linux/include/linux/hugetlb.h
<<
>>
Prefs
   1/* SPDX-License-Identifier: GPL-2.0 */
   2#ifndef _LINUX_HUGETLB_H
   3#define _LINUX_HUGETLB_H
   4
   5#include <linux/mm_types.h>
   6#include <linux/mmdebug.h>
   7#include <linux/fs.h>
   8#include <linux/hugetlb_inline.h>
   9#include <linux/cgroup.h>
  10#include <linux/list.h>
  11#include <linux/kref.h>
  12#include <linux/pgtable.h>
  13#include <linux/gfp.h>
  14#include <linux/userfaultfd_k.h>
  15
  16struct ctl_table;
  17struct user_struct;
  18struct mmu_gather;
  19
  20#ifndef is_hugepd
  21typedef struct { unsigned long pd; } hugepd_t;
  22#define is_hugepd(hugepd) (0)
  23#define __hugepd(x) ((hugepd_t) { (x) })
  24#endif
  25
  26#ifdef CONFIG_HUGETLB_PAGE
  27
  28#include <linux/mempolicy.h>
  29#include <linux/shm.h>
  30#include <asm/tlbflush.h>
  31
  32/*
  33 * For HugeTLB page, there are more metadata to save in the struct page. But
  34 * the head struct page cannot meet our needs, so we have to abuse other tail
  35 * struct page to store the metadata. In order to avoid conflicts caused by
  36 * subsequent use of more tail struct pages, we gather these discrete indexes
  37 * of tail struct page here.
  38 */
  39enum {
  40        SUBPAGE_INDEX_SUBPOOL = 1,      /* reuse page->private */
  41#ifdef CONFIG_CGROUP_HUGETLB
  42        SUBPAGE_INDEX_CGROUP,           /* reuse page->private */
  43        SUBPAGE_INDEX_CGROUP_RSVD,      /* reuse page->private */
  44        __MAX_CGROUP_SUBPAGE_INDEX = SUBPAGE_INDEX_CGROUP_RSVD,
  45#endif
  46        __NR_USED_SUBPAGE,
  47};
  48
  49struct hugepage_subpool {
  50        spinlock_t lock;
  51        long count;
  52        long max_hpages;        /* Maximum huge pages or -1 if no maximum. */
  53        long used_hpages;       /* Used count against maximum, includes */
  54                                /* both allocated and reserved pages. */
  55        struct hstate *hstate;
  56        long min_hpages;        /* Minimum huge pages or -1 if no minimum. */
  57        long rsv_hpages;        /* Pages reserved against global pool to */
  58                                /* satisfy minimum size. */
  59};
  60
  61struct resv_map {
  62        struct kref refs;
  63        spinlock_t lock;
  64        struct list_head regions;
  65        long adds_in_progress;
  66        struct list_head region_cache;
  67        long region_cache_count;
  68#ifdef CONFIG_CGROUP_HUGETLB
  69        /*
  70         * On private mappings, the counter to uncharge reservations is stored
  71         * here. If these fields are 0, then either the mapping is shared, or
  72         * cgroup accounting is disabled for this resv_map.
  73         */
  74        struct page_counter *reservation_counter;
  75        unsigned long pages_per_hpage;
  76        struct cgroup_subsys_state *css;
  77#endif
  78};
  79
  80/*
  81 * Region tracking -- allows tracking of reservations and instantiated pages
  82 *                    across the pages in a mapping.
  83 *
  84 * The region data structures are embedded into a resv_map and protected
  85 * by a resv_map's lock.  The set of regions within the resv_map represent
  86 * reservations for huge pages, or huge pages that have already been
  87 * instantiated within the map.  The from and to elements are huge page
  88 * indices into the associated mapping.  from indicates the starting index
  89 * of the region.  to represents the first index past the end of  the region.
  90 *
  91 * For example, a file region structure with from == 0 and to == 4 represents
  92 * four huge pages in a mapping.  It is important to note that the to element
  93 * represents the first element past the end of the region. This is used in
  94 * arithmetic as 4(to) - 0(from) = 4 huge pages in the region.
  95 *
  96 * Interval notation of the form [from, to) will be used to indicate that
  97 * the endpoint from is inclusive and to is exclusive.
  98 */
  99struct file_region {
 100        struct list_head link;
 101        long from;
 102        long to;
 103#ifdef CONFIG_CGROUP_HUGETLB
 104        /*
 105         * On shared mappings, each reserved region appears as a struct
 106         * file_region in resv_map. These fields hold the info needed to
 107         * uncharge each reservation.
