]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blob - include/linux/swap.h
KEYS: Fix race between updating and finding a negative key
[mirror_ubuntu-artful-kernel.git] / include / linux / swap.h
1 #ifndef _LINUX_SWAP_H
2 #define _LINUX_SWAP_H
3
4 #include <linux/spinlock.h>
5 #include <linux/linkage.h>
6 #include <linux/mmzone.h>
7 #include <linux/list.h>
8 #include <linux/memcontrol.h>
9 #include <linux/sched.h>
10 #include <linux/node.h>
11 #include <linux/fs.h>
12 #include <linux/atomic.h>
13 #include <linux/page-flags.h>
14 #include <asm/page.h>
15
16 struct notifier_block;
17
18 struct bio;
19
20 #define SWAP_FLAG_PREFER 0x8000 /* set if swap priority specified */
21 #define SWAP_FLAG_PRIO_MASK 0x7fff
22 #define SWAP_FLAG_PRIO_SHIFT 0
23 #define SWAP_FLAG_DISCARD 0x10000 /* enable discard for swap */
24 #define SWAP_FLAG_DISCARD_ONCE 0x20000 /* discard swap area at swapon-time */
25 #define SWAP_FLAG_DISCARD_PAGES 0x40000 /* discard page-clusters after use */
26
27 #define SWAP_FLAGS_VALID (SWAP_FLAG_PRIO_MASK | SWAP_FLAG_PREFER | \
28 SWAP_FLAG_DISCARD | SWAP_FLAG_DISCARD_ONCE | \
29 SWAP_FLAG_DISCARD_PAGES)
30 #define SWAP_BATCH 64
31
32 static inline int current_is_kswapd(void)
33 {
34 return current->flags & PF_KSWAPD;
35 }
36
37 /*
38 * MAX_SWAPFILES defines the maximum number of swaptypes: things which can
39 * be swapped to. The swap type and the offset into that swap type are
40 * encoded into pte's and into pgoff_t's in the swapcache. Using five bits
41 * for the type means that the maximum number of swapcache pages is 27 bits
42 * on 32-bit-pgoff_t architectures. And that assumes that the architecture packs
43 * the type/offset into the pte as 5/27 as well.
44 */
45 #define MAX_SWAPFILES_SHIFT 5
46
47 /*
48 * Use some of the swap files numbers for other purposes. This
49 * is a convenient way to hook into the VM to trigger special
50 * actions on faults.
51 */
52
53 /*
54 * NUMA node memory migration support
55 */
56 #ifdef CONFIG_MIGRATION
57 #define SWP_MIGRATION_NUM 2
58 #define SWP_MIGRATION_READ (MAX_SWAPFILES + SWP_HWPOISON_NUM)
59 #define SWP_MIGRATION_WRITE (MAX_SWAPFILES + SWP_HWPOISON_NUM + 1)
60 #else
61 #define SWP_MIGRATION_NUM 0
62 #endif
63
64 /*
65 * Handling of hardware poisoned pages with memory corruption.
66 */
67 #ifdef CONFIG_MEMORY_FAILURE
68 #define SWP_HWPOISON_NUM 1
69 #define SWP_HWPOISON MAX_SWAPFILES
70 #else
71 #define SWP_HWPOISON_NUM 0
72 #endif
73
74 #define MAX_SWAPFILES \
75 ((1 << MAX_SWAPFILES_SHIFT) - SWP_MIGRATION_NUM - SWP_HWPOISON_NUM)
76
77 /*
78 * Magic header for a swap area. The first part of the union is
79 * what the swap magic looks like for the old (limited to 128MB)
80 * swap area format, the second part of the union adds - in the
81 * old reserved area - some extra information. Note that the first
82 * kilobyte is reserved for boot loader or disk label stuff...
83 *
84 * Having the magic at the end of the PAGE_SIZE makes detecting swap
85 * areas somewhat tricky on machines that support multiple page sizes.
86 * For 2.5 we'll probably want to move the magic to just beyond the
87 * bootbits...
