blob: f5fa3082fd6a56c628730c67873429f64f8d9cbb [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001#ifndef _LINUX_MMZONE_H
2#define _LINUX_MMZONE_H
3
4#ifdef __KERNEL__
5#ifndef __ASSEMBLY__
6
7#include <linux/config.h>
8#include <linux/spinlock.h>
9#include <linux/list.h>
10#include <linux/wait.h>
11#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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070016#include <asm/atomic.h>
17
18/* Free memory management - zoned buddy allocator. */
19#ifndef CONFIG_FORCE_MAX_ZONEORDER
20#define MAX_ORDER 11
21#else
22#define MAX_ORDER CONFIG_FORCE_MAX_ZONEORDER
23#endif
24
25struct free_area {
26 struct list_head free_list;
27 unsigned long nr_free;
28};
29
30struct pglist_data;
31
32/*
33 * zone->lock and zone->lru_lock are two of the hottest locks in the kernel.
34 * So add a wild amount of padding here to ensure that they fall into separate
35 * cachelines. There are very few zone structures in the machine, so space
36 * consumption is not a concern here.
37 */
38#if defined(CONFIG_SMP)
39struct zone_padding {
40 char x[0];
41} ____cacheline_maxaligned_in_smp;
42#define ZONE_PADDING(name) struct zone_padding name;
43#else
44#define ZONE_PADDING(name)
45#endif
46
47struct per_cpu_pages {
48 int count; /* number of pages in the list */
49 int low; /* low watermark, refill needed */
50 int high; /* high watermark, emptying needed */
51 int batch; /* chunk size for buddy add/remove */
52 struct list_head list; /* the list of pages */
53};
54
55struct per_cpu_pageset {
56 struct per_cpu_pages pcp[2]; /* 0: hot. 1: cold */
57#ifdef CONFIG_NUMA
58 unsigned long numa_hit; /* allocated in intended node */
59 unsigned long numa_miss; /* allocated in non intended node */
60 unsigned long numa_foreign; /* was intended here, hit elsewhere */
61 unsigned long interleave_hit; /* interleaver prefered this zone */
62 unsigned long local_node; /* allocation from local node */
63 unsigned long other_node; /* allocation from other node */
64#endif
65} ____cacheline_aligned_in_smp;
66
Christoph Lametere7c8d5c2005-06-21 17:14:47 -070067#ifdef CONFIG_NUMA
68#define zone_pcp(__z, __cpu) ((__z)->pageset[(__cpu)])
69#else
70#define zone_pcp(__z, __cpu) (&(__z)->pageset[(__cpu)])
71#endif
72
Linus Torvalds1da177e2005-04-16 15:20:36 -070073#define ZONE_DMA 0
74#define ZONE_NORMAL 1
75#define ZONE_HIGHMEM 2
76
77#define MAX_NR_ZONES 3 /* Sync this with ZONES_SHIFT */
78#define ZONES_SHIFT 2 /* ceil(log2(MAX_NR_ZONES)) */
79
80
81/*
82 * When a memory allocation must conform to specific limitations (such
83 * as being suitable for DMA) the caller will pass in hints to the
84 * allocator in the gfp_mask, in the zone modifier bits. These bits
85 * are used to select a priority ordered list of memory zones which
86 * match the requested limits. GFP_ZONEMASK defines which bits within
87 * the gfp_mask should be considered as zone modifiers. Each valid
88 * combination of the zone modifier bits has a corresponding list
89 * of zones (in node_zonelists). Thus for two zone modifiers there
90 * will be a maximum of 4 (2 ** 2) zonelists, for 3 modifiers there will
91 * be 8 (2 ** 3) zonelists. GFP_ZONETYPES defines the number of possible
92 * combinations of zone modifiers in "zone modifier space".
