| 1 | /* SPDX-License-Identifier: GPL-2.0 */ | 
|---|
| 2 | #ifndef _LINUX_SWAP_H | 
|---|
| 3 | #define _LINUX_SWAP_H | 
|---|
| 4 |  | 
|---|
| 5 | #include <linux/spinlock.h> | 
|---|
| 6 | #include <linux/linkage.h> | 
|---|
| 7 | #include <linux/mmzone.h> | 
|---|
| 8 | #include <linux/list.h> | 
|---|
| 9 | #include <linux/memcontrol.h> | 
|---|
| 10 | #include <linux/sched.h> | 
|---|
| 11 | #include <linux/node.h> | 
|---|
| 12 | #include <linux/fs.h> | 
|---|
| 13 | #include <linux/pagemap.h> | 
|---|
| 14 | #include <linux/atomic.h> | 
|---|
| 15 | #include <linux/page-flags.h> | 
|---|
| 16 | #include <uapi/linux/mempolicy.h> | 
|---|
| 17 | #include <asm/page.h> | 
|---|
| 18 |  | 
|---|
| 19 | struct notifier_block; | 
|---|
| 20 |  | 
|---|
| 21 | struct bio; | 
|---|
| 22 |  | 
|---|
| 23 | struct pagevec; | 
|---|
| 24 |  | 
|---|
| 25 | #define SWAP_FLAG_PREFER	0x8000	/* set if swap priority specified */ | 
|---|
| 26 | #define SWAP_FLAG_PRIO_MASK	0x7fff | 
|---|
| 27 | #define SWAP_FLAG_DISCARD	0x10000 /* enable discard for swap */ | 
|---|
| 28 | #define SWAP_FLAG_DISCARD_ONCE	0x20000 /* discard swap area at swapon-time */ | 
|---|
| 29 | #define SWAP_FLAG_DISCARD_PAGES 0x40000 /* discard page-clusters after use */ | 
|---|
| 30 |  | 
|---|
| 31 | #define SWAP_FLAGS_VALID	(SWAP_FLAG_PRIO_MASK | SWAP_FLAG_PREFER | \ | 
|---|
| 32 | SWAP_FLAG_DISCARD | SWAP_FLAG_DISCARD_ONCE | \ | 
|---|
| 33 | SWAP_FLAG_DISCARD_PAGES) | 
|---|
| 34 | #define SWAP_BATCH 64 | 
|---|
| 35 |  | 
|---|
| 36 | static inline int current_is_kswapd(void) | 
|---|
| 37 | { | 
|---|
| 38 | return current->flags & PF_KSWAPD; | 
|---|
| 39 | } | 
|---|
| 40 |  | 
|---|
| 41 | /* | 
|---|
| 42 | * MAX_SWAPFILES defines the maximum number of swaptypes: things which can | 
|---|
| 43 | * be swapped to.  The swap type and the offset into that swap type are | 
|---|
| 44 | * encoded into pte's and into pgoff_t's in the swapcache.  Using five bits | 
|---|
| 45 | * for the type means that the maximum number of swapcache pages is 27 bits | 
|---|
| 46 | * on 32-bit-pgoff_t architectures.  And that assumes that the architecture packs | 
|---|
| 47 | * the type/offset into the pte as 5/27 as well. | 
|---|
| 48 | */ | 
|---|
| 49 | #define MAX_SWAPFILES_SHIFT	5 | 
|---|
| 50 |  | 
|---|
| 51 | /* | 
|---|
| 52 | * Use some of the swap files numbers for other purposes. This | 
|---|
| 53 | * is a convenient way to hook into the VM to trigger special | 
|---|
| 54 | * actions on faults. | 
|---|
| 55 | */ | 
|---|
| 56 |  | 
|---|
| 57 | /* | 
|---|
| 58 | * PTE markers are used to persist information onto PTEs that otherwise | 
|---|
| 59 | * should be a none pte.  As its name "PTE" hints, it should only be | 
|---|
| 60 | * applied to the leaves of pgtables. | 
|---|
| 61 | */ | 
|---|
| 62 | #define SWP_PTE_MARKER_NUM 1 | 
|---|
| 63 | #define SWP_PTE_MARKER     (MAX_SWAPFILES + SWP_HWPOISON_NUM + \ | 
|---|
| 64 | SWP_MIGRATION_NUM + SWP_DEVICE_NUM) | 
|---|
| 65 |  | 
|---|
| 66 | /* | 
|---|
| 67 | * Unaddressable device memory support. See include/linux/hmm.h and | 
|---|
| 68 | * Documentation/mm/hmm.rst. Short description is we need struct pages for | 
|---|
| 69 | * device memory that is unaddressable (inaccessible) by CPU, so that we can | 
|---|
| 70 | * migrate part of a process memory to device memory. | 
|---|
| 71 | * | 
|---|
| 72 | * When a page is migrated from CPU to device, we set the CPU page table entry | 
|---|
| 73 | * to a special SWP_DEVICE_{READ|WRITE} entry. | 
|---|
| 74 | * | 
|---|
| 75 | * When a page is mapped by the device for exclusive access we set the CPU page | 
|---|
| 76 | * table entries to a special SWP_DEVICE_EXCLUSIVE entry. | 
|---|
| 77 | */ | 
|---|
| 78 | #ifdef CONFIG_DEVICE_PRIVATE | 
|---|
| 79 | #define SWP_DEVICE_NUM 3 | 
|---|
| 80 | #define SWP_DEVICE_WRITE (MAX_SWAPFILES+SWP_HWPOISON_NUM+SWP_MIGRATION_NUM) | 
