at v6.5 22 kB view raw
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 19struct notifier_block; 20 21struct bio; 22 23struct pagevec; 24 25#define SWAP_FLAG_PREFER 0x8000 /* set if swap priority specified */ 26#define SWAP_FLAG_PRIO_MASK 0x7fff 27#define SWAP_FLAG_PRIO_SHIFT 0 28#define SWAP_FLAG_DISCARD 0x10000 /* enable discard for swap */ 29#define SWAP_FLAG_DISCARD_ONCE 0x20000 /* discard swap area at swapon-time */ 30#define SWAP_FLAG_DISCARD_PAGES 0x40000 /* discard page-clusters after use */ 31 32#define SWAP_FLAGS_VALID (SWAP_FLAG_PRIO_MASK | SWAP_FLAG_PREFER | \ 33 SWAP_FLAG_DISCARD | SWAP_FLAG_DISCARD_ONCE | \ 34 SWAP_FLAG_DISCARD_PAGES) 35#define SWAP_BATCH 64 36 37static inline int current_is_kswapd(void) 38{ 39 return current->flags & PF_KSWAPD; 40} 41 42/* 43 * MAX_SWAPFILES defines the maximum number of swaptypes: things which can 44 * be swapped to. The swap type and the offset into that swap type are 45 * encoded into pte's and into pgoff_t's in the swapcache. Using five bits 46 * for the type means that the maximum number of swapcache pages is 27 bits 47 * on 32-bit-pgoff_t architectures. And that assumes that the architecture packs 48 * the type/offset into the pte as 5/27 as well. 49 */ 50#define MAX_SWAPFILES_SHIFT 5 51 52/* 53 * Use some of the swap files numbers for other purposes. This 54 * is a convenient way to hook into the VM to trigger special 55 * actions on faults. 56 */ 57 58/* 59 * PTE markers are used to persist information onto PTEs that otherwise 60 * should be a none pte. As its name "PTE" hints, it should only be 61 * applied to the leaves of pgtables. 62 */ 63#define SWP_PTE_MARKER_NUM 1 64#define SWP_PTE_MARKER (MAX_SWAPFILES + SWP_HWPOISON_NUM + \ 65 SWP_MIGRATION_NUM + SWP_DEVICE_NUM) 66 67/* 68 * Unaddressable device memory support. See include/linux/hmm.h and 69 * Documentation/mm/hmm.rst. Short description is we need struct pages for 70 * device memory that is unaddressable (inaccessible) by CPU, so that we can 71 * migrate part of a process memory to device memory. 72 * 73 * When a page is migrated from CPU to device, we set the CPU page table entry 74 * to a special SWP_DEVICE_{READ|WRITE} entry. 75 * 76 * When a page is mapped by the device for exclusive access we set the CPU page 77 * table entries to special SWP_DEVICE_EXCLUSIVE_* entries. 78 */ 79#ifdef CONFIG_DEVICE_PRIVATE 80#define SWP_DEVICE_NUM 4 81#define SWP_DEVICE_WRITE (MAX_SWAPFILES+SWP_HWPOISON_NUM+SWP_MIGRATION_NUM) 82#define SWP_DEVICE_READ (MAX_SWAPFILES+SWP_HWPOISON_NUM+SWP_MIGRATION_NUM+1) 83#define SWP_DEVICE_EXCLUSIVE_WRITE (MAX_SWAPFILES+SWP_HWPOISON_NUM+SWP_MIGRATION_NUM+2) 84#define SWP_DEVICE_EXCLUSIVE_READ (MAX_SWAPFILES+SWP_HWPOISON_NUM+SWP_MIGRATION_NUM+3) 85#else 86#define SWP_DEVICE_NUM 0 87#endif 88 89/* 90 * Page migration support. 91 * 92 * SWP_MIGRATION_READ_EXCLUSIVE is only applicable to anonymous pages and 93 * indicates that the referenced (part of) an anonymous page is exclusive to 94 * a single process. For SWP_MIGRATION_WRITE, that information is implicit: 95 * (part of) an anonymous page that are mapped writable are exclusive to a 96 * single process. 