at v6.2-rc3 711 lines 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 unsigned long reclaimed_slab; 157#ifdef CONFIG_LRU_GEN 158 /* per-thread mm walk data */ 159 struct lru_gen_mm_walk *mm_walk; 160#endif 161}; 162 163#ifdef __KERNEL__ 164 165struct address_space; 166struct sysinfo; 167struct writeback_control; 168struct zone; 169 170/* 171 * A swap extent maps a range of a swapfile's PAGE_SIZE pages onto a range of 172 * disk blocks. A rbtree of swap extents maps the entire swapfile (Where the 173 * term `swapfile' refers to either a blockdevice or an IS_REG file). Apart 174 * from setup, they're handled identically. 175 * 176 * We always assume that blocks are of size PAGE_SIZE. 177 */ 178struct swap_extent { 179 struct rb_node rb_node; 180 pgoff_t start_page; 181 pgoff_t nr_pages; 182 sector_t start_block; 183}; 184 185/* 186 * Max bad pages in the new format.. 187 */ 188#define MAX_SWAP_BADPAGES \ 189 ((offsetof(union swap_header, magic.magic) - \ 190 offsetof(union swap_header, info.badpages)) / sizeof(int)) 191 192enum { 193 SWP_USED = (1 << 0), /* is slot in swap_info[] used? */ 194 SWP_WRITEOK = (1 << 1), /* ok to write to this swap? */ 195 SWP_DISCARDABLE = (1 << 2), /* blkdev support discard */ 196 SWP_DISCARDING = (1 << 3), /* now discarding a free cluster */ 197 SWP_SOLIDSTATE = (1 << 4), /* blkdev seeks are cheap */ 198 SWP_CONTINUED = (1 << 5), /* swap_map has count continuation */ 199 SWP_BLKDEV = (1 << 6), /* its a block device */ 200 SWP_ACTIVATED = (1 << 7), /* set after swap_activate success */ 201 SWP_FS_OPS = (1 << 8), /* swapfile operations go through fs */ 202 SWP_AREA_DISCARD = (1 << 9), /* single-time swap area discards */ 203 SWP_PAGE_DISCARD = (1 << 10), /* freed swap page-cluster discards */ 204 SWP_STABLE_WRITES = (1 << 11), /* no overwrite PG_writeback pages */ 205 SWP_SYNCHRONOUS_IO = (1 << 12), /* synchronous IO is efficient */ 206 /* add others here before... */ 207 SWP_SCANNING = (1 << 14), /* refcount in scan_swap_map */ 208}; 209 210#define SWAP_CLUSTER_MAX 32UL 211#define COMPACT_CLUSTER_MAX SWAP_CLUSTER_MAX 212 213/* Bit flag in swap_map */ 214#define SWAP_HAS_CACHE 0x40 /* Flag page is cached, in first swap_map */ 215#define COUNT_CONTINUED 0x80 /* Flag swap_map continuation for full count */ 216 217/* Special value in first swap_map */ 218#define SWAP_MAP_MAX 0x3e /* Max count */ 219#define SWAP_MAP_BAD 0x3f /* Note page is bad */ 220#define SWAP_MAP_SHMEM 0xbf /* Owned by shmem/tmpfs */ 221 222/* Special value in each swap_map continuation */ 223#define SWAP_CONT_MAX 0x7f /* Max count */ 224 225/* 226 * We use this to track usage of a cluster. A cluster is a block of swap disk 227 * space with SWAPFILE_CLUSTER pages long and naturally aligns in disk. All 228 * free clusters are organized into a list. We fetch an entry from the list to 229 * get a free cluster. 230 * 231 * The data field stores next cluster if the cluster is free or cluster usage 232 * counter otherwise. The flags field determines if a cluster is free. This is 233 * protected by swap_info_struct.lock. 234 */ 235struct swap_cluster_info { 236 spinlock_t lock; /* 237 * Protect swap_cluster_info fields 238 * and swap_info_struct->swap_map 239 * elements correspond to the swap 240 * cluster 241 */ 242 unsigned int data:24; 243 unsigned int flags:8; 244}; 245#define CLUSTER_FLAG_FREE 1 /* This cluster is free */ 246#define CLUSTER_FLAG_NEXT_NULL 2 /* This cluster has no next cluster */ 247#define CLUSTER_FLAG_HUGE 4 /* This cluster is backing a transparent huge page */ 248 249/* 250 * We assign a cluster to each CPU, so each CPU can allocate swap entry from 251 * its own cluster and swapout sequentially. The purpose is to optimize swapout 252 * throughput. 