|---|
| 81 | #define SWP_DEVICE_READ (MAX_SWAPFILES+SWP_HWPOISON_NUM+SWP_MIGRATION_NUM+1) | 
|---|
| 82 | #define SWP_DEVICE_EXCLUSIVE (MAX_SWAPFILES+SWP_HWPOISON_NUM+SWP_MIGRATION_NUM+2) | 
|---|
| 83 | #else | 
|---|
| 84 | #define SWP_DEVICE_NUM 0 | 
|---|
| 85 | #endif | 
|---|
| 86 |  | 
|---|
| 87 | /* | 
|---|
| 88 | * Page migration support. | 
|---|
| 89 | * | 
|---|
| 90 | * SWP_MIGRATION_READ_EXCLUSIVE is only applicable to anonymous pages and | 
|---|
| 91 | * indicates that the referenced (part of) an anonymous page is exclusive to | 
|---|
| 92 | * a single process. For SWP_MIGRATION_WRITE, that information is implicit: | 
|---|
| 93 | * (part of) an anonymous page that are mapped writable are exclusive to a | 
|---|
| 94 | * single process. | 
|---|
| 95 | */ | 
|---|
| 96 | #ifdef CONFIG_MIGRATION | 
|---|
| 97 | #define SWP_MIGRATION_NUM 3 | 
|---|
| 98 | #define SWP_MIGRATION_READ (MAX_SWAPFILES + SWP_HWPOISON_NUM) | 
|---|
| 99 | #define SWP_MIGRATION_READ_EXCLUSIVE (MAX_SWAPFILES + SWP_HWPOISON_NUM + 1) | 
|---|
| 100 | #define SWP_MIGRATION_WRITE (MAX_SWAPFILES + SWP_HWPOISON_NUM + 2) | 
|---|
| 101 | #else | 
|---|
| 102 | #define SWP_MIGRATION_NUM 0 | 
|---|
| 103 | #endif | 
|---|
| 104 |  | 
|---|
| 105 | /* | 
|---|
| 106 | * Handling of hardware poisoned pages with memory corruption. | 
|---|
| 107 | */ | 
|---|
| 108 | #ifdef CONFIG_MEMORY_FAILURE | 
|---|
| 109 | #define SWP_HWPOISON_NUM 1 | 
|---|
| 110 | #define SWP_HWPOISON		MAX_SWAPFILES | 
|---|
| 111 | #else | 
|---|
| 112 | #define SWP_HWPOISON_NUM 0 | 
|---|
| 113 | #endif | 
|---|
| 114 |  | 
|---|
| 115 | #define MAX_SWAPFILES \ | 
|---|
| 116 | ((1 << MAX_SWAPFILES_SHIFT) - SWP_DEVICE_NUM - \ | 
|---|
| 117 | SWP_MIGRATION_NUM - SWP_HWPOISON_NUM - \ | 
|---|
| 118 | SWP_PTE_MARKER_NUM) | 
|---|
| 119 |  | 
|---|
| 120 | /* | 
|---|
| 121 | * Magic header for a swap area. The first part of the union is | 
|---|
| 122 | * what the swap magic looks like for the old (limited to 128MB) | 
|---|
| 123 | * swap area format, the second part of the union adds - in the | 
|---|
| 124 | * old reserved area - some extra information. Note that the first | 
|---|
| 125 | * kilobyte is reserved for boot loader or disk label stuff... | 
|---|
| 126 | * | 
|---|
| 127 | * Having the magic at the end of the PAGE_SIZE makes detecting swap | 
|---|
| 128 | * areas somewhat tricky on machines that support multiple page sizes. | 
|---|
| 129 | * For 2.5 we'll probably want to move the magic to just beyond the | 
|---|
| 130 | * bootbits... | 
|---|
| 131 | */ | 
|---|
| 132 | union  { | 
|---|
| 133 | struct { | 
|---|
| 134 | char [PAGE_SIZE - 10]; | 
|---|
| 135 | char [10];			/* SWAP-SPACE or SWAPSPACE2 */ | 
|---|
| 136 | } ; | 
|---|
| 137 | struct { | 
|---|
| 138 | char		[1024];	/* Space for disklabel etc. */ | 
|---|
| 139 | __u32		; | 
|---|
| 140 | __u32		; | 
|---|
| 141 | __u32		; | 
|---|
| 142 | unsigned char	[16]; | 
|---|
| 143 | unsigned char	[16]; | 
|---|
| 144 | __u32		[117]; | 
|---|
| 145 | __u32		[1]; | 
|---|
| 146 | } ; | 
|---|
| 147 | }; | 
|---|
| 148 |  | 
|---|
| 149 | /* | 
|---|
| 150 | * current->reclaim_state points to one of these when a task is running | 
|---|
| 151 | * memory reclaim | 
|---|
| 152 | */ | 
|---|
| 153 | struct reclaim_state { | 
|---|
| 154 | /* pages reclaimed outside of LRU-based reclaim */ | 
|---|
| 155 | unsigned long reclaimed; | 
|---|
| 156 | #ifdef CONFIG_LRU_GEN | 
|---|
| 157 | /* per-thread mm walk data */ | 
|---|
| 158 | struct lru_gen_mm_walk *mm_walk; | 
|---|
| 159 | #endif | 
|---|
| 160 | }; | 
|---|
| 161 |  | 
|---|
| 162 | /* | 
|---|
| 163 | * mm_account_reclaimed_pages(): account reclaimed pages outside of LRU-based | 
|---|
| 164 | * reclaim | 
|---|
| 165 | * @pages: number of pages reclaimed | 
|---|
| 166 | * | 
|---|
