1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _LINUX_HUGETLB_H
3 #define _LINUX_HUGETLB_H
6 #include <linux/mm_types.h>
7 #include <linux/mmdebug.h>
9 #include <linux/hugetlb_inline.h>
10 #include <linux/cgroup.h>
11 #include <linux/page_ref.h>
12 #include <linux/list.h>
13 #include <linux/kref.h>
14 #include <linux/pgtable.h>
15 #include <linux/gfp.h>
16 #include <linux/userfaultfd_k.h>
23 #ifndef CONFIG_ARCH_HAS_HUGEPD
24 typedef struct { unsigned long pd; } hugepd_t;
25 #define is_hugepd(hugepd) (0)
26 #define __hugepd(x) ((hugepd_t) { (x) })
29 #ifdef CONFIG_HUGETLB_PAGE
31 #include <linux/mempolicy.h>
32 #include <linux/shm.h>
33 #include <asm/tlbflush.h>
36 * For HugeTLB page, there are more metadata to save in the struct page. But
37 * the head struct page cannot meet our needs, so we have to abuse other tail
38 * struct page to store the metadata.
40 #define __NR_USED_SUBPAGE 3
42 struct hugepage_subpool {
45 long max_hpages; /* Maximum huge pages or -1 if no maximum. */
46 long used_hpages; /* Used count against maximum, includes */
47 /* both allocated and reserved pages. */
48 struct hstate *hstate;
49 long min_hpages; /* Minimum huge pages or -1 if no minimum. */
50 long rsv_hpages; /* Pages reserved against global pool to */
51 /* satisfy minimum size. */
57 struct list_head regions;
58 long adds_in_progress;
59 struct list_head region_cache;
60 long region_cache_count;
61 #ifdef CONFIG_CGROUP_HUGETLB
63 * On private mappings, the counter to uncharge reservations is stored
64 * here. If these fields are 0, then either the mapping is shared, or
65 * cgroup accounting is disabled for this resv_map.
67 struct page_counter *reservation_counter;
68 unsigned long pages_per_hpage;
69 struct cgroup_subsys_state *css;
74 * Region tracking -- allows tracking of reservations and instantiated pages
75 * across the pages in a mapping.
77 * The region data structures are embedded into a resv_map and protected
78 * by a resv_map's lock. The set of regions within the resv_map represent
79 * reservations for huge pages, or huge pages that have already been
80 * instantiated within the map. The from and to elements are huge page
81 * indices into the associated mapping. from indicates the starting index
82 * of the region. to represents the first index past the end of the region.
84 * For example, a file region structure with from == 0 and to == 4 represents
85 * four huge pages in a mapping. It is important to note that the to element
86 * represents the first element past the end of the region. This is used in
87 * arithmetic as 4(to) - 0(from) = 4 huge pages in the region.
89 * Interval notation of the form [from, to) will be used to indicate that
90 * the endpoint from is inclusive and to is exclusive.
93 struct list_head link;
96 #ifdef CONFIG_CGROUP_HUGETLB
98 * On shared mappings, each reserved region appears as a struct
99 * file_region in resv_map. These fields hold the info needed to
100 * uncharge each reservation.
102 struct page_counter *reservation_counter;
103 struct cgroup_subsys_state *css;
107 struct hugetlb_vma_lock {
109 struct rw_semaphore rw_sema;
110 struct vm_area_struct *vma;
113 extern struct resv_map *resv_map_alloc(void);
114 void resv_map_release(struct kref *ref);
116 extern spinlock_t hugetlb_lock;
117 extern int hugetlb_max_hstate __read_mostly;
118 #define for_each_hstate(h) \
119 for ((h) = hstates; (h) < &hstates[hugetlb_max_hstate]; (h)++)
121 struct hugepage_subpool *hugepage_new_subpool(struct hstate *h, long max_hpages,
123 void hugepage_put_subpool(struct hugepage_subpool *spool);
125 void hugetlb_dup_vma_private(struct vm_area_struct *vma);
126 void clear_vma_resv_huge_pages(struct vm_area_struct *vma);
127 int move_hugetlb_page_tables(struct vm_area_struct *vma,
128 struct vm_area_struct *new_vma,
129 unsigned long old_addr, unsigned long new_addr,
131 int copy_hugetlb_page_range(struct mm_struct *, struct mm_struct *,
132 struct vm_area_struct *, struct vm_area_struct *);
133 struct page *hugetlb_follow_page_mask(struct vm_area_struct *vma,
134 unsigned long address, unsigned int flags);
135 long follow_hugetlb_page(struct mm_struct *, struct vm_area_struct *,
136 struct page **, struct vm_area_struct **,
137 unsigned long *, unsigned long *, long, unsigned int,
