]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | #ifndef _LINUX_MMZONE_H |
2 | #define _LINUX_MMZONE_H | |
3 | ||
4 | #ifdef __KERNEL__ | |
5 | #ifndef __ASSEMBLY__ | |
6 | ||
1da177e4 LT |
7 | #include <linux/spinlock.h> |
8 | #include <linux/list.h> | |
9 | #include <linux/wait.h> | |
10 | #include <linux/cache.h> | |
11 | #include <linux/threads.h> | |
12 | #include <linux/numa.h> | |
13 | #include <linux/init.h> | |
bdc8cb98 | 14 | #include <linux/seqlock.h> |
8357f869 | 15 | #include <linux/nodemask.h> |
1da177e4 | 16 | #include <asm/atomic.h> |
93ff66bf | 17 | #include <asm/page.h> |
1da177e4 LT |
18 | |
19 | /* Free memory management - zoned buddy allocator. */ | |
20 | #ifndef CONFIG_FORCE_MAX_ZONEORDER | |
21 | #define MAX_ORDER 11 | |
22 | #else | |
23 | #define MAX_ORDER CONFIG_FORCE_MAX_ZONEORDER | |
24 | #endif | |
e984bb43 | 25 | #define MAX_ORDER_NR_PAGES (1 << (MAX_ORDER - 1)) |
1da177e4 LT |
26 | |
27 | struct free_area { | |
28 | struct list_head free_list; | |
29 | unsigned long nr_free; | |
30 | }; | |
31 | ||
32 | struct pglist_data; | |
33 | ||
34 | /* | |
35 | * zone->lock and zone->lru_lock are two of the hottest locks in the kernel. | |
36 | * So add a wild amount of padding here to ensure that they fall into separate | |
37 | * cachelines. There are very few zone structures in the machine, so space | |
38 | * consumption is not a concern here. | |
39 | */ | |
40 | #if defined(CONFIG_SMP) | |
41 | struct zone_padding { | |
42 | char x[0]; | |
22fc6ecc | 43 | } ____cacheline_internodealigned_in_smp; |
1da177e4 LT |
44 | #define ZONE_PADDING(name) struct zone_padding name; |
45 | #else | |
46 | #define ZONE_PADDING(name) | |
47 | #endif | |
48 | ||
2244b95a | 49 | enum zone_stat_item { |
51ed4491 | 50 | /* First 128 byte cacheline (assuming 64 bit words) */ |
d23ad423 | 51 | NR_FREE_PAGES, |
c8785385 CL |
52 | NR_INACTIVE, |
53 | NR_ACTIVE, | |
f3dbd344 CL |
54 | NR_ANON_PAGES, /* Mapped anonymous pages */ |
55 | NR_FILE_MAPPED, /* pagecache pages mapped into pagetables. | |
65ba55f5 | 56 | only modified from process context */ |
347ce434 | 57 | NR_FILE_PAGES, |
b1e7a8fd | 58 | NR_FILE_DIRTY, |
ce866b34 | 59 | NR_WRITEBACK, |
51ed4491 CL |
60 | /* Second 128 byte cacheline */ |
61 | NR_SLAB_RECLAIMABLE, | |
62 | NR_SLAB_UNRECLAIMABLE, | |
63 | NR_PAGETABLE, /* used for pagetables */ | |
fd39fc85 | 64 | NR_UNSTABLE_NFS, /* NFS unstable pages */ |
d2c5e30c | 65 | NR_BOUNCE, |
e129b5c2 | 66 | NR_VMSCAN_WRITE, |
ca889e6c CL |
67 | #ifdef CONFIG_NUMA |
68 | NUMA_HIT, /* allocated in intended node */ | |
69 | NUMA_MISS, /* allocated in non intended node */ | |
70 | NUMA_FOREIGN, /* was intended here, hit elsewhere */ | |
71 | NUMA_INTERLEAVE_HIT, /* interleaver preferred this zone */ | |
72 | NUMA_LOCAL, /* allocation from local node */ | |
73 | NUMA_OTHER, /* allocation from other node */ | |
74 | #endif | |
2244b95a CL |
75 | NR_VM_ZONE_STAT_ITEMS }; |
76 | ||
1da177e4 LT |
77 | struct per_cpu_pages { |
78 | int count; /* number of pages in the list */ | |
1da177e4 LT |
79 | int high; /* high watermark, emptying needed */ |
80 | int batch; /* chunk size for buddy add/remove */ | |
81 | struct list_head list; /* the list of pages */ | |
82 | }; | |
83 | ||
84 | struct per_cpu_pageset { | |
85 | struct per_cpu_pages pcp[2]; /* 0: hot. 1: cold */ | |
2244b95a | 86 | #ifdef CONFIG_SMP |
df9ecaba | 87 | s8 stat_threshold; |
2244b95a CL |
88 | s8 vm_stat_diff[NR_VM_ZONE_STAT_ITEMS]; |
89 | #endif | |
1da177e4 LT |
90 | } ____cacheline_aligned_in_smp; |
91 | ||
e7c8d5c9 CL |
92 | #ifdef CONFIG_NUMA |
93 | #define zone_pcp(__z, __cpu) ((__z)->pageset[(__cpu)]) | |
94 | #else | |
95 | #define zone_pcp(__z, __cpu) (&(__z)->pageset[(__cpu)]) | |
96 | #endif | |
97 | ||
2f1b6248 | 98 | enum zone_type { |
4b51d669 | 99 | #ifdef CONFIG_ZONE_DMA |
2f1b6248 CL |
100 | /* |
101 | * ZONE_DMA is used when there are devices that are not able | |
102 | * to do DMA to all of addressable memory (ZONE_NORMAL). Then we | |
103 | * carve out the portion of memory that is needed for these devices. | |
104 | * The range is arch specific. | |
105 | * | |
106 | * Some examples | |
107 | * | |
108 | * Architecture Limit | |