| 167 | * If the current process is undergoing a reclaim operation, increment the | 
|---|
| 168 | * number of reclaimed pages by @pages. | 
|---|
| 169 | */ | 
|---|
| 170 | static inline void mm_account_reclaimed_pages(unsigned long pages) | 
|---|
| 171 | { | 
|---|
| 172 | if (current->reclaim_state) | 
|---|
| 173 | current->reclaim_state->reclaimed += pages; | 
|---|
| 174 | } | 
|---|
| 175 |  | 
|---|
| 176 | #ifdef __KERNEL__ | 
|---|
| 177 |  | 
|---|
| 178 | struct address_space; | 
|---|
| 179 | struct sysinfo; | 
|---|
| 180 | struct writeback_control; | 
|---|
| 181 | struct zone; | 
|---|
| 182 |  | 
|---|
| 183 | /* | 
|---|
| 184 | * A swap extent maps a range of a swapfile's PAGE_SIZE pages onto a range of | 
|---|
| 185 | * disk blocks.  A rbtree of swap extents maps the entire swapfile (Where the | 
|---|
| 186 | * term `swapfile' refers to either a blockdevice or an IS_REG file). Apart | 
|---|
| 187 | * from setup, they're handled identically. | 
|---|
| 188 | * | 
|---|
| 189 | * We always assume that blocks are of size PAGE_SIZE. | 
|---|
| 190 | */ | 
|---|
| 191 | struct swap_extent { | 
|---|
| 192 | struct rb_node rb_node; | 
|---|
| 193 | pgoff_t start_page; | 
|---|
| 194 | pgoff_t nr_pages; | 
|---|
| 195 | sector_t start_block; | 
|---|
| 196 | }; | 
|---|
| 197 |  | 
|---|
| 198 | /* | 
|---|
| 199 | * Max bad pages in the new format.. | 
|---|
| 200 | */ | 
|---|
| 201 | #define MAX_SWAP_BADPAGES \ | 
|---|
| 202 | ((offsetof(union swap_header, magic.magic) - \ | 
|---|
| 203 | offsetof(union swap_header, info.badpages)) / sizeof(int)) | 
|---|
| 204 |  | 
|---|
| 205 | enum { | 
|---|
| 206 | SWP_USED	= (1 << 0),	/* is slot in swap_info[] used? */ | 
|---|
| 207 | SWP_WRITEOK	= (1 << 1),	/* ok to write to this swap?	*/ | 
|---|
| 208 | SWP_DISCARDABLE = (1 << 2),	/* blkdev support discard */ | 
|---|
| 209 | SWP_DISCARDING	= (1 << 3),	/* now discarding a free cluster */ | 
|---|
| 210 | SWP_SOLIDSTATE	= (1 << 4),	/* blkdev seeks are cheap */ | 
|---|
| 211 | SWP_CONTINUED	= (1 << 5),	/* swap_map has count continuation */ | 
|---|
| 212 | SWP_BLKDEV	= (1 << 6),	/* its a block device */ | 
|---|
| 213 | SWP_ACTIVATED	= (1 << 7),	/* set after swap_activate success */ | 
|---|
| 214 | SWP_FS_OPS	= (1 << 8),	/* swapfile operations go through fs */ | 
|---|
| 215 | SWP_AREA_DISCARD = (1 << 9),	/* single-time swap area discards */ | 
|---|
| 216 | SWP_PAGE_DISCARD = (1 << 10),	/* freed swap page-cluster discards */ | 
|---|
| 217 | SWP_STABLE_WRITES = (1 << 11),	/* no overwrite PG_writeback pages */ | 
|---|
| 218 | SWP_SYNCHRONOUS_IO = (1 << 12),	/* synchronous IO is efficient */ | 
|---|
| 219 | /* add others here before... */ | 
|---|
| 220 | }; | 
|---|
| 221 |  | 
|---|
| 222 | #define SWAP_CLUSTER_MAX 32UL | 
|---|
| 223 | #define SWAP_CLUSTER_MAX_SKIPPED (SWAP_CLUSTER_MAX << 10) | 
|---|
| 224 | #define COMPACT_CLUSTER_MAX SWAP_CLUSTER_MAX | 
|---|
| 225 |  | 
|---|
| 226 | /* Bit flag in swap_map */ | 
|---|
| 227 | #define SWAP_HAS_CACHE	0x40	/* Flag page is cached, in first swap_map */ | 
|---|
| 228 | #define COUNT_CONTINUED	0x80	/* Flag swap_map continuation for full count */ | 
|---|
| 229 |  | 
|---|
| 230 | /* Special value in first swap_map */ | 
|---|
| 231 | #define SWAP_MAP_MAX	0x3e	/* Max count */ | 
|---|
| 232 | #define SWAP_MAP_BAD	0x3f	/* Note page is bad */ | 
|---|
| 233 | #define SWAP_MAP_SHMEM	0xbf	/* Owned by shmem/tmpfs */ | 
|---|
| 234 |  | 
|---|
| 235 | /* Special value in each swap_map continuation */ | 
|---|
| 236 | #define SWAP_CONT_MAX	0x7f	/* Max count */ | 
|---|
| 237 |  | 
|---|
| 238 | /* | 
|---|
| 239 | * The first page in the swap file is the swap header, which is always marked | 
|---|
| 240 | * bad to prevent it from being allocated as an entry. This also prevents the | 
|---|
| 241 | * cluster to which it belongs being marked free. Therefore 0 is safe to use as | 
|---|
