 | ||
550 | /* | 535 | /* |
@@ -639,7 +624,7 @@ static int __init noexec_setup(char *str) | |||
639 | } | 624 | } |
640 | early_param("noexec", noexec_setup); | 625 | early_param("noexec", noexec_setup); |
641 | 626 | ||
642 | static void __init set_nx(void) | 627 | void __init set_nx(void) |
643 | { | 628 | { |
644 | unsigned int v[4], l, h; | 629 | unsigned int v[4], l, h; |
645 | 630 | ||
@@ -675,75 +660,97 @@ static int __init parse_highmem(char *arg) | |||
675 | } | 660 | } |
676 | early_param("highmem", parse_highmem); | 661 | early_param("highmem", parse_highmem); |
677 | 662 | ||
663 | #define MSG_HIGHMEM_TOO_BIG \ | ||
664 | "highmem size (%luMB) is bigger than pages available (%luMB)!\n" | ||
665 | |||
666 | #define MSG_LOWMEM_TOO_SMALL \ | ||
667 | "highmem size (%luMB) results in <64MB lowmem, ignoring it!\n" | ||
678 | /* | 668 | /* |
679 | * Determine low and high memory ranges: | 669 | * All of RAM fits into lowmem - but if user wants highmem |
670 | * artificially via the highmem=x boot parameter then create | ||
671 | * it: | ||
680 | */ | 672 | */ |
681 | void __init find_low_pfn_range(void) | 673 | void __init lowmem_pfn_init(void) |
682 | { | 674 | { |
683 | /* it could update max_pfn */ | ||
684 | |||
685 | /* max_low_pfn is 0, we already have early_res support */ | 675 | /* max_low_pfn is 0, we already have early_res support */ |
686 | |||
687 | max_low_pfn = max_pfn; | 676 | max_low_pfn = max_pfn; |
688 | if (max_low_pfn > MAXMEM_PFN) { | 677 | |
689 | if (highmem_pages == -1) | 678 | if (highmem_pages == -1) |
690 | highmem_pages = max_pfn - MAXMEM_PFN; | 679 | highmem_pages = 0; |
691 | if (highmem_pages + MAXMEM_PFN < max_pfn) | 680 | #ifdef CONFIG_HIGHMEM |
692 | max_pfn = MAXMEM_PFN + highmem_pages; | 681 | if (highmem_pages >= max_pfn) { |
693 | if (highmem_pages + MAXMEM_PFN > max_pfn) { | 682 | printk(KERN_ERR MSG_HIGHMEM_TOO_BIG, |
694 | printk(KERN_WARNING "only %luMB highmem pages " | 683 | pages_to_mb(highmem_pages), pages_to_mb(max_pfn)); |
695 | "available, ignoring highmem size of %uMB.\n", | 684 | highmem_pages = 0; |
696 | pages_to_mb(max_pfn - MAXMEM_PFN), | 685 | } |
686 | if (highmem_pages) { | ||
687 | if (max_low_pfn - highmem_pages < 64*1024*1024/PAGE_SIZE) { | ||
688 | printk(KERN_ERR MSG_LOWMEM_TOO_SMALL, | ||
697 | pages_to_mb(highmem_pages)); | 689 | pages_to_mb(highmem_pages)); |
698 | highmem_pages = 0; | 690 | highmem_pages = 0; |
699 | } | 691 | } |
700 | max_low_pfn = MAXMEM_PFN; | 692 | max_low_pfn -= highmem_pages; |
693 | } | ||
694 | #else | ||
695 | if (highmem_pages) | ||
696 | printk(KERN_ERR "ignoring highmem size on non-highmem kernel!\n"); | ||
697 | #endif | ||
698 | } | ||
699 | |||
700 | #define MSG_HIGHMEM_TOO_SMALL \ | ||
701 | "only %luMB highmem pages available, ignoring highmem size of %luMB!\n" | ||
702 | |||
703 | #define MSG_HIGHMEM_TRIMMED \ | ||
704 | "Warning: only 4GB will be used. Use a HIGHMEM64G enabled kernel!\n" | ||
705 | /* | ||
