Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | #ifndef _LINUX_MMZONE_H |
| 2 | #define _LINUX_MMZONE_H |
| 3 | |
| 4 | #ifdef __KERNEL__ |
| 5 | #ifndef __ASSEMBLY__ |
| 6 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 7 | #include <linux/spinlock.h> |
| 8 | #include <linux/list.h> |
| 9 | #include <linux/wait.h> |
David Rientjes | e815af9 | 2007-10-16 23:25:54 -0700 | [diff] [blame] | 10 | #include <linux/bitops.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 11 | #include <linux/cache.h> |
| 12 | #include <linux/threads.h> |
| 13 | #include <linux/numa.h> |
| 14 | #include <linux/init.h> |
Dave Hansen | bdc8cb9 | 2005-10-29 18:16:53 -0700 | [diff] [blame] | 15 | #include <linux/seqlock.h> |
KAMEZAWA Hiroyuki | 8357f86 | 2006-03-27 01:15:57 -0800 | [diff] [blame] | 16 | #include <linux/nodemask.h> |
Mel Gorman | 835c134 | 2007-10-16 01:25:47 -0700 | [diff] [blame] | 17 | #include <linux/pageblock-flags.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 18 | #include <asm/atomic.h> |
Ralf Baechle | 93ff66b | 2006-06-04 02:51:29 -0700 | [diff] [blame] | 19 | #include <asm/page.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 20 | |
| 21 | /* Free memory management - zoned buddy allocator. */ |
| 22 | #ifndef CONFIG_FORCE_MAX_ZONEORDER |
| 23 | #define MAX_ORDER 11 |
| 24 | #else |
| 25 | #define MAX_ORDER CONFIG_FORCE_MAX_ZONEORDER |
| 26 | #endif |
Bob Picco | e984bb4 | 2006-05-20 15:00:31 -0700 | [diff] [blame] | 27 | #define MAX_ORDER_NR_PAGES (1 << (MAX_ORDER - 1)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 28 | |
Andy Whitcroft | 5ad333e | 2007-07-17 04:03:16 -0700 | [diff] [blame] | 29 | /* |
| 30 | * PAGE_ALLOC_COSTLY_ORDER is the order at which allocations are deemed |
| 31 | * costly to service. That is between allocation orders which should |
| 32 | * coelesce naturally under reasonable reclaim pressure and those which |
| 33 | * will not. |
| 34 | */ |
| 35 | #define PAGE_ALLOC_COSTLY_ORDER 3 |
| 36 | |
Mel Gorman | b2a0ac8 | 2007-10-16 01:25:48 -0700 | [diff] [blame] | 37 | #define MIGRATE_UNMOVABLE 0 |
Mel Gorman | e12ba74 | 2007-10-16 01:25:52 -0700 | [diff] [blame] | 38 | #define MIGRATE_RECLAIMABLE 1 |
| 39 | #define MIGRATE_MOVABLE 2 |
Mel Gorman | 64c5e13 | 2007-10-16 01:25:59 -0700 | [diff] [blame] | 40 | #define MIGRATE_RESERVE 3 |
KAMEZAWA Hiroyuki | a5d76b5 | 2007-10-16 01:26:11 -0700 | [diff] [blame] | 41 | #define MIGRATE_ISOLATE 4 /* can't allocate from here */ |
| 42 | #define MIGRATE_TYPES 5 |
Mel Gorman | b2a0ac8 | 2007-10-16 01:25:48 -0700 | [diff] [blame] | 43 | |
| 44 | #define for_each_migratetype_order(order, type) \ |
| 45 | for (order = 0; order < MAX_ORDER; order++) \ |
| 46 | for (type = 0; type < MIGRATE_TYPES; type++) |
| 47 | |
Mel Gorman | 467c996 | 2007-10-16 01:26:02 -0700 | [diff] [blame] | 48 | extern int page_group_by_mobility_disabled; |
| 49 | |
| 50 | static inline int get_pageblock_migratetype(struct page *page) |
| 51 | { |
| 52 | if (unlikely(page_group_by_mobility_disabled)) |
| 53 | return MIGRATE_UNMOVABLE; |
| 54 | |
| 55 | return get_pageblock_flags_group(page, PB_migrate, PB_migrate_end); |
| 56 | } |
| 57 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 58 | struct free_area { |
Mel Gorman | b2a0ac8 | 2007-10-16 01:25:48 -0700 | [diff] [blame] | 59 | struct list_head free_list[MIGRATE_TYPES]; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 60 | unsigned long nr_free; |
| 61 | }; |
| 62 | |
| 63 | struct pglist_data; |
| 64 | |
| 65 | /* |
| 66 | * zone->lock and zone->lru_lock are two of the hottest locks in the kernel. |
| 67 | * So add a wild amount of padding here to ensure that they fall into separate |
| 68 | * cachelines. There are very few zone structures in the machine, so space |
| 69 | * consumption is not a concern here. |
| 70 | */ |
| 71 | #if defined(CONFIG_SMP) |
| 72 | struct zone_padding { |
| 73 | char x[0]; |
Ravikiran G Thirumalai | 22fc6ec | 2006-01-08 01:01:27 -0800 | [diff] [blame] | 74 | } ____cacheline_internodealigned_in_smp; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 75 | #define ZONE_PADDING(name) struct zone_padding name; |
| 76 | #else |
| 77 | #define ZONE_PADDING(name) |
| 78 | #endif |
| 79 | |
Christoph Lameter | 2244b95 | 2006-06-30 01:55:33 -0700 | [diff] [blame] | 80 | enum zone_stat_item { |
Christoph Lameter | 51ed449 | 2007-02-10 01:43:02 -0800 | [diff] [blame] | 81 | /* First 128 byte cacheline (assuming 64 bit words) */ |
Christoph Lameter | d23ad42 | 2007-02-10 01:43:02 -0800 | [diff] [blame] | 82 | NR_FREE_PAGES, |
Christoph Lameter | c878538 | 2007-02-10 01:43:01 -0800 | [diff] [blame] | 83 | NR_INACTIVE, |
| 84 | NR_ACTIVE, |
Christoph Lameter | f3dbd34 | 2006-06-30 01:55:36 -0700 | [diff] [blame] | 85 | NR_ANON_PAGES, /* Mapped anonymous pages */ |
| 86 | NR_FILE_MAPPED, /* pagecache pages mapped into pagetables. |
Christoph Lameter | 65ba55f | 2006-06-30 01:55:34 -0700 | [diff] [blame] | 87 | only modified from process context */ |
Christoph Lameter | 347ce43 | 2006-06-30 01:55:35 -0700 | [diff] [blame] | 88 | NR_FILE_PAGES, |
Christoph Lameter | b1e7a8f | 2006-06-30 01:55:39 -0700 | [diff] [blame] | 89 | NR_FILE_DIRTY, |
Christoph Lameter | ce866b3 | 2006-06-30 01:55:40 -0700 | [diff] [blame] | 90 | NR_WRITEBACK, |
Christoph Lameter | 51ed449 | 2007-02-10 01:43:02 -0800 | [diff] [blame] | 91 | /* Second 128 byte cacheline */ |
| 92 | NR_SLAB_RECLAIMABLE, |
| 93 | NR_SLAB_UNRECLAIMABLE, |
| 94 | NR_PAGETABLE, /* used for pagetables */ |
Christoph Lameter | fd39fc8 | 2006-06-30 01:55:40 -0700 | [diff] [blame] | 95 | NR_UNSTABLE_NFS, /* NFS unstable pages */ |
Christoph Lameter | d2c5e30 | 2006-06-30 01:55:41 -0700 | [diff] [blame] | 96 | NR_BOUNCE, |
Andrew Morton | e129b5c | 2006-09-27 01:50:00 -0700 | [diff] [blame] | 97 | NR_VMSCAN_WRITE, |
Christoph Lameter | ca889e6 | 2006-06-30 01:55:44 -0700 | [diff] [blame] | 98 | #ifdef CONFIG_NUMA |
| 99 | NUMA_HIT, /* allocated in intended node */ |
| 100 | NUMA_MISS, /* allocated in non intended node */ |
| 101 | NUMA_FOREIGN, /* was intended here, hit elsewhere */ |
| 102 | NUMA_INTERLEAVE_HIT, /* interleaver preferred this zone */ |
| 103 | NUMA_LOCAL, /* allocation from local node */ |
| 104 | NUMA_OTHER, /* allocation from other node */ |
| 105 | #endif |
Christoph Lameter | 2244b95 | 2006-06-30 01:55:33 -0700 | [diff] [blame] | 106 | NR_VM_ZONE_STAT_ITEMS }; |
| 107 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 108 | struct per_cpu_pages { |
| 109 | int count; /* number of pages in the list */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 110 | int high; /* high watermark, emptying needed */ |
| 111 | int batch; /* chunk size for buddy add/remove */ |
| 112 | struct list_head list; /* the list of pages */ |
| 113 | }; |
| 114 | |
| 115 | struct per_cpu_pageset { |
Christoph Lameter | 3dfa572 | 2008-02-04 22:29:19 -0800 | [diff] [blame] | 116 | struct per_cpu_pages pcp; |
Christoph Lameter | 4037d45 | 2007-05-09 02:35:14 -0700 | [diff] [blame] | 117 | #ifdef CONFIG_NUMA |
| 118 | s8 expire; |
| 119 | #endif |
Christoph Lameter | 2244b95 | 2006-06-30 01:55:33 -0700 | [diff] [blame] | 120 | #ifdef CONFIG_SMP |
Christoph Lameter | df9ecab | 2006-08-31 21:27:35 -0700 | [diff] [blame] | 121 | s8 stat_threshold; |