253 */ 254struct percpu_cluster { 255 struct swap_cluster_info index; /* Current cluster index */ 256 unsigned int next; /* Likely next allocation offset */ 257}; 258 259struct swap_cluster_list { 260 struct swap_cluster_info head; 261 struct swap_cluster_info tail; 262}; 263 264/* 265 * The in-memory structure used to track swap areas. 266 */ 267struct swap_info_struct { 268 struct percpu_ref users; /* indicate and keep swap device valid. */ 269 unsigned long flags; /* SWP_USED etc: see above */ 270 signed short prio; /* swap priority of this type */ 271 struct plist_node list; /* entry in swap_active_head */ 272 signed char type; /* strange name for an index */ 273 unsigned int max; /* extent of the swap_map */ 274 unsigned char *swap_map; /* vmalloc'ed array of usage counts */ 275 struct swap_cluster_info *cluster_info; /* cluster info. Only for SSD */ 276 struct swap_cluster_list free_clusters; /* free clusters list */ 277 unsigned int lowest_bit; /* index of first free in swap_map */ 278 unsigned int highest_bit; /* index of last free in swap_map */ 279 unsigned int pages; /* total of usable pages of swap */ 280 unsigned int inuse_pages; /* number of those currently in use */ 281 unsigned int cluster_next; /* likely index for next allocation */ 282 unsigned int cluster_nr; /* countdown to next cluster search */ 283 unsigned int __percpu *cluster_next_cpu; /*percpu index for next allocation */ 284 struct percpu_cluster __percpu *percpu_cluster; /* per cpu's swap location */ 285 struct rb_root swap_extent_root;/* root of the swap extent rbtree */ 286 struct block_device *bdev; /* swap device or bdev of swap file */ 287 struct file *swap_file; /* seldom referenced */ 288 unsigned int old_block_size; /* seldom referenced */ 289 struct completion comp; /* seldom referenced */ 290#ifdef CONFIG_FRONTSWAP 291 unsigned long *frontswap_map; /* frontswap in-use, one bit per page */ 292 atomic_t frontswap_pages; /* frontswap pages in-use counter */ 293#endif 294 spinlock_t lock; /* 295 * protect map scan related fields like 296 * swap_map, lowest_bit, highest_bit, 297 * inuse_pages, cluster_next, 298 * cluster_nr, lowest_alloc, 299 * highest_alloc, free/discard cluster 300 * list. other fields are only changed 301 * at swapon/swapoff, so are protected 302 * by swap_lock. changing flags need 303 * hold this lock and swap_lock. If 304 * both locks need hold, hold swap_lock 305 * first. 306 */ 307 spinlock_t cont_lock; /* 308 * protect swap count continuation page 309 * list. 310 */ 311 struct work_struct discard_work; /* discard worker */ 312 struct swap_cluster_list discard_clusters; /* discard clusters list */ 313 struct plist_node avail_lists[]; /* 314 * entries in swap_avail_heads, one 315 * entry per node. 316 * Must be last as the number of the 317 * array is nr_node_ids, which is not 318 * a fixed value so have to allocate 319 * dynamically. 320 * And it has to be an array so that 321 * plist_for_each_* can work. 322 */ 323}; 324 325#ifdef CONFIG_64BIT 326#define SWAP_RA_ORDER_CEILING 5 327#else 328/* Avoid stack overflow, because we need to save part of page table */ 329#define SWAP_RA_ORDER_CEILING 3 330#define SWAP_RA_PTE_CACHE_SIZE (1 << SWAP_RA_ORDER_CEILING) 331#endif 332 333struct vma_swap_readahead { 334 unsigned short win; 335 unsigned short offset; 336 unsigned short nr_pte; 337#ifdef CONFIG_64BIT 338 pte_t *ptes; 339#else 340 pte_t ptes[SWAP_RA_PTE_CACHE_SIZE]; 341#endif 342}; 343 344static inline swp_entry_t folio_swap_entry(struct folio *folio) 345{ 346 swp_entry_t entry = { .val = page_private(&folio->page) }; 347 return entry; 348} 349 350static inline void folio_set_swap_entry(struct folio *folio, swp_entry_t entry) 351{ 352 folio->private = (void *)entry.val; 353} 354 355/* linux/mm/workingset.c */ 356void workingset_age_nonresident(struct lruvec *lruvec, unsigned long nr_pages); 357void *workingset_eviction(struct folio *folio, struct mem_cgroup *target_memcg); 358void workingset_refault(struct folio *folio, void *shadow); 359void workingset_activation(struct folio *folio); 360 361/* Only track the nodes of mappings with shadow entries */ 362void workingset_update_node(struct xa_node *node); 363extern struct list_lru shadow_nodes; 364#define mapping_set_update(xas, mapping) do { \ 365 if (!dax_mapping(mapping) && !shmem_mapping(mapping)) { \ 366 xas_set_update(xas, workingset_update_node); \ 367 xas_set_lru(xas, &shadow_nodes); \ 368 } \ 369} while (0) 370 371/* linux/mm/page_alloc.c */ 372extern unsigned long totalreserve_pages; 373 374/* Definition of global_zone_page_state not available yet */ 375#define nr_free_pages() global_zone_page_state(NR_FREE_PAGES) 376 377 378/* linux/mm/swap.c */ 379void lru_note_cost(struct lruvec *lruvec, bool file, 380 unsigned int nr_io, unsigned int nr_rotated); 381void lru_note_cost_refault(struct folio *); 382void folio_add_lru(struct folio *); 383void folio_add_lru_vma(struct folio *, struct vm_area_struct *); 384void mark_page_accessed(struct page *); 385void folio_mark_accessed(struct folio *); 386 387extern atomic_t lru_disable_count; 388 389static inline bool lru_cache_disabled(void) 390{ 391 return atomic_read(&lru_disable_count); 392} 393 394static inline void lru_cache_enable(void) 395{ 396 atomic_dec(&lru_disable_count); 397} 398 399extern void lru_cache_disable(void); 400extern void lru_add_drain(void); 401extern void lru_add_drain_cpu(int cpu); 402extern void lru_add_drain_cpu_zone(struct zone *zone); 403extern void lru_add_drain_all(void); 404extern void deactivate_page(struct page *page); 405extern void mark_page_lazyfree(struct page *page); 406extern void swap_setup(void); 407 408extern void lru_cache_add_inactive_or_unevictable(struct page *page, 409 struct vm_area_struct *vma); 410 411/* linux/mm/vmscan.c */ 412extern unsigned long zone_reclaimable_pages(struct zone *zone); 413extern unsigned long try_to_free_pages(struct zonelist *zonelist, int order, 414 gfp_t gfp_mask, nodemask_t *mask); 415 416#define MEMCG_RECLAIM_MAY_SWAP (1 << 1) 417#define MEMCG_RECLAIM_PROACTIVE (1 << 2) 418extern unsigned long try_to_free_mem_cgroup_pages(struct mem_cgroup *memcg, 419 unsigned long nr_pages, 420 gfp_t gfp_mask, 421 unsigned int reclaim_options, 422 nodemask_t *nodemask); 423extern unsigned long mem_cgroup_shrink_node(struct mem_cgroup *mem, 424 gfp_t gfp_mask, bool noswap, 425 pg_data_t *pgdat, 426 unsigned long *nr_scanned); 427extern unsigned long shrink_all_memory(unsigned long nr_pages); 428extern int vm_swappiness; 429long remove_mapping(struct address_space *mapping, struct folio *folio); 430 431extern unsigned long reclaim_pages(struct list_head *page_list); 432#ifdef CONFIG_NUMA 433extern int node_reclaim_mode; 434extern int sysctl_min_unmapped_ratio; 435extern int sysctl_min_slab_ratio; 436#else 437#define node_reclaim_mode 0 438#endif 439 440static inline bool node_reclaim_enabled(void) 441{ 442 /* Is any node_reclaim_mode bit set? */ 443 return node_reclaim_mode & (RECLAIM_ZONE|RECLAIM_WRITE|RECLAIM_UNMAP); 444} 445 446void check_move_unevictable_folios(struct folio_batch *fbatch); 447void check_move_unevictable_pages(struct pagevec *pvec); 448 449extern void kswapd_run(int nid); 450extern void kswapd_stop(int nid); 451 452#ifdef CONFIG_SWAP 453 454int add_swap_extent(struct swap_info_struct *sis, unsigned long start_page, 455 unsigned long nr_pages, sector_t