 108         */
 109        struct page_counter *reservation_counter;
 110        struct cgroup_subsys_state *css;
 111#endif
 112};
 113
 114extern struct resv_map *resv_map_alloc(void);
 115void resv_map_release(struct kref *ref);
 116
 117extern spinlock_t hugetlb_lock;
 118extern int hugetlb_max_hstate __read_mostly;
 119#define for_each_hstate(h) \
 120        for ((h) = hstates; (h) < &hstates[hugetlb_max_hstate]; (h)++)
 121
 122struct hugepage_subpool *hugepage_new_subpool(struct hstate *h, long max_hpages,
 123                                                long min_hpages);
 124void hugepage_put_subpool(struct hugepage_subpool *spool);
 125
 126void reset_vma_resv_huge_pages(struct vm_area_struct *vma);
 127int hugetlb_sysctl_handler(struct ctl_table *, int, void *, size_t *, loff_t *);
 128int hugetlb_overcommit_handler(struct ctl_table *, int, void *, size_t *,
 129                loff_t *);
 130int hugetlb_treat_movable_handler(struct ctl_table *, int, void *, size_t *,
 131                loff_t *);
 132int hugetlb_mempolicy_sysctl_handler(struct ctl_table *, int, void *, size_t *,
 133                loff_t *);
 134
 135int copy_hugetlb_page_range(struct mm_struct *, struct mm_struct *, struct vm_area_struct *);
 136long follow_hugetlb_page(struct mm_struct *, struct vm_area_struct *,
 137                         struct page **, struct vm_area_struct **,
 138                         unsigned long *, unsigned long *, long, unsigned int,
 139                         int *);
 140void unmap_hugepage_range(struct vm_area_struct *,
 141                          unsigned long, unsigned long, struct page *);
 142void __unmap_hugepage_range_final(struct mmu_gather *tlb,
 143                          struct vm_area_struct *vma,
 144                          unsigned long start, unsigned long end,
 145                          struct page *ref_page);
 146void __unmap_hugepage_range(struct mmu_gather *tlb, struct vm_area_struct *vma,
 147                                unsigned long start, unsigned long end,
 148                                struct page *ref_page);
 149void hugetlb_report_meminfo(struct seq_file *);
 150int hugetlb_report_node_meminfo(char *buf, int len, int nid);
 151void hugetlb_show_meminfo(void);
 152unsigned long hugetlb_total_pages(void);
 153vm_fault_t hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
 154                        unsigned long address, unsigned int flags);
 155#ifdef CONFIG_USERFAULTFD
 156int hugetlb_mcopy_atomic_pte(struct mm_struct *dst_mm, pte_t *dst_pte,
 157                                struct vm_area_struct *dst_vma,
 158                                unsigned long dst_addr,
 159                                unsigned long src_addr,
 160                                enum mcopy_atomic_mode mode,
 161                                struct page **pagep);
 162#endif /* CONFIG_USERFAULTFD */
 163bool hugetlb_reserve_pages(struct inode *inode, long from, long to,
 164                                                struct vm_area_struct *vma,
 165                                                vm_flags_t vm_flags);
 166long hugetlb_unreserve_pages(struct inode *inode, long start, long end,
 167                                                long freed);
 168bool isolate_huge_page(struct page *page, struct list_head *list);
 169int get_hwpoison_huge_page(struct page *page, bool *hugetlb);
 170void putback_active_hugepage(struct page *page);
 171void move_hugetlb_state(struct page *oldpage, struct page *newpage, int reason);
 172void free_huge_page(struct page *page);
 173void hugetlb_fix_reserve_counts(struct inode *inode);
 174extern struct mutex *hugetlb_fault_mutex_table;
 175u32 hugetlb_fault_mutex_hash(struct address_space *mapping, pgoff_t idx);
 176
 177pte_t *huge_pmd_share(struct mm_struct *mm, struct vm_area_struct *vma,
 178                      unsigned long addr, pud_t *pud);
 179
 180struct address_space *hugetlb_page_mapping_lock_write(struct page *hpage);
 181
 182extern int sysctl_hugetlb_shm_group;
 183extern struct list_head huge_boot_pages;
 184
 185/* arch callbacks */
 186
 187pte_t *huge_pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
 188                        unsigned long addr, unsigned long sz);
 189pte_t *huge_pte_offset(struct mm_struct *mm,
 190                       unsigned long addr, unsigned long sz);
 191int huge_pmd_unshare(struct mm_struct *mm, struct vm_area_struct *vma,
 192                                unsigned long *addr, pte_t *ptep);
 193void adjust_range_if_pmd_sharing_possible(struct vm_area_struct *vma,
 194                                unsigned long *start, unsigned long *end);
 195struct page *follow_huge_addr(struct mm_struct *mm, unsigned long address,
 196                              int write);
 197struct page *follow_huge_pd(struct vm_area_struct *vma,
 198                            unsigned long address, hugepd_t hpd,
 199                            int flags, int pdshift);
 200struct page *follow_huge_pmd(struct mm_struct *mm, unsigned long address,
 201                                pmd_t *pmd, int flags);
 202struct page *follow_huge_pud(struct mm_struct *mm, unsigned long address,
 203                                pud_t *pud, int flags);
 204struct page *follow_huge_pgd(struct mm_struct *mm, unsigned long address,
 205                             pgd_t *pgd, int flags);
 206
 207int pmd_huge(pmd_t pmd);
 208int pud_huge(pud_t pud);
 209unsigned long hugetlb_change_protection(struct vm_area_struct *vma,
 210                unsigned long address, unsigned long end, pgprot_t newprot);
 211
 212bool is_hugetlb_entry_migration(pte_t pte);
 213void hugetlb_unshare_all_pmds(struct vm_area_struct *vma);
 214
 215#else /* !CONFIG_HUGETLB_PAGE */
 216
 217static inline void reset_vma_resv_huge_pages(struct vm_area_struct *vma)
 218{
 219}
 220
 221static inline unsigned long hugetlb_total_pages(void)
 222{
 223        return 0;