Christoph Lameter | 2244b95 | 2006-06-30 01:55:33 -0700 | [diff] [blame] | 122 | s8 vm_stat_diff[NR_VM_ZONE_STAT_ITEMS]; |
| 123 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 124 | } ____cacheline_aligned_in_smp; |
| 125 | |
Christoph Lameter | e7c8d5c | 2005-06-21 17:14:47 -0700 | [diff] [blame] | 126 | #ifdef CONFIG_NUMA |
| 127 | #define zone_pcp(__z, __cpu) ((__z)->pageset[(__cpu)]) |
| 128 | #else |
| 129 | #define zone_pcp(__z, __cpu) (&(__z)->pageset[(__cpu)]) |
| 130 | #endif |
| 131 | |
Christoph Lameter | 2f1b624 | 2006-09-25 23:31:13 -0700 | [diff] [blame] | 132 | enum zone_type { |
Christoph Lameter | 4b51d66 | 2007-02-10 01:43:10 -0800 | [diff] [blame] | 133 | #ifdef CONFIG_ZONE_DMA |
Christoph Lameter | 2f1b624 | 2006-09-25 23:31:13 -0700 | [diff] [blame] | 134 | /* |
| 135 | * ZONE_DMA is used when there are devices that are not able |
| 136 | * to do DMA to all of addressable memory (ZONE_NORMAL). Then we |
| 137 | * carve out the portion of memory that is needed for these devices. |
| 138 | * The range is arch specific. |
| 139 | * |
| 140 | * Some examples |
| 141 | * |
| 142 | * Architecture Limit |
| 143 | * --------------------------- |
| 144 | * parisc, ia64, sparc <4G |
| 145 | * s390 <2G |
Christoph Lameter | 2f1b624 | 2006-09-25 23:31:13 -0700 | [diff] [blame] | 146 | * arm Various |
| 147 | * alpha Unlimited or 0-16MB. |
| 148 | * |
| 149 | * i386, x86_64 and multiple other arches |
| 150 | * <16M. |
| 151 | */ |
| 152 | ZONE_DMA, |
Christoph Lameter | 4b51d66 | 2007-02-10 01:43:10 -0800 | [diff] [blame] | 153 | #endif |
Christoph Lameter | fb0e794 | 2006-09-25 23:31:13 -0700 | [diff] [blame] | 154 | #ifdef CONFIG_ZONE_DMA32 |
Christoph Lameter | 2f1b624 | 2006-09-25 23:31:13 -0700 | [diff] [blame] | 155 | /* |
| 156 | * x86_64 needs two ZONE_DMAs because it supports devices that are |
| 157 | * only able to do DMA to the lower 16M but also 32 bit devices that |
| 158 | * can only do DMA areas below 4G. |
| 159 | */ |
| 160 | ZONE_DMA32, |
Christoph Lameter | fb0e794 | 2006-09-25 23:31:13 -0700 | [diff] [blame] | 161 | #endif |
Christoph Lameter | 2f1b624 | 2006-09-25 23:31:13 -0700 | [diff] [blame] | 162 | /* |
| 163 | * Normal addressable memory is in ZONE_NORMAL. DMA operations can be |
| 164 | * performed on pages in ZONE_NORMAL if the DMA devices support |
| 165 | * transfers to all addressable memory. |
| 166 | */ |
| 167 | ZONE_NORMAL, |
Christoph Lameter | e53ef38 | 2006-09-25 23:31:14 -0700 | [diff] [blame] | 168 | #ifdef CONFIG_HIGHMEM |
Christoph Lameter | 2f1b624 | 2006-09-25 23:31:13 -0700 | [diff] [blame] | 169 | /* |
| 170 | * A memory area that is only addressable by the kernel through |
| 171 | * mapping portions into its own address space. This is for example |
| 172 | * used by i386 to allow the kernel to address the memory beyond |
| 173 | * 900MB. The kernel will set up special mappings (page |
| 174 | * table entries on i386) for each page that the kernel needs to |
| 175 | * access. |
| 176 | */ |
| 177 | ZONE_HIGHMEM, |
Christoph Lameter | e53ef38 | 2006-09-25 23:31:14 -0700 | [diff] [blame] | 178 | #endif |
Mel Gorman | 2a1e274 | 2007-07-17 04:03:12 -0700 | [diff] [blame] | 179 | ZONE_MOVABLE, |
Christoph Lameter | 2f1b624 | 2006-09-25 23:31:13 -0700 | [diff] [blame] | 180 | MAX_NR_ZONES |
| 181 | }; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 182 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 183 | /* |
| 184 | * When a memory allocation must conform to specific limitations (such |
| 185 | * as being suitable for DMA) the caller will pass in hints to the |
| 186 | * allocator in the gfp_mask, in the zone modifier bits. These bits |
| 187 | * are used to select a priority ordered list of memory zones which |
Christoph Lameter | 19655d3 | 2006-09-25 23:31:19 -0700 | [diff] [blame] | 188 | * match the requested limits. See gfp_zone() in include/linux/gfp.h |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 189 | */ |
Christoph Lameter | fb0e794 | 2006-09-25 23:31:13 -0700 | [diff] [blame] | 190 | |
Christoph Lameter | 4b51d66 | 2007-02-10 01:43:10 -0800 | [diff] [blame] | 191 | /* |
| 192 | * Count the active zones. Note that the use of defined(X) outside |
| 193 | * #if and family is not necessarily defined so ensure we cannot use |
| 194 | * it later. Use __ZONE_COUNT to work out how many shift bits we need. |
| 195 | */ |
| 196 | #define __ZONE_COUNT ( \ |
| 197 | defined(CONFIG_ZONE_DMA) \ |
| 198 | + defined(CONFIG_ZONE_DMA32) \ |
| 199 | + 1 \ |
| 200 | + defined(CONFIG_HIGHMEM) \ |
Mel Gorman | 2a1e274 | 2007-07-17 04:03:12 -0700 | [diff] [blame] | 201 | + 1 \ |
Christoph Lameter | 4b51d66 | 2007-02-10 01:43:10 -0800 | [diff] [blame] | 202 | ) |
| 203 | #if __ZONE_COUNT < 2 |
| 204 | #define ZONES_SHIFT 0 |
| 205 | #elif __ZONE_COUNT <= 2 |
Christoph Lameter | 19655d3 | 2006-09-25 23:31:19 -0700 | [diff] [blame] | 206 | #define ZONES_SHIFT 1 |
Christoph Lameter | 4b51d66 | 2007-02-10 01:43:10 -0800 | [diff] [blame] | 207 | #elif __ZONE_COUNT <= 4 |
Christoph Lameter | 19655d3 | 2006-09-25 23:31:19 -0700 | [diff] [blame] | 208 | #define ZONES_SHIFT 2 |
Christoph Lameter | 4b51d66 | 2007-02-10 01:43:10 -0800 | [diff] [blame] | 209 | #else |
| 210 | #error ZONES_SHIFT -- too many zones configured adjust calculation |
Christoph Lameter | fb0e794 | 2006-09-25 23:31:13 -0700 | [diff] [blame] | 211 | #endif |
Christoph Lameter | 4b51d66 | 2007-02-10 01:43:10 -0800 | [diff] [blame] | 212 | #undef __ZONE_COUNT |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 213 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 214 | struct zone { |
| 215 | /* Fields commonly accessed by the page allocator */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 216 | unsigned long pages_min, pages_low, pages_high; |
| 217 | /* |
| 218 | * We don't know if the memory that we're going to allocate will be freeable |
| 219 | * or/and it will be released eventually, so to avoid totally wasting several |
| 220 | * GB of ram we must reserve some of the lower zone memory (otherwise we risk |
| 221 | * to run OOM on the lower zones despite there's tons of freeable ram |
| 222 | * on the higher zones). This array is recalculated at runtime if the |
| 223 | * sysctl_lowmem_reserve_ratio sysctl changes. |
| 224 | */ |
| 225 | unsigned long lowmem_reserve[MAX_NR_ZONES]; |
| 226 | |
Christoph Lameter | e7c8d5c | 2005-06-21 17:14:47 -0700 | [diff] [blame] | 227 | #ifdef CONFIG_NUMA |
Christoph Lameter | d5f541e | 2006-09-27 01:50:08 -0700 | [diff] [blame] | 228 | int node; |
Christoph Lameter | 9614634 | 2006-07-03 00:24:13 -0700 | [diff] [blame] | 229 | /* |
| 230 | * zone reclaim becomes active if more unmapped pages exist. |
| 231 | */ |
Christoph Lameter | 8417bba | 2006-09-25 23:31:51 -0700 | [diff] [blame] | 232 | unsigned long min_unmapped_pages; |
Christoph Lameter | 0ff3849 | 2006-09-25 23:31:52 -0700 | [diff] [blame] | 233 | unsigned long min_slab_pages; |
Christoph Lameter | e7c8d5c | 2005-06-21 17:14:47 -0700 | [diff] [blame] | 234 | struct per_cpu_pageset *pageset[NR_CPUS]; |
| 235 | #else |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 236 | struct per_cpu_pageset pageset[NR_CPUS]; |
Christoph Lameter | e7c8d5c | 2005-06-21 17:14:47 -0700 | [diff] [blame] | 237 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 238 | /* |
| 239 | * free areas of different sizes |
