diff options
Diffstat (limited to 'arch/x86/kernel/e820.c')
-rw-r--r-- | arch/x86/kernel/e820.c | 379 |
1 files changed, 58 insertions, 321 deletions
diff --git a/arch/x86/kernel/e820.c b/arch/x86/kernel/e820.c index 05ed7ab2ca48..7bca3c6a02fb 100644 --- a/arch/x86/kernel/e820.c +++ b/arch/x86/kernel/e820.c | |||
@@ -12,21 +12,13 @@ | |||
12 | #include <linux/types.h> | 12 | #include <linux/types.h> |
13 | #include <linux/init.h> | 13 | #include <linux/init.h> |
14 | #include <linux/bootmem.h> | 14 | #include <linux/bootmem.h> |
15 | #include <linux/ioport.h> | ||
16 | #include <linux/string.h> | ||
17 | #include <linux/kexec.h> | ||
18 | #include <linux/module.h> | ||
19 | #include <linux/mm.h> | ||
20 | #include <linux/pfn.h> | 15 | #include <linux/pfn.h> |
21 | #include <linux/suspend.h> | 16 | #include <linux/suspend.h> |
22 | #include <linux/firmware-map.h> | 17 | #include <linux/firmware-map.h> |
23 | 18 | ||
24 | #include <asm/pgtable.h> | ||
25 | #include <asm/page.h> | ||
26 | #include <asm/e820.h> | 19 | #include <asm/e820.h> |
27 | #include <asm/proto.h> | 20 | #include <asm/proto.h> |
28 | #include <asm/setup.h> | 21 | #include <asm/setup.h> |
29 | #include <asm/trampoline.h> | ||
30 | 22 | ||
31 | /* | 23 | /* |
32 | * The e820 map is the map that gets modified e.g. with command line parameters | 24 | * The e820 map is the map that gets modified e.g. with command line parameters |
@@ -517,31 +509,55 @@ u64 __init e820_remove_range(u64 start, u64 size, unsigned old_type, | |||
517 | int checktype) | 509 | int checktype) |
518 | { | 510 | { |
519 | int i; | 511 | int i; |
512 | u64 end; | ||
520 | u64 real_removed_size = 0; | 513 | u64 real_removed_size = 0; |
521 | 514 | ||
522 | if (size > (ULLONG_MAX - start)) | 515 | if (size > (ULLONG_MAX - start)) |
523 | size = ULLONG_MAX - start; | 516 | size = ULLONG_MAX - start; |
524 | 517 | ||
518 | end = start + size; | ||
519 | printk(KERN_DEBUG "e820 remove range: %016Lx - %016Lx ", | ||
520 | (unsigned long long) start, | ||
521 | (unsigned long long) end); | ||
522 | if (checktype) | ||
523 | e820_print_type(old_type); | ||
524 | printk(KERN_CONT "\n"); | ||
525 | |||
525 | for (i = 0; i < e820.nr_map; i++) { | 526 | for (i = 0; i < e820.nr_map; i++) { |
526 | struct e820entry *ei = &e820.map[i]; | 527 | struct e820entry *ei = &e820.map[i]; |
527 | u64 final_start, final_end; | 528 | u64 final_start, final_end; |
529 | u64 ei_end; | ||
528 | 530 | ||
529 | if (checktype && ei->type != old_type) | 531 | if (checktype && ei->type != old_type) |
530 | continue; | 532 | continue; |
533 | |||
534 | ei_end = ei->addr + ei->size; | ||
531 | /* totally covered? */ | 535 | /* totally covered? */ |
532 | if (ei->addr >= start && | 536 | if (ei->addr >= start && ei_end <= end) { |
533 | (ei->addr + ei->size) <= (start + size)) { | ||
534 | real_removed_size += ei->size; | 537 | real_removed_size += ei->size; |
535 | memset(ei, 0, sizeof(struct e820entry)); | 538 | memset(ei, 0, sizeof(struct e820entry)); |
536 | continue; | 539 | continue; |
537 | } | 540 | } |
541 | |||
542 | /* new range is totally covered? */ | ||
543 | if (ei->addr < start && ei_end > end) { | ||
544 | e820_add_region(end, ei_end - end, ei->type); | ||
545 | ei->size = start - ei->addr; | ||
546 | real_removed_size += size; | ||
547 | continue; | ||
548 | } | ||
549 | |||
538 | /* partially covered */ | 550 | /* partially covered */ |
539 | final_start = max(start, ei->addr); | 551 | final_start = max(start, ei->addr); |
540 | final_end = min(start + size, ei->addr + ei->size); | 552 | final_end = min(end, ei_end); |
