Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | #ifndef _LINUX_MMZONE_H |
| 2 | #define _LINUX_MMZONE_H |
| 3 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 4 | #ifndef __ASSEMBLY__ |
Christoph Lameter | 9796547 | 2008-04-28 02:12:54 -0700 | [diff] [blame] | 5 | #ifndef __GENERATING_BOUNDS_H |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 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> |
Peter Zijlstra | bbeae5b | 2013-02-22 16:34:30 -0800 | [diff] [blame] | 18 | #include <linux/page-flags-layout.h> |
Arun Sharma | 60063497 | 2011-07-26 16:09:06 -0700 | [diff] [blame] | 19 | #include <linux/atomic.h> |
Ralf Baechle | 93ff66b | 2006-06-04 02:51:29 -0700 | [diff] [blame] | 20 | #include <asm/page.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 21 | |
| 22 | /* Free memory management - zoned buddy allocator. */ |
| 23 | #ifndef CONFIG_FORCE_MAX_ZONEORDER |
| 24 | #define MAX_ORDER 11 |
| 25 | #else |
| 26 | #define MAX_ORDER CONFIG_FORCE_MAX_ZONEORDER |
| 27 | #endif |
Bob Picco | e984bb4 | 2006-05-20 15:00:31 -0700 | [diff] [blame] | 28 | #define MAX_ORDER_NR_PAGES (1 << (MAX_ORDER - 1)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 29 | |
Andy Whitcroft | 5ad333e | 2007-07-17 04:03:16 -0700 | [diff] [blame] | 30 | /* |
| 31 | * PAGE_ALLOC_COSTLY_ORDER is the order at which allocations are deemed |
| 32 | * costly to service. That is between allocation orders which should |
Wang YanQing | 35fca53 | 2012-04-15 20:42:28 +0800 | [diff] [blame] | 33 | * coalesce naturally under reasonable reclaim pressure and those which |
Andy Whitcroft | 5ad333e | 2007-07-17 04:03:16 -0700 | [diff] [blame] | 34 | * will not. |
| 35 | */ |
| 36 | #define PAGE_ALLOC_COSTLY_ORDER 3 |
| 37 | |
Michal Nazarewicz | 47118af | 2011-12-29 13:09:50 +0100 | [diff] [blame] | 38 | enum { |
| 39 | MIGRATE_UNMOVABLE, |
| 40 | MIGRATE_RECLAIMABLE, |
| 41 | MIGRATE_MOVABLE, |
| 42 | MIGRATE_PCPTYPES, /* the number of types on the pcp lists */ |
| 43 | MIGRATE_RESERVE = MIGRATE_PCPTYPES, |
| 44 | #ifdef CONFIG_CMA |
| 45 | /* |
| 46 | * MIGRATE_CMA migration type is designed to mimic the way |
| 47 | * ZONE_MOVABLE works. Only movable pages can be allocated |
| 48 | * from MIGRATE_CMA pageblocks and page allocator never |
| 49 | * implicitly change migration type of MIGRATE_CMA pageblock. |
| 50 | * |
| 51 | * The way to use it is to change migratetype of a range of |
| 52 | * pageblocks to MIGRATE_CMA which can be done by |
| 53 | * __free_pageblock_cma() function. What is important though |
| 54 | * is that a range of pageblocks must be aligned to |
| 55 | * MAX_ORDER_NR_PAGES should biggest page be bigger then |
| 56 | * a single pageblock. |
| 57 | */ |
| 58 | MIGRATE_CMA, |
| 59 | #endif |
Minchan Kim | 194159f | 2013-02-22 16:33:58 -0800 | [diff] [blame] | 60 | #ifdef CONFIG_MEMORY_ISOLATION |
Michal Nazarewicz | 47118af | 2011-12-29 13:09:50 +0100 | [diff] [blame] | 61 | MIGRATE_ISOLATE, /* can't allocate from here */ |
Minchan Kim | 194159f | 2013-02-22 16:33:58 -0800 | [diff] [blame] | 62 | #endif |
Michal Nazarewicz | 47118af | 2011-12-29 13:09:50 +0100 | [diff] [blame] | 63 | MIGRATE_TYPES |
| 64 | }; |
| 65 | |
| 66 | #ifdef CONFIG_CMA |
| 67 | # define is_migrate_cma(migratetype) unlikely((migratetype) == MIGRATE_CMA) |
| 68 | #else |
| 69 | # define is_migrate_cma(migratetype) false |
| 70 | #endif |
Mel Gorman | b2a0ac8 | 2007-10-16 01:25:48 -0700 | [diff] [blame] | 71 | |
| 72 | #define for_each_migratetype_order(order, type) \ |
| 73 | for (order = 0; order < MAX_ORDER; order++) \ |
| 74 | for (type = 0; type < MIGRATE_TYPES; type++) |
| 75 | |
Mel Gorman | 467c996 | 2007-10-16 01:26:02 -0700 | [diff] [blame] | 76 | extern int page_group_by_mobility_disabled; |
| 77 | |
Mel Gorman | e58469b | 2014-06-04 16:10:16 -0700 | [diff] [blame] | 78 | #define NR_MIGRATETYPE_BITS (PB_migrate_end - PB_migrate + 1) |
| 79 | #define MIGRATETYPE_MASK ((1UL << NR_MIGRATETYPE_BITS) - 1) |
| 80 | |
Mel Gorman | dc4b0ca | 2014-06-04 16:10:17 -0700 | [diff] [blame] | 81 | #define get_pageblock_migratetype(page) \ |
| 82 | get_pfnblock_flags_mask(page, page_to_pfn(page), \ |
| 83 | PB_migrate_end, MIGRATETYPE_MASK) |
| 84 | |
| 85 | static inline int get_pfnblock_migratetype(struct page *page, unsigned long pfn) |
Mel Gorman | 467c996 | 2007-10-16 01:26:02 -0700 | [diff] [blame] | 86 | { |
Mel Gorman | e58469b | 2014-06-04 16:10:16 -0700 | [diff] [blame] | 87 | BUILD_BUG_ON(PB_migrate_end - PB_migrate != 2); |
Mel Gorman | dc4b0ca | 2014-06-04 16:10:17 -0700 | [diff] [blame] | 88 | return get_pfnblock_flags_mask(page, pfn, PB_migrate_end, |
| 89 | MIGRATETYPE_MASK); |
Mel Gorman | 467c996 | 2007-10-16 01:26:02 -0700 | [diff] [blame] | 90 | } |
| 91 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 92 | struct free_area { |
Mel Gorman | b2a0ac8 | 2007-10-16 01:25:48 -0700 | [diff] [blame] | 93 | struct list_head free_list[MIGRATE_TYPES]; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 94 | unsigned long nr_free; |
| 95 | }; |
| 96 | |
| 97 | struct pglist_data; |
| 98 | |
| 99 | /* |
| 100 | * zone->lock and zone->lru_lock are two of the hottest locks in the kernel. |
| 101 | * So add a wild amount of padding here to ensure that they fall into separate |
| 102 | * cachelines. There are very few zone structures in the machine, so space |
| 103 | * consumption is not a concern here. |
| 104 | */ |
| 105 | #if defined(CONFIG_SMP) |
| 106 | struct zone_padding { |
| 107 | char x[0]; |
Ravikiran G Thirumalai | 22fc6ec | 2006-01-08 01:01:27 -0800 | [diff] [blame] | 108 | } ____cacheline_internodealigned_in_smp; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 109 | #define ZONE_PADDING(name) struct zone_padding name; |
| 110 | #else |
| 111 | #define ZONE_PADDING(name) |
| 112 | #endif |
| 113 | |
Christoph Lameter | 2244b95 | 2006-06-30 01:55:33 -0700 | [diff] [blame] | 114 | enum zone_stat_item { |
Christoph Lameter | 51ed449 | 2007-02-10 01:43:02 -0800 | [diff] [blame] | 115 | /* First 128 byte cacheline (assuming 64 bit words) */ |
Christoph Lameter | d23ad42 | 2007-02-10 01:43:02 -0800 | [diff] [blame] | 116 | NR_FREE_PAGES, |
Johannes Weiner | 81c0a2b | 2013-09-11 14:20:47 -0700 | [diff] [blame] | 117 | NR_ALLOC_BATCH, |
Christoph Lameter | b69408e | 2008-10-18 20:26:14 -0700 | [diff] [blame] | 118 | NR_LRU_BASE, |
Rik van Riel | 4f98a2f | 2008-10-18 20:26:32 -0700 | [diff] [blame] | 119 | NR_INACTIVE_ANON = NR_LRU_BASE, /* must match order of LRU_[IN]ACTIVE */ |
| 120 | NR_ACTIVE_ANON, /* " " " " " */ |
| 121 | NR_INACTIVE_FILE, /* " " " " " */ |
| 122 | NR_ACTIVE_FILE, /* " " " " " */ |
Lee Schermerhorn | 894bc31 | 2008-10-18 20:26:39 -0700 | [diff] [blame] | 123 | NR_UNEVICTABLE, /* " " " " " */ |
Nick Piggin | 5344b7e | 2008-10-18 20:26:51 -0700 | [diff] [blame] | 124 | NR_MLOCK, /* mlock()ed pages found and moved off LRU */ |
Christoph Lameter | f3dbd34 | 2006-06-30 01:55:36 -0700 | [diff] [blame] | 125 | NR_ANON_PAGES, /* Mapped anonymous pages */ |
| 126 | NR_FILE_MAPPED, /* pagecache pages mapped into pagetables. |
Christoph Lameter | 65ba55f | 2006-06-30 01:55:34 -0700 | [diff] [blame] | 127 | only modified from process context */ |
Christoph Lameter | 347ce43 | 2006-06-30 01:55:35 -0700 | [diff] [blame] | 128 | NR_FILE_PAGES, |
Christoph Lameter | b1e7a8f | 2006-06-30 01:55:39 -0700 | [diff] [blame] | 129 | NR_FILE_DIRTY, |
Christoph Lameter | ce866b3 | 2006-06-30 01:55:40 -0700 | [diff] [blame] | 130 | NR_WRITEBACK, |
Christoph Lameter | 51ed449 | 2007-02-10 01:43:02 -0800 | [diff] [blame] | 131 | NR_SLAB_RECLAIMABLE, |
| 132 | NR_SLAB_UNRECLAIMABLE, |
| 133 | NR_PAGETABLE, /* used for pagetables */ |
KOSAKI Motohiro | c6a7f57 | 2009-09-21 17:01:32 -0700 | [diff] [blame] | 134 | NR_KERNEL_STACK, |
| 135 | /* Second 128 byte cacheline */ |
Christoph Lameter | fd39fc8 | 2006-06-30 01:55:40 -0700 | [diff] [blame] | 136 | NR_UNSTABLE_NFS, /* NFS unstable pages */ |
Christoph Lameter | d2c5e30 | 2006-06-30 01:55:41 -0700 | [diff] [blame] | 137 | NR_BOUNCE, |
Andrew Morton | e129b5c | 2006-09-27 01:50:00 -0700 | [diff] [blame] | 138 | NR_VMSCAN_WRITE, |
Mel Gorman | 49ea7eb | 2011-10-31 17:07:59 -0700 | [diff] [blame] | 139 | NR_VMSCAN_IMMEDIATE, /* Prioritise for reclaim when writeback ends */ |
Miklos Szeredi | fc3ba69 | 2008-04-30 00:54:38 -0700 | [diff] [blame] | 140 | NR_WRITEBACK_TEMP, /* Writeback using temporary buffers */ |
KOSAKI Motohiro | a731286 | 2009-09-21 17:01:37 -0700 | [diff] [blame] | 141 | NR_ISOLATED_ANON, /* Temporary isolated pages from anon lru */ |
| 142 | NR_ISOLATED_FILE, /* Temporary isolated pages from file lru */ |
KOSAKI Motohiro | 4b02108 | 2009-09-21 17:01:33 -0700 | [diff] [blame] | 143 | NR_SHMEM, /* shmem pages (included tmpfs/GEM pages) */ |
Michael Rubin | ea941f0 | 2010-10-26 14:21:35 -0700 | [diff] [blame] | 144 | NR_DIRTIED, /* page dirtyings since bootup */ |
| 145 | NR_WRITTEN, /* page writings since bootup */ |
Mel Gorman | 0d5d823 | 2014-08-06 16:07:16 -0700 | [diff] [blame] | 146 | NR_PAGES_SCANNED, /* pages scanned since last reclaim */ |
Christoph Lameter | ca889e6 | 2006-06-30 01:55:44 -0700 | [diff] [blame] | 147 | #ifdef CONFIG_NUMA |
| 148 | NUMA_HIT, /* allocated in intended node */ |
| 149 | NUMA_MISS, /* allocated in non intended node */ |
| 150 | NUMA_FOREIGN, /* was intended here, hit elsewhere */ |
| 151 | NUMA_INTERLEAVE_HIT, /* interleaver preferred this zone */ |
| 152 | NUMA_LOCAL, /* allocation from local node */ |
| 153 | NUMA_OTHER, /* allocation from other node */ |
| 154 | #endif |
Johannes Weiner | a528910 | 2014-04-03 14:47:51 -0700 | [diff] [blame] | 155 | WORKINGSET_REFAULT, |
| 156 | WORKINGSET_ACTIVATE, |
