diff options
| -rw-r--r-- | arch/i386/Kconfig | 1 | ||||
| -rw-r--r-- | include/linux/mm.h | 92 | ||||
| -rw-r--r-- | include/linux/mmzone.h | 96 | ||||
| -rw-r--r-- | include/linux/numa.h | 2 | ||||
| -rw-r--r-- | mm/Kconfig | 38 | ||||
| -rw-r--r-- | mm/Makefile | 1 | ||||
| -rw-r--r-- | mm/bootmem.c | 9 | ||||
| -rw-r--r-- | mm/memory.c | 2 | ||||
| -rw-r--r-- | mm/page_alloc.c | 39 | ||||
| -rw-r--r-- | mm/sparse.c | 85 |
10 files changed, 332 insertions, 33 deletions
diff --git a/arch/i386/Kconfig b/arch/i386/Kconfig index 3b72481..f0064b5 100644 --- a/arch/i386/Kconfig +++ b/arch/i386/Kconfig @@ -813,6 +813,7 @@ source "mm/Kconfig" config HAVE_ARCH_EARLY_PFN_TO_NID bool default y + depends on NUMA config HIGHPTE bool "Allocate 3rd-level pagetables from highmem" diff --git a/include/linux/mm.h b/include/linux/mm.h index 57b2ead..6eb7f48 100644 --- a/include/linux/mm.h +++ b/include/linux/mm.h @@ -397,40 +397,80 @@ static inline void put_page(struct page *page) * sets it, so none of the operations on it need to be atomic. */ -/* Page flags: | NODE | ZONE | ... | FLAGS | */ -#define NODES_PGOFF ((sizeof(page_flags_t)*8) - NODES_SHIFT) -#define ZONES_PGOFF (NODES_PGOFF - ZONES_SHIFT) + +/* + * page->flags layout: + * + * There are three possibilities for how page->flags get + * laid out. The first is for the normal case, without + * sparsemem. The second is for sparsemem when there is + * plenty of space for node and section. The last is when + * we have run out of space and have to fall back to an + * alternate (slower) way of determining the node. + * + * No sparsemem: | NODE | ZONE | ... | FLAGS | + * with space for node: | SECTION | NODE | ZONE | ... | FLAGS | + * no space for node: | SECTION | ZONE | ... | FLAGS | + */ +#ifdef CONFIG_SPARSEMEM +#define SECTIONS_WIDTH SECTIONS_SHIFT +#else +#define SECTIONS_WIDTH 0 +#endif + +#define ZONES_WIDTH ZONES_SHIFT + +#if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= FLAGS_RESERVED +#define NODES_WIDTH NODES_SHIFT +#else +#define NODES_WIDTH 0 +#endif + +/* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */ +#define SECTIONS_PGOFF ((sizeof(page_flags_t)*8) - SECTIONS_WIDTH) +#define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH) +#define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH) + +/* + * We are going to use the flags for the page to node mapping if its in + * there. This includes the case where there is no node, so it is implicit. + */ +#define FLAGS_HAS_NODE (NODES_WIDTH > 0 || NODES_SHIFT == 0) + +#ifndef PFN_SECTION_SHIFT +#define PFN_SECTION_SHIFT 0 +#endif /* * Define the bit shifts to access each section. For non-existant * sections we define the shift as 0; that plus a 0 mask ensures * the compiler will optimise away reference to them. */ -#define NODES_PGSHIFT (NODES_PGOFF * (NODES_SHIFT != 0)) -#define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_SHIFT != 0)) +#define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0)) +#define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0)) +#define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0)) -/* NODE:ZONE is used to lookup the zone from a page. */ +/* NODE:ZONE or SECTION:ZONE is used to lookup the zone from a page. */ +#if FLAGS_HAS_NODE #define ZONETABLE_SHIFT (NODES_SHIFT + ZONES_SHIFT) +#else +#define ZONETABLE_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT) +#endif #define ZONETABLE_PGSHIFT ZONES_PGSHIFT -#if NODES_SHIFT+ZONES_SHIFT > FLAGS_RESERVED -#error NODES_SHIFT+ZONES_SHIFT > FLAGS_RESERVED +#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED +#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED #endif -#define NODEZONE(node, zone) ((node << ZONES_SHIFT) | zone) - -#define ZONES_MASK ((1UL << ZONES_SHIFT) - 1) -#define NODES_MASK ((1UL << NODES_SHIFT) - 1) +#define ZONES_MASK ((1UL << ZONES_WIDTH) - 1) +#define NODES_MASK ((1UL << NODES_WIDTH) - 1) +#define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1) #define ZONETABLE_MASK ((1UL << ZONETABLE_SHIFT) - 1) static inline unsigned long page_zonenum(struct page *page) { return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK; } -static inline unsigned long page_to_nid(struct page *page) -{ - return (page->flags >> NODES_PGSHIFT) & NODES_MASK; -} struct zone; extern struct zone *zone_table[]; @@ -441,6 +481,18 @@ static inline struct zone *page_zone(struct page *page) ZONETABLE_MASK]; } +static inline unsigned long page_to_nid(struct page *page) +{ + if (FLAGS_HAS_NODE) + return (page->flags >> NODES_PGSHIFT) & NODES_MASK; + else + return page_zone(page)->zone_pgdat->node_id; +} +static inline unsigned long page_to_section(struct page *page) +{ + return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK; +} + static inline void set_page_zone(struct page *page, unsigned long zone) { page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT); @@ -451,12 +503,18 @@ static inline void set_page_node(struct page *page, unsigned long node) page->flags &= ~(NODES_MASK << NODES_PGSHIFT); page->flags |= (node & NODES_MASK) << NODES_PGSHIFT; } +static inline void set_page_section(struct page *page, unsigned long section) +{ + page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT); + page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT; +} static inline void set_page_links(struct page *page, unsigned long zone, - unsigned long node) + unsigned long node, unsigned long pfn) { set_page_zone(page, zone); set_page_node(page, node); + set_page_section(page, pfn_to_section_nr(pfn)); } #ifndef CONFIG_DISCONTIGMEM diff --git a/include/linux/mmzone.h b/include/linux/mmzone.h index 6ef07de..19860d3 100644 --- a/include/linux/mmzone.h +++ b/include/linux/mmzone.h @@ -269,7 +269,9 @@ typedef struct pglist_data { struct zone node_zones[MAX_NR_ZONES]; struct zonelist node_zonelists[GFP_ZONETYPES]; int nr_zones; +#ifdef CONFIG_FLAT_NODE_MEM_MAP struct page *node_mem_map; +#endif struct bootmem_data *bdata; unsigned long node_start_pfn; unsigned long node_present_pages; /* total number of physical pages */ @@ -284,7 +286,11 @@ typedef struct pglist_data { #define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages) #define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages) +#ifdef CONFIG_FLAT_NODE_MEM_MAP #define pgdat_page_nr(pgdat, pagenr) ((pgdat)->node_mem_map + (pagenr)) +#else +#define pgdat_page_nr(pgdat, pagenr) pfn_to_page((pgdat)->node_start_pfn + (pagenr)) +#endif #define