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>
12 #include <linux/atomic.h>
13 #include <linux/page-flags.h>
16 struct notifier_block;
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 */
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)
31 static inline int current_is_kswapd(void)
33 return current->flags & PF_KSWAPD;
37 * MAX_SWAPFILES defines the maximum number of swaptypes: things which can
38 * be swapped to. The swap type and the offset into that swap type are
39 * encoded into pte's and into pgoff_t's in the swapcache. Using five bits
40 * for the type means that the maximum number of swapcache pages is 27 bits
41 * on 32-bit-pgoff_t architectures. And that assumes that the architecture packs
42 * the type/offset into the pte as 5/27 as well.
44 #define MAX_SWAPFILES_SHIFT 5
47 * Use some of the swap files numbers for other purposes. This
48 * is a convenient way to hook into the VM to trigger special
53 * NUMA node memory migration support
55 #ifdef CONFIG_MIGRATION
56 #define SWP_MIGRATION_NUM 2
57 #define SWP_MIGRATION_READ (MAX_SWAPFILES + SWP_HWPOISON_NUM)
58 #define SWP_MIGRATION_WRITE (MAX_SWAPFILES + SWP_HWPOISON_NUM + 1)
60 #define SWP_MIGRATION_NUM 0
64 * Handling of hardware poisoned pages with memory corruption.
66 #ifdef CONFIG_MEMORY_FAILURE
67 #define SWP_HWPOISON_NUM 1
68 #define SWP_HWPOISON MAX_SWAPFILES
70 #define SWP_HWPOISON_NUM 0
73 #define MAX_SWAPFILES \
74 ((1 << MAX_SWAPFILES_SHIFT) - SWP_MIGRATION_NUM - SWP_HWPOISON_NUM)
77 * Magic header for a swap area. The first part of the union is
78 * what the swap magic looks like for the old (limited to 128MB)
79 * swap area format, the second part of the union adds - in the
80 * old reserved area - some extra information. Note that the first
81 * kilobyte is reserved for boot loader or disk label stuff...
83 * Having the magic at the end of the PAGE_SIZE makes detecting swap
84 * areas somewhat tricky on machines that support multiple page sizes.
85 * For 2.5 we'll probably want to move the magic to just beyond the
90 char reserved[PAGE_SIZE - 10];
91 char magic[10]; /* SWAP-SPACE or SWAPSPACE2 */
94 char bootbits[1024]; /* Space for disklabel etc. */
98 unsigned char sws_uuid[16];
99 unsigned char sws_volume[16];
106 * current->reclaim_state points to one of these when a task is running
109 struct reclaim_state {
110 unsigned long reclaimed_slab;
115 struct address_space;
117 struct writeback_control;
121 * A swap extent maps a range of a swapfile's PAGE_SIZE pages onto a range of
122 * disk blocks. A list of swap extents maps the entire swapfile. (Where the
123 * term `swapfile' refers to either a blockdevice or an IS_REG file. Apart
124 * from setup, they're handled identically.
126 * We always assume that blocks are of size PAGE_SIZE.
129 struct list_head list;
132 sector_t start_block;
136 * Max bad pages in the new format..
138 #define MAX_SWAP_BADPAGES \
139 ((offsetof(union swap_header, magic.magic) - \
140 offsetof(union swap_header, info.badpages)) / sizeof(int))
143 SWP_USED = (1 << 0), /* is slot in swap_info[] used? */
144 SWP_WRITEOK = (1 << 1), /* ok to write to this swap? */
145 SWP_DISCARDABLE = (1 << 2), /* blkdev support discard */
146 SWP_DISCARDING = (1 << 3), /* now discarding a free cluster */
147 SWP_SOLIDSTATE = (1 << 4), /* blkdev seeks are cheap */
148 SWP_CONTINUED = (1 << 5), /* swap_map has count continuation */
149 SWP_BLKDEV = (1 << 6), /* its a block device */
150 SWP_FILE = (1 << 7), /* set after swap_activate success */
151 SWP_AREA_DISCARD = (1 << 8), /* single-time swap area discards */
152 SWP_PAGE_DISCARD = (1 << 9), /* freed swap page-cluster discards */
153 SWP_STABLE_WRITES = (1 << 10), /* no overwrite PG_writeback pages */
154 /* add others here before... */
155 SWP_FAST = (1 << 11), /* blkdev access is fast and cheap */
156 SWP_SCANNING = (1 << 12), /* refcount in scan_swap_map */
159 #define SWAP_CLUSTER_MAX 32UL
160 #define COMPACT_CLUSTER_MAX SWAP_CLUSTER_MAX
161 #define SWAPFILE_CLUSTER 256
164 * Ratio between zone->managed_pages and the "gap" that above the per-zone
165 * "high_wmark". While balancing nodes, We allow kswapd to shrink zones that
166 * do not meet the (high_wmark + gap) watermark, even which already met the
167 * high_wmark, in order to provide better per-zone lru behavior. We are ok to
168 * spend not more than 1% of the memory for this zone balancing "gap".