97 */ 98#ifdef CONFIG_MIGRATION 99#define SWP_MIGRATION_NUM 3 100#define SWP_MIGRATION_READ (MAX_SWAPFILES + SWP_HWPOISON_NUM) 101#define SWP_MIGRATION_READ_EXCLUSIVE (MAX_SWAPFILES + SWP_HWPOISON_NUM + 1) 102#define SWP_MIGRATION_WRITE (MAX_SWAPFILES + SWP_HWPOISON_NUM + 2) 103#else 104#define SWP_MIGRATION_NUM 0 105#endif 106 107/* 108 * Handling of hardware poisoned pages with memory corruption. 109 */ 110#ifdef CONFIG_MEMORY_FAILURE 111#define SWP_HWPOISON_NUM 1 112#define SWP_HWPOISON MAX_SWAPFILES 113#else 114#define SWP_HWPOISON_NUM 0 115#endif 116 117#define MAX_SWAPFILES \ 118 ((1 << MAX_SWAPFILES_SHIFT) - SWP_DEVICE_NUM - \ 119 SWP_MIGRATION_NUM - SWP_HWPOISON_NUM - \ 120 SWP_PTE_MARKER_NUM) 121 122/* 123 * Magic header for a swap area. The first part of the union is 124 * what the swap magic looks like for the old (limited to 128MB) 125 * swap area format, the second part of the union adds - in the 126 * old reserved area - some extra information. Note that the first 127 * kilobyte is reserved for boot loader or disk label stuff... 128 * 129 * Having the magic at the end of the PAGE_SIZE makes detecting swap 130 * areas somewhat tricky on machines that support multiple page sizes. 131 * For 2.5 we'll probably want to move the magic to just beyond the 132 * bootbits... 133 */ 134union swap_header { 135 struct { 136 char reserved[PAGE_SIZE - 10]; 137 char magic[10]; /* SWAP-SPACE or SWAPSPACE2 */ 138 } magic; 139 struct { 140 char bootbits[1024]; /* Space for disklabel etc. */ 141 __u32 version; 142 __u32 last_page; 143 __u32 nr_badpages; 144 unsigned char sws_uuid[16]; 145 unsigned char sws_volume[16]; 146 __u32 padding[117]; 147 __u32 badpages[1]; 148 } info; 149}; 150 151/* 152 * current->reclaim_state points to one of these when a task is running 153 * memory reclaim 154 */ 155struct reclaim_state { 156 /* pages reclaimed outside of LRU-based reclaim */ 157 unsigned long reclaimed; 158#ifdef CONFIG_LRU_GEN 159 /* per-thread mm walk data */ 160 struct lru_gen_mm_walk *mm_walk; 161#endif 162}; 163 164/* 165 * mm_account_reclaimed_pages(): account reclaimed pages outside of LRU-based 166 * reclaim 167 * @pages: number of pages reclaimed 168 * 169 * If the current process is undergoing a reclaim operation, increment the 170 * number of reclaimed pages by @pages. 171 */ 172static inline void mm_account_reclaimed_pages(unsigned long pages) 173{ 174 if (current->reclaim_state) 175 current->reclaim_state->reclaimed += pages; 176} 177 178#ifdef __KERNEL__ 179 180struct address_space; 181struct sysinfo; 182struct writeback_control; 183struct zone; 184 185/* 186 * A swap extent maps a range of a swapfile's PAGE_SIZE pages onto a range of 187 * disk blocks. A rbtree of swap extents maps the entire swapfile (Where the 188 * term `swapfile' refers to either a blockdevice or an IS_REG file). Apart 189 * from setup, they're handled identically. 190 * 191 * We always assume that blocks are of size PAGE_SIZE. 192 */ 193struct swap_extent { 194 struct rb_node rb_node; 195 pgoff_t start_page; 196 pgoff_t nr_pages; 197 sector_t start_block; 198}; 199 200/* 201 * Max bad pages in the new format.. 