start_block); 456int generic_swapfile_activate(struct swap_info_struct *, struct file *, 457 sector_t *); 458 459static inline unsigned long total_swapcache_pages(void) 460{ 461 return global_node_page_state(NR_SWAPCACHE); 462} 463 464extern void free_swap_cache(struct page *page); 465extern void free_page_and_swap_cache(struct page *); 466extern void free_pages_and_swap_cache(struct encoded_page **, int); 467/* linux/mm/swapfile.c */ 468extern atomic_long_t nr_swap_pages; 469extern long total_swap_pages; 470extern atomic_t nr_rotate_swap; 471extern bool has_usable_swap(void); 472 473/* Swap 50% full? Release swapcache more aggressively.. */ 474static inline bool vm_swap_full(void) 475{ 476 return atomic_long_read(&nr_swap_pages) * 2 < total_swap_pages; 477} 478 479static inline long get_nr_swap_pages(void) 480{ 481 return atomic_long_read(&nr_swap_pages); 482} 483 484extern void si_swapinfo(struct sysinfo *); 485swp_entry_t folio_alloc_swap(struct folio *folio); 486bool folio_free_swap(struct folio *folio); 487void put_swap_folio(struct folio *folio, swp_entry_t entry); 488extern swp_entry_t get_swap_page_of_type(int); 489extern int get_swap_pages(int n, swp_entry_t swp_entries[], int entry_size); 490extern int add_swap_count_continuation(swp_entry_t, gfp_t); 491extern void swap_shmem_alloc(swp_entry_t); 492extern int swap_duplicate(swp_entry_t); 493extern int swapcache_prepare(swp_entry_t); 494extern void swap_free(swp_entry_t); 495extern void swapcache_free_entries(swp_entry_t *entries, int n); 496extern int free_swap_and_cache(swp_entry_t); 497int swap_type_of(dev_t device, sector_t offset); 498int find_first_swap(dev_t *device); 499extern unsigned int count_swap_pages(int, int); 500extern sector_t swapdev_block(int, pgoff_t); 501extern int __swap_count(swp_entry_t entry); 502extern int __swp_swapcount(swp_entry_t entry); 503extern int swp_swapcount(swp_entry_t entry); 504extern struct swap_info_struct *page_swap_info(struct page *); 505extern struct swap_info_struct *swp_swap_info(swp_entry_t entry); 506struct backing_dev_info; 507extern int init_swap_address_space(unsigned int type, unsigned long nr_pages); 508extern void exit_swap_address_space(unsigned int type); 509extern struct swap_info_struct *get_swap_device(swp_entry_t entry); 510sector_t swap_page_sector(struct page *page); 511 512static inline void put_swap_device(struct swap_info_struct *si) 513{ 514 percpu_ref_put(&si->users); 515} 516 517#else /* CONFIG_SWAP */ 518static inline struct swap_info_struct *swp_swap_info(swp_entry_t entry) 519{ 520 return NULL; 521} 522 523static inline struct swap_info_struct *get_swap_device(swp_entry_t entry) 524{ 525 return NULL; 526} 527 528static inline void put_swap_device(struct swap_info_struct *si) 529{ 530} 531 532#define get_nr_swap_pages() 0L 533#define total_swap_pages 0L 534#define total_swapcache_pages() 0UL 535#define vm_swap_full() 0 536 537#define si_swapinfo(val) \ 538 do { (val)->freeswap = (val)->totalswap = 0; } while (0) 539/* only sparc can not include linux/pagemap.h in this file 540 * so leave put_page and release_pages undeclared... */ 541#define free_page_and_swap_cache(page) \ 542 put_page(page) 543#define free_pages_and_swap_cache(pages, nr) \ 544 release_pages((pages), (nr)); 545 546/* used to sanity check ptes in zap_pte_range when CONFIG_SWAP=0 */ 547#define free_swap_and_cache(e) is_pfn_swap_entry(e) 548 549static inline void free_swap_cache(struct page *page) 550{ 551} 552 553static inline int add_swap_count_continuation(swp_entry_t swp, gfp_t gfp_mask) 554{ 555 return 0; 556} 557 558static inline void swap_shmem_alloc(swp_entry_t swp) 559{ 560} 561 562static inline int swap_duplicate(swp_entry_t