blob: bd791e452ad7a96329d883afba92ffbdf7bbc9bc [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001#ifndef _LINUX_MMZONE_H
2#define _LINUX_MMZONE_H
3
Linus Torvalds1da177e2005-04-16 15:20:36 -07004#ifndef __ASSEMBLY__
Christoph Lameter97965472008-04-28 02:12:54 -07005#ifndef __GENERATING_BOUNDS_H
Linus Torvalds1da177e2005-04-16 15:20:36 -07006
Linus Torvalds1da177e2005-04-16 15:20:36 -07007#include <linux/spinlock.h>
8#include <linux/list.h>
9#include <linux/wait.h>
David Rientjese815af92007-10-16 23:25:54 -070010#include <linux/bitops.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070011#include <linux/cache.h>
12#include <linux/threads.h>
13#include <linux/numa.h>
14#include <linux/init.h>
Dave Hansenbdc8cb92005-10-29 18:16:53 -070015#include <linux/seqlock.h>
KAMEZAWA Hiroyuki8357f862006-03-27 01:15:57 -080016#include <linux/nodemask.h>
Mel Gorman835c1342007-10-16 01:25:47 -070017#include <linux/pageblock-flags.h>
Peter Zijlstrabbeae5b2013-02-22 16:34:30 -080018#include <linux/page-flags-layout.h>
Arun Sharma600634972011-07-26 16:09:06 -070019#include <linux/atomic.h>
Ralf Baechle93ff66b2006-06-04 02:51:29 -070020#include <asm/page.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070021
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 Piccoe984bb42006-05-20 15:00:31 -070028#define MAX_ORDER_NR_PAGES (1 << (MAX_ORDER - 1))
Linus Torvalds1da177e2005-04-16 15:20:36 -070029
Andy Whitcroft5ad333e2007-07-17 04:03:16 -070030/*
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 YanQing35fca532012-04-15 20:42:28 +080033 * coalesce naturally under reasonable reclaim pressure and those which
Andy Whitcroft5ad333e2007-07-17 04:03:16 -070034 * will not.
35 */
36#define PAGE_ALLOC_COSTLY_ORDER 3
37
Michal Nazarewicz47118af2011-12-29 13:09:50 +010038enum {
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 Kim194159f2013-02-22 16:33:58 -080060#ifdef CONFIG_MEMORY_ISOLATION
Michal Nazarewicz47118af2011-12-29 13:09:50 +010061 MIGRATE_ISOLATE, /* can't allocate from here */
Minchan Kim194159f2013-02-22 16:33:58 -080062#endif
Michal Nazarewicz47118af2011-12-29 13:09:50 +010063 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 Gormanb2a0ac82007-10-16 01:25:48 -070071
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 Gorman467c9962007-10-16 01:26:02 -070076extern int page_group_by_mobility_disabled;
77
78static inline int get_pageblock_migratetype(struct page *page)
79{
Mel Gorman467c9962007-10-16 01:26:02 -070080 return get_pageblock_flags_group(page, PB_migrate, PB_migrate_end);
81}
82
Linus Torvalds1da177e2005-04-16 15:20:36 -070083struct free_area {
Mel Gormanb2a0ac82007-10-16 01:25:48 -070084 struct list_head free_list[MIGRATE_TYPES];
Linus Torvalds1da177e2005-04-16 15:20:36 -070085 unsigned long nr_free;
86};
87
88struct pglist_data;
89
90/*
91 * zone->lock and zone->lru_lock are two of the hottest locks in the kernel.
92 * So add a wild amount of padding here to ensure that they fall into separate
93 * cachelines. There are very few zone structures in the machine, so space
94 * consumption is not a concern here.
95 */
96#if defined(CONFIG_SMP)
97struct zone_padding {
98 char x[0];
Ravikiran G Thirumalai22fc6ec2006-01-08 01:01:27 -080099} ____cacheline_internodealigned_in_smp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100#define ZONE_PADDING(name) struct zone_padding name;
101#else
102#define ZONE_PADDING(name)
103#endif
104
Christoph Lameter2244b952006-06-30 01:55:33 -0700105enum zone_stat_item {
Christoph Lameter51ed4492007-02-10 01:43:02 -0800106 /* First 128 byte cacheline (assuming 64 bit words) */
Christoph Lameterd23ad422007-02-10 01:43:02 -0800107 NR_FREE_PAGES,
Johannes Weiner81c0a2b2013-09-11 14:20:47 -0700108 NR_ALLOC_BATCH,
Christoph Lameterb69408e2008-10-18 20:26:14 -0700109 NR_LRU_BASE,
Rik van Riel4f98a2f2008-10-18 20:26:32 -0700110 NR_INACTIVE_ANON = NR_LRU_BASE, /* must match order of LRU_[IN]ACTIVE */
111 NR_ACTIVE_ANON, /* " " " " " */
112 NR_INACTIVE_FILE, /* " " " " " */
113 NR_ACTIVE_FILE, /* " " " " " */
Lee Schermerhorn894bc312008-10-18 20:26:39 -0700114 NR_UNEVICTABLE, /* " " " " " */
Nick Piggin5344b7e2008-10-18 20:26:51 -0700115 NR_MLOCK, /* mlock()ed pages found and moved off LRU */
Christoph Lameterf3dbd342006-06-30 01:55:36 -0700116 NR_ANON_PAGES, /* Mapped anonymous pages */
117 NR_FILE_MAPPED, /* pagecache pages mapped into pagetables.
Christoph Lameter65ba55f2006-06-30 01:55:34 -0700118 only modified from process context */
Christoph Lameter347ce432006-06-30 01:55:35 -0700119 NR_FILE_PAGES,
Christoph Lameterb1e7a8f2006-06-30 01:55:39 -0700120 NR_FILE_DIRTY,
Christoph Lameterce866b32006-06-30 01:55:40 -0700121 NR_WRITEBACK,
Christoph Lameter51ed4492007-02-10 01:43:02 -0800122 NR_SLAB_RECLAIMABLE,
123 NR_SLAB_UNRECLAIMABLE,
124 NR_PAGETABLE, /* used for pagetables */
KOSAKI Motohiroc6a7f572009-09-21 17:01:32 -0700125 NR_KERNEL_STACK,
126 /* Second 128 byte cacheline */
Christoph Lameterfd39fc82006-06-30 01:55:40 -0700127 NR_UNSTABLE_NFS, /* NFS unstable pages */
Christoph Lameterd2c5e302006-06-30 01:55:41 -0700128 NR_BOUNCE,
Andrew Mortone129b5c2006-09-27 01:50:00 -0700129 NR_VMSCAN_WRITE,
Mel Gorman49ea7eb2011-10-31 17:07:59 -0700130 NR_VMSCAN_IMMEDIATE, /* Prioritise for reclaim when writeback ends */
Miklos Szeredifc3ba692008-04-30 00:54:38 -0700131 NR_WRITEBACK_TEMP, /* Writeback using temporary buffers */
KOSAKI Motohiroa7312862009-09-21 17:01:37 -0700132 NR_ISOLATED_ANON, /* Temporary isolated pages from anon lru */
133 NR_ISOLATED_FILE, /* Temporary isolated pages from file lru */
KOSAKI Motohiro4b021082009-09-21 17:01:33 -0700134 NR_SHMEM, /* shmem pages (included tmpfs/GEM pages) */
Michael Rubinea941f02010-10-26 14:21:35 -0700135 NR_DIRTIED, /* page dirtyings since bootup */
136 NR_WRITTEN, /* page writings since bootup */
Christoph Lameterca889e62006-06-30 01:55:44 -0700137#ifdef CONFIG_NUMA
138 NUMA_HIT, /* allocated in intended node */
139 NUMA_MISS, /* allocated in non intended node */
140 NUMA_FOREIGN, /* was intended here, hit elsewhere */
141 NUMA_INTERLEAVE_HIT, /* interleaver preferred this zone */
142 NUMA_LOCAL, /* allocation from local node */
143 NUMA_OTHER, /* allocation from other node */
144#endif
Andrea Arcangeli79134172011-01-13 15:46:58 -0800145 NR_ANON_TRANSPARENT_HUGEPAGES,
Bartlomiej Zolnierkiewiczd1ce7492012-10-08 16:32:02 -0700146 NR_FREE_CMA_PAGES,
Christoph Lameter2244b952006-06-30 01:55:33 -0700147 NR_VM_ZONE_STAT_ITEMS };
148
Rik van Riel4f98a2f2008-10-18 20:26:32 -0700149/*
150 * We do arithmetic on the LRU lists in various places in the code,
151 * so it is important to keep the active lists LRU_ACTIVE higher in
152 * the array than the corresponding inactive lists, and to keep
153 * the *_FILE lists LRU_FILE higher than the corresponding _ANON lists.