541 | if (final_start >= final_end) | 553 | if (final_start >= final_end) |
542 | continue; | 554 | continue; |
543 | real_removed_size += final_end - final_start; | 555 | real_removed_size += final_end - final_start; |
544 | 556 | ||
557 | /* | ||
558 | * left range could be head or tail, so need to update | ||
559 | * size at first. | ||
560 | */ | ||
545 | ei->size -= final_end - final_start; | 561 | ei->size -= final_end - final_start; |
546 | if (ei->addr < final_start) | 562 | if (ei->addr < final_start) |
547 | continue; | 563 | continue; |
@@ -722,319 +738,44 @@ core_initcall(e820_mark_nvs_memory); | |||
722 | #endif | 738 | #endif |
723 | 739 | ||
724 | /* | 740 | /* |
725 | * Early reserved memory areas. | 741 | * Find a free area with specified alignment in a specific range. |
726 | */ | ||
727 | #define MAX_EARLY_RES 32 | ||
728 | |||
729 | struct early_res { | ||
730 | u64 start, end; | ||
731 | char name[16]; | ||
732 | char overlap_ok; | ||
733 | }; | ||
734 | static struct early_res early_res[MAX_EARLY_RES] __initdata = { | ||
735 | { 0, PAGE_SIZE, "BIOS data page", 1 }, /* BIOS data page */ | ||
736 | #ifdef CONFIG_X86_32 | ||
737 | /* | ||
738 | * But first pinch a few for the stack/trampoline stuff | ||
739 | * FIXME: Don't need the extra page at 4K, but need to fix | ||
740 | * trampoline before removing it. (see the GDT stuff) | ||
741 | */ | ||
742 | { PAGE_SIZE, PAGE_SIZE, "EX TRAMPOLINE", 1 }, | ||
743 | #endif | ||
744 | |||
745 | {} | ||
746 | }; | ||
747 | |||
748 | static int __init find_overlapped_early(u64 start, u64 end) | ||
749 | { | ||
750 | int i; | ||
751 | struct early_res *r; | ||
752 | |||
753 | for (i = 0; i < MAX_EARLY_RES && early_res[i].end; i++) { | ||
754 | r = &early_res[i]; | ||
755 | if (end > r->start && start < r->end) | ||
756 | break; | ||
757 | } | ||
758 | |||
759 | return i; | ||
760 | } | ||
761 | |||
762 | /* | ||
763 | * Drop the i-th range from the early reservation map, | ||
764 | * by copying any higher ranges down one over it, and | ||
765 | * clearing what had been the last slot. | ||
766 | */ | ||
767 | static void __init drop_range(int i) | ||
768 | { | ||
769 | int j; | ||
770 | |||
771 | for (j = i + 1; j < MAX_EARLY_RES && early_res[j].end; j++) | ||
772 | ; | ||
773 | |||
774 | memmove(&early_res[i], &early_res[i + 1], | ||
775 | (j - 1 - i) * sizeof(struct early_res)); | ||
776 | |||
777 | early_res[j - 1].end = 0; | ||
778 | } | ||
779 | |||
780 | /* | ||
781 | * Split any existing ranges that: | ||
782 | * 1) are marked 'overlap_ok', and | ||
783 | * 2) overlap with the stated range [start, end) | ||
784 | * into whatever portion (if any) of the existing range is entirely | ||
785 | * below or entirely above the stated range. Drop the portion | ||
786 | * of the existing range that overlaps with the stated range, | ||
787 | * which will allow the caller of this routine to then add that | ||
788 | * stated range without conflicting with any existing range. | ||
789 | */ | 742 | */ |
790 | static void __init drop_overlaps_that_are_ok(u64 start, u64 end) | 743 | u64 __init find_e820_area(u64 start, u64 end, u64 size, u64 align) |
791 | { | 744 | { |
792 | int i; | 745 | int i; |
793 | struct early_res *r; | ||
794 | u64 lower_start, lower_end; | ||
795 | u64 upper_start, upper_end; | ||
796 | char name[16]; | ||
797 | 746 | ||
798 | for (i = 0; i < MAX_EARLY_RES && early_res[i].end; i++) { | 747 | for (i = 0; i < e820.nr_map; i++) { |
799 | r = &early_res[i]; | 748 | struct e820entry *ei = &e820.map[i]; |