706 | * We have more RAM than fits into lowmem - we try to put it into | ||
707 | * highmem, also taking the highmem=x boot parameter into account: | ||
708 | */ | ||
709 | void __init highmem_pfn_init(void) | ||
710 | { | ||
711 | max_low_pfn = MAXMEM_PFN; | ||
712 | |||
713 | if (highmem_pages == -1) | ||
714 | highmem_pages = max_pfn - MAXMEM_PFN; | ||
715 | |||
716 | if (highmem_pages + MAXMEM_PFN < max_pfn) | ||
717 | max_pfn = MAXMEM_PFN + highmem_pages; | ||
718 | |||
719 | if (highmem_pages + MAXMEM_PFN > max_pfn) { | ||
720 | printk(KERN_WARNING MSG_HIGHMEM_TOO_SMALL, | ||
721 | pages_to_mb(max_pfn - MAXMEM_PFN), | ||
722 | pages_to_mb(highmem_pages)); | ||
723 | highmem_pages = 0; | ||
724 | } | ||
701 | #ifndef CONFIG_HIGHMEM | 725 | #ifndef CONFIG_HIGHMEM |
702 | /* Maximum memory usable is what is directly addressable */ | 726 | /* Maximum memory usable is what is directly addressable */ |
703 | printk(KERN_WARNING "Warning only %ldMB will be used.\n", | 727 | printk(KERN_WARNING "Warning only %ldMB will be used.\n", MAXMEM>>20); |
704 | MAXMEM>>20); | 728 | if (max_pfn > MAX_NONPAE_PFN) |
705 | if (max_pfn > MAX_NONPAE_PFN) | 729 | printk(KERN_WARNING "Use a HIGHMEM64G enabled kernel.\n"); |
706 | printk(KERN_WARNING | 730 | else |
707 | "Use a HIGHMEM64G enabled kernel.\n"); | 731 | printk(KERN_WARNING "Use a HIGHMEM enabled kernel.\n"); |
708 | else | 732 | max_pfn = MAXMEM_PFN; |
709 | printk(KERN_WARNING "Use a HIGHMEM enabled kernel.\n"); | ||
710 | max_pfn = MAXMEM_PFN; | ||
711 | #else /* !CONFIG_HIGHMEM */ | 733 | #else /* !CONFIG_HIGHMEM */ |
712 | #ifndef CONFIG_HIGHMEM64G | 734 | #ifndef CONFIG_HIGHMEM64G |
713 | if (max_pfn > MAX_NONPAE_PFN) { | 735 | if (max_pfn > MAX_NONPAE_PFN) { |
714 | max_pfn = MAX_NONPAE_PFN; | 736 | max_pfn = MAX_NONPAE_PFN; |
715 | printk(KERN_WARNING "Warning only 4GB will be used." | 737 | printk(KERN_WARNING MSG_HIGHMEM_TRIMMED); |
716 | "Use a HIGHMEM64G enabled kernel.\n"); | 738 | } |
717 | } | ||
718 | #endif /* !CONFIG_HIGHMEM64G */ | 739 | #endif /* !CONFIG_HIGHMEM64G */ |
719 | #endif /* !CONFIG_HIGHMEM */ | 740 | #endif /* !CONFIG_HIGHMEM */ |
720 | } else { | 741 | } |
721 | if (highmem_pages == -1) | 742 | |
722 | highmem_pages = 0; | 743 | /* |
723 | #ifdef CONFIG_HIGHMEM | 744 | * Determine low and high memory ranges: |
724 | if (highmem_pages >= max_pfn) { | 745 | */ |
725 | printk(KERN_ERR "highmem size specified (%uMB) is " | 746 | void __init find_low_pfn_range(void) |
726 | "bigger than pages available (%luMB)!.\n", | 747 | { |
727 | pages_to_mb(highmem_pages), | 748 | /* it could update max_pfn */ |
728 | pages_to_mb(max_pfn)); | 749 | |
729 | highmem_pages = 0; | 750 | if (max_pfn <= MAXMEM_PFN) |
730 | } | 751 | lowmem_pfn_init(); |
731 | if (highmem_pages) { | 752 | else |
732 | if (max_low_pfn - highmem_pages < | 753 | highmem_pfn_init(); |
733 | 64*1024*1024/PAGE_SIZE){ | ||
734 | printk(KERN_ERR "highmem size %uMB results in " | ||
