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];
105 /* A swap entry has to fit into a "unsigned long", as
106 * the entry is hidden in the "index" field of the
107 * swapper address space.
114 * current->reclaim_state points to one of these when a task is running
117 struct reclaim_state {
118 unsigned long reclaimed_slab;
123 struct address_space;
125 struct writeback_control;
129 * A swap extent maps a range of a swapfile's PAGE_SIZE pages onto a range of
130 * disk blocks. A list of swap extents maps the entire swapfile. (Where the
131 * term `swapfile' refers to either a blockdevice or an IS_REG file. Apart
132 * from setup, they're handled identically.
134 * We always assume that blocks are of size PAGE_SIZE.
137 struct list_head list;
140 sector_t start_block;
144 * Max bad pages in the new format..
146 #define __swapoffset(x) ((unsigned long)&((union swap_header *)0)->x)
147 #define MAX_SWAP_BADPAGES \
148 ((__swapoffset(magic.magic) - __swapoffset(info.badpages)) / sizeof(int))
151 SWP_USED = (1 << 0), /* is slot in swap_info[] used? */
152 SWP_WRITEOK = (1 << 1), /* ok to write to this swap? */
153 SWP_DISCARDABLE = (1 << 2), /* blkdev support discard */
154 SWP_DISCARDING = (1 << 3), /* now discarding a free cluster */
155 SWP_SOLIDSTATE = (1 << 4), /* blkdev seeks are cheap */
156 SWP_CONTINUED = (1 << 5), /* swap_map has count continuation */
157 SWP_BLKDEV = (1 << 6), /* its a block device */
158 SWP_FILE = (1 << 7), /* set after swap_activate success */
159 SWP_AREA_DISCARD = (1 << 8), /* single-time swap area discards */
160 SWP_PAGE_DISCARD = (1 << 9), /* freed swap page-cluster discards */
161 /* add others here before... */
162 SWP_SCANNING = (1 << 10), /* refcount in scan_swap_map */
165 #define SWAP_CLUSTER_MAX 32UL
166 #define COMPACT_CLUSTER_MAX SWAP_CLUSTER_MAX
169 * Ratio between zone->managed_pages and the "gap" that above the per-zone
170 * "high_wmark". While balancing nodes, We allow kswapd to shrink zones that
171 * do not meet the (high_wmark + gap) watermark, even which already met the
172 * high_wmark, in order to provide better per-zone lru behavior. We are ok to
173 * spend not more than 1% of the memory for this zone balancing "gap".
175 #define KSWAPD_ZONE_BALANCE_GAP_RATIO 100
177 #define SWAP_MAP_MAX 0x3e /* Max duplication count, in first swap_map */
178 #define SWAP_MAP_BAD 0x3f /* Note pageblock is bad, in first swap_map */
179 #define SWAP_HAS_CACHE 0x40 /* Flag page is cached, in first swap_map */
180 #define SWAP_CONT_MAX 0x7f /* Max count, in each swap_map continuation */
181 #define COUNT_CONTINUED 0x80 /* See swap_map continuation for full count */
182 #define SWAP_MAP_SHMEM 0xbf /* Owned by shmem/tmpfs, in first swap_map */
185 * We use this to track usage of a cluster. A cluster is a block of swap disk
186 * space with SWAPFILE_CLUSTER pages long and naturally aligns in disk. All
187 * free clusters are organized into a list. We fetch an entry from the list to
188 * get a free cluster.
190 * The data field stores next cluster if the cluster is free or cluster usage
191 * counter otherwise. The flags field determines if a cluster is free. This is
192 * protected by swap_info_struct.lock.
194 struct swap_cluster_info {
195 unsigned int data:24;
196 unsigned int flags:8;
198 #define CLUSTER_FLAG_FREE 1 /* This cluster is free */
199 #define CLUSTER_FLAG_NEXT_NULL 2 /* This cluster has no next cluster */
202 * We assign a cluster to each CPU, so each CPU can allocate swap entry from
203 * its own cluster and swapout sequentially. The purpose is to optimize swapout
206 struct percpu_cluster {
207 struct swap_cluster_info index; /* Current cluster index */
208 unsigned int next; /* Likely next allocation offset */
212 * The in-memory structure used to track swap areas.
