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)
32 static inline int current_is_kswapd(void)
34 return current
->flags
& PF_KSWAPD
;
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.
45 #define MAX_SWAPFILES_SHIFT 5
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
54 * NUMA node memory migration support
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)
61 #define SWP_MIGRATION_NUM 0
65 * Handling of hardware poisoned pages with memory corruption.
67 #ifdef CONFIG_MEMORY_FAILURE
68 #define SWP_HWPOISON_NUM 1
69 #define SWP_HWPOISON MAX_SWAPFILES
71 #define SWP_HWPOISON_NUM 0
74 #define MAX_SWAPFILES \
75 ((1 << MAX_SWAPFILES_SHIFT) - SWP_MIGRATION_NUM - SWP_HWPOISON_NUM)
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...
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
91 char reserved
[PAGE_SIZE
- 10];
92 char magic
[10]; /* SWAP-SPACE or SWAPSPACE2 */
95 char bootbits
[1024]; /* Space for disklabel etc. */
99 unsigned char sws_uuid
[16];
100 unsigned char sws_volume
[16];
107 * current->reclaim_state points to one of these when a task is running
110 struct reclaim_state
{
111 unsigned long reclaimed_slab
;
116 struct address_space
;
118 struct writeback_control
;
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.
127 * We always assume that blocks are of size PAGE_SIZE.
130 struct list_head list
;
133 sector_t start_block
;
137 * Max bad pages in the new format..
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))
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 */
159 #define SWAP_CLUSTER_MAX 32UL
160 #define COMPACT_CLUSTER_MAX SWAP_CLUSTER_MAX
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 */
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.
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.
179 struct swap_cluster_info
{
181 * Protect swap_cluster_info fields
182 * and swap_info_struct->swap_map
183 * elements correspond to the swap
186 unsigned int data
:24;
187 unsigned int flags
:8;
189 #define CLUSTER_FLAG_FREE 1 /* This cluster is free */
190 #define CLUSTER_FLAG_NEXT_NULL 2 /* This cluster has no next cluster */
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
197 struct percpu_cluster
{
198 struct swap_cluster_info index
; /* Current cluster index */
199 unsigned int next
; /* Likely next allocation offset */
202 struct swap_cluster_list
{
203 struct swap_cluster_info head
;
204 struct swap_cluster_info tail
;
208 * The in-memory structure used to track swap areas.
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 */
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_list discard_clusters
; /* discard clusters list */
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);
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);
265 /* Definition of global_page_state not available yet */
266 #define nr_free_pages() global_page_state(NR_FREE_PAGES)
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 rotate_reclaimable_page(struct page
*page
);
281 extern void deactivate_file_page(struct page
*page
);
282 extern void mark_page_lazyfree(struct page
*page
);
283 extern void swap_setup(void);
285 extern void add_page_to_unevictable_list(struct page
*page
);
287 extern void lru_cache_add_active_or_unevictable(struct page
*page
,
288 struct vm_area_struct
*vma
);
290 /* linux/mm/vmscan.c */
291 extern unsigned long zone_reclaimable_pages(struct zone
*zone
);
292 extern unsigned long pgdat_reclaimable_pages(struct pglist_data
*pgdat
);
293 extern unsigned long try_to_free_pages(struct zonelist
*zonelist
, int order
,
294 gfp_t gfp_mask
, nodemask_t
*mask
);
295 extern int __isolate_lru_page(struct page
*page
, isolate_mode_t mode
);
296 extern unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup
*memcg
,
297 unsigned long nr_pages
,
300 extern unsigned long mem_cgroup_shrink_node(struct mem_cgroup
*mem
,
301 gfp_t gfp_mask
, bool noswap
,
303 unsigned long *nr_scanned
);