735 | "smaller than 64MB lowmem, ignoring it.\n" | ||
736 | , pages_to_mb(highmem_pages)); | ||
737 | highmem_pages = 0; | ||
738 | } | ||
739 | max_low_pfn -= highmem_pages; | ||
740 | } | ||
741 | #else | ||
742 | if (highmem_pages) | ||
743 | printk(KERN_ERR "ignoring highmem size on non-highmem" | ||
744 | " kernel!\n"); | ||
745 | #endif | ||
746 | } | ||
747 | } | 754 | } |
748 | 755 | ||
749 | #ifndef CONFIG_NEED_MULTIPLE_NODES | 756 | #ifndef CONFIG_NEED_MULTIPLE_NODES |
@@ -769,6 +776,8 @@ void __init initmem_init(unsigned long start_pfn, | |||
769 | #ifdef CONFIG_FLATMEM | 776 | #ifdef CONFIG_FLATMEM |
770 | max_mapnr = num_physpages; | 777 | max_mapnr = num_physpages; |
771 | #endif | 778 | #endif |
779 | __vmalloc_start_set = true; | ||
780 | |||
772 | printk(KERN_NOTICE "%ldMB LOWMEM available.\n", | 781 | printk(KERN_NOTICE "%ldMB LOWMEM available.\n", |
773 | pages_to_mb(max_low_pfn)); | 782 | pages_to_mb(max_low_pfn)); |
774 | 783 | ||
@@ -790,176 +799,61 @@ static void __init zone_sizes_init(void) | |||
790 | free_area_init_nodes(max_zone_pfns); | 799 | free_area_init_nodes(max_zone_pfns); |
791 | } | 800 | } |
792 | 801 | ||
802 | static unsigned long __init setup_node_bootmem(int nodeid, | ||
803 | unsigned long start_pfn, | ||
804 | unsigned long end_pfn, | ||
805 | unsigned long bootmap) | ||
806 | { | ||
807 | unsigned long bootmap_size; | ||
808 | |||
809 | if (start_pfn > max_low_pfn) | ||
810 | return bootmap; | ||
811 | if (end_pfn > max_low_pfn) | ||
812 | end_pfn = max_low_pfn; | ||
813 | |||
814 | /* don't touch min_low_pfn */ | ||
815 | bootmap_size = init_bootmem_node(NODE_DATA(nodeid), | ||
816 | bootmap >> PAGE_SHIFT, | ||
817 | start_pfn, end_pfn); | ||
818 | printk(KERN_INFO " node %d low ram: %08lx - %08lx\n", | ||
819 | nodeid, start_pfn<<PAGE_SHIFT, end_pfn<<PAGE_SHIFT); | ||
820 | printk(KERN_INFO " node %d bootmap %08lx - %08lx\n", | ||
821 | nodeid, bootmap, bootmap + bootmap_size); | ||
822 | free_bootmem_with_active_regions(nodeid, end_pfn); | ||
823 | early_res_to_bootmem(start_pfn<<PAGE_SHIFT, end_pfn<<PAGE_SHIFT); | ||
824 | |||
825 | return bootmap + bootmap_size; | ||
826 | } | ||
827 | |||
793 | void __init setup_bootmem_allocator(void) | 828 | void __init setup_bootmem_allocator(void) |
794 | { | 829 | { |
795 | int i; | 830 | int nodeid; |
796 | unsigned long bootmap_size, bootmap; | 831 | unsigned long bootmap_size, bootmap; |
797 | /* | 832 | /* |
798 | * Initialize the boot-time allocator (with low memory only): | 833 | * Initialize the boot-time allocator (with low memory only): |
799 | */ | 834 | */ |
800 | bootmap_size = bootmem_bootmap_pages(max_low_pfn)<<PAGE_SHIFT; | 835 | bootmap_size = bootmem_bootmap_pages(max_low_pfn)<<PAGE_SHIFT; |
801 | bootmap = find_e820_area(min_low_pfn<<PAGE_SHIFT, | 836 | bootmap = find_e820_area(0, max_pfn_mapped<<PAGE_SHIFT, bootmap_size, |
802 | max_pfn_mapped<<PAGE_SHIFT, bootmap_size, | ||
803 | PAGE_SIZE); | 837 | PAGE_SIZE); |