214 struct swap_info_struct {
215 unsigned long flags; /* SWP_USED etc: see above */
216 signed short prio; /* swap priority of this type */
217 struct plist_node list; /* entry in swap_active_head */
218 struct plist_node avail_list; /* entry in swap_avail_head */
219 signed char type; /* strange name for an index */
220 unsigned int max; /* extent of the swap_map */
221 unsigned char *swap_map; /* vmalloc'ed array of usage counts */
222 struct swap_cluster_info *cluster_info; /* cluster info. Only for SSD */
223 struct swap_cluster_info free_cluster_head; /* free cluster list head */
224 struct swap_cluster_info free_cluster_tail; /* free cluster list tail */
225 unsigned int lowest_bit; /* index of first free in swap_map */
226 unsigned int highest_bit; /* index of last free in swap_map */
227 unsigned int pages; /* total of usable pages of swap */
228 unsigned int inuse_pages; /* number of those currently in use */
229 unsigned int cluster_next; /* likely index for next allocation */
230 unsigned int cluster_nr; /* countdown to next cluster search */
231 struct percpu_cluster __percpu *percpu_cluster; /* per cpu's swap location */
232 struct swap_extent *curr_swap_extent;
233 struct swap_extent first_swap_extent;
234 struct block_device *bdev; /* swap device or bdev of swap file */
235 struct file *swap_file; /* seldom referenced */
236 unsigned int old_block_size; /* seldom referenced */
237 #ifdef CONFIG_FRONTSWAP
238 unsigned long *frontswap_map; /* frontswap in-use, one bit per page */
239 atomic_t frontswap_pages; /* frontswap pages in-use counter */
242 * protect map scan related fields like
243 * swap_map, lowest_bit, highest_bit,
244 * inuse_pages, cluster_next,
245 * cluster_nr, lowest_alloc,
246 * highest_alloc, free/discard cluster
247 * list. other fields are only changed
248 * at swapon/swapoff, so are protected
249 * by swap_lock. changing flags need
250 * hold this lock and swap_lock. If
251 * both locks need hold, hold swap_lock
254 struct work_struct discard_work; /* discard worker */
255 struct swap_cluster_info discard_cluster_head; /* list head of discard clusters */
256 struct swap_cluster_info discard_cluster_tail; /* list tail of discard clusters */
259 /* linux/mm/workingset.c */
260 void *workingset_eviction(struct address_space *mapping, struct page *page);
261 bool workingset_refault(void *shadow);
262 void workingset_activation(struct page *page);
263 extern struct list_lru workingset_shadow_nodes;
265 static inline unsigned int workingset_node_pages(struct radix_tree_node *node)
267 return node->count & RADIX_TREE_COUNT_MASK;
270 static inline void workingset_node_pages_inc(struct radix_tree_node *node)
275 static inline void workingset_node_pages_dec(struct radix_tree_node *node)
280 static inline unsigned int workingset_node_shadows(struct radix_tree_node *node)
282 return node->count >> RADIX_TREE_COUNT_SHIFT;
285 static inline void workingset_node_shadows_inc(struct radix_tree_node *node)
287 node->count += 1U << RADIX_TREE_COUNT_SHIFT;
290 static inline void workingset_node_shadows_dec(struct radix_tree_node *node)
292 node->count -= 1U << RADIX_TREE_COUNT_SHIFT;
295 /* linux/mm/page_alloc.c */
296 extern unsigned long totalram_pages;
297 extern unsigned long totalreserve_pages;
298 extern unsigned long dirty_balance_reserve;
299 extern unsigned long nr_free_buffer_pages(void);
300 extern unsigned long nr_free_pagecache_pages(void);
