blob: 754c25966a0a7828901deaf2c69dc1d5243736f4 [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
Mel Gormane58469b2014-06-04 16:10:16 -070078#define NR_MIGRATETYPE_BITS (PB_migrate_end - PB_migrate + 1)
79#define MIGRATETYPE_MASK ((1UL << NR_MIGRATETYPE_BITS) - 1)
80
Mel Gormandc4b0ca2014-06-04 16:10:17 -070081#define get_pageblock_migratetype(page) \
82 get_pfnblock_flags_mask(page, page_to_pfn(page), \
83 PB_migrate_end, MIGRATETYPE_MASK)
84
85static inline int get_pfnblock_migratetype(struct page *page, unsigned long pfn)
Mel Gorman467c9962007-10-16 01:26:02 -070086{
Mel Gormane58469b2014-06-04 16:10:16 -070087 BUILD_BUG_ON(PB_migrate_end - PB_migrate != 2);
Mel Gormandc4b0ca2014-06-04 16:10:17 -070088 return get_pfnblock_flags_mask(page, pfn, PB_migrate_end,
89 MIGRATETYPE_MASK);
Mel Gorman467c9962007-10-16 01:26:02 -070090}
91
Linus Torvalds1da177e2005-04-16 15:20:36 -070092struct free_area {
Mel Gormanb2a0ac82007-10-16 01:25:48 -070093 struct list_head free_list[MIGRATE_TYPES];
Linus Torvalds1da177e2005-04-16 15:20:36 -070094 unsigned long nr_free;
95};
96
97struct 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)
106struct zone_padding {
107 char x[0];
Ravikiran G Thirumalai22fc6ec2006-01-08 01:01:27 -0800108} ____cacheline_internodealigned_in_smp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700109#define ZONE_PADDING(name) struct zone_padding name;
110#else
111#define ZONE_PADDING(name)
112#endif
113
Christoph Lameter2244b952006-06-30 01:55:33 -0700114enum zone_stat_item {
Christoph Lameter51ed4492007-02-10 01:43:02 -0800115 /* First 128 byte cacheline (assuming 64 bit words) */
Christoph Lameterd23ad422007-02-10 01:43:02 -0800116 NR_FREE_PAGES,
Johannes Weiner81c0a2b2013-09-11 14:20:47 -0700117 NR_ALLOC_BATCH,
Christoph Lameterb69408e2008-10-18 20:26:14 -0700118 NR_LRU_BASE,
Rik van Riel4f98a2f2008-10-18 20:26:32 -0700119 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 Schermerhorn894bc312008-10-18 20:26:39 -0700123 NR_UNEVICTABLE, /* " " " " " */
Nick Piggin5344b7e2008-10-18 20:26:51 -0700124 NR_MLOCK, /* mlock()ed pages found and moved off LRU */
Christoph Lameterf3dbd342006-06-30 01:55:36 -0700125 NR_ANON_PAGES, /* Mapped anonymous pages */
126 NR_FILE_MAPPED, /* pagecache pages mapped into pagetables.
Christoph Lameter65ba55f2006-06-30 01:55:34 -0700127 only modified from process context */
Christoph Lameter347ce432006-06-30 01:55:35 -0700128 NR_FILE_PAGES,
Christoph Lameterb1e7a8f2006-06-30 01:55:39 -0700129 NR_FILE_DIRTY,
Christoph Lameterce866b32006-06-30 01:55:40 -0700130 NR_WRITEBACK,
Christoph Lameter51ed4492007-02-10 01:43:02 -0800131 NR_SLAB_RECLAIMABLE,
132 NR_SLAB_UNRECLAIMABLE,
133 NR_PAGETABLE, /* used for pagetables */
KOSAKI Motohiroc6a7f572009-09-21 17:01:32 -0700134 NR_KERNEL_STACK,
135 /* Second 128 byte cacheline */
Christoph Lameterfd39fc82006-06-30 01:55:40 -0700136 NR_UNSTABLE_NFS, /* NFS unstable pages */
Christoph Lameterd2c5e302006-06-30 01:55:41 -0700137 NR_BOUNCE,
Andrew Mortone129b5c2006-09-27 01:50:00 -0700138 NR_VMSCAN_WRITE,
Mel Gorman49ea7eb2011-10-31 17:07:59 -0700139 NR_VMSCAN_IMMEDIATE, /* Prioritise for reclaim when writeback ends */
Miklos Szeredifc3ba692008-04-30 00:54:38 -0700140 NR_WRITEBACK_TEMP, /* Writeback using temporary buffers */
KOSAKI Motohiroa7312862009-09-21 17:01:37 -0700141 NR_ISOLATED_ANON, /* Temporary isolated pages from anon lru */
142 NR_ISOLATED_FILE, /* Temporary isolated pages from file lru */
KOSAKI Motohiro4b021082009-09-21 17:01:33 -0700143 NR_SHMEM, /* shmem pages (included tmpfs/GEM pages) */
Michael Rubinea941f02010-10-26 14:21:35 -0700144 NR_DIRTIED, /* page dirtyings since bootup */
145 NR_WRITTEN, /* page writings since bootup */
Mel Gorman0d5d8232014-08-06 16:07:16 -0700146 NR_PAGES_SCANNED, /* pages scanned since last reclaim */
Christoph Lameterca889e62006-06-30 01:55:44 -0700147#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 Weinera5289102014-04-03 14:47:51 -0700155 WORKINGSET_REFAULT,
156 WORKINGSET_ACTIVATE,
Johannes Weiner449dd692014-04-03 14:47:56 -0700157 WORKINGSET_NODERECLAIM,
Andrea Arcangeli79134172011-01-13 15:46:58 -0800158 NR_ANON_TRANSPARENT_HUGEPAGES,
Bartlomiej Zolnierkiewiczd1ce7492012-10-08 16:32:02 -0700159 NR_FREE_CMA_PAGES,
Christoph Lameter2244b952006-06-30 01:55:33 -0700160 NR_VM_ZONE_STAT_ITEMS };
161
Rik van Riel4f98a2f2008-10-18 20:26:32 -0700162/*
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 Lameterb69408e2008-10-18 20:26:14 -0700175enum lru_list {
Rik van Riel4f98a2f2008-10-18 20:26:32 -0700176 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 Schermerhorn894bc312008-10-18 20:26:39 -0700180 LRU_UNEVICTABLE,
Lee Schermerhorn894bc312008-10-18 20:26:39 -0700181 NR_LRU_LISTS
182};
Christoph Lameterb69408e2008-10-18 20:26:14 -0700183
Hugh Dickins41113042012-01-12 17:20:01 -0800184#define for_each_lru(lru) for (lru = 0; lru < NR_LRU_LISTS; lru++)
Christoph Lameterb69408e2008-10-18 20:26:14 -0700185
Hugh Dickins41113042012-01-12 17:20:01 -0800186#define for_each_evictable_lru(lru) for (lru = 0; lru <= LRU_ACTIVE_FILE; lru++)
Lee Schermerhorn894bc312008-10-18 20:26:39 -0700187
Hugh Dickins41113042012-01-12 17:20:01 -0800188static inline int is_file_lru(enum lru_list lru)
Rik van Riel4f98a2f2008-10-18 20:26:32 -0700189{
Hugh Dickins41113042012-01-12 17:20:01 -0800190 return (lru == LRU_INACTIVE_FILE || lru == LRU_ACTIVE_FILE);
Rik van Riel4f98a2f2008-10-18 20:26:32 -0700191}
192
Hugh Dickins41113042012-01-12 17:20:01 -0800193static inline int is_active_lru(enum lru_list lru)
Christoph Lameterb69408e2008-10-18 20:26:14 -0700194{
Hugh Dickins41113042012-01-12 17:20:01 -0800195 return (lru == LRU_ACTIVE_ANON || lru == LRU_ACTIVE_FILE);
Christoph Lameterb69408e2008-10-18 20:26:14 -0700196}
197
Hugh Dickins41113042012-01-12 17:20:01 -0800198static inline int is_unevictable_lru(enum lru_list lru)
Lee Schermerhorn894bc312008-10-18 20:26:39 -0700199{
Hugh Dickins41113042012-01-12 17:20:01 -0800200 return (lru == LRU_UNEVICTABLE);
Lee Schermerhorn894bc312008-10-18 20:26:39 -0700201}
202
Hugh Dickins89abfab2012-05-29 15:06:53 -0700203struct zone_reclaim_stat {
204 /*
205 * The pageout code in vmscan.c keeps track of how many of the
Jiri Kosina59f91e52012-06-29 14:45:58 +0200206 * mem/swap backed and file backed pages are referenced.