88 */
89 union swap_header {
90 struct {
91 char reserved[PAGE_SIZE - 10];
92 char magic[10]; /* SWAP-SPACE or SWAPSPACE2 */
93 } magic;
94 struct {
95 char bootbits[1024]; /* Space for disklabel etc. */
96 __u32 version;
97 __u32 last_page;
98 __u32 nr_badpages;
99 unsigned char sws_uuid[16];
100 unsigned char sws_volume[16];
101 __u32 padding[117];
102 __u32 badpages[1];
103 } info;
104 };
105
106 /*
107 * current->reclaim_state points to one of these when a task is running
108 * memory reclaim
109 */
110 struct reclaim_state {
111 unsigned long reclaimed_slab;
112 };
113
114 #ifdef __KERNEL__
115
116 struct address_space;
117 struct sysinfo;
118 struct writeback_control;
119 struct zone;
120
121 /*
122 * A swap extent maps a range of a swapfile's PAGE_SIZE pages onto a range of
123 * disk blocks. A list of swap extents maps the entire swapfile. (Where the
124 * term `swapfile' refers to either a blockdevice or an IS_REG file. Apart
125 * from setup, they're handled identically.
126 *
127 * We always assume that blocks are of size PAGE_SIZE.
128 */
129 struct swap_extent {
130 struct list_head list;
131 pgoff_t start_page;
132 pgoff_t nr_pages;
133 sector_t start_block;
134 };
135
136 /*
137 * Max bad pages in the new format..
138 */
139 #define __swapoffset(x) ((unsigned long)&((union swap_header *)0)->x)
140 #define MAX_SWAP_BADPAGES \
141 ((__swapoffset(magic.magic) - __swapoffset(info.badpages)) / sizeof(int))
142
143 enum {
144 SWP_USED = (1 << 0), /* is slot in swap_info[] used? */
145 SWP_WRITEOK = (1 << 1), /* ok to write to this swap? */
146 SWP_DISCARDABLE = (1 << 2), /* blkdev support discard */
147 SWP_DISCARDING = (1 << 3), /* now discarding a free cluster */
148 SWP_SOLIDSTATE = (1 << 4), /* blkdev seeks are cheap */
149 SWP_CONTINUED = (1 << 5), /* swap_map has count continuation */
150 SWP_BLKDEV = (1 << 6), /* its a block device */
151 SWP_FILE = (1 << 7), /* set after swap_activate success */
152 SWP_AREA_DISCARD = (1 << 8), /* single-time swap area discards */
153 SWP_PAGE_DISCARD = (1 << 9), /* freed swap page-cluster discards */
154 SWP_STABLE_WRITES = (1 << 10), /* no overwrite PG_writeback pages */
155 /* add others here before... */
156 SWP_SCANNING = (1 << 11), /* refcount in scan_swap_map */
157 };
158
159 #define SWAP_CLUSTER_MAX 32UL
160 #define COMPACT_CLUSTER_MAX SWAP_CLUSTER_MAX
161
162 #define SWAP_MAP_MAX 0x3e /* Max duplication count, in first swap_map */
163 #define SWAP_MAP_BAD 0x3f /* Note pageblock is bad, in first swap_map */
164 #define SWAP_HAS_CACHE 0x40 /* Flag page is cached, in first swap_map */
165 #define SWAP_CONT_MAX 0x7f /* Max count, in each swap_map continuation */
166 #define COUNT_CONTINUED 0x80 /* See swap_map continuation for full count */
167 #define SWAP_MAP_SHMEM 0xbf /* Owned by shmem/tmpfs, in first swap_map */
168
169 /*
170 * We use this to track usage of a cluster. A cluster is a block of swap disk
171 * space with SWAPFILE_CLUSTER pages long and naturally aligns in disk. All
172 * free clusters are organized into a list. We fetch an entry from the list to
173 * get a free cluster.
174 *
175 * The data field stores next cluster if the cluster is free or cluster usage
176 * counter otherwise. The flags field determines if a cluster is free. This is
177 * protected by swap_info_struct.lock.