nid_page_nr(nid, pagenr) pgdat_page_nr(NODE_DATA(nid),(pagenr)) extern struct pglist_data *pgdat_list; @@ -416,6 +422,10 @@ extern struct pglist_data contig_page_data; #endif /* !CONFIG_NEED_MULTIPLE_NODES */ +#ifdef CONFIG_SPARSEMEM +#include <asm/sparsemem.h> +#endif + #if BITS_PER_LONG == 32 || defined(ARCH_HAS_ATOMIC_UNSIGNED) /* * with 32 bit page->flags field, we reserve 8 bits for node/zone info. @@ -439,6 +449,92 @@ extern struct pglist_data contig_page_data; #define early_pfn_to_nid(nid) (0UL) #endif +#define pfn_to_section_nr(pfn) ((pfn) >> PFN_SECTION_SHIFT) +#define section_nr_to_pfn(sec) ((sec) << PFN_SECTION_SHIFT) + +#ifdef CONFIG_SPARSEMEM + +/* + * SECTION_SHIFT #bits space required to store a section # + * + * PA_SECTION_SHIFT physical address to/from section number + * PFN_SECTION_SHIFT pfn to/from section number + */ +#define SECTIONS_SHIFT (MAX_PHYSMEM_BITS - SECTION_SIZE_BITS) + +#define PA_SECTION_SHIFT (SECTION_SIZE_BITS) +#define PFN_SECTION_SHIFT (SECTION_SIZE_BITS - PAGE_SHIFT) + +#define NR_MEM_SECTIONS (1UL << SECTIONS_SHIFT) + +#define PAGES_PER_SECTION (1UL << PFN_SECTION_SHIFT) +#define PAGE_SECTION_MASK (~(PAGES_PER_SECTION-1)) + +#if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS +#error Allocator MAX_ORDER exceeds SECTION_SIZE +#endif + +struct page; +struct mem_section { + struct page *section_mem_map; +}; + +extern struct mem_section mem_section[NR_MEM_SECTIONS]; + +/* + * Given a kernel address, find the home node of the underlying memory. + */ +#define kvaddr_to_nid(kaddr) pfn_to_nid(__pa(kaddr) >> PAGE_SHIFT) + +static inline struct mem_section *__pfn_to_section(unsigned long pfn) +{ + return &mem_section[pfn_to_section_nr(pfn)]; +} + +#define pfn_to_page(pfn) \ +({ \ + unsigned long __pfn = (pfn); \ + __pfn_to_section(__pfn)->section_mem_map + __pfn; \ +}) +#define page_to_pfn(page) \ +({ \ + page - mem_section[page_to_section(page)].section_mem_map; \ +}) + +static inline int pfn_valid(unsigned long pfn) +{ + if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS) + return 0; + return mem_section[pfn_to_section_nr(pfn)].section_mem_map != 0; +} + +/* + * These are _only_ used during initialisation, therefore they + * can use __initdata ... They could have names to indicate + * this restriction. + */ +#ifdef CONFIG_NUMA +#define pfn_to_nid early_pfn_to_nid +#endif + +#define pfn_to_pgdat(pfn) \ +({ \ + NODE_DATA(pfn_to_nid(pfn)); \ +}) + +#define early_pfn_valid(pfn) pfn_valid(pfn) +void sparse_init(void); +#else +#define sparse_init() do {} while (0) +#endif /* CONFIG_SPARSEMEM */ + +#ifndef early_pfn_valid +#define early_pfn_valid(pfn) (1) +#endif + +void memory_present(int nid, unsigned long start, unsigned long end); +unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long); + #endif /* !