154 *
155 * This has to be kept in sync with the statistics in zone_stat_item
156 * above and the descriptions in vmstat_text in mm/vmstat.c
157 */
158#define LRU_BASE 0
159#define LRU_ACTIVE 1
160#define LRU_FILE 2
161
Christoph Lameterb69408e2008-10-18 20:26:14 -0700162enum lru_list {
Rik van Riel4f98a2f2008-10-18 20:26:32 -0700163 LRU_INACTIVE_ANON = LRU_BASE,
164 LRU_ACTIVE_ANON = LRU_BASE + LRU_ACTIVE,
165 LRU_INACTIVE_FILE = LRU_BASE + LRU_FILE,
166 LRU_ACTIVE_FILE = LRU_BASE + LRU_FILE + LRU_ACTIVE,
Lee Schermerhorn894bc312008-10-18 20:26:39 -0700167 LRU_UNEVICTABLE,
Lee Schermerhorn894bc312008-10-18 20:26:39 -0700168 NR_LRU_LISTS
169};
Christoph Lameterb69408e2008-10-18 20:26:14 -0700170
Hugh Dickins41113042012-01-12 17:20:01 -0800171#define for_each_lru(lru) for (lru = 0; lru < NR_LRU_LISTS; lru++)
Christoph Lameterb69408e2008-10-18 20:26:14 -0700172
Hugh Dickins41113042012-01-12 17:20:01 -0800173#define for_each_evictable_lru(lru) for (lru = 0; lru <= LRU_ACTIVE_FILE; lru++)
Lee Schermerhorn894bc312008-10-18 20:26:39 -0700174
Hugh Dickins41113042012-01-12 17:20:01 -0800175static inline int is_file_lru(enum lru_list lru)
Rik van Riel4f98a2f2008-10-18 20:26:32 -0700176{
Hugh Dickins41113042012-01-12 17:20:01 -0800177 return (lru == LRU_INACTIVE_FILE || lru == LRU_ACTIVE_FILE);
Rik van Riel4f98a2f2008-10-18 20:26:32 -0700178}
179
Hugh Dickins41113042012-01-12 17:20:01 -0800180static inline int is_active_lru(enum lru_list lru)
Christoph Lameterb69408e2008-10-18 20:26:14 -0700181{
Hugh Dickins41113042012-01-12 17:20:01 -0800182 return (lru == LRU_ACTIVE_ANON || lru == LRU_ACTIVE_FILE);
Christoph Lameterb69408e2008-10-18 20:26:14 -0700183}
184
Hugh Dickins41113042012-01-12 17:20:01 -0800185static inline int is_unevictable_lru(enum lru_list lru)
Lee Schermerhorn894bc312008-10-18 20:26:39 -0700186{
Hugh Dickins41113042012-01-12 17:20:01 -0800187 return (lru == LRU_UNEVICTABLE);
Lee Schermerhorn894bc312008-10-18 20:26:39 -0700188}
189
Hugh Dickins89abfab2012-05-29 15:06:53 -0700190struct zone_reclaim_stat {
191 /*
192 * The pageout code in vmscan.c keeps track of how many of the
Jiri Kosina59f91e52012-06-29 14:45:58 +0200193 * mem/swap backed and file backed pages are referenced.
Hugh Dickins89abfab2012-05-29 15:06:53 -0700194 * The higher the rotated/scanned ratio, the more valuable
195 * that cache is.
196 *
197 * The anon LRU stats live in [0], file LRU stats in [1]
198 */
199 unsigned long recent_rotated[2];
200 unsigned long recent_scanned[2];
201};
202
Johannes Weiner6290df52012-01-12 17:18:10 -0800203struct lruvec {
204 struct list_head lists[NR_LRU_LISTS];
Hugh Dickins89abfab2012-05-29 15:06:53 -0700205 struct zone_reclaim_stat reclaim_stat;
Andrew Mortonc255a452012-07-31 16:43:02 -0700206#ifdef CONFIG_MEMCG
Konstantin Khlebnikov7f5e86c2012-05-29 15:06:58 -0700207 struct zone *zone;
208#endif
Johannes Weiner6290df52012-01-12 17:18:10 -0800209};
210
KAMEZAWA Hiroyukibb2a0de2011-07-26 16:08:22 -0700211/* Mask used at gathering information at once (see memcontrol.c) */
212#define LRU_ALL_FILE (BIT(LRU_INACTIVE_FILE) | BIT(LRU_ACTIVE_FILE))
213#define LRU_ALL_ANON (BIT(LRU_INACTIVE_ANON) | BIT(LRU_ACTIVE_ANON))
KAMEZAWA Hiroyukibb2a0de2011-07-26 16:08:22 -0700214#define LRU_ALL ((1 << NR_LRU_LISTS) - 1)
215
Minchan Kim39deaf82011-10-31 17:06:51 -0700216/* Isolate clean file */
Konstantin Khlebnikovf3fd4a62012-05-29 15:06:54 -0700217#define ISOLATE_CLEAN ((__force isolate_mode_t)0x1)
Minchan Kimf80c0672011-10-31 17:06:55 -0700218/* Isolate unmapped file */
Konstantin Khlebnikovf3fd4a62012-05-29 15:06:54 -0700219#define ISOLATE_UNMAPPED ((__force isolate_mode_t)0x2)
Mel Gormanc8244932012-01-12 17:19:38 -0800220/* Isolate for asynchronous migration */
Konstantin Khlebnikovf3fd4a62012-05-29 15:06:54 -0700221#define ISOLATE_ASYNC_MIGRATE ((__force isolate_mode_t)0x4)
Minchan Kime46a2872012-10-08 16:33:48 -0700222/* Isolate unevictable pages */
223#define ISOLATE_UNEVICTABLE ((__force isolate_mode_t)0x8)
Minchan Kim4356f212011-10-31 17:06:47 -0700224
225/* LRU Isolation modes. */
226typedef unsigned __bitwise__ isolate_mode_t;
227
Mel Gorman41858962009-06-16 15:32:12 -0700228enum zone_watermarks {
229 WMARK_MIN,
230 WMARK_LOW,
231 WMARK_HIGH,
232 NR_WMARK
233};
234
235#define min_wmark_pages(z) (z->watermark[WMARK_MIN])
236#define low_wmark_pages(z) (z->watermark[WMARK_LOW])
237#define high_wmark_pages(z) (z->watermark[WMARK_HIGH])
238
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239struct per_cpu_pages {
240 int count; /* number of pages in the list */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241 int high; /* high watermark, emptying needed */
242 int batch; /* chunk size for buddy add/remove */
Mel Gorman5f8dcc22009-09-21 17:03:19 -0700243
244 /* Lists of pages, one per migrate type stored on the pcp-lists */
245 struct list_head lists[MIGRATE_PCPTYPES];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246};
247
248struct per_cpu_pageset {
Christoph Lameter3dfa5722008-02-04 22:29:19 -0800249 struct per_cpu_pages pcp;
Christoph Lameter4037d452007-05-09 02:35:14 -0700250#ifdef CONFIG_NUMA
251 s8 expire;
252#endif
Christoph Lameter2244b952006-06-30 01:55:33 -0700253#ifdef CONFIG_SMP
Christoph Lameterdf9ecab2006-08-31 21:27:35 -0700254 s8 stat_threshold;
Christoph Lameter2244b952006-06-30 01:55:33 -0700255 s8 vm_stat_diff[NR_VM_ZONE_STAT_ITEMS];
256#endif
Christoph Lameter99dcc3e2010-01-05 15:34:51 +0900257};
Christoph Lametere7c8d5c2005-06-21 17:14:47 -0700258
Christoph Lameter97965472008-04-28 02:12:54 -0700259#endif /* !__GENERATING_BOUNDS.H */
260
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700261enum zone_type {
Christoph Lameter4b51d662007-02-10 01:43:10 -0800262#ifdef CONFIG_ZONE_DMA
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700263 /*
264 * ZONE_DMA is used when there are devices that are not able
265 * to do DMA to all of addressable memory (ZONE_NORMAL). Then we
266 * carve out the portion of memory that is needed for these devices.
267 * The range is arch specific.
268 *
269 * Some examples
270 *
271 * Architecture Limit
272 * ---------------------------
273 * parisc, ia64, sparc <4G
274 * s390 <2G
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700275 * arm Various
276 * alpha Unlimited or 0-16MB.
277 *
278 * i386, x86_64 and multiple other arches
279 * <16M.
280 */
281 ZONE_DMA,
Christoph Lameter4b51d662007-02-10 01:43:10 -0800282#endif
Christoph Lameterfb0e7942006-09-25 23:31:13 -0700283#ifdef CONFIG_ZONE_DMA32
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700284 /*
285 * x86_64 needs two ZONE_DMAs because it supports devices that are
286 * only able to do DMA to the lower 16M but also 32 bit devices that
287 * can only do DMA areas below 4G.
288 */
289 ZONE_DMA32,
Christoph Lameterfb0e7942006-09-25 23:31:13 -0700290#endif
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700291 /*
292 * Normal addressable memory is in ZONE_NORMAL. DMA operations can be
293 * performed on pages in ZONE_NORMAL if the DMA devices support
294 * transfers to all addressable memory.