178 */
179 struct swap_cluster_info {
180 spinlock_t lock; /*
181 * Protect swap_cluster_info fields
182 * and swap_info_struct->swap_map
183 * elements correspond to the swap
184 * cluster
185 */
186 unsigned int data:24;
187 unsigned int flags:8;
188 };
189 #define CLUSTER_FLAG_FREE 1 /* This cluster is free */
190 #define CLUSTER_FLAG_NEXT_NULL 2 /* This cluster has no next cluster */
191
192 /*
193 * We assign a cluster to each CPU, so each CPU can allocate swap entry from
194 * its own cluster and swapout sequentially. The purpose is to optimize swapout
195 * throughput.
196 */
197 struct percpu_cluster {
198 struct swap_cluster_info index; /* Current cluster index */
199 unsigned int next; /* Likely next allocation offset */
200 };
201
202 struct swap_cluster_list {
203 struct swap_cluster_info head;
204 struct swap_cluster_info tail;
205 };
206
207 /*
208 * The in-memory structure used to track swap areas.
209 */
210 struct swap_info_struct {
211 unsigned long flags; /* SWP_USED etc: see above */
212 signed short prio; /* swap priority of this type */
213 struct plist_node list; /* entry in swap_active_head */
214 struct plist_node avail_list; /* entry in swap_avail_head */
215 signed char type; /* strange name for an index */
216 unsigned int max; /* extent of the swap_map */
217 unsigned char *swap_map; /* vmalloc'ed array of usage counts */
218 struct swap_cluster_info *cluster_info; /* cluster info. Only for SSD */
219 struct swap_cluster_list free_clusters; /* free clusters list */
220 unsigned int lowest_bit; /* index of first free in swap_map */
221 unsigned int highest_bit; /* index of last free in swap_map */
222 unsigned int pages; /* total of usable pages of swap */
223 unsigned int inuse_pages; /* number of those currently in use */
224 unsigned int cluster_next; /* likely index for next allocation */
225 unsigned int cluster_nr; /* countdown to next cluster search */
226 struct percpu_cluster __percpu *percpu_cluster; /* per cpu's swap location */
227 struct swap_extent *curr_swap_extent;
228 struct swap_extent first_swap_extent;
229 struct block_device *bdev; /* swap device or bdev of swap file */
230 struct file *swap_file; /* seldom referenced */
231 unsigned int old_block_size; /* seldom referenced */
232 #ifdef CONFIG_FRONTSWAP
233 unsigned long *frontswap_map; /* frontswap in-use, one bit per page */
234 atomic_t frontswap_pages; /* frontswap pages in-use counter */
235 #endif
236 spinlock_t lock; /*
237 * protect map scan related fields like
238 * swap_map, lowest_bit, highest_bit,
239 * inuse_pages, cluster_next,
240 * cluster_nr, lowest_alloc,
241 * highest_alloc, free/discard cluster
242 * list. other fields are only changed
243 * at swapon/swapoff, so are protected
244 * by swap_lock. changing flags need
245 * hold this lock and swap_lock. If
246 * both locks need hold, hold swap_lock
247 * first.