Hugh Dickins89abfab2012-05-29 15:06:53 -0700207 * 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 Weiner6290df52012-01-12 17:18:10 -0800216struct lruvec {
217 struct list_head lists[NR_LRU_LISTS];
Hugh Dickins89abfab2012-05-29 15:06:53 -0700218 struct zone_reclaim_stat reclaim_stat;
Andrew Mortonc255a452012-07-31 16:43:02 -0700219#ifdef CONFIG_MEMCG
Konstantin Khlebnikov7f5e86c2012-05-29 15:06:58 -0700220 struct zone *zone;
221#endif
Johannes Weiner6290df52012-01-12 17:18:10 -0800222};
223
KAMEZAWA Hiroyukibb2a0de2011-07-26 16:08:22 -0700224/* 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 Hiroyukibb2a0de2011-07-26 16:08:22 -0700227#define LRU_ALL ((1 << NR_LRU_LISTS) - 1)
228
Minchan Kim39deaf82011-10-31 17:06:51 -0700229/* Isolate clean file */
Konstantin Khlebnikovf3fd4a62012-05-29 15:06:54 -0700230#define ISOLATE_CLEAN ((__force isolate_mode_t)0x1)
Minchan Kimf80c0672011-10-31 17:06:55 -0700231/* Isolate unmapped file */
Konstantin Khlebnikovf3fd4a62012-05-29 15:06:54 -0700232#define ISOLATE_UNMAPPED ((__force isolate_mode_t)0x2)
Mel Gormanc8244932012-01-12 17:19:38 -0800233/* Isolate for asynchronous migration */
Konstantin Khlebnikovf3fd4a62012-05-29 15:06:54 -0700234#define ISOLATE_ASYNC_MIGRATE ((__force isolate_mode_t)0x4)
Minchan Kime46a2872012-10-08 16:33:48 -0700235/* Isolate unevictable pages */
236#define ISOLATE_UNEVICTABLE ((__force isolate_mode_t)0x8)
Minchan Kim4356f212011-10-31 17:06:47 -0700237
238/* LRU Isolation modes. */
239typedef unsigned __bitwise__ isolate_mode_t;
240
Mel Gorman41858962009-06-16 15:32:12 -0700241enum 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 Torvalds1da177e2005-04-16 15:20:36 -0700252struct per_cpu_pages {
253 int count; /* number of pages in the list */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254 int high; /* high watermark, emptying needed */
255 int batch; /* chunk size for buddy add/remove */
Mel Gorman5f8dcc22009-09-21 17:03:19 -0700256
257 /* Lists of pages, one per migrate type stored on the pcp-lists */
258 struct list_head lists[MIGRATE_PCPTYPES];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700259};
260
261struct per_cpu_pageset {
Christoph Lameter3dfa5722008-02-04 22:29:19 -0800262 struct per_cpu_pages pcp;
Christoph Lameter4037d452007-05-09 02:35:14 -0700263#ifdef CONFIG_NUMA
264 s8 expire;
265#endif
Christoph Lameter2244b952006-06-30 01:55:33 -0700266#ifdef CONFIG_SMP
Christoph Lameterdf9ecab2006-08-31 21:27:35 -0700267 s8 stat_threshold;
Christoph Lameter2244b952006-06-30 01:55:33 -0700268 s8 vm_stat_diff[NR_VM_ZONE_STAT_ITEMS];
269#endif
Christoph Lameter99dcc3e2010-01-05 15:34:51 +0900270};
Christoph Lametere7c8d5c2005-06-21 17:14:47 -0700271
Christoph Lameter97965472008-04-28 02:12:54 -0700272#endif /* !__GENERATING_BOUNDS.H */
273
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700274enum zone_type {
Christoph Lameter4b51d662007-02-10 01:43:10 -0800275#ifdef CONFIG_ZONE_DMA
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700276 /*
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 Lameter2f1b6242006-09-25 23:31:13 -0700288 * arm Various
289 * alpha Unlimited or 0-16MB.
290 *
291 * i386, x86_64 and multiple other arches
292 * <16M.
293 */
294 ZONE_DMA,
Christoph Lameter4b51d662007-02-10 01:43:10 -0800295#endif
Christoph Lameterfb0e7942006-09-25 23:31:13 -0700296#ifdef CONFIG_ZONE_DMA32
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700297 /*
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 Lameterfb0e7942006-09-25 23:31:13 -0700303#endif
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700304 /*
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 Lametere53ef382006-09-25 23:31:14 -0700310#ifdef CONFIG_HIGHMEM
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700311 /*
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 Lametere53ef382006-09-25 23:31:14 -0700320#endif
Mel Gorman2a1e2742007-07-17 04:03:12 -0700321 ZONE_MOVABLE,
Christoph Lameter97965472008-04-28 02:12:54 -0700322 __MAX_NR_ZONES
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700323};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324
Christoph Lameter97965472008-04-28 02:12:54 -0700325#ifndef __GENERATING_BOUNDS_H
326
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327struct zone {
Mel Gorman3484b2d2014-08-06 16:07:14 -0700328 /* Read-mostly fields */
Mel Gorman41858962009-06-16 15:32:12 -0700329
330 /* zone watermarks, access with *_wmark_pages(zone) macros */
331 unsigned long watermark[NR_WMARK];
332
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333 /*
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 Gorman3484b2d2014-08-06 16:07:14 -0700341 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 Torvalds1da177e2005-04-16 15:20:36 -0700355
Johannes Weinerab8fabd2012-01-10 15:07:42 -0800356 /*
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 Gorman835c1342007-10-16 01:25:47 -0700362#ifndef CONFIG_SPARSEMEM
363 /*
Mel Gormand9c23402007-10-16 01:26:01 -0700364 * Flags for a pageblock_nr_pages block. See pageblock-flags.h.