295 */
296 ZONE_NORMAL,
Christoph Lametere53ef382006-09-25 23:31:14 -0700297#ifdef CONFIG_HIGHMEM
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700298 /*
299 * A memory area that is only addressable by the kernel through
300 * mapping portions into its own address space. This is for example
301 * used by i386 to allow the kernel to address the memory beyond
302 * 900MB. The kernel will set up special mappings (page
303 * table entries on i386) for each page that the kernel needs to
304 * access.
305 */
306 ZONE_HIGHMEM,
Christoph Lametere53ef382006-09-25 23:31:14 -0700307#endif
Mel Gorman2a1e2742007-07-17 04:03:12 -0700308 ZONE_MOVABLE,
Christoph Lameter97965472008-04-28 02:12:54 -0700309 __MAX_NR_ZONES
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700310};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700311
Christoph Lameter97965472008-04-28 02:12:54 -0700312#ifndef __GENERATING_BOUNDS_H
313
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314struct zone {
315 /* Fields commonly accessed by the page allocator */
Mel Gorman41858962009-06-16 15:32:12 -0700316
317 /* zone watermarks, access with *_wmark_pages(zone) macros */
318 unsigned long watermark[NR_WMARK];
319
Linus Torvalds1da177e2005-04-16 15:20:36 -0700320 /*
Christoph Lameteraa454842010-09-09 16:38:17 -0700321 * When free pages are below this point, additional steps are taken
322 * when reading the number of free pages to avoid per-cpu counter
323 * drift allowing watermarks to be breached
324 */
325 unsigned long percpu_drift_mark;
326
327 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328 * We don't know if the memory that we're going to allocate will be freeable
329 * or/and it will be released eventually, so to avoid totally wasting several
330 * GB of ram we must reserve some of the lower zone memory (otherwise we risk
331 * to run OOM on the lower zones despite there's tons of freeable ram
332 * on the higher zones). This array is recalculated at runtime if the
333 * sysctl_lowmem_reserve_ratio sysctl changes.
334 */
335 unsigned long lowmem_reserve[MAX_NR_ZONES];
336
Johannes Weinerab8fabd2012-01-10 15:07:42 -0800337 /*
338 * This is a per-zone reserve of pages that should not be
339 * considered dirtyable memory.
340 */
341 unsigned long dirty_balance_reserve;
342
Christoph Lametere7c8d5c2005-06-21 17:14:47 -0700343#ifdef CONFIG_NUMA
Christoph Lameterd5f541e2006-09-27 01:50:08 -0700344 int node;
Christoph Lameter96146342006-07-03 00:24:13 -0700345 /*
346 * zone reclaim becomes active if more unmapped pages exist.
347 */
Christoph Lameter8417bba2006-09-25 23:31:51 -0700348 unsigned long min_unmapped_pages;
Christoph Lameter0ff38492006-09-25 23:31:52 -0700349 unsigned long min_slab_pages;
Christoph Lametere7c8d5c2005-06-21 17:14:47 -0700350#endif
Tejun Heo43cf38e2010-02-02 14:38:57 +0900351 struct per_cpu_pageset __percpu *pageset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352 /*
353 * free areas of different sizes
354 */
355 spinlock_t lock;
Mel Gormanbb13ffe2012-10-08 16:32:41 -0700356#if defined CONFIG_COMPACTION || defined CONFIG_CMA
Mel Gorman62997022012-10-08 16:32:47 -0700357 /* Set to true when the PG_migrate_skip bits should be cleared */
358 bool compact_blockskip_flush;
Mel Gormanc89511a2012-10-08 16:32:45 -0700359
360 /* pfns where compaction scanners should start */
361 unsigned long compact_cached_free_pfn;
362 unsigned long compact_cached_migrate_pfn;
Mel Gormanbb13ffe2012-10-08 16:32:41 -0700363#endif
Dave Hansenbdc8cb92005-10-29 18:16:53 -0700364#ifdef CONFIG_MEMORY_HOTPLUG
365 /* see spanned/present_pages for more description */
366 seqlock_t span_seqlock;
367#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368 struct free_area free_area[MAX_ORDER];
369
Mel Gorman835c1342007-10-16 01:25:47 -0700370#ifndef CONFIG_SPARSEMEM
371 /*
Mel Gormand9c23402007-10-16 01:26:01 -0700372 * Flags for a pageblock_nr_pages block. See pageblock-flags.h.
Mel Gorman835c1342007-10-16 01:25:47 -0700373 * In SPARSEMEM, this map is stored in struct mem_section
374 */
375 unsigned long *pageblock_flags;
376#endif /* CONFIG_SPARSEMEM */
377
Mel Gorman4f92e252010-05-24 14:32:32 -0700378#ifdef CONFIG_COMPACTION
379 /*
380 * On compaction failure, 1<<compact_defer_shift compactions
381 * are skipped before trying again. The number attempted since
382 * last failure is tracked with compact_considered.
383 */
384 unsigned int compact_considered;
385 unsigned int compact_defer_shift;
Rik van Rielaff62242012-03-21 16:33:52 -0700386 int compact_order_failed;
Mel Gorman4f92e252010-05-24 14:32:32 -0700387#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388
389 ZONE_PADDING(_pad1_)
390
391 /* Fields commonly accessed by the page reclaim scanner */
Johannes Weiner6290df52012-01-12 17:18:10 -0800392 spinlock_t lru_lock;
393 struct lruvec lruvec;
Rik van Riel4f98a2f2008-10-18 20:26:32 -0700394
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395 unsigned long pages_scanned; /* since last reclaim */
David Rientjese815af92007-10-16 23:25:54 -0700396 unsigned long flags; /* zone flags, see below */
Martin Hicks753ee722005-06-21 17:14:41 -0700397
Christoph Lameter2244b952006-06-30 01:55:33 -0700398 /* Zone statistics */
399 atomic_long_t vm_stat[NR_VM_ZONE_STAT_ITEMS];
Christoph Lameter9eeff232006-01-18 17:42:31 -0800400
401 /*
Rik van Riel556adec2008-10-18 20:26:34 -0700402 * The target ratio of ACTIVE_ANON to INACTIVE_ANON pages on
403 * this zone's LRU. Maintained by the pageout code.
404 */
405 unsigned int inactive_ratio;
406
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407
408 ZONE_PADDING(_pad2_)
409 /* Rarely used or read-mostly fields */
410
411 /*
412 * wait_table -- the array holding the hash table
Yasunori Goto02b694d2006-06-23 02:03:08 -0700413 * wait_table_hash_nr_entries -- the size of the hash table array
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414 * wait_table_bits -- wait_table_size == (1 << wait_table_bits)
415 *
416 * The purpose of all these is to keep track of the people
417 * waiting for a page to become available and make them
418 * runnable again when possible. The trouble is that this
419 * consumes a lot of space, especially when so few things
420 * wait on pages at a given time. So instead of using
421 * per-page waitqueues, we use a waitqueue hash table.
422 *
423 * The bucket discipline is to sleep on the same queue when
424 * colliding and wake all in that wait queue when removing.
425 * When something wakes, it must check to be sure its page is
426 * truly available, a la thundering herd. The cost of a
427 * collision is great, but given the expected load of the
428 * table, they should be so rare as to be outweighed by the
429 * benefits from the saved space.
430 *
431 * __wait_on_page_locked() and unlock_page() in mm/filemap.c, are the
432 * primary users of these fields, and in mm/page_alloc.c
433 * free_area_init_core() performs the initialization of them.
434 */
435 wait_queue_head_t * wait_table;
Yasunori Goto02b694d2006-06-23 02:03:08 -0700436 unsigned long wait_table_hash_nr_entries;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437 unsigned long wait_table_bits;
438
439 /*
440 * Discontig memory support fields.
441 */
442 struct pglist_data *zone_pgdat;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443 /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */
444 unsigned long zone_start_pfn;
445
Dave Hansenbdc8cb92005-10-29 18:16:53 -0700446 /*
Jiang Liu9feedc92012-12-12 13:52:12 -0800447 * spanned_pages is the total pages spanned by the zone, including
448 * holes, which is calculated as:
449 * spanned_pages = zone_end_pfn - zone_start_pfn;
Dave Hansenbdc8cb92005-10-29 18:16:53 -0700450 *
Jiang Liu9feedc92012-12-12 13:52:12 -0800451 * present_pages is physical pages existing within the zone, which
452 * is calculated as:
Cody P Schafer8761e312013-03-26 10:30:44 -0700453 * present_pages = spanned_pages - absent_pages(pages in holes);
Jiang Liu9feedc92012-12-12 13:52:12 -0800454 *
455 * managed_pages is present pages managed by the buddy system, which
456 * is calculated as (reserved_pages includes pages allocated by the
457 * bootmem allocator):
458 * managed_pages = present_pages - reserved_pages;
459 *
460 * So present_pages may be used by memory hotplug or memory power
461 * management logic to figure out unmanaged pages by checking
462 * (present_pages - managed_pages). And managed_pages should be used
463 * by page allocator and vm scanner to calculate all kinds of watermarks
464 * and thresholds.