304 extern unsigned long shrink_all_memory(unsigned long nr_pages
);
305 extern int vm_swappiness
;
306 extern int remove_mapping(struct address_space
*mapping
, struct page
*page
);
307 extern unsigned long vm_total_pages
;
310 extern int node_reclaim_mode
;
311 extern int sysctl_min_unmapped_ratio
;
312 extern int sysctl_min_slab_ratio
;
313 extern int node_reclaim(struct pglist_data
*, gfp_t
, unsigned int);
315 #define node_reclaim_mode 0
316 static inline int node_reclaim(struct pglist_data
*pgdat
, gfp_t mask
,
323 extern int page_evictable(struct page
*page
);
324 extern void check_move_unevictable_pages(struct page
**, int nr_pages
);
326 extern int kswapd_run(int nid
);
327 extern void kswapd_stop(int nid
);
331 #include <linux/blk_types.h> /* for bio_end_io_t */
333 /* linux/mm/page_io.c */
334 extern int swap_readpage(struct page
*);
335 extern int swap_writepage(struct page
*page
, struct writeback_control
*wbc
);
336 extern void end_swap_bio_write(struct bio
*bio
);
337 extern int __swap_writepage(struct page
*page
, struct writeback_control
*wbc
,
338 bio_end_io_t end_write_func
);
339 extern int swap_set_page_dirty(struct page
*page
);
341 int add_swap_extent(struct swap_info_struct
*sis
, unsigned long start_page
,
342 unsigned long nr_pages
, sector_t start_block
);
343 int generic_swapfile_activate(struct swap_info_struct
*, struct file
*,
346 /* linux/mm/swap_state.c */
347 /* One swap address space for each 64M swap space */
348 #define SWAP_ADDRESS_SPACE_SHIFT 14
349 #define SWAP_ADDRESS_SPACE_PAGES (1 << SWAP_ADDRESS_SPACE_SHIFT)
350 extern struct address_space
*swapper_spaces
[];
351 #define swap_address_space(entry) \
352 (&swapper_spaces[swp_type(entry)][swp_offset(entry) \
353 >> SWAP_ADDRESS_SPACE_SHIFT])
354 extern unsigned long total_swapcache_pages(void);
355 extern void show_swap_cache_info(void);
356 extern int add_to_swap(struct page
*page
);
357 extern int add_to_swap_cache(struct page
*, swp_entry_t
, gfp_t
);
358 extern int __add_to_swap_cache(struct page
*page
, swp_entry_t entry
);
359 extern void __delete_from_swap_cache(struct page
*);
360 extern void delete_from_swap_cache(struct page
*);
361 extern void free_page_and_swap_cache(struct page
*);
362 extern void free_pages_and_swap_cache(struct page
**, int);
363 extern struct page
*lookup_swap_cache(swp_entry_t
);
364 extern struct page
*read_swap_cache_async(swp_entry_t
, gfp_t
,
365 struct vm_area_struct
*vma
, unsigned long addr
);
366 extern struct page
*__read_swap_cache_async(swp_entry_t
, gfp_t
,
367 struct vm_area_struct
*vma
, unsigned long addr
,
368 bool *new_page_allocated
);
369 extern struct page
*swapin_readahead(swp_entry_t
, gfp_t
,
370 struct vm_area_struct
*vma
, unsigned long addr
);
372 /* linux/mm/swapfile.c */
373 extern atomic_long_t nr_swap_pages
;
374 extern long total_swap_pages
;
375 extern bool has_usable_swap(void);
377 /* Swap 50% full? Release swapcache more aggressively.. */
378 static inline bool vm_swap_full(void)
380 return atomic_long_read(&nr_swap_pages
) * 2 < total_swap_pages
;
383 static inline long get_nr_swap_pages(void)
385 return atomic_long_read(&nr_swap_pages
);
388 extern void si_swapinfo(struct sysinfo
*);
389 extern swp_entry_t
get_swap_page(struct page
*page
);
390 extern void put_swap_page(struct page
*page
, swp_entry_t entry
);
391 extern swp_entry_t
get_swap_page_of_type(int);
392 extern int get_swap_pages(int n
, bool cluster
, swp_entry_t swp_entries
[]);
393 extern int add_swap_count_continuation(swp_entry_t
, gfp_t
);
394 extern void swap_shmem_alloc(swp_entry_t
);
395 extern int swap_duplicate(swp_entry_t
);
396 extern int swapcache_prepare(swp_entry_t
);
397 extern void swap_free(swp_entry_t
);
398 extern void swapcache_free_entries(swp_entry_t
*entries
, int n
);
399 extern int free_swap_and_cache(swp_entry_t
);
400 extern int swap_type_of(dev_t
, sector_t
, struct block_device
**);
401 extern unsigned int count_swap_pages(int, int);
402 extern sector_t
