aboutsummaryrefslogtreecommitdiffstats
path: root/arch/x86/mm
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2010-10-21 21:52:11 -0400
committerLinus Torvalds <torvalds@linux-foundation.org>2010-10-21 21:52:11 -0400
commit3044100e58c84e133791c8b60a2f5bef69d732e4 (patch)
treef9ed0d1f3df89c31dd81ccaf0cf3478f57b08440 /arch/x86/mm
parentb5153163ed580e00c67bdfecb02b2e3843817b3e (diff)
parent67e87f0a1c5cbc750f81ebf6a128e8ff6f4376cc (diff)
Merge branch 'core-memblock-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip
* 'core-memblock-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (74 commits) x86-64: Only set max_pfn_mapped to 512 MiB if we enter via head_64.S xen: Cope with unmapped pages when initializing kernel pagetable memblock, bootmem: Round pfn properly for memory and reserved regions memblock: Annotate memblock functions with __init_memblock memblock: Allow memblock_init to be called early memblock/arm: Fix memblock_region_is_memory() typo x86, memblock: Remove __memblock_x86_find_in_range_size() memblock: Fix wraparound in find_region() x86-32, memblock: Make add_highpages honor early reserved ranges x86, memblock: Fix crashkernel allocation arm, memblock: Fix the sparsemem build memblock: Fix section mismatch warnings powerpc, memblock: Fix memblock API change fallout memblock, microblaze: Fix memblock API change fallout x86: Remove old bootmem code x86, memblock: Use memblock_memory_size()/memblock_free_memory_size() to get correct dma_reserve x86: Remove not used early_res code x86, memblock: Replace e820_/_early string with memblock_ x86: Use memblock to replace early_res x86, memblock: Use memblock_debug to control debug message print out ... Fix up trivial conflicts in arch/x86/kernel/setup.c and kernel/Makefile
Diffstat (limited to 'arch/x86/mm')
-rw-r--r--arch/x86/mm/Makefile2
-rw-r--r--arch/x86/mm/init.c10
-rw-r--r--arch/x86/mm/init_32.c119
-rw-r--r--arch/x86/mm/init_64.c67
-rw-r--r--arch/x86/mm/ioremap.c5
-rw-r--r--arch/x86/mm/k8topology_64.c4
-rw-r--r--arch/x86/mm/memblock.c348
-rw-r--r--arch/x86/mm/memtest.c7
-rw-r--r--arch/x86/mm/numa_32.c30
-rw-r--r--arch/x86/mm/numa_64.c84
-rw-r--r--arch/x86/mm/srat_32.c3
-rw-r--r--arch/x86/mm/srat_64.c11
12 files changed, 428 insertions, 262 deletions
diff --git a/arch/x86/mm/Makefile b/arch/x86/mm/Makefile
index a4c768397ba..55543397a8a 100644
--- a/arch/x86/mm/Makefile
+++ b/arch/x86/mm/Makefile
@@ -26,4 +26,6 @@ obj-$(CONFIG_NUMA) += numa.o numa_$(BITS).o
26obj-$(CONFIG_K8_NUMA) += k8topology_64.o 26obj-$(CONFIG_K8_NUMA) += k8topology_64.o
27obj-$(CONFIG_ACPI_NUMA) += srat_$(BITS).o 27obj-$(CONFIG_ACPI_NUMA) += srat_$(BITS).o
28 28
29obj-$(CONFIG_HAVE_MEMBLOCK) += memblock.o
30
29obj-$(CONFIG_MEMTEST) += memtest.o 31obj-$(CONFIG_MEMTEST) += memtest.o
diff --git a/arch/x86/mm/init.c b/arch/x86/mm/init.c
index b278535b14a..c0e28a13de7 100644
--- a/arch/x86/mm/init.c
+++ b/arch/x86/mm/init.c
@@ -2,6 +2,7 @@
2#include <linux/initrd.h> 2#include <linux/initrd.h>
3#include <linux/ioport.h> 3#include <linux/ioport.h>
4#include <linux/swap.h> 4#include <linux/swap.h>
5#include <linux/memblock.h>
5 6
6#include <asm/cacheflush.h> 7#include <asm/cacheflush.h>
7#include <asm/e820.h> 8#include <asm/e820.h>
@@ -33,6 +34,7 @@ static void __init find_early_table_space(unsigned long end, int use_pse,
33 int use_gbpages) 34 int use_gbpages)
34{ 35{
35 unsigned long puds, pmds, ptes, tables, start; 36 unsigned long puds, pmds, ptes, tables, start;
37 phys_addr_t base;
36 38
37 puds = (end + PUD_SIZE - 1) >> PUD_SHIFT; 39 puds = (end + PUD_SIZE - 1) >> PUD_SHIFT;
38 tables = roundup(puds * sizeof(pud_t), PAGE_SIZE); 40 tables = roundup(puds * sizeof(pud_t), PAGE_SIZE);
@@ -75,12 +77,12 @@ static void __init find_early_table_space(unsigned long end, int use_pse,
75#else 77#else
76 start = 0x8000; 78 start = 0x8000;
77#endif 79#endif
78 e820_table_start = find_e820_area(start, max_pfn_mapped<<PAGE_SHIFT, 80 base = memblock_find_in_range(start, max_pfn_mapped<<PAGE_SHIFT,
79 tables, PAGE_SIZE); 81 tables, PAGE_SIZE);
80 if (e820_table_start == -1UL) 82 if (base == MEMBLOCK_ERROR)
81 panic("Cannot find space for the kernel page tables"); 83 panic("Cannot find space for the kernel page tables");
82 84
83 e820_table_start >>= PAGE_SHIFT; 85 e820_table_start = base >> PAGE_SHIFT;
84 e820_table_end = e820_table_start; 86 e820_table_end = e820_table_start;
85 e820_table_top = e820_table_start + (tables >> PAGE_SHIFT); 87 e820_table_top = e820_table_start + (tables >> PAGE_SHIFT);
86 88
@@ -299,7 +301,7 @@ unsigned long __init_refok init_memory_mapping(unsigned long start,
299 __flush_tlb_all(); 301 __flush_tlb_all();
300 302
301 if (!after_bootmem && e820_table_end > e820_table_start) 303 if (!after_bootmem && e820_table_end > e820_table_start)
302 reserve_early(e820_table_start << PAGE_SHIFT, 304 memblock_x86_reserve_range(e820_table_start << PAGE_SHIFT,
303 e820_table_end << PAGE_SHIFT, "PGTABLE"); 305 e820_table_end << PAGE_SHIFT, "PGTABLE");
304 306
305 if (!after_bootmem) 307 if (!after_bootmem)
diff --git a/arch/x86/mm/init_32.c b/arch/x86/mm/init_32.c
index 558f2d33207..5d0a6711c28 100644
--- a/arch/x86/mm/init_32.c
+++ b/arch/x86/mm/init_32.c
@@ -25,6 +25,7 @@
25#include <linux/pfn.h> 25#include <linux/pfn.h>
26#include <linux/poison.h> 26#include <linux/poison.h>
27#include <linux/bootmem.h> 27#include <linux/bootmem.h>
28#include <linux/memblock.h>
28#include <linux/proc_fs.h> 29#include <linux/proc_fs.h>
29#include <linux/memory_hotplug.h> 30#include <linux/memory_hotplug.h>
30#include <linux/initrd.h> 31#include <linux/initrd.h>
@@ -422,49 +423,28 @@ static void __init add_one_highpage_init(struct page *page)
422 totalhigh_pages++; 423 totalhigh_pages++;
423} 424}
424 425
425struct add_highpages_data { 426void __init add_highpages_with_active_regions(int nid,
426 unsigned long start_pfn; 427 unsigned long start_pfn, unsigned long end_pfn)
427 unsigned long end_pfn;
428};
429
430static int __init add_highpages_work_fn(unsigned long start_pfn,
431 unsigned long end_pfn, void *datax)
432{ 428{
433 int node_pfn; 429 struct range *range;
434 struct page *page; 430 int nr_range;
435 unsigned long final_start_pfn, final_end_pfn; 431 int i;
436 struct add_highpages_data *data;
437 432
438 data = (struct add_highpages_data *)datax; 433 nr_range = __get_free_all_memory_range(&range, nid, start_pfn, end_pfn);