Mel Gorman835c1342007-10-16 01:25:47 -0700365 * In SPARSEMEM, this map is stored in struct mem_section
366 */
367 unsigned long *pageblock_flags;
368#endif /* CONFIG_SPARSEMEM */
369
Mel Gorman3484b2d2014-08-06 16:07:14 -0700370#ifdef CONFIG_NUMA
Mel Gorman4f92e252010-05-24 14:32:32 -0700371 /*
Mel Gorman3484b2d2014-08-06 16:07:14 -0700372 * zone reclaim becomes active if more unmapped pages exist.
Mel Gorman4f92e252010-05-24 14:32:32 -0700373 */
Mel Gorman3484b2d2014-08-06 16:07:14 -0700374 unsigned long min_unmapped_pages;
375 unsigned long min_slab_pages;
376#endif /* CONFIG_NUMA */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700377
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378 /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */
379 unsigned long zone_start_pfn;
380
Dave Hansenbdc8cb92005-10-29 18:16:53 -0700381 /*
Jiang Liu9feedc92012-12-12 13:52:12 -0800382 * 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 Hansenbdc8cb92005-10-29 18:16:53 -0700385 *
Jiang Liu9feedc92012-12-12 13:52:12 -0800386 * present_pages is physical pages existing within the zone, which
387 * is calculated as:
Cody P Schafer8761e312013-03-26 10:30:44 -0700388 * present_pages = spanned_pages - absent_pages(pages in holes);
Jiang Liu9feedc92012-12-12 13:52:12 -0800389 *
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 Hansenbdc8cb92005-10-29 18:16:53 -0700409 * frequently read in proximity to zone->lock. It's good to
410 * give them a chance of being in the same cacheline.
Jiang Liu9feedc92012-12-12 13:52:12 -0800411 *
Jiang Liuc3d5f5f2013-07-03 15:03:14 -0700412 * Write access to present_pages at runtime should be protected by
Vladimir Davydovbfc8c902014-06-04 16:07:18 -0700413 * 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 Liuc3d5f5f2013-07-03 15:03:14 -0700415 *
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 Hansenbdc8cb92005-10-29 18:16:53 -0700421 */
Mel Gorman3484b2d2014-08-06 16:07:14 -0700422 unsigned long managed_pages;
Jiang Liu9feedc92012-12-12 13:52:12 -0800423 unsigned long spanned_pages;
424 unsigned long present_pages;
Mel Gorman3484b2d2014-08-06 16:07:14 -0700425
426 const char *name;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427
428 /*
Baoquan He44628d92015-02-11 15:25:10 -0800429 * Number of MIGRATE_RESERVE page block. To maintain for just
Yasuaki Ishimatsu943dca12014-01-21 15:49:06 -0800430 * optimization. Protected by zone->lock.
431 */
432 int nr_migrate_reserve_block;
433
Joonsoo Kimad53f922014-11-13 15:19:11 -0800434#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 Gorman3484b2d2014-08-06 16:07:14 -0700443#ifdef CONFIG_MEMORY_HOTPLUG
444 /* see spanned/present_pages for more description */
445 seqlock_t span_seqlock;
446#endif
447
Yasuaki Ishimatsu943dca12014-01-21 15:49:06 -0800448 /*
Mel Gorman3484b2d2014-08-06 16:07:14 -0700449 * 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 Torvalds1da177e2005-04-16 15:20:36 -0700471 */
Mel Gorman3484b2d2014-08-06 16:07:14 -0700472 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_)
Mel Gorman3484b2d2014-08-06 16:07:14 -0700477 /* free areas of different sizes */
478 struct free_area free_area[MAX_ORDER];
479
480 /* zone flags, see below */
481 unsigned long flags;
482
Mel Gormana368ab62015-04-07 14:26:41 -0700483 /* Write-intensive fields used from the page allocator */
484 spinlock_t lock;
485
Mel Gorman3484b2d2014-08-06 16:07:14 -0700486 ZONE_PADDING(_pad2_)
487
488 /* Write-intensive fields used by page reclaim */
489
490 /* Fields commonly accessed by the page reclaim scanner */
491 spinlock_t lru_lock;
Mel Gorman3484b2d2014-08-06 16:07:14 -0700492 struct lruvec lruvec;
493
494 /* Evictions & activations on the inactive file list */
495 atomic_long_t inactive_age;
496
497 /*
498 * When free pages are below this point, additional steps are taken
499 * when reading the number of free pages to avoid per-cpu counter
500 * drift allowing watermarks to be breached
501 */
502 unsigned long percpu_drift_mark;
503
504#if defined CONFIG_COMPACTION || defined CONFIG_CMA
505 /* pfn where compaction free scanner should start */
506 unsigned long compact_cached_free_pfn;
507 /* pfn where async and sync compaction migration scanner should start */
508 unsigned long compact_cached_migrate_pfn[2];
509#endif
510
511#ifdef CONFIG_COMPACTION
512 /*
513 * On compaction failure, 1<<compact_defer_shift compactions
514 * are skipped before trying again. The number attempted since
515 * last failure is tracked with compact_considered.
516 */
517 unsigned int compact_considered;
518 unsigned int compact_defer_shift;
519 int compact_order_failed;
520#endif
521
522#if defined CONFIG_COMPACTION || defined CONFIG_CMA
523 /* Set to true when the PG_migrate_skip bits should be cleared */
524 bool compact_blockskip_flush;
525#endif
526
527 ZONE_PADDING(_pad3_)
528 /* Zone statistics */
529 atomic_long_t vm_stat[NR_VM_ZONE_STAT_ITEMS];
Ravikiran G Thirumalai22fc6ec2006-01-08 01:01:27 -0800530} ____cacheline_internodealigned_in_smp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531
Johannes Weiner57054652014-10-09 15:28:17 -0700532enum zone_flags {
David Rientjese815af92007-10-16 23:25:54 -0700533 ZONE_RECLAIM_LOCKED, /* prevents concurrent reclaim */
David Rientjes098d7f12007-10-16 23:25:55 -0700534 ZONE_OOM_LOCKED, /* zone is in OOM killer zonelist */
Mel Gorman0e093d992010-10-26 14:21:45 -0700535 ZONE_CONGESTED, /* zone has many dirty pages backed by
536 * a congested BDI
537 */
Johannes Weiner57054652014-10-09 15:28:17 -0700538 ZONE_DIRTY, /* reclaim scanning has recently found
Mel Gormand43006d2013-07-03 15:01:50 -0700539 * many dirty file pages at the tail
540 * of the LRU.