| 242 | * a sentinel to indicate an entry is not valid. | 
|---|
| 243 | */ | 
|---|
| 244 | #define SWAP_ENTRY_INVALID	0 | 
|---|
| 245 |  | 
|---|
| 246 | #ifdef CONFIG_THP_SWAP | 
|---|
| 247 | #define SWAP_NR_ORDERS		(PMD_ORDER + 1) | 
|---|
| 248 | #else | 
|---|
| 249 | #define SWAP_NR_ORDERS		1 | 
|---|
| 250 | #endif | 
|---|
| 251 |  | 
|---|
| 252 | /* | 
|---|
| 253 | * We keep using same cluster for rotational device so IO will be sequential. | 
|---|
| 254 | * The purpose is to optimize SWAP throughput on these device. | 
|---|
| 255 | */ | 
|---|
| 256 | struct swap_sequential_cluster { | 
|---|
| 257 | unsigned int next[SWAP_NR_ORDERS]; /* Likely next allocation offset */ | 
|---|
| 258 | }; | 
|---|
| 259 |  | 
|---|
| 260 | /* | 
|---|
| 261 | * The in-memory structure used to track swap areas. | 
|---|
| 262 | */ | 
|---|
| 263 | struct swap_info_struct { | 
|---|
| 264 | struct percpu_ref users;	/* indicate and keep swap device valid. */ | 
|---|
| 265 | unsigned long	flags;		/* SWP_USED etc: see above */ | 
|---|
| 266 | signed short	prio;		/* swap priority of this type */ | 
|---|
| 267 | struct plist_node list;		/* entry in swap_active_head */ | 
|---|
| 268 | signed char	type;		/* strange name for an index */ | 
|---|
| 269 | unsigned int	max;		/* extent of the swap_map */ | 
|---|
| 270 | unsigned char *swap_map;	/* vmalloc'ed array of usage counts */ | 
|---|
| 271 | unsigned long *zeromap;		/* kvmalloc'ed bitmap to track zero pages */ | 
|---|
| 272 | struct swap_cluster_info *cluster_info; /* cluster info. Only for SSD */ | 
|---|
| 273 | struct list_head free_clusters; /* free clusters list */ | 
|---|
| 274 | struct list_head full_clusters; /* full clusters list */ | 
|---|
| 275 | struct list_head nonfull_clusters[SWAP_NR_ORDERS]; | 
|---|
| 276 | /* list of cluster that contains at least one free slot */ | 
|---|
| 277 | struct list_head frag_clusters[SWAP_NR_ORDERS]; | 
|---|
| 278 | /* list of cluster that are fragmented or contented */ | 
|---|
| 279 | unsigned int pages;		/* total of usable pages of swap */ | 
|---|
| 280 | atomic_long_t inuse_pages;	/* number of those currently in use */ | 
|---|
| 281 | struct swap_sequential_cluster *global_cluster; /* Use one global cluster for rotating device */ | 
|---|
| 282 | spinlock_t global_cluster_lock;	/* Serialize usage of global cluster */ | 
|---|
| 283 | struct rb_root swap_extent_root;/* root of the swap extent rbtree */ | 
|---|
| 284 | struct block_device *bdev;	/* swap device or bdev of swap file */ | 
|---|
| 285 | struct file *swap_file;		/* seldom referenced */ | 
|---|
| 286 | struct completion comp;		/* seldom referenced */ | 
|---|
| 287 | spinlock_t lock;		/* | 
|---|
| 288 | * protect map scan related fields like | 
|---|
| 289 | * swap_map, inuse_pages and all cluster | 
|---|
| 290 | * lists. other fields are only changed | 
|---|
| 291 | * at swapon/swapoff, so are protected | 
|---|
| 292 | * by swap_lock. changing flags need | 
|---|
| 293 | * hold this lock and swap_lock. If | 
|---|
| 294 | * both locks need hold, hold swap_lock | 
|---|
| 295 | * first. | 
|---|
| 296 | */ | 
|---|
| 297 | spinlock_t cont_lock;		/* | 
|---|
| 298 | * protect swap count continuation page | 
|---|
| 299 | * list. | 
|---|
| 300 | */ | 
|---|
| 301 | struct work_struct discard_work; /* discard worker */ | 
|---|
| 302 | struct work_struct reclaim_work; /* reclaim worker */ | 
|---|
| 303 | struct list_head discard_clusters; /* discard clusters list */ | 
|---|
| 304 | struct plist_node avail_lists[]; /* | 
|---|
| 305 | * entries in swap_avail_heads, one | 
|---|
| 306 | * entry per node. | 
|---|
| 307 | * Must be last as the number of the | 
|---|
| 308 | * array is nr_node_ids, which is not | 
|---|
| 309 | * a fixed value so have to allocate | 
|---|
| 310 | * dynamically. | 
|---|
| 311 | * And it has to be an array so that | 
|---|
| 312 | * plist_for_each_* can work. | 
