blob: d572b78b65e14709d5fb8d8a60ee9d71147d0d81 [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,
Michal Nazarewicz47118af2011-12-29 13:09:50 +010040 MIGRATE_MOVABLE,
Mel Gorman016c13d2015-11-06 16:28:18 -080041 MIGRATE_RECLAIMABLE,
Mel Gorman0aaa29a2015-11-06 16:28:37 -080042 MIGRATE_PCPTYPES, /* the number of types on the pcp lists */
43 MIGRATE_HIGHATOMIC = MIGRATE_PCPTYPES,
Michal Nazarewicz47118af2011-12-29 13:09:50 +010044#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
Vlastimil Babka60f30352016-03-15 14:56:08 -070066/* In mm/page_alloc.c; keep in sync also with show_migration_types() there */
67extern char * const migratetype_names[MIGRATE_TYPES];
68
Michal Nazarewicz47118af2011-12-29 13:09:50 +010069#ifdef CONFIG_CMA
70# define is_migrate_cma(migratetype) unlikely((migratetype) == MIGRATE_CMA)
Laura Abbott7c15d9b2016-07-19 15:00:04 -070071# define is_migrate_cma_page(_page) (get_pageblock_migratetype(_page) == MIGRATE_CMA)
Michal Nazarewicz47118af2011-12-29 13:09:50 +010072#else
73# define is_migrate_cma(migratetype) false
Laura Abbott7c15d9b2016-07-19 15:00:04 -070074# define is_migrate_cma_page(_page) false
Michal Nazarewicz47118af2011-12-29 13:09:50 +010075#endif
Mel Gormanb2a0ac82007-10-16 01:25:48 -070076
77#define for_each_migratetype_order(order, type) \
78 for (order = 0; order < MAX_ORDER; order++) \
79 for (type = 0; type < MIGRATE_TYPES; type++)
80
Mel Gorman467c9962007-10-16 01:26:02 -070081extern int page_group_by_mobility_disabled;
82
Mel Gormane58469b2014-06-04 16:10:16 -070083#define NR_MIGRATETYPE_BITS (PB_migrate_end - PB_migrate + 1)
84#define MIGRATETYPE_MASK ((1UL << NR_MIGRATETYPE_BITS) - 1)
85
Mel Gormandc4b0ca2014-06-04 16:10:17 -070086#define get_pageblock_migratetype(page) \
87 get_pfnblock_flags_mask(page, page_to_pfn(page), \
88 PB_migrate_end, MIGRATETYPE_MASK)
89
Linus Torvalds1da177e2005-04-16 15:20:36 -070090struct free_area {
Mel Gormanb2a0ac82007-10-16 01:25:48 -070091 struct list_head free_list[MIGRATE_TYPES];
Linus Torvalds1da177e2005-04-16 15:20:36 -070092 unsigned long nr_free;
93};
94
95struct pglist_data;
96
97/*
Mel Gormana52633d2016-07-28 15:45:28 -070098 * zone->lock and the zone lru_lock are two of the hottest locks in the kernel.
Linus Torvalds1da177e2005-04-16 15:20:36 -070099 * So add a wild amount of padding here to ensure that they fall into separate
100 * cachelines. There are very few zone structures in the machine, so space
101 * consumption is not a concern here.
102 */
103#if defined(CONFIG_SMP)
104struct zone_padding {
105 char x[0];
Ravikiran G Thirumalai22fc6ec2006-01-08 01:01:27 -0800106} ____cacheline_internodealigned_in_smp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107#define ZONE_PADDING(name) struct zone_padding name;
108#else
109#define ZONE_PADDING(name)
110#endif
111
Christoph Lameter2244b952006-06-30 01:55:33 -0700112enum zone_stat_item {
Christoph Lameter51ed4492007-02-10 01:43:02 -0800113 /* First 128 byte cacheline (assuming 64 bit words) */
Christoph Lameterd23ad422007-02-10 01:43:02 -0800114 NR_FREE_PAGES,
Minchan Kim71c799f2016-07-28 15:47:26 -0700115 NR_ZONE_LRU_BASE, /* Used only for compaction and reclaim retry */
116 NR_ZONE_INACTIVE_ANON = NR_ZONE_LRU_BASE,
117 NR_ZONE_ACTIVE_ANON,
118 NR_ZONE_INACTIVE_FILE,
119 NR_ZONE_ACTIVE_FILE,
120 NR_ZONE_UNEVICTABLE,
Mel Gorman5a1c84b2016-07-28 15:47:31 -0700121 NR_ZONE_WRITE_PENDING, /* Count of dirty, writeback and unstable pages */
Nick Piggin5344b7e2008-10-18 20:26:51 -0700122 NR_MLOCK, /* mlock()ed pages found and moved off LRU */
Christoph Lameter51ed4492007-02-10 01:43:02 -0800123 NR_SLAB_RECLAIMABLE,
124 NR_SLAB_UNRECLAIMABLE,
125 NR_PAGETABLE, /* used for pagetables */
Andy Lutomirskid30dd8b2016-07-28 15:48:14 -0700126 NR_KERNEL_STACK_KB, /* measured in KiB */
KOSAKI Motohiroc6a7f572009-09-21 17:01:32 -0700127 /* Second 128 byte cacheline */
Christoph Lameterd2c5e302006-06-30 01:55:41 -0700128 NR_BOUNCE,
Minchan Kim91537fe2016-07-26 15:24:45 -0700129#if IS_ENABLED(CONFIG_ZSMALLOC)
130 NR_ZSPAGES, /* allocated in zsmalloc */
131#endif
Christoph Lameterca889e62006-06-30 01:55:44 -0700132#ifdef CONFIG_NUMA
133 NUMA_HIT, /* allocated in intended node */
134 NUMA_MISS, /* allocated in non intended node */
135 NUMA_FOREIGN, /* was intended here, hit elsewhere */
136 NUMA_INTERLEAVE_HIT, /* interleaver preferred this zone */
137 NUMA_LOCAL, /* allocation from local node */
138 NUMA_OTHER, /* allocation from other node */
139#endif
Bartlomiej Zolnierkiewiczd1ce7492012-10-08 16:32:02 -0700140 NR_FREE_CMA_PAGES,
Christoph Lameter2244b952006-06-30 01:55:33 -0700141 NR_VM_ZONE_STAT_ITEMS };
142
Mel Gorman75ef7182016-07-28 15:45:24 -0700143enum node_stat_item {
Mel Gorman599d0c92016-07-28 15:45:31 -0700144 NR_LRU_BASE,
145 NR_INACTIVE_ANON = NR_LRU_BASE, /* must match order of LRU_[IN]ACTIVE */
146 NR_ACTIVE_ANON, /* " " " " " */
147 NR_INACTIVE_FILE, /* " " " " " */
148 NR_ACTIVE_FILE, /* " " " " " */
149 NR_UNEVICTABLE, /* " " " " " */
150 NR_ISOLATED_ANON, /* Temporary isolated pages from anon lru */
151 NR_ISOLATED_FILE, /* Temporary isolated pages from file lru */
152 NR_PAGES_SCANNED, /* pages scanned since last reclaim */
Mel Gorman1e6b10852016-07-28 15:46:08 -0700153 WORKINGSET_REFAULT,
154 WORKINGSET_ACTIVATE,
155 WORKINGSET_NODERECLAIM,
Mel Gorman4b9d0fa2016-07-28 15:46:17 -0700156 NR_ANON_MAPPED, /* Mapped anonymous pages */
Mel Gorman50658e22016-07-28 15:46:14 -0700157 NR_FILE_MAPPED, /* pagecache pages mapped into pagetables.