465 *
466 * Locking rules:
467 *
468 * zone_start_pfn and spanned_pages are protected by span_seqlock.
469 * It is a seqlock because it has to be read outside of zone->lock,
470 * and it is done in the main allocator path. But, it is written
471 * quite infrequently.
472 *
473 * The span_seq lock is declared along with zone->lock because it is
Dave Hansenbdc8cb92005-10-29 18:16:53 -0700474 * frequently read in proximity to zone->lock. It's good to
475 * give them a chance of being in the same cacheline.
Jiang Liu9feedc92012-12-12 13:52:12 -0800476 *
Jiang Liuc3d5f5f2013-07-03 15:03:14 -0700477 * Write access to present_pages at runtime should be protected by
478 * lock_memory_hotplug()/unlock_memory_hotplug(). Any reader who can't
479 * tolerant drift of present_pages should hold memory hotplug lock to
480 * get a stable value.
481 *
482 * Read access to managed_pages should be safe because it's unsigned
483 * long. Write access to zone->managed_pages and totalram_pages are
484 * protected by managed_page_count_lock at runtime. Idealy only
485 * adjust_managed_page_count() should be used instead of directly
486 * touching zone->managed_pages and totalram_pages.
Dave Hansenbdc8cb92005-10-29 18:16:53 -0700487 */
Jiang Liu9feedc92012-12-12 13:52:12 -0800488 unsigned long spanned_pages;
489 unsigned long present_pages;
490 unsigned long managed_pages;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491
492 /*
493 * rarely used fields:
494 */
Helge Deller15ad7cd2006-12-06 20:40:36 -0800495 const char *name;
Ravikiran G Thirumalai22fc6ec2006-01-08 01:01:27 -0800496} ____cacheline_internodealigned_in_smp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497
David Rientjese815af92007-10-16 23:25:54 -0700498typedef enum {
David Rientjese815af92007-10-16 23:25:54 -0700499 ZONE_RECLAIM_LOCKED, /* prevents concurrent reclaim */
David Rientjes098d7f12007-10-16 23:25:55 -0700500 ZONE_OOM_LOCKED, /* zone is in OOM killer zonelist */
Mel Gorman0e093d992010-10-26 14:21:45 -0700501 ZONE_CONGESTED, /* zone has many dirty pages backed by
502 * a congested BDI
503 */
Mel Gormand43006d2013-07-03 15:01:50 -0700504 ZONE_TAIL_LRU_DIRTY, /* reclaim scanning has recently found
505 * many dirty file pages at the tail
506 * of the LRU.
507 */
Mel Gorman283aba92013-07-03 15:01:51 -0700508 ZONE_WRITEBACK, /* reclaim scanning has recently found
509 * many pages under writeback
510 */
David Rientjese815af92007-10-16 23:25:54 -0700511} zone_flags_t;
512
513static inline void zone_set_flag(struct zone *zone, zone_flags_t flag)
514{
515 set_bit(flag, &zone->flags);
516}
David Rientjesd773ed62007-10-16 23:26:01 -0700517
518static inline int zone_test_and_set_flag(struct zone *zone, zone_flags_t flag)
519{
520 return test_and_set_bit(flag, &zone->flags);
521}
522
David Rientjese815af92007-10-16 23:25:54 -0700523static inline void zone_clear_flag(struct zone *zone, zone_flags_t flag)
524{
525 clear_bit(flag, &zone->flags);
526}
527
Mel Gorman0e093d992010-10-26 14:21:45 -0700528static inline int zone_is_reclaim_congested(const struct zone *zone)
529{
530 return test_bit(ZONE_CONGESTED, &zone->flags);
531}
532
Mel Gormand43006d2013-07-03 15:01:50 -0700533static inline int zone_is_reclaim_dirty(const struct zone *zone)
534{
535 return test_bit(ZONE_TAIL_LRU_DIRTY, &zone->flags);
536}
537
Mel Gorman283aba92013-07-03 15:01:51 -0700538static inline int zone_is_reclaim_writeback(const struct zone *zone)
539{
540 return test_bit(ZONE_WRITEBACK, &zone->flags);
541}
542
David Rientjese815af92007-10-16 23:25:54 -0700543static inline int zone_is_reclaim_locked(const struct zone *zone)
544{
545 return test_bit(ZONE_RECLAIM_LOCKED, &zone->flags);
546}
David Rientjesd773ed62007-10-16 23:26:01 -0700547
David Rientjes098d7f12007-10-16 23:25:55 -0700548static inline int zone_is_oom_locked(const struct zone *zone)
549{
550 return test_bit(ZONE_OOM_LOCKED, &zone->flags);
551}
David Rientjese815af92007-10-16 23:25:54 -0700552
Russ Andersonf9228b22013-03-22 15:04:43 -0700553static inline unsigned long zone_end_pfn(const struct zone *zone)
Cody P Schafer108bcc92013-02-22 16:35:23 -0800554{
555 return zone->zone_start_pfn + zone->spanned_pages;
556}
557
558static inline bool zone_spans_pfn(const struct zone *zone, unsigned long pfn)
559{
560 return zone->zone_start_pfn <= pfn && pfn < zone_end_pfn(zone);
561}
562
Cody P Schafer2a6e3eb2013-02-22 16:35:24 -0800563static inline bool zone_is_initialized(struct zone *zone)
564{
565 return !!zone->wait_table;
566}
567
568static inline bool zone_is_empty(struct zone *zone)
569{
570 return zone->spanned_pages == 0;
571}
572
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573/*
574 * The "priority" of VM scanning is how much of the queues we will scan in one
575 * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the
576 * queues ("queue_length >> 12") during an aging round.
577 */
578#define DEF_PRIORITY 12
579
Paul Jackson9276b1bc2006-12-06 20:31:48 -0800580/* Maximum number of zones on a zonelist */
581#define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES)
582
583#ifdef CONFIG_NUMA
Christoph Lameter523b9452007-10-16 01:25:37 -0700584
585/*
Pete Zaitcev25a64ec2011-02-03 22:43:48 -0700586 * The NUMA zonelists are doubled because we need zonelists that restrict the
Christoph Lameter523b9452007-10-16 01:25:37 -0700587 * allocations to a single node for GFP_THISNODE.
588 *
Mel Gorman54a6eb52008-04-28 02:12:16 -0700589 * [0] : Zonelist with fallback
590 * [1] : No fallback (GFP_THISNODE)
Christoph Lameter523b9452007-10-16 01:25:37 -0700591 */
Mel Gorman54a6eb52008-04-28 02:12:16 -0700592#define MAX_ZONELISTS 2
Christoph Lameter523b9452007-10-16 01:25:37 -0700593
594
Paul Jackson9276b1bc2006-12-06 20:31:48 -0800595/*
596 * We cache key information from each zonelist for smaller cache
597 * footprint when scanning for free pages in get_page_from_freelist().
598 *
599 * 1) The BITMAP fullzones tracks which zones in a zonelist have come
600 * up short of free memory since the last time (last_fullzone_zap)
601 * we zero'd fullzones.
602 * 2) The array z_to_n[] maps each zone in the zonelist to its node
603 * id, so that we can efficiently evaluate whether that node is
604 * set in the current tasks mems_allowed.
605 *
606 * Both fullzones and z_to_n[] are one-to-one with the zonelist,
607 * indexed by a zones offset in the zonelist zones[] array.
608 *
609 * The get_page_from_freelist() routine does two scans. During the
610 * first scan, we skip zones whose corresponding bit in 'fullzones'
611 * is set or whose corresponding node in current->mems_allowed (which
612 * comes from cpusets) is not set. During the second scan, we bypass
613 * this zonelist_cache, to ensure we look methodically at each zone.
614 *
615 * Once per second, we zero out (zap) fullzones, forcing us to
616 * reconsider nodes that might have regained more free memory.
617 * The field last_full_zap is the time we last zapped fullzones.
618 *
619 * This mechanism reduces the amount of time we waste repeatedly
620 * reexaming zones for free memory when they just came up low on
621 * memory momentarilly ago.
622 *
623 * The zonelist_cache struct members logically belong in struct
624 * zonelist. However, the mempolicy zonelists constructed for
625 * MPOL_BIND are intentionally variable length (and usually much
626 * shorter). A general purpose mechanism for handling structs with
627 * multiple variable length members is more mechanism than we want
628 * here. We resort to some special case hackery instead.
629 *
630 * The MPOL_BIND zonelists don't need this zonelist_cache (in good
631 * part because they are shorter), so we put the fixed length stuff
632 * at the front of the zonelist struct, ending in a variable length
633 * zones[], as is needed by MPOL_BIND.