109 | * --------------------------- | |
110 | * parisc, ia64, sparc <4G | |
111 | * s390 <2G | |
112 | * arm26 <48M | |
113 | * arm Various | |
114 | * alpha Unlimited or 0-16MB. | |
115 | * | |
116 | * i386, x86_64 and multiple other arches | |
117 | * <16M. | |
118 | */ | |
119 | ZONE_DMA, | |
4b51d669 | 120 | #endif |
fb0e7942 | 121 | #ifdef CONFIG_ZONE_DMA32 |
2f1b6248 CL |
122 | /* |
123 | * x86_64 needs two ZONE_DMAs because it supports devices that are | |
124 | * only able to do DMA to the lower 16M but also 32 bit devices that | |
125 | * can only do DMA areas below 4G. | |
126 | */ | |
127 | ZONE_DMA32, | |
fb0e7942 | 128 | #endif |
2f1b6248 CL |
129 | /* |
130 | * Normal addressable memory is in ZONE_NORMAL. DMA operations can be | |
131 | * performed on pages in ZONE_NORMAL if the DMA devices support | |
132 | * transfers to all addressable memory. | |
133 | */ | |
134 | ZONE_NORMAL, | |
e53ef38d | 135 | #ifdef CONFIG_HIGHMEM |
2f1b6248 CL |
136 | /* |
137 | * A memory area that is only addressable by the kernel through | |
138 | * mapping portions into its own address space. This is for example | |
139 | * used by i386 to allow the kernel to address the memory beyond | |
140 | * 900MB. The kernel will set up special mappings (page | |
141 | * table entries on i386) for each page that the kernel needs to | |
142 | * access. | |
143 | */ | |
144 | ZONE_HIGHMEM, | |
e53ef38d | 145 | #endif |
2f1b6248 CL |
146 | MAX_NR_ZONES |
147 | }; | |
1da177e4 | 148 | |
1da177e4 LT |
149 | /* |
150 | * When a memory allocation must conform to specific limitations (such | |
151 | * as being suitable for DMA) the caller will pass in hints to the | |
152 | * allocator in the gfp_mask, in the zone modifier bits. These bits | |
153 | * are used to select a priority ordered list of memory zones which | |
19655d34 | 154 | * match the requested limits. See gfp_zone() in include/linux/gfp.h |
1da177e4 | 155 | */ |
fb0e7942 | 156 | |
4b51d669 CL |
157 | /* |
158 | * Count the active zones. Note that the use of defined(X) outside | |
159 | * #if and family is not necessarily defined so ensure we cannot use | |
160 | * it later. Use __ZONE_COUNT to work out how many shift bits we need. | |
161 | */ | |
162 | #define __ZONE_COUNT ( \ | |
163 | defined(CONFIG_ZONE_DMA) \ | |
164 | + defined(CONFIG_ZONE_DMA32) \ | |
165 | + 1 \ | |
166 | + defined(CONFIG_HIGHMEM) \ | |
167 | ) | |
168 | #if __ZONE_COUNT < 2 | |
169 | #define ZONES_SHIFT 0 | |
170 | #elif __ZONE_COUNT <= 2 | |
19655d34 | 171 | #define ZONES_SHIFT 1 |
4b51d669 | 172 | #elif __ZONE_COUNT <= 4 |
19655d34 | 173 | #define ZONES_SHIFT 2 |
4b51d669 CL |
174 | #else |
175 | #error ZONES_SHIFT -- too many zones configured adjust calculation | |
fb0e7942 | 176 | #endif |
4b51d669 | 177 | #undef __ZONE_COUNT |
1da177e4 | 178 | |
1da177e4 LT |
179 | struct zone { |
180 | /* Fields commonly accessed by the page allocator */ | |
1da177e4 LT |
181 | unsigned long pages_min, pages_low, pages_high; |
182 | /* | |
183 | * We don't know if the memory that we're going to allocate will be freeable | |
184 | * or/and it will be released eventually, so to avoid totally wasting several | |
185 | * GB of ram we must reserve some of the lower zone memory (otherwise we risk | |
186 | * to run OOM on the lower zones despite there's tons of freeable ram | |
187 | * on the higher zones). This array is recalculated at runtime if the | |
188 | * sysctl_lowmem_reserve_ratio sysctl changes. | |
189 | */ | |
190 | unsigned long lowmem_reserve[MAX_NR_ZONES]; | |
191 | ||
e7c8d5c9 | 192 | #ifdef CONFIG_NUMA |
d5f541ed | 193 | int node; |
9614634f CL |
194 | /* |
195 | * zone reclaim becomes active if more unmapped pages exist. | |
196 | */ | |
8417bba4 | 197 | unsigned long min_unmapped_pages; |
0ff38490 | 198 | unsigned long min_slab_pages; |
e7c8d5c9 CL |
199 | struct per_cpu_pageset *pageset[NR_CPUS]; |
200 | #else | |
1da177e4 | 201 | struct per_cpu_pageset pageset[NR_CPUS]; |
e7c8d5c9 | 202 | #endif |
1da177e4 LT |
203 | /* |
204 | * free areas of different sizes | |
205 | */ | |
206 | spinlock_t lock; | |
bdc8cb98 DH |
207 | #ifdef CONFIG_MEMORY_HOTPLUG |
208 | /* see spanned/present_pages for more description */ | |