158 only modified from process context */
Mel Gorman11fb9982016-07-28 15:46:20 -0700159 NR_FILE_PAGES,
160 NR_FILE_DIRTY,
161 NR_WRITEBACK,
162 NR_WRITEBACK_TEMP, /* Writeback using temporary buffers */
163 NR_SHMEM, /* shmem pages (included tmpfs/GEM pages) */
164 NR_SHMEM_THPS,
165 NR_SHMEM_PMDMAPPED,
166 NR_ANON_THPS,
167 NR_UNSTABLE_NFS, /* NFS unstable pages */
Mel Gormanc4a25632016-07-28 15:46:23 -0700168 NR_VMSCAN_WRITE,
169 NR_VMSCAN_IMMEDIATE, /* Prioritise for reclaim when writeback ends */
170 NR_DIRTIED, /* page dirtyings since bootup */
171 NR_WRITTEN, /* page writings since bootup */
Mel Gorman75ef7182016-07-28 15:45:24 -0700172 NR_VM_NODE_STAT_ITEMS
173};
174
Rik van Riel4f98a2f2008-10-18 20:26:32 -0700175/*
176 * We do arithmetic on the LRU lists in various places in the code,
177 * so it is important to keep the active lists LRU_ACTIVE higher in
178 * the array than the corresponding inactive lists, and to keep
179 * the *_FILE lists LRU_FILE higher than the corresponding _ANON lists.
180 *
181 * This has to be kept in sync with the statistics in zone_stat_item
182 * above and the descriptions in vmstat_text in mm/vmstat.c
183 */
184#define LRU_BASE 0
185#define LRU_ACTIVE 1
186#define LRU_FILE 2
187
Christoph Lameterb69408e2008-10-18 20:26:14 -0700188enum lru_list {
Rik van Riel4f98a2f2008-10-18 20:26:32 -0700189 LRU_INACTIVE_ANON = LRU_BASE,
190 LRU_ACTIVE_ANON = LRU_BASE + LRU_ACTIVE,
191 LRU_INACTIVE_FILE = LRU_BASE + LRU_FILE,
192 LRU_ACTIVE_FILE = LRU_BASE + LRU_FILE + LRU_ACTIVE,
Lee Schermerhorn894bc312008-10-18 20:26:39 -0700193 LRU_UNEVICTABLE,
Lee Schermerhorn894bc312008-10-18 20:26:39 -0700194 NR_LRU_LISTS
195};
Christoph Lameterb69408e2008-10-18 20:26:14 -0700196
Hugh Dickins41113042012-01-12 17:20:01 -0800197#define for_each_lru(lru) for (lru = 0; lru < NR_LRU_LISTS; lru++)
Christoph Lameterb69408e2008-10-18 20:26:14 -0700198
Hugh Dickins41113042012-01-12 17:20:01 -0800199#define for_each_evictable_lru(lru) for (lru = 0; lru <= LRU_ACTIVE_FILE; lru++)
Lee Schermerhorn894bc312008-10-18 20:26:39 -0700200
Hugh Dickins41113042012-01-12 17:20:01 -0800201static inline int is_file_lru(enum lru_list lru)
Rik van Riel4f98a2f2008-10-18 20:26:32 -0700202{
Hugh Dickins41113042012-01-12 17:20:01 -0800203 return (lru == LRU_INACTIVE_FILE || lru == LRU_ACTIVE_FILE);
Rik van Riel4f98a2f2008-10-18 20:26:32 -0700204}
205
Hugh Dickins41113042012-01-12 17:20:01 -0800206static inline int is_active_lru(enum lru_list lru)
Christoph Lameterb69408e2008-10-18 20:26:14 -0700207{
Hugh Dickins41113042012-01-12 17:20:01 -0800208 return (lru == LRU_ACTIVE_ANON || lru == LRU_ACTIVE_FILE);
Christoph Lameterb69408e2008-10-18 20:26:14 -0700209}
210
Hugh Dickins89abfab2012-05-29 15:06:53 -0700211struct zone_reclaim_stat {
212 /*
213 * The pageout code in vmscan.c keeps track of how many of the
Jiri Kosina59f91e52012-06-29 14:45:58 +0200214 * mem/swap backed and file backed pages are referenced.
Hugh Dickins89abfab2012-05-29 15:06:53 -0700215 * The higher the rotated/scanned ratio, the more valuable
216 * that cache is.
217 *
218 * The anon LRU stats live in [0], file LRU stats in [1]
219 */
220 unsigned long recent_rotated[2];
221 unsigned long recent_scanned[2];
222};
223
Johannes Weiner6290df52012-01-12 17:18:10 -0800224struct lruvec {
Johannes Weiner23047a92016-03-15 14:57:16 -0700225 struct list_head lists[NR_LRU_LISTS];
226 struct zone_reclaim_stat reclaim_stat;
227 /* Evictions & activations on the inactive file list */
228 atomic_long_t inactive_age;
Andrew Mortonc255a452012-07-31 16:43:02 -0700229#ifdef CONFIG_MEMCG
Mel Gorman599d0c92016-07-28 15:45:31 -0700230 struct pglist_data *pgdat;
Konstantin Khlebnikov7f5e86c2012-05-29 15:06:58 -0700231#endif
Johannes Weiner6290df52012-01-12 17:18:10 -0800232};
233
KAMEZAWA Hiroyukibb2a0de2011-07-26 16:08:22 -0700234/* Mask used at gathering information at once (see memcontrol.c) */
235#define LRU_ALL_FILE (BIT(LRU_INACTIVE_FILE) | BIT(LRU_ACTIVE_FILE))
236#define LRU_ALL_ANON (BIT(LRU_INACTIVE_ANON) | BIT(LRU_ACTIVE_ANON))
KAMEZAWA Hiroyukibb2a0de2011-07-26 16:08:22 -0700237#define LRU_ALL ((1 << NR_LRU_LISTS) - 1)
238
Minchan Kim39deaf82011-10-31 17:06:51 -0700239/* Isolate clean file */
Konstantin Khlebnikovf3fd4a62012-05-29 15:06:54 -0700240#define ISOLATE_CLEAN ((__force isolate_mode_t)0x1)
Minchan Kimf80c0672011-10-31 17:06:55 -0700241/* Isolate unmapped file */
Konstantin Khlebnikovf3fd4a62012-05-29 15:06:54 -0700242#define ISOLATE_UNMAPPED ((__force isolate_mode_t)0x2)
Mel Gormanc8244932012-01-12 17:19:38 -0800243/* Isolate for asynchronous migration */
Konstantin Khlebnikovf3fd4a62012-05-29 15:06:54 -0700244#define ISOLATE_ASYNC_MIGRATE ((__force isolate_mode_t)0x4)
Minchan Kime46a2872012-10-08 16:33:48 -0700245/* Isolate unevictable pages */
246#define ISOLATE_UNEVICTABLE ((__force isolate_mode_t)0x8)
Minchan Kim4356f212011-10-31 17:06:47 -0700247
248/* LRU Isolation modes. */
249typedef unsigned __bitwise__ isolate_mode_t;
250
Mel Gorman41858962009-06-16 15:32:12 -0700251enum zone_watermarks {
252 WMARK_MIN,
253 WMARK_LOW,
254 WMARK_HIGH,
255 NR_WMARK
256};
257
258#define min_wmark_pages(z) (z->watermark[WMARK_MIN])
259#define low_wmark_pages(z) (z->watermark[WMARK_LOW])
260#define high_wmark_pages(z) (z->watermark[WMARK_HIGH])
261
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262struct per_cpu_pages {
263 int count; /* number of pages in the list */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264 int high; /* high watermark, emptying needed */
265 int batch; /* chunk size for buddy add/remove */
Mel Gorman5f8dcc22009-09-21 17:03:19 -0700266
267 /* Lists of pages, one per migrate type stored on the pcp-lists */
268 struct list_head lists[MIGRATE_PCPTYPES];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700269};
270
271struct per_cpu_pageset {
Christoph Lameter3dfa5722008-02-04 22:29:19 -0800272 struct per_cpu_pages pcp;
Christoph Lameter4037d452007-05-09 02:35:14 -0700273#ifdef CONFIG_NUMA
274 s8 expire;
275#endif
Christoph Lameter2244b952006-06-30 01:55:33 -0700276#ifdef CONFIG_SMP
Christoph Lameterdf9ecab2006-08-31 21:27:35 -0700277 s8 stat_threshold;
Christoph Lameter2244b952006-06-30 01:55:33 -0700278 s8 vm_stat_diff[NR_VM_ZONE_STAT_ITEMS];
279#endif
Christoph Lameter99dcc3e2010-01-05 15:34:51 +0900280};
Christoph Lametere7c8d5c2005-06-21 17:14:47 -0700281
Mel Gorman75ef7182016-07-28 15:45:24 -0700282struct per_cpu_nodestat {
283 s8 stat_threshold;
284 s8 vm_node_stat_diff[NR_VM_NODE_STAT_ITEMS];
285};
286
Christoph Lameter97965472008-04-28 02:12:54 -0700287#endif /* !__GENERATING_BOUNDS.H */
288
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700289enum zone_type {
Christoph Lameter4b51d662007-02-10 01:43:10 -0800290#ifdef CONFIG_ZONE_DMA
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700291 /*
292 * ZONE_DMA is used when there are devices that are not able
293 * to do DMA to all of addressable memory (ZONE_NORMAL). Then we
294 * carve out the portion of memory that is needed for these devices.