93 */
94#define GFP_ZONEMASK 0x03
95/*
96 * As an optimisation any zone modifier bits which are only valid when
97 * no other zone modifier bits are set (loners) should be placed in
98 * the highest order bits of this field. This allows us to reduce the
99 * extent of the zonelists thus saving space. For example in the case
100 * of three zone modifier bits, we could require up to eight zonelists.
101 * If the left most zone modifier is a "loner" then the highest valid
102 * zonelist would be four allowing us to allocate only five zonelists.
103 * Use the first form when the left most bit is not a "loner", otherwise
104 * use the second.
105 */
106/* #define GFP_ZONETYPES (GFP_ZONEMASK + 1) */ /* Non-loner */
107#define GFP_ZONETYPES ((GFP_ZONEMASK + 1) / 2 + 1) /* Loner */
108
109/*
110 * On machines where it is needed (eg PCs) we divide physical memory
111 * into multiple physical zones. On a PC we have 3 zones:
112 *
113 * ZONE_DMA < 16 MB ISA DMA capable memory
114 * ZONE_NORMAL 16-896 MB direct mapped by the kernel
115 * ZONE_HIGHMEM > 896 MB only page cache and user processes
116 */
117
118struct zone {
119 /* Fields commonly accessed by the page allocator */
120 unsigned long free_pages;
121 unsigned long pages_min, pages_low, pages_high;
122 /*
123 * We don't know if the memory that we're going to allocate will be freeable
124 * or/and it will be released eventually, so to avoid totally wasting several
125 * GB of ram we must reserve some of the lower zone memory (otherwise we risk
126 * to run OOM on the lower zones despite there's tons of freeable ram
127 * on the higher zones). This array is recalculated at runtime if the
128 * sysctl_lowmem_reserve_ratio sysctl changes.
129 */
130 unsigned long lowmem_reserve[MAX_NR_ZONES];
131
Christoph Lametere7c8d5c2005-06-21 17:14:47 -0700132#ifdef CONFIG_NUMA
133 struct per_cpu_pageset *pageset[NR_CPUS];
134#else
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135 struct per_cpu_pageset pageset[NR_CPUS];
Christoph Lametere7c8d5c2005-06-21 17:14:47 -0700136#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137 /*
138 * free areas of different sizes
139 */
140 spinlock_t lock;
Dave Hansenbdc8cb92005-10-29 18:16:53 -0700141#ifdef CONFIG_MEMORY_HOTPLUG
142 /* see spanned/present_pages for more description */
143 seqlock_t span_seqlock;
144#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145 struct free_area free_area[MAX_ORDER];
146
147
148 ZONE_PADDING(_pad1_)
149
150 /* Fields commonly accessed by the page reclaim scanner */
151 spinlock_t lru_lock;
152 struct list_head active_list;
153 struct list_head inactive_list;
154 unsigned long nr_scan_active;
155 unsigned long nr_scan_inactive;
156 unsigned long nr_active;
157 unsigned long nr_inactive;
158 unsigned long pages_scanned; /* since last reclaim */
159 int all_unreclaimable; /* All pages pinned */
160
161 /*
Martin Hicks753ee722005-06-21 17:14:41 -0700162 * Does the allocator try to reclaim pages from the zone as soon
163 * as it fails a watermark_ok() in __alloc_pages?
164 */
165 int reclaim_pages;
Martin Hicks1e7e5a92005-06-21 17:14:43 -0700166 /* A count of how many reclaimers are scanning this zone */
167 atomic_t reclaim_in_progress;
Martin Hicks753ee722005-06-21 17:14:41 -0700168
169 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700170 * prev_priority holds the scanning priority for this zone. It is
171 * defined as the scanning priority at which we achieved our reclaim
172 * target at the previous try_to_free_pages() or balance_pgdat()
173 * invokation.
174 *
175 * We use prev_priority as a measure of how much stress page reclaim is
176 * under - it drives the swappiness decision: whether to unmap mapped
177 * pages.
178 *
179 * temp_priority is used to remember the scanning priority at which
180 * this zone was successfully refilled to free_pages == pages_high.