209 | seqlock_t span_seqlock; | |
210 | #endif | |
1da177e4 LT |
211 | struct free_area free_area[MAX_ORDER]; |
212 | ||
213 | ||
214 | ZONE_PADDING(_pad1_) | |
215 | ||
216 | /* Fields commonly accessed by the page reclaim scanner */ | |
217 | spinlock_t lru_lock; | |
218 | struct list_head active_list; | |
219 | struct list_head inactive_list; | |
220 | unsigned long nr_scan_active; | |
221 | unsigned long nr_scan_inactive; | |
1da177e4 LT |
222 | unsigned long pages_scanned; /* since last reclaim */ |
223 | int all_unreclaimable; /* All pages pinned */ | |
224 | ||
1e7e5a90 MH |
225 | /* A count of how many reclaimers are scanning this zone */ |
226 | atomic_t reclaim_in_progress; | |
753ee728 | 227 | |
2244b95a CL |
228 | /* Zone statistics */ |
229 | atomic_long_t vm_stat[NR_VM_ZONE_STAT_ITEMS]; | |
9eeff239 | 230 | |
1da177e4 LT |
231 | /* |
232 | * prev_priority holds the scanning priority for this zone. It is | |
233 | * defined as the scanning priority at which we achieved our reclaim | |
234 | * target at the previous try_to_free_pages() or balance_pgdat() | |
235 | * invokation. | |
236 | * | |
237 | * We use prev_priority as a measure of how much stress page reclaim is | |
238 | * under - it drives the swappiness decision: whether to unmap mapped | |
239 | * pages. | |
240 | * | |
3bb1a852 | 241 | * Access to both this field is quite racy even on uniprocessor. But |
1da177e4 LT |
242 | * it is expected to average out OK. |
243 | */ | |
1da177e4 LT |
244 | int prev_priority; |
245 | ||
246 | ||
247 | ZONE_PADDING(_pad2_) | |
248 | /* Rarely used or read-mostly fields */ | |
249 | ||
250 | /* | |
251 | * wait_table -- the array holding the hash table | |
02b694de | 252 | * wait_table_hash_nr_entries -- the size of the hash table array |
1da177e4 LT |
253 | * wait_table_bits -- wait_table_size == (1 << wait_table_bits) |
254 | * | |
255 | * The purpose of all these is to keep track of the people | |
256 | * waiting for a page to become available and make them | |
257 | * runnable again when possible. The trouble is that this | |
258 | * consumes a lot of space, especially when so few things | |
259 | * wait on pages at a given time. So instead of using | |
260 | * per-page waitqueues, we use a waitqueue hash table. | |
261 | * | |
262 | * The bucket discipline is to sleep on the same queue when | |
263 | * colliding and wake all in that wait queue when removing. | |
264 | * When something wakes, it must check to be sure its page is | |
265 | * truly available, a la thundering herd. The cost of a | |
266 | * collision is great, but given the expected load of the | |
267 | * table, they should be so rare as to be outweighed by the | |
268 | * benefits from the saved space. | |
269 | * | |
270 | * __wait_on_page_locked() and unlock_page() in mm/filemap.c, are the | |
271 | * primary users of these fields, and in mm/page_alloc.c | |
272 | * free_area_init_core() performs the initialization of them. | |
273 | */ | |
274 | wait_queue_head_t * wait_table; | |
02b694de | 275 | unsigned long wait_table_hash_nr_entries; |
1da177e4 LT |
276 | unsigned long wait_table_bits; |
277 | ||
278 | /* | |
279 | * Discontig memory support fields. | |
280 | */ | |
281 | struct pglist_data *zone_pgdat; | |
1da177e4 LT |
282 | /* zone_start_pfn == zone_start_paddr >> PAGE_SHIFT */ |
283 | unsigned long zone_start_pfn; | |
284 | ||
bdc8cb98 DH |
285 | /* |
286 | * zone_start_pfn, spanned_pages and present_pages are all | |
287 | * protected by span_seqlock. It is a seqlock because it has | |
288 | * to be read outside of zone->lock, and it is done in the main | |
289 | * allocator path. But, it is written quite infrequently. | |
290 | * | |
291 | * The lock is declared along with zone->lock because it is | |
292 | * frequently read in proximity to zone->lock. It's good to | |
293 | * give them a chance of being in the same cacheline. | |
294 | */ | |
1da177e4 LT |
295 | unsigned long spanned_pages; /* total size, including holes */ |
296 | unsigned long present_pages; /* amount of memory (excluding holes) */ | |
297 | ||
298 | /* | |
299 | * rarely used fields: | |
300 | */ | |
15ad7cdc | 301 | const char *name; |