 224}
 225
 226static inline struct address_space *hugetlb_page_mapping_lock_write(
 227                                                        struct page *hpage)
 228{
 229        return NULL;
 230}
 231
 232static inline int huge_pmd_unshare(struct mm_struct *mm,
 233                                        struct vm_area_struct *vma,
 234                                        unsigned long *addr, pte_t *ptep)
 235{
 236        return 0;
 237}
 238
 239static inline void adjust_range_if_pmd_sharing_possible(
 240                                struct vm_area_struct *vma,
 241                                unsigned long *start, unsigned long *end)
 242{
 243}
 244
 245static inline long follow_hugetlb_page(struct mm_struct *mm,
 246                        struct vm_area_struct *vma, struct page **pages,
 247                        struct vm_area_struct **vmas, unsigned long *position,
 248                        unsigned long *nr_pages, long i, unsigned int flags,
 249                        int *nonblocking)
 250{
 251        BUG();
 252        return 0;
 253}
 254
 255static inline struct page *follow_huge_addr(struct mm_struct *mm,
 256                                        unsigned long address, int write)
 257{
 258        return ERR_PTR(-EINVAL);
 259}
 260
 261static inline int copy_hugetlb_page_range(struct mm_struct *dst,
 262                        struct mm_struct *src, struct vm_area_struct *vma)
 263{
 264        BUG();
 265        return 0;
 266}
 267
 268static inline void hugetlb_report_meminfo(struct seq_file *m)
 269{
 270}
 271
 272static inline int hugetlb_report_node_meminfo(char *buf, int len, int nid)
 273{
 274        return 0;
 275}
 276
 277static inline void hugetlb_show_meminfo(void)
 278{
 279}
 280
 281static inline struct page *follow_huge_pd(struct vm_area_struct *vma,
 282                                unsigned long address, hugepd_t hpd, int flags,
 283                                int pdshift)
 284{
 285        return NULL;
 286}
 287
 288static inline struct page *follow_huge_pmd(struct mm_struct *mm,
 289                                unsigned long address, pmd_t *pmd, int flags)
 290{
 291        return NULL;
 292}
 293
 294static inline struct page *follow_huge_pud(struct mm_struct *mm,
 295                                unsigned long address, pud_t *pud, int flags)
 296{
 297        return NULL;
 298}
 299
 300static inline struct page *follow_huge_pgd(struct mm_struct *mm,
 301                                unsigned long address, pgd_t *pgd, int flags)
 302{
 303        return NULL;
 304}
 305
 306static inline int prepare_hugepage_range(struct file *file,
 307                                unsigned long addr, unsigned long len)
 308{
 309        return -EINVAL;
 310}
 311
 312static inline int pmd_huge(pmd_t pmd)
 313{
 314        return 0;
 315}
 316
 317static inline int pud_huge(pud_t pud)
 318{
 319        return 0;
 320}
 321
 322static inline int is_hugepage_only_range(struct mm_struct *mm,
 323                                        unsigned long addr, unsigned long len)
 324{
 325        return 0;
 326}
 327
 328static inline void hugetlb_free_pgd_range(struct mmu_gather *tlb,
 329                                unsigned long addr, unsigned long end,
 330                                unsigned long floor, unsigned long ceiling)
 331{
 332        BUG();
 333}
 334
 335#ifdef CONFIG_USERFAULTFD
 336static inline int hugetlb_mcopy_atomic_pte(struct mm_struct *dst_mm,
 337                                                pte_t *dst_pte,
 338                                                struct vm_area_struct *dst_vma,
 339                                                unsigned long dst_addr,
 340                                                unsigned long src_addr,
 341                                                enum mcopy_atomic_mode mode,
 342                                                struct page **pagep)
 343{
 344        BUG();
 345        return 0;
 346}
 347#endif /* CONFIG_USERFAULTFD */
 348
 349static inline pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr,
 350                                        unsigned long sz)
 351{
 352        return NULL;
 353}
 354
 355static inline bool isolate_huge_page(struct page *page, struct list_head *list)
 356{
 357        return false;
 358}
 359
 360static inline int get_hwpoison_huge_page(struct page *page, bool *hugetlb)
 361{
 362        return 0;
 363}
 364
 365static inline void putback_active_hugepage(struct page *page)
 366{
 367}
 368
 369static inline void move_hugetlb_state(struct page *oldpage,
 370                                        struct page *newpage, int reason)
 371{
 372}
 373
 374static inline unsigned long hugetlb_change_protection(
 375                        struct vm_area_struct *vma, unsigned long address,
 376                        unsigned long end, pgprot_t newprot)
 377{
 378        return 0;
 379}
 380
 381static inline void __unmap_hugepage_range_final(struct mmu_gather *tlb,
 382                        struct vm_area_struct *vma, unsigned long start,
 383                        unsigned long end, struct page *ref_page)
 384{
 385        BUG();
 386}
 387
 388static inline void __unmap_hugepage_range(struct mmu_gather *tlb,
 389                        struct vm_area_struct *vma, unsigned long start,
 390                        unsigned long end, struct page *ref_page)
 391{
 392        BUG();
 393}
 394
 395static inline vm_fault_t hugetlb_fault(struct mm_struct *mm,
 396                        struct vm_area_struct *vma, unsigned long address,
 397                        unsigned int flags)
 398{
 399        BUG();
 400        return 0;
 401}
 402
 403static inline void hugetlb_unshare_all_pmds(struct vm_area_struct *vma) { }
 404
 405#endif /* !CONFIG_HUGETLB_PAGE */
 406/*
 407 * hugepages at page global directory. If arch support
 408 * hugepages at pgd level, they need to define this.