__ASSEMBLY__ */ #endif /* __KERNEL__ */ #endif /* _LINUX_MMZONE_H */ diff --git a/include/linux/numa.h b/include/linux/numa.h index bd0c8c4..f0c539b 100644 --- a/include/linux/numa.h +++ b/include/linux/numa.h @@ -3,7 +3,7 @@ #include <linux/config.h> -#ifdef CONFIG_DISCONTIGMEM +#ifndef CONFIG_FLATMEM #include <asm/numnodes.h> #endif @@ -6,6 +6,7 @@ choice prompt "Memory model" depends on SELECT_MEMORY_MODEL default DISCONTIGMEM_MANUAL if ARCH_DISCONTIGMEM_DEFAULT + default SPARSEMEM_MANUAL if ARCH_SPARSEMEM_DEFAULT default FLATMEM_MANUAL config FLATMEM_MANUAL @@ -17,7 +18,15 @@ config FLATMEM_MANUAL only have one option here: FLATMEM. This is normal and a correct option. - If unsure, choose this option over any other. + Some users of more advanced features like NUMA and + memory hotplug may have different options here. + DISCONTIGMEM is an more mature, better tested system, + but is incompatible with memory hotplug and may suffer + decreased performance over SPARSEMEM. If unsure between + "Sparse Memory" and "Discontiguous Memory", choose + "Discontiguous Memory". + + If unsure, choose this option (Flat Memory) over any other. config DISCONTIGMEM_MANUAL bool "Discontigious Memory" @@ -35,15 +44,38 @@ config DISCONTIGMEM_MANUAL If unsure, choose "Flat Memory" over this option. +config SPARSEMEM_MANUAL + bool "Sparse Memory" + depends on ARCH_SPARSEMEM_ENABLE + help + This will be the only option for some systems, including + memory hotplug systems. This is normal. + + For many other systems, this will be an alternative to + "Discontigious Memory". This option provides some potential + performance benefits, along with decreased code complexity, + but it is newer, and more experimental. + + If unsure, choose "Discontiguous Memory" or "Flat Memory" + over this option. + endchoice config DISCONTIGMEM def_bool y depends on (!SELECT_MEMORY_MODEL && ARCH_DISCONTIGMEM_ENABLE) || DISCONTIGMEM_MANUAL +config SPARSEMEM + def_bool y + depends on SPARSEMEM_MANUAL + config FLATMEM def_bool y - depends on !DISCONTIGMEM || FLATMEM_MANUAL + depends on (!DISCONTIGMEM && !SPARSEMEM) || FLATMEM_MANUAL + +config FLAT_NODE_MEM_MAP + def_bool y + depends on !SPARSEMEM # # Both the NUMA code and DISCONTIGMEM use arrays of pg_data_t's @@ -56,4 +88,4 @@ config NEED_MULTIPLE_NODES config HAVE_MEMORY_PRESENT def_bool y - depends on ARCH_HAVE_MEMORY_PRESENT + depends on ARCH_HAVE_MEMORY_PRESENT || SPARSEMEM diff --git a/mm/Makefile b/mm/Makefile index 0974080..8f70ffd 100644 --- a/mm/Makefile +++ b/mm/Makefile @@ -15,6 +15,7 @@ obj-y := bootmem.o filemap.o mempool.o oom_kill.o fadvise.o \ obj-$(CONFIG_SWAP) += page_io.o swap_state.o swapfile.o thrash.o obj-$(CONFIG_HUGETLBFS) += hugetlb.o obj-$(CONFIG_NUMA) += mempolicy.o +obj-$(CONFIG_SPARSEMEM) += sparse.o obj-$(CONFIG_SHMEM) += shmem.o obj-$(CONFIG_TINY_SHMEM) += tiny-shmem.o diff --git a/mm/bootmem.c b/mm/bootmem.c index 260e703..f82f7ae 100644 --- a/mm/bootmem.c +++ b/mm/bootmem.c @@ -256,6 +256,7 @@ found: static unsigned long __init free_all_bootmem_core(pg_data_t *pgdat) { struct page *page; + unsigned long pfn; bootmem_data_t *bdata = pgdat->bdata; unsigned long i, count, total = 0; unsigned long idx; @@ -266,7 +267,7 @@ static unsigned long __init free_all_bootmem_core(pg_data_t *pgdat) count = 0; /* first extant page of the node */ - page = virt_to_page(phys_to_virt(bdata->node_boot_start)); + pfn = bdata->node_boot_start >> PAGE_SHIFT; idx = bdata->node_low_pfn - (bdata->node_boot_start >> PAGE_SHIFT); map = bdata->node_bootmem_map; /* Check physaddr is O(LOG2(BITS_PER_LONG)) page aligned */ @@ -275,9 +276,11 @@ static unsigned long __init free_all_bootmem_core(pg_data_t *pgdat) gofast = 1; for (i = 0; i < idx; ) { unsigned long v = ~map[i / BITS_PER_LONG]; + if (gofast && v == ~0UL) { int j, order; + page = pfn_to_page(pfn); count += BITS_PER_LONG; __ClearPageReserved(page); order = ffs(BITS_PER_LONG) - 1; @@ -292,6 +295,8 @@ static unsigned long __init free_all_bootmem_core(pg_data_t *pgdat) page += BITS_PER_LONG; } else if (v) { unsigned long m; + + page = pfn_to_page(pfn); for (m = 1; m && i < idx; m<<=1, page++, i++) { if (v & m) { count++; @@ -302,8 +307,8 @@ static unsigned long __init free_all_bootmem_core(pg_data_t *pgdat) } } else { i+=BITS_PER_LONG; - page += BITS_PER_LONG; } + pfn += BITS_PER_LONG; } total += count; diff --git a/mm/memory.c b/mm/memory.c index da91b7b..30975ef 100644 --- a/mm/memory.c +++ b/mm/memory.c @@ -58,7 +58,7 @@ #include <linux/swapops.h> #include <linux/elf.h> -#ifndef CONFIG_DISCONTIGMEM +#ifndef CONFIG_NEED_MULTIPLE_NODES /* use the per-pgdat data instead for discontigmem - mbligh */ unsigned long max_mapnr; struct page *mem_map; diff --git a/mm/page_alloc.c b/mm/page_alloc.c index 20e2395..5c1b898 100644 --- a/mm/page_alloc.c +++ b/mm/page_alloc.c @@ -68,7 +68,7 @@ EXPORT_SYMBOL(nr_swap_pages); * Used by page_zone() to look up the address of the struct zone whose * id is encoded in the upper bits of page->flags */ -struct zone *zone_table[1 << (ZONES_SHIFT + NODES_SHIFT)]; +struct zone *zone_table[1 << ZONETABLE_SHIFT]; EXPORT_SYMBOL(zone_table); static char *zone_names[MAX_NR_ZONES] = { "DMA", "Normal", "HighMem" }; @@ -1649,11 +1649,15 @@ static void __init calculate_zone_totalpages(struct pglist_data *pgdat, void __init memmap_init_zone(unsigned long size, int nid, unsigned long zone, unsigned long start_pfn) { - struct page *start = pfn_to_page(start_pfn); struct page *page; + int end_pfn = start_pfn + size; + int pfn; - for (page = start; page < (start + size); page++) { - set_page_links(page, zone, nid); + for (pfn = start_pfn; pfn < end_pfn; pfn++, page++) { + if (!early_pfn_valid(pfn)) + continue; + page = pfn_to_page(pfn); + set_page_links(page, zone, nid, pfn); set_page_count(page, 0); reset_page_mapcount(page); SetPageReserved(page); @@ -1677,6 +1681,20 @@ void zone_init_free_lists(struct pglist_data *pgdat, struct zone *zone, } } +#define ZONETABLE_INDEX(x, zone_nr) ((x << ZONES_SHIFT) | zone_nr) +void zonetable_add(struct zone *zone, int nid, int zid, unsigned long pfn, + unsigned long size) +{ + unsigned long snum = pfn_to_section_nr(pfn); + unsigned long end = pfn_to_section_nr(pfn + size); + + if (FLAGS_HAS_NODE) + zone_table[ZONETABLE_INDEX(nid, zid)] = zone; + else + for (; snum <= end; snum++) + zone_table[ZONETABLE_INDEX(snum, zid)] = zone; +} + #ifndef __HAVE_ARCH_MEMMAP_INIT #define memmap_init(size, nid, zone, start_pfn) \ memmap_init_zone((size), (nid), (zone), (start_pfn)) @@ -1861,7 +1879,6 @@ static void __init free_area_init_core(struct pglist_data *pgdat, unsigned long size, realsize; unsigned long batch; - zone_table[NODEZONE(nid, j)] = zone; realsize = size = zones_size[j]; if (zholes_size) realsize -= zholes_size[j]; @@ -1927,6 +1944,8 @@ static void __init free_area_init_core(struct pglist_data *pgdat, memmap_init(size, nid, j, zone_start_pfn); + zonetable_add(zone, nid, j, zone_start_pfn, size); + zone_start_pfn += size; zone_init_free_lists(pgdat, zone, zone->spanned_pages); @@ -1935,28 +1954,30 @@ static void __init free_area_init_core(struct pglist_data *pgdat, static void __init alloc_node_mem_map(struct pglist_data *pgdat) { - unsigned long size; - struct page *map; - /* Skip empty nodes */ if (!pgdat->node_spanned_pages) return; +#ifdef CONFIG_FLAT_NODE_MEM_MAP /* ia64 gets its own node_mem_map, before this, without bootmem */ if (!pgdat->node_mem_map) { + unsigned long size; + struct page *map; + size = (pgdat->node_spanned_pages + 1) * sizeof(struct page); map = alloc_remap(pgdat->node_id, size); if (!map) map = alloc_bootmem_node(pgdat, size); pgdat->node_mem_map = map; } -#ifndef CONFIG_DISCONTIGMEM +#ifdef CONFIG_FLATMEM /* * With no DISCONTIG, the global mem_map is just set as node 0's */ if (pgdat == NODE_DATA(0)) mem_map = NODE_DATA(0)->node_mem_map; #endif +#endif /* CONFIG_FLAT_NODE_MEM_MAP */ } void __init free_area_init_node(int nid, struct pglist_data *pgdat, diff --git a/mm/sparse.c b/mm/sparse.c new file mode 100644 index 0000000..f888385 --- /dev/null +++ b/mm/sparse.c @@ -0,0 +1,85 @@ +/* + * sparse memory mappings. + */ +#include <linux/config.h> +#include <linux/mm.h> +#include <linux/mmzone.h> +#include <linux/bootmem.h> +#include <linux/module.h> +#include <asm/dma.h> + +/* + * Permanent SPARSEMEM data: + * + * 1) mem_section - memory sections, mem_map's for valid memory + */ +struct mem_section mem_section[NR_MEM_SECTIONS]; +EXPORT_SYMBOL(mem_section); + +/* Record a memory area against a node. */ +void memory_present(int nid, unsigned long start, unsigned long end) +{ + unsigned long pfn; + + start &= PAGE_SECTION_MASK; + for (pfn = start; pfn < end; pfn += PAGES_PER_SECTION) { + unsigned long section = pfn_to_section_nr(pfn); + if (!mem_section[section].section_mem_map) + mem_section[section].section_mem_map = (void *) -1; + } +} + +/* + * Only used by the i386 NUMA architecures, but relatively + * generic code. + */ +unsigned long __init node_memmap_size_bytes(int nid, unsigned long start_pfn, + unsigned long end_pfn) +{ + unsigned long pfn; + unsigned long nr_pages = 0; + + for (pfn = start_pfn; pfn < end_pfn; pfn += PAGES_PER_SECTION) { + if (nid != early_pfn_to_nid(pfn)) + continue; + + if (pfn_valid(pfn)) + nr_pages += PAGES_PER_SECTION; + } + + return nr_pages * sizeof(struct page); +} + +/* + * Allocate the accumulated non-linear sections, allocate a mem_map + * for each and record the physical to section mapping. + */ +void sparse_init(void) +{ + unsigned long pnum; + struct page *map; + int nid; + + for (pnum = 0; pnum < NR_MEM_SECTIONS; pnum++) { + if (!mem_section[pnum].section_mem_map) + continue; + + nid = early_pfn_to_nid(section_nr_to_pfn(pnum)); + map = alloc_remap(nid, sizeof(struct page) * PAGES_PER_SECTION); + if (!map) + map = alloc_bootmem_node(NODE_DATA(nid), + sizeof(struct page) * PAGES_PER_SECTION); + if (!map) { + mem_section[pnum].section_mem_map = 0; + continue; + } + + /* + * Subtle, we encode the real pfn into the mem_map such that + * the identity pfn - section_mem_map will return the actual + * physical page frame number. + */ + mem_section[pnum].section_mem_map = map - + section_nr_to_pfn(pnum); + } +} |