181 *
182 * Access to both these fields is quite racy even on uniprocessor. But
183 * it is expected to average out OK.
184 */
185 int temp_priority;
186 int prev_priority;
187
188
189 ZONE_PADDING(_pad2_)
190 /* Rarely used or read-mostly fields */
191
192 /*
193 * wait_table -- the array holding the hash table
194 * wait_table_size -- the size of the hash table array
195 * wait_table_bits -- wait_table_size == (1 << wait_table_bits)
196 *
197 * The purpose of all these is to keep track of the people
198 * waiting for a page to become available and make them
199 * runnable again when possible. The trouble is that this
200 * consumes a lot of space, especially when so few things
201 * wait on pages at a given time. So instead of using
202 * per-page waitqueues, we use a waitqueue hash table.
203 *
204 * The bucket discipline is to sleep on the same queue when
205 * colliding and wake all in that wait queue when removing.
206 * When something wakes, it must check to be sure its page is
207 * truly available, a la thundering herd. The cost of a
208 * collision is great, but given the expected load of the
209 * table, they should be so rare as to be outweighed by the
210 * benefits from the saved space.
211 *
212 * __wait_on_page_locked() and unlock_page() in mm/filemap.c, are the
213 * primary users of these fields, and in mm/page_alloc.c
214 * free_area_init_core() performs the initialization of them.
215 */
216 wait_queue_head_t * wait_table;
217 unsigned long wait_table_size;
218 unsigned long wait_table_bits;
219
220 /*
221 * Discontig memory support fields.
222 */
223 struct pglist_data *zone_pgdat;
224 struct page *zone_mem_map;
225 /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */
226 unsigned long zone_start_pfn;
227
Dave Hansenbdc8cb92005-10-29 18:16:53 -0700228 /*
229 * zone_start_pfn, spanned_pages and present_pages are all
230 * protected by span_seqlock. It is a seqlock because it has
231 * to be read outside of zone->lock, and it is done in the main
232 * allocator path. But, it is written quite infrequently.
233 *
234 * The lock is declared along with zone->lock because it is
235 * frequently read in proximity to zone->lock. It's good to
236 * give them a chance of being in the same cacheline.
237 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238 unsigned long spanned_pages; /* total size, including holes */
239 unsigned long present_pages; /* amount of memory (excluding holes) */
240
241 /*
242 * rarely used fields:
243 */
244 char *name;
245} ____cacheline_maxaligned_in_smp;
246
247
248/*
249 * The "priority" of VM scanning is how much of the queues we will scan in one
250 * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the
251 * queues ("queue_length >> 12") during an aging round.
252 */
253#define DEF_PRIORITY 12
254
255/*
256 * One allocation request operates on a zonelist. A zonelist
257 * is a list of zones, the first one is the 'goal' of the
258 * allocation, the other zones are fallback zones, in decreasing
259 * priority.
260 *
261 * Right now a zonelist takes up less than a cacheline. We never
262 * modify it apart from boot-up, and only a few indices are used,
263 * so despite the zonelist table being relatively big, the cache
264 * footprint of this construct is very small.
265 */
266struct zonelist {
267 struct zone *zones[MAX_NUMNODES * MAX_NR_ZONES + 1]; // NULL delimited
268};
269
270
271/*
272 * The pg_data_t structure is used in machines with CONFIG_DISCONTIGMEM
273 * (mostly NUMA machines?) to denote a higher-level memory zone than the
274 * zone denotes.
275 *
276 * On NUMA machines, each NUMA node would have a pg_data_t to describe
277 * it's memory layout.
278 *
279 * Memory statistics and page replacement data structures are maintained on a
280 * per-zone basis.