swp) 563{ 564 return 0; 565} 566 567static inline void swap_free(swp_entry_t swp) 568{ 569} 570 571static inline void put_swap_folio(struct folio *folio, swp_entry_t swp) 572{ 573} 574 575static inline int __swap_count(swp_entry_t entry) 576{ 577 return 0; 578} 579 580static inline int __swp_swapcount(swp_entry_t entry) 581{ 582 return 0; 583} 584 585static inline int swp_swapcount(swp_entry_t entry) 586{ 587 return 0; 588} 589 590static inline swp_entry_t folio_alloc_swap(struct folio *folio) 591{ 592 swp_entry_t entry; 593 entry.val = 0; 594 return entry; 595} 596 597static inline bool folio_free_swap(struct folio *folio) 598{ 599 return false; 600} 601 602static inline int add_swap_extent(struct swap_info_struct *sis, 603 unsigned long start_page, 604 unsigned long nr_pages, sector_t start_block) 605{ 606 return -EINVAL; 607} 608#endif /* CONFIG_SWAP */ 609 610#ifdef CONFIG_THP_SWAP 611extern int split_swap_cluster(swp_entry_t entry); 612#else 613static inline int split_swap_cluster(swp_entry_t entry) 614{ 615 return 0; 616} 617#endif 618 619#ifdef CONFIG_MEMCG 620static inline int mem_cgroup_swappiness(struct mem_cgroup *memcg) 621{ 622 /* Cgroup2 doesn't have per-cgroup swappiness */ 623 if (cgroup_subsys_on_dfl(memory_cgrp_subsys)) 624 return vm_swappiness; 625 626 /* root ? */ 627 if (mem_cgroup_disabled() || mem_cgroup_is_root(memcg)) 628 return vm_swappiness; 629 630 return memcg->swappiness; 631} 632#else 633static inline int mem_cgroup_swappiness(struct mem_cgroup *mem) 634{ 635 return vm_swappiness; 636} 637#endif 638 639#ifdef CONFIG_ZSWAP 640extern u64 zswap_pool_total_size; 641extern atomic_t zswap_stored_pages; 642#endif 643 644#if defined(CONFIG_SWAP) && defined(CONFIG_MEMCG) && defined(CONFIG_BLK_CGROUP) 645extern void __cgroup_throttle_swaprate(struct page *page, gfp_t gfp_mask); 646static inline void cgroup_throttle_swaprate(struct page *page, gfp_t gfp_mask) 647{ 648 if (mem_cgroup_disabled()) 649 return; 650 __cgroup_throttle_swaprate(page, gfp_mask); 651} 652#else 653static inline void cgroup_throttle_swaprate(struct page *page, gfp_t gfp_mask) 654{ 655} 656#endif 657static inline void folio_throttle_swaprate(struct folio *folio, gfp_t gfp) 658{ 659 cgroup_throttle_swaprate(&folio->page, gfp); 660} 661 662#if defined(CONFIG_MEMCG) && defined(CONFIG_SWAP) 663void mem_cgroup_swapout(struct folio *folio, swp_entry_t entry); 664int __mem_cgroup_try_charge_swap(struct folio *folio, swp_entry_t entry); 665static inline int mem_cgroup_try_charge_swap(struct folio *folio, 666 swp_entry_t entry) 667{ 668 if (mem_cgroup_disabled()) 669 return 0; 670 return __mem_cgroup_try_charge_swap(folio, entry); 671} 672 673extern void __mem_cgroup_uncharge_swap(swp_entry_t entry, unsigned int nr_pages); 674static inline void mem_cgroup_uncharge_swap(swp_entry_t entry, unsigned int nr_pages) 675{ 676 if (mem_cgroup_disabled()) 677 return; 678 __mem_cgroup_uncharge_swap(entry, nr_pages); 679} 680 681extern long mem_cgroup_get_nr_swap_pages(struct mem_cgroup *memcg); 682extern bool mem_cgroup_swap_full(struct folio *folio); 683#else 684static inline void mem_cgroup_swapout(struct folio *folio, swp_entry_t entry) 685{ 686} 687 688static inline int mem_cgroup_try_charge_swap(struct folio *folio, 689 swp_entry_t entry) 690{ 691 return 0; 692} 693 694static inline void mem_cgroup_uncharge_swap(swp_entry_t entry, 695 unsigned int nr_pages) 696{ 697} 698 699static inline long mem_cgroup_get_nr_swap_pages(struct mem_cgroup *memcg) 700{ 701 return get_nr_swap_pages(); 702} 703 704static inline bool mem_cgroup_swap_full(struct folio *folio) 705{ 706 return vm_swap_full(); 707} 708#endif 709 710#endif /* __KERNEL__*/ 711#endif /* _LINUX_SWAP_H */