634 *
635 * Then we put the optional zonelist cache on the end of the zonelist
636 * struct. This optional stuff is found by a 'zlcache_ptr' pointer in
637 * the fixed length portion at the front of the struct. This pointer
638 * both enables us to find the zonelist cache, and in the case of
639 * MPOL_BIND zonelists, (which will just set the zlcache_ptr to NULL)
640 * to know that the zonelist cache is not there.
641 *
642 * The end result is that struct zonelists come in two flavors:
643 * 1) The full, fixed length version, shown below, and
644 * 2) The custom zonelists for MPOL_BIND.
645 * The custom MPOL_BIND zonelists have a NULL zlcache_ptr and no zlcache.
646 *
647 * Even though there may be multiple CPU cores on a node modifying
648 * fullzones or last_full_zap in the same zonelist_cache at the same
649 * time, we don't lock it. This is just hint data - if it is wrong now
650 * and then, the allocator will still function, perhaps a bit slower.
651 */
652
653
654struct zonelist_cache {
Paul Jackson9276b1bc2006-12-06 20:31:48 -0800655 unsigned short z_to_n[MAX_ZONES_PER_ZONELIST]; /* zone->nid */
Paul Jackson7253f4e2006-12-06 20:31:49 -0800656 DECLARE_BITMAP(fullzones, MAX_ZONES_PER_ZONELIST); /* zone full? */
Paul Jackson9276b1bc2006-12-06 20:31:48 -0800657 unsigned long last_full_zap; /* when last zap'd (jiffies) */
658};
659#else
Mel Gorman54a6eb52008-04-28 02:12:16 -0700660#define MAX_ZONELISTS 1
Paul Jackson9276b1bc2006-12-06 20:31:48 -0800661struct zonelist_cache;
662#endif
663
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664/*
Mel Gormandd1a2392008-04-28 02:12:17 -0700665 * This struct contains information about a zone in a zonelist. It is stored
666 * here to avoid dereferences into large structures and lookups of tables
667 */
668struct zoneref {
669 struct zone *zone; /* Pointer to actual zone */
670 int zone_idx; /* zone_idx(zoneref->zone) */
671};
672
673/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674 * One allocation request operates on a zonelist. A zonelist
675 * is a list of zones, the first one is the 'goal' of the
676 * allocation, the other zones are fallback zones, in decreasing
677 * priority.
678 *
Paul Jackson9276b1bc2006-12-06 20:31:48 -0800679 * If zlcache_ptr is not NULL, then it is just the address of zlcache,
680 * as explained above. If zlcache_ptr is NULL, there is no zlcache.
Mel Gormandd1a2392008-04-28 02:12:17 -0700681 * *
682 * To speed the reading of the zonelist, the zonerefs contain the zone index
683 * of the entry being read. Helper functions to access information given
684 * a struct zoneref are
685 *
686 * zonelist_zone() - Return the struct zone * for an entry in _zonerefs
687 * zonelist_zone_idx() - Return the index of the zone for an entry
688 * zonelist_node_idx() - Return the index of the node for an entry
Linus Torvalds1da177e2005-04-16 15:20:36 -0700689 */
690struct zonelist {
Paul Jackson9276b1bc2006-12-06 20:31:48 -0800691 struct zonelist_cache *zlcache_ptr; // NULL or &zlcache
Mel Gormandd1a2392008-04-28 02:12:17 -0700692 struct zoneref _zonerefs[MAX_ZONES_PER_ZONELIST + 1];
Paul Jackson9276b1bc2006-12-06 20:31:48 -0800693#ifdef CONFIG_NUMA
694 struct zonelist_cache zlcache; // optional ...
695#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700696};
697
Tejun Heo0ee332c2011-12-08 10:22:09 -0800698#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
Mel Gormanc7132162006-09-27 01:49:43 -0700699struct node_active_region {
700 unsigned long start_pfn;
701 unsigned long end_pfn;
702 int nid;
703};
Tejun Heo0ee332c2011-12-08 10:22:09 -0800704#endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705
Heiko Carstens5b99cd02006-09-27 01:50:01 -0700706#ifndef CONFIG_DISCONTIGMEM
707/* The array of struct pages - for discontigmem use pgdat->lmem_map */
708extern struct page *mem_map;
709#endif
710
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711/*
712 * The pg_data_t structure is used in machines with CONFIG_DISCONTIGMEM
713 * (mostly NUMA machines?) to denote a higher-level memory zone than the
714 * zone denotes.
715 *
716 * On NUMA machines, each NUMA node would have a pg_data_t to describe
717 * it's memory layout.
718 *
719 * Memory statistics and page replacement data structures are maintained on a
720 * per-zone basis.
721 */
722struct bootmem_data;
723typedef struct pglist_data {
724 struct zone node_zones[MAX_NR_ZONES];
Christoph Lameter523b9452007-10-16 01:25:37 -0700725 struct zonelist node_zonelists[MAX_ZONELISTS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726 int nr_zones;
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -0700727#ifdef CONFIG_FLAT_NODE_MEM_MAP /* means !SPARSEMEM */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700728 struct page *node_mem_map;
Andrew Mortonc255a452012-07-31 16:43:02 -0700729#ifdef CONFIG_MEMCG
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -0700730 struct page_cgroup *node_page_cgroup;
731#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700732#endif
Yinghai Lu08677212010-02-10 01:20:20 -0800733#ifndef CONFIG_NO_BOOTMEM
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734 struct bootmem_data *bdata;
Yinghai Lu08677212010-02-10 01:20:20 -0800735#endif
Dave Hansen208d54e2005-10-29 18:16:52 -0700736#ifdef CONFIG_MEMORY_HOTPLUG
737 /*
738 * Must be held any time you expect node_start_pfn, node_present_pages
739 * or node_spanned_pages stay constant. Holding this will also
740 * guarantee that any pfn_valid() stays that way.
741 *
Cody P Schafer114d4b72013-07-03 15:02:09 -0700742 * pgdat_resize_lock() and pgdat_resize_unlock() are provided to
743 * manipulate node_size_lock without checking for CONFIG_MEMORY_HOTPLUG.
744 *
Cody P Schafer72c3b512013-07-03 15:02:08 -0700745 * Nests above zone->lock and zone->span_seqlock
Dave Hansen208d54e2005-10-29 18:16:52 -0700746 */
747 spinlock_t node_size_lock;
748#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749 unsigned long node_start_pfn;
750 unsigned long node_present_pages; /* total number of physical pages */
751 unsigned long node_spanned_pages; /* total size of physical page
752 range, including holes */
753 int node_id;
David Rientjes957f8222012-10-08 16:33:24 -0700754 nodemask_t reclaim_nodes; /* Nodes allowed to reclaim from */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755 wait_queue_head_t kswapd_wait;
Mel Gorman55150612012-07-31 16:44:35 -0700756 wait_queue_head_t pfmemalloc_wait;
Jiang Liud8adde12012-07-11 14:01:52 -0700757 struct task_struct *kswapd; /* Protected by lock_memory_hotplug() */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758 int kswapd_max_order;
Mel Gorman99504742011-01-13 15:46:20 -0800759 enum zone_type classzone_idx;
Andrea Arcangeli8177a422012-03-23 20:56:34 +0100760#ifdef CONFIG_NUMA_BALANCING
761 /*
762 * Lock serializing the per destination node AutoNUMA memory
763 * migration rate limiting data.