295 * The range is arch specific.
296 *
297 * Some examples
298 *
299 * Architecture Limit
300 * ---------------------------
301 * parisc, ia64, sparc <4G
302 * s390 <2G
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700303 * arm Various
304 * alpha Unlimited or 0-16MB.
305 *
306 * i386, x86_64 and multiple other arches
307 * <16M.
308 */
309 ZONE_DMA,
Christoph Lameter4b51d662007-02-10 01:43:10 -0800310#endif
Christoph Lameterfb0e7942006-09-25 23:31:13 -0700311#ifdef CONFIG_ZONE_DMA32
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700312 /*
313 * x86_64 needs two ZONE_DMAs because it supports devices that are
314 * only able to do DMA to the lower 16M but also 32 bit devices that
315 * can only do DMA areas below 4G.
316 */
317 ZONE_DMA32,
Christoph Lameterfb0e7942006-09-25 23:31:13 -0700318#endif
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700319 /*
320 * Normal addressable memory is in ZONE_NORMAL. DMA operations can be
321 * performed on pages in ZONE_NORMAL if the DMA devices support
322 * transfers to all addressable memory.
323 */
324 ZONE_NORMAL,
Christoph Lametere53ef382006-09-25 23:31:14 -0700325#ifdef CONFIG_HIGHMEM
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700326 /*
327 * A memory area that is only addressable by the kernel through
328 * mapping portions into its own address space. This is for example
329 * used by i386 to allow the kernel to address the memory beyond
330 * 900MB. The kernel will set up special mappings (page
331 * table entries on i386) for each page that the kernel needs to
332 * access.
333 */
334 ZONE_HIGHMEM,
Christoph Lametere53ef382006-09-25 23:31:14 -0700335#endif
Mel Gorman2a1e2742007-07-17 04:03:12 -0700336 ZONE_MOVABLE,
Dan Williams033fbae2015-08-09 15:29:06 -0400337#ifdef CONFIG_ZONE_DEVICE
338 ZONE_DEVICE,
339#endif
Christoph Lameter97965472008-04-28 02:12:54 -0700340 __MAX_NR_ZONES
Dan Williams033fbae2015-08-09 15:29:06 -0400341
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700342};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700343
Christoph Lameter97965472008-04-28 02:12:54 -0700344#ifndef __GENERATING_BOUNDS_H
345
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346struct zone {
Mel Gorman3484b2d2014-08-06 16:07:14 -0700347 /* Read-mostly fields */
Mel Gorman41858962009-06-16 15:32:12 -0700348
349 /* zone watermarks, access with *_wmark_pages(zone) macros */
350 unsigned long watermark[NR_WMARK];
351
Mel Gorman0aaa29a2015-11-06 16:28:37 -0800352 unsigned long nr_reserved_highatomic;
353
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354 /*
Andrew Morton89903322015-11-06 16:28:46 -0800355 * We don't know if the memory that we're going to allocate will be
356 * freeable or/and it will be released eventually, so to avoid totally
357 * wasting several GB of ram we must reserve some of the lower zone
358 * memory (otherwise we risk to run OOM on the lower zones despite
359 * there being tons of freeable ram on the higher zones). This array is
360 * recalculated at runtime if the sysctl_lowmem_reserve_ratio sysctl
361 * changes.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700362 */
Mel Gorman3484b2d2014-08-06 16:07:14 -0700363 long lowmem_reserve[MAX_NR_ZONES];
364
365#ifdef CONFIG_NUMA
366 int node;
367#endif
Mel Gorman3484b2d2014-08-06 16:07:14 -0700368 struct pglist_data *zone_pgdat;
369 struct per_cpu_pageset __percpu *pageset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700370
Mel Gorman835c1342007-10-16 01:25:47 -0700371#ifndef CONFIG_SPARSEMEM
372 /*
Mel Gormand9c23402007-10-16 01:26:01 -0700373 * Flags for a pageblock_nr_pages block. See pageblock-flags.h.
Mel Gorman835c1342007-10-16 01:25:47 -0700374 * In SPARSEMEM, this map is stored in struct mem_section
375 */
376 unsigned long *pageblock_flags;
377#endif /* CONFIG_SPARSEMEM */
378
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379 /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */
380 unsigned long zone_start_pfn;
381
Dave Hansenbdc8cb92005-10-29 18:16:53 -0700382 /*
Jiang Liu9feedc92012-12-12 13:52:12 -0800383 * spanned_pages is the total pages spanned by the zone, including
384 * holes, which is calculated as:
385 * spanned_pages = zone_end_pfn - zone_start_pfn;
Dave Hansenbdc8cb92005-10-29 18:16:53 -0700386 *
Jiang Liu9feedc92012-12-12 13:52:12 -0800387 * present_pages is physical pages existing within the zone, which
388 * is calculated as:
Cody P Schafer8761e312013-03-26 10:30:44 -0700389 * present_pages = spanned_pages - absent_pages(pages in holes);
Jiang Liu9feedc92012-12-12 13:52:12 -0800390 *
391 * managed_pages is present pages managed by the buddy system, which
392 * is calculated as (reserved_pages includes pages allocated by the
393 * bootmem allocator):
394 * managed_pages = present_pages - reserved_pages;
395 *
396 * So present_pages may be used by memory hotplug or memory power
397 * management logic to figure out unmanaged pages by checking
398 * (present_pages - managed_pages). And managed_pages should be used
399 * by page allocator and vm scanner to calculate all kinds of watermarks
400 * and thresholds.
401 *
402 * Locking rules:
403 *
404 * zone_start_pfn and spanned_pages are protected by span_seqlock.
405 * It is a seqlock because it has to be read outside of zone->lock,
406 * and it is done in the main allocator path. But, it is written
407 * quite infrequently.
408 *
409 * The span_seq lock is declared along with zone->lock because it is
Dave Hansenbdc8cb92005-10-29 18:16:53 -0700410 * frequently read in proximity to zone->lock. It's good to
411 * give them a chance of being in the same cacheline.