248 */
249 struct work_struct discard_work; /* discard worker */
250 struct swap_cluster_list discard_clusters; /* discard clusters list */
251 };
252
253 /* linux/mm/workingset.c */
254 void *workingset_eviction(struct address_space *mapping, struct page *page);
255 bool workingset_refault(void *shadow);
256 void workingset_activation(struct page *page);
257 void workingset_update_node(struct radix_tree_node *node, void *private);
258
259 /* linux/mm/page_alloc.c */
260 extern unsigned long totalram_pages;
261 extern unsigned long totalreserve_pages;
262 extern unsigned long nr_free_buffer_pages(void);
263 extern unsigned long nr_free_pagecache_pages(void);
264
265 /* Definition of global_page_state not available yet */
266 #define nr_free_pages() global_page_state(NR_FREE_PAGES)
267
268
269 /* linux/mm/swap.c */
270 extern void lru_cache_add(struct page *);
271 extern void lru_cache_add_anon(struct page *page);
272 extern void lru_cache_add_file(struct page *page);
273 extern void lru_add_page_tail(struct page *page, struct page *page_tail,
274 struct lruvec *lruvec, struct list_head *head);
275 extern void activate_page(struct page *);
276 extern void mark_page_accessed(struct page *);
277 extern void lru_add_drain(void);
278 extern void lru_add_drain_cpu(int cpu);
279 extern void lru_add_drain_all(void);
280 extern void lru_add_drain_all_cpuslocked(void);
281 extern void rotate_reclaimable_page(struct page *page);
282 extern void deactivate_file_page(struct page *page);
283 extern void mark_page_lazyfree(struct page *page);
284 extern void swap_setup(void);
285
286 extern void add_page_to_unevictable_list(struct page *page);
287
288 extern void lru_cache_add_active_or_unevictable(struct page *page,
289 struct vm_area_struct *vma);
290
291 /* linux/mm/vmscan.c */
292 extern unsigned long zone_reclaimable_pages(struct zone *zone);
293 extern unsigned long pgdat_reclaimable_pages(struct pglist_data *pgdat);
294 extern unsigned long try_to_free_pages(struct zonelist *zonelist, int order,
295 gfp_t gfp_mask, nodemask_t *mask);
296 extern int __isolate_lru_page(struct page *page, isolate_mode_t mode);
297 extern unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *memcg,
298 unsigned long nr_pages,
299 gfp_t gfp_mask,
300 bool may_swap);
301 extern unsigned long mem_cgroup_shrink_node(struct mem_cgroup *mem,
302 gfp_t gfp_mask, bool noswap,
303 pg_data_t *pgdat,
304 unsigned long *nr_scanned);
305 extern unsigned long shrink_all_memory(unsigned long nr_pages);
306 extern int vm_swappiness;
307 extern int remove_mapping(struct address_space *mapping, struct page *page);
308 extern unsigned long vm_total_pages;
309
310 #ifdef CONFIG_NUMA
311 extern int node_reclaim_mode;
312 extern int sysctl_min_unmapped_ratio;
313 extern int sysctl_min_slab_ratio;
314 extern int node_reclaim(struct pglist_data *, gfp_t, unsigned int);
315 #else
316 #define node_reclaim_mode 0
317 static inline int node_reclaim(struct pglist_data *pgdat, gfp_t mask,
318 unsigned int order)
319 {
320 return 0;
321 }
322 #endif
323
324 extern int page_evictable(struct page *page);
325 extern void check_move_unevictable_pages(struct page **, int nr_pages);
326
327 extern int kswapd_run(int nid);
328 extern void kswapd_stop(int nid);
329
330 #ifdef CONFIG_SWAP
331
332 #include <linux/blk_types.h> /* for bio_end_io_t */
333
334 /* linux/mm/page_io.c */
335 extern int swap_readpage(struct page *page, bool do_poll);
336 extern int swap_writepage(struct page *page, struct writeback_control *wbc);
337 extern void end_swap_bio_write(struct bio *bio);
338 extern int __swap_writepage(struct page *page, struct writeback_control *wbc,
339 bio_end_io_t end_write_func);
340 extern int swap_set_page_dirty(struct page *page);
341