 409 */
 410#ifndef pgd_huge
 411#define pgd_huge(x)     0
 412#endif
 413#ifndef p4d_huge
 414#define p4d_huge(x)     0
 415#endif
 416
 417#ifndef pgd_write
 418static inline int pgd_write(pgd_t pgd)
 419{
 420        BUG();
 421        return 0;
 422}
 423#endif
 424
 425#define HUGETLB_ANON_FILE "anon_hugepage"
 426
 427enum {
 428        /*
 429         * The file will be used as an shm file so shmfs accounting rules
 430         * apply
 431         */
 432        HUGETLB_SHMFS_INODE     = 1,
 433        /*
 434         * The file is being created on the internal vfs mount and shmfs
 435         * accounting rules do not apply
 436         */
 437        HUGETLB_ANONHUGE_INODE  = 2,
 438};
 439
 440#ifdef CONFIG_HUGETLBFS
 441struct hugetlbfs_sb_info {
 442        long    max_inodes;   /* inodes allowed */
 443        long    free_inodes;  /* inodes free */
 444        spinlock_t      stat_lock;
 445        struct hstate *hstate;
 446        struct hugepage_subpool *spool;
 447        kuid_t  uid;
 448        kgid_t  gid;
 449        umode_t mode;
 450};
 451
 452static inline struct hugetlbfs_sb_info *HUGETLBFS_SB(struct super_block *sb)
 453{
 454        return sb->s_fs_info;
 455}
 456
 457struct hugetlbfs_inode_info {
 458        struct shared_policy policy;
 459        struct inode vfs_inode;
 460        unsigned int seals;
 461};
 462
 463static inline struct hugetlbfs_inode_info *HUGETLBFS_I(struct inode *inode)
 464{
 465        return container_of(inode, struct hugetlbfs_inode_info, vfs_inode);
 466}
 467
 468extern const struct file_operations hugetlbfs_file_operations;
 469extern const struct vm_operations_struct hugetlb_vm_ops;
 470struct file *hugetlb_file_setup(const char *name, size_t size, vm_flags_t acct,
 471                                struct ucounts **ucounts, int creat_flags,
 472                                int page_size_log);
 473
 474static inline bool is_file_hugepages(struct file *file)
 475{
 476        if (file->f_op == &hugetlbfs_file_operations)
 477                return true;
 478
 479        return is_file_shm_hugepages(file);
 480}
 481
 482static inline struct hstate *hstate_inode(struct inode *i)
 483{
 484        return HUGETLBFS_SB(i->i_sb)->hstate;
 485}
 486#else /* !CONFIG_HUGETLBFS */
 487
 488#define is_file_hugepages(file)                 false
 489static inline struct file *
 490hugetlb_file_setup(const char *name, size_t size, vm_flags_t acctflag,
 491                struct ucounts **ucounts, int creat_flags,
 492                int page_size_log)
 493{
 494        return ERR_PTR(-ENOSYS);
 495}
 496
 497static inline struct hstate *hstate_inode(struct inode *i)
 498{
 499        return NULL;
 500}
 501#endif /* !CONFIG_HUGETLBFS */
 502
 503#ifdef HAVE_ARCH_HUGETLB_UNMAPPED_AREA
 504unsigned long hugetlb_get_unmapped_area(struct file *file, unsigned long addr,
 505                                        unsigned long len, unsigned long pgoff,
 506                                        unsigned long flags);
 507#endif /* HAVE_ARCH_HUGETLB_UNMAPPED_AREA */
 508
 509/*
 510 * huegtlb page specific state flags.  These flags are located in page.private
 511 * of the hugetlb head page.  Functions created via the below macros should be
 512 * used to manipulate these flags.
 513 *
 514 * HPG_restore_reserve - Set when a hugetlb page consumes a reservation at
 515 *      allocation time.  Cleared when page is fully instantiated.  Free
 516 *      routine checks flag to restore a reservation on error paths.
 517 *      Synchronization:  Examined or modified by code that knows it has
 518 *      the only reference to page.  i.e. After allocation but before use
 519 *      or when the page is being freed.
 520 * HPG_migratable  - Set after a newly allocated page is added to the page
 521 *      cache and/or page tables.  Indicates the page is a candidate for
 522 *      migration.
 523 *      Synchronization:  Initially set after new page allocation with no
 524 *      locking.  When examined and modified during migration processing
 525 *      (isolate, migrate, putback) the hugetlb_lock is held.