22fc6ecc | 302 | } ____cacheline_internodealigned_in_smp; |
1da177e4 | 303 | |
1da177e4 LT |
304 | /* |
305 | * The "priority" of VM scanning is how much of the queues we will scan in one | |
306 | * go. A value of 12 for DEF_PRIORITY implies that we will scan 1/4096th of the | |
307 | * queues ("queue_length >> 12") during an aging round. | |
308 | */ | |
309 | #define DEF_PRIORITY 12 | |
310 | ||
9276b1bc PJ |
311 | /* Maximum number of zones on a zonelist */ |
312 | #define MAX_ZONES_PER_ZONELIST (MAX_NUMNODES * MAX_NR_ZONES) | |
313 | ||
314 | #ifdef CONFIG_NUMA | |
315 | /* | |
316 | * We cache key information from each zonelist for smaller cache | |
317 | * footprint when scanning for free pages in get_page_from_freelist(). | |
318 | * | |
319 | * 1) The BITMAP fullzones tracks which zones in a zonelist have come | |
320 | * up short of free memory since the last time (last_fullzone_zap) | |
321 | * we zero'd fullzones. | |
322 | * 2) The array z_to_n[] maps each zone in the zonelist to its node | |
323 | * id, so that we can efficiently evaluate whether that node is | |
324 | * set in the current tasks mems_allowed. | |
325 | * | |
326 | * Both fullzones and z_to_n[] are one-to-one with the zonelist, | |
327 | * indexed by a zones offset in the zonelist zones[] array. | |
328 | * | |
329 | * The get_page_from_freelist() routine does two scans. During the | |
330 | * first scan, we skip zones whose corresponding bit in 'fullzones' | |
331 | * is set or whose corresponding node in current->mems_allowed (which | |
332 | * comes from cpusets) is not set. During the second scan, we bypass | |
333 | * this zonelist_cache, to ensure we look methodically at each zone. | |
334 | * | |
335 | * Once per second, we zero out (zap) fullzones, forcing us to | |
336 | * reconsider nodes that might have regained more free memory. | |
337 | * The field last_full_zap is the time we last zapped fullzones. | |
338 | * | |
339 | * This mechanism reduces the amount of time we waste repeatedly | |
340 | * reexaming zones for free memory when they just came up low on | |
341 | * memory momentarilly ago. | |
342 | * | |
343 | * The zonelist_cache struct members logically belong in struct | |
344 | * zonelist. However, the mempolicy zonelists constructed for | |
345 | * MPOL_BIND are intentionally variable length (and usually much | |
346 | * shorter). A general purpose mechanism for handling structs with | |
347 | * multiple variable length members is more mechanism than we want | |
348 | * here. We resort to some special case hackery instead. | |
349 | * | |
350 | * The MPOL_BIND zonelists don't need this zonelist_cache (in good | |
351 | * part because they are shorter), so we put the fixed length stuff | |
352 | * at the front of the zonelist struct, ending in a variable length | |
353 | * zones[], as is needed by MPOL_BIND. | |
354 | * | |
355 | * Then we put the optional zonelist cache on the end of the zonelist | |
356 | * struct. This optional stuff is found by a 'zlcache_ptr' pointer in | |
357 | * the fixed length portion at the front of the struct. This pointer | |
358 | * both enables us to find the zonelist cache, and in the case of | |
359 | * MPOL_BIND zonelists, (which will just set the zlcache_ptr to NULL) | |
360 | * to know that the zonelist cache is not there. | |
361 | * | |
362 | * The end result is that struct zonelists come in two flavors: | |
363 | * 1) The full, fixed length version, shown below, and | |
364 | * 2) The custom zonelists for MPOL_BIND. | |
365 | * The custom MPOL_BIND zonelists have a NULL zlcache_ptr and no zlcache. | |
366 | * | |
367 | * Even though there may be multiple CPU cores on a node modifying | |
368 | * fullzones or last_full_zap in the same zonelist_cache at the same | |
369 | * time, we don't lock it. This is just hint data - if it is wrong now | |
370 | * and then, the allocator will still function, perhaps a bit slower. | |
371 | */ | |
372 | ||
373 | ||
374 | struct zonelist_cache { | |
9276b1bc | 375 | unsigned short z_to_n[MAX_ZONES_PER_ZONELIST]; /* zone->nid */ |
7253f4ef | 376 | DECLARE_BITMAP(fullzones, MAX_ZONES_PER_ZONELIST); /* zone full? */ |
9276b1bc PJ |