Jiang Liu9feedc92012-12-12 13:52:12 -0800412 *
Jiang Liuc3d5f5f2013-07-03 15:03:14 -0700413 * Write access to present_pages at runtime should be protected by
Vladimir Davydovbfc8c902014-06-04 16:07:18 -0700414 * mem_hotplug_begin/end(). Any reader who can't tolerant drift of
415 * present_pages should get_online_mems() to get a stable value.
Jiang Liuc3d5f5f2013-07-03 15:03:14 -0700416 *
417 * Read access to managed_pages should be safe because it's unsigned
418 * long. Write access to zone->managed_pages and totalram_pages are
419 * protected by managed_page_count_lock at runtime. Idealy only
420 * adjust_managed_page_count() should be used instead of directly
421 * touching zone->managed_pages and totalram_pages.
Dave Hansenbdc8cb92005-10-29 18:16:53 -0700422 */
Mel Gorman3484b2d2014-08-06 16:07:14 -0700423 unsigned long managed_pages;
Jiang Liu9feedc92012-12-12 13:52:12 -0800424 unsigned long spanned_pages;
425 unsigned long present_pages;
Mel Gorman3484b2d2014-08-06 16:07:14 -0700426
427 const char *name;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428
Joonsoo Kimad53f922014-11-13 15:19:11 -0800429#ifdef CONFIG_MEMORY_ISOLATION
430 /*
431 * Number of isolated pageblock. It is used to solve incorrect
432 * freepage counting problem due to racy retrieving migratetype
433 * of pageblock. Protected by zone->lock.
434 */
435 unsigned long nr_isolate_pageblock;
436#endif
437
Mel Gorman3484b2d2014-08-06 16:07:14 -0700438#ifdef CONFIG_MEMORY_HOTPLUG
439 /* see spanned/present_pages for more description */
440 seqlock_t span_seqlock;
441#endif
442
Yasuaki Ishimatsu943dca12014-01-21 15:49:06 -0800443 /*
Mel Gorman3484b2d2014-08-06 16:07:14 -0700444 * wait_table -- the array holding the hash table
445 * wait_table_hash_nr_entries -- the size of the hash table array
446 * wait_table_bits -- wait_table_size == (1 << wait_table_bits)
447 *
448 * The purpose of all these is to keep track of the people
449 * waiting for a page to become available and make them
450 * runnable again when possible. The trouble is that this
451 * consumes a lot of space, especially when so few things
452 * wait on pages at a given time. So instead of using
453 * per-page waitqueues, we use a waitqueue hash table.
454 *
455 * The bucket discipline is to sleep on the same queue when
456 * colliding and wake all in that wait queue when removing.
457 * When something wakes, it must check to be sure its page is
458 * truly available, a la thundering herd. The cost of a
459 * collision is great, but given the expected load of the
460 * table, they should be so rare as to be outweighed by the
461 * benefits from the saved space.
462 *
463 * __wait_on_page_locked() and unlock_page() in mm/filemap.c, are the
464 * primary users of these fields, and in mm/page_alloc.c
465 * free_area_init_core() performs the initialization of them.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466 */
Mel Gorman3484b2d2014-08-06 16:07:14 -0700467 wait_queue_head_t *wait_table;
468 unsigned long wait_table_hash_nr_entries;
469 unsigned long wait_table_bits;
470
Mel Gorman0f661142016-07-28 15:45:34 -0700471 /* Write-intensive fields used from the page allocator */
Mel Gorman3484b2d2014-08-06 16:07:14 -0700472 ZONE_PADDING(_pad1_)
Mel Gorman0f661142016-07-28 15:45:34 -0700473
Mel Gorman3484b2d2014-08-06 16:07:14 -0700474 /* free areas of different sizes */
475 struct free_area free_area[MAX_ORDER];
476
477 /* zone flags, see below */
478 unsigned long flags;
479
Mel Gorman0f661142016-07-28 15:45:34 -0700480 /* Primarily protects free_area */
Mel Gormana368ab62015-04-07 14:26:41 -0700481 spinlock_t lock;
482
Mel Gorman0f661142016-07-28 15:45:34 -0700483 /* Write-intensive fields used by compaction and vmstats. */
Mel Gorman3484b2d2014-08-06 16:07:14 -0700484 ZONE_PADDING(_pad2_)
485
Mel Gorman3484b2d2014-08-06 16:07:14 -0700486 /*
487 * When free pages are below this point, additional steps are taken
488 * when reading the number of free pages to avoid per-cpu counter
489 * drift allowing watermarks to be breached
490 */
491 unsigned long percpu_drift_mark;
492
493#if defined CONFIG_COMPACTION || defined CONFIG_CMA
494 /* pfn where compaction free scanner should start */
495 unsigned long compact_cached_free_pfn;
496 /* pfn where async and sync compaction migration scanner should start */
497 unsigned long compact_cached_migrate_pfn[2];
498#endif
499
500#ifdef CONFIG_COMPACTION
501 /*
502 * On compaction failure, 1<<compact_defer_shift compactions
503 * are skipped before trying again. The number attempted since
504 * last failure is tracked with compact_considered.
505 */
506 unsigned int compact_considered;
507 unsigned int compact_defer_shift;
508 int compact_order_failed;
509#endif
510
511#if defined CONFIG_COMPACTION || defined CONFIG_CMA
512 /* Set to true when the PG_migrate_skip bits should be cleared */
513 bool compact_blockskip_flush;
514#endif
515
Joonsoo Kim7cf91a92016-03-15 14:57:51 -0700516 bool contiguous;
517
Mel Gorman3484b2d2014-08-06 16:07:14 -0700518 ZONE_PADDING(_pad3_)
519 /* Zone statistics */
520 atomic_long_t vm_stat[NR_VM_ZONE_STAT_ITEMS];
Ravikiran G Thirumalai22fc6ec2006-01-08 01:01:27 -0800521} ____cacheline_internodealigned_in_smp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522
Mel Gorman599d0c92016-07-28 15:45:31 -0700523enum pgdat_flags {
524 PGDAT_CONGESTED, /* pgdat has many dirty pages backed by
Mel Gorman0e093d992010-10-26 14:21:45 -0700525 * a congested BDI
526 */
Mel Gorman599d0c92016-07-28 15:45:31 -0700527 PGDAT_DIRTY, /* reclaim scanning has recently found
Mel Gormand43006d2013-07-03 15:01:50 -0700528 * many dirty file pages at the tail
529 * of the LRU.
530 */
Mel Gorman599d0c92016-07-28 15:45:31 -0700531 PGDAT_WRITEBACK, /* reclaim scanning has recently found
Mel Gorman283aba92013-07-03 15:01:51 -0700532 * many pages under writeback
533 */
Mel Gormana5f5f912016-07-28 15:46:32 -0700534 PGDAT_RECLAIM_LOCKED, /* prevents concurrent reclaim */
Johannes Weiner57054652014-10-09 15:28:17 -0700535};
David Rientjese815af92007-10-16 23:25:54 -0700536
Russ Andersonf9228b22013-03-22 15:04:43 -0700537static inline unsigned long zone_end_pfn(const struct zone *zone)
Cody P Schafer108bcc92013-02-22 16:35:23 -0800538{
539 return zone->zone_start_pfn + zone->spanned_pages;
540}
541
542static inline bool zone_spans_pfn(const struct zone *zone, unsigned long pfn)
543{
544 return zone->zone_start_pfn <= pfn && pfn < zone_end_pfn(zone);
545}
546
Cody P Schafer2a6e3eb2013-02-22 16:35:24 -0800547static inline bool zone_is_initialized(struct zone *zone)
548{
549 return !!zone->wait_table;
550}
551
552static inline bool zone_is_empty(struct zone *zone)
553{
554 return zone->spanned_pages == 0;
555}
556
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557/*
558 * The "priority" of VM scanning is how much of the queues we will scan in one
559 * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the
560 * queues ("queue_length >> 12") during an aging round.