202 */ 203#define MAX_SWAP_BADPAGES \ 204 ((offsetof(union swap_header, magic.magic) - \ 205 offsetof(union swap_header, info.badpages)) / sizeof(int)) 206 207enum { 208 SWP_USED = (1 << 0), /* is slot in swap_info[] used? */ 209 SWP_WRITEOK = (1 << 1), /* ok to write to this swap? */ 210 SWP_DISCARDABLE = (1 << 2), /* blkdev support discard */ 211 SWP_DISCARDING = (1 << 3), /* now discarding a free cluster */ 212 SWP_SOLIDSTATE = (1 << 4), /* blkdev seeks are cheap */ 213 SWP_CONTINUED = (1 << 5), /* swap_map has count continuation */ 214 SWP_BLKDEV = (1 << 6), /* its a block device */ 215 SWP_ACTIVATED = (1 << 7), /* set after swap_activate success */ 216 SWP_FS_OPS = (1 << 8), /* swapfile operations go through fs */ 217 SWP_AREA_DISCARD = (1 << 9), /* single-time swap area discards */ 218 SWP_PAGE_DISCARD = (1 << 10), /* freed swap page-cluster discards */ 219 SWP_STABLE_WRITES = (1 << 11), /* no overwrite PG_writeback pages */ 220 SWP_SYNCHRONOUS_IO = (1 << 12), /* synchronous IO is efficient */ 221 /* add others here before... */ 222 SWP_SCANNING = (1 << 14), /* refcount in scan_swap_map */ 223}; 224 225#define SWAP_CLUSTER_MAX 32UL 226#define COMPACT_CLUSTER_MAX SWAP_CLUSTER_MAX 227 228/* Bit flag in swap_map */ 229#define SWAP_HAS_CACHE 0x40 /* Flag page is cached, in first swap_map */ 230#define COUNT_CONTINUED 0x80 /* Flag swap_map continuation for full count */ 231 232/* Special value in first swap_map */ 233#define SWAP_MAP_MAX 0x3e /* Max count */ 234#define SWAP_MAP_BAD 0x3f /* Note page is bad */ 235#define SWAP_MAP_SHMEM 0xbf /* Owned by shmem/tmpfs */ 236 237/* Special value in each swap_map continuation */ 238#define SWAP_CONT_MAX 0x7f /* Max count */ 239 240/* 241 * We use this to track usage of a cluster. A cluster is a block of swap disk 242 * space with SWAPFILE_CLUSTER pages long and naturally aligns in disk. All 243 * free clusters are organized into a list. We fetch an entry from the list to 244 * get a free cluster. 245 * 246 * The data field stores next cluster if the cluster is free or cluster usage 247 * counter otherwise. The flags field determines if a cluster is free. This is 248 * protected by swap_info_struct.lock. 249 */ 250struct swap_cluster_info { 251 spinlock_t lock; /* 252 * Protect swap_cluster_info fields 253 * and swap_info_struct->swap_map 254 * elements correspond to the swap 255 * cluster 256 */ 257 unsigned int data:24; 258 unsigned int flags:8; 259}; 260#define CLUSTER_FLAG_FREE 1 /* This cluster is free */ 261#define CLUSTER_FLAG_NEXT_NULL 2 /* This cluster has no next cluster */ 262#define CLUSTER_FLAG_HUGE 4 /* This cluster is backing a transparent huge page */ 263 264/* 265 * We assign a cluster to each CPU, so each CPU can allocate swap entry from 266 * its own cluster and swapout sequentially. The purpose is to optimize swapout 267 * throughput. 268 */ 269struct percpu_cluster { 270 struct swap_cluster_info index; /* Current cluster index */ 271 unsigned int next; /* Likely next allocation offset */ 272}; 273 274struct swap_cluster_list { 275 struct swap_cluster_info head; 276 struct swap_cluster_info tail; 277}; 278 279/* 280 * The in-memory structure used to track swap areas. 281 */ 282struct swap_info_struct { 283 struct percpu_ref users; /* indicate and keep swap device valid. */ 284 unsigned long flags; /* SWP_USED etc: see above */ 285 signed short prio; /* swap priority of this type */ 286 struct plist_node list; /* entry in swap_active_head */ 287 signed char type; /* strange name for an index */ 288 unsigned int max; /* extent of the swap_map */ 289 unsigned char *swap_map; /* vmalloc'ed array of usage counts */ 290 struct swap_cluster_info *cluster_info; /* cluster