804 | if (bootmap == -1L) | 838 | if (bootmap == -1L) |
805 | panic("Cannot find bootmem map of size %ld\n", bootmap_size); | 839 | panic("Cannot find bootmem map of size %ld\n", bootmap_size); |
806 | reserve_early(bootmap, bootmap + bootmap_size, "BOOTMAP"); | 840 | reserve_early(bootmap, bootmap + bootmap_size, "BOOTMAP"); |
807 | 841 | ||
808 | /* don't touch min_low_pfn */ | ||
809 | bootmap_size = init_bootmem_node(NODE_DATA(0), bootmap >> PAGE_SHIFT, | ||
810 | min_low_pfn, max_low_pfn); | ||
811 | printk(KERN_INFO " mapped low ram: 0 - %08lx\n", | 842 | printk(KERN_INFO " mapped low ram: 0 - %08lx\n", |
812 | max_pfn_mapped<<PAGE_SHIFT); | 843 | max_pfn_mapped<<PAGE_SHIFT); |
813 | printk(KERN_INFO " low ram: %08lx - %08lx\n", | 844 | printk(KERN_INFO " low ram: 0 - %08lx\n", max_low_pfn<<PAGE_SHIFT); |
814 | min_low_pfn<<PAGE_SHIFT, max_low_pfn<<PAGE_SHIFT); | ||
815 | printk(KERN_INFO " bootmap %08lx - %08lx\n", | ||
816 | bootmap, bootmap + bootmap_size); | ||
817 | for_each_online_node(i) | ||
818 | free_bootmem_with_active_regions(i, max_low_pfn); | ||
819 | early_res_to_bootmem(0, max_low_pfn<<PAGE_SHIFT); | ||
820 | |||
821 | after_init_bootmem = 1; | ||
822 | } | ||
823 | |||
824 | static void __init find_early_table_space(unsigned long end, int use_pse) | ||
825 | { | ||
826 | unsigned long puds, pmds, ptes, tables, start; | ||
827 | |||
828 | puds = (end + PUD_SIZE - 1) >> PUD_SHIFT; | ||
829 | tables = PAGE_ALIGN(puds * sizeof(pud_t)); | ||
830 | |||
831 | pmds = (end + PMD_SIZE - 1) >> PMD_SHIFT; | ||
832 | tables += PAGE_ALIGN(pmds * sizeof(pmd_t)); | ||
833 | |||
834 | if (use_pse) { | ||
835 | unsigned long extra; | ||
836 | |||
837 | extra = end - ((end>>PMD_SHIFT) << PMD_SHIFT); | ||
838 | extra += PMD_SIZE; | ||
839 | ptes = (extra + PAGE_SIZE - 1) >> PAGE_SHIFT; | ||
840 | } else | ||
841 | ptes = (end + PAGE_SIZE - 1) >> PAGE_SHIFT; | ||
842 | 845 | ||
843 | tables += PAGE_ALIGN(ptes * sizeof(pte_t)); | 846 | #ifdef CONFIG_NEED_MULTIPLE_NODES |
844 | 847 | for_each_online_node(nodeid) | |
845 | /* for fixmap */ | 848 | bootmap = setup_node_bootmem(nodeid, node_start_pfn[nodeid], |
846 | tables += PAGE_ALIGN(__end_of_fixed_addresses * sizeof(pte_t)); | 849 | node_end_pfn[nodeid], bootmap); |
847 | |||
848 | /* | ||
849 | * RED-PEN putting page tables only on node 0 could | ||
850 | * cause a hotspot and fill up ZONE_DMA. The page tables | ||
851 | * need roughly 0.5KB per GB. | ||
852 | */ | ||
853 | start = 0x7000; | ||
854 | table_start = find_e820_area(start, max_pfn_mapped<<PAGE_SHIFT, | ||
855 | tables, PAGE_SIZE); | ||
856 | if (table_start == -1UL) | ||
857 | panic("Cannot find space for the kernel page tables"); | ||
858 | |||
859 | table_start >>= PAGE_SHIFT; | ||
860 | table_end = table_start; | ||
861 | table_top = table_start + (tables>>PAGE_SHIFT); | ||
862 | |||
863 | printk(KERN_DEBUG "kernel direct mapping tables up to %lx @ %lx-%lx\n", | ||