|---|
| 313 | */ | 
|---|
| 314 | }; | 
|---|
| 315 |  | 
|---|
| 316 | static inline swp_entry_t page_swap_entry(struct page *page) | 
|---|
| 317 | { | 
|---|
| 318 | struct folio *folio = page_folio(page); | 
|---|
| 319 | swp_entry_t entry = folio->swap; | 
|---|
| 320 |  | 
|---|
| 321 | entry.val += folio_page_idx(folio, page); | 
|---|
| 322 | return entry; | 
|---|
| 323 | } | 
|---|
| 324 |  | 
|---|
| 325 | /* linux/mm/workingset.c */ | 
|---|
| 326 | bool workingset_test_recent(void *shadow, bool file, bool *workingset, | 
|---|
| 327 | bool flush); | 
|---|
| 328 | void workingset_age_nonresident(struct lruvec *lruvec, unsigned long nr_pages); | 
|---|
| 329 | void *workingset_eviction(struct folio *folio, struct mem_cgroup *target_memcg); | 
|---|
| 330 | void workingset_refault(struct folio *folio, void *shadow); | 
|---|
| 331 | void workingset_activation(struct folio *folio); | 
|---|
| 332 |  | 
|---|
| 333 | /* linux/mm/page_alloc.c */ | 
|---|
| 334 | extern unsigned long totalreserve_pages; | 
|---|
| 335 |  | 
|---|
| 336 | /* Definition of global_zone_page_state not available yet */ | 
|---|
| 337 | #define nr_free_pages() global_zone_page_state(NR_FREE_PAGES) | 
|---|
| 338 |  | 
|---|
| 339 |  | 
|---|
| 340 | /* linux/mm/swap.c */ | 
|---|
| 341 | void lru_note_cost_unlock_irq(struct lruvec *lruvec, bool file, | 
|---|
| 342 | unsigned int nr_io, unsigned int nr_rotated) | 
|---|
| 343 | __releases(lruvec->lru_lock); | 
|---|
| 344 | void lru_note_cost_refault(struct folio *); | 
|---|
| 345 | void folio_add_lru(struct folio *); | 
|---|
| 346 | void folio_add_lru_vma(struct folio *, struct vm_area_struct *); | 
|---|
| 347 | void mark_page_accessed(struct page *); | 
|---|
| 348 | void folio_mark_accessed(struct folio *); | 
|---|
| 349 |  | 
|---|
| 350 | static inline bool folio_may_be_lru_cached(struct folio *folio) | 
|---|
| 351 | { | 
|---|
| 352 | /* | 
|---|
| 353 | * Holding PMD-sized folios in per-CPU LRU cache unbalances accounting. | 
|---|
| 354 | * Holding small numbers of low-order mTHP folios in per-CPU LRU cache | 
|---|
| 355 | * will be sensible, but nobody has implemented and tested that yet. | 
|---|
| 356 | */ | 
|---|
| 357 | return !folio_test_large(folio); | 
|---|
| 358 | } | 
|---|
| 359 |  | 
|---|
| 360 | extern atomic_t lru_disable_count; | 
|---|
| 361 |  | 
|---|
| 362 | static inline bool lru_cache_disabled(void) | 
|---|
| 363 | { | 
|---|
| 364 | return atomic_read(v: &lru_disable_count); | 
|---|
| 365 | } | 
|---|
| 366 |  | 
|---|
| 367 | static inline void lru_cache_enable(void) | 
|---|
| 368 | { | 
|---|
| 369 | atomic_dec(v: &lru_disable_count); | 
|---|
| 370 | } | 
|---|
| 371 |  | 
|---|
| 372 | extern void lru_cache_disable(void); | 
|---|
| 373 | extern void lru_add_drain(void); | 
|---|
| 374 | extern void lru_add_drain_cpu(int cpu); | 
|---|
| 375 | extern void lru_add_drain_cpu_zone(struct zone *zone); | 
|---|
| 376 | extern void lru_add_drain_all(void); | 
|---|
| 377 | void folio_deactivate(struct folio *folio); | 
|---|
| 378 | void folio_mark_lazyfree(struct folio *folio); | 
|---|
| 379 | extern void swap_setup(void); | 
|---|
| 380 |  | 
|---|
| 381 | /* linux/mm/vmscan.c */ | 
|---|
| 382 | extern unsigned long zone_reclaimable_pages(struct zone *zone); | 
|---|
| 383 | extern unsigned long try_to_free_pages(struct zonelist *zonelist, int order, | 
|---|
| 384 | gfp_t gfp_mask, nodemask_t *mask); | 
|---|
| 385 |  | 
|---|
| 386 | #define MEMCG_RECLAIM_MAY_SWAP (1 << 1) | 
|---|
| 387 | #define MEMCG_RECLAIM_PROACTIVE (1 << 2) | 
|---|
| 388 | #define MIN_SWAPPINESS 0 | 
|---|
| 389 | #define MAX_SWAPPINESS 200 | 
|---|
| 390 |  | 
|---|
| 391 | /* Just reclaim from anon folios in proactive memory reclaim */ | 
|---|
| 392 | #define SWAPPINESS_ANON_ONLY (MAX_SWAPPINESS + 1) | 
|---|
| 393 |  | 
|---|