 526 * HPG_temporary - - Set on a page that is temporarily allocated from the buddy
 527 *      allocator.  Typically used for migration target pages when no pages
 528 *      are available in the pool.  The hugetlb free page path will
 529 *      immediately free pages with this flag set to the buddy allocator.
 530 *      Synchronization: Can be set after huge page allocation from buddy when
 531 *      code knows it has only reference.  All other examinations and
 532 *      modifications require hugetlb_lock.
 533 * HPG_freed - Set when page is on the free lists.
 534 *      Synchronization: hugetlb_lock held for examination and modification.
 535 * HPG_vmemmap_optimized - Set when the vmemmap pages of the page are freed.
 536 */
 537enum hugetlb_page_flags {
 538        HPG_restore_reserve = 0,
 539        HPG_migratable,
 540        HPG_temporary,
 541        HPG_freed,
 542        HPG_vmemmap_optimized,
 543        __NR_HPAGEFLAGS,
 544};
 545
 546/*
 547 * Macros to create test, set and clear function definitions for
 548 * hugetlb specific page flags.
 549 */
 550#ifdef CONFIG_HUGETLB_PAGE
 551#define TESTHPAGEFLAG(uname, flname)                            \
 552static inline int HPage##uname(struct page *page)               \
 553        { return test_bit(HPG_##flname, &(page->private)); }
 554
 555#define SETHPAGEFLAG(uname, flname)                             \
 556static inline void SetHPage##uname(struct page *page)           \
 557        { set_bit(HPG_##flname, &(page->private)); }
 558
 559#define CLEARHPAGEFLAG(uname, flname)                           \
 560static inline void ClearHPage##uname(struct page *page)         \
 561        { clear_bit(HPG_##flname, &(page->private)); }
 562#else
 563#define TESTHPAGEFLAG(uname, flname)                            \
 564static inline int HPage##uname(struct page *page)               \
 565        { return 0; }
 566
 567#define SETHPAGEFLAG(uname, flname)                             \
 568static inline void SetHPage##uname(struct page *page)           \
 569        { }
 570
 571#define CLEARHPAGEFLAG(uname, flname)                           \
 572static inline void ClearHPage##uname(struct page *page)         \
 573        { }
 574#endif
 575
 576#define HPAGEFLAG(uname, flname)                                \
 577        TESTHPAGEFLAG(uname, flname)                            \
 578        SETHPAGEFLAG(uname, flname)                             \
 579        CLEARHPAGEFLAG(uname, flname)                           \
 580
 581/*
 582 * Create functions associated with hugetlb page flags
 583 */
 584HPAGEFLAG(RestoreReserve, restore_reserve)
 585HPAGEFLAG(Migratable, migratable)
 586HPAGEFLAG(Temporary, temporary)
 587HPAGEFLAG(Freed, freed)
 588HPAGEFLAG(VmemmapOptimized, vmemmap_optimized)
 589
 590#ifdef CONFIG_HUGETLB_PAGE
 591
 592#define HSTATE_NAME_LEN 32
 593/* Defines one hugetlb page size */
 594struct hstate {
 595        struct mutex resize_lock;
 596        int next_nid_to_alloc;
 597        int next_nid_to_free;
 598        unsigned int order;
 599        unsigned long mask;
 600        unsigned long max_huge_pages;
 601        unsigned long nr_huge_pages;
 602        unsigned long free_huge_pages;
 603        unsigned long resv_huge_pages;
 604        unsigned long surplus_huge_pages;
 605        unsigned long nr_overcommit_huge_pages;
 606        struct list_head hugepage_activelist;
 607        struct list_head hugepage_freelists[MAX_NUMNODES];
 608        unsigned int nr_huge_pages_node[MAX_NUMNODES];
 609        unsigned int free_huge_pages_node[MAX_NUMNODES];
 610        unsigned int surplus_huge_pages_node[MAX_NUMNODES];
 611#ifdef CONFIG_HUGETLB_PAGE_FREE_VMEMMAP
 612        unsigned int nr_free_vmemmap_pages;
 613#endif
 614#ifdef CONFIG_CGROUP_HUGETLB
 615        /* cgroup control files */
 616        struct cftype cgroup_files_dfl[7];
 617        struct cftype cgroup_files_legacy[9];
 618#endif
 619        char name[HSTATE_NAME_LEN];
 620};
 621
 622struct huge_bootmem_page {
 623        struct list_head list;
 624        struct hstate *hstate;
 625};
 626
 627int isolate_or_dissolve_huge_page(struct page *page, struct list_head *list);
 628struct page *alloc_huge_page(struct vm_area_struct *vma,
 629                                unsigned long addr, int avoid_reserve);
 630struct page *alloc_huge_page_nodemask(struct hstate *h, int preferred_nid,
 631                                nodemask_t *nmask, gfp_t gfp_mask);
 632struct page *alloc_huge_page_vma(struct hstate *h, struct vm_area_struct *vma,
 633                                unsigned long address);
 634int huge_add_to_page_cache(struct page *page, struct address_space *mapping,
 635                        pgoff_t idx);
 636void restore_reserve_on_error(struct hstate *h, struct vm_area_struct *vma,