Johannes Weiner | 449dd69 | 2014-04-03 14:47:56 -0700 | [diff] [blame] | 157 | WORKINGSET_NODERECLAIM, |
Andrea Arcangeli | 7913417 | 2011-01-13 15:46:58 -0800 | [diff] [blame] | 158 | NR_ANON_TRANSPARENT_HUGEPAGES, |
Bartlomiej Zolnierkiewicz | d1ce749 | 2012-10-08 16:32:02 -0700 | [diff] [blame] | 159 | NR_FREE_CMA_PAGES, |
Christoph Lameter | 2244b95 | 2006-06-30 01:55:33 -0700 | [diff] [blame] | 160 | NR_VM_ZONE_STAT_ITEMS }; |
| 161 | |
Rik van Riel | 4f98a2f | 2008-10-18 20:26:32 -0700 | [diff] [blame] | 162 | /* |
| 163 | * We do arithmetic on the LRU lists in various places in the code, |
| 164 | * so it is important to keep the active lists LRU_ACTIVE higher in |
| 165 | * the array than the corresponding inactive lists, and to keep |
| 166 | * the *_FILE lists LRU_FILE higher than the corresponding _ANON lists. |
| 167 | * |
| 168 | * This has to be kept in sync with the statistics in zone_stat_item |
| 169 | * above and the descriptions in vmstat_text in mm/vmstat.c |
| 170 | */ |
| 171 | #define LRU_BASE 0 |
| 172 | #define LRU_ACTIVE 1 |
| 173 | #define LRU_FILE 2 |
| 174 | |
Christoph Lameter | b69408e | 2008-10-18 20:26:14 -0700 | [diff] [blame] | 175 | enum lru_list { |
Rik van Riel | 4f98a2f | 2008-10-18 20:26:32 -0700 | [diff] [blame] | 176 | LRU_INACTIVE_ANON = LRU_BASE, |
| 177 | LRU_ACTIVE_ANON = LRU_BASE + LRU_ACTIVE, |
| 178 | LRU_INACTIVE_FILE = LRU_BASE + LRU_FILE, |
| 179 | LRU_ACTIVE_FILE = LRU_BASE + LRU_FILE + LRU_ACTIVE, |
Lee Schermerhorn | 894bc31 | 2008-10-18 20:26:39 -0700 | [diff] [blame] | 180 | LRU_UNEVICTABLE, |
Lee Schermerhorn | 894bc31 | 2008-10-18 20:26:39 -0700 | [diff] [blame] | 181 | NR_LRU_LISTS |
| 182 | }; |
Christoph Lameter | b69408e | 2008-10-18 20:26:14 -0700 | [diff] [blame] | 183 | |
Hugh Dickins | 4111304 | 2012-01-12 17:20:01 -0800 | [diff] [blame] | 184 | #define for_each_lru(lru) for (lru = 0; lru < NR_LRU_LISTS; lru++) |
Christoph Lameter | b69408e | 2008-10-18 20:26:14 -0700 | [diff] [blame] | 185 | |
Hugh Dickins | 4111304 | 2012-01-12 17:20:01 -0800 | [diff] [blame] | 186 | #define for_each_evictable_lru(lru) for (lru = 0; lru <= LRU_ACTIVE_FILE; lru++) |
Lee Schermerhorn | 894bc31 | 2008-10-18 20:26:39 -0700 | [diff] [blame] | 187 | |
Hugh Dickins | 4111304 | 2012-01-12 17:20:01 -0800 | [diff] [blame] | 188 | static inline int is_file_lru(enum lru_list lru) |
Rik van Riel | 4f98a2f | 2008-10-18 20:26:32 -0700 | [diff] [blame] | 189 | { |
Hugh Dickins | 4111304 | 2012-01-12 17:20:01 -0800 | [diff] [blame] | 190 | return (lru == LRU_INACTIVE_FILE || lru == LRU_ACTIVE_FILE); |
Rik van Riel | 4f98a2f | 2008-10-18 20:26:32 -0700 | [diff] [blame] | 191 | } |
| 192 | |
Hugh Dickins | 4111304 | 2012-01-12 17:20:01 -0800 | [diff] [blame] | 193 | static inline int is_active_lru(enum lru_list lru) |
Christoph Lameter | b69408e | 2008-10-18 20:26:14 -0700 | [diff] [blame] | 194 | { |
Hugh Dickins | 4111304 | 2012-01-12 17:20:01 -0800 | [diff] [blame] | 195 | return (lru == LRU_ACTIVE_ANON || lru == LRU_ACTIVE_FILE); |
Christoph Lameter | b69408e | 2008-10-18 20:26:14 -0700 | [diff] [blame] | 196 | } |
| 197 | |
Hugh Dickins | 4111304 | 2012-01-12 17:20:01 -0800 | [diff] [blame] | 198 | static inline int is_unevictable_lru(enum lru_list lru) |
Lee Schermerhorn | 894bc31 | 2008-10-18 20:26:39 -0700 | [diff] [blame] | 199 | { |
Hugh Dickins | 4111304 | 2012-01-12 17:20:01 -0800 | [diff] [blame] | 200 | return (lru == LRU_UNEVICTABLE); |
Lee Schermerhorn | 894bc31 | 2008-10-18 20:26:39 -0700 | [diff] [blame] | 201 | } |
| 202 | |
Hugh Dickins | 89abfab | 2012-05-29 15:06:53 -0700 | [diff] [blame] | 203 | struct zone_reclaim_stat { |
| 204 | /* |
| 205 | * The pageout code in vmscan.c keeps track of how many of the |
Jiri Kosina | 59f91e5 | 2012-06-29 14:45:58 +0200 | [diff] [blame] | 206 | * mem/swap backed and file backed pages are referenced. |
Hugh Dickins | 89abfab | 2012-05-29 15:06:53 -0700 | [diff] [blame] | 207 | * The higher the rotated/scanned ratio, the more valuable |
| 208 | * that cache is. |
| 209 | * |
| 210 | * The anon LRU stats live in [0], file LRU stats in [1] |
| 211 | */ |
| 212 | unsigned long recent_rotated[2]; |
| 213 | unsigned long recent_scanned[2]; |
| 214 | }; |
| 215 | |
Johannes Weiner | 6290df5 | 2012-01-12 17:18:10 -0800 | [diff] [blame] | 216 | struct lruvec { |
| 217 | struct list_head lists[NR_LRU_LISTS]; |
Hugh Dickins | 89abfab | 2012-05-29 15:06:53 -0700 | [diff] [blame] | 218 | struct zone_reclaim_stat reclaim_stat; |
Andrew Morton | c255a45 | 2012-07-31 16:43:02 -0700 | [diff] [blame] | 219 | #ifdef CONFIG_MEMCG |
Konstantin Khlebnikov | 7f5e86c | 2012-05-29 15:06:58 -0700 | [diff] [blame] | 220 | struct zone *zone; |
| 221 | #endif |
Johannes Weiner | 6290df5 | 2012-01-12 17:18:10 -0800 | [diff] [blame] | 222 | }; |
| 223 | |
KAMEZAWA Hiroyuki | bb2a0de | 2011-07-26 16:08:22 -0700 | [diff] [blame] | 224 | /* Mask used at gathering information at once (see memcontrol.c) */ |
| 225 | #define LRU_ALL_FILE (BIT(LRU_INACTIVE_FILE) | BIT(LRU_ACTIVE_FILE)) |
| 226 | #define LRU_ALL_ANON (BIT(LRU_INACTIVE_ANON) | BIT(LRU_ACTIVE_ANON)) |
KAMEZAWA Hiroyuki | bb2a0de | 2011-07-26 16:08:22 -0700 | [diff] [blame] | 227 | #define LRU_ALL ((1 << NR_LRU_LISTS) - 1) |
| 228 | |
Minchan Kim | 39deaf8 | 2011-10-31 17:06:51 -0700 | [diff] [blame] | 229 | /* Isolate clean file */ |
Konstantin Khlebnikov | f3fd4a6 | 2012-05-29 15:06:54 -0700 | [diff] [blame] | 230 | #define ISOLATE_CLEAN ((__force isolate_mode_t)0x1) |
Minchan Kim | f80c067 | 2011-10-31 17:06:55 -0700 | [diff] [blame] | 231 | /* Isolate unmapped file */ |
Konstantin Khlebnikov | f3fd4a6 | 2012-05-29 15:06:54 -0700 | [diff] [blame] | 232 | #define ISOLATE_UNMAPPED ((__force isolate_mode_t)0x2) |
Mel Gorman | c824493 | 2012-01-12 17:19:38 -0800 | [diff] [blame] | 233 | /* Isolate for asynchronous migration */ |
Konstantin Khlebnikov | f3fd4a6 | 2012-05-29 15:06:54 -0700 | [diff] [blame] | 234 | #define ISOLATE_ASYNC_MIGRATE ((__force isolate_mode_t)0x4) |
Minchan Kim | e46a287 | 2012-10-08 16:33:48 -0700 | [diff] [blame] | 235 | /* Isolate unevictable pages */ |
| 236 | #define ISOLATE_UNEVICTABLE ((__force isolate_mode_t)0x8) |
Minchan Kim | 4356f21 | 2011-10-31 17:06:47 -0700 | [diff] [blame] | 237 | |
| 238 | /* LRU Isolation modes. */ |
| 239 | typedef unsigned __bitwise__ isolate_mode_t; |
| 240 | |
Mel Gorman | 4185896 | 2009-06-16 15:32:12 -0700 | [diff] [blame] | 241 | enum zone_watermarks { |
| 242 | WMARK_MIN, |
| 243 | WMARK_LOW, |
| 244 | WMARK_HIGH, |
| 245 | NR_WMARK |
| 246 | }; |
| 247 | |
| 248 | #define min_wmark_pages(z) (z->watermark[WMARK_MIN]) |
| 249 | #define low_wmark_pages(z) (z->watermark[WMARK_LOW]) |
| 250 | #define high_wmark_pages(z) (z->watermark[WMARK_HIGH]) |
| 251 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 252 | struct per_cpu_pages { |
| 253 | int count; /* number of pages in the list */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 254 | int high; /* high watermark, emptying needed */ |
| 255 | int batch; /* chunk size for buddy add/remove */ |
Mel Gorman | 5f8dcc2 | 2009-09-21 17:03:19 -0700 | [diff] [blame] | 256 | |
| 257 | /* Lists of pages, one per migrate type stored on the pcp-lists */ |
| 258 | struct list_head lists[MIGRATE_PCPTYPES]; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 259 | }; |
| 260 | |
| 261 | struct per_cpu_pageset { |
Christoph Lameter | 3dfa572 | 2008-02-04 22:29:19 -0800 | [diff] [blame] | 262 | struct per_cpu_pages pcp; |
Christoph Lameter | 4037d45 | 2007-05-09 02:35:14 -0700 | [diff] [blame] | 263 | #ifdef CONFIG_NUMA |
| 264 | s8 expire; |
| 265 | #endif |
Christoph Lameter | 2244b95 | 2006-06-30 01:55:33 -0700 | [diff] [blame] | 266 | #ifdef CONFIG_SMP |
Christoph Lameter | df9ecab | 2006-08-31 21:27:35 -0700 | [diff] [blame] | 267 | s8 stat_threshold; |
Christoph Lameter | 2244b95 | 2006-06-30 01:55:33 -0700 | [diff] [blame] | 268 | s8 vm_stat_diff[NR_VM_ZONE_STAT_ITEMS]; |
| 269 | #endif |
Christoph Lameter | 99dcc3e | 2010-01-05 15:34:51 +0900 | [diff] [blame] | 270 | }; |
Christoph Lameter | e7c8d5c | 2005-06-21 17:14:47 -0700 | [diff] [blame] | 271 | |
Christoph Lameter | 9796547 | 2008-04-28 02:12:54 -0700 | [diff] [blame] | 272 | #endif /* !__GENERATING_BOUNDS.H */ |
| 273 | |
Christoph Lameter | 2f1b624 | 2006-09-25 23:31:13 -0700 | [diff] [blame] | 274 | enum zone_type { |
Christoph Lameter | 4b51d66 | 2007-02-10 01:43:10 -0800 | [diff] [blame] | 275 | #ifdef CONFIG_ZONE_DMA |
Christoph Lameter | 2f1b624 | 2006-09-25 23:31:13 -0700 | [diff] [blame] | 276 | /* |
| 277 | * ZONE_DMA is used when there are devices that are not able |
| 278 | * to do DMA to all of addressable memory (ZONE_NORMAL). Then we |
| 279 | * carve out the portion of memory that is needed for these devices. |
| 280 | * The range is arch specific. |
| 281 | * |
| 282 | * Some examples |
| 283 | * |
| 284 | * Architecture Limit |
| 285 | * --------------------------- |
| 286 | * parisc, ia64, sparc <4G |
| 287 | * s390 <2G |
Christoph Lameter | 2f1b624 | 2006-09-25 23:31:13 -0700 | [diff] [blame] | 288 | * arm Various |
| 289 | * alpha Unlimited or 0-16MB. |
| 290 | * |
| 291 | * i386, x86_64 and multiple other arches |
| 292 | * <16M. |
| 293 | */ |
| 294 | ZONE_DMA, |
Christoph Lameter | 4b51d66 | 2007-02-10 01:43:10 -0800 | [diff] [blame] | 295 | #endif |
Christoph Lameter | fb0e794 | 2006-09-25 23:31:13 -0700 | [diff] [blame] | 296 | #ifdef CONFIG_ZONE_DMA32 |
Christoph Lameter | 2f1b624 | 2006-09-25 23:31:13 -0700 | [diff] [blame] | 297 | /* |
| 298 | * x86_64 needs two ZONE_DMAs because it supports devices that are |
| 299 | * only able to do DMA to the lower 16M but also 32 bit devices that |
| 300 | * can only do DMA areas below 4G. |
| 301 | */ |
| 302 | ZONE_DMA32, |