| 394 | extern unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *memcg, | 
|---|
| 395 | unsigned long nr_pages, | 
|---|
| 396 | gfp_t gfp_mask, | 
|---|
| 397 | unsigned int reclaim_options, | 
|---|
| 398 | int *swappiness); | 
|---|
| 399 | extern unsigned long mem_cgroup_shrink_node(struct mem_cgroup *mem, | 
|---|
| 400 | gfp_t gfp_mask, bool noswap, | 
|---|
| 401 | pg_data_t *pgdat, | 
|---|
| 402 | unsigned long *nr_scanned); | 
|---|
| 403 | extern unsigned long shrink_all_memory(unsigned long nr_pages); | 
|---|
| 404 | extern int vm_swappiness; | 
|---|
| 405 | long remove_mapping(struct address_space *mapping, struct folio *folio); | 
|---|
| 406 |  | 
|---|
| 407 | #if defined(CONFIG_SYSFS) && defined(CONFIG_NUMA) | 
|---|
| 408 | extern int reclaim_register_node(struct node *node); | 
|---|
| 409 | extern void reclaim_unregister_node(struct node *node); | 
|---|
| 410 |  | 
|---|
| 411 | #else | 
|---|
| 412 |  | 
|---|
| 413 | static inline int reclaim_register_node(struct node *node) | 
|---|
| 414 | { | 
|---|
| 415 | return 0; | 
|---|
| 416 | } | 
|---|
| 417 |  | 
|---|
| 418 | static inline void reclaim_unregister_node(struct node *node) | 
|---|
| 419 | { | 
|---|
| 420 | } | 
|---|
| 421 | #endif /* CONFIG_SYSFS && CONFIG_NUMA */ | 
|---|
| 422 |  | 
|---|
| 423 | #ifdef CONFIG_NUMA | 
|---|
| 424 | extern int sysctl_min_unmapped_ratio; | 
|---|
| 425 | extern int sysctl_min_slab_ratio; | 
|---|
| 426 | #endif | 
|---|
| 427 |  | 
|---|
| 428 | void check_move_unevictable_folios(struct folio_batch *fbatch); | 
|---|
| 429 |  | 
|---|
| 430 | extern void __meminit kswapd_run(int nid); | 
|---|
| 431 | extern void __meminit kswapd_stop(int nid); | 
|---|
| 432 |  | 
|---|
| 433 | #ifdef CONFIG_SWAP | 
|---|
| 434 |  | 
|---|
| 435 | int add_swap_extent(struct swap_info_struct *sis, unsigned long start_page, | 
|---|
| 436 | unsigned long nr_pages, sector_t start_block); | 
|---|
| 437 | int generic_swapfile_activate(struct swap_info_struct *, struct file *, | 
|---|
| 438 | sector_t *); | 
|---|
| 439 |  | 
|---|
| 440 | static inline unsigned long total_swapcache_pages(void) | 
|---|
| 441 | { | 
|---|
| 442 | return global_node_page_state(item: NR_SWAPCACHE); | 
|---|
| 443 | } | 
|---|
| 444 |  | 
|---|
| 445 | void free_swap_cache(struct folio *folio); | 
|---|
| 446 | void free_folio_and_swap_cache(struct folio *folio); | 
|---|
| 447 | void free_pages_and_swap_cache(struct encoded_page **, int); | 
|---|
| 448 | /* linux/mm/swapfile.c */ | 
|---|
| 449 | extern atomic_long_t nr_swap_pages; | 
|---|
| 450 | extern long total_swap_pages; | 
|---|
| 451 | extern atomic_t nr_rotate_swap; | 
|---|
| 452 |  | 
|---|
| 453 | /* Swap 50% full? Release swapcache more aggressively.. */ | 
|---|
| 454 | static inline bool vm_swap_full(void) | 
|---|
| 455 | { | 
|---|
| 456 | return atomic_long_read(v: &nr_swap_pages) * 2 < total_swap_pages; | 
|---|
| 457 | } | 
|---|
| 458 |  | 
|---|
| 459 | static inline long get_nr_swap_pages(void) | 
|---|
| 460 | { | 
|---|
| 461 | return atomic_long_read(v: &nr_swap_pages); | 
|---|
| 462 | } | 
|---|
| 463 |  | 
|---|
| 464 | extern void si_swapinfo(struct sysinfo *); | 
|---|
| 465 | int folio_alloc_swap(struct folio *folio, gfp_t gfp_mask); | 
|---|
| 466 | bool folio_free_swap(struct folio *folio); | 
|---|
| 467 | void put_swap_folio(struct folio *folio, swp_entry_t entry); | 
|---|
| 468 | extern swp_entry_t get_swap_page_of_type(int); | 
|---|
| 469 | extern int add_swap_count_continuation(swp_entry_t, gfp_t); | 
|---|
| 470 | extern void swap_shmem_alloc(swp_entry_t, int); | 
|---|
| 471 | extern int swap_duplicate(swp_entry_t); | 
|---|
| 472 | extern int swapcache_prepare(swp_entry_t entry, int nr); | 
|---|
| 473 | extern void swap_free_nr(swp_entry_t entry, int nr_pages); | 
|---|
| 474 | extern void free_swap_and_cache_nr(swp_entry_t entry, int nr); | 
|---|