377 | unsigned long last_full_zap; /* when last zap'd (jiffies) */ |
378 | }; | |
379 | #else | |
380 | struct zonelist_cache; | |
381 | #endif | |
382 | ||
1da177e4 LT |
383 | /* |
384 | * One allocation request operates on a zonelist. A zonelist | |
385 | * is a list of zones, the first one is the 'goal' of the | |
386 | * allocation, the other zones are fallback zones, in decreasing | |
387 | * priority. | |
388 | * | |
9276b1bc PJ |
389 | * If zlcache_ptr is not NULL, then it is just the address of zlcache, |
390 | * as explained above. If zlcache_ptr is NULL, there is no zlcache. | |
1da177e4 | 391 | */ |
9276b1bc | 392 | |
1da177e4 | 393 | struct zonelist { |
9276b1bc PJ |
394 | struct zonelist_cache *zlcache_ptr; // NULL or &zlcache |
395 | struct zone *zones[MAX_ZONES_PER_ZONELIST + 1]; // NULL delimited | |
396 | #ifdef CONFIG_NUMA | |
397 | struct zonelist_cache zlcache; // optional ... | |
398 | #endif | |
1da177e4 LT |
399 | }; |
400 | ||
c713216d MG |
401 | #ifdef CONFIG_ARCH_POPULATES_NODE_MAP |
402 | struct node_active_region { | |
403 | unsigned long start_pfn; | |
404 | unsigned long end_pfn; | |
405 | int nid; | |
406 | }; | |
407 | #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */ | |
1da177e4 | 408 | |
5b99cd0e HC |
409 | #ifndef CONFIG_DISCONTIGMEM |
410 | /* The array of struct pages - for discontigmem use pgdat->lmem_map */ | |
411 | extern struct page *mem_map; | |
412 | #endif | |
413 | ||
1da177e4 LT |
414 | /* |
415 | * The pg_data_t structure is used in machines with CONFIG_DISCONTIGMEM | |
416 | * (mostly NUMA machines?) to denote a higher-level memory zone than the | |
417 | * zone denotes. | |
418 | * | |
419 | * On NUMA machines, each NUMA node would have a pg_data_t to describe | |
420 | * it's memory layout. | |
421 | * | |
422 | * Memory statistics and page replacement data structures are maintained on a | |
423 | * per-zone basis. | |
424 | */ | |
425 | struct bootmem_data; | |
426 | typedef struct pglist_data { | |
427 | struct zone node_zones[MAX_NR_ZONES]; | |
19655d34 | 428 | struct zonelist node_zonelists[MAX_NR_ZONES]; |
1da177e4 | 429 | int nr_zones; |
d41dee36 | 430 | #ifdef CONFIG_FLAT_NODE_MEM_MAP |
1da177e4 | 431 | struct page *node_mem_map; |
d41dee36 | 432 | #endif |
1da177e4 | 433 | struct bootmem_data *bdata; |
208d54e5 DH |
434 | #ifdef CONFIG_MEMORY_HOTPLUG |
435 | /* | |
436 | * Must be held any time you expect node_start_pfn, node_present_pages | |
437 | * or node_spanned_pages stay constant. Holding this will also | |
438 | * guarantee that any pfn_valid() stays that way. | |
439 | * | |
440 | * Nests above zone->lock and zone->size_seqlock. | |
441 | */ | |
442 | spinlock_t node_size_lock; | |
443 | #endif | |
1da177e4 LT |
444 | unsigned long node_start_pfn; |
445 | unsigned long node_present_pages; /* total number of physical pages */ | |
446 | unsigned long node_spanned_pages; /* total size of physical page | |
447 | range, including holes */ | |
448 | int node_id; | |
1da177e4 LT |
449 | wait_queue_head_t kswapd_wait; |
450 | struct task_struct *kswapd; | |
451 | int kswapd_max_order; | |
452 | } pg_data_t; | |
453 | ||
454 | #define node_present_pages(nid) (NODE_DATA(nid)->node_present_pages) | |
455 | #define node_spanned_pages(nid) (NODE_DATA(nid)->node_spanned_pages) | |
d41dee36 | 456 | #ifdef CONFIG_FLAT_NODE_MEM_MAP |
408fde81 | 457 | #define pgdat_page_nr(pgdat, pagenr) ((pgdat)->node_mem_map + (pagenr)) |
d41dee36 AW |
458 | #else |
459 | #define pgdat_page_nr(pgdat, pagenr) pfn_to_page((pgdat)->node_start_pfn + (pagenr)) | |
460 | #endif | |
408fde81 | 461 | #define nid_page_nr(nid, pagenr) pgdat_page_nr(NODE_DATA(nid),(pagenr)) |
1da177e4 | 462 | |
208d54e5 DH |
463 | #include <linux/memory_hotplug.h> |
464 | ||
1da177e4 LT |
465 | void get_zone_counts(unsigned long *active, unsigned long *inactive, |
466 | unsigned long *free); | |
467 | void build_all_zonelists(void); | |
468 | void wakeup_kswapd(struct zone *zone, int order); | |
469 | int zone_watermark_ok(struct zone *z, int order, unsigned long mark, | |
7fb1d9fc | 470 | int classzone_idx, int alloc_flags); |
a2f3aa02 DH |
471 | enum memmap_context { |
472 | MEMMAP_EARLY, | |
473 | MEMMAP_HOTPLUG, | |