541 */
Mel Gorman283aba92013-07-03 15:01:51 -0700542 ZONE_WRITEBACK, /* reclaim scanning has recently found
543 * many pages under writeback
544 */
Mel Gorman4ffeaf32014-08-06 16:07:22 -0700545 ZONE_FAIR_DEPLETED, /* fair zone policy batch depleted */
Johannes Weiner57054652014-10-09 15:28:17 -0700546};
David Rientjese815af92007-10-16 23:25:54 -0700547
Russ Andersonf9228b22013-03-22 15:04:43 -0700548static inline unsigned long zone_end_pfn(const struct zone *zone)
Cody P Schafer108bcc92013-02-22 16:35:23 -0800549{
550 return zone->zone_start_pfn + zone->spanned_pages;
551}
552
553static inline bool zone_spans_pfn(const struct zone *zone, unsigned long pfn)
554{
555 return zone->zone_start_pfn <= pfn && pfn < zone_end_pfn(zone);
556}
557
Cody P Schafer2a6e3eb2013-02-22 16:35:24 -0800558static inline bool zone_is_initialized(struct zone *zone)
559{
560 return !!zone->wait_table;
561}
562
563static inline bool zone_is_empty(struct zone *zone)
564{
565 return zone->spanned_pages == 0;
566}
567
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568/*
569 * The "priority" of VM scanning is how much of the queues we will scan in one
570 * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the
571 * queues ("queue_length >> 12") during an aging round.
572 */
573#define DEF_PRIORITY 12
574
Paul Jackson9276b1bc2006-12-06 20:31:48 -0800575/* Maximum number of zones on a zonelist */
576#define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES)
577
578#ifdef CONFIG_NUMA
Christoph Lameter523b9452007-10-16 01:25:37 -0700579
580/*
Pete Zaitcev25a64ec2011-02-03 22:43:48 -0700581 * The NUMA zonelists are doubled because we need zonelists that restrict the
Johannes Weinere97ca8e2014-03-10 15:49:43 -0700582 * allocations to a single node for __GFP_THISNODE.
Christoph Lameter523b9452007-10-16 01:25:37 -0700583 *
Mel Gorman54a6eb52008-04-28 02:12:16 -0700584 * [0] : Zonelist with fallback
Johannes Weinere97ca8e2014-03-10 15:49:43 -0700585 * [1] : No fallback (__GFP_THISNODE)
Christoph Lameter523b9452007-10-16 01:25:37 -0700586 */
Mel Gorman54a6eb52008-04-28 02:12:16 -0700587#define MAX_ZONELISTS 2
Christoph Lameter523b9452007-10-16 01:25:37 -0700588
589
Paul Jackson9276b1bc2006-12-06 20:31:48 -0800590/*
591 * We cache key information from each zonelist for smaller cache
592 * footprint when scanning for free pages in get_page_from_freelist().
593 *
594 * 1) The BITMAP fullzones tracks which zones in a zonelist have come
595 * up short of free memory since the last time (last_fullzone_zap)
596 * we zero'd fullzones.
597 * 2) The array z_to_n[] maps each zone in the zonelist to its node
598 * id, so that we can efficiently evaluate whether that node is
599 * set in the current tasks mems_allowed.
600 *
601 * Both fullzones and z_to_n[] are one-to-one with the zonelist,
602 * indexed by a zones offset in the zonelist zones[] array.
603 *
604 * The get_page_from_freelist() routine does two scans. During the
605 * first scan, we skip zones whose corresponding bit in 'fullzones'
606 * is set or whose corresponding node in current->mems_allowed (which
607 * comes from cpusets) is not set. During the second scan, we bypass
608 * this zonelist_cache, to ensure we look methodically at each zone.
609 *
610 * Once per second, we zero out (zap) fullzones, forcing us to
611 * reconsider nodes that might have regained more free memory.
612 * The field last_full_zap is the time we last zapped fullzones.
613 *
614 * This mechanism reduces the amount of time we waste repeatedly
615 * reexaming zones for free memory when they just came up low on
616 * memory momentarilly ago.
617 *
618 * The zonelist_cache struct members logically belong in struct
619 * zonelist. However, the mempolicy zonelists constructed for
620 * MPOL_BIND are intentionally variable length (and usually much
621 * shorter). A general purpose mechanism for handling structs with
622 * multiple variable length members is more mechanism than we want
623 * here. We resort to some special case hackery instead.
624 *
625 * The MPOL_BIND zonelists don't need this zonelist_cache (in good
626 * part because they are shorter), so we put the fixed length stuff
627 * at the front of the zonelist struct, ending in a variable length
628 * zones[], as is needed by MPOL_BIND.
629 *
630 * Then we put the optional zonelist cache on the end of the zonelist
631 * struct. This optional stuff is found by a 'zlcache_ptr' pointer in
632 * the fixed length portion at the front of the struct. This pointer
633 * both enables us to find the zonelist cache, and in the case of
634 * MPOL_BIND zonelists, (which will just set the zlcache_ptr to NULL)
635 * to know that the zonelist cache is not there.
636 *
637 * The end result is that struct zonelists come in two flavors:
638 * 1) The full, fixed length version, shown below, and
639 * 2) The custom zonelists for MPOL_BIND.
640 * The custom MPOL_BIND zonelists have a NULL zlcache_ptr and no zlcache.
641 *
642 * Even though there may be multiple CPU cores on a node modifying
643 * fullzones or last_full_zap in the same zonelist_cache at the same
644 * time, we don't lock it. This is just hint data - if it is wrong now
645 * and then, the allocator will still function, perhaps a bit slower.
646 */
647
648
649struct zonelist_cache {
Paul Jackson9276b1bc2006-12-06 20:31:48 -0800650 unsigned short z_to_n[MAX_ZONES_PER_ZONELIST]; /* zone->nid */
Paul Jackson7253f4e2006-12-06 20:31:49 -0800651 DECLARE_BITMAP(fullzones, MAX_ZONES_PER_ZONELIST); /* zone full? */
Paul Jackson9276b1bc2006-12-06 20:31:48 -0800652 unsigned long last_full_zap; /* when last zap'd (jiffies) */
653};
654#else
Mel Gorman54a6eb52008-04-28 02:12:16 -0700655#define MAX_ZONELISTS 1
Paul Jackson9276b1bc2006-12-06 20:31:48 -0800656struct zonelist_cache;
657#endif
658
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659/*
Mel Gormandd1a2392008-04-28 02:12:17 -0700660 * This struct contains information about a zone in a zonelist. It is stored
661 * here to avoid dereferences into large structures and lookups of tables
662 */
663struct zoneref {
664 struct zone *zone; /* Pointer to actual zone */
665 int zone_idx; /* zone_idx(zoneref->zone) */
666};
667
668/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669 * One allocation request operates on a zonelist. A zonelist
670 * is a list of zones, the first one is the 'goal' of the
671 * allocation, the other zones are fallback zones, in decreasing
672 * priority.
673 *
Paul Jackson9276b1bc2006-12-06 20:31:48 -0800674 * If zlcache_ptr is not NULL, then it is just the address of zlcache,
675 * as explained above. If zlcache_ptr is NULL, there is no zlcache.