map_swap_page(struct page
*, struct block_device
**);
403 extern sector_t
swapdev_block(int, pgoff_t
);
404 extern int page_swapcount(struct page
*);
405 extern int __swp_swapcount(swp_entry_t entry
);
406 extern int swp_swapcount(swp_entry_t entry
);
407 extern struct swap_info_struct
*page_swap_info(struct page
*);
408 extern bool reuse_swap_page(struct page
*, int *);
409 extern int try_to_free_swap(struct page
*);
410 struct backing_dev_info
;
411 extern int init_swap_address_space(unsigned int type
, unsigned long nr_pages
);
412 extern void exit_swap_address_space(unsigned int type
);
414 #else /* CONFIG_SWAP */
416 #define swap_address_space(entry) (NULL)
417 #define get_nr_swap_pages() 0L
418 #define total_swap_pages 0L
419 #define total_swapcache_pages() 0UL
420 #define vm_swap_full() 0
422 #define si_swapinfo(val) \
423 do { (val)->freeswap = (val)->totalswap = 0; } while (0)
424 /* only sparc can not include linux/pagemap.h in this file
425 * so leave put_page and release_pages undeclared... */
426 #define free_page_and_swap_cache(page) \
428 #define free_pages_and_swap_cache(pages, nr) \
429 release_pages((pages), (nr), false);
431 static inline void show_swap_cache_info(void)
435 #define free_swap_and_cache(swp) is_migration_entry(swp)
436 #define swapcache_prepare(swp) is_migration_entry(swp)
438 static inline int add_swap_count_continuation(swp_entry_t swp
, gfp_t gfp_mask
)
443 static inline void swap_shmem_alloc(swp_entry_t swp
)
447 static inline int swap_duplicate(swp_entry_t swp
)
452 static inline void swap_free(swp_entry_t swp
)
456 static inline void put_swap_page(struct page
*page
, swp_entry_t swp
)
460 static inline struct page
*swapin_readahead(swp_entry_t swp
, gfp_t gfp_mask
,
461 struct vm_area_struct
*vma
, unsigned long addr
)
466 static inline int swap_writepage(struct page
*p
, struct writeback_control
*wbc
)
471 static inline struct page
*lookup_swap_cache(swp_entry_t swp
)
476 static inline int add_to_swap(struct page
*page
)
481 static inline int add_to_swap_cache(struct page
*page
, swp_entry_t entry
,
487 static inline void __delete_from_swap_cache(struct page
*page
)
491 static inline void delete_from_swap_cache(struct page
*page
)
495 static inline int page_swapcount(struct page
*page
)
500 static inline int __swp_swapcount(swp_entry_t entry
)
505 static inline int swp_swapcount(swp_entry_t entry
)
510 #define reuse_swap_page(page, total_mapcount) \
511 (page_trans_huge_mapcount(page, total_mapcount) == 1)
513 static inline int try_to_free_swap(struct page
*page
)
518 static inline swp_entry_t
get_swap_page(struct page
*page
)
525 #endif /* CONFIG_SWAP */
528 static inline int mem_cgroup_swappiness(struct mem_cgroup
*memcg
)
530 /* Cgroup2 doesn't have per-cgroup swappiness */
531 if (cgroup_subsys_on_dfl(memory_cgrp_subsys
))
532 return vm_swappiness
;
535 if (mem_cgroup_disabled() || !memcg
->css
.parent
)
536 return vm_swappiness
;
538 return memcg
->swappiness
;
542 static inline int mem_cgroup_swappiness(struct mem_cgroup
*mem
)
544 return vm_swappiness
;
548 #ifdef CONFIG_MEMCG_SWAP
549 extern void mem_cgroup_swapout(struct page
*page
, swp_entry_t entry
);
550 extern int mem_cgroup_try_charge_swap(struct page
*page
, swp_entry_t entry
);
551 extern void mem_cgroup_uncharge_swap(swp_entry_t entry
, unsigned int nr_pages
);
552 extern long mem_cgroup_get_nr_swap_pages(struct mem_cgroup
*memcg
);
553 extern bool mem_cgroup_swap_full(struct page
*page
);
555 static inline void mem_cgroup_swapout(struct page
*page
, swp_entry_t entry
)
559 static inline int mem_cgroup_try_charge_swap(struct page
*page
,
565 static inline void mem_cgroup_uncharge_swap(swp_entry_t entry
,
566 unsigned int nr_pages
)
570 static inline long mem_cgroup_get_nr_swap_pages(struct mem_cgroup
*memcg
)
572 return get_nr_swap_pages();
575 static inline bool mem_cgroup_swap_full(struct page
*page
)
577 return vm_swap_full();
581 #endif /* __KERNEL__*/
582 #endif /* _LINUX_SWAP_H */