170 #define KSWAPD_ZONE_BALANCE_GAP_RATIO 100
172 #define SWAP_MAP_MAX 0x3e /* Max duplication count, in first swap_map */
173 #define SWAP_MAP_BAD 0x3f /* Note pageblock is bad, in first swap_map */
174 #define SWAP_HAS_CACHE 0x40 /* Flag page is cached, in first swap_map */
175 #define SWAP_CONT_MAX 0x7f /* Max count, in each swap_map continuation */
176 #define COUNT_CONTINUED 0x80 /* See swap_map continuation for full count */
177 #define SWAP_MAP_SHMEM 0xbf /* Owned by shmem/tmpfs, in first swap_map */
180 * We use this to track usage of a cluster. A cluster is a block of swap disk
181 * space with SWAPFILE_CLUSTER pages long and naturally aligns in disk. All
182 * free clusters are organized into a list. We fetch an entry from the list to
183 * get a free cluster.
185 * The data field stores next cluster if the cluster is free or cluster usage
186 * counter otherwise. The flags field determines if a cluster is free. This is
187 * protected by swap_info_struct.lock.
189 struct swap_cluster_info {
190 unsigned int data:24;
191 unsigned int flags:8;
193 #define CLUSTER_FLAG_FREE 1 /* This cluster is free */
194 #define CLUSTER_FLAG_NEXT_NULL 2 /* This cluster has no next cluster */
197 * We assign a cluster to each CPU, so each CPU can allocate swap entry from
198 * its own cluster and swapout sequentially. The purpose is to optimize swapout
201 struct percpu_cluster {
202 struct swap_cluster_info index; /* Current cluster index */
203 unsigned int next; /* Likely next allocation offset */
207 * The in-memory structure used to track swap areas.
209 struct swap_info_struct {
210 unsigned long flags; /* SWP_USED etc: see above */
211 signed short prio; /* swap priority of this type */
212 struct plist_node list; /* entry in swap_active_head */
213 struct plist_node avail_list; /* entry in swap_avail_head */
214 signed char type; /* strange name for an index */
215 unsigned int max; /* extent of the swap_map */
216 unsigned char *swap_map; /* vmalloc'ed array of usage counts */
217 struct swap_cluster_info *cluster_info; /* cluster info. Only for SSD */
218 struct swap_cluster_info free_cluster_head; /* free cluster list head */
219 struct swap_cluster_info free_cluster_tail; /* free cluster list tail */
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 */
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
249 struct work_struct discard_work; /* discard worker */
250 struct swap_cluster_info discard_cluster_head; /* list head of discard clusters */
251 struct swap_cluster_info discard_cluster_tail; /* list tail of discard clusters */
252 unsigned int write_pending;
253 unsigned int max_writes;
256 /* linux/mm/workingset.c */
257 void *workingset_eviction(struct address_space *mapping, struct page *page);
258 bool workingset_refault(void *shadow);
259 void workingset_activation(struct page *page);
260 extern struct list_lru workingset_shadow_nodes;
262 static inline unsigned int workingset_node_pages(struct radix_tree_node *node)
264 return node->count & RADIX_TREE_COUNT_MASK;
267 static inline void workingset_node_pages_inc(struct radix_tree_node *node)
272 static inline void workingset_node_pages_dec(struct radix_tree_node *node)
274 VM_WARN_ON_ONCE(!workingset_node_pages(node));
278 static inline unsigned int workingset_node_shadows(struct radix_tree_node *node)
280 return node->count >> RADIX_TREE_COUNT_SHIFT;
283 static inline void workingset_node_shadows_inc(struct radix_tree_node *node)
285 node->count += 1U << RADIX_TREE_COUNT_SHIFT;
288 static inline void workingset_node_shadows_dec(struct radix_tree_node *node)
290 VM_WARN_ON_ONCE(!workingset_node_shadows(node));
291 node->count -= 1U << RADIX_TREE_COUNT_SHIFT;
294 /* linux/mm/page_alloc.c */
295 extern unsigned long totalram_pages;
296 extern unsigned long totalreserve_pages;
297 extern unsigned long nr_free_buffer_pages(void);
298 extern unsigned long nr_free_pagecache_pages(void);
300 /* Definition of global_page_state not available yet */
301 #define nr_free_pages() global_page_state(NR_FREE_PAGES)
304 /* linux/mm/swap.c */
305 extern void lru_cache_add(struct page *);
306 extern void lru_cache_add_anon(struct page *page);
307 extern void lru_cache_add_file(struct page *page);
308 extern void lru_add_page_tail(struct page *page, struct page *page_tail,
309 struct lruvec *lruvec, struct list_head *head);
310 extern void activate_page(struct page *);