 637                                unsigned long address, struct page *page);
 638
 639/* arch callback */
 640int __init __alloc_bootmem_huge_page(struct hstate *h);
 641int __init alloc_bootmem_huge_page(struct hstate *h);
 642
 643void __init hugetlb_add_hstate(unsigned order);
 644bool __init arch_hugetlb_valid_size(unsigned long size);
 645struct hstate *size_to_hstate(unsigned long size);
 646
 647#ifndef HUGE_MAX_HSTATE
 648#define HUGE_MAX_HSTATE 1
 649#endif
 650
 651extern struct hstate hstates[HUGE_MAX_HSTATE];
 652extern unsigned int default_hstate_idx;
 653
 654#define default_hstate (hstates[default_hstate_idx])
 655
 656/*
 657 * hugetlb page subpool pointer located in hpage[1].private
 658 */
 659static inline struct hugepage_subpool *hugetlb_page_subpool(struct page *hpage)
 660{
 661        return (void *)page_private(hpage + SUBPAGE_INDEX_SUBPOOL);
 662}
 663
 664static inline void hugetlb_set_page_subpool(struct page *hpage,
 665                                        struct hugepage_subpool *subpool)
 666{
 667        set_page_private(hpage + SUBPAGE_INDEX_SUBPOOL, (unsigned long)subpool);
 668}
 669
 670static inline struct hstate *hstate_file(struct file *f)
 671{
 672        return hstate_inode(file_inode(f));
 673}
 674
 675static inline struct hstate *hstate_sizelog(int page_size_log)
 676{
 677        if (!page_size_log)
 678                return &default_hstate;
 679
 680        return size_to_hstate(1UL << page_size_log);
 681}
 682
 683static inline struct hstate *hstate_vma(struct vm_area_struct *vma)
 684{
 685        return hstate_file(vma->vm_file);
 686}
 687
 688static inline unsigned long huge_page_size(struct hstate *h)
 689{
 690        return (unsigned long)PAGE_SIZE << h->order;
 691}
 692
 693extern unsigned long vma_kernel_pagesize(struct vm_area_struct *vma);
 694
 695extern unsigned long vma_mmu_pagesize(struct vm_area_struct *vma);
 696
 697static inline unsigned long huge_page_mask(struct hstate *h)
 698{
 699        return h->mask;
 700}
 701
 702static inline unsigned int huge_page_order(struct hstate *h)
 703{
 704        return h->order;
 705}
 706
 707static inline unsigned huge_page_shift(struct hstate *h)
 708{
 709        return h->order + PAGE_SHIFT;
 710}
 711
 712static inline bool hstate_is_gigantic(struct hstate *h)
 713{
 714        return huge_page_order(h) >= MAX_ORDER;
 715}
 716
 717static inline unsigned int pages_per_huge_page(struct hstate *h)
 718{
 719        return 1 << h->order;
 720}
 721
 722static inline unsigned int blocks_per_huge_page(struct hstate *h)
 723{
 724        return huge_page_size(h) / 512;
 725}
 726
 727#include <asm/hugetlb.h>
 728
 729#ifndef is_hugepage_only_range
 730static inline int is_hugepage_only_range(struct mm_struct *mm,
 731                                        unsigned long addr, unsigned long len)
 732{
 733        return 0;
 734}
 735#define is_hugepage_only_range is_hugepage_only_range
 736#endif
 737
 738#ifndef arch_clear_hugepage_flags
 739static inline void arch_clear_hugepage_flags(struct page *page) { }
 740#define arch_clear_hugepage_flags arch_clear_hugepage_flags
 741#endif
 742
 743#ifndef arch_make_huge_pte
 744static inline pte_t arch_make_huge_pte(pte_t entry, unsigned int shift,
 745                                       vm_flags_t flags)
 746{
 747        return entry;
 748}
 749#endif
 750
 751static inline struct hstate *page_hstate(struct page *page)
 752{
 753        VM_BUG_ON_PAGE(!PageHuge(page), page);
 754        return size_to_hstate(page_size(page));
 755}
 756
 757static inline unsigned hstate_index_to_shift(unsigned index)
 758{
 759        return hstates[index].order + PAGE_SHIFT;
 760}
 761
 762static inline int hstate_index(struct hstate *h)
 763{
 764        return h - hstates;
 765}
 766
 767extern int dissolve_free_huge_page(struct page *page);
 768extern int dissolve_free_huge_pages(unsigned long start_pfn,
 769                                    unsigned long end_pfn);
 770
 771#ifdef CONFIG_ARCH_ENABLE_HUGEPAGE_MIGRATION
 772#ifndef arch_hugetlb_migration_supported
 773static inline bool arch_hugetlb_migration_supported(struct hstate *h)
 774{
 775        if ((huge_page_shift(h) == PMD_SHIFT) ||
 776                (huge_page_shift(h) == PUD_SHIFT) ||
 777                        (huge_page_shift(h) == PGDIR_SHIFT))
 778                return true;
 779        else
 780                return false;
 781}
 782#endif
 783#else
 784static inline bool arch_hugetlb_migration_supported(struct hstate *h)
 785{
 786        return false;
 787}
 788#endif
 789
 790static inline bool hugepage_migration_supported(struct hstate *h)
 791{
 792        return arch_hugetlb_migration_supported(h);
 793}
 794
 795/*
 796 * Movability check is different as compared to migration check.