info. Only for SSD */ 291 struct swap_cluster_list free_clusters; /* free clusters list */ 292 unsigned int lowest_bit; /* index of first free in swap_map */ 293 unsigned int highest_bit; /* index of last free in swap_map */ 294 unsigned int pages; /* total of usable pages of swap */ 295 unsigned int inuse_pages; /* number of those currently in use */ 296 unsigned int cluster_next; /* likely index for next allocation */ 297 unsigned int cluster_nr; /* countdown to next cluster search */ 298 unsigned int __percpu *cluster_next_cpu; /*percpu index for next allocation */ 299 struct percpu_cluster __percpu *percpu_cluster; /* per cpu's swap location */ 300 struct rb_root swap_extent_root;/* root of the swap extent rbtree */ 301 struct block_device *bdev; /* swap device or bdev of swap file */ 302 struct file *swap_file; /* seldom referenced */ 303 unsigned int old_block_size; /* seldom referenced */ 304 struct completion comp; /* seldom referenced */ 305#ifdef CONFIG_FRONTSWAP 306 unsigned long *frontswap_map; /* frontswap in-use, one bit per page */ 307 atomic_t frontswap_pages; /* frontswap pages in-use counter */ 308#endif 309 spinlock_t lock; /* 310 * protect map scan related fields like 311 * swap_map, lowest_bit, highest_bit, 312 * inuse_pages, cluster_next, 313 * cluster_nr, lowest_alloc, 314 * highest_alloc, free/discard cluster 315 * list. other fields are only changed 316 * at swapon/swapoff, so are protected 317 * by swap_lock. changing flags need 318 * hold this lock and swap_lock. If 319 * both locks need hold, hold swap_lock 320 * first. 321 */ 322 spinlock_t cont_lock; /* 323 * protect swap count continuation page 324 * list. 325 */ 326 struct work_struct discard_work; /* discard worker */ 327 struct swap_cluster_list discard_clusters; /* discard clusters list */ 328 struct plist_node avail_lists[]; /* 329 * entries in swap_avail_heads, one 330 * entry per node. 331 * Must be last as the number of the 332 * array is nr_node_ids, which is not 333 * a fixed value so have to allocate 334 * dynamically. 335 * And it has to be an array so that 336 * plist_for_each_* can work. 337 */ 338}; 339 340static inline swp_entry_t folio_swap_entry(struct folio *folio) 341{ 342 swp_entry_t entry = { .val = page_private(&folio->page) }; 343 return entry; 344} 345 346static inline void folio_set_swap_entry(struct folio *folio, swp_entry_t entry) 347{ 348 folio->private = (void *)entry.val; 349} 350 351/* linux/mm/workingset.c */ 352bool workingset_test_recent(void *shadow, bool file, bool *workingset); 353void workingset_age_nonresident(struct lruvec *lruvec, unsigned long nr_pages); 354void *workingset_eviction(struct folio *folio, struct mem_cgroup *target_memcg); 355void workingset_refault(struct folio *folio, void *shadow); 356void workingset_activation(struct folio *folio); 357 358/* Only track the nodes of mappings with shadow entries */ 359void workingset_update_node(struct xa_node *node); 360extern struct list_lru shadow_nodes; 361#define mapping_set_update(xas, mapping) do { \ 362 if (!dax_mapping(mapping) && !shmem_mapping(mapping)) { \ 363 xas_set_update(xas, workingset_update_node); \ 364 xas_set_lru(xas, &shadow_nodes); \ 365 } \ 366} while (0) 367 368/* linux/mm/page_alloc.c */ 369extern unsigned long totalreserve_pages; 370 371/* Definition of global_zone_page_state