311 extern void mark_page_accessed(struct page *);
312 extern void lru_add_drain(void);
313 extern void lru_add_drain_cpu(int cpu);
314 extern void lru_add_drain_all(void);
315 extern void rotate_reclaimable_page(struct page *page);
316 extern void deactivate_file_page(struct page *page);
317 extern void swap_setup(void);
319 extern void add_page_to_unevictable_list(struct page *page);
321 extern void lru_cache_add_active_or_unevictable(struct page *page,
322 struct vm_area_struct *vma);
324 /* linux/mm/vmscan.c */
325 extern unsigned long try_to_free_pages(struct zonelist *zonelist, int order,
326 gfp_t gfp_mask, nodemask_t *mask);
327 extern int __isolate_lru_page(struct page *page, isolate_mode_t mode);
328 extern unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *memcg,
329 unsigned long nr_pages,
332 extern unsigned long mem_cgroup_shrink_node_zone(struct mem_cgroup *mem,
333 gfp_t gfp_mask, bool noswap,
335 unsigned long *nr_scanned);
336 extern unsigned long shrink_all_memory(unsigned long nr_pages);
337 extern int vm_swappiness;
338 extern int sysctl_swap_ratio;
339 extern int sysctl_swap_ratio_enable;
340 extern int remove_mapping(struct address_space *mapping, struct page *page);
341 extern unsigned long vm_total_pages;
344 extern int zone_reclaim_mode;
345 extern int sysctl_min_unmapped_ratio;
346 extern int sysctl_min_slab_ratio;
347 extern int zone_reclaim(struct zone *, gfp_t, unsigned int);
349 #define zone_reclaim_mode 0
350 static inline int zone_reclaim(struct zone *z, gfp_t mask, unsigned int order)
356 extern int page_evictable(struct page *page);
357 extern void check_move_unevictable_pages(struct page **, int nr_pages);
359 extern int kswapd_run(int nid);
360 extern void kswapd_stop(int nid);
362 static inline int mem_cgroup_swappiness(struct mem_cgroup *memcg)
365 if (mem_cgroup_disabled() || !memcg->css.parent)
366 return vm_swappiness;
368 return memcg->swappiness;
372 static inline int mem_cgroup_swappiness(struct mem_cgroup *mem)
374 return vm_swappiness;
377 #ifdef CONFIG_MEMCG_SWAP
378 extern void mem_cgroup_swapout(struct page *page, swp_entry_t entry);
379 extern void mem_cgroup_uncharge_swap(swp_entry_t entry);
381 static inline void mem_cgroup_swapout(struct page *page, swp_entry_t entry)
384 static inline void mem_cgroup_uncharge_swap(swp_entry_t entry)
389 /* linux/mm/page_io.c */
390 extern int swap_readpage(struct page *);
391 extern int swap_writepage(struct page *page, struct writeback_control *wbc);
392 extern void end_swap_bio_write(struct bio *bio);
393 extern int __swap_writepage(struct page *page, struct writeback_control *wbc,
394 bio_end_io_t end_write_func);
395 extern int swap_set_page_dirty(struct page *page);
397 int add_swap_extent(struct swap_info_struct *sis, unsigned long start_page,
398 unsigned long nr_pages, sector_t start_block);
399 int generic_swapfile_activate(struct swap_info_struct *, struct file *,
402 /* linux/mm/swap_state.c */
403 extern struct address_space swapper_spaces[];
404 #define swap_address_space(entry) (&swapper_spaces[swp_type(entry)])
405 extern unsigned long total_swapcache_pages(void);
406 extern void show_swap_cache_info(void);
407 extern int add_to_swap(struct page *, struct list_head *list);
408 extern int add_to_swap_cache(struct page *, swp_entry_t, gfp_t);
409 extern int __add_to_swap_cache(struct page *page, swp_entry_t entry);
410 extern void __delete_from_swap_cache(struct page *);
411 extern void delete_from_swap_cache(struct page *);
412 extern void free_page_and_swap_cache(struct page *);
413 extern void free_pages_and_swap_cache(struct page **, int);
414 extern struct page *lookup_swap_cache(swp_entry_t);
415 extern struct page *read_swap_cache_async(swp_entry_t, gfp_t,
416 struct vm_area_struct *vma, unsigned long addr);
417 extern struct page *__read_swap_cache_async(swp_entry_t, gfp_t,
418 struct vm_area_struct *vma, unsigned long addr,
419 bool *new_page_allocated);
420 extern struct page *swapin_readahead(swp_entry_t, gfp_t,
421 struct vm_area_struct *vma, unsigned long addr);
423 /* linux/mm/swapfile.c */
424 extern atomic_long_t nr_swap_pages;
425 extern long total_swap_pages;
426 extern bool is_swap_fast(swp_entry_t entry);
428 /* Swap 50% full? Release swapcache more aggressively.. */
429 static inline bool vm_swap_full(struct swap_info_struct *si)
432 * If the swap device is fast, return true
433 * not to delay swap free.