439 434
440 final_start_pfn = max(start_pfn, data->start_pfn); 435 for (i = 0; i < nr_range; i++) {
441 final_end_pfn = min(end_pfn, data->end_pfn); 436 struct page *page;
442 if (final_start_pfn >= final_end_pfn) 437 int node_pfn;
443 return 0;
444 438
445 for (node_pfn = final_start_pfn; node_pfn < final_end_pfn; 439 for (node_pfn = range[i].start; node_pfn < range[i].end;
446 node_pfn++) { 440 node_pfn++) {
447 if (!pfn_valid(node_pfn)) 441 if (!pfn_valid(node_pfn))
448 continue; 442 continue;
449 page = pfn_to_page(node_pfn); 443 page = pfn_to_page(node_pfn);
450 add_one_highpage_init(page); 444 add_one_highpage_init(page);
445 }
451 } 446 }
452
453 return 0;
454
455} 447}
456
457void __init add_highpages_with_active_regions(int nid, unsigned long start_pfn,
458 unsigned long end_pfn)
459{
460 struct add_highpages_data data;
461
462 data.start_pfn = start_pfn;
463 data.end_pfn = end_pfn;
464
465 work_with_active_regions(nid, add_highpages_work_fn, &data);
466}
467
468#else 448#else
469static inline void permanent_kmaps_init(pgd_t *pgd_base) 449static inline void permanent_kmaps_init(pgd_t *pgd_base)
470{ 450{
@@ -712,14 +692,14 @@ void __init initmem_init(unsigned long start_pfn, unsigned long end_pfn,
712 highstart_pfn = highend_pfn = max_pfn; 692 highstart_pfn = highend_pfn = max_pfn;
713 if (max_pfn > max_low_pfn) 693 if (max_pfn > max_low_pfn)
714 highstart_pfn = max_low_pfn; 694 highstart_pfn = max_low_pfn;
715 e820_register_active_regions(0, 0, highend_pfn); 695 memblock_x86_register_active_regions(0, 0, highend_pfn);
716 sparse_memory_present_with_active_regions(0); 696 sparse_memory_present_with_active_regions(0);
717 printk(KERN_NOTICE "%ldMB HIGHMEM available.\n", 697 printk(KERN_NOTICE "%ldMB HIGHMEM available.\n",
718 pages_to_mb(highend_pfn - highstart_pfn)); 698 pages_to_mb(highend_pfn - highstart_pfn));
719 num_physpages = highend_pfn; 699 num_physpages = highend_pfn;
720 high_memory = (void *) __va(highstart_pfn * PAGE_SIZE - 1) + 1; 700 high_memory = (void *) __va(highstart_pfn * PAGE_SIZE - 1) + 1;
721#else 701#else
722 e820_register_active_regions(0, 0, max_low_pfn); 702 memblock_x86_register_active_regions(0, 0, max_low_pfn);
723 sparse_memory_present_with_active_regions(0); 703 sparse_memory_present_with_active_regions(0);
724 num_physpages = max_low_pfn; 704 num_physpages = max_low_pfn;
725 high_memory = (void *) __va(max_low_pfn * PAGE_SIZE - 1) + 1; 705 high_memory = (void *) __va(max_low_pfn * PAGE_SIZE - 1) + 1;
@@ -750,68 +730,12 @@ static void __init zone_sizes_init(void)
750 free_area_init_nodes(max_zone_pfns); 730 free_area_init_nodes(max_zone_pfns);
751} 731}
752 732
753#ifndef CONFIG_NO_BOOTMEM
754static unsigned long __init setup_node_bootmem(int nodeid,
755 unsigned long start_pfn,
756 unsigned long end_pfn,
757 unsigned long bootmap)
758{
759 unsigned long bootmap_size;
760
761 /* don't touch min_low_pfn */
762 bootmap_size = init_bootmem_node(NODE_DATA(nodeid),
763 bootmap >> PAGE_SHIFT,
764 start_pfn, end_pfn);
765 printk(KERN_INFO " node %d low ram: %08lx - %08lx\n",
766 nodeid, start_pfn<<PAGE_SHIFT, end_pfn<<PAGE_SHIFT);
767 printk(KERN_INFO " node %d bootmap %08lx - %08lx\n",
768 nodeid, bootmap, bootmap + bootmap_size);
769 free_bootmem_with_active_regions(nodeid, end_pfn);
770
771 return bootmap + bootmap_size;
772}
773#endif
774
775void __init setup_bootmem_allocator(void) 733void __init setup_bootmem_allocator(void)
776{ 734{
777#ifndef CONFIG_NO_BOOTMEM
778 int nodeid;
779 unsigned long bootmap_size, bootmap;
780 /*
781 * Initialize the boot-time allocator (with low memory only):
782 */
783 bootmap_size = bootmem_bootmap_pages(max_low_pfn)<<PAGE_SHIFT;
784 bootmap = find_e820_area(0, max_pfn_mapped<<PAGE_SHIFT, bootmap_size,
785 PAGE_SIZE);
786 if (bootmap == -1L)
787 panic("Cannot find bootmem map of size %ld\n", bootmap_size);
788 reserve_early(bootmap, bootmap + bootmap_size, "BOOTMAP");
789#endif
790
791 printk(KERN_INFO " mapped low ram: 0 - %08lx\n", 735 printk(KERN_INFO " mapped low ram: 0 - %08lx\n",
792 max_pfn_mapped<<PAGE_SHIFT); 736 max_pfn_mapped<<PAGE_SHIFT);
793 printk(KERN_INFO " low ram: 0 - %08lx\n", max_low_pfn<<PAGE_SHIFT); 737 printk(KERN_INFO " low ram: 0 - %08lx\n", max_low_pfn<<PAGE_SHIFT);
794 738
795#ifndef CONFIG_NO_BOOTMEM
796 for_each_online_node(nodeid) {
797 unsigned long start_pfn, end_pfn;
798
799#ifdef CONFIG_NEED_MULTIPLE_NODES
800 start_pfn = node_start_pfn[nodeid];
801 end_pfn = node_end_pfn[nodeid];
802 if (start_pfn > max_low_pfn)
803 continue;
804 if (end_pfn > max_low_pfn)
805 end_pfn = max_low_pfn;
806#else
807 start_pfn = 0;
808 end_pfn = max_low_pfn;
809#endif
810 bootmap = setup_node_bootmem(nodeid, start_pfn, end_pfn,
811 bootmap);
812 }
813#endif
814
815 after_bootmem = 1; 739 after_bootmem = 1;
816} 740}
817 741
@@ -1070,8 +994,3 @@ void mark_rodata_ro(void)
1070} 994}
1071#endif 995#endif
1072 996
1073int __init reserve_bootmem_generic(unsigned long phys, unsigned long len,
1074 int flags)
1075{
1076 return reserve_bootmem(phys, len, flags);
1077}
diff --git a/arch/x86/mm/init_64.c b/arch/x86/mm/init_64.c
index c55f900fbf8..84346200e78 100644
--- a/arch/x86/mm/init_64.c
+++ b/arch/x86/mm/init_64.c
@@ -21,6 +21,7 @@
21#include <linux/initrd.h> 21#include <linux/initrd.h>
22#include <linux/pagemap.h> 22#include <linux/pagemap.h>
23#include <linux/bootmem.h> 23#include <linux/bootmem.h>
24#include <linux/memblock.h>
24#include <linux/proc_fs.h> 25#include <linux/proc_fs.h>
25#include <linux/pci.h> 26#include <linux/pci.h>
26#include <linux/pfn.h> 27#include <linux/pfn.h>
@@ -52,8 +53,6 @@
52#include <asm/init.h> 53#include <asm/init.h>
53#include <linux/bootmem.h> 54#include <linux/bootmem.h>
54 55
55static unsigned long dma_reserve __initdata;
56
57static int __init parse_direct_gbpages_off(char *arg) 56static int __init parse_direct_gbpages_off(char *arg)
58{ 57{
59 direct_gbpages = 0; 58 direct_gbpages = 0;
@@ -617,23 +616,7 @@ kernel_physical_mapping_init(unsigned long start,
617void __init initmem_init(unsigned long start_pfn, unsigned long end_pfn, 616void __init initmem_init(unsigned long start_pfn, unsigned long end_pfn,
618 int acpi, int k8) 617 int acpi, int k8)
619{ 618{
620#ifndef CONFIG_NO_BOOTMEM 619 memblock_x86_register_active_regions(0, start_pfn, end_pfn);
621 unsigned long bootmap_size, bootmap;
622
623 bootmap_size = bootmem_bootmap_pages(end_pfn)<<PAGE_SHIFT;