342 int add_swap_extent(struct swap_info_struct *sis, unsigned long start_page,
343 unsigned long nr_pages, sector_t start_block);
344 int generic_swapfile_activate(struct swap_info_struct *, struct file *,
345 sector_t *);
346
347 /* linux/mm/swap_state.c */
348 /* One swap address space for each 64M swap space */
349 #define SWAP_ADDRESS_SPACE_SHIFT 14
350 #define SWAP_ADDRESS_SPACE_PAGES (1 << SWAP_ADDRESS_SPACE_SHIFT)
351 extern struct address_space *swapper_spaces[];
352 #define swap_address_space(entry) \
353 (&swapper_spaces[swp_type(entry)][swp_offset(entry) \
354 >> SWAP_ADDRESS_SPACE_SHIFT])
355 extern unsigned long total_swapcache_pages(void);
356 extern void show_swap_cache_info(void);
357 extern int add_to_swap(struct page *page);
358 extern int add_to_swap_cache(struct page *, swp_entry_t, gfp_t);
359 extern int __add_to_swap_cache(struct page *page, swp_entry_t entry);
360 extern void __delete_from_swap_cache(struct page *);
361 extern void delete_from_swap_cache(struct page *);
362 extern void free_page_and_swap_cache(struct page *);
363 extern void free_pages_and_swap_cache(struct page **, int);
364 extern struct page *lookup_swap_cache(swp_entry_t);
365 extern struct page *read_swap_cache_async(swp_entry_t, gfp_t,
366 struct vm_area_struct *vma, unsigned long addr,
367 bool do_poll);
368 extern struct page *__read_swap_cache_async(swp_entry_t, gfp_t,
369 struct vm_area_struct *vma, unsigned long addr,
370 bool *new_page_allocated);
371 extern struct page *swapin_readahead(swp_entry_t, gfp_t,
372 struct vm_area_struct *vma, unsigned long addr);
373
374 /* linux/mm/swapfile.c */
375 extern atomic_long_t nr_swap_pages;
376 extern long total_swap_pages;
377 extern bool has_usable_swap(void);
378
379 /* Swap 50% full? Release swapcache more aggressively.. */
380 static inline bool vm_swap_full(void)
381 {
382 return atomic_long_read(&nr_swap_pages) * 2 < total_swap_pages;
383 }
384
385 static inline long get_nr_swap_pages(void)
386 {
387 return atomic_long_read(&nr_swap_pages);
388 }
389
390 extern void si_swapinfo(struct sysinfo *);
391 extern swp_entry_t get_swap_page(struct page *page);
392 extern void put_swap_page(struct page *page, swp_entry_t entry);
393 extern swp_entry_t get_swap_page_of_type(int);
394 extern int get_swap_pages(int n, bool cluster, swp_entry_t swp_entries[]);
395 extern int add_swap_count_continuation(swp_entry_t, gfp_t);
396 extern void swap_shmem_alloc(swp_entry_t);
397 extern int swap_duplicate(swp_entry_t);
398 extern int swapcache_prepare(swp_entry_t);
399 extern void swap_free(swp_entry_t);
400 extern void swapcache_free_entries(swp_entry_t *entries, int n);
401 extern int free_swap_and_cache(swp_entry_t);
402 extern int swap_type_of(dev_t, sector_t, struct block_device **);
403 extern unsigned int count_swap_pages(int, int);
404 extern sector_t map_swap_page(struct page *, struct block_device **);
405 extern sector_t swapdev_block(int, pgoff_t);
406 extern int page_swapcount(struct page *);
407 extern int __swp_swapcount(swp_entry_t entry);
408 extern int swp_swapcount(swp_entry_t entry);
409 extern struct swap_info_struct *page_swap_info(struct page *);
410 extern bool reuse_swap_page(struct page *, int *);
411 extern int try_to_free_swap(struct page *);
412 struct backing_dev_info;
413 extern int init_swap_address_space(unsigned int type, unsigned long nr_pages);
414 extern void exit_swap_address_space(unsigned int type);
415
416 #else /* CONFIG_SWAP */
417
418 #define swap_address_space(entry) (NULL)
419 #define get_nr_swap_pages() 0L
420 #define total_swap_pages 0L
421 #define total_swapcache_pages() 0UL
422 #define vm_swap_full() 0
423
424 #define si_swapinfo(val) \
425 do { (val)->freeswap = (val)->totalswap = 0; } while (0)
426 /* only sparc can not include linux/pagemap.h in this file