764 */
765 spinlock_t numabalancing_migrate_lock;
766
767 /* Rate limiting time interval */
768 unsigned long numabalancing_migrate_next_window;
769
770 /* Number of pages migrated during the rate limiting time interval */
771 unsigned long numabalancing_migrate_nr_pages;
772#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773} pg_data_t;
774
775#define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages)
776#define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700777#ifdef CONFIG_FLAT_NODE_MEM_MAP
Dave Hansen408fde82005-06-23 00:07:37 -0700778#define pgdat_page_nr(pgdat, pagenr) ((pgdat)->node_mem_map + (pagenr))
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700779#else
780#define pgdat_page_nr(pgdat, pagenr) pfn_to_page((pgdat)->node_start_pfn + (pagenr))
781#endif
Dave Hansen408fde82005-06-23 00:07:37 -0700782#define nid_page_nr(nid, pagenr) pgdat_page_nr(NODE_DATA(nid),(pagenr))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783
KAMEZAWA Hiroyukic6830c22011-06-16 17:28:07 +0900784#define node_start_pfn(nid) (NODE_DATA(nid)->node_start_pfn)
Cody P Schaferda3649e2013-02-22 16:35:27 -0800785#define node_end_pfn(nid) pgdat_end_pfn(NODE_DATA(nid))
KAMEZAWA Hiroyukic6830c22011-06-16 17:28:07 +0900786
Cody P Schaferda3649e2013-02-22 16:35:27 -0800787static inline unsigned long pgdat_end_pfn(pg_data_t *pgdat)
788{
789 return pgdat->node_start_pfn + pgdat->node_spanned_pages;
790}
791
792static inline bool pgdat_is_empty(pg_data_t *pgdat)
793{
794 return !pgdat->node_start_pfn && !pgdat->node_spanned_pages;
795}
KAMEZAWA Hiroyukic6830c22011-06-16 17:28:07 +0900796
Dave Hansen208d54e2005-10-29 18:16:52 -0700797#include <linux/memory_hotplug.h>
798
Haicheng Li4eaf3f62010-05-24 14:32:52 -0700799extern struct mutex zonelists_mutex;
Jiang Liu9adb62a2012-07-31 16:43:28 -0700800void build_all_zonelists(pg_data_t *pgdat, struct zone *zone);
Mel Gorman99504742011-01-13 15:46:20 -0800801void wakeup_kswapd(struct zone *zone, int order, enum zone_type classzone_idx);
Mel Gorman88f5acf2011-01-13 15:45:41 -0800802bool zone_watermark_ok(struct zone *z, int order, unsigned long mark,
803 int classzone_idx, int alloc_flags);
804bool zone_watermark_ok_safe(struct zone *z, int order, unsigned long mark,
Rohit Seth7fb1d9f2005-11-13 16:06:43 -0800805 int classzone_idx, int alloc_flags);
Dave Hansena2f3aa022007-01-10 23:15:30 -0800806enum memmap_context {
807 MEMMAP_EARLY,
808 MEMMAP_HOTPLUG,
809};
Yasunori Goto718127c2006-06-23 02:03:10 -0700810extern int init_currently_empty_zone(struct zone *zone, unsigned long start_pfn,
Dave Hansena2f3aa022007-01-10 23:15:30 -0800811 unsigned long size,
812 enum memmap_context context);
Yasunori Goto718127c2006-06-23 02:03:10 -0700813
Hugh Dickinsbea8c152012-11-16 14:14:54 -0800814extern void lruvec_init(struct lruvec *lruvec);
Konstantin Khlebnikov7f5e86c2012-05-29 15:06:58 -0700815
816static inline struct zone *lruvec_zone(struct lruvec *lruvec)
817{
Andrew Mortonc255a452012-07-31 16:43:02 -0700818#ifdef CONFIG_MEMCG
Konstantin Khlebnikov7f5e86c2012-05-29 15:06:58 -0700819 return lruvec->zone;
820#else
821 return container_of(lruvec, struct zone, lruvec);
822#endif
823}
824
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825#ifdef CONFIG_HAVE_MEMORY_PRESENT
826void memory_present(int nid, unsigned long start, unsigned long end);
827#else
828static inline void memory_present(int nid, unsigned long start, unsigned long end) {}
829#endif
830
Lee Schermerhorn7aac7892010-05-26 14:45:00 -0700831#ifdef CONFIG_HAVE_MEMORYLESS_NODES
832int local_memory_node(int node_id);
833#else
834static inline int local_memory_node(int node_id) { return node_id; };
835#endif
836
Linus Torvalds1da177e2005-04-16 15:20:36 -0700837#ifdef CONFIG_NEED_NODE_MEMMAP_SIZE
838unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
839#endif
840
841/*
842 * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc.
843 */
844#define zone_idx(zone) ((zone) - (zone)->zone_pgdat->node_zones)
845
Con Kolivasf3fe6512006-01-06 00:11:15 -0800846static inline int populated_zone(struct zone *zone)
847{
848 return (!!zone->present_pages);
849}
850
Mel Gorman2a1e2742007-07-17 04:03:12 -0700851extern int movable_zone;
852
853static inline int zone_movable_is_highmem(void)
854{
Rabin Vincentfe030252012-07-31 16:43:14 -0700855#if defined(CONFIG_HIGHMEM) && defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP)
Mel Gorman2a1e2742007-07-17 04:03:12 -0700856 return movable_zone == ZONE_HIGHMEM;
857#else
858 return 0;
859#endif
860}
861
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700862static inline int is_highmem_idx(enum zone_type idx)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863{
Christoph Lametere53ef382006-09-25 23:31:14 -0700864#ifdef CONFIG_HIGHMEM
Mel Gorman2a1e2742007-07-17 04:03:12 -0700865 return (idx == ZONE_HIGHMEM ||
866 (idx == ZONE_MOVABLE && zone_movable_is_highmem()));
Christoph Lametere53ef382006-09-25 23:31:14 -0700867#else
868 return 0;
869#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700870}
871
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872/**
873 * is_highmem - helper function to quickly check if a struct zone is a
874 * highmem zone or not. This is an attempt to keep references
875 * to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum.
876 * @zone - pointer to struct zone variable
877 */
878static inline int is_highmem(struct zone *zone)
879{
Christoph Lametere53ef382006-09-25 23:31:14 -0700880#ifdef CONFIG_HIGHMEM
Harvey Harrisonddc81ed2008-04-28 02:12:07 -0700881 int zone_off = (char *)zone - (char *)zone->zone_pgdat->node_zones;
882 return zone_off == ZONE_HIGHMEM * sizeof(*zone) ||
883 (zone_off == ZONE_MOVABLE * sizeof(*zone) &&
884 zone_movable_is_highmem());
Christoph Lametere53ef382006-09-25 23:31:14 -0700885#else
886 return 0;
887#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888}
889
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890/* These two functions are used to setup the per zone pages min values */
891struct ctl_table;
Alexey Dobriyan8d65af72009-09-23 15:57:19 -0700892int min_free_kbytes_sysctl_handler(struct ctl_table *, int,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893 void __user *, size_t *, loff_t *);
894extern int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1];
Alexey Dobriyan8d65af72009-09-23 15:57:19 -0700895int lowmem_reserve_ratio_sysctl_handler(struct ctl_table *, int,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896 void __user *, size_t *, loff_t *);
Alexey Dobriyan8d65af72009-09-23 15:57:19 -0700897int percpu_pagelist_fraction_sysctl_handler(struct ctl_table *, int,
Rohit Seth8ad4b1f2006-01-08 01:00:40 -0800898 void __user *, size_t *, loff_t *);
Christoph Lameter96146342006-07-03 00:24:13 -0700899int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table *, int,
Alexey Dobriyan8d65af72009-09-23 15:57:19 -0700900 void __user *, size_t *, loff_t *);
Christoph Lameter0ff38492006-09-25 23:31:52 -0700901int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table *, int,
Alexey Dobriyan8d65af72009-09-23 15:57:19 -0700902 void __user *, size_t *, loff_t *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903
KAMEZAWA Hiroyukif0c0b2b2007-07-15 23:38:01 -0700904extern int numa_zonelist_order_handler(struct ctl_table *, int,
Alexey Dobriyan8d65af72009-09-23 15:57:19 -0700905 void __user *, size_t *, loff_t *);
KAMEZAWA Hiroyukif0c0b2b2007-07-15 23:38:01 -0700906extern char numa_zonelist_order[];
907#define NUMA_ZONELIST_ORDER_LEN 16 /* string buffer size */
908
Dave Hansen93b75042005-06-23 00:07:47 -0700909#ifndef CONFIG_NEED_MULTIPLE_NODES
Linus Torvalds1da177e2005-04-16 15:20:36 -0700910
911extern struct pglist_data contig_page_data;
912#define NODE_DATA(nid) (&contig_page_data)
913#define NODE_MEM_MAP(nid) mem_map
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914
Dave Hansen93b75042005-06-23 00:07:47 -0700915#else /* CONFIG_NEED_MULTIPLE_NODES */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916
917#include <asm/mmzone.h>
918
Dave Hansen93b75042005-06-23 00:07:47 -0700919#endif /* !CONFIG_NEED_MULTIPLE_NODES */
Dave Hansen348f8b62005-06-23 00:07:40 -0700920
KAMEZAWA Hiroyuki95144c72006-03-27 01:16:02 -0800921extern struct pglist_data *first_online_pgdat(void);
922extern struct pglist_data *next_online_pgdat(struct pglist_data *pgdat);
923extern struct zone *next_zone(struct zone *zone);
KAMEZAWA Hiroyuki8357f862006-03-27 01:15:57 -0800924
925/**
Fernando Luis Vazquez Cao12d15f02008-05-23 13:05:01 -0700926 * for_each_online_pgdat - helper macro to iterate over all online nodes
KAMEZAWA Hiroyuki8357f862006-03-27 01:15:57 -0800927 * @pgdat - pointer to a pg_data_t variable
928 */
929#define for_each_online_pgdat(pgdat) \
930 for (pgdat = first_online_pgdat(); \
931 pgdat; \
932 pgdat = next_online_pgdat(pgdat))
KAMEZAWA Hiroyuki8357f862006-03-27 01:15:57 -0800933/**
934 * for_each_zone - helper macro to iterate over all memory zones
935 * @zone - pointer to struct zone variable
936 *
937 * The user only needs to declare the zone variable, for_each_zone
938 * fills it in.