561 */
562#define DEF_PRIORITY 12
563
Paul Jackson9276b1bc2006-12-06 20:31:48 -0800564/* Maximum number of zones on a zonelist */
565#define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES)
566
Yaowei Baic00eb152016-01-14 15:19:00 -0800567enum {
568 ZONELIST_FALLBACK, /* zonelist with fallback */
Paul Jackson9276b1bc2006-12-06 20:31:48 -0800569#ifdef CONFIG_NUMA
Yaowei Baic00eb152016-01-14 15:19:00 -0800570 /*
571 * The NUMA zonelists are doubled because we need zonelists that
572 * restrict the allocations to a single node for __GFP_THISNODE.
573 */
574 ZONELIST_NOFALLBACK, /* zonelist without fallback (__GFP_THISNODE) */
Paul Jackson9276b1bc2006-12-06 20:31:48 -0800575#endif
Yaowei Baic00eb152016-01-14 15:19:00 -0800576 MAX_ZONELISTS
577};
Paul Jackson9276b1bc2006-12-06 20:31:48 -0800578
Linus Torvalds1da177e2005-04-16 15:20:36 -0700579/*
Mel Gormandd1a2392008-04-28 02:12:17 -0700580 * This struct contains information about a zone in a zonelist. It is stored
581 * here to avoid dereferences into large structures and lookups of tables
582 */
583struct zoneref {
584 struct zone *zone; /* Pointer to actual zone */
585 int zone_idx; /* zone_idx(zoneref->zone) */
586};
587
588/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589 * One allocation request operates on a zonelist. A zonelist
590 * is a list of zones, the first one is the 'goal' of the
591 * allocation, the other zones are fallback zones, in decreasing
592 * priority.
593 *
Mel Gormandd1a2392008-04-28 02:12:17 -0700594 * To speed the reading of the zonelist, the zonerefs contain the zone index
595 * of the entry being read. Helper functions to access information given
596 * a struct zoneref are
597 *
598 * zonelist_zone() - Return the struct zone * for an entry in _zonerefs
599 * zonelist_zone_idx() - Return the index of the zone for an entry
600 * zonelist_node_idx() - Return the index of the node for an entry
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601 */
602struct zonelist {
Mel Gormandd1a2392008-04-28 02:12:17 -0700603 struct zoneref _zonerefs[MAX_ZONES_PER_ZONELIST + 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604};
605
Heiko Carstens5b99cd02006-09-27 01:50:01 -0700606#ifndef CONFIG_DISCONTIGMEM
607/* The array of struct pages - for discontigmem use pgdat->lmem_map */
608extern struct page *mem_map;
609#endif
610
Linus Torvalds1da177e2005-04-16 15:20:36 -0700611/*
612 * The pg_data_t structure is used in machines with CONFIG_DISCONTIGMEM
613 * (mostly NUMA machines?) to denote a higher-level memory zone than the
614 * zone denotes.
615 *
616 * On NUMA machines, each NUMA node would have a pg_data_t to describe
617 * it's memory layout.
618 *
619 * Memory statistics and page replacement data structures are maintained on a
620 * per-zone basis.
621 */
622struct bootmem_data;
623typedef struct pglist_data {
624 struct zone node_zones[MAX_NR_ZONES];
Christoph Lameter523b9452007-10-16 01:25:37 -0700625 struct zonelist node_zonelists[MAX_ZONELISTS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626 int nr_zones;
KAMEZAWA Hiroyuki52d4b9a2008-10-18 20:28:16 -0700627#ifdef CONFIG_FLAT_NODE_MEM_MAP /* means !SPARSEMEM */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700628 struct page *node_mem_map;
Joonsoo Kimeefa8642014-12-12 16:55:46 -0800629#ifdef CONFIG_PAGE_EXTENSION
630 struct page_ext *node_page_ext;
631#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700632#endif
Yinghai Lu08677212010-02-10 01:20:20 -0800633#ifndef CONFIG_NO_BOOTMEM
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634 struct bootmem_data *bdata;
Yinghai Lu08677212010-02-10 01:20:20 -0800635#endif
Dave Hansen208d54e2005-10-29 18:16:52 -0700636#ifdef CONFIG_MEMORY_HOTPLUG
637 /*
638 * Must be held any time you expect node_start_pfn, node_present_pages
639 * or node_spanned_pages stay constant. Holding this will also
640 * guarantee that any pfn_valid() stays that way.
641 *
Cody P Schafer114d4b72013-07-03 15:02:09 -0700642 * pgdat_resize_lock() and pgdat_resize_unlock() are provided to
643 * manipulate node_size_lock without checking for CONFIG_MEMORY_HOTPLUG.
644 *
Cody P Schafer72c3b512013-07-03 15:02:08 -0700645 * Nests above zone->lock and zone->span_seqlock
Dave Hansen208d54e2005-10-29 18:16:52 -0700646 */
647 spinlock_t node_size_lock;
648#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649 unsigned long node_start_pfn;
650 unsigned long node_present_pages; /* total number of physical pages */
651 unsigned long node_spanned_pages; /* total size of physical page
652 range, including holes */
653 int node_id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654 wait_queue_head_t kswapd_wait;
Mel Gorman55150612012-07-31 16:44:35 -0700655 wait_queue_head_t pfmemalloc_wait;
Vladimir Davydovbfc8c902014-06-04 16:07:18 -0700656 struct task_struct *kswapd; /* Protected by
657 mem_hotplug_begin/end() */
Mel Gorman38087d92016-07-28 15:45:49 -0700658 int kswapd_order;
659 enum zone_type kswapd_classzone_idx;
660
Vlastimil Babka698b1b32016-03-17 14:18:08 -0700661#ifdef CONFIG_COMPACTION
662 int kcompactd_max_order;
663 enum zone_type kcompactd_classzone_idx;
664 wait_queue_head_t kcompactd_wait;
665 struct task_struct *kcompactd;
666#endif
Andrea Arcangeli8177a422012-03-23 20:56:34 +0100667#ifdef CONFIG_NUMA_BALANCING
Mel Gorman1c5e9c22014-01-21 15:50:59 -0800668 /* Lock serializing the migrate rate limiting window */
Andrea Arcangeli8177a422012-03-23 20:56:34 +0100669 spinlock_t numabalancing_migrate_lock;
670
671 /* Rate limiting time interval */
672 unsigned long numabalancing_migrate_next_window;
673
674 /* Number of pages migrated during the rate limiting time interval */
675 unsigned long numabalancing_migrate_nr_pages;
676#endif
Mel Gorman281e3722016-07-28 15:46:11 -0700677 /*
678 * This is a per-node reserve of pages that are not available
679 * to userspace allocations.
680 */
681 unsigned long totalreserve_pages;
682
Mel Gormana5f5f912016-07-28 15:46:32 -0700683#ifdef CONFIG_NUMA
684 /*
685 * zone reclaim becomes active if more unmapped pages exist.
686 */
687 unsigned long min_unmapped_pages;
688 unsigned long min_slab_pages;
689#endif /* CONFIG_NUMA */
690
Mel Gormana52633d2016-07-28 15:45:28 -0700691 /* Write-intensive fields used by page reclaim */
692 ZONE_PADDING(_pad1_)
693 spinlock_t lru_lock;
Mel Gorman3a80a7f2015-06-30 14:57:02 -0700694
695#ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT
696 /*
697 * If memory initialisation on large machines is deferred then this
698 * is the first PFN that needs to be initialised.