Christoph Lameter | fb0e794 | 2006-09-25 23:31:13 -0700 | [diff] [blame] | 303 | #endif |
Christoph Lameter | 2f1b624 | 2006-09-25 23:31:13 -0700 | [diff] [blame] | 304 | /* |
| 305 | * Normal addressable memory is in ZONE_NORMAL. DMA operations can be |
| 306 | * performed on pages in ZONE_NORMAL if the DMA devices support |
| 307 | * transfers to all addressable memory. |
| 308 | */ |
| 309 | ZONE_NORMAL, |
Christoph Lameter | e53ef38 | 2006-09-25 23:31:14 -0700 | [diff] [blame] | 310 | #ifdef CONFIG_HIGHMEM |
Christoph Lameter | 2f1b624 | 2006-09-25 23:31:13 -0700 | [diff] [blame] | 311 | /* |
| 312 | * A memory area that is only addressable by the kernel through |
| 313 | * mapping portions into its own address space. This is for example |
| 314 | * used by i386 to allow the kernel to address the memory beyond |
| 315 | * 900MB. The kernel will set up special mappings (page |
| 316 | * table entries on i386) for each page that the kernel needs to |
| 317 | * access. |
| 318 | */ |
| 319 | ZONE_HIGHMEM, |
Christoph Lameter | e53ef38 | 2006-09-25 23:31:14 -0700 | [diff] [blame] | 320 | #endif |
Mel Gorman | 2a1e274 | 2007-07-17 04:03:12 -0700 | [diff] [blame] | 321 | ZONE_MOVABLE, |
Christoph Lameter | 9796547 | 2008-04-28 02:12:54 -0700 | [diff] [blame] | 322 | __MAX_NR_ZONES |
Christoph Lameter | 2f1b624 | 2006-09-25 23:31:13 -0700 | [diff] [blame] | 323 | }; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 324 | |
Christoph Lameter | 9796547 | 2008-04-28 02:12:54 -0700 | [diff] [blame] | 325 | #ifndef __GENERATING_BOUNDS_H |
| 326 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 327 | struct zone { |
Mel Gorman | 3484b2d | 2014-08-06 16:07:14 -0700 | [diff] [blame] | 328 | /* Read-mostly fields */ |
Mel Gorman | 4185896 | 2009-06-16 15:32:12 -0700 | [diff] [blame] | 329 | |
| 330 | /* zone watermarks, access with *_wmark_pages(zone) macros */ |
| 331 | unsigned long watermark[NR_WMARK]; |
| 332 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 333 | /* |
| 334 | * We don't know if the memory that we're going to allocate will be freeable |
| 335 | * or/and it will be released eventually, so to avoid totally wasting several |
| 336 | * GB of ram we must reserve some of the lower zone memory (otherwise we risk |
| 337 | * to run OOM on the lower zones despite there's tons of freeable ram |
| 338 | * on the higher zones). This array is recalculated at runtime if the |
| 339 | * sysctl_lowmem_reserve_ratio sysctl changes. |
| 340 | */ |
Mel Gorman | 3484b2d | 2014-08-06 16:07:14 -0700 | [diff] [blame] | 341 | long lowmem_reserve[MAX_NR_ZONES]; |
| 342 | |
| 343 | #ifdef CONFIG_NUMA |
| 344 | int node; |
| 345 | #endif |
| 346 | |
| 347 | /* |
| 348 | * The target ratio of ACTIVE_ANON to INACTIVE_ANON pages on |
| 349 | * this zone's LRU. Maintained by the pageout code. |
| 350 | */ |
| 351 | unsigned int inactive_ratio; |
| 352 | |
| 353 | struct pglist_data *zone_pgdat; |
| 354 | struct per_cpu_pageset __percpu *pageset; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 355 | |
Johannes Weiner | ab8fabd | 2012-01-10 15:07:42 -0800 | [diff] [blame] | 356 | /* |
| 357 | * This is a per-zone reserve of pages that should not be |
| 358 | * considered dirtyable memory. |
| 359 | */ |
| 360 | unsigned long dirty_balance_reserve; |
| 361 | |
Mel Gorman | 835c134 | 2007-10-16 01:25:47 -0700 | [diff] [blame] | 362 | #ifndef CONFIG_SPARSEMEM |
| 363 | /* |
Mel Gorman | d9c2340 | 2007-10-16 01:26:01 -0700 | [diff] [blame] | 364 | * Flags for a pageblock_nr_pages block. See pageblock-flags.h. |
Mel Gorman | 835c134 | 2007-10-16 01:25:47 -0700 | [diff] [blame] | 365 | * In SPARSEMEM, this map is stored in struct mem_section |
| 366 | */ |
| 367 | unsigned long *pageblock_flags; |
| 368 | #endif /* CONFIG_SPARSEMEM */ |
| 369 | |
Mel Gorman | 3484b2d | 2014-08-06 16:07:14 -0700 | [diff] [blame] | 370 | #ifdef CONFIG_NUMA |
Mel Gorman | 4f92e25 | 2010-05-24 14:32:32 -0700 | [diff] [blame] | 371 | /* |
Mel Gorman | 3484b2d | 2014-08-06 16:07:14 -0700 | [diff] [blame] | 372 | * zone reclaim becomes active if more unmapped pages exist. |
Mel Gorman | 4f92e25 | 2010-05-24 14:32:32 -0700 | [diff] [blame] | 373 | */ |
Mel Gorman | 3484b2d | 2014-08-06 16:07:14 -0700 | [diff] [blame] | 374 | unsigned long min_unmapped_pages; |
| 375 | unsigned long min_slab_pages; |
| 376 | #endif /* CONFIG_NUMA */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 377 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 378 | /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */ |
| 379 | unsigned long zone_start_pfn; |
| 380 | |
Dave Hansen | bdc8cb9 | 2005-10-29 18:16:53 -0700 | [diff] [blame] | 381 | /* |
Jiang Liu | 9feedc9 | 2012-12-12 13:52:12 -0800 | [diff] [blame] | 382 | * spanned_pages is the total pages spanned by the zone, including |
| 383 | * holes, which is calculated as: |
| 384 | * spanned_pages = zone_end_pfn - zone_start_pfn; |
Dave Hansen | bdc8cb9 | 2005-10-29 18:16:53 -0700 | [diff] [blame] | 385 | * |
Jiang Liu | 9feedc9 | 2012-12-12 13:52:12 -0800 | [diff] [blame] | 386 | * present_pages is physical pages existing within the zone, which |
| 387 | * is calculated as: |
Cody P Schafer | 8761e31 | 2013-03-26 10:30:44 -0700 | [diff] [blame] | 388 | * present_pages = spanned_pages - absent_pages(pages in holes); |
Jiang Liu | 9feedc9 | 2012-12-12 13:52:12 -0800 | [diff] [blame] | 389 | * |
| 390 | * managed_pages is present pages managed by the buddy system, which |
| 391 | * is calculated as (reserved_pages includes pages allocated by the |
| 392 | * bootmem allocator): |
| 393 | * managed_pages = present_pages - reserved_pages; |
| 394 | * |
| 395 | * So present_pages may be used by memory hotplug or memory power |
| 396 | * management logic to figure out unmanaged pages by checking |
| 397 | * (present_pages - managed_pages). And managed_pages should be used |
| 398 | * by page allocator and vm scanner to calculate all kinds of watermarks |
| 399 | * and thresholds. |
| 400 | * |
| 401 | * Locking rules: |
| 402 | * |
| 403 | * zone_start_pfn and spanned_pages are protected by span_seqlock. |
| 404 | * It is a seqlock because it has to be read outside of zone->lock, |
| 405 | * and it is done in the main allocator path. But, it is written |
| 406 | * quite infrequently. |
| 407 | * |
| 408 | * The span_seq lock is declared along with zone->lock because it is |
Dave Hansen | bdc8cb9 | 2005-10-29 18:16:53 -0700 | [diff] [blame] | 409 | * frequently read in proximity to zone->lock. It's good to |
| 410 | * give them a chance of being in the same cacheline. |
Jiang Liu | 9feedc9 | 2012-12-12 13:52:12 -0800 | [diff] [blame] | 411 | * |
Jiang Liu | c3d5f5f | 2013-07-03 15:03:14 -0700 | [diff] [blame] | 412 | * Write access to present_pages at runtime should be protected by |
Vladimir Davydov | bfc8c90 | 2014-06-04 16:07:18 -0700 | [diff] [blame] | 413 | * mem_hotplug_begin/end(). Any reader who can't tolerant drift of |
| 414 | * present_pages should get_online_mems() to get a stable value. |
Jiang Liu | c3d5f5f | 2013-07-03 15:03:14 -0700 | [diff] [blame] | 415 | * |
| 416 | * Read access to managed_pages should be safe because it's unsigned |
| 417 | * long. Write access to zone->managed_pages and totalram_pages are |
| 418 | * protected by managed_page_count_lock at runtime. Idealy only |
| 419 | * adjust_managed_page_count() should be used instead of directly |
| 420 | * touching zone->managed_pages and totalram_pages. |
Dave Hansen | bdc8cb9 | 2005-10-29 18:16:53 -0700 | [diff] [blame] | 421 | */ |
Mel Gorman | 3484b2d | 2014-08-06 16:07:14 -0700 | [diff] [blame] | 422 | unsigned long managed_pages; |
Jiang Liu | 9feedc9 | 2012-12-12 13:52:12 -0800 | [diff] [blame] | 423 | unsigned long spanned_pages; |
| 424 | unsigned long present_pages; |
Mel Gorman | 3484b2d | 2014-08-06 16:07:14 -0700 | [diff] [blame] | 425 | |
| 426 | const char *name; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 427 | |
| 428 | /* |
Yasuaki Ishimatsu | 943dca1 | 2014-01-21 15:49:06 -0800 | [diff] [blame] | 429 | * Number of MIGRATE_RESEVE page block. To maintain for just |
| 430 | * optimization. Protected by zone->lock. |
| 431 | */ |
| 432 | int nr_migrate_reserve_block; |
| 433 | |
Joonsoo Kim | ad53f92 | 2014-11-13 15:19:11 -0800 | [diff] [blame] | 434 | #ifdef CONFIG_MEMORY_ISOLATION |
| 435 | /* |
| 436 | * Number of isolated pageblock. It is used to solve incorrect |
| 437 | * freepage counting problem due to racy retrieving migratetype |
| 438 | * of pageblock. Protected by zone->lock. |
| 439 | */ |
| 440 | unsigned long nr_isolate_pageblock; |
| 441 | #endif |
| 442 | |
Mel Gorman | 3484b2d | 2014-08-06 16:07:14 -0700 | [diff] [blame] | 443 | #ifdef CONFIG_MEMORY_HOTPLUG |
| 444 | /* see spanned/present_pages for more description */ |
| 445 | seqlock_t span_seqlock; |
| 446 | #endif |
| 447 | |
Yasuaki Ishimatsu | 943dca1 | 2014-01-21 15:49:06 -0800 | [diff] [blame] | 448 | /* |
Mel Gorman | 3484b2d | 2014-08-06 16:07:14 -0700 | [diff] [blame] | 449 | * wait_table -- the array holding the hash table |
| 450 | * wait_table_hash_nr_entries -- the size of the hash table array |
| 451 | * wait_table_bits -- wait_table_size == (1 << wait_table_bits) |
| 452 | * |
| 453 | * The purpose of all these is to keep track of the people |
| 454 | * waiting for a page to become available and make them |
| 455 | * runnable again when possible. The trouble is that this |
| 456 | * consumes a lot of space, especially when so few things |
| 457 | * wait on pages at a given time. So instead of using |
| 458 | * per-page waitqueues, we use a waitqueue hash table. |
| 459 | * |
| 460 | * The bucket discipline is to sleep on the same queue when |