281 */
282struct bootmem_data;
283typedef struct pglist_data {
284 struct zone node_zones[MAX_NR_ZONES];
285 struct zonelist node_zonelists[GFP_ZONETYPES];
286 int nr_zones;
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700287#ifdef CONFIG_FLAT_NODE_MEM_MAP
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288 struct page *node_mem_map;
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700289#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700290 struct bootmem_data *bdata;
Dave Hansen208d54e2005-10-29 18:16:52 -0700291#ifdef CONFIG_MEMORY_HOTPLUG
292 /*
293 * Must be held any time you expect node_start_pfn, node_present_pages
294 * or node_spanned_pages stay constant. Holding this will also
295 * guarantee that any pfn_valid() stays that way.
296 *
297 * Nests above zone->lock and zone->size_seqlock.
298 */
299 spinlock_t node_size_lock;
300#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700301 unsigned long node_start_pfn;
302 unsigned long node_present_pages; /* total number of physical pages */
303 unsigned long node_spanned_pages; /* total size of physical page
304 range, including holes */
305 int node_id;
306 struct pglist_data *pgdat_next;
307 wait_queue_head_t kswapd_wait;
308 struct task_struct *kswapd;
309 int kswapd_max_order;
310} pg_data_t;
311
312#define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages)
313#define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700314#ifdef CONFIG_FLAT_NODE_MEM_MAP
Dave Hansen408fde82005-06-23 00:07:37 -0700315#define pgdat_page_nr(pgdat, pagenr) ((pgdat)->node_mem_map + (pagenr))
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700316#else
317#define pgdat_page_nr(pgdat, pagenr) pfn_to_page((pgdat)->node_start_pfn + (pagenr))
318#endif
Dave Hansen408fde82005-06-23 00:07:37 -0700319#define nid_page_nr(nid, pagenr) pgdat_page_nr(NODE_DATA(nid),(pagenr))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700320
Dave Hansen208d54e2005-10-29 18:16:52 -0700321#include <linux/memory_hotplug.h>
322
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323extern struct pglist_data *pgdat_list;
324
325void __get_zone_counts(unsigned long *active, unsigned long *inactive,
326 unsigned long *free, struct pglist_data *pgdat);
327void get_zone_counts(unsigned long *active, unsigned long *inactive,
328 unsigned long *free);
329void build_all_zonelists(void);
330void wakeup_kswapd(struct zone *zone, int order);
331int zone_watermark_ok(struct zone *z, int order, unsigned long mark,
Al Viro260b2362005-10-21 03:22:44 -0400332 int alloc_type, int can_try_harder, gfp_t gfp_high);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333
334#ifdef CONFIG_HAVE_MEMORY_PRESENT
335void memory_present(int nid, unsigned long start, unsigned long end);
336#else
337static inline void memory_present(int nid, unsigned long start, unsigned long end) {}
338#endif
339
340#ifdef CONFIG_NEED_NODE_MEMMAP_SIZE
341unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
342#endif
343
344/*
345 * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc.
346 */
347#define zone_idx(zone) ((zone) - (zone)->zone_pgdat->node_zones)
348
349/**
350 * for_each_pgdat - helper macro to iterate over all nodes
351 * @pgdat - pointer to a pg_data_t variable
352 *
353 * Meant to help with common loops of the form
354 * pgdat = pgdat_list;
355 * while(pgdat) {
356 * ...
357 * pgdat = pgdat->pgdat_next;
358 * }
359 */
360#define for_each_pgdat(pgdat) \
361 for (pgdat = pgdat_list; pgdat; pgdat = pgdat->pgdat_next)
362
363/*
364 * next_zone - helper magic for for_each_zone()
365 * Thanks to William Lee Irwin III for this piece of ingenuity.