749 | u64 addr; | ||
750 | u64 ei_start, ei_last; | ||
800 | 751 | ||
801 | /* Continue past non-overlapping ranges */ | 752 | if (ei->type != E820_RAM) |
802 | if (end <= r->start || start >= r->end) | ||
803 | continue; | 753 | continue; |
804 | 754 | ||
805 | /* | 755 | ei_last = ei->addr + ei->size; |
806 | * Leave non-ok overlaps as is; let caller | 756 | ei_start = ei->addr; |
807 | * panic "Overlapping early reservations" | 757 | addr = find_early_area(ei_start, ei_last, start, end, |
808 | * when it hits this overlap. | 758 | size, align); |
809 | */ | ||
810 | if (!r->overlap_ok) | ||
811 | return; | ||
812 | |||
813 | /* | ||
814 | * We have an ok overlap. We will drop it from the early | ||
815 | * reservation map, and add back in any non-overlapping | ||
816 | * portions (lower or upper) as separate, overlap_ok, | ||
817 | * non-overlapping ranges. | ||
818 | */ | ||
819 | |||
820 | /* 1. Note any non-overlapping (lower or upper) ranges. */ | ||
821 | strncpy(name, r->name, sizeof(name) - 1); | ||
822 | |||
823 | lower_start = lower_end = 0; | ||
824 | upper_start = upper_end = 0; | ||
825 | if (r->start < start) { | ||
826 | lower_start = r->start; | ||
827 | lower_end = start; | ||
828 | } | ||
829 | if (r->end > end) { | ||
830 | upper_start = end; | ||
831 | upper_end = r->end; | ||
832 | } | ||
833 | |||
834 | /* 2. Drop the original ok overlapping range */ | ||
835 | drop_range(i); | ||
836 | |||
837 | i--; /* resume for-loop on copied down entry */ | ||
838 | |||
839 | /* 3. Add back in any non-overlapping ranges. */ | ||
840 | if (lower_end) | ||
841 | reserve_early_overlap_ok(lower_start, lower_end, name); | ||
842 | if (upper_end) | ||
843 | reserve_early_overlap_ok(upper_start, upper_end, name); | ||
844 | } | ||
845 | } | ||
846 | |||
847 | static void __init __reserve_early(u64 start, u64 end, char *name, | ||
848 | int overlap_ok) | ||
849 | { | ||
850 | int i; | ||
851 | struct early_res *r; | ||
852 | |||
853 | i = find_overlapped_early(start, end); | ||
854 | if (i >= MAX_EARLY_RES) | ||
855 | panic("Too many early reservations"); | ||
856 | r = &early_res[i]; | ||
857 | if (r->end) | ||
858 | panic("Overlapping early reservations " | ||
859 | "%llx-%llx %s to %llx-%llx %s\n", | ||
860 | start, end - 1, name?name:"", r->start, | ||
861 | r->end - 1, r->name); | ||
862 | r->start = start; | ||
863 | r->end = end; | ||
864 | r->overlap_ok = overlap_ok; | ||
865 | if (name) | ||
866 | strncpy(r->name, name, sizeof(r->name) - 1); | ||
867 | } | ||
868 | |||
869 | /* | ||
870 | * A few early reservtations come here. | ||
871 | * | ||
872 | * The 'overlap_ok' in the name of this routine does -not- mean it | ||
873 | * is ok for these reservations to overlap an earlier reservation. | ||
874 | * Rather it means that it is ok for subsequent reservations to | ||
875 | * overlap this one. | ||
876 | * | ||
877 | * Use this entry point to reserve early ranges when you are doing | ||
878 | * so out of "Paranoia", reserving perhaps more memory than you need, | ||
879 | * just in case, and don't mind a subsequent overlapping reservation | ||
880 | * that is known to be needed. | ||
881 | * | ||
882 | * The drop_overlaps_that_are_ok() call here isn't really needed. | ||
883 | * It would be needed if we had two colliding 'overlap_ok' | ||
884 | * reservations, so that the second such would not panic on the | ||
885 | * overlap with the first. We don't have any such as of this | ||
886 | * writing, but might as well tolerate such if it happens in | ||
887 | * the future. | ||
888 | */ | ||