| 240 | */ |
| 241 | spinlock_t lock; |
Dave Hansen | bdc8cb9 | 2005-10-29 18:16:53 -0700 | [diff] [blame] | 242 | #ifdef CONFIG_MEMORY_HOTPLUG |
| 243 | /* see spanned/present_pages for more description */ |
| 244 | seqlock_t span_seqlock; |
| 245 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 246 | struct free_area free_area[MAX_ORDER]; |
| 247 | |
Mel Gorman | 835c134 | 2007-10-16 01:25:47 -0700 | [diff] [blame] | 248 | #ifndef CONFIG_SPARSEMEM |
| 249 | /* |
Mel Gorman | d9c2340 | 2007-10-16 01:26:01 -0700 | [diff] [blame] | 250 | * Flags for a pageblock_nr_pages block. See pageblock-flags.h. |
Mel Gorman | 835c134 | 2007-10-16 01:25:47 -0700 | [diff] [blame] | 251 | * In SPARSEMEM, this map is stored in struct mem_section |
| 252 | */ |
| 253 | unsigned long *pageblock_flags; |
| 254 | #endif /* CONFIG_SPARSEMEM */ |
| 255 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 256 | |
| 257 | ZONE_PADDING(_pad1_) |
| 258 | |
| 259 | /* Fields commonly accessed by the page reclaim scanner */ |
| 260 | spinlock_t lru_lock; |
| 261 | struct list_head active_list; |
| 262 | struct list_head inactive_list; |
| 263 | unsigned long nr_scan_active; |
| 264 | unsigned long nr_scan_inactive; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 265 | unsigned long pages_scanned; /* since last reclaim */ |
David Rientjes | e815af9 | 2007-10-16 23:25:54 -0700 | [diff] [blame] | 266 | unsigned long flags; /* zone flags, see below */ |
Martin Hicks | 753ee72 | 2005-06-21 17:14:41 -0700 | [diff] [blame] | 267 | |
Christoph Lameter | 2244b95 | 2006-06-30 01:55:33 -0700 | [diff] [blame] | 268 | /* Zone statistics */ |
| 269 | atomic_long_t vm_stat[NR_VM_ZONE_STAT_ITEMS]; |
Christoph Lameter | 9eeff23 | 2006-01-18 17:42:31 -0800 | [diff] [blame] | 270 | |
| 271 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 272 | * prev_priority holds the scanning priority for this zone. It is |
| 273 | * defined as the scanning priority at which we achieved our reclaim |
| 274 | * target at the previous try_to_free_pages() or balance_pgdat() |
| 275 | * invokation. |
| 276 | * |
| 277 | * We use prev_priority as a measure of how much stress page reclaim is |
| 278 | * under - it drives the swappiness decision: whether to unmap mapped |
| 279 | * pages. |
| 280 | * |
Martin Bligh | 3bb1a85 | 2006-10-28 10:38:24 -0700 | [diff] [blame] | 281 | * Access to both this field is quite racy even on uniprocessor. But |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 282 | * it is expected to average out OK. |
| 283 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 284 | int prev_priority; |
| 285 | |
| 286 | |
| 287 | ZONE_PADDING(_pad2_) |
| 288 | /* Rarely used or read-mostly fields */ |
| 289 | |
| 290 | /* |
| 291 | * wait_table -- the array holding the hash table |
Yasunori Goto | 02b694d | 2006-06-23 02:03:08 -0700 | [diff] [blame] | 292 | * wait_table_hash_nr_entries -- the size of the hash table array |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 293 | * wait_table_bits -- wait_table_size == (1 << wait_table_bits) |
| 294 | * |
| 295 | * The purpose of all these is to keep track of the people |
| 296 | * waiting for a page to become available and make them |
| 297 | * runnable again when possible. The trouble is that this |
| 298 | * consumes a lot of space, especially when so few things |
| 299 | * wait on pages at a given time. So instead of using |
| 300 | * per-page waitqueues, we use a waitqueue hash table. |
| 301 | * |
| 302 | * The bucket discipline is to sleep on the same queue when |
| 303 | * colliding and wake all in that wait queue when removing. |
| 304 | * When something wakes, it must check to be sure its page is |
| 305 | * truly available, a la thundering herd. The cost of a |
| 306 | * collision is great, but given the expected load of the |
| 307 | * table, they should be so rare as to be outweighed by the |
| 308 | * benefits from the saved space. |
| 309 | * |
| 310 | * __wait_on_page_locked() and unlock_page() in mm/filemap.c, are the |
| 311 | * primary users of these fields, and in mm/page_alloc.c |
| 312 | * free_area_init_core() performs the initialization of them. |
| 313 | */ |
| 314 | wait_queue_head_t * wait_table; |
Yasunori Goto | 02b694d | 2006-06-23 02:03:08 -0700 | [diff] [blame] | 315 | unsigned long wait_table_hash_nr_entries; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 316 | unsigned long wait_table_bits; |
| 317 | |
| 318 | /* |
| 319 | * Discontig memory support fields. |
| 320 | */ |
| 321 | struct pglist_data *zone_pgdat; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 322 | /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */ |
| 323 | unsigned long zone_start_pfn; |
| 324 | |
Dave Hansen | bdc8cb9 | 2005-10-29 18:16:53 -0700 | [diff] [blame] | 325 | /* |
| 326 | * zone_start_pfn, spanned_pages and present_pages are all |
| 327 | * protected by span_seqlock. It is a seqlock because it has |
| 328 | * to be read outside of zone->lock, and it is done in the main |
| 329 | * allocator path. But, it is written quite infrequently. |
| 330 | * |
| 331 | * The lock is declared along with zone->lock because it is |
| 332 | * frequently read in proximity to zone->lock. It's good to |
| 333 | * give them a chance of being in the same cacheline. |
| 334 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 335 | unsigned long spanned_pages; /* total size, including holes */ |
| 336 | unsigned long present_pages; /* amount of memory (excluding holes) */ |
| 337 | |
| 338 | /* |
| 339 | * rarely used fields: |
| 340 | */ |
Helge Deller | 15ad7cd | 2006-12-06 20:40:36 -0800 | [diff] [blame] | 341 | const char *name; |
Ravikiran G Thirumalai | 22fc6ec | 2006-01-08 01:01:27 -0800 | [diff] [blame] | 342 | } ____cacheline_internodealigned_in_smp; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 343 | |
David Rientjes | e815af9 | 2007-10-16 23:25:54 -0700 | [diff] [blame] | 344 | typedef enum { |
| 345 | ZONE_ALL_UNRECLAIMABLE, /* all pages pinned */ |
| 346 | ZONE_RECLAIM_LOCKED, /* prevents concurrent reclaim */ |
David Rientjes | 098d7f1 | 2007-10-16 23:25:55 -0700 | [diff] [blame] | 347 | ZONE_OOM_LOCKED, /* zone is in OOM killer zonelist */ |
David Rientjes | e815af9 | 2007-10-16 23:25:54 -0700 | [diff] [blame] | 348 | } zone_flags_t; |
| 349 | |
| 350 | static inline void zone_set_flag(struct zone *zone, zone_flags_t flag) |
| 351 | { |
| 352 | set_bit(flag, &zone->flags); |
| 353 | } |
David Rientjes | d773ed6 | 2007-10-16 23:26:01 -0700 | [diff] [blame] | 354 | |
| 355 | static inline int zone_test_and_set_flag(struct zone *zone, zone_flags_t flag) |
| 356 | { |
| 357 | return test_and_set_bit(flag, &zone->flags); |
| 358 | } |
| 359 | |
David Rientjes | e815af9 | 2007-10-16 23:25:54 -0700 | [diff] [blame] | 360 | static inline void zone_clear_flag(struct zone *zone, zone_flags_t flag) |
| 361 | { |
| 362 | clear_bit(flag, &zone->flags); |
| 363 | } |
| 364 | |
| 365 | static inline int zone_is_all_unreclaimable(const struct zone *zone) |
| 366 | { |
| 367 | return test_bit(ZONE_ALL_UNRECLAIMABLE, &zone->flags); |
| 368 | } |
David Rientjes | d773ed6 | 2007-10-16 23:26:01 -0700 | [diff] [blame] | 369 | |
David Rientjes | e815af9 | 2007-10-16 23:25:54 -0700 | [diff] [blame] | 370 | static inline int zone_is_reclaim_locked(const struct zone *zone) |