| 461 | * colliding and wake all in that wait queue when removing. |
| 462 | * When something wakes, it must check to be sure its page is |
| 463 | * truly available, a la thundering herd. The cost of a |
| 464 | * collision is great, but given the expected load of the |
| 465 | * table, they should be so rare as to be outweighed by the |
| 466 | * benefits from the saved space. |
| 467 | * |
| 468 | * __wait_on_page_locked() and unlock_page() in mm/filemap.c, are the |
| 469 | * primary users of these fields, and in mm/page_alloc.c |
| 470 | * free_area_init_core() performs the initialization of them. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 471 | */ |
Mel Gorman | 3484b2d | 2014-08-06 16:07:14 -0700 | [diff] [blame] | 472 | wait_queue_head_t *wait_table; |
| 473 | unsigned long wait_table_hash_nr_entries; |
| 474 | unsigned long wait_table_bits; |
| 475 | |
| 476 | ZONE_PADDING(_pad1_) |
| 477 | |
| 478 | /* Write-intensive fields used from the page allocator */ |
| 479 | spinlock_t lock; |
| 480 | |
| 481 | /* free areas of different sizes */ |
| 482 | struct free_area free_area[MAX_ORDER]; |
| 483 | |
| 484 | /* zone flags, see below */ |
| 485 | unsigned long flags; |
| 486 | |
| 487 | ZONE_PADDING(_pad2_) |
| 488 | |
| 489 | /* Write-intensive fields used by page reclaim */ |
| 490 | |
| 491 | /* Fields commonly accessed by the page reclaim scanner */ |
| 492 | spinlock_t lru_lock; |
Mel Gorman | 3484b2d | 2014-08-06 16:07:14 -0700 | [diff] [blame] | 493 | struct lruvec lruvec; |
| 494 | |
| 495 | /* Evictions & activations on the inactive file list */ |
| 496 | atomic_long_t inactive_age; |
| 497 | |
| 498 | /* |
| 499 | * When free pages are below this point, additional steps are taken |
| 500 | * when reading the number of free pages to avoid per-cpu counter |
| 501 | * drift allowing watermarks to be breached |
| 502 | */ |
| 503 | unsigned long percpu_drift_mark; |
| 504 | |
| 505 | #if defined CONFIG_COMPACTION || defined CONFIG_CMA |
| 506 | /* pfn where compaction free scanner should start */ |
| 507 | unsigned long compact_cached_free_pfn; |
| 508 | /* pfn where async and sync compaction migration scanner should start */ |
| 509 | unsigned long compact_cached_migrate_pfn[2]; |
| 510 | #endif |
| 511 | |
| 512 | #ifdef CONFIG_COMPACTION |
| 513 | /* |
| 514 | * On compaction failure, 1<<compact_defer_shift compactions |
| 515 | * are skipped before trying again. The number attempted since |
| 516 | * last failure is tracked with compact_considered. |
| 517 | */ |
| 518 | unsigned int compact_considered; |
| 519 | unsigned int compact_defer_shift; |
| 520 | int compact_order_failed; |
| 521 | #endif |
| 522 | |
| 523 | #if defined CONFIG_COMPACTION || defined CONFIG_CMA |
| 524 | /* Set to true when the PG_migrate_skip bits should be cleared */ |
| 525 | bool compact_blockskip_flush; |
| 526 | #endif |
| 527 | |
| 528 | ZONE_PADDING(_pad3_) |
| 529 | /* Zone statistics */ |
| 530 | atomic_long_t vm_stat[NR_VM_ZONE_STAT_ITEMS]; |
Ravikiran G Thirumalai | 22fc6ec | 2006-01-08 01:01:27 -0800 | [diff] [blame] | 531 | } ____cacheline_internodealigned_in_smp; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 532 | |
Johannes Weiner | 5705465 | 2014-10-09 15:28:17 -0700 | [diff] [blame] | 533 | enum zone_flags { |
David Rientjes | e815af9 | 2007-10-16 23:25:54 -0700 | [diff] [blame] | 534 | ZONE_RECLAIM_LOCKED, /* prevents concurrent reclaim */ |
David Rientjes | 098d7f1 | 2007-10-16 23:25:55 -0700 | [diff] [blame] | 535 | ZONE_OOM_LOCKED, /* zone is in OOM killer zonelist */ |
Mel Gorman | 0e093d99 | 2010-10-26 14:21:45 -0700 | [diff] [blame] | 536 | ZONE_CONGESTED, /* zone has many dirty pages backed by |
| 537 | * a congested BDI |
| 538 | */ |
Johannes Weiner | 5705465 | 2014-10-09 15:28:17 -0700 | [diff] [blame] | 539 | ZONE_DIRTY, /* reclaim scanning has recently found |
Mel Gorman | d43006d | 2013-07-03 15:01:50 -0700 | [diff] [blame] | 540 | * many dirty file pages at the tail |
| 541 | * of the LRU. |
| 542 | */ |
Mel Gorman | 283aba9 | 2013-07-03 15:01:51 -0700 | [diff] [blame] | 543 | ZONE_WRITEBACK, /* reclaim scanning has recently found |
| 544 | * many pages under writeback |
| 545 | */ |
Mel Gorman | 4ffeaf3 | 2014-08-06 16:07:22 -0700 | [diff] [blame] | 546 | ZONE_FAIR_DEPLETED, /* fair zone policy batch depleted */ |
Johannes Weiner | 5705465 | 2014-10-09 15:28:17 -0700 | [diff] [blame] | 547 | }; |
David Rientjes | e815af9 | 2007-10-16 23:25:54 -0700 | [diff] [blame] | 548 | |
Russ Anderson | f9228b2 | 2013-03-22 15:04:43 -0700 | [diff] [blame] | 549 | static inline unsigned long zone_end_pfn(const struct zone *zone) |
Cody P Schafer | 108bcc9 | 2013-02-22 16:35:23 -0800 | [diff] [blame] | 550 | { |
| 551 | return zone->zone_start_pfn + zone->spanned_pages; |
| 552 | } |
| 553 | |
| 554 | static inline bool zone_spans_pfn(const struct zone *zone, unsigned long pfn) |
| 555 | { |
| 556 | return zone->zone_start_pfn <= pfn && pfn < zone_end_pfn(zone); |
| 557 | } |
| 558 | |
Cody P Schafer | 2a6e3eb | 2013-02-22 16:35:24 -0800 | [diff] [blame] | 559 | static inline bool zone_is_initialized(struct zone *zone) |
| 560 | { |
| 561 | return !!zone->wait_table; |
| 562 | } |
| 563 | |
| 564 | static inline bool zone_is_empty(struct zone *zone) |
| 565 | { |
| 566 | return zone->spanned_pages == 0; |
| 567 | } |
| 568 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 569 | /* |
| 570 | * The "priority" of VM scanning is how much of the queues we will scan in one |
| 571 | * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the |
| 572 | * queues ("queue_length >> 12") during an aging round. |
| 573 | */ |
| 574 | #define DEF_PRIORITY 12 |
| 575 | |
Paul Jackson | 9276b1bc | 2006-12-06 20:31:48 -0800 | [diff] [blame] | 576 | /* Maximum number of zones on a zonelist */ |
| 577 | #define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES) |
| 578 | |
| 579 | #ifdef CONFIG_NUMA |
Christoph Lameter | 523b945 | 2007-10-16 01:25:37 -0700 | [diff] [blame] | 580 | |
| 581 | /* |
Pete Zaitcev | 25a64ec | 2011-02-03 22:43:48 -0700 | [diff] [blame] | 582 | * The NUMA zonelists are doubled because we need zonelists that restrict the |
Johannes Weiner | e97ca8e | 2014-03-10 15:49:43 -0700 | [diff] [blame] | 583 | * allocations to a single node for __GFP_THISNODE. |
Christoph Lameter | 523b945 | 2007-10-16 01:25:37 -0700 | [diff] [blame] | 584 | * |
Mel Gorman | 54a6eb5 | 2008-04-28 02:12:16 -0700 | [diff] [blame] | 585 | * [0] : Zonelist with fallback |
Johannes Weiner | e97ca8e | 2014-03-10 15:49:43 -0700 | [diff] [blame] | 586 | * [1] : No fallback (__GFP_THISNODE) |
Christoph Lameter | 523b945 | 2007-10-16 01:25:37 -0700 | [diff] [blame] | 587 | */ |
Mel Gorman | 54a6eb5 | 2008-04-28 02:12:16 -0700 | [diff] [blame] | 588 | #define MAX_ZONELISTS 2 |
Christoph Lameter | 523b945 | 2007-10-16 01:25:37 -0700 | [diff] [blame] | 589 | |
| 590 | |
Paul Jackson | 9276b1bc | 2006-12-06 20:31:48 -0800 | [diff] [blame] | 591 | /* |
| 592 | * We cache key information from each zonelist for smaller cache |
| 593 | * footprint when scanning for free pages in get_page_from_freelist(). |
| 594 | * |
| 595 | * 1) The BITMAP fullzones tracks which zones in a zonelist have come |
| 596 | * up short of free memory since the last time (last_fullzone_zap) |
| 597 | * we zero'd fullzones. |
| 598 | * 2) The array z_to_n[] maps each zone in the zonelist to its node |
| 599 | * id, so that we can efficiently evaluate whether that node is |
| 600 | * set in the current tasks mems_allowed. |
| 601 | * |
| 602 | * Both fullzones and z_to_n[] are one-to-one with the zonelist, |
| 603 | * indexed by a zones offset in the zonelist zones[] array. |
| 604 | * |
| 605 | * The get_page_from_freelist() routine does two scans. During the |
| 606 | * first scan, we skip zones whose corresponding bit in 'fullzones' |
| 607 | * is set or whose corresponding node in current->mems_allowed (which |
| 608 | * comes from cpusets) is not set. During the second scan, we bypass |
| 609 | * this zonelist_cache, to ensure we look methodically at each zone. |
| 610 | * |
| 611 | * Once per second, we zero out (zap) fullzones, forcing us to |
| 612 | * reconsider nodes that might have regained more free memory. |
| 613 | * The field last_full_zap is the time we last zapped fullzones. |
| 614 | * |
| 615 | * This mechanism reduces the amount of time we waste repeatedly |
| 616 | * reexaming zones for free memory when they just came up low on |
| 617 | * memory momentarilly ago. |
| 618 | * |
| 619 | * The zonelist_cache struct members logically belong in struct |
| 620 | * zonelist. However, the mempolicy zonelists constructed for |
| 621 | * MPOL_BIND are intentionally variable length (and usually much |
| 622 | * shorter). A general purpose mechanism for handling structs with |
| 623 | * multiple variable length members is more mechanism than we want |
| 624 | * here. We resort to some special case hackery instead. |
| 625 | * |
| 626 | * The MPOL_BIND zonelists don't need this zonelist_cache (in good |
| 627 | * part because they are shorter), so we put the fixed length stuff |
| 628 | * at the front of the zonelist struct, ending in a variable length |
| 629 | * zones[], as is needed by MPOL_BIND. |
| 630 | * |
| 631 | * Then we put the optional zonelist cache on the end of the zonelist |
| 632 | * struct. This optional stuff is found by a 'zlcache_ptr' pointer in |
| 633 | * the fixed length portion at the front of the struct. This pointer |
| 634 | * both enables us to find the zonelist cache, and in the case of |
| 635 | * MPOL_BIND zonelists, (which will just set the zlcache_ptr to NULL) |
| 636 | * to know that the zonelist cache is not there. |
| 637 | * |