366 */
367static inline struct zone *next_zone(struct zone *zone)
368{
369 pg_data_t *pgdat = zone->zone_pgdat;
370
371 if (zone < pgdat->node_zones + MAX_NR_ZONES - 1)
372 zone++;
373 else if (pgdat->pgdat_next) {
374 pgdat = pgdat->pgdat_next;
375 zone = pgdat->node_zones;
376 } else
377 zone = NULL;
378
379 return zone;
380}
381
382/**
383 * for_each_zone - helper macro to iterate over all memory zones
384 * @zone - pointer to struct zone variable
385 *
386 * The user only needs to declare the zone variable, for_each_zone
387 * fills it in. This basically means for_each_zone() is an
388 * easier to read version of this piece of code:
389 *
390 * for (pgdat = pgdat_list; pgdat; pgdat = pgdat->node_next)
391 * for (i = 0; i < MAX_NR_ZONES; ++i) {
392 * struct zone * z = pgdat->node_zones + i;
393 * ...
394 * }
395 * }
396 */
397#define for_each_zone(zone) \
398 for (zone = pgdat_list->node_zones; zone; zone = next_zone(zone))
399
400static inline int is_highmem_idx(int idx)
401{
402 return (idx == ZONE_HIGHMEM);
403}
404
405static inline int is_normal_idx(int idx)
406{
407 return (idx == ZONE_NORMAL);
408}
409/**
410 * is_highmem - helper function to quickly check if a struct zone is a
411 * highmem zone or not. This is an attempt to keep references
412 * to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum.
413 * @zone - pointer to struct zone variable
414 */
415static inline int is_highmem(struct zone *zone)
416{
417 return zone == zone->zone_pgdat->node_zones + ZONE_HIGHMEM;
418}
419
420static inline int is_normal(struct zone *zone)
421{
422 return zone == zone->zone_pgdat->node_zones + ZONE_NORMAL;
423}
424
425/* These two functions are used to setup the per zone pages min values */
426struct ctl_table;
427struct file;
428int min_free_kbytes_sysctl_handler(struct ctl_table *, int, struct file *,
429 void __user *, size_t *, loff_t *);
430extern int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1];
431int lowmem_reserve_ratio_sysctl_handler(struct ctl_table *, int, struct file *,
432 void __user *, size_t *, loff_t *);
433
434#include <linux/topology.h>
435/* Returns the number of the current Node. */
Ingo Molnar39c715b2005-06-21 17:14:34 -0700436#define numa_node_id() (cpu_to_node(raw_smp_processor_id()))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437
Dave Hansen93b75042005-06-23 00:07:47 -0700438#ifndef CONFIG_NEED_MULTIPLE_NODES
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439
440extern struct pglist_data contig_page_data;
441#define NODE_DATA(nid) (&contig_page_data)
442#define NODE_MEM_MAP(nid) mem_map
443#define MAX_NODES_SHIFT 1
444#define pfn_to_nid(pfn) (0)
445
Dave Hansen93b75042005-06-23 00:07:47 -0700446#else /* CONFIG_NEED_MULTIPLE_NODES */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447
448#include <asm/mmzone.h>
449
Dave Hansen93b75042005-06-23 00:07:47 -0700450#endif /* !CONFIG_NEED_MULTIPLE_NODES */
Dave Hansen348f8b62005-06-23 00:07:40 -0700451
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700452#ifdef CONFIG_SPARSEMEM
453#include <asm/sparsemem.h>
454#endif
455
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456#if BITS_PER_LONG == 32 || defined(ARCH_HAS_ATOMIC_UNSIGNED)
457/*
458 * with 32 bit page->flags field, we reserve 8 bits for node/zone info.
459 * there are 3 zones (2 bits) and this leaves 8-2=6 bits for nodes.
460 */
Dave Hansen348f8b62005-06-23 00:07:40 -0700461#define FLAGS_RESERVED 8
462
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463#elif BITS_PER_LONG == 64
464/*
465 * with 64 bit flags field, there's plenty of room.