474 | }; | |
718127cc | 475 | extern int init_currently_empty_zone(struct zone *zone, unsigned long start_pfn, |
a2f3aa02 DH |
476 | unsigned long size, |
477 | enum memmap_context context); | |
718127cc | 478 | |
1da177e4 LT |
479 | #ifdef CONFIG_HAVE_MEMORY_PRESENT |
480 | void memory_present(int nid, unsigned long start, unsigned long end); | |
481 | #else | |
482 | static inline void memory_present(int nid, unsigned long start, unsigned long end) {} | |
483 | #endif | |
484 | ||
485 | #ifdef CONFIG_NEED_NODE_MEMMAP_SIZE | |
486 | unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long); | |
487 | #endif | |
488 | ||
489 | /* | |
490 | * zone_idx() returns 0 for the ZONE_DMA zone, 1 for the ZONE_NORMAL zone, etc. | |
491 | */ | |
492 | #define zone_idx(zone) ((zone) - (zone)->zone_pgdat->node_zones) | |
493 | ||
f3fe6512 CK |
494 | static inline int populated_zone(struct zone *zone) |
495 | { | |
496 | return (!!zone->present_pages); | |
497 | } | |
498 | ||
2f1b6248 | 499 | static inline int is_highmem_idx(enum zone_type idx) |
1da177e4 | 500 | { |
e53ef38d | 501 | #ifdef CONFIG_HIGHMEM |
1da177e4 | 502 | return (idx == ZONE_HIGHMEM); |
e53ef38d CL |
503 | #else |
504 | return 0; | |
505 | #endif | |
1da177e4 LT |
506 | } |
507 | ||
2f1b6248 | 508 | static inline int is_normal_idx(enum zone_type idx) |
1da177e4 LT |
509 | { |
510 | return (idx == ZONE_NORMAL); | |
511 | } | |
9328b8fa | 512 | |
1da177e4 LT |
513 | /** |
514 | * is_highmem - helper function to quickly check if a struct zone is a | |
515 | * highmem zone or not. This is an attempt to keep references | |
516 | * to ZONE_{DMA/NORMAL/HIGHMEM/etc} in general code to a minimum. | |
517 | * @zone - pointer to struct zone variable | |
518 | */ | |
519 | static inline int is_highmem(struct zone *zone) | |
520 | { | |
e53ef38d | 521 | #ifdef CONFIG_HIGHMEM |
1da177e4 | 522 | return zone == zone->zone_pgdat->node_zones + ZONE_HIGHMEM; |
e53ef38d CL |
523 | #else |
524 | return 0; | |
525 | #endif | |
1da177e4 LT |
526 | } |
527 | ||
528 | static inline int is_normal(struct zone *zone) | |
529 | { | |
530 | return zone == zone->zone_pgdat->node_zones + ZONE_NORMAL; | |
531 | } | |
532 | ||
9328b8fa NP |
533 | static inline int is_dma32(struct zone *zone) |
534 | { | |
fb0e7942 | 535 | #ifdef CONFIG_ZONE_DMA32 |
9328b8fa | 536 | return zone == zone->zone_pgdat->node_zones + ZONE_DMA32; |
fb0e7942 CL |
537 | #else |
538 | return 0; | |
539 | #endif | |
9328b8fa NP |
540 | } |
541 | ||
542 | static inline int is_dma(struct zone *zone) | |
543 | { | |
4b51d669 | 544 | #ifdef CONFIG_ZONE_DMA |
9328b8fa | 545 | return zone == zone->zone_pgdat->node_zones + ZONE_DMA; |
4b51d669 CL |
546 | #else |
547 | return 0; | |
548 | #endif | |
9328b8fa NP |
549 | } |
550 | ||
1da177e4 LT |
551 | /* These two functions are used to setup the per zone pages min values */ |
552 | struct ctl_table; | |
553 | struct file; | |
554 | int min_free_kbytes_sysctl_handler(struct ctl_table *, int, struct file *, | |
555 | void __user *, size_t *, loff_t *); | |
556 | extern int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1]; | |
557 | int lowmem_reserve_ratio_sysctl_handler(struct ctl_table *, int, struct file *, | |
558 | void __user *, size_t *, loff_t *); | |
8ad4b1fb RS |
559 | int percpu_pagelist_fraction_sysctl_handler(struct ctl_table *, int, struct file *, |
560 | void __user *, size_t *, loff_t *); | |
9614634f CL |
561 | int sysctl_min_unmapped_ratio_sysctl_handler(struct ctl_table *, int, |
562 | struct file *, void __user *, size_t *, loff_t *); | |
0ff38490 CL |
563 | int sysctl_min_slab_ratio_sysctl_handler(struct ctl_table *, int, |
564 | struct file *, void __user *, size_t *, loff_t *); | |
1da177e4 LT |
565 | |
566 | #include <linux/topology.h> | |
567 | /* Returns the number of the current Node. */ | |
69d81fcd | 568 | #ifndef numa_node_id |
39c715b7 | 569 | #define numa_node_id() (cpu_to_node(raw_smp_processor_id())) |
69d81fcd | 570 | #endif |
1da177e4 | 571 | |
93b7504e | 572 | #ifndef CONFIG_NEED_MULTIPLE_NODES |
1da177e4 LT |
573 | |
574 | extern struct pglist_data contig_page_data; | |
575 | #define NODE_DATA(nid) (&contig_page_data) | |