| 638 | * The end result is that struct zonelists come in two flavors: |
| 639 | * 1) The full, fixed length version, shown below, and |
| 640 | * 2) The custom zonelists for MPOL_BIND. |
| 641 | * The custom MPOL_BIND zonelists have a NULL zlcache_ptr and no zlcache. |
| 642 | * |
| 643 | * Even though there may be multiple CPU cores on a node modifying |
| 644 | * fullzones or last_full_zap in the same zonelist_cache at the same |
| 645 | * time, we don't lock it. This is just hint data - if it is wrong now |
| 646 | * and then, the allocator will still function, perhaps a bit slower. |
| 647 | */ |
| 648 | |
| 649 | |
| 650 | struct zonelist_cache { |
Paul Jackson | 9276b1bc | 2006-12-06 20:31:48 -0800 | [diff] [blame] | 651 | unsigned short z_to_n[MAX_ZONES_PER_ZONELIST]; /* zone->nid */ |
Paul Jackson | 7253f4e | 2006-12-06 20:31:49 -0800 | [diff] [blame] | 652 | DECLARE_BITMAP(fullzones, MAX_ZONES_PER_ZONELIST); /* zone full? */ |
Paul Jackson | 9276b1bc | 2006-12-06 20:31:48 -0800 | [diff] [blame] | 653 | unsigned long last_full_zap; /* when last zap'd (jiffies) */ |
| 654 | }; |
| 655 | #else |
Mel Gorman | 54a6eb5 | 2008-04-28 02:12:16 -0700 | [diff] [blame] | 656 | #define MAX_ZONELISTS 1 |
Paul Jackson | 9276b1bc | 2006-12-06 20:31:48 -0800 | [diff] [blame] | 657 | struct zonelist_cache; |
| 658 | #endif |
| 659 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 660 | /* |
Mel Gorman | dd1a239 | 2008-04-28 02:12:17 -0700 | [diff] [blame] | 661 | * This struct contains information about a zone in a zonelist. It is stored |
| 662 | * here to avoid dereferences into large structures and lookups of tables |
| 663 | */ |
| 664 | struct zoneref { |
| 665 | struct zone *zone; /* Pointer to actual zone */ |
| 666 | int zone_idx; /* zone_idx(zoneref->zone) */ |
| 667 | }; |
| 668 | |
| 669 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 670 | * One allocation request operates on a zonelist. A zonelist |
| 671 | * is a list of zones, the first one is the 'goal' of the |
| 672 | * allocation, the other zones are fallback zones, in decreasing |
| 673 | * priority. |
| 674 | * |
Paul Jackson | 9276b1bc | 2006-12-06 20:31:48 -0800 | [diff] [blame] | 675 | * If zlcache_ptr is not NULL, then it is just the address of zlcache, |
| 676 | * as explained above. If zlcache_ptr is NULL, there is no zlcache. |
Mel Gorman | dd1a239 | 2008-04-28 02:12:17 -0700 | [diff] [blame] | 677 | * * |
| 678 | * To speed the reading of the zonelist, the zonerefs contain the zone index |
| 679 | * of the entry being read. Helper functions to access information given |
| 680 | * a struct zoneref are |
| 681 | * |
| 682 | * zonelist_zone() - Return the struct zone * for an entry in _zonerefs |
| 683 | * zonelist_zone_idx() - Return the index of the zone for an entry |
| 684 | * zonelist_node_idx() - Return the index of the node for an entry |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 685 | */ |
| 686 | struct zonelist { |
Paul Jackson | 9276b1bc | 2006-12-06 20:31:48 -0800 | [diff] [blame] | 687 | struct zonelist_cache *zlcache_ptr; // NULL or &zlcache |
Mel Gorman | dd1a239 | 2008-04-28 02:12:17 -0700 | [diff] [blame] | 688 | struct zoneref _zonerefs[MAX_ZONES_PER_ZONELIST + 1]; |
Paul Jackson | 9276b1bc | 2006-12-06 20:31:48 -0800 | [diff] [blame] | 689 | #ifdef CONFIG_NUMA |
| 690 | struct zonelist_cache zlcache; // optional ... |
| 691 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 692 | }; |
| 693 | |
Tejun Heo | 0ee332c | 2011-12-08 10:22:09 -0800 | [diff] [blame] | 694 | #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP |
Mel Gorman | c713216 | 2006-09-27 01:49:43 -0700 | [diff] [blame] | 695 | struct node_active_region { |
| 696 | unsigned long start_pfn; |
| 697 | unsigned long end_pfn; |
| 698 | int nid; |
| 699 | }; |
Tejun Heo | 0ee332c | 2011-12-08 10:22:09 -0800 | [diff] [blame] | 700 | #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 701 | |
Heiko Carstens | 5b99cd0 | 2006-09-27 01:50:01 -0700 | [diff] [blame] | 702 | #ifndef CONFIG_DISCONTIGMEM |
| 703 | /* The array of struct pages - for discontigmem use pgdat->lmem_map */ |
| 704 | extern struct page *mem_map; |
| 705 | #endif |
| 706 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 707 | /* |
| 708 | * The pg_data_t structure is used in machines with CONFIG_DISCONTIGMEM |
| 709 | * (mostly NUMA machines?) to denote a higher-level memory zone than the |
| 710 | * zone denotes. |
| 711 | * |
| 712 | * On NUMA machines, each NUMA node would have a pg_data_t to describe |
| 713 | * it's memory layout. |
| 714 | * |
| 715 | * Memory statistics and page replacement data structures are maintained on a |
| 716 | * per-zone basis. |
| 717 | */ |
| 718 | struct bootmem_data; |
| 719 | typedef struct pglist_data { |
| 720 | struct zone node_zones[MAX_NR_ZONES]; |
Christoph Lameter | 523b945 | 2007-10-16 01:25:37 -0700 | [diff] [blame] | 721 | struct zonelist node_zonelists[MAX_ZONELISTS]; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 722 | int nr_zones; |
KAMEZAWA Hiroyuki | 52d4b9a | 2008-10-18 20:28:16 -0700 | [diff] [blame] | 723 | #ifdef CONFIG_FLAT_NODE_MEM_MAP /* means !SPARSEMEM */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 724 | struct page *node_mem_map; |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 725 | #endif |
Yinghai Lu | 0867721 | 2010-02-10 01:20:20 -0800 | [diff] [blame] | 726 | #ifndef CONFIG_NO_BOOTMEM |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 727 | struct bootmem_data *bdata; |
Yinghai Lu | 0867721 | 2010-02-10 01:20:20 -0800 | [diff] [blame] | 728 | #endif |
Dave Hansen | 208d54e | 2005-10-29 18:16:52 -0700 | [diff] [blame] | 729 | #ifdef CONFIG_MEMORY_HOTPLUG |
| 730 | /* |
| 731 | * Must be held any time you expect node_start_pfn, node_present_pages |
| 732 | * or node_spanned_pages stay constant. Holding this will also |
| 733 | * guarantee that any pfn_valid() stays that way. |
| 734 | * |
Cody P Schafer | 114d4b7 | 2013-07-03 15:02:09 -0700 | [diff] [blame] | 735 | * pgdat_resize_lock() and pgdat_resize_unlock() are provided to |
| 736 | * manipulate node_size_lock without checking for CONFIG_MEMORY_HOTPLUG. |
| 737 | * |
Cody P Schafer | 72c3b51 | 2013-07-03 15:02:08 -0700 | [diff] [blame] | 738 | * Nests above zone->lock and zone->span_seqlock |
Dave Hansen | 208d54e | 2005-10-29 18:16:52 -0700 | [diff] [blame] | 739 | */ |
| 740 | spinlock_t node_size_lock; |
| 741 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 742 | unsigned long node_start_pfn; |
| 743 | unsigned long node_present_pages; /* total number of physical pages */ |
| 744 | unsigned long node_spanned_pages; /* total size of physical page |
| 745 | range, including holes */ |
| 746 | int node_id; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 747 | wait_queue_head_t kswapd_wait; |
Mel Gorman | 5515061 | 2012-07-31 16:44:35 -0700 | [diff] [blame] | 748 | wait_queue_head_t pfmemalloc_wait; |
Vladimir Davydov | bfc8c90 | 2014-06-04 16:07:18 -0700 | [diff] [blame] | 749 | struct task_struct *kswapd; /* Protected by |
| 750 | mem_hotplug_begin/end() */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 751 | int kswapd_max_order; |
Mel Gorman | 9950474 | 2011-01-13 15:46:20 -0800 | [diff] [blame] | 752 | enum zone_type classzone_idx; |
Andrea Arcangeli | 8177a42 | 2012-03-23 20:56:34 +0100 | [diff] [blame] | 753 | #ifdef CONFIG_NUMA_BALANCING |
Mel Gorman | 1c5e9c2 | 2014-01-21 15:50:59 -0800 | [diff] [blame] | 754 | /* Lock serializing the migrate rate limiting window */ |
Andrea Arcangeli | 8177a42 | 2012-03-23 20:56:34 +0100 | [diff] [blame] | 755 | spinlock_t numabalancing_migrate_lock; |
| 756 | |
| 757 | /* Rate limiting time interval */ |
| 758 | unsigned long numabalancing_migrate_next_window; |
| 759 | |
| 760 | /* Number of pages migrated during the rate limiting time interval */ |
| 761 | unsigned long numabalancing_migrate_nr_pages; |
| 762 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 763 | } pg_data_t; |
| 764 | |
| 765 | #define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages) |
| 766 | #define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages) |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 767 | #ifdef CONFIG_FLAT_NODE_MEM_MAP |
Dave Hansen | 408fde8 | 2005-06-23 00:07:37 -0700 | [diff] [blame] | 768 | #define pgdat_page_nr(pgdat, pagenr) ((pgdat)->node_mem_map + (pagenr)) |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 769 | #else |
| 770 | #define pgdat_page_nr(pgdat, pagenr) pfn_to_page((pgdat)->node_start_pfn + (pagenr)) |
| 771 | #endif |
Dave Hansen | 408fde8 | 2005-06-23 00:07:37 -0700 | [diff] [blame] | 772 | #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] | 773 | |
KAMEZAWA Hiroyuki | c6830c2 | 2011-06-16 17:28:07 +0900 | [diff] [blame] | 774 | #define node_start_pfn(nid) (NODE_DATA(nid)->node_start_pfn) |
Cody P Schafer | da3649e | 2013-02-22 16:35:27 -0800 | [diff] [blame] | 775 | #define node_end_pfn(nid) pgdat_end_pfn(NODE_DATA(nid)) |
KAMEZAWA Hiroyuki | c6830c2 | 2011-06-16 17:28:07 +0900 | [diff] [blame] | 776 | |
Cody P Schafer | da3649e | 2013-02-22 16:35:27 -0800 | [diff] [blame] | 777 | static inline unsigned long pgdat_end_pfn(pg_data_t *pgdat) |
| 778 | { |
| 779 | return pgdat->node_start_pfn + pgdat->node_spanned_pages; |
| 780 | } |
| 781 | |
| 782 | static inline bool pgdat_is_empty(pg_data_t *pgdat) |
| 783 | { |
| 784 | return !pgdat->node_start_pfn && !pgdat->node_spanned_pages; |
| 785 | } |
KAMEZAWA Hiroyuki | c6830c2 | 2011-06-16 17:28:07 +0900 | [diff] [blame] | 786 | |
Dave Hansen | 208d54e | 2005-10-29 18:16:52 -0700 | [diff] [blame] | 787 | #include <linux/memory_hotplug.h> |
| 788 | |