624 bootmap = find_e820_area(0, end_pfn<<PAGE_SHIFT, bootmap_size,
625 PAGE_SIZE);
626 if (bootmap == -1L)
627 panic("Cannot find bootmem map of size %ld\n", bootmap_size);
628 reserve_early(bootmap, bootmap + bootmap_size, "BOOTMAP");
629 /* don't touch min_low_pfn */
630 bootmap_size = init_bootmem_node(NODE_DATA(0), bootmap >> PAGE_SHIFT,
631 0, end_pfn);
632 e820_register_active_regions(0, start_pfn, end_pfn);
633 free_bootmem_with_active_regions(0, end_pfn);
634#else
635 e820_register_active_regions(0, start_pfn, end_pfn);
636#endif
637} 620}
638#endif 621#endif
639 622
@@ -843,52 +826,6 @@ void mark_rodata_ro(void)
843 826
844#endif 827#endif
845 828
846int __init reserve_bootmem_generic(unsigned long phys, unsigned long len,
847 int flags)
848{
849#ifdef CONFIG_NUMA
850 int nid, next_nid;
851 int ret;
852#endif
853 unsigned long pfn = phys >> PAGE_SHIFT;
854
855 if (pfn >= max_pfn) {
856 /*
857 * This can happen with kdump kernels when accessing
858 * firmware tables:
859 */
860 if (pfn < max_pfn_mapped)
861 return -EFAULT;
862
863 printk(KERN_ERR "reserve_bootmem: illegal reserve %lx %lu\n",
864 phys, len);
865 return -EFAULT;
866 }
867
868 /* Should check here against the e820 map to avoid double free */
869#ifdef CONFIG_NUMA
870 nid = phys_to_nid(phys);
871 next_nid = phys_to_nid(phys + len - 1);
872 if (nid == next_nid)
873 ret = reserve_bootmem_node(NODE_DATA(nid), phys, len, flags);
874 else
875 ret = reserve_bootmem(phys, len, flags);
876
877 if (ret != 0)
878 return ret;
879
880#else
881 reserve_bootmem(phys, len, flags);
882#endif
883
884 if (phys+len <= MAX_DMA_PFN*PAGE_SIZE) {
885 dma_reserve += len / PAGE_SIZE;
886 set_dma_reserve(dma_reserve);
887 }
888
889 return 0;
890}
891
892int kern_addr_valid(unsigned long addr) 829int kern_addr_valid(unsigned long addr)
893{ 830{
894 unsigned long above = ((long)addr) >> __VIRTUAL_MASK_SHIFT; 831 unsigned long above = ((long)addr) >> __VIRTUAL_MASK_SHIFT;
diff --git a/arch/x86/mm/ioremap.c b/arch/x86/mm/ioremap.c
index 3ba6e0608c5..0369843511d 100644
--- a/arch/x86/mm/ioremap.c
+++ b/arch/x86/mm/ioremap.c
@@ -362,6 +362,11 @@ static inline pte_t * __init early_ioremap_pte(unsigned long addr)
362 return &bm_pte[pte_index(addr)]; 362 return &bm_pte[pte_index(addr)];
363} 363}
364 364
365bool __init is_early_ioremap_ptep(pte_t *ptep)
366{
367 return ptep >= &bm_pte[0] && ptep < &bm_pte[PAGE_SIZE/sizeof(pte_t)];
368}
369
365static unsigned long slot_virt[FIX_BTMAPS_SLOTS] __initdata; 370static unsigned long slot_virt[FIX_BTMAPS_SLOTS] __initdata;
366 371
367void __init early_ioremap_init(void) 372void __init early_ioremap_init(void)
diff --git a/arch/x86/mm/k8topology_64.c b/arch/x86/mm/k8topology_64.c
index 52d54bfc1eb..804a3b6c6e1 100644
--- a/arch/x86/mm/k8topology_64.c
+++ b/arch/x86/mm/k8topology_64.c
@@ -11,6 +11,8 @@
11#include <linux/string.h> 11#include <linux/string.h>
12#include <linux/module.h> 12#include <linux/module.h>
13#include <linux/nodemask.h> 13#include <linux/nodemask.h>
14#include <linux/memblock.h>
15
14#include <asm/io.h> 16#include <asm/io.h>
15#include <linux/pci_ids.h> 17#include <linux/pci_ids.h>
16#include <linux/acpi.h> 18#include <linux/acpi.h>
@@ -222,7 +224,7 @@ int __init k8_scan_nodes(void)
222 for_each_node_mask(i, node_possible_map) { 224 for_each_node_mask(i, node_possible_map) {
223 int j; 225 int j;
224 226
225 e820_register_active_regions(i, 227 memblock_x86_register_active_regions(i,
226 nodes[i].start >> PAGE_SHIFT, 228 nodes[i].start >> PAGE_SHIFT,
227 nodes[i].end >> PAGE_SHIFT); 229 nodes[i].end >> PAGE_SHIFT);
228 for (j = apicid_base; j < cores + apicid_base; j++) 230 for (j = apicid_base; j < cores + apicid_base; j++)
diff --git a/arch/x86/mm/memblock.c b/arch/x86/mm/memblock.c
new file mode 100644
index 00000000000..aa1169392b8
--- /dev/null
+++ b/arch/x86/mm/memblock.c
@@ -0,0 +1,348 @@
1#include <linux/kernel.h>
2#include <linux/types.h>
3#include <linux/init.h>
4#include <linux/bitops.h>
5#include <linux/memblock.h>
6#include <linux/bootmem.h>
7#include <linux/mm.h>
8#include <linux/range.h>
9
10/* Check for already reserved areas */
11static bool __init check_with_memblock_reserved_size(u64 *addrp, u64 *sizep, u64 align)
12{
13 struct memblock_region *r;
14 u64 addr = *addrp, last;
15 u64 size = *sizep;
16 bool changed = false;
17
18again:
19 last = addr + size;
20 for_each_memblock(reserved, r) {
21 if (last > r->base && addr < r->base) {
22 size = r->base - addr;
23 changed = true;
24 goto again;
25 }
26 if (last > (r->base + r->size) && addr < (r->base + r->size)) {
27 addr = round_up(r->base + r->size, align);
28 size = last - addr;
29 changed = true;
30 goto again;
31 }
32 if (last <= (r->base + r->size) && addr >= r->base) {
33 *sizep = 0;
34 return false;
35 }
36 }
37 if (changed) {
38 *addrp = addr;
39 *sizep = size;
40 }
41 return changed;
42}
43
44/*
45 * Find next free range after start, and size is returned in *sizep
46 */
47u64 __init memblock_x86_find_in_range_size(u64 start, u64 *sizep, u64 align)
48{
49 struct memblock_region *r;
50
51 for_each_memblock(memory, r) {
52 u64 ei_start = r->base;
53 u64 ei_last = ei_start + r->size;
54 u64 addr;
55
56 addr = round_up(ei_start, align);
57 if (addr < start)
58 addr = round_up(start, align);
59 if (addr >= ei_last)
60 continue;
61 *sizep = ei_last - addr;
62 while (check_with_memblock_reserved_size(&addr, sizep, align))
63 ;
64
65 if (*sizep)
66 return addr;
67 }
68
69 return MEMBLOCK_ERROR;
70}
71
72static __init struct range *find_range_array(int count)
73{
74 u64 end, size, mem;
75 struct range *range;
76
77 size = sizeof(struct range) * count;
78 end = memblock.current_limit;
79
80 mem = memblock_find_in_range(0, end, size, sizeof(struct range));
81 if (mem == MEMBLOCK_ERROR)
82 panic("can not find more space for range array");
83
84 /*
85 * This range is tempoaray, so don't reserve it, it will not be
86 * overlapped because We will not alloccate new buffer before
87 * We discard this one
88 */
89 range = __va(mem);
90 memset(range, 0, size);
91
92 return range;
93}
94
95static void __init memblock_x86_subtract_reserved(struct range *range, int az)
96{
97 u64 final_start, final_end;
98 struct memblock_region *r;
99
100 /* Take out region array itself at first*/
101 memblock_free_reserved_regions();
102
103 memblock_dbg("Subtract (%ld early reservations)\n", memblock.reserved.cnt);