| 371 | { |
| 372 | return test_bit(ZONE_RECLAIM_LOCKED, &zone->flags); |
| 373 | } |
David Rientjes | d773ed6 | 2007-10-16 23:26:01 -0700 | [diff] [blame] | 374 | |
David Rientjes | 098d7f1 | 2007-10-16 23:25:55 -0700 | [diff] [blame] | 375 | static inline int zone_is_oom_locked(const struct zone *zone) |
| 376 | { |
| 377 | return test_bit(ZONE_OOM_LOCKED, &zone->flags); |
| 378 | } |
David Rientjes | e815af9 | 2007-10-16 23:25:54 -0700 | [diff] [blame] | 379 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 380 | /* |
| 381 | * The "priority" of VM scanning is how much of the queues we will scan in one |
| 382 | * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the |
| 383 | * queues ("queue_length >> 12") during an aging round. |
| 384 | */ |
| 385 | #define DEF_PRIORITY 12 |
| 386 | |
Paul Jackson | 9276b1bc | 2006-12-06 20:31:48 -0800 | [diff] [blame] | 387 | /* Maximum number of zones on a zonelist */ |
| 388 | #define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES) |
| 389 | |
| 390 | #ifdef CONFIG_NUMA |
Christoph Lameter | 523b945 | 2007-10-16 01:25:37 -0700 | [diff] [blame] | 391 | |
| 392 | /* |
| 393 | * The NUMA zonelists are doubled becausse we need zonelists that restrict the |
| 394 | * allocations to a single node for GFP_THISNODE. |
| 395 | * |
Mel Gorman | 54a6eb5 | 2008-04-28 02:12:16 -0700 | [diff] [blame] | 396 | * [0] : Zonelist with fallback |
| 397 | * [1] : No fallback (GFP_THISNODE) |
Christoph Lameter | 523b945 | 2007-10-16 01:25:37 -0700 | [diff] [blame] | 398 | */ |
Mel Gorman | 54a6eb5 | 2008-04-28 02:12:16 -0700 | [diff] [blame] | 399 | #define MAX_ZONELISTS 2 |
Christoph Lameter | 523b945 | 2007-10-16 01:25:37 -0700 | [diff] [blame] | 400 | |
| 401 | |
Paul Jackson | 9276b1bc | 2006-12-06 20:31:48 -0800 | [diff] [blame] | 402 | /* |
| 403 | * We cache key information from each zonelist for smaller cache |
| 404 | * footprint when scanning for free pages in get_page_from_freelist(). |
| 405 | * |
| 406 | * 1) The BITMAP fullzones tracks which zones in a zonelist have come |
| 407 | * up short of free memory since the last time (last_fullzone_zap) |
| 408 | * we zero'd fullzones. |
| 409 | * 2) The array z_to_n[] maps each zone in the zonelist to its node |
| 410 | * id, so that we can efficiently evaluate whether that node is |
| 411 | * set in the current tasks mems_allowed. |
| 412 | * |
| 413 | * Both fullzones and z_to_n[] are one-to-one with the zonelist, |
| 414 | * indexed by a zones offset in the zonelist zones[] array. |
| 415 | * |
| 416 | * The get_page_from_freelist() routine does two scans. During the |
| 417 | * first scan, we skip zones whose corresponding bit in 'fullzones' |
| 418 | * is set or whose corresponding node in current->mems_allowed (which |
| 419 | * comes from cpusets) is not set. During the second scan, we bypass |
| 420 | * this zonelist_cache, to ensure we look methodically at each zone. |
| 421 | * |
| 422 | * Once per second, we zero out (zap) fullzones, forcing us to |
| 423 | * reconsider nodes that might have regained more free memory. |
| 424 | * The field last_full_zap is the time we last zapped fullzones. |
| 425 | * |
| 426 | * This mechanism reduces the amount of time we waste repeatedly |
| 427 | * reexaming zones for free memory when they just came up low on |
| 428 | * memory momentarilly ago. |
| 429 | * |
| 430 | * The zonelist_cache struct members logically belong in struct |
| 431 | * zonelist. However, the mempolicy zonelists constructed for |
| 432 | * MPOL_BIND are intentionally variable length (and usually much |
| 433 | * shorter). A general purpose mechanism for handling structs with |
| 434 | * multiple variable length members is more mechanism than we want |
| 435 | * here. We resort to some special case hackery instead. |
| 436 | * |
| 437 | * The MPOL_BIND zonelists don't need this zonelist_cache (in good |
| 438 | * part because they are shorter), so we put the fixed length stuff |
| 439 | * at the front of the zonelist struct, ending in a variable length |
| 440 | * zones[], as is needed by MPOL_BIND. |
| 441 | * |
| 442 | * Then we put the optional zonelist cache on the end of the zonelist |
| 443 | * struct. This optional stuff is found by a 'zlcache_ptr' pointer in |
| 444 | * the fixed length portion at the front of the struct. This pointer |
| 445 | * both enables us to find the zonelist cache, and in the case of |
| 446 | * MPOL_BIND zonelists, (which will just set the zlcache_ptr to NULL) |
| 447 | * to know that the zonelist cache is not there. |
| 448 | * |
| 449 | * The end result is that struct zonelists come in two flavors: |
| 450 | * 1) The full, fixed length version, shown below, and |
| 451 | * 2) The custom zonelists for MPOL_BIND. |
| 452 | * The custom MPOL_BIND zonelists have a NULL zlcache_ptr and no zlcache. |
| 453 | * |
| 454 | * Even though there may be multiple CPU cores on a node modifying |
| 455 | * fullzones or last_full_zap in the same zonelist_cache at the same |
| 456 | * time, we don't lock it. This is just hint data - if it is wrong now |
| 457 | * and then, the allocator will still function, perhaps a bit slower. |
| 458 | */ |
| 459 | |
| 460 | |
| 461 | struct zonelist_cache { |
Paul Jackson | 9276b1bc | 2006-12-06 20:31:48 -0800 | [diff] [blame] | 462 | unsigned short z_to_n[MAX_ZONES_PER_ZONELIST]; /* zone->nid */ |
Paul Jackson | 7253f4e | 2006-12-06 20:31:49 -0800 | [diff] [blame] | 463 | DECLARE_BITMAP(fullzones, MAX_ZONES_PER_ZONELIST); /* zone full? */ |
Paul Jackson | 9276b1bc | 2006-12-06 20:31:48 -0800 | [diff] [blame] | 464 | unsigned long last_full_zap; /* when last zap'd (jiffies) */ |
| 465 | }; |
| 466 | #else |
Mel Gorman | 54a6eb5 | 2008-04-28 02:12:16 -0700 | [diff] [blame] | 467 | #define MAX_ZONELISTS 1 |
Paul Jackson | 9276b1bc | 2006-12-06 20:31:48 -0800 | [diff] [blame] | 468 | struct zonelist_cache; |
| 469 | #endif |
| 470 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 471 | /* |
Mel Gorman | dd1a239 | 2008-04-28 02:12:17 -0700 | [diff] [blame^] | 472 | * This struct contains information about a zone in a zonelist. It is stored |
| 473 | * here to avoid dereferences into large structures and lookups of tables |
| 474 | */ |
| 475 | struct zoneref { |
| 476 | struct zone *zone; /* Pointer to actual zone */ |
| 477 | int zone_idx; /* zone_idx(zoneref->zone) */ |
| 478 | }; |
| 479 | |
| 480 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 481 | * One allocation request operates on a zonelist. A zonelist |
| 482 | * is a list of zones, the first one is the 'goal' of the |
| 483 | * allocation, the other zones are fallback zones, in decreasing |
| 484 | * priority. |
| 485 | * |
Paul Jackson | 9276b1bc | 2006-12-06 20:31:48 -0800 | [diff] [blame] | 486 | * If zlcache_ptr is not NULL, then it is just the address of zlcache, |
| 487 | * as explained above. If zlcache_ptr is NULL, there is no zlcache. |
Mel Gorman | dd1a239 | 2008-04-28 02:12:17 -0700 | [diff] [blame^] | 488 | * * |
| 489 | * To speed the reading of the zonelist, the zonerefs contain the zone index |
| 490 | * of the entry being read. Helper functions to access information given |
| 491 | * a struct zoneref are |
| 492 | * |
| 493 | * zonelist_zone() - Return the struct zone * for an entry in _zonerefs |