435 if (si->flags & SWP_FAST)
438 return atomic_long_read(&nr_swap_pages) * 2 < total_swap_pages;
441 static inline long get_nr_swap_pages(void)
443 return atomic_long_read(&nr_swap_pages);
446 extern void si_swapinfo(struct sysinfo *);
447 extern swp_entry_t get_swap_page(void);
448 extern swp_entry_t get_swap_page_of_type(int);
449 extern int add_swap_count_continuation(swp_entry_t, gfp_t);
450 extern void swap_shmem_alloc(swp_entry_t);
451 extern int swap_duplicate(swp_entry_t);
452 extern int swapcache_prepare(swp_entry_t);
453 extern void swap_free(swp_entry_t);
454 extern void swapcache_free(swp_entry_t);
455 extern int free_swap_and_cache(swp_entry_t);
456 extern int swap_type_of(dev_t, sector_t, struct block_device **);
457 extern unsigned int count_swap_pages(int, int);
458 extern sector_t map_swap_page(struct page *, struct block_device **);
459 extern sector_t swapdev_block(int, pgoff_t);
460 extern int page_swapcount(struct page *);
461 extern int swp_swapcount(swp_entry_t entry);
462 extern struct swap_info_struct *page_swap_info(struct page *);
463 extern int reuse_swap_page(struct page *);
464 extern int try_to_free_swap(struct page *);
465 struct backing_dev_info;
467 #else /* CONFIG_SWAP */
469 #define swap_address_space(entry) (NULL)
470 #define get_nr_swap_pages() 0L
471 #define total_swap_pages 0L
472 #define total_swapcache_pages() 0UL
473 #define vm_swap_full(si) 0
475 #define si_swapinfo(val) \
476 do { (val)->freeswap = (val)->totalswap = 0; } while (0)
477 /* only sparc can not include linux/pagemap.h in this file
478 * so leave page_cache_release and release_pages undeclared... */
479 #define free_page_and_swap_cache(page) \
480 page_cache_release(page)
481 #define free_pages_and_swap_cache(pages, nr) \
482 release_pages((pages), (nr), false);
484 static inline void show_swap_cache_info(void)
488 #define free_swap_and_cache(swp) is_migration_entry(swp)
489 #define swapcache_prepare(swp) is_migration_entry(swp)
491 static inline int add_swap_count_continuation(swp_entry_t swp, gfp_t gfp_mask)
496 static inline void swap_shmem_alloc(swp_entry_t swp)
500 static inline int swap_duplicate(swp_entry_t swp)
505 static inline void swap_free(swp_entry_t swp)
509 static inline void swapcache_free(swp_entry_t swp)
513 static inline struct page *swapin_readahead(swp_entry_t swp, gfp_t gfp_mask,
514 struct vm_area_struct *vma, unsigned long addr)
519 static inline int swap_writepage(struct page *p, struct writeback_control *wbc)
524 static inline struct page *lookup_swap_cache(swp_entry_t swp)
529 static inline int add_to_swap(struct page *page, struct list_head *list)
534 static inline int add_to_swap_cache(struct page *page, swp_entry_t entry,
540 static inline void __delete_from_swap_cache(struct page *page)
544 static inline void delete_from_swap_cache(struct page *page)
548 static inline int page_swapcount(struct page *page)
553 static inline int swp_swapcount(swp_entry_t entry)
558 #define reuse_swap_page(page) (page_mapcount(page) == 1)
560 static inline int try_to_free_swap(struct page *page)
565 static inline swp_entry_t get_swap_page(void)
572 #endif /* CONFIG_SWAP */
573 #endif /* __KERNEL__*/
574 #endif /* _LINUX_SWAP_H */