Haicheng Li | 4eaf3f6 | 2010-05-24 14:32:52 -0700 | [diff] [blame] | 789 | extern struct mutex zonelists_mutex; |
Jiang Liu | 9adb62a | 2012-07-31 16:43:28 -0700 | [diff] [blame] | 790 | void build_all_zonelists(pg_data_t *pgdat, struct zone *zone); |
Mel Gorman | 9950474 | 2011-01-13 15:46:20 -0800 | [diff] [blame] | 791 | void wakeup_kswapd(struct zone *zone, int order, enum zone_type classzone_idx); |
Mel Gorman | 7aeb09f | 2014-06-04 16:10:21 -0700 | [diff] [blame] | 792 | bool zone_watermark_ok(struct zone *z, unsigned int order, |
| 793 | unsigned long mark, int classzone_idx, int alloc_flags); |
| 794 | bool zone_watermark_ok_safe(struct zone *z, unsigned int order, |
| 795 | unsigned long mark, int classzone_idx, int alloc_flags); |
Dave Hansen | a2f3aa02 | 2007-01-10 23:15:30 -0800 | [diff] [blame] | 796 | enum memmap_context { |
| 797 | MEMMAP_EARLY, |
| 798 | MEMMAP_HOTPLUG, |
| 799 | }; |
Yasunori Goto | 718127c | 2006-06-23 02:03:10 -0700 | [diff] [blame] | 800 | extern int init_currently_empty_zone(struct zone *zone, unsigned long start_pfn, |
Dave Hansen | a2f3aa02 | 2007-01-10 23:15:30 -0800 | [diff] [blame] | 801 | unsigned long size, |
| 802 | enum memmap_context context); |
Yasunori Goto | 718127c | 2006-06-23 02:03:10 -0700 | [diff] [blame] | 803 | |
Hugh Dickins | bea8c15 | 2012-11-16 14:14:54 -0800 | [diff] [blame] | 804 | extern void lruvec_init(struct lruvec *lruvec); |
Konstantin Khlebnikov | 7f5e86c | 2012-05-29 15:06:58 -0700 | [diff] [blame] | 805 | |
| 806 | static inline struct zone *lruvec_zone(struct lruvec *lruvec) |
| 807 | { |
Andrew Morton | c255a45 | 2012-07-31 16:43:02 -0700 | [diff] [blame] | 808 | #ifdef CONFIG_MEMCG |
Konstantin Khlebnikov | 7f5e86c | 2012-05-29 15:06:58 -0700 | [diff] [blame] | 809 | return lruvec->zone; |
| 810 | #else |
| 811 | return container_of(lruvec, struct zone, lruvec); |
| 812 | #endif |
| 813 | } |
| 814 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 815 | #ifdef CONFIG_HAVE_MEMORY_PRESENT |
| 816 | void memory_present(int nid, unsigned long start, unsigned long end); |
| 817 | #else |
| 818 | static inline void memory_present(int nid, unsigned long start, unsigned long end) {} |
| 819 | #endif |
| 820 | |
Lee Schermerhorn | 7aac789 | 2010-05-26 14:45:00 -0700 | [diff] [blame] | 821 | #ifdef CONFIG_HAVE_MEMORYLESS_NODES |
| 822 | int local_memory_node(int node_id); |
| 823 | #else |
| 824 | static inline int local_memory_node(int node_id) { return node_id; }; |
| 825 | #endif |
| 826 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 827 | #ifdef CONFIG_NEED_NODE_MEMMAP_SIZE |
| 828 | unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long); |
| 829 | #endif |
| 830 | |
| 831 | /* |
| 832 | * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc. |
| 833 | */ |
| 834 | #define zone_idx(zone) ((zone) - (zone)->zone_pgdat->node_zones) |
| 835 | |
Con Kolivas | f3fe651 | 2006-01-06 00:11:15 -0800 | [diff] [blame] | 836 | static inline int populated_zone(struct zone *zone) |
| 837 | { |
| 838 | return (!!zone->present_pages); |
| 839 | } |
| 840 | |
Mel Gorman | 2a1e274 | 2007-07-17 04:03:12 -0700 | [diff] [blame] | 841 | extern int movable_zone; |
| 842 | |
| 843 | static inline int zone_movable_is_highmem(void) |
| 844 | { |
Rabin Vincent | fe03025 | 2012-07-31 16:43:14 -0700 | [diff] [blame] | 845 | #if defined(CONFIG_HIGHMEM) && defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) |
Mel Gorman | 2a1e274 | 2007-07-17 04:03:12 -0700 | [diff] [blame] | 846 | return movable_zone == ZONE_HIGHMEM; |
Wang Nan | 1a4dc5b | 2014-08-06 16:06:08 -0700 | [diff] [blame] | 847 | #elif defined(CONFIG_HIGHMEM) |
| 848 | return (ZONE_MOVABLE - 1) == ZONE_HIGHMEM; |
Mel Gorman | 2a1e274 | 2007-07-17 04:03:12 -0700 | [diff] [blame] | 849 | #else |
| 850 | return 0; |
| 851 | #endif |
| 852 | } |
| 853 | |
Christoph Lameter | 2f1b624 | 2006-09-25 23:31:13 -0700 | [diff] [blame] | 854 | static inline int is_highmem_idx(enum zone_type idx) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 855 | { |
Christoph Lameter | e53ef38 | 2006-09-25 23:31:14 -0700 | [diff] [blame] | 856 | #ifdef CONFIG_HIGHMEM |
Mel Gorman | 2a1e274 | 2007-07-17 04:03:12 -0700 | [diff] [blame] | 857 | return (idx == ZONE_HIGHMEM || |
| 858 | (idx == ZONE_MOVABLE && zone_movable_is_highmem())); |
Christoph Lameter | e53ef38 | 2006-09-25 23:31:14 -0700 | [diff] [blame] | 859 | #else |
| 860 | return 0; |
| 861 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 862 | } |
| 863 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 864 | /** |
| 865 | * is_highmem - helper function to quickly check if a struct zone is a |
| 866 | * highmem zone or not. This is an attempt to keep references |
| 867 | * to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum. |
| 868 | * @zone - pointer to struct zone variable |
| 869 | */ |
| 870 | static inline int is_highmem(struct zone *zone) |
| 871 | { |
Christoph Lameter | e53ef38 | 2006-09-25 23:31:14 -0700 | [diff] [blame] | 872 | #ifdef CONFIG_HIGHMEM |
Harvey Harrison | ddc81ed | 2008-04-28 02:12:07 -0700 | [diff] [blame] | 873 | int zone_off = (char *)zone - (char *)zone->zone_pgdat->node_zones; |
| 874 | return zone_off == ZONE_HIGHMEM * sizeof(*zone) || |
| 875 | (zone_off == ZONE_MOVABLE * sizeof(*zone) && |
| 876 | zone_movable_is_highmem()); |
Christoph Lameter | e53ef38 | 2006-09-25 23:31:14 -0700 | [diff] [blame] | 877 | #else |
| 878 | return 0; |
| 879 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 880 | } |
| 881 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 882 | /* These two functions are used to setup the per zone pages min values */ |
| 883 | struct ctl_table; |
Alexey Dobriyan | 8d65af7 | 2009-09-23 15:57:19 -0700 | [diff] [blame] | 884 | int min_free_kbytes_sysctl_handler(struct ctl_table *, int, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 885 | void __user *, size_t *, loff_t *); |
| 886 | extern int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1]; |
Alexey Dobriyan | 8d65af7 | 2009-09-23 15:57:19 -0700 | [diff] [blame] | 887 | int lowmem_reserve_ratio_sysctl_handler(struct ctl_table *, int, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 888 | void __user *, size_t *, loff_t *); |
Alexey Dobriyan | 8d65af7 | 2009-09-23 15:57:19 -0700 | [diff] [blame] | 889 | int percpu_pagelist_fraction_sysctl_handler(struct ctl_table *, int, |
Rohit Seth | 8ad4b1f | 2006-01-08 01:00:40 -0800 | [diff] [blame] | 890 | void __user *, size_t *, loff_t *); |
Christoph Lameter | 9614634 | 2006-07-03 00:24:13 -0700 | [diff] [blame] | 891 | int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table *, int, |
Alexey Dobriyan | 8d65af7 | 2009-09-23 15:57:19 -0700 | [diff] [blame] | 892 | void __user *, size_t *, loff_t *); |
Christoph Lameter | 0ff3849 | 2006-09-25 23:31:52 -0700 | [diff] [blame] | 893 | int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table *, int, |
Alexey Dobriyan | 8d65af7 | 2009-09-23 15:57:19 -0700 | [diff] [blame] | 894 | void __user *, size_t *, loff_t *); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 895 | |
KAMEZAWA Hiroyuki | f0c0b2b | 2007-07-15 23:38:01 -0700 | [diff] [blame] | 896 | extern int numa_zonelist_order_handler(struct ctl_table *, int, |
Alexey Dobriyan | 8d65af7 | 2009-09-23 15:57:19 -0700 | [diff] [blame] | 897 | void __user *, size_t *, loff_t *); |
KAMEZAWA Hiroyuki | f0c0b2b | 2007-07-15 23:38:01 -0700 | [diff] [blame] | 898 | extern char numa_zonelist_order[]; |
| 899 | #define NUMA_ZONELIST_ORDER_LEN 16 /* string buffer size */ |
| 900 | |
Dave Hansen | 93b7504 | 2005-06-23 00:07:47 -0700 | [diff] [blame] | 901 | #ifndef CONFIG_NEED_MULTIPLE_NODES |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 902 | |
| 903 | extern struct pglist_data contig_page_data; |
| 904 | #define NODE_DATA(nid) (&contig_page_data) |
| 905 | #define NODE_MEM_MAP(nid) mem_map |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 906 | |
Dave Hansen | 93b7504 | 2005-06-23 00:07:47 -0700 | [diff] [blame] | 907 | #else /* CONFIG_NEED_MULTIPLE_NODES */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 908 | |
| 909 | #include <asm/mmzone.h> |
| 910 | |
Dave Hansen | 93b7504 | 2005-06-23 00:07:47 -0700 | [diff] [blame] | 911 | #endif /* !CONFIG_NEED_MULTIPLE_NODES */ |
Dave Hansen | 348f8b6 | 2005-06-23 00:07:40 -0700 | [diff] [blame] | 912 | |
KAMEZAWA Hiroyuki | 95144c7 | 2006-03-27 01:16:02 -0800 | [diff] [blame] | 913 | extern struct pglist_data *first_online_pgdat(void); |
| 914 | extern struct pglist_data *next_online_pgdat(struct pglist_data *pgdat); |
| 915 | extern struct zone *next_zone(struct zone *zone); |
KAMEZAWA Hiroyuki | 8357f86 | 2006-03-27 01:15:57 -0800 | [diff] [blame] | 916 | |
| 917 | /** |
Fernando Luis Vazquez Cao | 12d15f0 | 2008-05-23 13:05:01 -0700 | [diff] [blame] | 918 | * for_each_online_pgdat - helper macro to iterate over all online nodes |
KAMEZAWA Hiroyuki | 8357f86 | 2006-03-27 01:15:57 -0800 | [diff] [blame] | 919 | * @pgdat - pointer to a pg_data_t variable |
| 920 | */ |
| 921 | #define for_each_online_pgdat(pgdat) \ |
| 922 | for (pgdat = first_online_pgdat(); \ |
| 923 | pgdat; \ |
| 924 | pgdat = next_online_pgdat(pgdat)) |
KAMEZAWA Hiroyuki | 8357f86 | 2006-03-27 01:15:57 -0800 | [diff] [blame] | 925 | /** |
| 926 | * for_each_zone - helper macro to iterate over all memory zones |
| 927 | * @zone - pointer to struct zone variable |
| 928 | * |
| 929 | * The user only needs to declare the zone variable, for_each_zone |
| 930 | * fills it in. |
| 931 | */ |
| 932 | #define for_each_zone(zone) \ |
| 933 | for (zone = (first_online_pgdat())->node_zones; \ |
| 934 | zone; \ |
| 935 | zone = next_zone(zone)) |
| 936 | |
KOSAKI Motohiro | ee99c71 | 2009-03-31 15:19:31 -0700 | [diff] [blame] | 937 | #define for_each_populated_zone(zone) \ |