not available yet */ 372#define nr_free_pages() global_zone_page_state(NR_FREE_PAGES) 373 374 375/* linux/mm/swap.c */ 376void lru_note_cost(struct lruvec *lruvec, bool file, 377 unsigned int nr_io, unsigned int nr_rotated); 378void lru_note_cost_refault(struct folio *); 379void folio_add_lru(struct folio *); 380void folio_add_lru_vma(struct folio *, struct vm_area_struct *); 381void mark_page_accessed(struct page *); 382void folio_mark_accessed(struct folio *); 383 384extern atomic_t lru_disable_count; 385 386static inline bool lru_cache_disabled(void) 387{ 388 return atomic_read(&lru_disable_count); 389} 390 391static inline void lru_cache_enable(void) 392{ 393 atomic_dec(&lru_disable_count); 394} 395 396extern void lru_cache_disable(void); 397extern void lru_add_drain(void); 398extern void lru_add_drain_cpu(int cpu); 399extern void lru_add_drain_cpu_zone(struct zone *zone); 400extern void lru_add_drain_all(void); 401void folio_deactivate(struct folio *folio); 402void folio_mark_lazyfree(struct folio *folio); 403extern void swap_setup(void); 404 405extern void lru_cache_add_inactive_or_unevictable(struct page *page, 406 struct vm_area_struct *vma); 407 408/* linux/mm/vmscan.c */ 409extern unsigned long zone_reclaimable_pages(struct zone *zone); 410extern unsigned long try_to_free_pages(struct zonelist *zonelist, int order, 411 gfp_t gfp_mask, nodemask_t *mask); 412 413#define MEMCG_RECLAIM_MAY_SWAP (1 << 1) 414#define MEMCG_RECLAIM_PROACTIVE (1 << 2) 415extern unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *memcg, 416 unsigned long nr_pages, 417 gfp_t gfp_mask, 418 unsigned int reclaim_options); 419extern unsigned long mem_cgroup_shrink_node(struct mem_cgroup *mem, 420 gfp_t gfp_mask, bool noswap, 421 pg_data_t *pgdat, 422 unsigned long *nr_scanned); 423extern unsigned long shrink_all_memory(unsigned long nr_pages); 424extern int vm_swappiness; 425long remove_mapping(struct address_space *mapping, struct folio *folio); 426 427#ifdef CONFIG_NUMA 428extern int node_reclaim_mode; 429extern int sysctl_min_unmapped_ratio; 430extern int sysctl_min_slab_ratio; 431#else 432#define node_reclaim_mode 0 433#endif 434 435static inline bool node_reclaim_enabled(void) 436{ 437 /* Is any node_reclaim_mode bit set? */ 438 return node_reclaim_mode & (RECLAIM_ZONE|RECLAIM_WRITE|RECLAIM_UNMAP); 439} 440 441void check_move_unevictable_folios(struct folio_batch *fbatch); 442 443extern void __meminit kswapd_run(int nid); 444extern void __meminit kswapd_stop(int nid); 445 446#ifdef CONFIG_SWAP 447 448int add_swap_extent(struct swap_info_struct *sis, unsigned long start_page, 449 unsigned long nr_pages, sector_t start_block); 450int generic_swapfile_activate(struct swap_info_struct *, struct file *, 451 sector_t *); 452 453static inline unsigned long total_swapcache_pages(void) 454{ 455 return global_node_page_state(NR_SWAPCACHE); 456} 457 458extern void free_swap_cache(struct page *page); 459extern void free_page_and_swap_cache(struct page *); 460extern void free_pages_and_swap_cache(struct encoded_page **, int); 461/* linux/mm/swapfile.c */ 462extern atomic_long_t nr_swap_pages; 463extern long total_swap_pages; 464extern atomic_t nr_rotate_swap; 465extern bool has_usable_swap(void); 466 467/* Swap 50% full? Release swapcache more aggressively.. */ 468static inline bool