Mel Gormandd1a2392008-04-28 02:12:17 -0700676 * *
677 * To speed the reading of the zonelist, the zonerefs contain the zone index
678 * of the entry being read. Helper functions to access information given
679 * a struct zoneref are
680 *
681 * zonelist_zone() - Return the struct zone * for an entry in _zonerefs
682 * zonelist_zone_idx() - Return the index of the zone for an entry
683 * zonelist_node_idx() - Return the index of the node for an entry
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684 */
685struct zonelist {
Paul Jackson9276b1bc2006-12-06 20:31:48 -0800686 struct zonelist_cache *zlcache_ptr; // NULL or &zlcache
Mel Gormandd1a2392008-04-28 02:12:17 -0700687 struct zoneref _zonerefs[MAX_ZONES_PER_ZONELIST + 1];
Paul Jackson9276b1bc2006-12-06 20:31:48 -0800688#ifdef CONFIG_NUMA
689 struct zonelist_cache zlcache; // optional ...
690#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700691};
692
Tejun Heo0ee332c2011-12-08 10:22:09 -0800693#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
Mel Gormanc7132162006-09-27 01:49:43 -0700694struct node_active_region {
695 unsigned long start_pfn;
696 unsigned long end_pfn;
697 int nid;
698};
Tejun Heo0ee332c2011-12-08 10:22:09 -0800699#endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700700
Heiko Carstens5b99cd02006-09-27 01:50:01 -0700701#ifndef CONFIG_DISCONTIGMEM
702/* The array of struct pages - for discontigmem use pgdat->lmem_map */
703extern struct page *mem_map;
704#endif
705
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706/*
707 * The pg_data_t structure is used in machines with CONFIG_DISCONTIGMEM
708 * (mostly NUMA machines?) to denote a higher-level memory zone than the
709 * zone denotes.
710 *
711 * On NUMA machines, each NUMA node would have a pg_data_t to describe
712 * it's memory layout.
713 *
714 * Memory statistics and page replacement data structures are maintained on a
715 * per-zone basis.
716 */
717struct bootmem_data;
718typedef struct pglist_data {
719 struct zone node_zones[MAX_NR_ZONES];
Christoph Lameter523b9452007-10-16 01:25:37 -0700720 struct zonelist node_zonelists[MAX_ZONELISTS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721 int nr_zones;
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -0700722#ifdef CONFIG_FLAT_NODE_MEM_MAP /* means !SPARSEMEM */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723 struct page *node_mem_map;
Joonsoo Kimeefa8642014-12-12 16:55:46 -0800724#ifdef CONFIG_PAGE_EXTENSION
725 struct page_ext *node_page_ext;
726#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700727#endif
Yinghai Lu08677212010-02-10 01:20:20 -0800728#ifndef CONFIG_NO_BOOTMEM
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729 struct bootmem_data *bdata;
Yinghai Lu08677212010-02-10 01:20:20 -0800730#endif
Dave Hansen208d54e2005-10-29 18:16:52 -0700731#ifdef CONFIG_MEMORY_HOTPLUG
732 /*
733 * Must be held any time you expect node_start_pfn, node_present_pages
734 * or node_spanned_pages stay constant. Holding this will also
735 * guarantee that any pfn_valid() stays that way.
736 *
Cody P Schafer114d4b72013-07-03 15:02:09 -0700737 * pgdat_resize_lock() and pgdat_resize_unlock() are provided to
738 * manipulate node_size_lock without checking for CONFIG_MEMORY_HOTPLUG.
739 *
Cody P Schafer72c3b512013-07-03 15:02:08 -0700740 * Nests above zone->lock and zone->span_seqlock
Dave Hansen208d54e2005-10-29 18:16:52 -0700741 */
742 spinlock_t node_size_lock;
743#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744 unsigned long node_start_pfn;
745 unsigned long node_present_pages; /* total number of physical pages */
746 unsigned long node_spanned_pages; /* total size of physical page
747 range, including holes */
748 int node_id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749 wait_queue_head_t kswapd_wait;
Mel Gorman55150612012-07-31 16:44:35 -0700750 wait_queue_head_t pfmemalloc_wait;
Vladimir Davydovbfc8c902014-06-04 16:07:18 -0700751 struct task_struct *kswapd; /* Protected by
752 mem_hotplug_begin/end() */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753 int kswapd_max_order;
Mel Gorman99504742011-01-13 15:46:20 -0800754 enum zone_type classzone_idx;
Andrea Arcangeli8177a422012-03-23 20:56:34 +0100755#ifdef CONFIG_NUMA_BALANCING
Mel Gorman1c5e9c22014-01-21 15:50:59 -0800756 /* Lock serializing the migrate rate limiting window */
Andrea Arcangeli8177a422012-03-23 20:56:34 +0100757 spinlock_t numabalancing_migrate_lock;
758
759 /* Rate limiting time interval */
760 unsigned long numabalancing_migrate_next_window;
761
762 /* Number of pages migrated during the rate limiting time interval */
763 unsigned long numabalancing_migrate_nr_pages;
764#endif
Mel Gorman3a80a7f2015-06-30 14:57:02 -0700765
766#ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT
767 /*
768 * If memory initialisation on large machines is deferred then this
769 * is the first PFN that needs to be initialised.
770 */
771 unsigned long first_deferred_pfn;
772#endif /* CONFIG_DEFERRED_STRUCT_PAGE_INIT */
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 Gorman7aeb09f2014-06-04 16:10:21 -0700802bool zone_watermark_ok(struct zone *z, unsigned int order,
803 unsigned long mark, int classzone_idx, int alloc_flags);
804bool zone_watermark_ok_safe(struct zone *z, unsigned int order,
805 unsigned long mark, 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
Zhang Zhend7e4a2e2015-04-15 16:12:57 -0700853#ifdef CONFIG_HIGHMEM
Mel Gorman2a1e2742007-07-17 04:03:12 -0700854static inline int zone_movable_is_highmem(void)
855{
Zhang Zhend7e4a2e2015-04-15 16:12:57 -0700856#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
Mel Gorman2a1e2742007-07-17 04:03:12 -0700857 return movable_zone == ZONE_HIGHMEM;
858#else
Zhang Zhend7e4a2e2015-04-15 16:12:57 -0700859 return (ZONE_MOVABLE - 1) == ZONE_HIGHMEM;
Mel Gorman2a1e2742007-07-17 04:03:12 -0700860#endif
861}
Zhang Zhend7e4a2e2015-04-15 16:12:57 -0700862#endif
Mel Gorman2a1e2742007-07-17 04:03:12 -0700863
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700864static inline int is_highmem_idx(enum zone_type idx)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865{
Christoph Lametere53ef382006-09-25 23:31:14 -0700866#ifdef CONFIG_HIGHMEM
Mel Gorman2a1e2742007-07-17 04:03:12 -0700867 return (idx == ZONE_HIGHMEM ||
868 (idx == ZONE_MOVABLE && zone_movable_is_highmem()));
Christoph Lametere53ef382006-09-25 23:31:14 -0700869#else
870 return 0;
871#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872}
873
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874/**
875 * is_highmem - helper function to quickly check if a struct zone is a
876 * highmem zone or not. This is an attempt to keep references
877 * to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum.