104
105 for_each_memblock(reserved, r) {
106 memblock_dbg(" [%010llx-%010llx]\n", (u64)r->base, (u64)r->base + r->size - 1);
107 final_start = PFN_DOWN(r->base);
108 final_end = PFN_UP(r->base + r->size);
109 if (final_start >= final_end)
110 continue;
111 subtract_range(range, az, final_start, final_end);
112 }
113
114 /* Put region array back ? */
115 memblock_reserve_reserved_regions();
116}
117
118struct count_data {
119 int nr;
120};
121
122static int __init count_work_fn(unsigned long start_pfn,
123 unsigned long end_pfn, void *datax)
124{
125 struct count_data *data = datax;
126
127 data->nr++;
128
129 return 0;
130}
131
132static int __init count_early_node_map(int nodeid)
133{
134 struct count_data data;
135
136 data.nr = 0;
137 work_with_active_regions(nodeid, count_work_fn, &data);
138
139 return data.nr;
140}
141
142int __init __get_free_all_memory_range(struct range **rangep, int nodeid,
143 unsigned long start_pfn, unsigned long end_pfn)
144{
145 int count;
146 struct range *range;
147 int nr_range;
148
149 count = (memblock.reserved.cnt + count_early_node_map(nodeid)) * 2;
150
151 range = find_range_array(count);
152 nr_range = 0;
153
154 /*
155 * Use early_node_map[] and memblock.reserved.region to get range array
156 * at first
157 */
158 nr_range = add_from_early_node_map(range, count, nr_range, nodeid);
159 subtract_range(range, count, 0, start_pfn);
160 subtract_range(range, count, end_pfn, -1ULL);
161
162 memblock_x86_subtract_reserved(range, count);
163 nr_range = clean_sort_range(range, count);
164
165 *rangep = range;
166 return nr_range;
167}
168
169int __init get_free_all_memory_range(struct range **rangep, int nodeid)
170{
171 unsigned long end_pfn = -1UL;
172
173#ifdef CONFIG_X86_32
174 end_pfn = max_low_pfn;
175#endif
176 return __get_free_all_memory_range(rangep, nodeid, 0, end_pfn);
177}
178
179static u64 __init __memblock_x86_memory_in_range(u64 addr, u64 limit, bool get_free)
180{
181 int i, count;
182 struct range *range;
183 int nr_range;
184 u64 final_start, final_end;
185 u64 free_size;
186 struct memblock_region *r;
187
188 count = (memblock.reserved.cnt + memblock.memory.cnt) * 2;
189
190 range = find_range_array(count);
191 nr_range = 0;
192
193 addr = PFN_UP(addr);
194 limit = PFN_DOWN(limit);
195
196 for_each_memblock(memory, r) {
197 final_start = PFN_UP(r->base);
198 final_end = PFN_DOWN(r->base + r->size);
199 if (final_start >= final_end)
200 continue;
201 if (final_start >= limit || final_end <= addr)
202 continue;
203
204 nr_range = add_range(range, count, nr_range, final_start, final_end);
205 }
206 subtract_range(range, count, 0, addr);
207 subtract_range(range, count, limit, -1ULL);
208
209 /* Subtract memblock.reserved.region in range ? */
210 if (!get_free)
211 goto sort_and_count_them;
212 for_each_memblock(reserved, r) {
213 final_start = PFN_DOWN(r->base);
214 final_end = PFN_UP(r->base + r->size);
215 if (final_start >= final_end)
216 continue;
217 if (final_start >= limit || final_end <= addr)
218 continue;
219
220 subtract_range(range, count, final_start, final_end);
221 }
222
223sort_and_count_them:
224 nr_range = clean_sort_range(range, count);
225
226 free_size = 0;
227 for (i = 0; i < nr_range; i++)
228 free_size += range[i].end - range[i].start;
229
230 return free_size << PAGE_SHIFT;
231}
232
233u64 __init memblock_x86_free_memory_in_range(u64 addr, u64 limit)
234{
235 return __memblock_x86_memory_in_range(addr, limit, true);
236}
237
238u64 __init memblock_x86_memory_in_range(u64 addr, u64 limit)
239{
240 return __memblock_x86_memory_in_range(addr, limit, false);
241}
242
243void __init memblock_x86_reserve_range(u64 start, u64 end, char *name)
244{
245 if (start == end)
246 return;
247
248 if (WARN_ONCE(start > end, "memblock_x86_reserve_range: wrong range [%#llx, %#llx)\n", start, end))
249 return;
250
251 memblock_dbg(" memblock_x86_reserve_range: [%#010llx-%#010llx] %16s\n", start, end - 1, name);
252
253 memblock_reserve(start, end - start);
254}
255
256void __init memblock_x86_free_range(u64 start, u64 end)
257{
258 if (start == end)
259 return;
260
261 if (WARN_ONCE(start > end, "memblock_x86_free_range: wrong range [%#llx, %#llx)\n", start, end))
262 return;
263
264 memblock_dbg(" memblock_x86_free_range: [%#010llx-%#010llx]\n", start, end - 1);
265
266 memblock_free(start, end - start);
267}
268
269/*
270 * Need to call this function after memblock_x86_register_active_regions,
271 * so early_node_map[] is filled already.
272 */
273u64 __init memblock_x86_find_in_range_node(int nid, u64 start, u64 end, u64 size, u64 align)
274{
275 u64 addr;
276 addr = find_memory_core_early(nid, size, align, start, end);
277 if (addr != MEMBLOCK_ERROR)
278 return addr;
279
280 /* Fallback, should already have start end within node range */
281 return memblock_find_in_range(start, end, size, align);
282}
283
284/*
285 * Finds an active region in the address range from start_pfn to last_pfn and
286 * returns its range in ei_startpfn and ei_endpfn for the memblock entry.
287 */
288static int __init memblock_x86_find_active_region(const struct memblock_region *ei,
289 unsigned long start_pfn,
290 unsigned long last_pfn,
291 unsigned long *ei_startpfn,
292 unsigned long *ei_endpfn)
293{
294 u64 align = PAGE_SIZE;
295
296 *ei_startpfn = round_up(ei->base, align) >> PAGE_SHIFT;
297 *ei_endpfn = round_down(ei->base + ei->size, align) >> PAGE_SHIFT;
298
299 /* Skip map entries smaller than a page */
300 if (*ei_startpfn >= *ei_endpfn)
301 return 0;
302
303 /* Skip if map is outside the node */
304 if (*ei_endpfn <= start_pfn || *ei_startpfn >= last_pfn)
305 return 0;
306
307 /* Check for overlaps */
308 if (*ei_startpfn < start_pfn)
309 *ei_startpfn = start_pfn;
310 if (*ei_endpfn > last_pfn)
311 *ei_endpfn = last_pfn;
312
313 return 1;
314}
315
316/* Walk the memblock.memory map and register active regions within a node */
317void __init memblock_x86_register_active_regions(int nid, unsigned long start_pfn,
318 unsigned long last_pfn)
319{
320 unsigned long ei_startpfn;
321 unsigned long ei_endpfn;
322 struct memblock_region *r;
323
324 for_each_memblock(memory, r)
325 if (memblock_x86_find_active_region(r, start_pfn, last_pfn,
326 &ei_startpfn, &ei_endpfn))
327 add_active_range(nid, ei_startpfn, ei_endpfn);
328}
329
330/*
331 * Find the hole size (in bytes) in the memory range.