864 | end, table_start << PAGE_SHIFT, | ||
865 | (table_start << PAGE_SHIFT) + tables); | ||
866 | } | ||
867 | |||
868 | unsigned long __init_refok init_memory_mapping(unsigned long start, | ||
869 | unsigned long end) | ||
870 | { | ||
871 | pgd_t *pgd_base = swapper_pg_dir; | ||
872 | unsigned long start_pfn, end_pfn; | ||
873 | unsigned long big_page_start; | ||
874 | #ifdef CONFIG_DEBUG_PAGEALLOC | ||
875 | /* | ||
876 | * For CONFIG_DEBUG_PAGEALLOC, identity mapping will use small pages. | ||
877 | * This will simplify cpa(), which otherwise needs to support splitting | ||
878 | * large pages into small in interrupt context, etc. | ||
879 | */ | ||
880 | int use_pse = 0; | ||
881 | #else | 850 | #else |
882 | int use_pse = cpu_has_pse; | 851 | bootmap = setup_node_bootmem(0, 0, max_low_pfn, bootmap); |
883 | #endif | ||
884 | |||
885 | /* | ||
886 | * Find space for the kernel direct mapping tables. | ||
887 | */ | ||
888 | if (!after_init_bootmem) | ||
889 | find_early_table_space(end, use_pse); | ||
890 | |||
891 | #ifdef CONFIG_X86_PAE | ||
892 | set_nx(); | ||
893 | if (nx_enabled) | ||
894 | printk(KERN_INFO "NX (Execute Disable) protection: active\n"); | ||
895 | #endif | 852 | #endif |
896 | 853 | ||
897 | /* Enable PSE if available */ | 854 | after_bootmem = 1; |
898 | if (cpu_has_pse) | ||
899 | set_in_cr4(X86_CR4_PSE); | ||
900 | |||
901 | /* Enable PGE if available */ | ||
902 | if (cpu_has_pge) { | ||
903 | set_in_cr4(X86_CR4_PGE); | ||
904 | __supported_pte_mask |= _PAGE_GLOBAL; | ||
905 | } | ||
906 | |||
907 | /* | ||
908 | * Don't use a large page for the first 2/4MB of memory | ||
909 | * because there are often fixed size MTRRs in there | ||
910 | * and overlapping MTRRs into large pages can cause | ||
911 | * slowdowns. | ||
912 | */ | ||
913 | big_page_start = PMD_SIZE; | ||
914 | |||
915 | if (start < big_page_start) { | ||
916 | start_pfn = start >> PAGE_SHIFT; | ||
917 | end_pfn = min(big_page_start>>PAGE_SHIFT, end>>PAGE_SHIFT); | ||
918 | } else { | ||
919 | /* head is not big page alignment ? */ | ||
920 | start_pfn = start >> PAGE_SHIFT; | ||
921 | end_pfn = ((start + (PMD_SIZE - 1))>>PMD_SHIFT) | ||
922 | << (PMD_SHIFT - PAGE_SHIFT); | ||
923 | } | ||
924 | if (start_pfn < end_pfn) | ||
925 | kernel_physical_mapping_init(pgd_base, start_pfn, end_pfn, 0); | ||
926 | |||
927 | /* big page range */ | ||
928 | start_pfn = ((start + (PMD_SIZE - 1))>>PMD_SHIFT) | ||
929 | << (PMD_SHIFT - PAGE_SHIFT); | ||
930 | if (start_pfn < (big_page_start >> PAGE_SHIFT)) | ||
931 | start_pfn = big_page_start >> PAGE_SHIFT; | ||
932 | end_pfn = (end>>PMD_SHIFT) << (PMD_SHIFT - PAGE_SHIFT); | ||
933 | if (start_pfn < end_pfn) | ||
934 | kernel_physical_mapping_init(pgd_base, start_pfn, end_pfn, | ||
935 | use_pse); | ||
936 | |||
937 | /* tail is not big page alignment ? */ | ||
938 | start_pfn = end_pfn; | ||
939 | if (start_pfn > (big_page_start>>PAGE_SHIFT)) { | ||
940 | end_pfn = end >> PAGE_SHIFT; | ||