889 | void __init reserve_early_overlap_ok(u64 start, u64 end, char *name) | ||
890 | { | ||
891 | drop_overlaps_that_are_ok(start, end); | ||
892 | __reserve_early(start, end, name, 1); | ||
893 | } | ||
894 | |||
895 | /* | ||
896 | * Most early reservations come here. | ||
897 | * | ||
898 | * We first have drop_overlaps_that_are_ok() drop any pre-existing | ||
899 | * 'overlap_ok' ranges, so that we can then reserve this memory | ||
900 | * range without risk of panic'ing on an overlapping overlap_ok | ||
901 | * early reservation. | ||
902 | */ | ||
903 | void __init reserve_early(u64 start, u64 end, char *name) | ||
904 | { | ||
905 | if (start >= end) | ||
906 | return; | ||
907 | |||
908 | drop_overlaps_that_are_ok(start, end); | ||
909 | __reserve_early(start, end, name, 0); | ||
910 | } | ||
911 | |||
912 | void __init free_early(u64 start, u64 end) | ||
913 | { | ||
914 | struct early_res *r; | ||
915 | int i; | ||
916 | |||
917 | i = find_overlapped_early(start, end); | ||
918 | r = &early_res[i]; | ||
919 | if (i >= MAX_EARLY_RES || r->end != end || r->start != start) | ||
920 | panic("free_early on not reserved area: %llx-%llx!", | ||
921 | start, end - 1); | ||
922 | |||
923 | drop_range(i); | ||
924 | } | ||
925 | |||
926 | void __init early_res_to_bootmem(u64 start, u64 end) | ||
927 | { | ||
928 | int i, count; | ||
929 | u64 final_start, final_end; | ||
930 | |||
931 | count = 0; | ||
932 | for (i = 0; i < MAX_EARLY_RES && early_res[i].end; i++) | ||
933 | count++; | ||
934 | |||
935 | printk(KERN_INFO "(%d early reservations) ==> bootmem [%010llx - %010llx]\n", | ||
936 | count, start, end); | ||
937 | for (i = 0; i < count; i++) { | ||
938 | struct early_res *r = &early_res[i]; | ||
939 | printk(KERN_INFO " #%d [%010llx - %010llx] %16s", i, | ||
940 | r->start, r->end, r->name); | ||
941 | final_start = max(start, r->start); | ||
942 | final_end = min(end, r->end); | ||
943 | if (final_start >= final_end) { | ||
944 | printk(KERN_CONT "\n"); | ||
945 | continue; | ||
946 | } | ||
947 | printk(KERN_CONT " ==> [%010llx - %010llx]\n", | ||
948 | final_start, final_end); | ||
949 | reserve_bootmem_generic(final_start, final_end - final_start, | ||
950 | BOOTMEM_DEFAULT); | ||
951 | } | ||
952 | } | ||
953 | 759 | ||
954 | /* Check for already reserved areas */ | 760 | if (addr != -1ULL) |
955 | static inline int __init bad_addr(u64 *addrp, u64 size, u64 align) | 761 | return addr; |
956 | { | ||
957 | int i; | ||
958 | u64 addr = *addrp; | ||
959 | int changed = 0; | ||
960 | struct early_res *r; | ||
961 | again: | ||
962 | i = find_overlapped_early(addr, addr + size); | ||
963 | r = &early_res[i]; | ||
964 | if (i < MAX_EARLY_RES && r->end) { | ||
965 | *addrp = addr = round_up(r->end, align); | ||
966 | changed = 1; | ||
967 | goto again; | ||
968 | } | 762 | } |
969 | return changed; | 763 | return -1ULL; |
970 | } | 764 | } |
971 | 765 | ||
972 | /* Check for already reserved areas */ | 766 | u64 __init find_fw_memmap_area(u64 start, u64 end, u64 size, u64 align) |
973 | static inline int __init bad_addr_size(u64 *addrp, u64 *sizep, u64 align) | ||
974 | { | 767 | { |
975 | int i; | 768 | return find_e820_area(start, end, size, align); |
976 | u64 addr = *addrp, last; | ||
977 | u64 size = *sizep; | ||
978 | int changed = 0; | ||
979 | again: | ||
980 | last = addr + size; | ||
981 | for (i = 0; i < MAX_EARLY_RES && early_res[i].end; i++) { | ||
982 | struct early_res *r = &early_res[i]; | ||
983 | if (last > r->start && addr < r->start) { | ||
984 | size = r->start - addr; | ||