576 | #define NODE_MEM_MAP(nid) mem_map | |
577 | #define MAX_NODES_SHIFT 1 | |
1da177e4 | 578 | |
93b7504e | 579 | #else /* CONFIG_NEED_MULTIPLE_NODES */ |
1da177e4 LT |
580 | |
581 | #include <asm/mmzone.h> | |
582 | ||
93b7504e | 583 | #endif /* !CONFIG_NEED_MULTIPLE_NODES */ |
348f8b6c | 584 | |
95144c78 KH |
585 | extern struct pglist_data *first_online_pgdat(void); |
586 | extern struct pglist_data *next_online_pgdat(struct pglist_data *pgdat); | |
587 | extern struct zone *next_zone(struct zone *zone); | |
8357f869 KH |
588 | |
589 | /** | |
590 | * for_each_pgdat - helper macro to iterate over all nodes | |
591 | * @pgdat - pointer to a pg_data_t variable | |
592 | */ | |
593 | #define for_each_online_pgdat(pgdat) \ | |
594 | for (pgdat = first_online_pgdat(); \ | |
595 | pgdat; \ | |
596 | pgdat = next_online_pgdat(pgdat)) | |
8357f869 KH |
597 | /** |
598 | * for_each_zone - helper macro to iterate over all memory zones | |
599 | * @zone - pointer to struct zone variable | |
600 | * | |
601 | * The user only needs to declare the zone variable, for_each_zone | |
602 | * fills it in. | |
603 | */ | |
604 | #define for_each_zone(zone) \ | |
605 | for (zone = (first_online_pgdat())->node_zones; \ | |
606 | zone; \ | |
607 | zone = next_zone(zone)) | |
608 | ||
d41dee36 AW |
609 | #ifdef CONFIG_SPARSEMEM |
610 | #include <asm/sparsemem.h> | |
611 | #endif | |
612 | ||
07808b74 | 613 | #if BITS_PER_LONG == 32 |
1da177e4 | 614 | /* |
a2f1b424 AK |
615 | * with 32 bit page->flags field, we reserve 9 bits for node/zone info. |
616 | * there are 4 zones (3 bits) and this leaves 9-3=6 bits for nodes. | |
1da177e4 | 617 | */ |
a2f1b424 | 618 | #define FLAGS_RESERVED 9 |
348f8b6c | 619 | |
1da177e4 LT |
620 | #elif BITS_PER_LONG == 64 |
621 | /* | |
622 | * with 64 bit flags field, there's plenty of room. | |
623 | */ | |
348f8b6c | 624 | #define FLAGS_RESERVED 32 |
1da177e4 | 625 | |
348f8b6c | 626 | #else |
1da177e4 | 627 | |
348f8b6c | 628 | #error BITS_PER_LONG not defined |
1da177e4 | 629 | |
1da177e4 LT |
630 | #endif |
631 | ||
c713216d MG |
632 | #if !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) && \ |
633 | !defined(CONFIG_ARCH_POPULATES_NODE_MAP) | |
b159d43f AW |
634 | #define early_pfn_to_nid(nid) (0UL) |
635 | #endif | |
636 | ||
2bdaf115 AW |
637 | #ifdef CONFIG_FLATMEM |
638 | #define pfn_to_nid(pfn) (0) | |
639 | #endif | |
640 | ||
d41dee36 AW |
641 | #define pfn_to_section_nr(pfn) ((pfn) >> PFN_SECTION_SHIFT) |
642 | #define section_nr_to_pfn(sec) ((sec) << PFN_SECTION_SHIFT) | |
643 | ||
644 | #ifdef CONFIG_SPARSEMEM | |
645 | ||
646 | /* | |
647 | * SECTION_SHIFT #bits space required to store a section # | |
648 | * | |
649 | * PA_SECTION_SHIFT physical address to/from section number | |
650 | * PFN_SECTION_SHIFT pfn to/from section number | |
651 | */ | |
652 | #define SECTIONS_SHIFT (MAX_PHYSMEM_BITS - SECTION_SIZE_BITS) | |
653 | ||
654 | #define PA_SECTION_SHIFT (SECTION_SIZE_BITS) | |
655 | #define PFN_SECTION_SHIFT (SECTION_SIZE_BITS - PAGE_SHIFT) | |
656 | ||
657 | #define NR_MEM_SECTIONS (1UL << SECTIONS_SHIFT) | |
658 | ||
659 | #define PAGES_PER_SECTION (1UL << PFN_SECTION_SHIFT) | |
660 | #define PAGE_SECTION_MASK (~(PAGES_PER_SECTION-1)) | |
661 | ||
662 | #if (MAX_ORDER - 1 + PAGE_SHIFT) > SECTION_SIZE_BITS | |
663 | #error Allocator MAX_ORDER exceeds SECTION_SIZE | |
664 | #endif | |
665 | ||
666 | struct page; | |
667 | struct mem_section { | |
29751f69 AW |
668 | /* |
669 | * This is, logically, a pointer to an array of struct | |
670 | * pages. However, it is stored with some other magic. | |
671 | * (see sparse.c::sparse_init_one_section()) | |
672 | * | |
30c253e6 AW |
673 | * Additionally during early boot we encode node id of |
674 | * the location of the section here to guide allocation. | |
675 | * (see sparse.c::memory_present()) | |
676 | * | |
29751f69 AW |
677 | * Making it a UL at least makes someone do a cast |
678 | * before using it wrong. | |
679 | */ | |
680 | unsigned long section_mem_map; | |
d41dee36 AW |
681 | }; |
682 | ||
3e347261 BP |
683 | #ifdef CONFIG_SPARSEMEM_EXTREME |
684 | #define SECTIONS_PER_ROOT (PAGE_SIZE / sizeof (struct mem_section)) | |
685 | #else | |