332 * @start: starting address of the memory range to scan
333 * @end: ending address of the memory range to scan
334 */
335u64 __init memblock_x86_hole_size(u64 start, u64 end)
336{
337 unsigned long start_pfn = start >> PAGE_SHIFT;
338 unsigned long last_pfn = end >> PAGE_SHIFT;
339 unsigned long ei_startpfn, ei_endpfn, ram = 0;
340 struct memblock_region *r;
341
342 for_each_memblock(memory, r)
343 if (memblock_x86_find_active_region(r, start_pfn, last_pfn,
344 &ei_startpfn, &ei_endpfn))
345 ram += ei_endpfn - ei_startpfn;
346
347 return end - start - ((u64)ram << PAGE_SHIFT);
348}
diff --git a/arch/x86/mm/memtest.c b/arch/x86/mm/memtest.c
index 18d244f7020..92faf3a1c53 100644
--- a/arch/x86/mm/memtest.c
+++ b/arch/x86/mm/memtest.c
@@ -6,8 +6,7 @@
6#include <linux/smp.h> 6#include <linux/smp.h>
7#include <linux/init.h> 7#include <linux/init.h>
8#include <linux/pfn.h> 8#include <linux/pfn.h>
9 9#include <linux/memblock.h>
10#include <asm/e820.h>
11 10
12static u64 patterns[] __initdata = { 11static u64 patterns[] __initdata = {
13 0, 12 0,
@@ -35,7 +34,7 @@ static void __init reserve_bad_mem(u64 pattern, u64 start_bad, u64 end_bad)
35 (unsigned long long) pattern, 34 (unsigned long long) pattern,
36 (unsigned long long) start_bad, 35 (unsigned long long) start_bad,
37 (unsigned long long) end_bad); 36 (unsigned long long) end_bad);
38 reserve_early(start_bad, end_bad, "BAD RAM"); 37 memblock_x86_reserve_range(start_bad, end_bad, "BAD RAM");
39} 38}
40 39
41static void __init memtest(u64 pattern, u64 start_phys, u64 size) 40static void __init memtest(u64 pattern, u64 start_phys, u64 size)
@@ -74,7 +73,7 @@ static void __init do_one_pass(u64 pattern, u64 start, u64 end)
74 u64 size = 0; 73 u64 size = 0;
75 74
76 while (start < end) { 75 while (start < end) {
77 start = find_e820_area_size(start, &size, 1); 76 start = memblock_x86_find_in_range_size(start, &size, 1);
78 77
79 /* done ? */ 78 /* done ? */
80 if (start >= end) 79 if (start >= end)
diff --git a/arch/x86/mm/numa_32.c b/arch/x86/mm/numa_32.c
index 809baaaf48b..84a3e4c9f27 100644
--- a/arch/x86/mm/numa_32.c
+++ b/arch/x86/mm/numa_32.c
@@ -24,6 +24,7 @@
24 24
25#include <linux/mm.h> 25#include <linux/mm.h>
26#include <linux/bootmem.h> 26#include <linux/bootmem.h>
27#include <linux/memblock.h>
27#include <linux/mmzone.h> 28#include <linux/mmzone.h>
28#include <linux/highmem.h> 29#include <linux/highmem.h>
29#include <linux/initrd.h> 30#include <linux/initrd.h>
@@ -120,7 +121,7 @@ int __init get_memcfg_numa_flat(void)
120 121
121 node_start_pfn[0] = 0; 122 node_start_pfn[0] = 0;
122 node_end_pfn[0] = max_pfn; 123 node_end_pfn[0] = max_pfn;
123 e820_register_active_regions(0, 0, max_pfn); 124 memblock_x86_register_active_regions(0, 0, max_pfn);
124 memory_present(0, 0, max_pfn); 125 memory_present(0, 0, max_pfn);
125 node_remap_size[0] = node_memmap_size_bytes(0, 0, max_pfn); 126 node_remap_size[0] = node_memmap_size_bytes(0, 0, max_pfn);
126 127
@@ -161,14 +162,14 @@ static void __init allocate_pgdat(int nid)
161 NODE_DATA(nid) = (pg_data_t *)node_remap_start_vaddr[nid]; 162 NODE_DATA(nid) = (pg_data_t *)node_remap_start_vaddr[nid];
162 else { 163 else {
163 unsigned long pgdat_phys; 164 unsigned long pgdat_phys;
164 pgdat_phys = find_e820_area(min_low_pfn<<PAGE_SHIFT, 165 pgdat_phys = memblock_find_in_range(min_low_pfn<<PAGE_SHIFT,
165 max_pfn_mapped<<PAGE_SHIFT, 166 max_pfn_mapped<<PAGE_SHIFT,
166 sizeof(pg_data_t), 167 sizeof(pg_data_t),
167 PAGE_SIZE); 168 PAGE_SIZE);
168 NODE_DATA(nid) = (pg_data_t *)(pfn_to_kaddr(pgdat_phys>>PAGE_SHIFT)); 169 NODE_DATA(nid) = (pg_data_t *)(pfn_to_kaddr(pgdat_phys>>PAGE_SHIFT));
169 memset(buf, 0, sizeof(buf)); 170 memset(buf, 0, sizeof(buf));
170 sprintf(buf, "NODE_DATA %d", nid); 171 sprintf(buf, "NODE_DATA %d", nid);
171 reserve_early(pgdat_phys, pgdat_phys + sizeof(pg_data_t), buf); 172 memblock_x86_reserve_range(pgdat_phys, pgdat_phys + sizeof(pg_data_t), buf);
172 } 173 }
173 printk(KERN_DEBUG "allocate_pgdat: node %d NODE_DATA %08lx\n", 174 printk(KERN_DEBUG "allocate_pgdat: node %d NODE_DATA %08lx\n",
174 nid, (unsigned long)NODE_DATA(nid)); 175 nid, (unsigned long)NODE_DATA(nid));
@@ -291,15 +292,15 @@ static __init unsigned long calculate_numa_remap_pages(void)
291 PTRS_PER_PTE); 292 PTRS_PER_PTE);
292 node_kva_target <<= PAGE_SHIFT; 293 node_kva_target <<= PAGE_SHIFT;
293 do { 294 do {
294 node_kva_final = find_e820_area(node_kva_target, 295 node_kva_final = memblock_find_in_range(node_kva_target,
295 ((u64)node_end_pfn[nid])<<PAGE_SHIFT, 296 ((u64)node_end_pfn[nid])<<PAGE_SHIFT,
296 ((u64)size)<<PAGE_SHIFT, 297 ((u64)size)<<PAGE_SHIFT,
297 LARGE_PAGE_BYTES); 298 LARGE_PAGE_BYTES);
298 node_kva_target -= LARGE_PAGE_BYTES; 299 node_kva_target -= LARGE_PAGE_BYTES;
299 } while (node_kva_final == -1ULL && 300 } while (node_kva_final == MEMBLOCK_ERROR &&
300 (node_kva_target>>PAGE_SHIFT) > (node_start_pfn[nid])); 301 (node_kva_target>>PAGE_SHIFT) > (node_start_pfn[nid]));
301 302
302 if (node_kva_final == -1ULL) 303 if (node_kva_final == MEMBLOCK_ERROR)
303 panic("Can not get kva ram\n"); 304 panic("Can not get kva ram\n");
304 305
305 node_remap_size[nid] = size; 306 node_remap_size[nid] = size;
@@ -318,15 +319,13 @@ static __init unsigned long calculate_numa_remap_pages(void)
318 * but we could have some hole in high memory, and it will only 319 * but we could have some hole in high memory, and it will only
319 * check page_is_ram(pfn) && !page_is_reserved_early(pfn) to decide 320 * check page_is_ram(pfn) && !page_is_reserved_early(pfn) to decide
320 * to use it as free. 321 * to use it as free.