939 */
940#define for_each_zone(zone) \
941 for (zone = (first_online_pgdat())->node_zones; \
942 zone; \
943 zone = next_zone(zone))
944
KOSAKI Motohiroee99c712009-03-31 15:19:31 -0700945#define for_each_populated_zone(zone) \
946 for (zone = (first_online_pgdat())->node_zones; \
947 zone; \
948 zone = next_zone(zone)) \
949 if (!populated_zone(zone)) \
950 ; /* do nothing */ \
951 else
952
Mel Gormandd1a2392008-04-28 02:12:17 -0700953static inline struct zone *zonelist_zone(struct zoneref *zoneref)
954{
955 return zoneref->zone;
956}
957
958static inline int zonelist_zone_idx(struct zoneref *zoneref)
959{
960 return zoneref->zone_idx;
961}
962
963static inline int zonelist_node_idx(struct zoneref *zoneref)
964{
965#ifdef CONFIG_NUMA
966 /* zone_to_nid not available in this context */
967 return zoneref->zone->node;
968#else
969 return 0;
970#endif /* CONFIG_NUMA */
971}
972
Mel Gorman19770b32008-04-28 02:12:18 -0700973/**
974 * 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
975 * @z - The cursor used as a starting point for the search
976 * @highest_zoneidx - The zone index of the highest zone to return
977 * @nodes - An optional nodemask to filter the zonelist with
978 * @zone - The first suitable zone found is returned via this parameter
979 *
980 * This function returns the next zone at or below a given zone index that is
981 * within the allowed nodemask using a cursor as the starting point for the
Mel Gorman5bead2a2008-09-13 02:33:19 -0700982 * search. The zoneref returned is a cursor that represents the current zone
983 * being examined. It should be advanced by one before calling
984 * next_zones_zonelist again.
Mel Gorman19770b32008-04-28 02:12:18 -0700985 */
986struct zoneref *next_zones_zonelist(struct zoneref *z,
987 enum zone_type highest_zoneidx,
988 nodemask_t *nodes,
989 struct zone **zone);
Mel Gormandd1a2392008-04-28 02:12:17 -0700990
Mel Gorman19770b32008-04-28 02:12:18 -0700991/**
992 * first_zones_zonelist - Returns the first zone at or below highest_zoneidx within the allowed nodemask in a zonelist
993 * @zonelist - The zonelist to search for a suitable zone
994 * @highest_zoneidx - The zone index of the highest zone to return
995 * @nodes - An optional nodemask to filter the zonelist with
996 * @zone - The first suitable zone found is returned via this parameter
997 *
998 * This function returns the first zone at or below a given zone index that is
999 * within the allowed nodemask. The zoneref returned is a cursor that can be
Mel Gorman5bead2a2008-09-13 02:33:19 -07001000 * used to iterate the zonelist with next_zones_zonelist by advancing it by
1001 * one before calling.
Mel Gorman19770b32008-04-28 02:12:18 -07001002 */
Mel Gormandd1a2392008-04-28 02:12:17 -07001003static inline struct zoneref *first_zones_zonelist(struct zonelist *zonelist,
Mel Gorman19770b32008-04-28 02:12:18 -07001004 enum zone_type highest_zoneidx,
1005 nodemask_t *nodes,
1006 struct zone **zone)
Mel Gorman54a6eb52008-04-28 02:12:16 -07001007{
Mel Gorman19770b32008-04-28 02:12:18 -07001008 return next_zones_zonelist(zonelist->_zonerefs, highest_zoneidx, nodes,
1009 zone);
Mel Gorman54a6eb52008-04-28 02:12:16 -07001010}
1011
Mel Gorman19770b32008-04-28 02:12:18 -07001012/**
1013 * 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
1014 * @zone - The current zone in the iterator
1015 * @z - The current pointer within zonelist->zones being iterated
1016 * @zlist - The zonelist being iterated
1017 * @highidx - The zone index of the highest zone to return
1018 * @nodemask - Nodemask allowed by the allocator
1019 *
1020 * This iterator iterates though all zones at or below a given zone index and
1021 * within a given nodemask
1022 */
1023#define for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \
1024 for (z = first_zones_zonelist(zlist, highidx, nodemask, &zone); \
1025 zone; \
Mel Gorman5bead2a2008-09-13 02:33:19 -07001026 z = next_zones_zonelist(++z, highidx, nodemask, &zone)) \
Mel Gorman54a6eb52008-04-28 02:12:16 -07001027
1028/**
1029 * for_each_zone_zonelist - helper macro to iterate over valid zones in a zonelist at or below a given zone index
1030 * @zone - The current zone in the iterator
1031 * @z - The current pointer within zonelist->zones being iterated
1032 * @zlist - The zonelist being iterated
1033 * @highidx - The zone index of the highest zone to return
1034 *
1035 * This iterator iterates though all zones at or below a given zone index.
1036 */
1037#define for_each_zone_zonelist(zone, z, zlist, highidx) \
Mel Gorman19770b32008-04-28 02:12:18 -07001038 for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, NULL)
Mel Gorman54a6eb52008-04-28 02:12:16 -07001039
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001040#ifdef CONFIG_SPARSEMEM
1041#include <asm/sparsemem.h>
1042#endif
1043
Mel Gormanc7132162006-09-27 01:49:43 -07001044#if !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) && \
Tejun Heo0ee332c2011-12-08 10:22:09 -08001045 !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP)
Andrew Mortonb4544562008-04-28 02:12:39 -07001046static inline unsigned long early_pfn_to_nid(unsigned long pfn)
1047{
1048 return 0;
1049}
Andy Whitcroftb159d432005-06-23 00:07:52 -07001050#endif
1051
Andy Whitcroft2bdaf112006-01-06 00:10:53 -08001052#ifdef CONFIG_FLATMEM
1053#define pfn_to_nid(pfn) (0)
1054#endif
1055
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001056#ifdef CONFIG_SPARSEMEM
1057
1058/*
1059 * SECTION_SHIFT #bits space required to store a section #
1060 *
1061 * PA_SECTION_SHIFT physical address to/from section number
1062 * PFN_SECTION_SHIFT pfn to/from section number
1063 */
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001064#define PA_SECTION_SHIFT (SECTION_SIZE_BITS)
1065#define PFN_SECTION_SHIFT (SECTION_SIZE_BITS - PAGE_SHIFT)
1066
1067#define NR_MEM_SECTIONS (1UL << SECTIONS_SHIFT)
1068
1069#define PAGES_PER_SECTION (1UL << PFN_SECTION_SHIFT)
1070#define PAGE_SECTION_MASK (~(PAGES_PER_SECTION-1))
1071
Mel Gorman835c1342007-10-16 01:25:47 -07001072#define SECTION_BLOCKFLAGS_BITS \
Mel Gormand9c23402007-10-16 01:26:01 -07001073 ((1UL << (PFN_SECTION_SHIFT - pageblock_order)) * NR_PAGEBLOCK_BITS)
Mel Gorman835c1342007-10-16 01:25:47 -07001074
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001075#if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
1076#error Allocator MAX_ORDER exceeds SECTION_SIZE
1077#endif
1078
Daniel Kipere3c40f32011-05-24 17:12:33 -07001079#define pfn_to_section_nr(pfn) ((pfn) >> PFN_SECTION_SHIFT)
1080#define section_nr_to_pfn(sec) ((sec) << PFN_SECTION_SHIFT)
1081
Daniel Kipera539f352011-05-24 17:12:51 -07001082#define SECTION_ALIGN_UP(pfn) (((pfn) + PAGES_PER_SECTION - 1) & PAGE_SECTION_MASK)
1083#define SECTION_ALIGN_DOWN(pfn) ((pfn) & PAGE_SECTION_MASK)
1084
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001085struct page;
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -07001086struct page_cgroup;
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001087struct mem_section {
Andy Whitcroft29751f62005-06-23 00:08:00 -07001088 /*
1089 * This is, logically, a pointer to an array of struct
1090 * pages. However, it is stored with some other magic.
1091 * (see sparse.c::sparse_init_one_section())
1092 *
Andy Whitcroft30c253e2006-06-23 02:03:41 -07001093 * Additionally during early boot we encode node id of
1094 * the location of the section here to guide allocation.
1095 * (see sparse.c::memory_present())
1096 *
Andy Whitcroft29751f62005-06-23 00:08:00 -07001097 * Making it a UL at least makes someone do a cast
1098 * before using it wrong.
1099 */
1100 unsigned long section_mem_map;
Mel Gorman5c0e3062007-10-16 01:25:56 -07001101
1102 /* See declaration of similar field in struct zone */
1103 unsigned long *pageblock_flags;
Andrew Mortonc255a452012-07-31 16:43:02 -07001104#ifdef CONFIG_MEMCG
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -07001105 /*
1106 * If !SPARSEMEM, pgdat doesn't have page_cgroup pointer. We use
1107 * section. (see memcontrol.h/page_cgroup.h about this.)
1108 */
1109 struct page_cgroup *page_cgroup;
1110 unsigned long pad;
1111#endif
Cody P Schafer55878e82013-07-03 15:04:44 -07001112 /*
1113 * WARNING: mem_section must be a power-of-2 in size for the
1114 * calculation and use of SECTION_ROOT_MASK to make sense.