466 */
Dave Hansen348f8b62005-06-23 00:07:40 -0700467#define FLAGS_RESERVED 32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468
Dave Hansen348f8b62005-06-23 00:07:40 -0700469#else
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470
Dave Hansen348f8b62005-06-23 00:07:40 -0700471#error BITS_PER_LONG not defined
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473#endif
474
Andy Whitcroftb159d432005-06-23 00:07:52 -0700475#ifndef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
476#define early_pfn_to_nid(nid) (0UL)
477#endif
478
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700479#define pfn_to_section_nr(pfn) ((pfn) >> PFN_SECTION_SHIFT)
480#define section_nr_to_pfn(sec) ((sec) << PFN_SECTION_SHIFT)
481
482#ifdef CONFIG_SPARSEMEM
483
484/*
485 * SECTION_SHIFT #bits space required to store a section #
486 *
487 * PA_SECTION_SHIFT physical address to/from section number
488 * PFN_SECTION_SHIFT pfn to/from section number
489 */
490#define SECTIONS_SHIFT (MAX_PHYSMEM_BITS - SECTION_SIZE_BITS)
491
492#define PA_SECTION_SHIFT (SECTION_SIZE_BITS)
493#define PFN_SECTION_SHIFT (SECTION_SIZE_BITS - PAGE_SHIFT)
494
495#define NR_MEM_SECTIONS (1UL << SECTIONS_SHIFT)
496
497#define PAGES_PER_SECTION (1UL << PFN_SECTION_SHIFT)
498#define PAGE_SECTION_MASK (~(PAGES_PER_SECTION-1))
499
500#if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS
501#error Allocator MAX_ORDER exceeds SECTION_SIZE
502#endif
503
504struct page;
505struct mem_section {
Andy Whitcroft29751f62005-06-23 00:08:00 -0700506 /*
507 * This is, logically, a pointer to an array of struct
508 * pages. However, it is stored with some other magic.
509 * (see sparse.c::sparse_init_one_section())
510 *
511 * Making it a UL at least makes someone do a cast
512 * before using it wrong.
513 */
514 unsigned long section_mem_map;
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700515};
516
Bob Picco3e347262005-09-03 15:54:28 -0700517#ifdef CONFIG_SPARSEMEM_EXTREME
518#define SECTIONS_PER_ROOT (PAGE_SIZE / sizeof (struct mem_section))
Bob Picco802f1922005-09-03 15:54:26 -0700519#else
Bob Picco3e347262005-09-03 15:54:28 -0700520#define SECTIONS_PER_ROOT 1
521#endif
Bob Picco802f1922005-09-03 15:54:26 -0700522
Bob Picco3e347262005-09-03 15:54:28 -0700523#define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT)
524#define NR_SECTION_ROOTS (NR_MEM_SECTIONS / SECTIONS_PER_ROOT)
525#define SECTION_ROOT_MASK (SECTIONS_PER_ROOT - 1)
526
527#ifdef CONFIG_SPARSEMEM_EXTREME
528extern struct mem_section *mem_section[NR_SECTION_ROOTS];
529#else
530extern struct mem_section mem_section[NR_SECTION_ROOTS][SECTIONS_PER_ROOT];
531#endif
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700532
Andy Whitcroft29751f62005-06-23 00:08:00 -0700533static inline struct mem_section *__nr_to_section(unsigned long nr)
534{
Bob Picco3e347262005-09-03 15:54:28 -0700535 if (!mem_section[SECTION_NR_TO_ROOT(nr)])
536 return NULL;
537 return &mem_section[SECTION_NR_TO_ROOT(nr)][nr & SECTION_ROOT_MASK];
Andy Whitcroft29751f62005-06-23 00:08:00 -0700538}
Dave Hansen4ca644d2005-10-29 18:16:51 -0700539extern int __section_nr(struct mem_section* ms);
Andy Whitcroft29751f62005-06-23 00:08:00 -0700540
541/*
542 * We use the lower bits of the mem_map pointer to store
543 * a little bit of information. There should be at least
544 * 3 bits here due to 32-bit alignment.