| 494 | * zonelist_zone_idx() - Return the index of the zone for an entry |
| 495 | * zonelist_node_idx() - Return the index of the node for an entry |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 496 | */ |
| 497 | struct zonelist { |
Paul Jackson | 9276b1bc | 2006-12-06 20:31:48 -0800 | [diff] [blame] | 498 | struct zonelist_cache *zlcache_ptr; // NULL or &zlcache |
Mel Gorman | dd1a239 | 2008-04-28 02:12:17 -0700 | [diff] [blame^] | 499 | struct zoneref _zonerefs[MAX_ZONES_PER_ZONELIST + 1]; |
Paul Jackson | 9276b1bc | 2006-12-06 20:31:48 -0800 | [diff] [blame] | 500 | #ifdef CONFIG_NUMA |
| 501 | struct zonelist_cache zlcache; // optional ... |
| 502 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 503 | }; |
| 504 | |
Mel Gorman | c713216 | 2006-09-27 01:49:43 -0700 | [diff] [blame] | 505 | #ifdef CONFIG_ARCH_POPULATES_NODE_MAP |
| 506 | struct node_active_region { |
| 507 | unsigned long start_pfn; |
| 508 | unsigned long end_pfn; |
| 509 | int nid; |
| 510 | }; |
| 511 | #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 512 | |
Heiko Carstens | 5b99cd0 | 2006-09-27 01:50:01 -0700 | [diff] [blame] | 513 | #ifndef CONFIG_DISCONTIGMEM |
| 514 | /* The array of struct pages - for discontigmem use pgdat->lmem_map */ |
| 515 | extern struct page *mem_map; |
| 516 | #endif |
| 517 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 518 | /* |
| 519 | * The pg_data_t structure is used in machines with CONFIG_DISCONTIGMEM |
| 520 | * (mostly NUMA machines?) to denote a higher-level memory zone than the |
| 521 | * zone denotes. |
| 522 | * |
| 523 | * On NUMA machines, each NUMA node would have a pg_data_t to describe |
| 524 | * it's memory layout. |
| 525 | * |
| 526 | * Memory statistics and page replacement data structures are maintained on a |
| 527 | * per-zone basis. |
| 528 | */ |
| 529 | struct bootmem_data; |
| 530 | typedef struct pglist_data { |
| 531 | struct zone node_zones[MAX_NR_ZONES]; |
Christoph Lameter | 523b945 | 2007-10-16 01:25:37 -0700 | [diff] [blame] | 532 | struct zonelist node_zonelists[MAX_ZONELISTS]; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 533 | int nr_zones; |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 534 | #ifdef CONFIG_FLAT_NODE_MEM_MAP |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 535 | struct page *node_mem_map; |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 536 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 537 | struct bootmem_data *bdata; |
Dave Hansen | 208d54e | 2005-10-29 18:16:52 -0700 | [diff] [blame] | 538 | #ifdef CONFIG_MEMORY_HOTPLUG |
| 539 | /* |
| 540 | * Must be held any time you expect node_start_pfn, node_present_pages |
| 541 | * or node_spanned_pages stay constant. Holding this will also |
| 542 | * guarantee that any pfn_valid() stays that way. |
| 543 | * |
| 544 | * Nests above zone->lock and zone->size_seqlock. |
| 545 | */ |
| 546 | spinlock_t node_size_lock; |
| 547 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 548 | unsigned long node_start_pfn; |
| 549 | unsigned long node_present_pages; /* total number of physical pages */ |
| 550 | unsigned long node_spanned_pages; /* total size of physical page |
| 551 | range, including holes */ |
| 552 | int node_id; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 553 | wait_queue_head_t kswapd_wait; |
| 554 | struct task_struct *kswapd; |
| 555 | int kswapd_max_order; |
| 556 | } pg_data_t; |
| 557 | |
| 558 | #define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages) |
| 559 | #define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages) |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 560 | #ifdef CONFIG_FLAT_NODE_MEM_MAP |
Dave Hansen | 408fde8 | 2005-06-23 00:07:37 -0700 | [diff] [blame] | 561 | #define pgdat_page_nr(pgdat, pagenr) ((pgdat)->node_mem_map + (pagenr)) |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 562 | #else |
| 563 | #define pgdat_page_nr(pgdat, pagenr) pfn_to_page((pgdat)->node_start_pfn + (pagenr)) |
| 564 | #endif |
Dave Hansen | 408fde8 | 2005-06-23 00:07:37 -0700 | [diff] [blame] | 565 | #define nid_page_nr(nid, pagenr) pgdat_page_nr(NODE_DATA(nid),(pagenr)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 566 | |
Dave Hansen | 208d54e | 2005-10-29 18:16:52 -0700 | [diff] [blame] | 567 | #include <linux/memory_hotplug.h> |
| 568 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 569 | void get_zone_counts(unsigned long *active, unsigned long *inactive, |
| 570 | unsigned long *free); |
| 571 | void build_all_zonelists(void); |
| 572 | void wakeup_kswapd(struct zone *zone, int order); |
| 573 | int zone_watermark_ok(struct zone *z, int order, unsigned long mark, |
Rohit Seth | 7fb1d9f | 2005-11-13 16:06:43 -0800 | [diff] [blame] | 574 | int classzone_idx, int alloc_flags); |
Dave Hansen | a2f3aa0 | 2007-01-10 23:15:30 -0800 | [diff] [blame] | 575 | enum memmap_context { |
| 576 | MEMMAP_EARLY, |
| 577 | MEMMAP_HOTPLUG, |
| 578 | }; |
Yasunori Goto | 718127c | 2006-06-23 02:03:10 -0700 | [diff] [blame] | 579 | extern int init_currently_empty_zone(struct zone *zone, unsigned long start_pfn, |
Dave Hansen | a2f3aa0 | 2007-01-10 23:15:30 -0800 | [diff] [blame] | 580 | unsigned long size, |
| 581 | enum memmap_context context); |
Yasunori Goto | 718127c | 2006-06-23 02:03:10 -0700 | [diff] [blame] | 582 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 583 | #ifdef CONFIG_HAVE_MEMORY_PRESENT |
| 584 | void memory_present(int nid, unsigned long start, unsigned long end); |
| 585 | #else |
| 586 | static inline void memory_present(int nid, unsigned long start, unsigned long end) {} |
| 587 | #endif |
| 588 | |
| 589 | #ifdef CONFIG_NEED_NODE_MEMMAP_SIZE |
| 590 | unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long); |
| 591 | #endif |
| 592 | |
| 593 | /* |
| 594 | * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc. |
| 595 | */ |
| 596 | #define zone_idx(zone) ((zone) - (zone)->zone_pgdat->node_zones) |
| 597 | |
Con Kolivas | f3fe651 | 2006-01-06 00:11:15 -0800 | [diff] [blame] | 598 | static inline int populated_zone(struct zone *zone) |
| 599 | { |
| 600 | return (!!zone->present_pages); |
| 601 | } |
| 602 | |
Mel Gorman | 2a1e274 | 2007-07-17 04:03:12 -0700 | [diff] [blame] | 603 | extern int movable_zone; |
| 604 | |
| 605 | static inline int zone_movable_is_highmem(void) |
| 606 | { |
| 607 | #if defined(CONFIG_HIGHMEM) && defined(CONFIG_ARCH_POPULATES_NODE_MAP) |
| 608 | return movable_zone == ZONE_HIGHMEM; |
| 609 | #else |
| 610 | return 0; |
| 611 | #endif |
| 612 | } |
| 613 | |
Christoph Lameter | 2f1b624 | 2006-09-25 23:31:13 -0700 | [diff] [blame] | 614 | static inline int is_highmem_idx(enum zone_type idx) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 615 | { |
Christoph Lameter | e53ef38 | 2006-09-25 23:31:14 -0700 | [diff] [blame] | 616 | #ifdef CONFIG_HIGHMEM |
Mel Gorman | 2a1e274 | 2007-07-17 04:03:12 -0700 | [diff] [blame] | 617 | return (idx == ZONE_HIGHMEM || |
| 618 | (idx == ZONE_MOVABLE && zone_movable_is_highmem())); |
Christoph Lameter | e53ef38 | 2006-09-25 23:31:14 -0700 | [diff] [blame] | 619 | #else |
| 620 | return 0; |
| 621 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 622 | } |
| 623 | |