| 938 | for (zone = (first_online_pgdat())->node_zones; \ |
| 939 | zone; \ |
| 940 | zone = next_zone(zone)) \ |
| 941 | if (!populated_zone(zone)) \ |
| 942 | ; /* do nothing */ \ |
| 943 | else |
| 944 | |
Mel Gorman | dd1a239 | 2008-04-28 02:12:17 -0700 | [diff] [blame] | 945 | static inline struct zone *zonelist_zone(struct zoneref *zoneref) |
| 946 | { |
| 947 | return zoneref->zone; |
| 948 | } |
| 949 | |
| 950 | static inline int zonelist_zone_idx(struct zoneref *zoneref) |
| 951 | { |
| 952 | return zoneref->zone_idx; |
| 953 | } |
| 954 | |
| 955 | static inline int zonelist_node_idx(struct zoneref *zoneref) |
| 956 | { |
| 957 | #ifdef CONFIG_NUMA |
| 958 | /* zone_to_nid not available in this context */ |
| 959 | return zoneref->zone->node; |
| 960 | #else |
| 961 | return 0; |
| 962 | #endif /* CONFIG_NUMA */ |
| 963 | } |
| 964 | |
Mel Gorman | 19770b3 | 2008-04-28 02:12:18 -0700 | [diff] [blame] | 965 | /** |
| 966 | * next_zones_zonelist - Returns the next zone at or below highest_zoneidx within the allowed nodemask using a cursor within a zonelist as a starting point |
| 967 | * @z - The cursor used as a starting point for the search |
| 968 | * @highest_zoneidx - The zone index of the highest zone to return |
| 969 | * @nodes - An optional nodemask to filter the zonelist with |
| 970 | * @zone - The first suitable zone found is returned via this parameter |
| 971 | * |
| 972 | * This function returns the next zone at or below a given zone index that is |
| 973 | * within the allowed nodemask using a cursor as the starting point for the |
Mel Gorman | 5bead2a | 2008-09-13 02:33:19 -0700 | [diff] [blame] | 974 | * search. The zoneref returned is a cursor that represents the current zone |
| 975 | * being examined. It should be advanced by one before calling |
| 976 | * next_zones_zonelist again. |
Mel Gorman | 19770b3 | 2008-04-28 02:12:18 -0700 | [diff] [blame] | 977 | */ |
| 978 | struct zoneref *next_zones_zonelist(struct zoneref *z, |
| 979 | enum zone_type highest_zoneidx, |
| 980 | nodemask_t *nodes, |
| 981 | struct zone **zone); |
Mel Gorman | dd1a239 | 2008-04-28 02:12:17 -0700 | [diff] [blame] | 982 | |
Mel Gorman | 19770b3 | 2008-04-28 02:12:18 -0700 | [diff] [blame] | 983 | /** |
| 984 | * first_zones_zonelist - Returns the first zone at or below highest_zoneidx within the allowed nodemask in a zonelist |
| 985 | * @zonelist - The zonelist to search for a suitable zone |
| 986 | * @highest_zoneidx - The zone index of the highest zone to return |
| 987 | * @nodes - An optional nodemask to filter the zonelist with |
| 988 | * @zone - The first suitable zone found is returned via this parameter |
| 989 | * |
| 990 | * This function returns the first zone at or below a given zone index that is |
| 991 | * within the allowed nodemask. The zoneref returned is a cursor that can be |
Mel Gorman | 5bead2a | 2008-09-13 02:33:19 -0700 | [diff] [blame] | 992 | * used to iterate the zonelist with next_zones_zonelist by advancing it by |
| 993 | * one before calling. |
Mel Gorman | 19770b3 | 2008-04-28 02:12:18 -0700 | [diff] [blame] | 994 | */ |
Mel Gorman | dd1a239 | 2008-04-28 02:12:17 -0700 | [diff] [blame] | 995 | static inline struct zoneref *first_zones_zonelist(struct zonelist *zonelist, |
Mel Gorman | 19770b3 | 2008-04-28 02:12:18 -0700 | [diff] [blame] | 996 | enum zone_type highest_zoneidx, |
| 997 | nodemask_t *nodes, |
| 998 | struct zone **zone) |
Mel Gorman | 54a6eb5 | 2008-04-28 02:12:16 -0700 | [diff] [blame] | 999 | { |
Mel Gorman | 19770b3 | 2008-04-28 02:12:18 -0700 | [diff] [blame] | 1000 | return next_zones_zonelist(zonelist->_zonerefs, highest_zoneidx, nodes, |
| 1001 | zone); |
Mel Gorman | 54a6eb5 | 2008-04-28 02:12:16 -0700 | [diff] [blame] | 1002 | } |
| 1003 | |
Mel Gorman | 19770b3 | 2008-04-28 02:12:18 -0700 | [diff] [blame] | 1004 | /** |
| 1005 | * for_each_zone_zonelist_nodemask - helper macro to iterate over valid zones in a zonelist at or below a given zone index and within a nodemask |
| 1006 | * @zone - The current zone in the iterator |
| 1007 | * @z - The current pointer within zonelist->zones being iterated |
| 1008 | * @zlist - The zonelist being iterated |
| 1009 | * @highidx - The zone index of the highest zone to return |
| 1010 | * @nodemask - Nodemask allowed by the allocator |
| 1011 | * |
| 1012 | * This iterator iterates though all zones at or below a given zone index and |
| 1013 | * within a given nodemask |
| 1014 | */ |
| 1015 | #define for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \ |
| 1016 | for (z = first_zones_zonelist(zlist, highidx, nodemask, &zone); \ |
| 1017 | zone; \ |
Mel Gorman | 5bead2a | 2008-09-13 02:33:19 -0700 | [diff] [blame] | 1018 | z = next_zones_zonelist(++z, highidx, nodemask, &zone)) \ |
Mel Gorman | 54a6eb5 | 2008-04-28 02:12:16 -0700 | [diff] [blame] | 1019 | |
| 1020 | /** |
| 1021 | * for_each_zone_zonelist - helper macro to iterate over valid zones in a zonelist at or below a given zone index |
| 1022 | * @zone - The current zone in the iterator |
| 1023 | * @z - The current pointer within zonelist->zones being iterated |
| 1024 | * @zlist - The zonelist being iterated |
| 1025 | * @highidx - The zone index of the highest zone to return |
| 1026 | * |
| 1027 | * This iterator iterates though all zones at or below a given zone index. |
| 1028 | */ |
| 1029 | #define for_each_zone_zonelist(zone, z, zlist, highidx) \ |
Mel Gorman | 19770b3 | 2008-04-28 02:12:18 -0700 | [diff] [blame] | 1030 | for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, NULL) |
Mel Gorman | 54a6eb5 | 2008-04-28 02:12:16 -0700 | [diff] [blame] | 1031 | |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 1032 | #ifdef CONFIG_SPARSEMEM |
| 1033 | #include <asm/sparsemem.h> |
| 1034 | #endif |
| 1035 | |
Mel Gorman | c713216 | 2006-09-27 01:49:43 -0700 | [diff] [blame] | 1036 | #if !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) && \ |
Tejun Heo | 0ee332c | 2011-12-08 10:22:09 -0800 | [diff] [blame] | 1037 | !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) |
Andrew Morton | b454456 | 2008-04-28 02:12:39 -0700 | [diff] [blame] | 1038 | static inline unsigned long early_pfn_to_nid(unsigned long pfn) |
| 1039 | { |
| 1040 | return 0; |
| 1041 | } |
Andy Whitcroft | b159d43 | 2005-06-23 00:07:52 -0700 | [diff] [blame] | 1042 | #endif |
| 1043 | |
Andy Whitcroft | 2bdaf11 | 2006-01-06 00:10:53 -0800 | [diff] [blame] | 1044 | #ifdef CONFIG_FLATMEM |
| 1045 | #define pfn_to_nid(pfn) (0) |
| 1046 | #endif |
| 1047 | |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 1048 | #ifdef CONFIG_SPARSEMEM |
| 1049 | |
| 1050 | /* |
| 1051 | * SECTION_SHIFT #bits space required to store a section # |
| 1052 | * |
| 1053 | * PA_SECTION_SHIFT physical address to/from section number |
| 1054 | * PFN_SECTION_SHIFT pfn to/from section number |
| 1055 | */ |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 1056 | #define PA_SECTION_SHIFT (SECTION_SIZE_BITS) |
| 1057 | #define PFN_SECTION_SHIFT (SECTION_SIZE_BITS - PAGE_SHIFT) |
| 1058 | |
| 1059 | #define NR_MEM_SECTIONS (1UL << SECTIONS_SHIFT) |
| 1060 | |
| 1061 | #define PAGES_PER_SECTION (1UL << PFN_SECTION_SHIFT) |
| 1062 | #define PAGE_SECTION_MASK (~(PAGES_PER_SECTION-1)) |
| 1063 | |
Mel Gorman | 835c134 | 2007-10-16 01:25:47 -0700 | [diff] [blame] | 1064 | #define SECTION_BLOCKFLAGS_BITS \ |
Mel Gorman | d9c2340 | 2007-10-16 01:26:01 -0700 | [diff] [blame] | 1065 | ((1UL << (PFN_SECTION_SHIFT - pageblock_order)) * NR_PAGEBLOCK_BITS) |
Mel Gorman | 835c134 | 2007-10-16 01:25:47 -0700 | [diff] [blame] | 1066 | |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 1067 | #if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS |
| 1068 | #error Allocator MAX_ORDER exceeds SECTION_SIZE |
| 1069 | #endif |
| 1070 | |
Daniel Kiper | e3c40f3 | 2011-05-24 17:12:33 -0700 | [diff] [blame] | 1071 | #define pfn_to_section_nr(pfn) ((pfn) >> PFN_SECTION_SHIFT) |
| 1072 | #define section_nr_to_pfn(sec) ((sec) << PFN_SECTION_SHIFT) |
| 1073 | |
Daniel Kiper | a539f35 | 2011-05-24 17:12:51 -0700 | [diff] [blame] | 1074 | #define SECTION_ALIGN_UP(pfn) (((pfn) + PAGES_PER_SECTION - 1) & PAGE_SECTION_MASK) |
| 1075 | #define SECTION_ALIGN_DOWN(pfn) ((pfn) & PAGE_SECTION_MASK) |
| 1076 | |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 1077 | struct page; |
| 1078 | struct mem_section { |
Andy Whitcroft | 29751f6 | 2005-06-23 00:08:00 -0700 | [diff] [blame] | 1079 | /* |
| 1080 | * This is, logically, a pointer to an array of struct |
| 1081 | * pages. However, it is stored with some other magic. |
| 1082 | * (see sparse.c::sparse_init_one_section()) |
| 1083 | * |
Andy Whitcroft | 30c253e | 2006-06-23 02:03:41 -0700 | [diff] [blame] | 1084 | * Additionally during early boot we encode node id of |
| 1085 | * the location of the section here to guide allocation. |
| 1086 | * (see sparse.c::memory_present()) |
| 1087 | * |
Andy Whitcroft | 29751f6 | 2005-06-23 00:08:00 -0700 | [diff] [blame] | 1088 | * Making it a UL at least makes someone do a cast |
| 1089 | * before using it wrong. |
| 1090 | */ |
| 1091 | unsigned long section_mem_map; |
Mel Gorman | 5c0e306 | 2007-10-16 01:25:56 -0700 | [diff] [blame] | 1092 | |
| 1093 | /* See declaration of similar field in struct zone */ |
| 1094 | unsigned long *pageblock_flags; |
Cody P Schafer | 55878e8 | 2013-07-03 15:04:44 -0700 | [diff] [blame] | 1095 | /* |
| 1096 | * WARNING: mem_section must be a power-of-2 in size for the |
| 1097 | * calculation and use of SECTION_ROOT_MASK to make sense. |
| 1098 | */ |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 1099 | }; |
| 1100 | |