321 * So reserve_early here, hope we don't run out of that array 322 * So memblock_x86_reserve_range here, hope we don't run out of that array
322 */ 323 */
323 reserve_early(node_kva_final, 324 memblock_x86_reserve_range(node_kva_final,
324 node_kva_final+(((u64)size)<<PAGE_SHIFT), 325 node_kva_final+(((u64)size)<<PAGE_SHIFT),
325 "KVA RAM"); 326 "KVA RAM");
326 327
327 node_remap_start_pfn[nid] = node_kva_final>>PAGE_SHIFT; 328 node_remap_start_pfn[nid] = node_kva_final>>PAGE_SHIFT;
328 remove_active_range(nid, node_remap_start_pfn[nid],
329 node_remap_start_pfn[nid] + size);
330 } 329 }
331 printk(KERN_INFO "Reserving total of %lx pages for numa KVA remap\n", 330 printk(KERN_INFO "Reserving total of %lx pages for numa KVA remap\n",
332 reserve_pages); 331 reserve_pages);
@@ -367,14 +366,14 @@ void __init initmem_init(unsigned long start_pfn, unsigned long end_pfn,
367 366
368 kva_target_pfn = round_down(max_low_pfn - kva_pages, PTRS_PER_PTE); 367 kva_target_pfn = round_down(max_low_pfn - kva_pages, PTRS_PER_PTE);
369 do { 368 do {
370 kva_start_pfn = find_e820_area(kva_target_pfn<<PAGE_SHIFT, 369 kva_start_pfn = memblock_find_in_range(kva_target_pfn<<PAGE_SHIFT,
371 max_low_pfn<<PAGE_SHIFT, 370 max_low_pfn<<PAGE_SHIFT,
372 kva_pages<<PAGE_SHIFT, 371 kva_pages<<PAGE_SHIFT,
373 PTRS_PER_PTE<<PAGE_SHIFT) >> PAGE_SHIFT; 372 PTRS_PER_PTE<<PAGE_SHIFT) >> PAGE_SHIFT;
374 kva_target_pfn -= PTRS_PER_PTE; 373 kva_target_pfn -= PTRS_PER_PTE;
375 } while (kva_start_pfn == -1UL && kva_target_pfn > min_low_pfn); 374 } while (kva_start_pfn == MEMBLOCK_ERROR && kva_target_pfn > min_low_pfn);
376 375
377 if (kva_start_pfn == -1UL) 376 if (kva_start_pfn == MEMBLOCK_ERROR)
378 panic("Can not get kva space\n"); 377 panic("Can not get kva space\n");
379 378
380 printk(KERN_INFO "kva_start_pfn ~ %lx max_low_pfn ~ %lx\n", 379 printk(KERN_INFO "kva_start_pfn ~ %lx max_low_pfn ~ %lx\n",
@@ -382,7 +381,7 @@ void __init initmem_init(unsigned long start_pfn, unsigned long end_pfn,
382 printk(KERN_INFO "max_pfn = %lx\n", max_pfn); 381 printk(KERN_INFO "max_pfn = %lx\n", max_pfn);
383 382
384 /* avoid clash with initrd */ 383 /* avoid clash with initrd */
385 reserve_early(kva_start_pfn<<PAGE_SHIFT, 384 memblock_x86_reserve_range(kva_start_pfn<<PAGE_SHIFT,
386 (kva_start_pfn + kva_pages)<<PAGE_SHIFT, 385 (kva_start_pfn + kva_pages)<<PAGE_SHIFT,
387 "KVA PG"); 386 "KVA PG");
388#ifdef CONFIG_HIGHMEM 387#ifdef CONFIG_HIGHMEM
@@ -419,9 +418,6 @@ void __init initmem_init(unsigned long start_pfn, unsigned long end_pfn,
419 for_each_online_node(nid) { 418 for_each_online_node(nid) {
420 memset(NODE_DATA(nid), 0, sizeof(struct pglist_data)); 419 memset(NODE_DATA(nid), 0, sizeof(struct pglist_data));
421 NODE_DATA(nid)->node_id = nid; 420 NODE_DATA(nid)->node_id = nid;
422#ifndef CONFIG_NO_BOOTMEM
423 NODE_DATA(nid)->bdata = &bootmem_node_data[nid];
424#endif
425 } 421 }
426 422
427 setup_bootmem_allocator(); 423 setup_bootmem_allocator();
diff --git a/arch/x86/mm/numa_64.c b/arch/x86/mm/numa_64.c
index 4962f1aeda6..60f498511dd 100644
--- a/arch/x86/mm/numa_64.c
+++ b/arch/x86/mm/numa_64.c
@@ -7,6 +7,7 @@
7#include <linux/string.h> 7#include <linux/string.h>
8#include <linux/init.h> 8#include <linux/init.h>
9#include <linux/bootmem.h> 9#include <linux/bootmem.h>
10#include <linux/memblock.h>
10#include <linux/mmzone.h> 11#include <linux/mmzone.h>
11#include <linux/ctype.h> 12#include <linux/ctype.h>
12#include <linux/module.h> 13#include <linux/module.h>
@@ -86,16 +87,16 @@ static int __init allocate_cachealigned_memnodemap(void)
86 87
87 addr = 0x8000; 88 addr = 0x8000;
88 nodemap_size = roundup(sizeof(s16) * memnodemapsize, L1_CACHE_BYTES); 89 nodemap_size = roundup(sizeof(s16) * memnodemapsize, L1_CACHE_BYTES);
89 nodemap_addr = find_e820_area(addr, max_pfn<<PAGE_SHIFT, 90 nodemap_addr = memblock_find_in_range(addr, max_pfn<<PAGE_SHIFT,
90 nodemap_size, L1_CACHE_BYTES); 91 nodemap_size, L1_CACHE_BYTES);
91 if (nodemap_addr == -1UL) { 92 if (nodemap_addr == MEMBLOCK_ERROR) {
92 printk(KERN_ERR 93 printk(KERN_ERR
93 "NUMA: Unable to allocate Memory to Node hash map\n"); 94 "NUMA: Unable to allocate Memory to Node hash map\n");
94 nodemap_addr = nodemap_size = 0; 95 nodemap_addr = nodemap_size = 0;
95 return -1; 96 return -1;
96 } 97 }
97 memnodemap = phys_to_virt(nodemap_addr); 98 memnodemap = phys_to_virt(nodemap_addr);
98 reserve_early(nodemap_addr, nodemap_addr + nodemap_size, "MEMNODEMAP"); 99 memblock_x86_reserve_range(nodemap_addr, nodemap_addr + nodemap_size, "MEMNODEMAP");
99 100
100 printk(KERN_DEBUG "NUMA: Allocated memnodemap from %lx - %lx\n", 101 printk(KERN_DEBUG "NUMA: Allocated memnodemap from %lx - %lx\n",
101 nodemap_addr, nodemap_addr + nodemap_size); 102 nodemap_addr, nodemap_addr + nodemap_size);
@@ -171,8 +172,8 @@ static void * __init early_node_mem(int nodeid, unsigned long start,
171 if (start < (MAX_DMA32_PFN<<PAGE_SHIFT) && 172 if (start < (MAX_DMA32_PFN<<PAGE_SHIFT) &&
172 end > (MAX_DMA32_PFN<<PAGE_SHIFT)) 173 end > (MAX_DMA32_PFN<<PAGE_SHIFT))
173 start = MAX_DMA32_PFN<<PAGE_SHIFT; 174 start = MAX_DMA32_PFN<<PAGE_SHIFT;
174 mem = find_e820_area(start, end, size, align); 175 mem = memblock_x86_find_in_range_node(nodeid, start, end, size, align);
175 if (mem != -1L) 176 if (mem != MEMBLOCK_ERROR)
176 return __va(mem); 177 return __va(mem);
177 178
178 /* extend the search scope */ 179 /* extend the search scope */
@@ -181,8 +182,8 @@ static void * __init early_node_mem(int nodeid, unsigned long start,
181 start = MAX_DMA32_PFN<<PAGE_SHIFT; 182 start = MAX_DMA32_PFN<<PAGE_SHIFT;
182 else 183 else
183 start = MAX_DMA_PFN<<PAGE_SHIFT; 184 start = MAX_DMA_PFN<<PAGE_SHIFT;
184 mem = find_e820_area(start, end, size, align); 185 mem = memblock_x86_find_in_range_node(nodeid, start, end, size, align);
185 if (mem != -1L) 186 if (mem != MEMBLOCK_ERROR)
186 return __va(mem); 187 return __va(mem);
187 188
188 printk(KERN_ERR "Cannot find %lu bytes in node %d\n", 189 printk(KERN_ERR "Cannot find %lu bytes in node %d\n",
@@ -198,10 +199,6 @@ setup_node_bootmem(int nodeid, unsigned long start, unsigned long end)
198 unsigned long start_pfn, last_pfn, nodedata_phys; 199 unsigned long start_pfn, last_pfn, nodedata_phys;