 797 * It determines whether or not a huge page should be placed on
 798 * movable zone or not. Movability of any huge page should be
 799 * required only if huge page size is supported for migration.
 800 * There won't be any reason for the huge page to be movable if
 801 * it is not migratable to start with. Also the size of the huge
 802 * page should be large enough to be placed under a movable zone
 803 * and still feasible enough to be migratable. Just the presence
 804 * in movable zone does not make the migration feasible.
 805 *
 806 * So even though large huge page sizes like the gigantic ones
 807 * are migratable they should not be movable because its not
 808 * feasible to migrate them from movable zone.
 809 */
 810static inline bool hugepage_movable_supported(struct hstate *h)
 811{
 812        if (!hugepage_migration_supported(h))
 813                return false;
 814
 815        if (hstate_is_gigantic(h))
 816                return false;
 817        return true;
 818}
 819
 820/* Movability of hugepages depends on migration support. */
 821static inline gfp_t htlb_alloc_mask(struct hstate *h)
 822{
 823        if (hugepage_movable_supported(h))
 824                return GFP_HIGHUSER_MOVABLE;
 825        else
 826                return GFP_HIGHUSER;
 827}
 828
 829static inline gfp_t htlb_modify_alloc_mask(struct hstate *h, gfp_t gfp_mask)
 830{
 831        gfp_t modified_mask = htlb_alloc_mask(h);
 832
 833        /* Some callers might want to enforce node */
 834        modified_mask |= (gfp_mask & __GFP_THISNODE);
 835
 836        modified_mask |= (gfp_mask & __GFP_NOWARN);
 837
 838        return modified_mask;
 839}
 840
 841static inline spinlock_t *huge_pte_lockptr(struct hstate *h,
 842                                           struct mm_struct *mm, pte_t *pte)
 843{
 844        if (huge_page_size(h) == PMD_SIZE)
 845                return pmd_lockptr(mm, (pmd_t *) pte);
 846        VM_BUG_ON(huge_page_size(h) == PAGE_SIZE);
 847        return &mm->page_table_lock;
 848}
 849
 850#ifndef hugepages_supported
 851/*
 852 * Some platform decide whether they support huge pages at boot
 853 * time. Some of them, such as powerpc, set HPAGE_SHIFT to 0
 854 * when there is no such support
 855 */
 856#define hugepages_supported() (HPAGE_SHIFT != 0)
 857#endif
 858
 859void hugetlb_report_usage(struct seq_file *m, struct mm_struct *mm);
 860
 861static inline void hugetlb_count_add(long l, struct mm_struct *mm)
 862{
 863        atomic_long_add(l, &mm->hugetlb_usage);
 864}
 865
 866static inline void hugetlb_count_sub(long l, struct mm_struct *mm)
 867{
 868        atomic_long_sub(l, &mm->hugetlb_usage);
 869}
 870
 871#ifndef set_huge_swap_pte_at
 872static inline void set_huge_swap_pte_at(struct mm_struct *mm, unsigned long addr,
 873                                        pte_t *ptep, pte_t pte, unsigned long sz)
 874{
 875        set_huge_pte_at(mm, addr, ptep, pte);
 876}
 877#endif
 878
 879#ifndef huge_ptep_modify_prot_start
 880#define huge_ptep_modify_prot_start huge_ptep_modify_prot_start
 881static inline pte_t huge_ptep_modify_prot_start(struct vm_area_struct *vma,
 882                                                unsigned long addr, pte_t *ptep)
 883{
 884        return huge_ptep_get_and_clear(vma->vm_mm, addr, ptep);
 885}
 886#endif
 887
 888#ifndef huge_ptep_modify_prot_commit
 889#define huge_ptep_modify_prot_commit huge_ptep_modify_prot_commit
 890static inline void huge_ptep_modify_prot_commit(struct vm_area_struct *vma,
 891                                                unsigned long addr, pte_t *ptep,
 892                                                pte_t old_pte, pte_t pte)
 893{
 894        set_huge_pte_at(vma->vm_mm, addr, ptep, pte);
 895}
 896#endif
 897
 898#else   /* CONFIG_HUGETLB_PAGE */
 899struct hstate {};
 900
 901static inline struct hugepage_subpool *hugetlb_page_subpool(struct page *hpage)
 902{
 903        return NULL;
 904}
 905
 906static inline int isolate_or_dissolve_huge_page(struct page *page,
 907                                                struct list_head *list)
 908{
 909        return -ENOMEM;
 910}
 911
 912static inline struct page *alloc_huge_page(struct vm_area_struct *vma,
 913                                           unsigned long addr,
 914                                           int avoid_reserve)
 915{
 916        return NULL;
 917}
 918
 919static inline struct page *
 920alloc_huge_page_nodemask(struct hstate *h, int preferred_nid,