| 475 | int swap_type_of(dev_t device, sector_t offset); | 
|---|
| 476 | int find_first_swap(dev_t *device); | 
|---|
| 477 | extern unsigned int count_swap_pages(int, int); | 
|---|
| 478 | extern sector_t swapdev_block(int, pgoff_t); | 
|---|
| 479 | extern int __swap_count(swp_entry_t entry); | 
|---|
| 480 | extern bool swap_entry_swapped(struct swap_info_struct *si, swp_entry_t entry); | 
|---|
| 481 | extern int swp_swapcount(swp_entry_t entry); | 
|---|
| 482 | struct backing_dev_info; | 
|---|
| 483 | extern struct swap_info_struct *get_swap_device(swp_entry_t entry); | 
|---|
| 484 | sector_t swap_folio_sector(struct folio *folio); | 
|---|
| 485 |  | 
|---|
| 486 | static inline void put_swap_device(struct swap_info_struct *si) | 
|---|
| 487 | { | 
|---|
| 488 | percpu_ref_put(ref: &si->users); | 
|---|
| 489 | } | 
|---|
| 490 |  | 
|---|
| 491 | #else /* CONFIG_SWAP */ | 
|---|
| 492 | static inline struct swap_info_struct *get_swap_device(swp_entry_t entry) | 
|---|
| 493 | { | 
|---|
| 494 | return NULL; | 
|---|
| 495 | } | 
|---|
| 496 |  | 
|---|
| 497 | static inline void put_swap_device(struct swap_info_struct *si) | 
|---|
| 498 | { | 
|---|
| 499 | } | 
|---|
| 500 |  | 
|---|
| 501 | #define get_nr_swap_pages()			0L | 
|---|
| 502 | #define total_swap_pages			0L | 
|---|
| 503 | #define total_swapcache_pages()			0UL | 
|---|
| 504 | #define vm_swap_full()				0 | 
|---|
| 505 |  | 
|---|
| 506 | #define si_swapinfo(val) \ | 
|---|
| 507 | do { (val)->freeswap = (val)->totalswap = 0; } while (0) | 
|---|
| 508 | #define free_folio_and_swap_cache(folio) \ | 
|---|
| 509 | folio_put(folio) | 
|---|
| 510 | #define free_pages_and_swap_cache(pages, nr) \ | 
|---|
| 511 | release_pages((pages), (nr)); | 
|---|
| 512 |  | 
|---|
| 513 | static inline void free_swap_and_cache_nr(swp_entry_t entry, int nr) | 
|---|
| 514 | { | 
|---|
| 515 | } | 
|---|
| 516 |  | 
|---|
| 517 | static inline void free_swap_cache(struct folio *folio) | 
|---|
| 518 | { | 
|---|
| 519 | } | 
|---|
| 520 |  | 
|---|
| 521 | static inline int add_swap_count_continuation(swp_entry_t swp, gfp_t gfp_mask) | 
|---|
| 522 | { | 
|---|
| 523 | return 0; | 
|---|
| 524 | } | 
|---|
| 525 |  | 
|---|
| 526 | static inline void swap_shmem_alloc(swp_entry_t swp, int nr) | 
|---|
| 527 | { | 
|---|
| 528 | } | 
|---|
| 529 |  | 
|---|
| 530 | static inline int swap_duplicate(swp_entry_t swp) | 
|---|
| 531 | { | 
|---|
| 532 | return 0; | 
|---|
| 533 | } | 
|---|
| 534 |  | 
|---|
| 535 | static inline int swapcache_prepare(swp_entry_t swp, int nr) | 
|---|
| 536 | { | 
|---|
| 537 | return 0; | 
|---|
| 538 | } | 
|---|
| 539 |  | 
|---|
| 540 | static inline void swap_free_nr(swp_entry_t entry, int nr_pages) | 
|---|
| 541 | { | 
|---|
| 542 | } | 
|---|
| 543 |  | 
|---|
| 544 | static inline void put_swap_folio(struct folio *folio, swp_entry_t swp) | 
|---|
| 545 | { | 
|---|
| 546 | } | 
|---|
| 547 |  | 
|---|
| 548 | static inline int __swap_count(swp_entry_t entry) | 
|---|
| 549 | { | 
|---|
| 550 | return 0; | 
|---|
| 551 | } | 
|---|
| 552 |  | 
|---|
| 553 | static inline bool swap_entry_swapped(struct swap_info_struct *si, swp_entry_t entry) | 
|---|
| 554 | { | 
|---|
| 555 | return false; | 
|---|
| 556 | } | 
|---|
| 557 |  | 
|---|
| 558 | static inline int swp_swapcount(swp_entry_t entry) | 
|---|
| 559 | { | 
|---|
| 560 | return 0; | 
|---|
| 561 | } | 
|---|
| 562 |  | 
|---|
| 563 | static inline int folio_alloc_swap(struct folio *folio, gfp_t gfp_mask) | 
|---|
| 564 | { | 
|---|
| 565 | return -EINVAL; | 
|---|
| 566 | } | 
|---|
| 567 |  | 
|---|
| 568 | static inline bool folio_free_swap(struct folio *folio) | 
|---|
| 569 | { | 
|---|
| 570 | return false; | 
|---|
| 571 | } | 
|---|
| 572 |  | 
|---|
| 573 | static inline int add_swap_extent(struct swap_info_struct *sis, | 
|---|
| 574 | unsigned long start_page, | 
|---|