1115 */
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001116};
1117
Bob Picco3e347262005-09-03 15:54:28 -07001118#ifdef CONFIG_SPARSEMEM_EXTREME
1119#define SECTIONS_PER_ROOT (PAGE_SIZE / sizeof (struct mem_section))
Bob Picco802f1922005-09-03 15:54:26 -07001120#else
Bob Picco3e347262005-09-03 15:54:28 -07001121#define SECTIONS_PER_ROOT 1
1122#endif
Bob Picco802f1922005-09-03 15:54:26 -07001123
Bob Picco3e347262005-09-03 15:54:28 -07001124#define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT)
Marcelo Roberto Jimenez0faa5632010-05-24 14:32:47 -07001125#define NR_SECTION_ROOTS DIV_ROUND_UP(NR_MEM_SECTIONS, SECTIONS_PER_ROOT)
Bob Picco3e347262005-09-03 15:54:28 -07001126#define SECTION_ROOT_MASK (SECTIONS_PER_ROOT - 1)
1127
1128#ifdef CONFIG_SPARSEMEM_EXTREME
1129extern struct mem_section *mem_section[NR_SECTION_ROOTS];
1130#else
1131extern struct mem_section mem_section[NR_SECTION_ROOTS][SECTIONS_PER_ROOT];
1132#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001133
Andy Whitcroft29751f62005-06-23 00:08:00 -07001134static inline struct mem_section *__nr_to_section(unsigned long nr)
1135{
Bob Picco3e347262005-09-03 15:54:28 -07001136 if (!mem_section[SECTION_NR_TO_ROOT(nr)])
1137 return NULL;
1138 return &mem_section[SECTION_NR_TO_ROOT(nr)][nr & SECTION_ROOT_MASK];
Andy Whitcroft29751f62005-06-23 00:08:00 -07001139}
Dave Hansen4ca644d2005-10-29 18:16:51 -07001140extern int __section_nr(struct mem_section* ms);
Yasunori Goto04753272008-04-28 02:13:31 -07001141extern unsigned long usemap_size(void);
Andy Whitcroft29751f62005-06-23 00:08:00 -07001142
1143/*
1144 * We use the lower bits of the mem_map pointer to store
1145 * a little bit of information. There should be at least
1146 * 3 bits here due to 32-bit alignment.
1147 */
1148#define SECTION_MARKED_PRESENT (1UL<<0)
1149#define SECTION_HAS_MEM_MAP (1UL<<1)
1150#define SECTION_MAP_LAST_BIT (1UL<<2)
1151#define SECTION_MAP_MASK (~(SECTION_MAP_LAST_BIT-1))
Andy Whitcroft30c253e2006-06-23 02:03:41 -07001152#define SECTION_NID_SHIFT 2
Andy Whitcroft29751f62005-06-23 00:08:00 -07001153
1154static inline struct page *__section_mem_map_addr(struct mem_section *section)
1155{
1156 unsigned long map = section->section_mem_map;
1157 map &= SECTION_MAP_MASK;
1158 return (struct page *)map;
1159}
1160
Andy Whitcroft540557b2007-10-16 01:24:11 -07001161static inline int present_section(struct mem_section *section)
Andy Whitcroft29751f62005-06-23 00:08:00 -07001162{
Bob Picco802f1922005-09-03 15:54:26 -07001163 return (section && (section->section_mem_map & SECTION_MARKED_PRESENT));
Andy Whitcroft29751f62005-06-23 00:08:00 -07001164}
1165
Andy Whitcroft540557b2007-10-16 01:24:11 -07001166static inline int present_section_nr(unsigned long nr)
1167{
1168 return present_section(__nr_to_section(nr));
1169}
1170
1171static inline int valid_section(struct mem_section *section)
Andy Whitcroft29751f62005-06-23 00:08:00 -07001172{
Bob Picco802f1922005-09-03 15:54:26 -07001173 return (section && (section->section_mem_map & SECTION_HAS_MEM_MAP));
Andy Whitcroft29751f62005-06-23 00:08:00 -07001174}
1175
1176static inline int valid_section_nr(unsigned long nr)
1177{
1178 return valid_section(__nr_to_section(nr));
1179}
1180
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001181static inline struct mem_section *__pfn_to_section(unsigned long pfn)
1182{
Andy Whitcroft29751f62005-06-23 00:08:00 -07001183 return __nr_to_section(pfn_to_section_nr(pfn));
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001184}
1185
Will Deacon7b7bf492011-05-19 13:21:14 +01001186#ifndef CONFIG_HAVE_ARCH_PFN_VALID
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001187static inline int pfn_valid(unsigned long pfn)
1188{
1189 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1190 return 0;
Andy Whitcroft29751f62005-06-23 00:08:00 -07001191 return valid_section(__nr_to_section(pfn_to_section_nr(pfn)));
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001192}
Will Deacon7b7bf492011-05-19 13:21:14 +01001193#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001194
Andy Whitcroft540557b2007-10-16 01:24:11 -07001195static inline int pfn_present(unsigned long pfn)
1196{
1197 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1198 return 0;
1199 return present_section(__nr_to_section(pfn_to_section_nr(pfn)));
1200}
1201
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001202/*
1203 * These are _only_ used during initialisation, therefore they
1204 * can use __initdata ... They could have names to indicate
1205 * this restriction.
1206 */
1207#ifdef CONFIG_NUMA
Andy Whitcroft161599f2006-01-06 00:10:54 -08001208#define pfn_to_nid(pfn) \
1209({ \
1210 unsigned long __pfn_to_nid_pfn = (pfn); \
1211 page_to_nid(pfn_to_page(__pfn_to_nid_pfn)); \
1212})
Andy Whitcroft2bdaf112006-01-06 00:10:53 -08001213#else
1214#define pfn_to_nid(pfn) (0)
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001215#endif
1216
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001217#define early_pfn_valid(pfn) pfn_valid(pfn)
1218void sparse_init(void);
1219#else
1220#define sparse_init() do {} while (0)
Dave Hansen28ae55c2005-09-03 15:54:29 -07001221#define sparse_index_init(_sec, _nid) do {} while (0)
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001222#endif /* CONFIG_SPARSEMEM */
1223
Andy Whitcroft75167952006-10-21 10:24:14 -07001224#ifdef CONFIG_NODES_SPAN_OTHER_NODES
KAMEZAWA Hiroyukicc2559b2009-02-18 14:48:33 -08001225bool early_pfn_in_nid(unsigned long pfn, int nid);
Andy Whitcroft75167952006-10-21 10:24:14 -07001226#else
1227#define early_pfn_in_nid(pfn, nid) (1)
1228#endif
1229
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001230#ifndef early_pfn_valid
1231#define early_pfn_valid(pfn) (1)
1232#endif
1233
1234void memory_present(int nid, unsigned long start, unsigned long end);
1235unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
1236
Andy Whitcroft14e07292007-05-06 14:49:14 -07001237/*
1238 * If it is possible to have holes within a MAX_ORDER_NR_PAGES, then we
1239 * need to check pfn validility within that MAX_ORDER_NR_PAGES block.
1240 * pfn_valid_within() should be used in this case; we optimise this away
1241 * when we have no holes within a MAX_ORDER_NR_PAGES block.
1242 */
1243#ifdef CONFIG_HOLES_IN_ZONE
1244#define pfn_valid_within(pfn) pfn_valid(pfn)
1245#else
1246#define pfn_valid_within(pfn) (1)
1247#endif
1248
Mel Gormaneb335752009-05-13 17:34:48 +01001249#ifdef CONFIG_ARCH_HAS_HOLES_MEMORYMODEL
1250/*
1251 * pfn_valid() is meant to be able to tell if a given PFN has valid memmap
1252 * associated with it or not. In FLATMEM, it is expected that holes always
1253 * have valid memmap as long as there is valid PFNs either side of the hole.
1254 * In SPARSEMEM, it is assumed that a valid section has a memmap for the
1255 * entire section.
1256 *
1257 * However, an ARM, and maybe other embedded architectures in the future
1258 * free memmap backing holes to save memory on the assumption the memmap is
1259 * never used. The page_zone linkages are then broken even though pfn_valid()
1260 * returns true. A walker of the full memmap must then do this additional
1261 * check to ensure the memmap they are looking at is sane by making sure
1262 * the zone and PFN linkages are still valid. This is expensive, but walkers
1263 * of the full memmap are extremely rare.
1264 */
1265int memmap_valid_within(unsigned long pfn,
1266 struct page *page, struct zone *zone);
1267#else
1268static inline int memmap_valid_within(unsigned long pfn,
1269 struct page *page, struct zone *zone)
1270{
1271 return 1;
1272}
1273#endif /* CONFIG_ARCH_HAS_HOLES_MEMORYMODEL */
1274
Christoph Lameter97965472008-04-28 02:12:54 -07001275#endif /* !__GENERATING_BOUNDS.H */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276#endif /* !__ASSEMBLY__ */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001277#endif /* _LINUX_MMZONE_H */