427 * so leave put_page and release_pages undeclared... */
428 #define free_page_and_swap_cache(page) \
429 put_page(page)
430 #define free_pages_and_swap_cache(pages, nr) \
431 release_pages((pages), (nr), false);
432
433 static inline void show_swap_cache_info(void)
434 {
435 }
436
437 #define free_swap_and_cache(swp) is_migration_entry(swp)
438 #define swapcache_prepare(swp) is_migration_entry(swp)
439
440 static inline int add_swap_count_continuation(swp_entry_t swp, gfp_t gfp_mask)
441 {
442 return 0;
443 }
444
445 static inline void swap_shmem_alloc(swp_entry_t swp)
446 {
447 }
448
449 static inline int swap_duplicate(swp_entry_t swp)
450 {
451 return 0;
452 }
453
454 static inline void swap_free(swp_entry_t swp)
455 {
456 }
457
458 static inline void put_swap_page(struct page *page, swp_entry_t swp)
459 {
460 }
461
462 static inline struct page *swapin_readahead(swp_entry_t swp, gfp_t gfp_mask,
463 struct vm_area_struct *vma, unsigned long addr)
464 {
465 return NULL;
466 }
467
468 static inline int swap_writepage(struct page *p, struct writeback_control *wbc)
469 {
470 return 0;
471 }
472
473 static inline struct page *lookup_swap_cache(swp_entry_t swp)
474 {
475 return NULL;
476 }
477
478 static inline int add_to_swap(struct page *page)
479 {
480 return 0;
481 }
482
483 static inline int add_to_swap_cache(struct page *page, swp_entry_t entry,
484 gfp_t gfp_mask)
485 {
486 return -1;
487 }
488
489 static inline void __delete_from_swap_cache(struct page *page)
490 {
491 }
492
493 static inline void delete_from_swap_cache(struct page *page)
494 {
495 }
496
497 static inline int page_swapcount(struct page *page)
498 {
499 return 0;
500 }
501
502 static inline int __swp_swapcount(swp_entry_t entry)
503 {
504 return 0;
505 }
506
507 static inline int swp_swapcount(swp_entry_t entry)
508 {
509 return 0;
510 }
511
512 #define reuse_swap_page(page, total_mapcount) \
513 (page_trans_huge_mapcount(page, total_mapcount) == 1)
514
515 static inline int try_to_free_swap(struct page *page)
516 {
517 return 0;
518 }
519
520 static inline swp_entry_t get_swap_page(struct page *page)
521 {
522 swp_entry_t entry;
523 entry.val = 0;
524 return entry;
525 }
526
527 #endif /* CONFIG_SWAP */
528
529 #ifdef CONFIG_MEMCG
530 static inline int mem_cgroup_swappiness(struct mem_cgroup *memcg)
531 {
532 /* Cgroup2 doesn't have per-cgroup swappiness */
533 if (cgroup_subsys_on_dfl(memory_cgrp_subsys))
534 return vm_swappiness;
535
536 /* root ? */
537 if (mem_cgroup_disabled() || !memcg->css.parent)
538 return vm_swappiness;
539
540 return memcg->swappiness;
541 }
542
543 #else
544 static inline int mem_cgroup_swappiness(struct mem_cgroup *mem)
545 {
546 return vm_swappiness;
547 }
548 #endif
549
550 #ifdef CONFIG_MEMCG_SWAP
551 extern void mem_cgroup_swapout(struct page *page, swp_entry_t entry);
552 extern int mem_cgroup_try_charge_swap(struct page *page, swp_entry_t entry);
553 extern void mem_cgroup_uncharge_swap(swp_entry_t entry, unsigned int nr_pages);
554 extern long mem_cgroup_get_nr_swap_pages(struct mem_cgroup *memcg);
555 extern bool mem_cgroup_swap_full(struct page *page);
556 #else
557 static inline void mem_cgroup_swapout(struct page *page, swp_entry_t entry)
558 {
559 }
560
561 static inline int mem_cgroup_try_charge_swap(struct page *page,
562 swp_entry_t entry)
563 {
564 return 0;
565 }
566
567 static inline void mem_cgroup_uncharge_swap(swp_entry_t entry,
568 unsigned int nr_pages)
569 {
570 }
571
572 static inline long mem_cgroup_get_nr_swap_pages(struct mem_cgroup *memcg)
573 {
574 return get_nr_swap_pages();
575 }
576
577 static inline bool mem_cgroup_swap_full(struct page *page)
578 {
579 return vm_swap_full();
580 }
581 #endif
582
583 #endif /* __KERNEL__*/
584 #endif /* _LINUX_SWAP_H */