878 * @zone - pointer to struct zone variable
879 */
880static inline int is_highmem(struct zone *zone)
881{
Christoph Lametere53ef382006-09-25 23:31:14 -0700882#ifdef CONFIG_HIGHMEM
Harvey Harrisonddc81ed2008-04-28 02:12:07 -0700883 int zone_off = (char *)zone - (char *)zone->zone_pgdat->node_zones;
884 return zone_off == ZONE_HIGHMEM * sizeof(*zone) ||
885 (zone_off == ZONE_MOVABLE * sizeof(*zone) &&
886 zone_movable_is_highmem());
Christoph Lametere53ef382006-09-25 23:31:14 -0700887#else
888 return 0;
889#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890}
891
Linus Torvalds1da177e2005-04-16 15:20:36 -0700892/* These two functions are used to setup the per zone pages min values */
893struct ctl_table;
Alexey Dobriyan8d65af72009-09-23 15:57:19 -0700894int min_free_kbytes_sysctl_handler(struct ctl_table *, int,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895 void __user *, size_t *, loff_t *);
896extern int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1];
Alexey Dobriyan8d65af72009-09-23 15:57:19 -0700897int lowmem_reserve_ratio_sysctl_handler(struct ctl_table *, int,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898 void __user *, size_t *, loff_t *);
Alexey Dobriyan8d65af72009-09-23 15:57:19 -0700899int percpu_pagelist_fraction_sysctl_handler(struct ctl_table *, int,
Rohit Seth8ad4b1f2006-01-08 01:00:40 -0800900 void __user *, size_t *, loff_t *);
Christoph Lameter96146342006-07-03 00:24:13 -0700901int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table *, int,
Alexey Dobriyan8d65af72009-09-23 15:57:19 -0700902 void __user *, size_t *, loff_t *);
Christoph Lameter0ff38492006-09-25 23:31:52 -0700903int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table *, int,
Alexey Dobriyan8d65af72009-09-23 15:57:19 -0700904 void __user *, size_t *, loff_t *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905
KAMEZAWA Hiroyukif0c0b2b2007-07-15 23:38:01 -0700906extern int numa_zonelist_order_handler(struct ctl_table *, int,
Alexey Dobriyan8d65af72009-09-23 15:57:19 -0700907 void __user *, size_t *, loff_t *);
KAMEZAWA Hiroyukif0c0b2b2007-07-15 23:38:01 -0700908extern char numa_zonelist_order[];
909#define NUMA_ZONELIST_ORDER_LEN 16 /* string buffer size */
910
Dave Hansen93b75042005-06-23 00:07:47 -0700911#ifndef CONFIG_NEED_MULTIPLE_NODES
Linus Torvalds1da177e2005-04-16 15:20:36 -0700912
913extern struct pglist_data contig_page_data;
914#define NODE_DATA(nid) (&contig_page_data)
915#define NODE_MEM_MAP(nid) mem_map
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916
Dave Hansen93b75042005-06-23 00:07:47 -0700917#else /* CONFIG_NEED_MULTIPLE_NODES */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700918
919#include <asm/mmzone.h>
920
Dave Hansen93b75042005-06-23 00:07:47 -0700921#endif /* !CONFIG_NEED_MULTIPLE_NODES */
Dave Hansen348f8b62005-06-23 00:07:40 -0700922
KAMEZAWA Hiroyuki95144c72006-03-27 01:16:02 -0800923extern struct pglist_data *first_online_pgdat(void);
924extern struct pglist_data *next_online_pgdat(struct pglist_data *pgdat);
925extern struct zone *next_zone(struct zone *zone);
KAMEZAWA Hiroyuki8357f862006-03-27 01:15:57 -0800926
927/**
Fernando Luis Vazquez Cao12d15f02008-05-23 13:05:01 -0700928 * for_each_online_pgdat - helper macro to iterate over all online nodes
KAMEZAWA Hiroyuki8357f862006-03-27 01:15:57 -0800929 * @pgdat - pointer to a pg_data_t variable
930 */
931#define for_each_online_pgdat(pgdat) \
932 for (pgdat = first_online_pgdat(); \
933 pgdat; \
934 pgdat = next_online_pgdat(pgdat))
KAMEZAWA Hiroyuki8357f862006-03-27 01:15:57 -0800935/**
936 * for_each_zone - helper macro to iterate over all memory zones
937 * @zone - pointer to struct zone variable
938 *
939 * The user only needs to declare the zone variable, for_each_zone
940 * fills it in.
941 */
942#define for_each_zone(zone) \
943 for (zone = (first_online_pgdat())->node_zones; \
944 zone; \
945 zone = next_zone(zone))
946
KOSAKI Motohiroee99c712009-03-31 15:19:31 -0700947#define for_each_populated_zone(zone) \
948 for (zone = (first_online_pgdat())->node_zones; \
949 zone; \
950 zone = next_zone(zone)) \
951 if (!populated_zone(zone)) \
952 ; /* do nothing */ \
953 else
954
Mel Gormandd1a2392008-04-28 02:12:17 -0700955static inline struct zone *zonelist_zone(struct zoneref *zoneref)
956{
957 return zoneref->zone;
958}
959
960static inline int zonelist_zone_idx(struct zoneref *zoneref)
961{
962 return zoneref->zone_idx;
963}
964
965static inline int zonelist_node_idx(struct zoneref *zoneref)
966{
967#ifdef CONFIG_NUMA
968 /* zone_to_nid not available in this context */
969 return zoneref->zone->node;
970#else
971 return 0;
972#endif /* CONFIG_NUMA */
973}
974
Mel Gorman19770b32008-04-28 02:12:18 -0700975/**
976 * 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
977 * @z - The cursor used as a starting point for the search
978 * @highest_zoneidx - The zone index of the highest zone to return
979 * @nodes - An optional nodemask to filter the zonelist with
Mel Gorman19770b32008-04-28 02:12:18 -0700980 *
981 * This function returns the next zone at or below a given zone index that is
982 * within the allowed nodemask using a cursor as the starting point for the
Mel Gorman5bead2a2008-09-13 02:33:19 -0700983 * search. The zoneref returned is a cursor that represents the current zone
984 * being examined. It should be advanced by one before calling
985 * next_zones_zonelist again.
Mel Gorman19770b32008-04-28 02:12:18 -0700986 */
987struct zoneref *next_zones_zonelist(struct zoneref *z,
988 enum zone_type highest_zoneidx,
Vlastimil Babka05891fb2015-02-11 15:25:47 -0800989 nodemask_t *nodes);
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{
Vlastimil Babka05891fb2015-02-11 15:25:47 -08001008 struct zoneref *z = next_zones_zonelist(zonelist->_zonerefs,
1009 highest_zoneidx, nodes);
1010 *zone = zonelist_zone(z);
1011 return z;
Mel Gorman54a6eb52008-04-28 02:12:16 -07001012}
1013
Mel Gorman19770b32008-04-28 02:12:18 -07001014/**
1015 * 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
1016 * @zone - The current zone in the iterator
1017 * @z - The current pointer within zonelist->zones being iterated
1018 * @zlist - The zonelist being iterated
1019 * @highidx - The zone index of the highest zone to return
1020 * @nodemask - Nodemask allowed by the allocator
1021 *
1022 * This iterator iterates though all zones at or below a given zone index and
1023 * within a given nodemask
1024 */
1025#define for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \
1026 for (z = first_zones_zonelist(zlist, highidx, nodemask, &zone); \
1027 zone; \
Vlastimil Babka05891fb2015-02-11 15:25:47 -08001028 z = next_zones_zonelist(++z, highidx, nodemask), \
1029 zone = zonelist_zone(z)) \
Mel Gorman54a6eb52008-04-28 02:12:16 -07001030
1031/**
1032 * for_each_zone_zonelist - helper macro to iterate over valid zones in a zonelist at or below a given zone index
1033 * @zone - The current zone in the iterator
1034 * @z - The current pointer within zonelist->zones being iterated
1035 * @zlist - The zonelist being iterated
1036 * @highidx - The zone index of the highest zone to return
1037 *
1038 * This iterator iterates though all zones at or below a given zone index.