| 575 | unsigned long nr_pages, sector_t start_block) | 
|---|
| 576 | { | 
|---|
| 577 | return -EINVAL; | 
|---|
| 578 | } | 
|---|
| 579 | #endif /* CONFIG_SWAP */ | 
|---|
| 580 |  | 
|---|
| 581 | static inline void free_swap_and_cache(swp_entry_t entry) | 
|---|
| 582 | { | 
|---|
| 583 | free_swap_and_cache_nr(entry, nr: 1); | 
|---|
| 584 | } | 
|---|
| 585 |  | 
|---|
| 586 | static inline void swap_free(swp_entry_t entry) | 
|---|
| 587 | { | 
|---|
| 588 | swap_free_nr(entry, nr_pages: 1); | 
|---|
| 589 | } | 
|---|
| 590 |  | 
|---|
| 591 | #ifdef CONFIG_MEMCG | 
|---|
| 592 | static inline int mem_cgroup_swappiness(struct mem_cgroup *memcg) | 
|---|
| 593 | { | 
|---|
| 594 | /* Cgroup2 doesn't have per-cgroup swappiness */ | 
|---|
| 595 | if (cgroup_subsys_on_dfl(memory_cgrp_subsys)) | 
|---|
| 596 | return READ_ONCE(vm_swappiness); | 
|---|
| 597 |  | 
|---|
| 598 | /* root ? */ | 
|---|
| 599 | if (mem_cgroup_disabled() || mem_cgroup_is_root(memcg)) | 
|---|
| 600 | return READ_ONCE(vm_swappiness); | 
|---|
| 601 |  | 
|---|
| 602 | return READ_ONCE(memcg->swappiness); | 
|---|
| 603 | } | 
|---|
| 604 | #else | 
|---|
| 605 | static inline int mem_cgroup_swappiness(struct mem_cgroup *mem) | 
|---|
| 606 | { | 
|---|
| 607 | return READ_ONCE(vm_swappiness); | 
|---|
| 608 | } | 
|---|
| 609 | #endif | 
|---|
| 610 |  | 
|---|
| 611 | #if defined(CONFIG_SWAP) && defined(CONFIG_MEMCG) && defined(CONFIG_BLK_CGROUP) | 
|---|
| 612 | void __folio_throttle_swaprate(struct folio *folio, gfp_t gfp); | 
|---|
| 613 | static inline void folio_throttle_swaprate(struct folio *folio, gfp_t gfp) | 
|---|
| 614 | { | 
|---|
| 615 | if (mem_cgroup_disabled()) | 
|---|
| 616 | return; | 
|---|
| 617 | __folio_throttle_swaprate(folio, gfp); | 
|---|
| 618 | } | 
|---|
| 619 | #else | 
|---|
| 620 | static inline void folio_throttle_swaprate(struct folio *folio, gfp_t gfp) | 
|---|
| 621 | { | 
|---|
| 622 | } | 
|---|
| 623 | #endif | 
|---|
| 624 |  | 
|---|
| 625 | #if defined(CONFIG_MEMCG) && defined(CONFIG_SWAP) | 
|---|
| 626 | int __mem_cgroup_try_charge_swap(struct folio *folio, swp_entry_t entry); | 
|---|
| 627 | static inline int mem_cgroup_try_charge_swap(struct folio *folio, | 
|---|
| 628 | swp_entry_t entry) | 
|---|
| 629 | { | 
|---|
| 630 | if (mem_cgroup_disabled()) | 
|---|
| 631 | return 0; | 
|---|
| 632 | return __mem_cgroup_try_charge_swap(folio, entry); | 
|---|
| 633 | } | 
|---|
| 634 |  | 
|---|
| 635 | extern void __mem_cgroup_uncharge_swap(swp_entry_t entry, unsigned int nr_pages); | 
|---|
| 636 | static inline void mem_cgroup_uncharge_swap(swp_entry_t entry, unsigned int nr_pages) | 
|---|
| 637 | { | 
|---|
| 638 | if (mem_cgroup_disabled()) | 
|---|
| 639 | return; | 
|---|
| 640 | __mem_cgroup_uncharge_swap(entry, nr_pages); | 
|---|
| 641 | } | 
|---|
| 642 |  | 
|---|
| 643 | extern long mem_cgroup_get_nr_swap_pages(struct mem_cgroup *memcg); | 
|---|
| 644 | extern bool mem_cgroup_swap_full(struct folio *folio); | 
|---|
| 645 | #else | 
|---|
| 646 | static inline int mem_cgroup_try_charge_swap(struct folio *folio, | 
|---|
| 647 | swp_entry_t entry) | 
|---|
| 648 | { | 
|---|
| 649 | return 0; | 
|---|
| 650 | } | 
|---|
| 651 |  | 
|---|
| 652 | static inline void mem_cgroup_uncharge_swap(swp_entry_t entry, | 
|---|
| 653 | unsigned int nr_pages) | 
|---|
| 654 | { | 
|---|
| 655 | } | 
|---|
| 656 |  | 
|---|
| 657 | static inline long mem_cgroup_get_nr_swap_pages(struct mem_cgroup *memcg) | 
|---|
| 658 | { | 
|---|
| 659 | return get_nr_swap_pages(); | 
|---|
| 660 | } | 
|---|
| 661 |  | 
|---|
| 662 | static inline bool mem_cgroup_swap_full(struct folio *folio) | 
|---|
| 663 | { | 
|---|
| 664 | return vm_swap_full(); | 
|---|
| 665 | } | 
|---|
| 666 | #endif | 
|---|
| 667 |  | 
|---|
| 668 | #endif /* __KERNEL__*/ | 
|---|
| 669 | #endif /* _LINUX_SWAP_H */ | 
|---|
| 670 |  | 
|---|