199 const int pgdat_size = roundup(sizeof(pg_data_t), PAGE_SIZE); 200 const int pgdat_size = roundup(sizeof(pg_data_t), PAGE_SIZE);
200 int nid; 201 int nid;
201#ifndef CONFIG_NO_BOOTMEM
202 unsigned long bootmap_start, bootmap_pages, bootmap_size;
203 void *bootmap;
204#endif
205 202
206 if (!end) 203 if (!end)
207 return; 204 return;
@@ -226,7 +223,7 @@ setup_node_bootmem(int nodeid, unsigned long start, unsigned long end)
226 if (node_data[nodeid] == NULL) 223 if (node_data[nodeid] == NULL)
227 return; 224 return;
228 nodedata_phys = __pa(node_data[nodeid]); 225 nodedata_phys = __pa(node_data[nodeid]);
229 reserve_early(nodedata_phys, nodedata_phys + pgdat_size, "NODE_DATA"); 226 memblock_x86_reserve_range(nodedata_phys, nodedata_phys + pgdat_size, "NODE_DATA");
230 printk(KERN_INFO " NODE_DATA [%016lx - %016lx]\n", nodedata_phys, 227 printk(KERN_INFO " NODE_DATA [%016lx - %016lx]\n", nodedata_phys,
231 nodedata_phys + pgdat_size - 1); 228 nodedata_phys + pgdat_size - 1);
232 nid = phys_to_nid(nodedata_phys); 229 nid = phys_to_nid(nodedata_phys);
@@ -238,47 +235,6 @@ setup_node_bootmem(int nodeid, unsigned long start, unsigned long end)
238 NODE_DATA(nodeid)->node_start_pfn = start_pfn; 235 NODE_DATA(nodeid)->node_start_pfn = start_pfn;
239 NODE_DATA(nodeid)->node_spanned_pages = last_pfn - start_pfn; 236 NODE_DATA(nodeid)->node_spanned_pages = last_pfn - start_pfn;
240 237
241#ifndef CONFIG_NO_BOOTMEM
242 NODE_DATA(nodeid)->bdata = &bootmem_node_data[nodeid];
243
244 /*
245 * Find a place for the bootmem map
246 * nodedata_phys could be on other nodes by alloc_bootmem,
247 * so need to sure bootmap_start not to be small, otherwise
248 * early_node_mem will get that with find_e820_area instead
249 * of alloc_bootmem, that could clash with reserved range
250 */
251 bootmap_pages = bootmem_bootmap_pages(last_pfn - start_pfn);
252 bootmap_start = roundup(nodedata_phys + pgdat_size, PAGE_SIZE);
253 /*
254 * SMP_CACHE_BYTES could be enough, but init_bootmem_node like
255 * to use that to align to PAGE_SIZE
256 */
257 bootmap = early_node_mem(nodeid, bootmap_start, end,
258 bootmap_pages<<PAGE_SHIFT, PAGE_SIZE);
259 if (bootmap == NULL) {
260 free_early(nodedata_phys, nodedata_phys + pgdat_size);
261 node_data[nodeid] = NULL;
262 return;
263 }
264 bootmap_start = __pa(bootmap);
265 reserve_early(bootmap_start, bootmap_start+(bootmap_pages<<PAGE_SHIFT),
266 "BOOTMAP");
267
268 bootmap_size = init_bootmem_node(NODE_DATA(nodeid),
269 bootmap_start >> PAGE_SHIFT,
270 start_pfn, last_pfn);
271
272 printk(KERN_INFO " bootmap [%016lx - %016lx] pages %lx\n",
273 bootmap_start, bootmap_start + bootmap_size - 1,
274 bootmap_pages);
275 nid = phys_to_nid(bootmap_start);
276 if (nid != nodeid)
277 printk(KERN_INFO " bootmap(%d) on node %d\n", nodeid, nid);
278
279 free_bootmem_with_active_regions(nodeid, end);
280#endif
281
282 node_set_online(nodeid); 238 node_set_online(nodeid);
283} 239}
284 240
@@ -416,7 +372,7 @@ static int __init split_nodes_interleave(u64 addr, u64 max_addr,
416 nr_nodes = MAX_NUMNODES; 372 nr_nodes = MAX_NUMNODES;
417 } 373 }
418 374
419 size = (max_addr - addr - e820_hole_size(addr, max_addr)) / nr_nodes; 375 size = (max_addr - addr - memblock_x86_hole_size(addr, max_addr)) / nr_nodes;
420 /* 376 /*
421 * Calculate the number of big nodes that can be allocated as a result 377 * Calculate the number of big nodes that can be allocated as a result
422 * of consolidating the remainder. 378 * of consolidating the remainder.
@@ -452,7 +408,7 @@ static int __init split_nodes_interleave(u64 addr, u64 max_addr,
452 * non-reserved memory is less than the per-node size. 408 * non-reserved memory is less than the per-node size.
453 */ 409 */
454 while (end - physnodes[i].start - 410 while (end - physnodes[i].start -
455 e820_hole_size(physnodes[i].start, end) < size) { 411 memblock_x86_hole_size(physnodes[i].start, end) < size) {
456 end += FAKE_NODE_MIN_SIZE; 412 end += FAKE_NODE_MIN_SIZE;
457 if (end > physnodes[i].end) { 413 if (end > physnodes[i].end) {
458 end = physnodes[i].end; 414 end = physnodes[i].end;
@@ -466,7 +422,7 @@ static int __init split_nodes_interleave(u64 addr, u64 max_addr,
466 * this one must extend to the boundary. 422 * this one must extend to the boundary.
467 */ 423 */
468 if (end < dma32_end && dma32_end - end - 424 if (end < dma32_end && dma32_end - end -
469 e820_hole_size(end, dma32_end) < FAKE_NODE_MIN_SIZE) 425 memblock_x86_hole_size(end, dma32_end) < FAKE_NODE_MIN_SIZE)
470 end = dma32_end; 426 end = dma32_end;
471 427
472 /* 428 /*
@@ -475,7 +431,7 @@ static int __init split_nodes_interleave(u64 addr, u64 max_addr,
475 * physical node. 431 * physical node.
476 */ 432 */
477 if (physnodes[i].end - end - 433 if (physnodes[i].end - end -
478 e820_hole_size(end, physnodes[i].end) < size) 434 memblock_x86_hole_size(end, physnodes[i].end) < size)
479 end = physnodes[i].end; 435 end = physnodes[i].end;
480 436
481 /* 437 /*
@@ -503,7 +459,7 @@ static u64 __init find_end_of_node(u64 start, u64 max_addr, u64 size)
503{ 459{
504 u64 end = start + size; 460 u64 end = start + size;
505 461
506 while (end - start - e820_hole_size(start, end) < size) { 462 while (end - start - memblock_x86_hole_size(start, end) < size) {
507 end += FAKE_NODE_MIN_SIZE; 463 end += FAKE_NODE_MIN_SIZE;
508 if (end > max_addr) { 464 if (end > max_addr) {
509 end = max_addr; 465 end = max_addr;
@@ -532,7 +488,7 @@ static int __init split_nodes_size_interleave(u64 addr, u64 max_addr, u64 size)
532 * creates a uniform distribution of node sizes across the entire 488 * creates a uniform distribution of node sizes across the entire
533 * machine (but not necessarily over physical nodes). 489 * machine (but not necessarily over physical nodes).
534 */ 490 */
535 min_size = (max_addr - addr - e820_hole_size(addr, max_addr)) / 491 min_size = (max_addr - addr - memblock_x86_hole_size(addr, max_addr)) /
536 MAX_NUMNODES; 492 MAX_NUMNODES;
537 min_size = max(min_size, FAKE_NODE_MIN_SIZE); 493 min_size = max(min_size, FAKE_NODE_MIN_SIZE);
538 if ((min_size & FAKE_NODE_MIN_HASH_MASK) < min_size) 494 if ((min_size & FAKE_NODE_MIN_HASH_MASK) < min_size)
@@ -565,7 +521,7 @@ static int __init split_nodes_size_interleave(u64 addr, u64 max_addr, u64 size)
565 * this one must extend to the boundary. 521 * this one must extend to the boundary.