Christoph Lameter | 2f1b624 | 2006-09-25 23:31:13 -0700 | [diff] [blame] | 624 | static inline int is_normal_idx(enum zone_type idx) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 625 | { |
| 626 | return (idx == ZONE_NORMAL); |
| 627 | } |
Nick Piggin | 9328b8f | 2006-01-06 00:11:10 -0800 | [diff] [blame] | 628 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 629 | /** |
| 630 | * is_highmem - helper function to quickly check if a struct zone is a |
| 631 | * highmem zone or not. This is an attempt to keep references |
| 632 | * to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum. |
| 633 | * @zone - pointer to struct zone variable |
| 634 | */ |
| 635 | static inline int is_highmem(struct zone *zone) |
| 636 | { |
Christoph Lameter | e53ef38 | 2006-09-25 23:31:14 -0700 | [diff] [blame] | 637 | #ifdef CONFIG_HIGHMEM |
Harvey Harrison | ddc81ed | 2008-04-28 02:12:07 -0700 | [diff] [blame] | 638 | int zone_off = (char *)zone - (char *)zone->zone_pgdat->node_zones; |
| 639 | return zone_off == ZONE_HIGHMEM * sizeof(*zone) || |
| 640 | (zone_off == ZONE_MOVABLE * sizeof(*zone) && |
| 641 | zone_movable_is_highmem()); |
Christoph Lameter | e53ef38 | 2006-09-25 23:31:14 -0700 | [diff] [blame] | 642 | #else |
| 643 | return 0; |
| 644 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 645 | } |
| 646 | |
| 647 | static inline int is_normal(struct zone *zone) |
| 648 | { |
| 649 | return zone == zone->zone_pgdat->node_zones + ZONE_NORMAL; |
| 650 | } |
| 651 | |
Nick Piggin | 9328b8f | 2006-01-06 00:11:10 -0800 | [diff] [blame] | 652 | static inline int is_dma32(struct zone *zone) |
| 653 | { |
Christoph Lameter | fb0e794 | 2006-09-25 23:31:13 -0700 | [diff] [blame] | 654 | #ifdef CONFIG_ZONE_DMA32 |
Nick Piggin | 9328b8f | 2006-01-06 00:11:10 -0800 | [diff] [blame] | 655 | return zone == zone->zone_pgdat->node_zones + ZONE_DMA32; |
Christoph Lameter | fb0e794 | 2006-09-25 23:31:13 -0700 | [diff] [blame] | 656 | #else |
| 657 | return 0; |
| 658 | #endif |
Nick Piggin | 9328b8f | 2006-01-06 00:11:10 -0800 | [diff] [blame] | 659 | } |
| 660 | |
| 661 | static inline int is_dma(struct zone *zone) |
| 662 | { |
Christoph Lameter | 4b51d66 | 2007-02-10 01:43:10 -0800 | [diff] [blame] | 663 | #ifdef CONFIG_ZONE_DMA |
Nick Piggin | 9328b8f | 2006-01-06 00:11:10 -0800 | [diff] [blame] | 664 | return zone == zone->zone_pgdat->node_zones + ZONE_DMA; |
Christoph Lameter | 4b51d66 | 2007-02-10 01:43:10 -0800 | [diff] [blame] | 665 | #else |
| 666 | return 0; |
| 667 | #endif |
Nick Piggin | 9328b8f | 2006-01-06 00:11:10 -0800 | [diff] [blame] | 668 | } |
| 669 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 670 | /* These two functions are used to setup the per zone pages min values */ |
| 671 | struct ctl_table; |
| 672 | struct file; |
| 673 | int min_free_kbytes_sysctl_handler(struct ctl_table *, int, struct file *, |
| 674 | void __user *, size_t *, loff_t *); |
| 675 | extern int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1]; |
| 676 | int lowmem_reserve_ratio_sysctl_handler(struct ctl_table *, int, struct file *, |
| 677 | void __user *, size_t *, loff_t *); |
Rohit Seth | 8ad4b1f | 2006-01-08 01:00:40 -0800 | [diff] [blame] | 678 | int percpu_pagelist_fraction_sysctl_handler(struct ctl_table *, int, struct file *, |
| 679 | void __user *, size_t *, loff_t *); |
Christoph Lameter | 9614634 | 2006-07-03 00:24:13 -0700 | [diff] [blame] | 680 | int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table *, int, |
| 681 | struct file *, void __user *, size_t *, loff_t *); |
Christoph Lameter | 0ff3849 | 2006-09-25 23:31:52 -0700 | [diff] [blame] | 682 | int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table *, int, |
| 683 | struct file *, void __user *, size_t *, loff_t *); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 684 | |
KAMEZAWA Hiroyuki | f0c0b2b | 2007-07-15 23:38:01 -0700 | [diff] [blame] | 685 | extern int numa_zonelist_order_handler(struct ctl_table *, int, |
| 686 | struct file *, void __user *, size_t *, loff_t *); |
| 687 | extern char numa_zonelist_order[]; |
| 688 | #define NUMA_ZONELIST_ORDER_LEN 16 /* string buffer size */ |
| 689 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 690 | #include <linux/topology.h> |
| 691 | /* Returns the number of the current Node. */ |
Andi Kleen | 69d81fc | 2005-11-05 17:25:53 +0100 | [diff] [blame] | 692 | #ifndef numa_node_id |
Ingo Molnar | 39c715b | 2005-06-21 17:14:34 -0700 | [diff] [blame] | 693 | #define numa_node_id() (cpu_to_node(raw_smp_processor_id())) |
Andi Kleen | 69d81fc | 2005-11-05 17:25:53 +0100 | [diff] [blame] | 694 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 695 | |
Dave Hansen | 93b7504 | 2005-06-23 00:07:47 -0700 | [diff] [blame] | 696 | #ifndef CONFIG_NEED_MULTIPLE_NODES |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 697 | |
| 698 | extern struct pglist_data contig_page_data; |
| 699 | #define NODE_DATA(nid) (&contig_page_data) |
| 700 | #define NODE_MEM_MAP(nid) mem_map |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 701 | |
Dave Hansen | 93b7504 | 2005-06-23 00:07:47 -0700 | [diff] [blame] | 702 | #else /* CONFIG_NEED_MULTIPLE_NODES */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 703 | |
| 704 | #include <asm/mmzone.h> |
| 705 | |
Dave Hansen | 93b7504 | 2005-06-23 00:07:47 -0700 | [diff] [blame] | 706 | #endif /* !CONFIG_NEED_MULTIPLE_NODES */ |
Dave Hansen | 348f8b6 | 2005-06-23 00:07:40 -0700 | [diff] [blame] | 707 | |
KAMEZAWA Hiroyuki | 95144c7 | 2006-03-27 01:16:02 -0800 | [diff] [blame] | 708 | extern struct pglist_data *first_online_pgdat(void); |
| 709 | extern struct pglist_data *next_online_pgdat(struct pglist_data *pgdat); |
| 710 | extern struct zone *next_zone(struct zone *zone); |
KAMEZAWA Hiroyuki | 8357f86 | 2006-03-27 01:15:57 -0800 | [diff] [blame] | 711 | |
| 712 | /** |
| 713 | * for_each_pgdat - helper macro to iterate over all nodes |
| 714 | * @pgdat - pointer to a pg_data_t variable |
| 715 | */ |
| 716 | #define for_each_online_pgdat(pgdat) \ |
| 717 | for (pgdat = first_online_pgdat(); \ |
| 718 | pgdat; \ |
| 719 | pgdat = next_online_pgdat(pgdat)) |
KAMEZAWA Hiroyuki | 8357f86 | 2006-03-27 01:15:57 -0800 | [diff] [blame] | 720 | /** |
| 721 | * for_each_zone - helper macro to iterate over all memory zones |
| 722 | * @zone - pointer to struct zone variable |
| 723 | * |
| 724 | * The user only needs to declare the zone variable, for_each_zone |
| 725 | * fills it in. |
| 726 | */ |
| 727 | #define for_each_zone(zone) \ |
| 728 | for (zone = (first_online_pgdat())->node_zones; \ |
| 729 | zone; \ |
| 730 | zone = next_zone(zone)) |
| 731 | |
Mel Gorman | dd1a239 | 2008-04-28 02:12:17 -0700 | [diff] [blame^] | 732 | static inline struct zone *zonelist_zone(struct zoneref *zoneref) |
| 733 | { |
| 734 | return zoneref->zone; |
| 735 | } |
| 736 | |
| 737 | static inline int zonelist_zone_idx(struct zoneref *zoneref) |
| 738 | { |
| 739 | return zoneref->zone_idx; |
| 740 | } |
| 741 | |
| 742 | static inline int zonelist_node_idx(struct zoneref *zoneref) |
| 743 | { |
| 744 | #ifdef CONFIG_NUMA |
| 745 | /* zone_to_nid not available in this context */ |
| 746 | return zoneref->zone->node; |
| 747 | #else |
| 748 | return 0; |
| 749 | #endif /* CONFIG_NUMA */ |
| 750 | } |
| 751 | |
| 752 | static inline void zoneref_set_zone(struct zone *zone, struct zoneref *zoneref) |
| 753 | { |
| 754 | zoneref->zone = zone; |