Bob Picco | 3e34726 | 2005-09-03 15:54:28 -0700 | [diff] [blame] | 1101 | #ifdef CONFIG_SPARSEMEM_EXTREME |
| 1102 | #define SECTIONS_PER_ROOT (PAGE_SIZE / sizeof (struct mem_section)) |
Bob Picco | 802f192 | 2005-09-03 15:54:26 -0700 | [diff] [blame] | 1103 | #else |
Bob Picco | 3e34726 | 2005-09-03 15:54:28 -0700 | [diff] [blame] | 1104 | #define SECTIONS_PER_ROOT 1 |
| 1105 | #endif |
Bob Picco | 802f192 | 2005-09-03 15:54:26 -0700 | [diff] [blame] | 1106 | |
Bob Picco | 3e34726 | 2005-09-03 15:54:28 -0700 | [diff] [blame] | 1107 | #define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT) |
Marcelo Roberto Jimenez | 0faa563 | 2010-05-24 14:32:47 -0700 | [diff] [blame] | 1108 | #define NR_SECTION_ROOTS DIV_ROUND_UP(NR_MEM_SECTIONS, SECTIONS_PER_ROOT) |
Bob Picco | 3e34726 | 2005-09-03 15:54:28 -0700 | [diff] [blame] | 1109 | #define SECTION_ROOT_MASK (SECTIONS_PER_ROOT - 1) |
| 1110 | |
| 1111 | #ifdef CONFIG_SPARSEMEM_EXTREME |
| 1112 | extern struct mem_section *mem_section[NR_SECTION_ROOTS]; |
| 1113 | #else |
| 1114 | extern struct mem_section mem_section[NR_SECTION_ROOTS][SECTIONS_PER_ROOT]; |
| 1115 | #endif |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 1116 | |
Andy Whitcroft | 29751f6 | 2005-06-23 00:08:00 -0700 | [diff] [blame] | 1117 | static inline struct mem_section *__nr_to_section(unsigned long nr) |
| 1118 | { |
Bob Picco | 3e34726 | 2005-09-03 15:54:28 -0700 | [diff] [blame] | 1119 | if (!mem_section[SECTION_NR_TO_ROOT(nr)]) |
| 1120 | return NULL; |
| 1121 | return &mem_section[SECTION_NR_TO_ROOT(nr)][nr & SECTION_ROOT_MASK]; |
Andy Whitcroft | 29751f6 | 2005-06-23 00:08:00 -0700 | [diff] [blame] | 1122 | } |
Dave Hansen | 4ca644d | 2005-10-29 18:16:51 -0700 | [diff] [blame] | 1123 | extern int __section_nr(struct mem_section* ms); |
Yasunori Goto | 0475327 | 2008-04-28 02:13:31 -0700 | [diff] [blame] | 1124 | extern unsigned long usemap_size(void); |
Andy Whitcroft | 29751f6 | 2005-06-23 00:08:00 -0700 | [diff] [blame] | 1125 | |
| 1126 | /* |
| 1127 | * We use the lower bits of the mem_map pointer to store |
| 1128 | * a little bit of information. There should be at least |
| 1129 | * 3 bits here due to 32-bit alignment. |
| 1130 | */ |
| 1131 | #define SECTION_MARKED_PRESENT (1UL<<0) |
| 1132 | #define SECTION_HAS_MEM_MAP (1UL<<1) |
| 1133 | #define SECTION_MAP_LAST_BIT (1UL<<2) |
| 1134 | #define SECTION_MAP_MASK (~(SECTION_MAP_LAST_BIT-1)) |
Andy Whitcroft | 30c253e | 2006-06-23 02:03:41 -0700 | [diff] [blame] | 1135 | #define SECTION_NID_SHIFT 2 |
Andy Whitcroft | 29751f6 | 2005-06-23 00:08:00 -0700 | [diff] [blame] | 1136 | |
| 1137 | static inline struct page *__section_mem_map_addr(struct mem_section *section) |
| 1138 | { |
| 1139 | unsigned long map = section->section_mem_map; |
| 1140 | map &= SECTION_MAP_MASK; |
| 1141 | return (struct page *)map; |
| 1142 | } |
| 1143 | |
Andy Whitcroft | 540557b | 2007-10-16 01:24:11 -0700 | [diff] [blame] | 1144 | static inline int present_section(struct mem_section *section) |
Andy Whitcroft | 29751f6 | 2005-06-23 00:08:00 -0700 | [diff] [blame] | 1145 | { |
Bob Picco | 802f192 | 2005-09-03 15:54:26 -0700 | [diff] [blame] | 1146 | return (section && (section->section_mem_map & SECTION_MARKED_PRESENT)); |
Andy Whitcroft | 29751f6 | 2005-06-23 00:08:00 -0700 | [diff] [blame] | 1147 | } |
| 1148 | |
Andy Whitcroft | 540557b | 2007-10-16 01:24:11 -0700 | [diff] [blame] | 1149 | static inline int present_section_nr(unsigned long nr) |
| 1150 | { |
| 1151 | return present_section(__nr_to_section(nr)); |
| 1152 | } |
| 1153 | |
| 1154 | static inline int valid_section(struct mem_section *section) |
Andy Whitcroft | 29751f6 | 2005-06-23 00:08:00 -0700 | [diff] [blame] | 1155 | { |
Bob Picco | 802f192 | 2005-09-03 15:54:26 -0700 | [diff] [blame] | 1156 | return (section && (section->section_mem_map & SECTION_HAS_MEM_MAP)); |
Andy Whitcroft | 29751f6 | 2005-06-23 00:08:00 -0700 | [diff] [blame] | 1157 | } |
| 1158 | |
| 1159 | static inline int valid_section_nr(unsigned long nr) |
| 1160 | { |
| 1161 | return valid_section(__nr_to_section(nr)); |
| 1162 | } |
| 1163 | |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 1164 | static inline struct mem_section *__pfn_to_section(unsigned long pfn) |
| 1165 | { |
Andy Whitcroft | 29751f6 | 2005-06-23 00:08:00 -0700 | [diff] [blame] | 1166 | return __nr_to_section(pfn_to_section_nr(pfn)); |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 1167 | } |
| 1168 | |
Will Deacon | 7b7bf49 | 2011-05-19 13:21:14 +0100 | [diff] [blame] | 1169 | #ifndef CONFIG_HAVE_ARCH_PFN_VALID |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 1170 | static inline int pfn_valid(unsigned long pfn) |
| 1171 | { |
| 1172 | if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS) |
| 1173 | return 0; |
Andy Whitcroft | 29751f6 | 2005-06-23 00:08:00 -0700 | [diff] [blame] | 1174 | return valid_section(__nr_to_section(pfn_to_section_nr(pfn))); |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 1175 | } |
Will Deacon | 7b7bf49 | 2011-05-19 13:21:14 +0100 | [diff] [blame] | 1176 | #endif |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 1177 | |
Andy Whitcroft | 540557b | 2007-10-16 01:24:11 -0700 | [diff] [blame] | 1178 | static inline int pfn_present(unsigned long pfn) |
| 1179 | { |
| 1180 | if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS) |
| 1181 | return 0; |
| 1182 | return present_section(__nr_to_section(pfn_to_section_nr(pfn))); |
| 1183 | } |
| 1184 | |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 1185 | /* |
| 1186 | * These are _only_ used during initialisation, therefore they |
| 1187 | * can use __initdata ... They could have names to indicate |
| 1188 | * this restriction. |
| 1189 | */ |
| 1190 | #ifdef CONFIG_NUMA |
Andy Whitcroft | 161599f | 2006-01-06 00:10:54 -0800 | [diff] [blame] | 1191 | #define pfn_to_nid(pfn) \ |
| 1192 | ({ \ |
| 1193 | unsigned long __pfn_to_nid_pfn = (pfn); \ |
| 1194 | page_to_nid(pfn_to_page(__pfn_to_nid_pfn)); \ |
| 1195 | }) |
Andy Whitcroft | 2bdaf11 | 2006-01-06 00:10:53 -0800 | [diff] [blame] | 1196 | #else |
| 1197 | #define pfn_to_nid(pfn) (0) |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 1198 | #endif |
| 1199 | |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 1200 | #define early_pfn_valid(pfn) pfn_valid(pfn) |
| 1201 | void sparse_init(void); |
| 1202 | #else |
| 1203 | #define sparse_init() do {} while (0) |
Dave Hansen | 28ae55c | 2005-09-03 15:54:29 -0700 | [diff] [blame] | 1204 | #define sparse_index_init(_sec, _nid) do {} while (0) |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 1205 | #endif /* CONFIG_SPARSEMEM */ |
| 1206 | |
Andy Whitcroft | 7516795 | 2006-10-21 10:24:14 -0700 | [diff] [blame] | 1207 | #ifdef CONFIG_NODES_SPAN_OTHER_NODES |
KAMEZAWA Hiroyuki | cc2559b | 2009-02-18 14:48:33 -0800 | [diff] [blame] | 1208 | bool early_pfn_in_nid(unsigned long pfn, int nid); |
Andy Whitcroft | 7516795 | 2006-10-21 10:24:14 -0700 | [diff] [blame] | 1209 | #else |
| 1210 | #define early_pfn_in_nid(pfn, nid) (1) |
| 1211 | #endif |
| 1212 | |
Andy Whitcroft | d41dee3 | 2005-06-23 00:07:54 -0700 | [diff] [blame] | 1213 | #ifndef early_pfn_valid |
| 1214 | #define early_pfn_valid(pfn) (1) |
| 1215 | #endif |
| 1216 | |
| 1217 | void memory_present(int nid, unsigned long start, unsigned long end); |
| 1218 | unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long); |
| 1219 | |
Andy Whitcroft | 14e0729 | 2007-05-06 14:49:14 -0700 | [diff] [blame] | 1220 | /* |
| 1221 | * If it is possible to have holes within a MAX_ORDER_NR_PAGES, then we |
| 1222 | * need to check pfn validility within that MAX_ORDER_NR_PAGES block. |
| 1223 | * pfn_valid_within() should be used in this case; we optimise this away |
| 1224 | * when we have no holes within a MAX_ORDER_NR_PAGES block. |
| 1225 | */ |
| 1226 | #ifdef CONFIG_HOLES_IN_ZONE |
| 1227 | #define pfn_valid_within(pfn) pfn_valid(pfn) |
| 1228 | #else |
| 1229 | #define pfn_valid_within(pfn) (1) |
| 1230 | #endif |
| 1231 | |
Mel Gorman | eb33575 | 2009-05-13 17:34:48 +0100 | [diff] [blame] | 1232 | #ifdef CONFIG_ARCH_HAS_HOLES_MEMORYMODEL |
| 1233 | /* |
| 1234 | * pfn_valid() is meant to be able to tell if a given PFN has valid memmap |
| 1235 | * associated with it or not. In FLATMEM, it is expected that holes always |
| 1236 | * have valid memmap as long as there is valid PFNs either side of the hole. |
| 1237 | * In SPARSEMEM, it is assumed that a valid section has a memmap for the |
| 1238 | * entire section. |
| 1239 | * |
| 1240 | * However, an ARM, and maybe other embedded architectures in the future |
| 1241 | * free memmap backing holes to save memory on the assumption the memmap is |
| 1242 | * never used. The page_zone linkages are then broken even though pfn_valid() |
| 1243 | * returns true. A walker of the full memmap must then do this additional |
| 1244 | * check to ensure the memmap they are looking at is sane by making sure |
| 1245 | * the zone and PFN linkages are still valid. This is expensive, but walkers |
| 1246 | * of the full memmap are extremely rare. |
| 1247 | */ |
| 1248 | int memmap_valid_within(unsigned long pfn, |
| 1249 | struct page *page, struct zone *zone); |
| 1250 | #else |
| 1251 | static inline int memmap_valid_within(unsigned long pfn, |
| 1252 | struct page *page, struct zone *zone) |
| 1253 | { |
| 1254 | return 1; |
| 1255 | } |
| 1256 | #endif /* CONFIG_ARCH_HAS_HOLES_MEMORYMODEL */ |
| 1257 | |
Christoph Lameter | 9796547 | 2008-04-28 02:12:54 -0700 | [diff] [blame] | 1258 | #endif /* !__GENERATING_BOUNDS.H */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1259 | #endif /* !__ASSEMBLY__ */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1260 | #endif /* _LINUX_MMZONE_H */ |