 921                        nodemask_t *nmask, gfp_t gfp_mask)
 922{
 923        return NULL;
 924}
 925
 926static inline struct page *alloc_huge_page_vma(struct hstate *h,
 927                                               struct vm_area_struct *vma,
 928                                               unsigned long address)
 929{
 930        return NULL;
 931}
 932
 933static inline int __alloc_bootmem_huge_page(struct hstate *h)
 934{
 935        return 0;
 936}
 937
 938static inline struct hstate *hstate_file(struct file *f)
 939{
 940        return NULL;
 941}
 942
 943static inline struct hstate *hstate_sizelog(int page_size_log)
 944{
 945        return NULL;
 946}
 947
 948static inline struct hstate *hstate_vma(struct vm_area_struct *vma)
 949{
 950        return NULL;
 951}
 952
 953static inline struct hstate *page_hstate(struct page *page)
 954{
 955        return NULL;
 956}
 957
 958static inline unsigned long huge_page_size(struct hstate *h)
 959{
 960        return PAGE_SIZE;
 961}
 962
 963static inline unsigned long huge_page_mask(struct hstate *h)
 964{
 965        return PAGE_MASK;
 966}
 967
 968static inline unsigned long vma_kernel_pagesize(struct vm_area_struct *vma)
 969{
 970        return PAGE_SIZE;
 971}
 972
 973static inline unsigned long vma_mmu_pagesize(struct vm_area_struct *vma)
 974{
 975        return PAGE_SIZE;
 976}
 977
 978static inline unsigned int huge_page_order(struct hstate *h)
 979{
 980        return 0;
 981}
 982
 983static inline unsigned int huge_page_shift(struct hstate *h)
 984{
 985        return PAGE_SHIFT;
 986}
 987
 988static inline bool hstate_is_gigantic(struct hstate *h)
 989{
 990        return false;
 991}
 992
 993static inline unsigned int pages_per_huge_page(struct hstate *h)
 994{
 995        return 1;
 996}
 997
 998static inline unsigned hstate_index_to_shift(unsigned index)
 999{
1000        return 0;
1001}
1002
1003static inline int hstate_index(struct hstate *h)
1004{
1005        return 0;
1006}
1007
1008static inline int dissolve_free_huge_page(struct page *page)
1009{
1010        return 0;
1011}
1012
1013static inline int dissolve_free_huge_pages(unsigned long start_pfn,
1014                                           unsigned long end_pfn)
1015{
1016        return 0;
1017}
1018
1019static inline bool hugepage_migration_supported(struct hstate *h)
1020{
1021        return false;
1022}
1023
1024static inline bool hugepage_movable_supported(struct hstate *h)
1025{
1026        return false;
1027}
1028
1029static inline gfp_t htlb_alloc_mask(struct hstate *h)
1030{
1031        return 0;
1032}
1033
1034static inline gfp_t htlb_modify_alloc_mask(struct hstate *h, gfp_t gfp_mask)
1035{
1036        return 0;
1037}
1038
1039static inline spinlock_t *huge_pte_lockptr(struct hstate *h,
1040                                           struct mm_struct *mm, pte_t *pte)
1041{
1042        return &mm->page_table_lock;
1043}
1044
1045static inline void hugetlb_report_usage(struct seq_file *f, struct mm_struct *m)
1046{
1047}
1048
1049static inline void hugetlb_count_sub(long l, struct mm_struct *mm)
1050{
1051}
1052
1053static inline void set_huge_swap_pte_at(struct mm_struct *mm, unsigned long addr,
1054                                        pte_t *ptep, pte_t pte, unsigned long sz)
1055{
1056}
1057#endif  /* CONFIG_HUGETLB_PAGE */
1058
1059#ifdef CONFIG_HUGETLB_PAGE_FREE_VMEMMAP
1060extern bool hugetlb_free_vmemmap_enabled;
1061#else
1062#define hugetlb_free_vmemmap_enabled    false
1063#endif
1064
1065static inline spinlock_t *huge_pte_lock(struct hstate *h,
1066                                        struct mm_struct *mm, pte_t *pte)
1067{
1068        spinlock_t *ptl;
1069
1070        ptl = huge_pte_lockptr(h, mm, pte);
1071        spin_lock(ptl);
1072        return ptl;
1073}
1074
1075#if defined(CONFIG_HUGETLB_PAGE) && defined(CONFIG_CMA)
1076extern void __init hugetlb_cma_reserve(int order);
1077extern void __init hugetlb_cma_check(void);
1078#else
1079static inline __init void hugetlb_cma_reserve(int order)
1080{
1081}
1082static inline __init void hugetlb_cma_check(void)
1083{
1084}
1085#endif
1086
1087bool want_pmd_share(struct vm_area_struct *vma, unsigned long addr);
1088
1089#ifndef __HAVE_ARCH_FLUSH_HUGETLB_TLB_RANGE
1090/*
1091 * ARCHes with special requirements for evicting HUGETLB backing TLB entries can
1092 * implement this.
1093 */
1094#define flush_hugetlb_tlb_range(vma, addr, end) flush_tlb_range(vma, addr, end)
1095#endif
1096
1097#endif /* _LINUX_HUGETLB_H */
1098