| 755 | zoneref->zone_idx = zone_idx(zone); |
| 756 | } |
| 757 | |
Mel Gorman | 54a6eb5 | 2008-04-28 02:12:16 -0700 | [diff] [blame] | 758 | /* Returns the first zone at or below highest_zoneidx in a zonelist */ |
Mel Gorman | dd1a239 | 2008-04-28 02:12:17 -0700 | [diff] [blame^] | 759 | static inline struct zoneref *first_zones_zonelist(struct zonelist *zonelist, |
Mel Gorman | 54a6eb5 | 2008-04-28 02:12:16 -0700 | [diff] [blame] | 760 | enum zone_type highest_zoneidx) |
| 761 | { |
Mel Gorman | dd1a239 | 2008-04-28 02:12:17 -0700 | [diff] [blame^] | 762 | struct zoneref *z; |
Mel Gorman | 54a6eb5 | 2008-04-28 02:12:16 -0700 | [diff] [blame] | 763 | |
| 764 | /* Find the first suitable zone to use for the allocation */ |
Mel Gorman | dd1a239 | 2008-04-28 02:12:17 -0700 | [diff] [blame^] | 765 | z = zonelist->_zonerefs; |
| 766 | while (zonelist_zone_idx(z) > highest_zoneidx) |
Mel Gorman | 54a6eb5 | 2008-04-28 02:12:16 -0700 | [diff] [blame] | 767 | z++; |
| 768 | |
| 769 | return z; |
| 770 | } |
| 771 | |
| 772 | /* Returns the next zone at or below highest_zoneidx in a zonelist */ |
Mel Gorman | dd1a239 | 2008-04-28 02:12:17 -0700 | [diff] [blame^] | 773 | static inline struct zoneref *next_zones_zonelist(struct zoneref *z, |
Mel Gorman | 54a6eb5 | 2008-04-28 02:12:16 -0700 | [diff] [blame] | 774 | enum zone_type highest_zoneidx) |
| 775 | { |
| 776 | /* Find the next suitable zone to use for the allocation */ |
Mel Gorman | dd1a239 | 2008-04-28 02:12:17 -0700 | [diff] [blame^] | 777 | while (zonelist_zone_idx(z) > highest_zoneidx) |
Mel Gorman | 54a6eb5 | 2008-04-28 02:12:16 -0700 | [diff] [blame] | 778 | z++; |
| 779 | |
| 780 | return z; |
| 781 | } |
| 782 | |
| 783 | /** |
| 784 | * for_each_zone_zonelist - helper macro to iterate over valid zones in a zonelist at or below a given zone index |
| 785 | * @zone - The current zone in the iterator |
| 786 | * @z - The current pointer within zonelist->zones being iterated |
| 787 | * @zlist - The zonelist being iterated |
| 788 | * @highidx - The zone index of the highest zone to return |
| 789 | * |
| 790 | * This iterator iterates though all zones at or below a given zone index. |
| 791 | */ |
| 792 | #define for_each_zone_zonelist(zone, z, zlist, highidx) \ |
Mel Gorman | dd1a239 | 2008-04-28 02:12:17 -0700 | [diff] [blame^] | 793 | for (z = first_zones_zonelist(zlist, highidx), \ |
| 794 | zone = zonelist_zone(z++); \ |
Mel Gorman | 54a6eb5 | 2008-04-28 02:12:16 -0700 | [diff] [blame] | 795 | zone; \ |
Mel Gorman | dd1a239 | 2008-04-28 02:12:17 -0700 | [diff] [blame^] | 796 | z = next_zones_zonelist(z, highidx), \ |
| 797 | zone = zonelist_zone(z++)) |
Mel Gorman | 54a6eb5 | 2008-04-28 02:12:16 -0700 | [diff] [blame] | 798 | |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 799 | #ifdef CONFIG_SPARSEMEM |
| 800 | #include <asm/sparsemem.h> |
| 801 | #endif |
| 802 | |
Andi Kleen | 07808b7 | 2005-11-05 17:25:53 +0100 | [diff] [blame] | 803 | #if BITS_PER_LONG == 32 |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 804 | /* |
Andi Kleen | a2f1b42 | 2005-11-05 17:25:53 +0100 | [diff] [blame] | 805 | * with 32 bit page->flags field, we reserve 9 bits for node/zone info. |
| 806 | * there are 4 zones (3 bits) and this leaves 9-3=6 bits for nodes. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 807 | */ |
Andi Kleen | a2f1b42 | 2005-11-05 17:25:53 +0100 | [diff] [blame] | 808 | #define FLAGS_RESERVED 9 |
Dave Hansen | 348f8b6 | 2005-06-23 00:07:40 -0700 | [diff] [blame] | 809 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 810 | #elif BITS_PER_LONG == 64 |
| 811 | /* |
| 812 | * with 64 bit flags field, there's plenty of room. |
| 813 | */ |
Dave Hansen | 348f8b6 | 2005-06-23 00:07:40 -0700 | [diff] [blame] | 814 | #define FLAGS_RESERVED 32 |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 815 | |
Dave Hansen | 348f8b6 | 2005-06-23 00:07:40 -0700 | [diff] [blame] | 816 | #else |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 817 | |
Dave Hansen | 348f8b6 | 2005-06-23 00:07:40 -0700 | [diff] [blame] | 818 | #error BITS_PER_LONG not defined |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 819 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 820 | #endif |
| 821 | |
Mel Gorman | c713216 | 2006-09-27 01:49:43 -0700 | [diff] [blame] | 822 | #if !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) && \ |
| 823 | !defined(CONFIG_ARCH_POPULATES_NODE_MAP) |
Andy Whitcroft | b159d43 | 2005-06-23 00:07:52 -0700 | [diff] [blame] | 824 | #define early_pfn_to_nid(nid) (0UL) |
| 825 | #endif |
| 826 | |
Andy Whitcroft | 2bdaf11 | 2006-01-06 00:10:53 -0800 | [diff] [blame] | 827 | #ifdef CONFIG_FLATMEM |
| 828 | #define pfn_to_nid(pfn) (0) |
| 829 | #endif |
| 830 | |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 831 | #define pfn_to_section_nr(pfn) ((pfn) >> PFN_SECTION_SHIFT) |
| 832 | #define section_nr_to_pfn(sec) ((sec) << PFN_SECTION_SHIFT) |
| 833 | |
| 834 | #ifdef CONFIG_SPARSEMEM |
| 835 | |
| 836 | /* |
| 837 | * SECTION_SHIFT #bits space required to store a section # |
| 838 | * |
| 839 | * PA_SECTION_SHIFT physical address to/from section number |
| 840 | * PFN_SECTION_SHIFT pfn to/from section number |
| 841 | */ |
| 842 | #define SECTIONS_SHIFT (MAX_PHYSMEM_BITS - SECTION_SIZE_BITS) |
| 843 | |
| 844 | #define PA_SECTION_SHIFT (SECTION_SIZE_BITS) |
| 845 | #define PFN_SECTION_SHIFT (SECTION_SIZE_BITS - PAGE_SHIFT) |
| 846 | |
| 847 | #define NR_MEM_SECTIONS (1UL << SECTIONS_SHIFT) |
| 848 | |
| 849 | #define PAGES_PER_SECTION (1UL << PFN_SECTION_SHIFT) |
| 850 | #define PAGE_SECTION_MASK (~(PAGES_PER_SECTION-1)) |
| 851 | |
Mel Gorman | 835c134 | 2007-10-16 01:25:47 -0700 | [diff] [blame] | 852 | #define SECTION_BLOCKFLAGS_BITS \ |
Mel Gorman | d9c2340 | 2007-10-16 01:26:01 -0700 | [diff] [blame] | 853 | ((1UL << (PFN_SECTION_SHIFT - pageblock_order)) * NR_PAGEBLOCK_BITS) |
Mel Gorman | 835c134 | 2007-10-16 01:25:47 -0700 | [diff] [blame] | 854 | |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 855 | #if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS |
| 856 | #error Allocator MAX_ORDER exceeds SECTION_SIZE |
| 857 | #endif |
| 858 | |
| 859 | struct page; |
| 860 | struct mem_section { |
Andy Whitcroft | 29751f6 | 2005-06-23 00:08:00 -0700 | [diff] [blame] | 861 | /* |
| 862 | * This is, logically, a pointer to an array of struct |
| 863 | * pages. However, it is stored with some other magic. |
| 864 | * (see sparse.c::sparse_init_one_section()) |
| 865 | * |
Andy Whitcroft | 30c253e | 2006-06-23 02:03:41 -0700 | [diff] [blame] | 866 | * Additionally during early boot we encode node id of |
| 867 | * the location of the section here to guide allocation. |
| 868 | * (see sparse.c::memory_present()) |
| 869 | * |
Andy Whitcroft | 29751f6 | 2005-06-23 00:08:00 -0700 | [diff] [blame] | 870 | * Making it a UL at least makes someone do a cast |
| 871 | * before using it wrong. |
| 872 | */ |
| 873 | unsigned long section_mem_map; |
Mel Gorman | 5c0e306 | 2007-10-16 01:25:56 -0700 | [diff] [blame] | 874 | |
| 875 | /* See declaration of similar field in struct zone */ |
| 876 | unsigned long *pageblock_flags; |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 877 | }; |
| 878 | |
Bob Picco | 3e34726 | 2005-09-03 15:54:28 -0700 | [diff] [blame] | 879 | #ifdef CONFIG_SPARSEMEM_EXTREME |