vm_swap_full(void) 469{ 470 return atomic_long_read(&nr_swap_pages) * 2 < total_swap_pages; 471} 472 473static inline long get_nr_swap_pages(void) 474{ 475 return atomic_long_read(&nr_swap_pages); 476} 477 478extern void si_swapinfo(struct sysinfo *); 479swp_entry_t folio_alloc_swap(struct folio *folio); 480bool folio_free_swap(struct folio *folio); 481void put_swap_folio(struct folio *folio, swp_entry_t entry); 482extern swp_entry_t get_swap_page_of_type(int); 483extern int get_swap_pages(int n, swp_entry_t swp_entries[], int entry_size); 484extern int add_swap_count_continuation(swp_entry_t, gfp_t); 485extern void swap_shmem_alloc(swp_entry_t); 486extern int swap_duplicate(swp_entry_t); 487extern int swapcache_prepare(swp_entry_t); 488extern void swap_free(swp_entry_t); 489extern void swapcache_free_entries(swp_entry_t *entries, int n); 490extern int free_swap_and_cache(swp_entry_t); 491int swap_type_of(dev_t device, sector_t offset); 492int find_first_swap(dev_t *device); 493extern unsigned int count_swap_pages(int, int); 494extern sector_t swapdev_block(int, pgoff_t); 495extern int __swap_count(swp_entry_t entry); 496extern int swap_swapcount(struct swap_info_struct *si, swp_entry_t entry); 497extern int swp_swapcount(swp_entry_t entry); 498extern struct swap_info_struct *page_swap_info(struct page *); 499extern struct swap_info_struct *swp_swap_info(swp_entry_t entry); 500struct backing_dev_info; 501extern int init_swap_address_space(unsigned int type, unsigned long nr_pages); 502extern void exit_swap_address_space(unsigned int type); 503extern struct swap_info_struct *get_swap_device(swp_entry_t entry); 504sector_t swap_page_sector(struct page *page); 505 506static inline void put_swap_device(struct swap_info_struct *si) 507{ 508 percpu_ref_put(&si->users); 509} 510 511#else /* CONFIG_SWAP */ 512static inline struct swap_info_struct *swp_swap_info(swp_entry_t entry) 513{ 514 return NULL; 515} 516 517static inline struct swap_info_struct *get_swap_device(swp_entry_t entry) 518{ 519 return NULL; 520} 521 522static inline void put_swap_device(struct swap_info_struct *si) 523{ 524} 525 526#define get_nr_swap_pages() 0L 527#define total_swap_pages 0L 528#define total_swapcache_pages() 0UL 529#define vm_swap_full() 0 530 531#define si_swapinfo(val) \ 532 do { (val)->freeswap = (val)->totalswap = 0; } while (0) 533/* only sparc can not include linux/pagemap.h in this file 534 * so leave put_page and release_pages undeclared... */ 535#define free_page_and_swap_cache(page) \ 536 put_page(page) 537#define free_pages_and_swap_cache(pages, nr) \ 538 release_pages((pages), (nr)); 539 540/* used to sanity check ptes in zap_pte_range when CONFIG_SWAP=0 */ 541#define free_swap_and_cache(e) is_pfn_swap_entry(e) 542 543static inline void free_swap_cache(struct page *page) 544{ 545} 546 547static inline int add_swap_count_continuation(swp_entry_t swp, gfp_t gfp_mask) 548{ 549 return 0; 550} 551 552static inline void swap_shmem_alloc(swp_entry_t swp) 553{ 554} 555 556static inline int swap_duplicate(swp_entry_t swp) 557{ 558 return 0; 559} 560 561static inline void swap_free(swp_entry_t swp) 562{ 563} 564 565static inline void put_swap_folio(struct folio *folio, swp_entry_t swp) 566{ 567} 568 569static inline int __swap_count(swp_entry_t entry) 570{ 571 return 0; 572} 573 574static inline