566 */ 522 */
567 if (end < dma32_end && dma32_end - end - 523 if (end < dma32_end && dma32_end - end -
568 e820_hole_size(end, dma32_end) < FAKE_NODE_MIN_SIZE) 524 memblock_x86_hole_size(end, dma32_end) < FAKE_NODE_MIN_SIZE)
569 end = dma32_end; 525 end = dma32_end;
570 526
571 /* 527 /*
@@ -574,7 +530,7 @@ static int __init split_nodes_size_interleave(u64 addr, u64 max_addr, u64 size)
574 * physical node. 530 * physical node.
575 */ 531 */
576 if (physnodes[i].end - end - 532 if (physnodes[i].end - end -
577 e820_hole_size(end, physnodes[i].end) < size) 533 memblock_x86_hole_size(end, physnodes[i].end) < size)
578 end = physnodes[i].end; 534 end = physnodes[i].end;
579 535
580 /* 536 /*
@@ -638,7 +594,7 @@ static int __init numa_emulation(unsigned long start_pfn,
638 */ 594 */
639 remove_all_active_ranges(); 595 remove_all_active_ranges();
640 for_each_node_mask(i, node_possible_map) { 596 for_each_node_mask(i, node_possible_map) {
641 e820_register_active_regions(i, nodes[i].start >> PAGE_SHIFT, 597 memblock_x86_register_active_regions(i, nodes[i].start >> PAGE_SHIFT,
642 nodes[i].end >> PAGE_SHIFT); 598 nodes[i].end >> PAGE_SHIFT);
643 setup_node_bootmem(i, nodes[i].start, nodes[i].end); 599 setup_node_bootmem(i, nodes[i].start, nodes[i].end);
644 } 600 }
@@ -691,7 +647,7 @@ void __init initmem_init(unsigned long start_pfn, unsigned long last_pfn,
691 node_set(0, node_possible_map); 647 node_set(0, node_possible_map);
692 for (i = 0; i < nr_cpu_ids; i++) 648 for (i = 0; i < nr_cpu_ids; i++)
693 numa_set_node(i, 0); 649 numa_set_node(i, 0);
694 e820_register_active_regions(0, start_pfn, last_pfn); 650 memblock_x86_register_active_regions(0, start_pfn, last_pfn);
695 setup_node_bootmem(0, start_pfn << PAGE_SHIFT, last_pfn << PAGE_SHIFT); 651 setup_node_bootmem(0, start_pfn << PAGE_SHIFT, last_pfn << PAGE_SHIFT);
696} 652}
697 653
@@ -703,9 +659,7 @@ unsigned long __init numa_free_all_bootmem(void)
703 for_each_online_node(i) 659 for_each_online_node(i)
704 pages += free_all_bootmem_node(NODE_DATA(i)); 660 pages += free_all_bootmem_node(NODE_DATA(i));
705 661
706#ifdef CONFIG_NO_BOOTMEM
707 pages += free_all_memory_core_early(MAX_NUMNODES); 662 pages += free_all_memory_core_early(MAX_NUMNODES);
708#endif
709 663
710 return pages; 664 return pages;
711} 665}
diff --git a/arch/x86/mm/srat_32.c b/arch/x86/mm/srat_32.c
index 9324f13492d..a17dffd136c 100644
--- a/arch/x86/mm/srat_32.c
+++ b/arch/x86/mm/srat_32.c
@@ -25,6 +25,7 @@
25 */ 25 */
26#include <linux/mm.h> 26#include <linux/mm.h>
27#include <linux/bootmem.h> 27#include <linux/bootmem.h>
28#include <linux/memblock.h>
28#include <linux/mmzone.h> 29#include <linux/mmzone.h>
29#include <linux/acpi.h> 30#include <linux/acpi.h>
30#include <linux/nodemask.h> 31#include <linux/nodemask.h>
@@ -264,7 +265,7 @@ int __init get_memcfg_from_srat(void)
264 if (node_read_chunk(chunk->nid, chunk)) 265 if (node_read_chunk(chunk->nid, chunk))
265 continue; 266 continue;
266 267
267 e820_register_active_regions(chunk->nid, chunk->start_pfn, 268 memblock_x86_register_active_regions(chunk->nid, chunk->start_pfn,
268 min(chunk->end_pfn, max_pfn)); 269 min(chunk->end_pfn, max_pfn));
269 } 270 }
270 /* for out of order entries in SRAT */ 271 /* for out of order entries in SRAT */
diff --git a/arch/x86/mm/srat_64.c b/arch/x86/mm/srat_64.c
index 9c0d0d399c3..a35cb9d8b06 100644
--- a/arch/x86/mm/srat_64.c
+++ b/arch/x86/mm/srat_64.c
@@ -16,6 +16,7 @@
16#include <linux/module.h> 16#include <linux/module.h>
17#include <linux/topology.h> 17#include <linux/topology.h>
18#include <linux/bootmem.h> 18#include <linux/bootmem.h>
19#include <linux/memblock.h>
19#include <linux/mm.h> 20#include <linux/mm.h>
20#include <asm/proto.h> 21#include <asm/proto.h>
21#include <asm/numa.h> 22#include <asm/numa.h>
@@ -98,15 +99,15 @@ void __init acpi_numa_slit_init(struct acpi_table_slit *slit)
98 unsigned long phys; 99 unsigned long phys;
99 100
100 length = slit->header.length; 101 length = slit->header.length;
101 phys = find_e820_area(0, max_pfn_mapped<<PAGE_SHIFT, length, 102 phys = memblock_find_in_range(0, max_pfn_mapped<<PAGE_SHIFT, length,
102 PAGE_SIZE); 103 PAGE_SIZE);
103 104
104 if (phys == -1L) 105 if (phys == MEMBLOCK_ERROR)
105 panic(" Can not save slit!\n"); 106 panic(" Can not save slit!\n");
106 107
107 acpi_slit = __va(phys); 108 acpi_slit = __va(phys);
108 memcpy(acpi_slit, slit, length); 109 memcpy(acpi_slit, slit, length);
109 reserve_early(phys, phys + length, "ACPI SLIT"); 110 memblock_x86_reserve_range(phys, phys + length, "ACPI SLIT");
110} 111}
111 112
112/* Callback for Proximity Domain -> x2APIC mapping */ 113/* Callback for Proximity Domain -> x2APIC mapping */
@@ -324,7 +325,7 @@ static int __init nodes_cover_memory(const struct bootnode *nodes)
324 pxmram = 0; 325 pxmram = 0;
325 } 326 }
326 327
327 e820ram = max_pfn - (e820_hole_size(0, max_pfn<<PAGE_SHIFT)>>PAGE_SHIFT); 328 e820ram = max_pfn - (memblock_x86_hole_size(0, max_pfn<<PAGE_SHIFT)>>PAGE_SHIFT);
328 /* We seem to lose 3 pages somewhere. Allow 1M of slack. */ 329 /* We seem to lose 3 pages somewhere. Allow 1M of slack. */
329 if ((long)(e820ram - pxmram) >= (1<<(20 - PAGE_SHIFT))) { 330 if ((long)(e820ram - pxmram) >= (1<<(20 - PAGE_SHIFT))) {
330 printk(KERN_ERR 331 printk(KERN_ERR
@@ -421,7 +422,7 @@ int __init acpi_scan_nodes(unsigned long start, unsigned long end)
421 } 422 }
422 423
423 for (i = 0; i < num_node_memblks; i++) 424 for (i = 0; i < num_node_memblks; i++)
424 e820_register_active_regions(memblk_nodeid[i], 425 memblock_x86_register_active_regions(memblk_nodeid[i],
425 node_memblk_range[i].start >> PAGE_SHIFT, 426 node_memblk_range[i].start >> PAGE_SHIFT,
426 node_memblk_range[i].end >> PAGE_SHIFT); 427 node_memblk_range[i].end >> PAGE_SHIFT);
427 428