699 */
700 unsigned long first_deferred_pfn;
701#endif /* CONFIG_DEFERRED_STRUCT_PAGE_INIT */
Kirill A. Shutemova3d0a9182016-02-02 16:57:08 -0800702
703#ifdef CONFIG_TRANSPARENT_HUGEPAGE
704 spinlock_t split_queue_lock;
705 struct list_head split_queue;
706 unsigned long split_queue_len;
707#endif
Mel Gorman75ef7182016-07-28 15:45:24 -0700708
Mel Gorman599d0c92016-07-28 15:45:31 -0700709 /* Fields commonly accessed by the page reclaim scanner */
710 struct lruvec lruvec;
711
712 /*
713 * The target ratio of ACTIVE_ANON to INACTIVE_ANON pages on
714 * this node's LRU. Maintained by the pageout code.
715 */
716 unsigned int inactive_ratio;
717
718 unsigned long flags;
719
720 ZONE_PADDING(_pad2_)
721
Mel Gorman75ef7182016-07-28 15:45:24 -0700722 /* Per-node vmstats */
723 struct per_cpu_nodestat __percpu *per_cpu_nodestats;
724 atomic_long_t vm_stat[NR_VM_NODE_STAT_ITEMS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700725} pg_data_t;
726
727#define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages)
728#define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700729#ifdef CONFIG_FLAT_NODE_MEM_MAP
Dave Hansen408fde82005-06-23 00:07:37 -0700730#define pgdat_page_nr(pgdat, pagenr) ((pgdat)->node_mem_map + (pagenr))
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700731#else
732#define pgdat_page_nr(pgdat, pagenr) pfn_to_page((pgdat)->node_start_pfn + (pagenr))
733#endif
Dave Hansen408fde82005-06-23 00:07:37 -0700734#define nid_page_nr(nid, pagenr) pgdat_page_nr(NODE_DATA(nid),(pagenr))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735
KAMEZAWA Hiroyukic6830c22011-06-16 17:28:07 +0900736#define node_start_pfn(nid) (NODE_DATA(nid)->node_start_pfn)
Cody P Schaferda3649e2013-02-22 16:35:27 -0800737#define node_end_pfn(nid) pgdat_end_pfn(NODE_DATA(nid))
Mel Gormana52633d2016-07-28 15:45:28 -0700738static inline spinlock_t *zone_lru_lock(struct zone *zone)
739{
740 return &zone->zone_pgdat->lru_lock;
741}
KAMEZAWA Hiroyukic6830c22011-06-16 17:28:07 +0900742
Mel Gormana9dd0a82016-07-28 15:46:02 -0700743static inline struct lruvec *node_lruvec(struct pglist_data *pgdat)
Mel Gorman599d0c92016-07-28 15:45:31 -0700744{
Mel Gormana9dd0a82016-07-28 15:46:02 -0700745 return &pgdat->lruvec;
Mel Gorman599d0c92016-07-28 15:45:31 -0700746}
747
Cody P Schaferda3649e2013-02-22 16:35:27 -0800748static inline unsigned long pgdat_end_pfn(pg_data_t *pgdat)
749{
750 return pgdat->node_start_pfn + pgdat->node_spanned_pages;
751}
752
753static inline bool pgdat_is_empty(pg_data_t *pgdat)
754{
755 return !pgdat->node_start_pfn && !pgdat->node_spanned_pages;
756}
KAMEZAWA Hiroyukic6830c22011-06-16 17:28:07 +0900757
Dan Williams033fbae2015-08-09 15:29:06 -0400758static inline int zone_id(const struct zone *zone)
759{
760 struct pglist_data *pgdat = zone->zone_pgdat;
761
762 return zone - pgdat->node_zones;
763}
764
765#ifdef CONFIG_ZONE_DEVICE
766static inline bool is_dev_zone(const struct zone *zone)
767{
768 return zone_id(zone) == ZONE_DEVICE;
769}
770#else
771static inline bool is_dev_zone(const struct zone *zone)
772{
773 return false;
774}
775#endif
776
Dave Hansen208d54e2005-10-29 18:16:52 -0700777#include <linux/memory_hotplug.h>
778
Haicheng Li4eaf3f62010-05-24 14:32:52 -0700779extern struct mutex zonelists_mutex;
Jiang Liu9adb62a2012-07-31 16:43:28 -0700780void build_all_zonelists(pg_data_t *pgdat, struct zone *zone);
Mel Gorman99504742011-01-13 15:46:20 -0800781void wakeup_kswapd(struct zone *zone, int order, enum zone_type classzone_idx);
Michal Hocko86a294a2016-05-20 16:57:12 -0700782bool __zone_watermark_ok(struct zone *z, unsigned int order, unsigned long mark,
783 int classzone_idx, unsigned int alloc_flags,
784 long free_pages);
Mel Gorman7aeb09f2014-06-04 16:10:21 -0700785bool zone_watermark_ok(struct zone *z, unsigned int order,
Mel Gormanc6038442016-05-19 17:13:38 -0700786 unsigned long mark, int classzone_idx,
787 unsigned int alloc_flags);
Mel Gorman7aeb09f2014-06-04 16:10:21 -0700788bool zone_watermark_ok_safe(struct zone *z, unsigned int order,
Mel Gormane2b19192015-11-06 16:28:09 -0800789 unsigned long mark, int classzone_idx);
Dave Hansena2f3aa022007-01-10 23:15:30 -0800790enum memmap_context {
791 MEMMAP_EARLY,
792 MEMMAP_HOTPLUG,
793};
Yasunori Goto718127c2006-06-23 02:03:10 -0700794extern int init_currently_empty_zone(struct zone *zone, unsigned long start_pfn,
Yaowei Baib171e402015-11-05 18:47:06 -0800795 unsigned long size);
Yasunori Goto718127c2006-06-23 02:03:10 -0700796
Hugh Dickinsbea8c152012-11-16 14:14:54 -0800797extern void lruvec_init(struct lruvec *lruvec);
Konstantin Khlebnikov7f5e86c2012-05-29 15:06:58 -0700798
Mel Gorman599d0c92016-07-28 15:45:31 -0700799static inline struct pglist_data *lruvec_pgdat(struct lruvec *lruvec)
Konstantin Khlebnikov7f5e86c2012-05-29 15:06:58 -0700800{
Andrew Mortonc255a452012-07-31 16:43:02 -0700801#ifdef CONFIG_MEMCG
Mel Gorman599d0c92016-07-28 15:45:31 -0700802 return lruvec->pgdat;
Konstantin Khlebnikov7f5e86c2012-05-29 15:06:58 -0700803#else
Mel Gorman599d0c92016-07-28 15:45:31 -0700804 return container_of(lruvec, struct pglist_data, lruvec);
Konstantin Khlebnikov7f5e86c2012-05-29 15:06:58 -0700805#endif
806}
807
Johannes Weiner23047a92016-03-15 14:57:16 -0700808extern unsigned long lruvec_lru_size(struct lruvec *lruvec, enum lru_list lru);
809
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810#ifdef CONFIG_HAVE_MEMORY_PRESENT
811void memory_present(int nid, unsigned long start, unsigned long end);
812#else
813static inline void memory_present(int nid, unsigned long start, unsigned long end) {}
814#endif
815
Lee Schermerhorn7aac7892010-05-26 14:45:00 -0700816#ifdef CONFIG_HAVE_MEMORYLESS_NODES
817int local_memory_node(int node_id);
818#else
819static inline int local_memory_node(int node_id) { return node_id; };
820#endif
821
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822#ifdef CONFIG_NEED_NODE_MEMMAP_SIZE
823unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
824#endif
825
826/*
827 * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc.