int swap_swapcount(struct swap_info_struct *si, swp_entry_t entry) 575{ 576 return 0; 577} 578 579static inline int swp_swapcount(swp_entry_t entry) 580{ 581 return 0; 582} 583 584static inline swp_entry_t folio_alloc_swap(struct folio *folio) 585{ 586 swp_entry_t entry; 587 entry.val = 0; 588 return entry; 589} 590 591static inline bool folio_free_swap(struct folio *folio) 592{ 593 return false; 594} 595 596static inline int add_swap_extent(struct swap_info_struct *sis, 597 unsigned long start_page, 598 unsigned long nr_pages, sector_t start_block) 599{ 600 return -EINVAL; 601} 602#endif /* CONFIG_SWAP */ 603 604#ifdef CONFIG_THP_SWAP 605extern int split_swap_cluster(swp_entry_t entry); 606#else 607static inline int split_swap_cluster(swp_entry_t entry) 608{ 609 return 0; 610} 611#endif 612 613#ifdef CONFIG_MEMCG 614static inline int mem_cgroup_swappiness(struct mem_cgroup *memcg) 615{ 616 /* Cgroup2 doesn't have per-cgroup swappiness */ 617 if (cgroup_subsys_on_dfl(memory_cgrp_subsys)) 618 return READ_ONCE(vm_swappiness); 619 620 /* root ? */ 621 if (mem_cgroup_disabled() || mem_cgroup_is_root(memcg)) 622 return READ_ONCE(vm_swappiness); 623 624 return READ_ONCE(memcg->swappiness); 625} 626#else 627static inline int mem_cgroup_swappiness(struct mem_cgroup *mem) 628{ 629 return READ_ONCE(vm_swappiness); 630} 631#endif 632 633#ifdef CONFIG_ZSWAP 634extern u64 zswap_pool_total_size; 635extern atomic_t zswap_stored_pages; 636#endif 637 638#if defined(CONFIG_SWAP) && defined(CONFIG_MEMCG) && defined(CONFIG_BLK_CGROUP) 639void __folio_throttle_swaprate(struct folio *folio, gfp_t gfp); 640static inline void folio_throttle_swaprate(struct folio *folio, gfp_t gfp) 641{ 642 if (mem_cgroup_disabled()) 643 return; 644 __folio_throttle_swaprate(folio, gfp); 645} 646#else 647static inline void folio_throttle_swaprate(struct folio *folio, gfp_t gfp) 648{ 649} 650#endif 651 652#if defined(CONFIG_MEMCG) && defined(CONFIG_SWAP) 653void mem_cgroup_swapout(struct folio *folio, swp_entry_t entry); 654int __mem_cgroup_try_charge_swap(struct folio *folio, swp_entry_t entry); 655static inline int mem_cgroup_try_charge_swap(struct folio *folio, 656 swp_entry_t entry) 657{ 658 if (mem_cgroup_disabled()) 659 return 0; 660 return __mem_cgroup_try_charge_swap(folio, entry); 661} 662 663extern void __mem_cgroup_uncharge_swap(swp_entry_t entry, unsigned int nr_pages); 664static inline void mem_cgroup_uncharge_swap(swp_entry_t entry, unsigned int nr_pages) 665{ 666 if (mem_cgroup_disabled()) 667 return; 668 __mem_cgroup_uncharge_swap(entry, nr_pages); 669} 670 671extern long mem_cgroup_get_nr_swap_pages(struct mem_cgroup *memcg); 672extern bool mem_cgroup_swap_full(struct folio *folio); 673#else 674static inline void mem_cgroup_swapout(struct folio *folio, swp_entry_t entry) 675{ 676} 677 678static inline int mem_cgroup_try_charge_swap(struct folio *folio, 679 swp_entry_t entry) 680{ 681 return 0; 682} 683 684static inline void mem_cgroup_uncharge_swap(swp_entry_t entry, 685 unsigned int nr_pages) 686{ 687} 688 689static inline long mem_cgroup_get_nr_swap_pages(struct mem_cgroup *memcg) 690{ 691 return get_nr_swap_pages(); 692} 693 694static inline bool mem_cgroup_swap_full(struct folio *folio) 695{ 696 return vm_swap_full(); 697} 698#endif 699 700#endif /* __KERNEL__*/ 701#endif /* _LINUX_SWAP_H */