985 | changed = 1; | ||
986 | goto again; | ||
987 | } | ||
988 | if (last > r->end && addr < r->end) { | ||
989 | addr = round_up(r->end, align); | ||
990 | size = last - addr; | ||
991 | changed = 1; | ||
992 | goto again; | ||
993 | } | ||
994 | if (last <= r->end && addr >= r->start) { | ||
995 | (*sizep)++; | ||
996 | return 0; | ||
997 | } | ||
998 | } | ||
999 | if (changed) { | ||
1000 | *addrp = addr; | ||
1001 | *sizep = size; | ||
1002 | } | ||
1003 | return changed; | ||
1004 | } | 769 | } |
1005 | 770 | ||
1006 | /* | 771 | u64 __init get_max_mapped(void) |
1007 | * Find a free area with specified alignment in a specific range. | ||
1008 | */ | ||
1009 | u64 __init find_e820_area(u64 start, u64 end, u64 size, u64 align) | ||
1010 | { | 772 | { |
1011 | int i; | 773 | u64 end = max_pfn_mapped; |
1012 | 774 | ||
1013 | for (i = 0; i < e820.nr_map; i++) { | 775 | end <<= PAGE_SHIFT; |
1014 | struct e820entry *ei = &e820.map[i]; | ||
1015 | u64 addr, last; | ||
1016 | u64 ei_last; | ||
1017 | 776 | ||
1018 | if (ei->type != E820_RAM) | 777 | return end; |
1019 | continue; | ||
1020 | addr = round_up(ei->addr, align); | ||
1021 | ei_last = ei->addr + ei->size; | ||
1022 | if (addr < start) | ||
1023 | addr = round_up(start, align); | ||
1024 | if (addr >= ei_last) | ||
1025 | continue; | ||
1026 | while (bad_addr(&addr, size, align) && addr+size <= ei_last) | ||
1027 | ; | ||
1028 | last = addr + size; | ||
1029 | if (last > ei_last) | ||
1030 | continue; | ||
1031 | if (last > end) | ||
1032 | continue; | ||
1033 | return addr; | ||
1034 | } | ||
1035 | return -1ULL; | ||
1036 | } | 778 | } |
1037 | |||
1038 | /* | 779 | /* |
1039 | * Find next free range after *start | 780 | * Find next free range after *start |
1040 | */ | 781 | */ |
@@ -1044,25 +785,19 @@ u64 __init find_e820_area_size(u64 start, u64 *sizep, u64 align) | |||
1044 | 785 | ||
1045 | for (i = 0; i < e820.nr_map; i++) { | 786 | for (i = 0; i < e820.nr_map; i++) { |
1046 | struct e820entry *ei = &e820.map[i]; | 787 | struct e820entry *ei = &e820.map[i]; |
1047 | u64 addr, last; | 788 | u64 addr; |
1048 | u64 ei_last; | 789 | u64 ei_start, ei_last; |
1049 | 790 | ||
1050 | if (ei->type != E820_RAM) | 791 | if (ei->type != E820_RAM) |
1051 | continue; | 792 | continue; |
1052 | addr = round_up(ei->addr, align); | 793 | |
1053 | ei_last = ei->addr + ei->size; | 794 | ei_last = ei->addr + ei->size; |
1054 | if (addr < start) | 795 | ei_start = ei->addr; |
1055 | addr = round_up(start, align); | 796 | addr = find_early_area_size(ei_start, ei_last, start, |
1056 | if (addr >= ei_last) | 797 | sizep, align); |
1057 | continue; | 798 | |
1058 | *sizep = ei_last - addr; | 799 | if (addr != -1ULL) |
1059 | while (bad_addr_size(&addr, sizep, align) && | 800 | return addr; |
1060 | addr + *sizep <= ei_last) | ||
1061 | ; | ||
1062 | last = addr + *sizep; | ||
1063 | if (last > ei_last) | ||
1064 | continue; | ||
1065 | return addr; | ||
1066 | } | 801 | } |
1067 | 802 | ||
1068 | return -1ULL; | 803 | return -1ULL; |
@@ -1421,6 +1156,8 @@ void __init e820_reserve_resources_late(void) | |||
1421 | end = MAX_RESOURCE_SIZE; | 1156 | end = MAX_RESOURCE_SIZE; |
1422 | if (start >= end) | 1157 | if (start >= end) |
1423 | continue; | 1158 | continue; |
1159 | printk(KERN_DEBUG "reserve RAM buffer: %016llx - %016llx ", | ||
1160 | start, end); | ||
1424 | reserve_region_with_split(&iomem_resource, start, end, | 1161 | reserve_region_with_split(&iomem_resource, start, end, |
1425 | "RAM buffer"); | 1162 | "RAM buffer"); |
1426 | } | 1163 | } |