828 */
829#define zone_idx(zone) ((zone) - (zone)->zone_pgdat->node_zones)
830
Con Kolivasf3fe6512006-01-06 00:11:15 -0800831static inline int populated_zone(struct zone *zone)
832{
833 return (!!zone->present_pages);
834}
835
Mel Gorman2a1e2742007-07-17 04:03:12 -0700836extern int movable_zone;
837
Zhang Zhend7e4a2e2015-04-15 16:12:57 -0700838#ifdef CONFIG_HIGHMEM
Mel Gorman2a1e2742007-07-17 04:03:12 -0700839static inline int zone_movable_is_highmem(void)
840{
Zhang Zhend7e4a2e2015-04-15 16:12:57 -0700841#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
Mel Gorman2a1e2742007-07-17 04:03:12 -0700842 return movable_zone == ZONE_HIGHMEM;
843#else
Zhang Zhend7e4a2e2015-04-15 16:12:57 -0700844 return (ZONE_MOVABLE - 1) == ZONE_HIGHMEM;
Mel Gorman2a1e2742007-07-17 04:03:12 -0700845#endif
846}
Zhang Zhend7e4a2e2015-04-15 16:12:57 -0700847#endif
Mel Gorman2a1e2742007-07-17 04:03:12 -0700848
Christoph Lameter2f1b6242006-09-25 23:31:13 -0700849static inline int is_highmem_idx(enum zone_type idx)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850{
Christoph Lametere53ef382006-09-25 23:31:14 -0700851#ifdef CONFIG_HIGHMEM
Mel Gorman2a1e2742007-07-17 04:03:12 -0700852 return (idx == ZONE_HIGHMEM ||
853 (idx == ZONE_MOVABLE && zone_movable_is_highmem()));
Christoph Lametere53ef382006-09-25 23:31:14 -0700854#else
855 return 0;
856#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857}
858
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859/**
860 * is_highmem - helper function to quickly check if a struct zone is a
861 * highmem zone or not. This is an attempt to keep references
862 * to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum.
863 * @zone - pointer to struct zone variable
864 */
865static inline int is_highmem(struct zone *zone)
866{
Christoph Lametere53ef382006-09-25 23:31:14 -0700867#ifdef CONFIG_HIGHMEM
Chanho Min29f9cb52016-05-19 17:11:57 -0700868 return is_highmem_idx(zone_idx(zone));
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/* These two functions are used to setup the per zone pages min values */
875struct ctl_table;
Alexey Dobriyan8d65af72009-09-23 15:57:19 -0700876int min_free_kbytes_sysctl_handler(struct ctl_table *, int,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877 void __user *, size_t *, loff_t *);
Johannes Weiner795ae7a2016-03-17 14:19:14 -0700878int watermark_scale_factor_sysctl_handler(struct ctl_table *, int,
879 void __user *, size_t *, loff_t *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880extern int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1];
Alexey Dobriyan8d65af72009-09-23 15:57:19 -0700881int lowmem_reserve_ratio_sysctl_handler(struct ctl_table *, int,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700882 void __user *, size_t *, loff_t *);
Alexey Dobriyan8d65af72009-09-23 15:57:19 -0700883int percpu_pagelist_fraction_sysctl_handler(struct ctl_table *, int,
Rohit Seth8ad4b1f2006-01-08 01:00:40 -0800884 void __user *, size_t *, loff_t *);
Christoph Lameter96146342006-07-03 00:24:13 -0700885int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table *, int,
Alexey Dobriyan8d65af72009-09-23 15:57:19 -0700886 void __user *, size_t *, loff_t *);
Christoph Lameter0ff38492006-09-25 23:31:52 -0700887int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table *, int,
Alexey Dobriyan8d65af72009-09-23 15:57:19 -0700888 void __user *, size_t *, loff_t *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889
KAMEZAWA Hiroyukif0c0b2b2007-07-15 23:38:01 -0700890extern int numa_zonelist_order_handler(struct ctl_table *, int,
Alexey Dobriyan8d65af72009-09-23 15:57:19 -0700891 void __user *, size_t *, loff_t *);
KAMEZAWA Hiroyukif0c0b2b2007-07-15 23:38:01 -0700892extern char numa_zonelist_order[];
893#define NUMA_ZONELIST_ORDER_LEN 16 /* string buffer size */
894
Dave Hansen93b75042005-06-23 00:07:47 -0700895#ifndef CONFIG_NEED_MULTIPLE_NODES
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896
897extern struct pglist_data contig_page_data;
898#define NODE_DATA(nid) (&contig_page_data)
899#define NODE_MEM_MAP(nid) mem_map
Linus Torvalds1da177e2005-04-16 15:20:36 -0700900
Dave Hansen93b75042005-06-23 00:07:47 -0700901#else /* CONFIG_NEED_MULTIPLE_NODES */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902
903#include <asm/mmzone.h>
904
Dave Hansen93b75042005-06-23 00:07:47 -0700905#endif /* !CONFIG_NEED_MULTIPLE_NODES */
Dave Hansen348f8b62005-06-23 00:07:40 -0700906
KAMEZAWA Hiroyuki95144c72006-03-27 01:16:02 -0800907extern struct pglist_data *first_online_pgdat(void);
908extern struct pglist_data *next_online_pgdat(struct pglist_data *pgdat);
909extern struct zone *next_zone(struct zone *zone);
KAMEZAWA Hiroyuki8357f862006-03-27 01:15:57 -0800910
911/**
Fernando Luis Vazquez Cao12d15f02008-05-23 13:05:01 -0700912 * for_each_online_pgdat - helper macro to iterate over all online nodes
KAMEZAWA Hiroyuki8357f862006-03-27 01:15:57 -0800913 * @pgdat - pointer to a pg_data_t variable
914 */
915#define for_each_online_pgdat(pgdat) \
916 for (pgdat = first_online_pgdat(); \
917 pgdat; \
918 pgdat = next_online_pgdat(pgdat))
KAMEZAWA Hiroyuki8357f862006-03-27 01:15:57 -0800919/**
920 * for_each_zone - helper macro to iterate over all memory zones
921 * @zone - pointer to struct zone variable
922 *
923 * The user only needs to declare the zone variable, for_each_zone
924 * fills it in.
925 */
926#define for_each_zone(zone) \
927 for (zone = (first_online_pgdat())->node_zones; \
928 zone; \
929 zone = next_zone(zone))
930
KOSAKI Motohiroee99c712009-03-31 15:19:31 -0700931#define for_each_populated_zone(zone) \
932 for (zone = (first_online_pgdat())->node_zones; \
933 zone; \
934 zone = next_zone(zone)) \
935 if (!populated_zone(zone)) \
936 ; /* do nothing */ \
937 else
938
Mel Gormandd1a2392008-04-28 02:12:17 -0700939static inline struct zone *zonelist_zone(struct zoneref *zoneref)
940{
941 return zoneref->zone;
942}
943
944static inline int zonelist_zone_idx(struct zoneref *zoneref)
945{
946 return zoneref->zone_idx;
947}
948
949static inline int zonelist_node_idx(struct zoneref *zoneref)
950{
951#ifdef CONFIG_NUMA
952 /* zone_to_nid not available in this context */
953 return zoneref->zone->node;
954#else
955 return 0;
956#endif /* CONFIG_NUMA */
957}
958
Mel Gorman682a3382016-05-19 17:13:30 -0700959struct zoneref *__next_zones_zonelist(struct zoneref *z,
960 enum zone_type highest_zoneidx,
961 nodemask_t *nodes);
962
Mel Gorman19770b32008-04-28 02:12:18 -0700963/**
964 * 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
965 * @z - The cursor used as a starting point for the search
966 * @highest_zoneidx - The zone index of the highest zone to return
967 * @nodes - An optional nodemask to filter the zonelist with
Mel Gorman19770b32008-04-28 02:12:18 -0700968 *
969 * This function returns the next zone at or below a given zone index that is
970 * within the allowed nodemask using a cursor as the starting point for the
Mel Gorman5bead2a2008-09-13 02:33:19 -0700971 * search. The zoneref returned is a cursor that represents the current zone
972 * being examined. It should be advanced by one before calling
973 * next_zones_zonelist again.