1039 */
1040#define for_each_zone_zonelist(zone, z, zlist, highidx) \
Mel Gorman19770b32008-04-28 02:12:18 -07001041 for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, NULL)
Mel Gorman54a6eb52008-04-28 02:12:16 -07001042
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001043#ifdef CONFIG_SPARSEMEM
1044#include <asm/sparsemem.h>
1045#endif
1046
Mel Gormanc7132162006-09-27 01:49:43 -07001047#if !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) && \
Tejun Heo0ee332c2011-12-08 10:22:09 -08001048 !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP)
Andrew Mortonb4544562008-04-28 02:12:39 -07001049static inline unsigned long early_pfn_to_nid(unsigned long pfn)
1050{
1051 return 0;
1052}
Andy Whitcroftb159d432005-06-23 00:07:52 -07001053#endif
1054
Andy Whitcroft2bdaf112006-01-06 00:10:53 -08001055#ifdef CONFIG_FLATMEM
1056#define pfn_to_nid(pfn) (0)
1057#endif
1058
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001059#ifdef CONFIG_SPARSEMEM
1060
1061/*
1062 * SECTION_SHIFT #bits space required to store a section #
1063 *
1064 * PA_SECTION_SHIFT physical address to/from section number
1065 * PFN_SECTION_SHIFT pfn to/from section number
1066 */
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001067#define PA_SECTION_SHIFT (SECTION_SIZE_BITS)
1068#define PFN_SECTION_SHIFT (SECTION_SIZE_BITS - PAGE_SHIFT)
1069
1070#define NR_MEM_SECTIONS (1UL << SECTIONS_SHIFT)
1071
1072#define PAGES_PER_SECTION (1UL << PFN_SECTION_SHIFT)
1073#define PAGE_SECTION_MASK (~(PAGES_PER_SECTION-1))
1074
Mel Gorman835c1342007-10-16 01:25:47 -07001075#define SECTION_BLOCKFLAGS_BITS \
Mel Gormand9c23402007-10-16 01:26:01 -07001076 ((1UL << (PFN_SECTION_SHIFT - pageblock_order)) * NR_PAGEBLOCK_BITS)
Mel Gorman835c1342007-10-16 01:25:47 -07001077
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001078#if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
1079#error Allocator MAX_ORDER exceeds SECTION_SIZE
1080#endif
1081
Daniel Kipere3c40f32011-05-24 17:12:33 -07001082#define pfn_to_section_nr(pfn) ((pfn) >> PFN_SECTION_SHIFT)
1083#define section_nr_to_pfn(sec) ((sec) << PFN_SECTION_SHIFT)
1084
Daniel Kipera539f352011-05-24 17:12:51 -07001085#define SECTION_ALIGN_UP(pfn) (((pfn) + PAGES_PER_SECTION - 1) & PAGE_SECTION_MASK)
1086#define SECTION_ALIGN_DOWN(pfn) ((pfn) & PAGE_SECTION_MASK)
1087
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001088struct page;
Joonsoo Kimeefa8642014-12-12 16:55:46 -08001089struct page_ext;
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001090struct mem_section {
Andy Whitcroft29751f62005-06-23 00:08:00 -07001091 /*
1092 * This is, logically, a pointer to an array of struct
1093 * pages. However, it is stored with some other magic.
1094 * (see sparse.c::sparse_init_one_section())
1095 *
Andy Whitcroft30c253e2006-06-23 02:03:41 -07001096 * Additionally during early boot we encode node id of
1097 * the location of the section here to guide allocation.
1098 * (see sparse.c::memory_present())
1099 *
Andy Whitcroft29751f62005-06-23 00:08:00 -07001100 * Making it a UL at least makes someone do a cast
1101 * before using it wrong.
1102 */
1103 unsigned long section_mem_map;
Mel Gorman5c0e3062007-10-16 01:25:56 -07001104
1105 /* See declaration of similar field in struct zone */
1106 unsigned long *pageblock_flags;
Joonsoo Kimeefa8642014-12-12 16:55:46 -08001107#ifdef CONFIG_PAGE_EXTENSION
1108 /*
1109 * If !SPARSEMEM, pgdat doesn't have page_ext pointer. We use
1110 * section. (see page_ext.h about this.)
1111 */
1112 struct page_ext *page_ext;
1113 unsigned long pad;
1114#endif
Cody P Schafer55878e82013-07-03 15:04:44 -07001115 /*
1116 * WARNING: mem_section must be a power-of-2 in size for the
1117 * calculation and use of SECTION_ROOT_MASK to make sense.
1118 */
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001119};
1120
Bob Picco3e347262005-09-03 15:54:28 -07001121#ifdef CONFIG_SPARSEMEM_EXTREME
1122#define SECTIONS_PER_ROOT (PAGE_SIZE / sizeof (struct mem_section))
Bob Picco802f1922005-09-03 15:54:26 -07001123#else
Bob Picco3e347262005-09-03 15:54:28 -07001124#define SECTIONS_PER_ROOT 1
1125#endif
Bob Picco802f1922005-09-03 15:54:26 -07001126
Bob Picco3e347262005-09-03 15:54:28 -07001127#define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT)
Marcelo Roberto Jimenez0faa5632010-05-24 14:32:47 -07001128#define NR_SECTION_ROOTS DIV_ROUND_UP(NR_MEM_SECTIONS, SECTIONS_PER_ROOT)
Bob Picco3e347262005-09-03 15:54:28 -07001129#define SECTION_ROOT_MASK (SECTIONS_PER_ROOT - 1)
1130
1131#ifdef CONFIG_SPARSEMEM_EXTREME
1132extern struct mem_section *mem_section[NR_SECTION_ROOTS];
1133#else
1134extern struct mem_section mem_section[NR_SECTION_ROOTS][SECTIONS_PER_ROOT];
1135#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001136
Andy Whitcroft29751f62005-06-23 00:08:00 -07001137static inline struct mem_section *__nr_to_section(unsigned long nr)
1138{
Bob Picco3e347262005-09-03 15:54:28 -07001139 if (!mem_section[SECTION_NR_TO_ROOT(nr)])
1140 return NULL;
1141 return &mem_section[SECTION_NR_TO_ROOT(nr)][nr & SECTION_ROOT_MASK];
Andy Whitcroft29751f62005-06-23 00:08:00 -07001142}
Dave Hansen4ca644d2005-10-29 18:16:51 -07001143extern int __section_nr(struct mem_section* ms);
Yasunori Goto04753272008-04-28 02:13:31 -07001144extern unsigned long usemap_size(void);
Andy Whitcroft29751f62005-06-23 00:08:00 -07001145
1146/*
1147 * We use the lower bits of the mem_map pointer to store
1148 * a little bit of information. There should be at least
1149 * 3 bits here due to 32-bit alignment.