545 */
546#define SECTION_MARKED_PRESENT (1UL<<0)
547#define SECTION_HAS_MEM_MAP (1UL<<1)
548#define SECTION_MAP_LAST_BIT (1UL<<2)
549#define SECTION_MAP_MASK (~(SECTION_MAP_LAST_BIT-1))
550
551static inline struct page *__section_mem_map_addr(struct mem_section *section)
552{
553 unsigned long map = section->section_mem_map;
554 map &= SECTION_MAP_MASK;
555 return (struct page *)map;
556}
557
558static inline int valid_section(struct mem_section *section)
559{
Bob Picco802f1922005-09-03 15:54:26 -0700560 return (section && (section->section_mem_map & SECTION_MARKED_PRESENT));
Andy Whitcroft29751f62005-06-23 00:08:00 -0700561}
562
563static inline int section_has_mem_map(struct mem_section *section)
564{
Bob Picco802f1922005-09-03 15:54:26 -0700565 return (section && (section->section_mem_map & SECTION_HAS_MEM_MAP));
Andy Whitcroft29751f62005-06-23 00:08:00 -0700566}
567
568static inline int valid_section_nr(unsigned long nr)
569{
570 return valid_section(__nr_to_section(nr));
571}
572
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700573/*
574 * Given a kernel address, find the home node of the underlying memory.
575 */
576#define kvaddr_to_nid(kaddr) pfn_to_nid(__pa(kaddr) >> PAGE_SHIFT)
577
578static inline struct mem_section *__pfn_to_section(unsigned long pfn)
579{
Andy Whitcroft29751f62005-06-23 00:08:00 -0700580 return __nr_to_section(pfn_to_section_nr(pfn));
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700581}
582
583#define pfn_to_page(pfn) \
584({ \
585 unsigned long __pfn = (pfn); \
Andy Whitcroft29751f62005-06-23 00:08:00 -0700586 __section_mem_map_addr(__pfn_to_section(__pfn)) + __pfn; \
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700587})
588#define page_to_pfn(page) \
589({ \
Andy Whitcroft29751f62005-06-23 00:08:00 -0700590 page - __section_mem_map_addr(__nr_to_section( \
591 page_to_section(page))); \
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700592})
593
594static inline int pfn_valid(unsigned long pfn)
595{
596 if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS)
597 return 0;
Andy Whitcroft29751f62005-06-23 00:08:00 -0700598 return valid_section(__nr_to_section(pfn_to_section_nr(pfn)));
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700599}
600
601/*
602 * These are _only_ used during initialisation, therefore they
603 * can use __initdata ... They could have names to indicate
604 * this restriction.
605 */
606#ifdef CONFIG_NUMA
607#define pfn_to_nid early_pfn_to_nid
608#endif
609
610#define pfn_to_pgdat(pfn) \
611({ \
612 NODE_DATA(pfn_to_nid(pfn)); \
613})
614
615#define early_pfn_valid(pfn) pfn_valid(pfn)
616void sparse_init(void);
617#else
618#define sparse_init() do {} while (0)
Dave Hansen28ae55c2005-09-03 15:54:29 -0700619#define sparse_index_init(_sec, _nid) do {} while (0)
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700620#endif /* CONFIG_SPARSEMEM */
621
Andy Whitcroft641c7672005-06-23 00:07:59 -0700622#ifdef CONFIG_NODES_SPAN_OTHER_NODES
623#define early_pfn_in_nid(pfn, nid) (early_pfn_to_nid(pfn) == (nid))
624#else
625#define early_pfn_in_nid(pfn, nid) (1)
626#endif
627
Andy Whitcroftd41dee32005-06-23 00:07:54 -0700628#ifndef early_pfn_valid
629#define early_pfn_valid(pfn) (1)
630#endif
631
632void memory_present(int nid, unsigned long start, unsigned long end);
633unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long);
634
Linus Torvalds1da177e2005-04-16 15:20:36 -0700635#endif /* !__ASSEMBLY__ */
636#endif /* __KERNEL__ */
637#endif /* _LINUX_MMZONE_H */