941 | if (start_pfn < end_pfn) | ||
942 | kernel_physical_mapping_init(pgd_base, start_pfn, | ||
943 | end_pfn, 0); | ||
944 | } | ||
945 | |||
946 | early_ioremap_page_table_range_init(pgd_base); | ||
947 | |||
948 | load_cr3(swapper_pg_dir); | ||
949 | |||
950 | __flush_tlb_all(); | ||
951 | |||
952 | if (!after_init_bootmem) | ||
953 | reserve_early(table_start << PAGE_SHIFT, | ||
954 | table_end << PAGE_SHIFT, "PGTABLE"); | ||
955 | |||
956 | if (!after_init_bootmem) | ||
957 | early_memtest(start, end); | ||
958 | |||
959 | return end >> PAGE_SHIFT; | ||
960 | } | 855 | } |
961 | 856 | ||
962 | |||
963 | /* | 857 | /* |
964 | * paging_init() sets up the page tables - note that the first 8MB are | 858 | * paging_init() sets up the page tables - note that the first 8MB are |
965 | * already mapped by head.S. | 859 | * already mapped by head.S. |
@@ -1222,52 +1116,6 @@ void mark_rodata_ro(void) | |||
1222 | } | 1116 | } |
1223 | #endif | 1117 | #endif |
1224 | 1118 | ||
1225 | void free_init_pages(char *what, unsigned long begin, unsigned long end) | ||
1226 | { | ||
1227 | #ifdef CONFIG_DEBUG_PAGEALLOC | ||
1228 | /* | ||
1229 | * If debugging page accesses then do not free this memory but | ||
1230 | * mark them not present - any buggy init-section access will | ||
1231 | * create a kernel page fault: | ||
1232 | */ | ||
1233 | printk(KERN_INFO "debug: unmapping init memory %08lx..%08lx\n", | ||
1234 | begin, PAGE_ALIGN(end)); | ||
1235 | set_memory_np(begin, (end - begin) >> PAGE_SHIFT); | ||
1236 | #else | ||
1237 | unsigned long addr; | ||
1238 | |||
1239 | /* | ||
1240 | * We just marked the kernel text read only above, now that | ||
1241 | * we are going to free part of that, we need to make that | ||
1242 | * writeable first. | ||
1243 | */ | ||
1244 | set_memory_rw(begin, (end - begin) >> PAGE_SHIFT); | ||
1245 | |||
1246 | for (addr = begin; addr < end; addr += PAGE_SIZE) { | ||
1247 | ClearPageReserved(virt_to_page(addr)); | ||
1248 | init_page_count(virt_to_page(addr)); | ||
1249 | memset((void *)addr, POISON_FREE_INITMEM, PAGE_SIZE); | ||
1250 | free_page(addr); | ||
1251 | totalram_pages++; | ||
1252 | } | ||
1253 | printk(KERN_INFO "Freeing %s: %luk freed\n", what, (end - begin) >> 10); | ||
1254 | #endif | ||
1255 | } | ||
1256 | |||
1257 | void free_initmem(void) | ||
1258 | { | ||
1259 | free_init_pages("unused kernel memory", | ||
1260 | (unsigned long)(&__init_begin), | ||
1261 | (unsigned long)(&__init_end)); | ||
1262 | } | ||
1263 | |||
1264 | #ifdef CONFIG_BLK_DEV_INITRD | ||
1265 | void free_initrd_mem(unsigned long start, unsigned long end) | ||
1266 | { | ||
1267 | free_init_pages("initrd memory", start, end); | ||
1268 | } | ||
1269 | #endif | ||
1270 | |||
1271 | int __init reserve_bootmem_generic(unsigned long phys, unsigned long len, | 1119 | int __init reserve_bootmem_generic(unsigned long phys, unsigned long len, |
1272 | int flags) | 1120 | int flags) |
1273 | { | 1121 | { |