1150 */
1151#define SECTION_MARKED_PRESENT (1UL<<0)
1152#define SECTION_HAS_MEM_MAP (1UL<<1)
1153#define SECTION_MAP_LAST_BIT (1UL<<2)
1154#define SECTION_MAP_MASK (~(SECTION_MAP_LAST_BIT-1))
Andy Whitcroft30c253e2006-06-23 02:03:41 -07001155#define SECTION_NID_SHIFT 2
Andy Whitcroft29751f62005-06-23 00:08:00 -07001156
1157static inline struct page *__section_mem_map_addr(struct mem_section *section)
1158{
1159 unsigned long map = section->section_mem_map;
1160 map &= SECTION_MAP_MASK;
1161 return (struct page *)map;
1162}
1163
Andy Whitcroft540557b2007-10-16 01:24:11 -07001164static inline int present_section(struct mem_section *section)
Andy Whitcroft29751f62005-06-23 00:08:00 -07001165{
Bob Picco802f1922005-09-03 15:54:26 -07001166 return (section && (section->section_mem_map & SECTION_MARKED_PRESENT));
Andy Whitcroft29751f62005-06-23 00:08:00 -07001167}
1168
Andy Whitcroft540557b2007-10-16 01:24:11 -07001169static inline int present_section_nr(unsigned long nr)
1170{
1171 return present_section(__nr_to_section(nr));
1172}
1173
1174static inline int valid_section(struct mem_section *section)
Andy Whitcroft29751f62005-06-23 00:08:00 -07001175{
Bob Picco802f1922005-09-03 15:54:26 -07001176 return (section && (section->section_mem_map & SECTION_HAS_MEM_MAP));
Andy Whitcroft29751f62005-06-23 00:08:00 -07001177}
1178
1179static inline int valid_section_nr(unsigned long nr)
1180{
1181 return valid_section(__nr_to_section(nr));
1182}
1183
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001184static inline struct mem_section *__pfn_to_section(unsigned long pfn)
1185{
Andy Whitcroft29751f62005-06-23 00:08:00 -07001186 return __nr_to_section(pfn_to_section_nr(pfn));
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001187}
1188
Will Deacon7b7bf492011-05-19 13:21:14 +01001189#ifndef CONFIG_HAVE_ARCH_PFN_VALID
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001190static inline int pfn_valid(unsigned long pfn)
1191{
1192 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1193 return 0;
Andy Whitcroft29751f62005-06-23 00:08:00 -07001194 return valid_section(__nr_to_section(pfn_to_section_nr(pfn)));
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001195}
Will Deacon7b7bf492011-05-19 13:21:14 +01001196#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001197
Andy Whitcroft540557b2007-10-16 01:24:11 -07001198static inline int pfn_present(unsigned long pfn)
1199{
1200 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
1201 return 0;
1202 return present_section(__nr_to_section(pfn_to_section_nr(pfn)));
1203}
1204
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001205/*
1206 * These are _only_ used during initialisation, therefore they
1207 * can use __initdata ... They could have names to indicate
1208 * this restriction.
1209 */
1210#ifdef CONFIG_NUMA
Andy Whitcroft161599f2006-01-06 00:10:54 -08001211#define pfn_to_nid(pfn) \
1212({ \
1213 unsigned long __pfn_to_nid_pfn = (pfn); \
1214 page_to_nid(pfn_to_page(__pfn_to_nid_pfn)); \
1215})
Andy Whitcroft2bdaf112006-01-06 00:10:53 -08001216#else
1217#define pfn_to_nid(pfn) (0)
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001218#endif
1219
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001220#define early_pfn_valid(pfn) pfn_valid(pfn)
1221void sparse_init(void);
1222#else
1223#define sparse_init() do {} while (0)
Dave Hansen28ae55c2005-09-03 15:54:29 -07001224#define sparse_index_init(_sec, _nid) do {} while (0)
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001225#endif /* CONFIG_SPARSEMEM */
1226
Mel Gorman8a942fd2015-06-30 14:56:55 -07001227/*
1228 * During memory init memblocks map pfns to nids. The search is expensive and
1229 * this caches recent lookups. The implementation of __early_pfn_to_nid
1230 * may treat start/end as pfns or sections.
1231 */
1232struct mminit_pfnnid_cache {
1233 unsigned long last_start;
1234 unsigned long last_end;
1235 int last_nid;
1236};
1237
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001238#ifndef early_pfn_valid
1239#define early_pfn_valid(pfn) (1)
1240#endif
1241
1242void memory_present(int nid, unsigned long start, unsigned long end);
1243unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
1244
Andy Whitcroft14e07292007-05-06 14:49:14 -07001245/*
1246 * If it is possible to have holes within a MAX_ORDER_NR_PAGES, then we
1247 * need to check pfn validility within that MAX_ORDER_NR_PAGES block.
1248 * pfn_valid_within() should be used in this case; we optimise this away
1249 * when we have no holes within a MAX_ORDER_NR_PAGES block.
1250 */
1251#ifdef CONFIG_HOLES_IN_ZONE
1252#define pfn_valid_within(pfn) pfn_valid(pfn)
1253#else
1254#define pfn_valid_within(pfn) (1)
1255#endif
1256
Mel Gormaneb335752009-05-13 17:34:48 +01001257#ifdef CONFIG_ARCH_HAS_HOLES_MEMORYMODEL
1258/*
1259 * pfn_valid() is meant to be able to tell if a given PFN has valid memmap
1260 * associated with it or not. In FLATMEM, it is expected that holes always
1261 * have valid memmap as long as there is valid PFNs either side of the hole.
1262 * In SPARSEMEM, it is assumed that a valid section has a memmap for the
1263 * entire section.
1264 *
1265 * However, an ARM, and maybe other embedded architectures in the future
1266 * free memmap backing holes to save memory on the assumption the memmap is
1267 * never used. The page_zone linkages are then broken even though pfn_valid()
1268 * returns true. A walker of the full memmap must then do this additional
1269 * check to ensure the memmap they are looking at is sane by making sure
1270 * the zone and PFN linkages are still valid. This is expensive, but walkers
1271 * of the full memmap are extremely rare.
1272 */
1273int memmap_valid_within(unsigned long pfn,
1274 struct page *page, struct zone *zone);
1275#else
1276static inline int memmap_valid_within(unsigned long pfn,
1277 struct page *page, struct zone *zone)
1278{
1279 return 1;
1280}
1281#endif /* CONFIG_ARCH_HAS_HOLES_MEMORYMODEL */
1282
Christoph Lameter97965472008-04-28 02:12:54 -07001283#endif /* !__GENERATING_BOUNDS.H */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284#endif /* !__ASSEMBLY__ */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001285#endif /* _LINUX_MMZONE_H */