Mel Gorman19770b32008-04-28 02:12:18 -0700974 */
Mel Gorman682a3382016-05-19 17:13:30 -0700975static __always_inline struct zoneref *next_zones_zonelist(struct zoneref *z,
Mel Gorman19770b32008-04-28 02:12:18 -0700976 enum zone_type highest_zoneidx,
Mel Gorman682a3382016-05-19 17:13:30 -0700977 nodemask_t *nodes)
978{
979 if (likely(!nodes && zonelist_zone_idx(z) <= highest_zoneidx))
980 return z;
981 return __next_zones_zonelist(z, highest_zoneidx, nodes);
982}
Mel Gormandd1a2392008-04-28 02:12:17 -0700983
Mel Gorman19770b32008-04-28 02:12:18 -0700984/**
985 * first_zones_zonelist - Returns the first zone at or below highest_zoneidx within the allowed nodemask in a zonelist
986 * @zonelist - The zonelist to search for a suitable zone
987 * @highest_zoneidx - The zone index of the highest zone to return
988 * @nodes - An optional nodemask to filter the zonelist with
989 * @zone - The first suitable zone found is returned via this parameter
990 *
991 * This function returns the first zone at or below a given zone index that is
992 * within the allowed nodemask. The zoneref returned is a cursor that can be
Mel Gorman5bead2a2008-09-13 02:33:19 -0700993 * used to iterate the zonelist with next_zones_zonelist by advancing it by
994 * one before calling.
Mel Gorman19770b32008-04-28 02:12:18 -0700995 */
Mel Gormandd1a2392008-04-28 02:12:17 -0700996static inline struct zoneref *first_zones_zonelist(struct zonelist *zonelist,
Mel Gorman19770b32008-04-28 02:12:18 -0700997 enum zone_type highest_zoneidx,
Mel Gormanc33d6c02016-05-19 17:14:10 -0700998 nodemask_t *nodes)
Mel Gorman54a6eb52008-04-28 02:12:16 -0700999{
Mel Gormanc33d6c02016-05-19 17:14:10 -07001000 return next_zones_zonelist(zonelist->_zonerefs,
Vlastimil Babka05891fb2015-02-11 15:25:47 -08001001 highest_zoneidx, nodes);
Mel Gorman54a6eb52008-04-28 02:12:16 -07001002}
1003
Mel Gorman19770b32008-04-28 02:12:18 -07001004/**
1005 * for_each_zone_zonelist_nodemask - helper macro to iterate over valid zones in a zonelist at or below a given zone index and within a nodemask
1006 * @zone - The current zone in the iterator
1007 * @z - The current pointer within zonelist->zones being iterated
1008 * @zlist - The zonelist being iterated
1009 * @highidx - The zone index of the highest zone to return
1010 * @nodemask - Nodemask allowed by the allocator
1011 *
1012 * This iterator iterates though all zones at or below a given zone index and
1013 * within a given nodemask
1014 */
1015#define for_each_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \
Mel Gormanc33d6c02016-05-19 17:14:10 -07001016 for (z = first_zones_zonelist(zlist, highidx, nodemask), zone = zonelist_zone(z); \
Mel Gorman19770b32008-04-28 02:12:18 -07001017 zone; \
Vlastimil Babka05891fb2015-02-11 15:25:47 -08001018 z = next_zones_zonelist(++z, highidx, nodemask), \
Mel Gormanc33d6c02016-05-19 17:14:10 -07001019 zone = zonelist_zone(z))
1020
1021#define for_next_zone_zonelist_nodemask(zone, z, zlist, highidx, nodemask) \
1022 for (zone = z->zone; \
1023 zone; \
1024 z = next_zones_zonelist(++z, highidx, nodemask), \
1025 zone = zonelist_zone(z))
1026
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;
Joonsoo Kimeefa8642014-12-12 16:55:46 -08001086struct page_ext;
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;
Joonsoo Kimeefa8642014-12-12 16:55:46 -08001104#ifdef CONFIG_PAGE_EXTENSION
1105 /*
Weijie Yang0c9ad802016-05-20 16:58:04 -07001106 * If SPARSEMEM, pgdat doesn't have page_ext pointer. We use
Joonsoo Kimeefa8642014-12-12 16:55:46 -08001107 * section. (see page_ext.h about this.)
1108 */
1109 struct page_ext *page_ext;
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
Mel Gorman8a942fd2015-06-30 14:56:55 -07001224/*
1225 * During memory init memblocks map pfns to nids. The search is expensive and
1226 * this caches recent lookups. The implementation of __early_pfn_to_nid
1227 * may treat start/end as pfns or sections.
1228 */
1229struct mminit_pfnnid_cache {
1230 unsigned long last_start;
1231 unsigned long last_end;
1232 int last_nid;
1233};
1234
Andy Whitcroftd41dee32005-06-23 00:07:54 -07001235#ifndef early_pfn_valid
1236#define early_pfn_valid(pfn) (1)
1237#endif
1238
1239void memory_present(int nid, unsigned long start, unsigned long end);
1240unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
1241
Andy Whitcroft14e07292007-05-06 14:49:14 -07001242/*
1243 * If it is possible to have holes within a MAX_ORDER_NR_PAGES, then we
1244 * need to check pfn validility within that MAX_ORDER_NR_PAGES block.
1245 * pfn_valid_within() should be used in this case; we optimise this away
1246 * when we have no holes within a MAX_ORDER_NR_PAGES block.
1247 */
1248#ifdef CONFIG_HOLES_IN_ZONE
1249#define pfn_valid_within(pfn) pfn_valid(pfn)
1250#else
1251#define pfn_valid_within(pfn) (1)
1252#endif
1253
Mel Gormaneb335752009-05-13 17:34:48 +01001254#ifdef CONFIG_ARCH_HAS_HOLES_MEMORYMODEL
1255/*
1256 * pfn_valid() is meant to be able to tell if a given PFN has valid memmap
1257 * associated with it or not. In FLATMEM, it is expected that holes always
1258 * have valid memmap as long as there is valid PFNs either side of the hole.
1259 * In SPARSEMEM, it is assumed that a valid section has a memmap for the
1260 * entire section.
1261 *
1262 * However, an ARM, and maybe other embedded architectures in the future
1263 * free memmap backing holes to save memory on the assumption the memmap is
1264 * never used. The page_zone linkages are then broken even though pfn_valid()
1265 * returns true. A walker of the full memmap must then do this additional
1266 * check to ensure the memmap they are looking at is sane by making sure
1267 * the zone and PFN linkages are still valid. This is expensive, but walkers
1268 * of the full memmap are extremely rare.
1269 */
Yaowei Bai5b802872016-01-14 15:19:11 -08001270bool memmap_valid_within(unsigned long pfn,
Mel Gormaneb335752009-05-13 17:34:48 +01001271 struct page *page, struct zone *zone);
1272#else
Yaowei Bai5b802872016-01-14 15:19:11 -08001273static inline bool memmap_valid_within(unsigned long pfn,
Mel Gormaneb335752009-05-13 17:34:48 +01001274 struct page *page, struct zone *zone)
1275{
Yaowei Bai5b802872016-01-14 15:19:11 -08001276 return true;
Mel Gormaneb335752009-05-13 17:34:48 +01001277}
1278#endif /* CONFIG_ARCH_HAS_HOLES_MEMORYMODEL */
1279
Christoph Lameter97965472008-04-28 02:12:54 -07001280#endif /* !__GENERATING_BOUNDS.H */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001281#endif /* !__ASSEMBLY__ */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001282#endif /* _LINUX_MMZONE_H */