| 880 | #define SECTIONS_PER_ROOT (PAGE_SIZE / sizeof (struct mem_section)) |
Bob Picco | 802f192 | 2005-09-03 15:54:26 -0700 | [diff] [blame] | 881 | #else |
Bob Picco | 3e34726 | 2005-09-03 15:54:28 -0700 | [diff] [blame] | 882 | #define SECTIONS_PER_ROOT 1 |
| 883 | #endif |
Bob Picco | 802f192 | 2005-09-03 15:54:26 -0700 | [diff] [blame] | 884 | |
Bob Picco | 3e34726 | 2005-09-03 15:54:28 -0700 | [diff] [blame] | 885 | #define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT) |
| 886 | #define NR_SECTION_ROOTS (NR_MEM_SECTIONS / SECTIONS_PER_ROOT) |
| 887 | #define SECTION_ROOT_MASK (SECTIONS_PER_ROOT - 1) |
| 888 | |
| 889 | #ifdef CONFIG_SPARSEMEM_EXTREME |
| 890 | extern struct mem_section *mem_section[NR_SECTION_ROOTS]; |
| 891 | #else |
| 892 | extern struct mem_section mem_section[NR_SECTION_ROOTS][SECTIONS_PER_ROOT]; |
| 893 | #endif |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 894 | |
Andy Whitcroft | 29751f6 | 2005-06-23 00:08:00 -0700 | [diff] [blame] | 895 | static inline struct mem_section *__nr_to_section(unsigned long nr) |
| 896 | { |
Bob Picco | 3e34726 | 2005-09-03 15:54:28 -0700 | [diff] [blame] | 897 | if (!mem_section[SECTION_NR_TO_ROOT(nr)]) |
| 898 | return NULL; |
| 899 | return &mem_section[SECTION_NR_TO_ROOT(nr)][nr & SECTION_ROOT_MASK]; |
Andy Whitcroft | 29751f6 | 2005-06-23 00:08:00 -0700 | [diff] [blame] | 900 | } |
Dave Hansen | 4ca644d | 2005-10-29 18:16:51 -0700 | [diff] [blame] | 901 | extern int __section_nr(struct mem_section* ms); |
Andy Whitcroft | 29751f6 | 2005-06-23 00:08:00 -0700 | [diff] [blame] | 902 | |
| 903 | /* |
| 904 | * We use the lower bits of the mem_map pointer to store |
| 905 | * a little bit of information. There should be at least |
| 906 | * 3 bits here due to 32-bit alignment. |
| 907 | */ |
| 908 | #define SECTION_MARKED_PRESENT (1UL<<0) |
| 909 | #define SECTION_HAS_MEM_MAP (1UL<<1) |
| 910 | #define SECTION_MAP_LAST_BIT (1UL<<2) |
| 911 | #define SECTION_MAP_MASK (~(SECTION_MAP_LAST_BIT-1)) |
Andy Whitcroft | 30c253e | 2006-06-23 02:03:41 -0700 | [diff] [blame] | 912 | #define SECTION_NID_SHIFT 2 |
Andy Whitcroft | 29751f6 | 2005-06-23 00:08:00 -0700 | [diff] [blame] | 913 | |
| 914 | static inline struct page *__section_mem_map_addr(struct mem_section *section) |
| 915 | { |
| 916 | unsigned long map = section->section_mem_map; |
| 917 | map &= SECTION_MAP_MASK; |
| 918 | return (struct page *)map; |
| 919 | } |
| 920 | |
Andy Whitcroft | 540557b | 2007-10-16 01:24:11 -0700 | [diff] [blame] | 921 | static inline int present_section(struct mem_section *section) |
Andy Whitcroft | 29751f6 | 2005-06-23 00:08:00 -0700 | [diff] [blame] | 922 | { |
Bob Picco | 802f192 | 2005-09-03 15:54:26 -0700 | [diff] [blame] | 923 | return (section && (section->section_mem_map & SECTION_MARKED_PRESENT)); |
Andy Whitcroft | 29751f6 | 2005-06-23 00:08:00 -0700 | [diff] [blame] | 924 | } |
| 925 | |
Andy Whitcroft | 540557b | 2007-10-16 01:24:11 -0700 | [diff] [blame] | 926 | static inline int present_section_nr(unsigned long nr) |
| 927 | { |
| 928 | return present_section(__nr_to_section(nr)); |
| 929 | } |
| 930 | |
| 931 | static inline int valid_section(struct mem_section *section) |
Andy Whitcroft | 29751f6 | 2005-06-23 00:08:00 -0700 | [diff] [blame] | 932 | { |
Bob Picco | 802f192 | 2005-09-03 15:54:26 -0700 | [diff] [blame] | 933 | return (section && (section->section_mem_map & SECTION_HAS_MEM_MAP)); |
Andy Whitcroft | 29751f6 | 2005-06-23 00:08:00 -0700 | [diff] [blame] | 934 | } |
| 935 | |
| 936 | static inline int valid_section_nr(unsigned long nr) |
| 937 | { |
| 938 | return valid_section(__nr_to_section(nr)); |
| 939 | } |
| 940 | |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 941 | static inline struct mem_section *__pfn_to_section(unsigned long pfn) |
| 942 | { |
Andy Whitcroft | 29751f6 | 2005-06-23 00:08:00 -0700 | [diff] [blame] | 943 | return __nr_to_section(pfn_to_section_nr(pfn)); |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 944 | } |
| 945 | |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 946 | static inline int pfn_valid(unsigned long pfn) |
| 947 | { |
| 948 | if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS) |
| 949 | return 0; |
Andy Whitcroft | 29751f6 | 2005-06-23 00:08:00 -0700 | [diff] [blame] | 950 | return valid_section(__nr_to_section(pfn_to_section_nr(pfn))); |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 951 | } |
| 952 | |
Andy Whitcroft | 540557b | 2007-10-16 01:24:11 -0700 | [diff] [blame] | 953 | static inline int pfn_present(unsigned long pfn) |
| 954 | { |
| 955 | if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS) |
| 956 | return 0; |
| 957 | return present_section(__nr_to_section(pfn_to_section_nr(pfn))); |
| 958 | } |
| 959 | |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 960 | /* |
| 961 | * These are _only_ used during initialisation, therefore they |
| 962 | * can use __initdata ... They could have names to indicate |
| 963 | * this restriction. |
| 964 | */ |
| 965 | #ifdef CONFIG_NUMA |
Andy Whitcroft | 161599f | 2006-01-06 00:10:54 -0800 | [diff] [blame] | 966 | #define pfn_to_nid(pfn) \ |
| 967 | ({ \ |
| 968 | unsigned long __pfn_to_nid_pfn = (pfn); \ |
| 969 | page_to_nid(pfn_to_page(__pfn_to_nid_pfn)); \ |
| 970 | }) |
Andy Whitcroft | 2bdaf11 | 2006-01-06 00:10:53 -0800 | [diff] [blame] | 971 | #else |
| 972 | #define pfn_to_nid(pfn) (0) |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 973 | #endif |
| 974 | |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 975 | #define early_pfn_valid(pfn) pfn_valid(pfn) |
| 976 | void sparse_init(void); |
| 977 | #else |
| 978 | #define sparse_init() do {} while (0) |
Dave Hansen | 28ae55c | 2005-09-03 15:54:29 -0700 | [diff] [blame] | 979 | #define sparse_index_init(_sec, _nid) do {} while (0) |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 980 | #endif /* CONFIG_SPARSEMEM */ |
| 981 | |
Andy Whitcroft | 7516795 | 2006-10-21 10:24:14 -0700 | [diff] [blame] | 982 | #ifdef CONFIG_NODES_SPAN_OTHER_NODES |
| 983 | #define early_pfn_in_nid(pfn, nid) (early_pfn_to_nid(pfn) == (nid)) |
| 984 | #else |
| 985 | #define early_pfn_in_nid(pfn, nid) (1) |
| 986 | #endif |
| 987 | |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 988 | #ifndef early_pfn_valid |
| 989 | #define early_pfn_valid(pfn) (1) |
| 990 | #endif |
| 991 | |
| 992 | void memory_present(int nid, unsigned long start, unsigned long end); |
| 993 | unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long); |
| 994 | |
Andy Whitcroft | 14e0729 | 2007-05-06 14:49:14 -0700 | [diff] [blame] | 995 | /* |
| 996 | * If it is possible to have holes within a MAX_ORDER_NR_PAGES, then we |
| 997 | * need to check pfn validility within that MAX_ORDER_NR_PAGES block. |
| 998 | * pfn_valid_within() should be used in this case; we optimise this away |
| 999 | * when we have no holes within a MAX_ORDER_NR_PAGES block. |
| 1000 | */ |
| 1001 | #ifdef CONFIG_HOLES_IN_ZONE |
| 1002 | #define pfn_valid_within(pfn) pfn_valid(pfn) |
| 1003 | #else |
| 1004 | #define pfn_valid_within(pfn) (1) |
| 1005 | #endif |
| 1006 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1007 | #endif /* !__ASSEMBLY__ */ |
| 1008 | #endif /* __KERNEL__ */ |
| 1009 | #endif /* _LINUX_MMZONE_H */ |