686 | #define SECTIONS_PER_ROOT 1 | |
687 | #endif | |
802f192e | 688 | |
3e347261 BP |
689 | #define SECTION_NR_TO_ROOT(sec) ((sec) / SECTIONS_PER_ROOT) |
690 | #define NR_SECTION_ROOTS (NR_MEM_SECTIONS / SECTIONS_PER_ROOT) | |
691 | #define SECTION_ROOT_MASK (SECTIONS_PER_ROOT - 1) | |
802f192e | 692 | |
3e347261 BP |
693 | #ifdef CONFIG_SPARSEMEM_EXTREME |
694 | extern struct mem_section *mem_section[NR_SECTION_ROOTS]; | |
802f192e | 695 | #else |
3e347261 BP |
696 | extern struct mem_section mem_section[NR_SECTION_ROOTS][SECTIONS_PER_ROOT]; |
697 | #endif | |
d41dee36 | 698 | |
29751f69 AW |
699 | static inline struct mem_section *__nr_to_section(unsigned long nr) |
700 | { | |
3e347261 BP |
701 | if (!mem_section[SECTION_NR_TO_ROOT(nr)]) |
702 | return NULL; | |
703 | return &mem_section[SECTION_NR_TO_ROOT(nr)][nr & SECTION_ROOT_MASK]; | |
29751f69 | 704 | } |
4ca644d9 | 705 | extern int __section_nr(struct mem_section* ms); |
29751f69 AW |
706 | |
707 | /* | |
708 | * We use the lower bits of the mem_map pointer to store | |
709 | * a little bit of information. There should be at least | |
710 | * 3 bits here due to 32-bit alignment. | |
711 | */ | |
712 | #define SECTION_MARKED_PRESENT (1UL<<0) | |
713 | #define SECTION_HAS_MEM_MAP (1UL<<1) | |
714 | #define SECTION_MAP_LAST_BIT (1UL<<2) | |
715 | #define SECTION_MAP_MASK (~(SECTION_MAP_LAST_BIT-1)) | |
30c253e6 | 716 | #define SECTION_NID_SHIFT 2 |
29751f69 AW |
717 | |
718 | static inline struct page *__section_mem_map_addr(struct mem_section *section) | |
719 | { | |
720 | unsigned long map = section->section_mem_map; | |
721 | map &= SECTION_MAP_MASK; | |
722 | return (struct page *)map; | |
723 | } | |
724 | ||
725 | static inline int valid_section(struct mem_section *section) | |
726 | { | |
802f192e | 727 | return (section && (section->section_mem_map & SECTION_MARKED_PRESENT)); |
29751f69 AW |
728 | } |
729 | ||
730 | static inline int section_has_mem_map(struct mem_section *section) | |
731 | { | |
802f192e | 732 | return (section && (section->section_mem_map & SECTION_HAS_MEM_MAP)); |
29751f69 AW |
733 | } |
734 | ||
735 | static inline int valid_section_nr(unsigned long nr) | |
736 | { | |
737 | return valid_section(__nr_to_section(nr)); | |
738 | } | |
739 | ||
d41dee36 AW |
740 | static inline struct mem_section *__pfn_to_section(unsigned long pfn) |
741 | { | |
29751f69 | 742 | return __nr_to_section(pfn_to_section_nr(pfn)); |
d41dee36 AW |
743 | } |
744 | ||
d41dee36 AW |
745 | static inline int pfn_valid(unsigned long pfn) |
746 | { | |
747 | if (pfn_to_section_nr(pfn) >= NR_MEM_SECTIONS) | |
748 | return 0; | |
29751f69 | 749 | return valid_section(__nr_to_section(pfn_to_section_nr(pfn))); |
d41dee36 AW |
750 | } |
751 | ||
752 | /* | |
753 | * These are _only_ used during initialisation, therefore they | |
754 | * can use __initdata ... They could have names to indicate | |
755 | * this restriction. | |
756 | */ | |
757 | #ifdef CONFIG_NUMA | |
161599ff AW |
758 | #define pfn_to_nid(pfn) \ |
759 | ({ \ | |
760 | unsigned long __pfn_to_nid_pfn = (pfn); \ | |
761 | page_to_nid(pfn_to_page(__pfn_to_nid_pfn)); \ | |
762 | }) | |
2bdaf115 AW |
763 | #else |
764 | #define pfn_to_nid(pfn) (0) | |
d41dee36 AW |
765 | #endif |
766 | ||
d41dee36 AW |
767 | #define early_pfn_valid(pfn) pfn_valid(pfn) |
768 | void sparse_init(void); | |
769 | #else | |
770 | #define sparse_init() do {} while (0) | |
28ae55c9 | 771 | #define sparse_index_init(_sec, _nid) do {} while (0) |
d41dee36 AW |
772 | #endif /* CONFIG_SPARSEMEM */ |
773 | ||
75167957 AW |
774 | #ifdef CONFIG_NODES_SPAN_OTHER_NODES |
775 | #define early_pfn_in_nid(pfn, nid) (early_pfn_to_nid(pfn) == (nid)) | |
776 | #else | |
777 | #define early_pfn_in_nid(pfn, nid) (1) | |
778 | #endif | |
779 | ||
d41dee36 AW |
780 | #ifndef early_pfn_valid |
781 | #define early_pfn_valid(pfn) (1) | |
782 | #endif | |
783 | ||
784 | void memory_present(int nid, unsigned long start, unsigned long end); | |
785 | unsigned long __init node_memmap_size_bytes(int, unsigned long, unsigned long); | |
786 | ||
1da177e4 LT |
787 | #endif /* !__ASSEMBLY__ */ |
788 | #endif /* __KERNEL__ */ | |
789 | #endif /* _LINUX_MMZONE_H */ |