302 /* Definition of global_page_state not available yet */
303 #define nr_free_pages() global_page_state(NR_FREE_PAGES)
306 /* linux/mm/swap.c */
307 extern void lru_cache_add(struct page *);
308 extern void lru_cache_add_anon(struct page *page);
309 extern void lru_cache_add_file(struct page *page);
310 extern void lru_add_page_tail(struct page *page, struct page *page_tail,
311 struct lruvec *lruvec, struct list_head *head);
312 extern void activate_page(struct page *);
313 extern void mark_page_accessed(struct page *);
314 extern void lru_add_drain(void);
315 extern void lru_add_drain_cpu(int cpu);
316 extern void lru_add_drain_all(void);
317 extern void rotate_reclaimable_page(struct page *page);
318 extern void deactivate_page(struct page *page);
319 extern void swap_setup(void);
321 extern void add_page_to_unevictable_list(struct page *page);
323 /* linux/mm/vmscan.c */
324 extern unsigned long try_to_free_pages(struct zonelist *zonelist, int order,
325 gfp_t gfp_mask, nodemask_t *mask);
326 extern int __isolate_lru_page(struct page *page, isolate_mode_t mode);
327 extern unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *mem,
328 gfp_t gfp_mask, bool noswap);
329 extern unsigned long mem_cgroup_shrink_node_zone(struct mem_cgroup *mem,
330 gfp_t gfp_mask, bool noswap,
332 unsigned long *nr_scanned);
333 extern unsigned long shrink_all_memory(unsigned long nr_pages);
334 extern int vm_swappiness;
335 extern int remove_mapping(struct address_space *mapping, struct page *page);
336 extern unsigned long vm_total_pages;
339 extern int zone_reclaim_mode;
340 extern int sysctl_min_unmapped_ratio;
341 extern int sysctl_min_slab_ratio;
342 extern int zone_reclaim(struct zone *, gfp_t, unsigned int);
344 #define zone_reclaim_mode 0
345 static inline int zone_reclaim(struct zone *z, gfp_t mask, unsigned int order)
351 extern int page_evictable(struct page *page);
352 extern void check_move_unevictable_pages(struct page **, int nr_pages);
354 extern unsigned long scan_unevictable_pages;
355 extern int scan_unevictable_handler(struct ctl_table *, int,
356 void __user *, size_t *, loff_t *);
358 extern int scan_unevictable_register_node(struct node *node);
359 extern void scan_unevictable_unregister_node(struct node *node);
361 static inline int scan_unevictable_register_node(struct node *node)
365 static inline void scan_unevictable_unregister_node(struct node *node)
370 extern int kswapd_run(int nid);
371 extern void kswapd_stop(int nid);
373 extern int mem_cgroup_swappiness(struct mem_cgroup *mem);
375 static inline int mem_cgroup_swappiness(struct mem_cgroup *mem)
377 return vm_swappiness;
380 #ifdef CONFIG_MEMCG_SWAP
381 extern void mem_cgroup_uncharge_swap(swp_entry_t ent);
383 static inline void mem_cgroup_uncharge_swap(swp_entry_t ent)
388 /* linux/mm/page_io.c */
389 extern int swap_readpage(struct page *);
390 extern int swap_writepage(struct page *page, struct writeback_control *wbc);
391 extern void end_swap_bio_write(struct bio *bio, int err);
392 extern int __swap_writepage(struct page *page, struct writeback_control *wbc,
393 void (*end_write_func)(struct bio *, int));
394 extern int swap_set_page_dirty(struct page *page);
395 extern void end_swap_bio_read(struct bio *bio, int err);
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 *swapin_readahead(swp_entry_t, gfp_t,
418 struct vm_area_struct *vma, unsigned long addr);
420 /* linux/mm/swapfile.c */
421 extern atomic_long_t nr_swap_pages;
422 extern long total_swap_pages;
424 /* Swap 50% full? Release swapcache more aggressively.. */
425 static inline bool vm_swap_full(void)
427 return atomic_long_read(&nr_swap_pages) * 2 < total_swap_pages;
430 static inline long get_nr_swap_pages(void)
432 return atomic_long_read(&nr_swap_pages);
435 extern void si_swapinfo(struct sysinfo *);
436 extern swp_entry_t get_swap_page(void);
437 extern swp_entry_t get_swap_page_of_type(int);
438 extern int add_swap_count_continuation(swp_entry_t, gfp_t);
439 extern void swap_shmem_alloc(swp_entry_t);
440 extern int swap_duplicate(swp_entry_t);
441 extern int swapcache_prepare(swp_entry_t);
442 extern void swap_free(swp_entry_t);
443 extern void swapcache_free(swp_entry_t, struct page *page);
444 extern int free_swap_and_cache(swp_entry_t);
445 extern int swap_type_of(dev_t, sector_t, struct block_device **);
446 extern unsigned int count_swap_pages(int, int);
447 extern sector_t map_swap_page(struct page *, struct block_device **);
448 extern sector_t swapdev_block(int, pgoff_t);
449 extern int page_swapcount(struct page *);
450 extern struct swap_info_struct *page_swap_info(struct page *);
451 extern int reuse_swap_page(struct page *);
452 extern int try_to_free_swap(struct page *);
453 struct backing_dev_info;
457 mem_cgroup_uncharge_swapcache(struct page *page, swp_entry_t ent, bool swapout);
460 mem_cgroup_uncharge_swapcache(struct page *page, swp_entry_t ent, bool swapout)
465 #else /* CONFIG_SWAP */
467 #define swap_address_space(entry) (NULL)
468 #define get_nr_swap_pages() 0L
469 #define total_swap_pages 0L
470 #define total_swapcache_pages() 0UL
471 #define vm_swap_full() 0
473 #define si_swapinfo(val) \
474 do { (val)->freeswap = (val)->totalswap = 0; } while (0)
475 /* only sparc can not include linux/pagemap.h in this file
476 * so leave page_cache_release and release_pages undeclared... */
477 #define free_page_and_swap_cache(page) \
478 page_cache_release(page)
479 #define free_pages_and_swap_cache(pages, nr) \
480 release_pages((pages), (nr), false);
482 static inline void show_swap_cache_info(void)
486 #define free_swap_and_cache(swp) is_migration_entry(swp)
487 #define swapcache_prepare(swp) is_migration_entry(swp)
489 static inline int add_swap_count_continuation(swp_entry_t swp, gfp_t gfp_mask)
494 static inline void swap_shmem_alloc(swp_entry_t swp)
498 static inline int swap_duplicate(swp_entry_t swp)
503 static inline void swap_free(swp_entry_t swp)
507 static inline void swapcache_free(swp_entry_t swp, struct page *page)
511 static inline struct page *swapin_readahead(swp_entry_t swp, gfp_t gfp_mask,
512 struct vm_area_struct *vma, unsigned long addr)
517 static inline int swap_writepage(struct page *p, struct writeback_control *wbc)
522 static inline struct page *lookup_swap_cache(swp_entry_t swp)
527 static inline int add_to_swap(struct page *page, struct list_head *list)
532 static inline int add_to_swap_cache(struct page *page, swp_entry_t entry,
538 static inline void __delete_from_swap_cache(struct page *page)
542 static inline void delete_from_swap_cache(struct page *page)
546 static inline int page_swapcount(struct page *page)
551 #define reuse_swap_page(page) (page_mapcount(page) == 1)
553 static inline int try_to_free_swap(struct page *page)
558 static inline swp_entry_t get_swap_page(void)
566 mem_cgroup_uncharge_swapcache(struct page *page, swp_entry_t ent)
570 #endif /* CONFIG_SWAP */
571 #endif /* __KERNEL__*/
572 #endif /* _LINUX_SWAP_H */