139 void unmap_hugepage_range(struct vm_area_struct *,
140 unsigned long, unsigned long, struct page *,
142 void __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, zap_flags_t zap_flags);
146 void hugetlb_report_meminfo(struct seq_file *);
147 int hugetlb_report_node_meminfo(char *buf, int len, int nid);
148 void hugetlb_show_meminfo_node(int nid);
149 unsigned long hugetlb_total_pages(void);
150 vm_fault_t hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
151 unsigned long address, unsigned int flags);
152 #ifdef CONFIG_USERFAULTFD
153 int hugetlb_mfill_atomic_pte(pte_t *dst_pte,
154 struct vm_area_struct *dst_vma,
155 unsigned long dst_addr,
156 unsigned long src_addr,
158 struct folio **foliop);
159 #endif /* CONFIG_USERFAULTFD */
160 bool hugetlb_reserve_pages(struct inode *inode, long from, long to,
161 struct vm_area_struct *vma,
162 vm_flags_t vm_flags);
163 long hugetlb_unreserve_pages(struct inode *inode, long start, long end,
165 bool isolate_hugetlb(struct folio *folio, struct list_head *list);
166 int get_hwpoison_hugetlb_folio(struct folio *folio, bool *hugetlb, bool unpoison);
167 int get_huge_page_for_hwpoison(unsigned long pfn, int flags,
168 bool *migratable_cleared);
169 void folio_putback_active_hugetlb(struct folio *folio);
170 void move_hugetlb_state(struct folio *old_folio, struct folio *new_folio, int reason);
171 void free_huge_page(struct page *page);
172 void hugetlb_fix_reserve_counts(struct inode *inode);
173 extern struct mutex *hugetlb_fault_mutex_table;
174 u32 hugetlb_fault_mutex_hash(struct address_space *mapping, pgoff_t idx);
176 pte_t *huge_pmd_share(struct mm_struct *mm, struct vm_area_struct *vma,
177 unsigned long addr, pud_t *pud);
179 struct address_space *hugetlb_page_mapping_lock_write(struct page *hpage);
181 extern int sysctl_hugetlb_shm_group;
182 extern struct list_head huge_boot_pages;
186 #ifndef CONFIG_HIGHPTE
188 * pte_offset_huge() and pte_alloc_huge() are helpers for those architectures
189 * which may go down to the lowest PTE level in their huge_pte_offset() and
190 * huge_pte_alloc(): to avoid reliance on pte_offset_map() without pte_unmap().
192 static inline pte_t *pte_offset_huge(pmd_t *pmd, unsigned long address)
194 return pte_offset_kernel(pmd, address);
196 static inline pte_t *pte_alloc_huge(struct mm_struct *mm, pmd_t *pmd,
197 unsigned long address)
199 return pte_alloc(mm, pmd) ? NULL : pte_offset_huge(pmd, address);
203 pte_t *huge_pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
204 unsigned long addr, unsigned long sz);
206 * huge_pte_offset(): Walk the hugetlb pgtable until the last level PTE.
207 * Returns the pte_t* if found, or NULL if the address is not mapped.
209 * IMPORTANT: we should normally not directly call this function, instead
210 * this is only a common interface to implement arch-specific
211 * walker. Please use hugetlb_walk() instead, because that will attempt to
212 * verify the locking for you.
214 * Since this function will walk all the pgtable pages (including not only
215 * high-level pgtable page, but also PUD entry that can be unshared
216 * concurrently for VM_SHARED), the caller of this function should be
217 * responsible of its thread safety. One can follow this rule:
219 * (1) For private mappings: pmd unsharing is not possible, so holding the
220 * mmap_lock for either read or write is sufficient. Most callers
221 * already hold the mmap_lock, so normally, no special action is
224 * (2) For shared mappings: pmd unsharing is possible (so the PUD-ranged
225 * pgtable page can go away from under us! It can be done by a pmd
226 * unshare with a follow up munmap() on the other process), then we
229 * (2.1) hugetlb vma lock read or write held, to make sure pmd unshare
230 * won't happen upon the range (it also makes sure the pte_t we
231 * read is the right and stable one), or,
233 * (2.2) hugetlb mapping i_mmap_rwsem lock held read or write, to make
234 * sure even if unshare happened the racy unmap() will wait until
235 * i_mmap_rwsem is released.
237 * Option (2.1) is the safest, which guarantees pte stability from pmd
238 * sharing pov, until the vma lock released. Option (2.2) doesn't protect
239 * a concurrent pmd unshare, but it makes sure the pgtable page is safe to
242 pte_t *huge_pte_offset(struct mm_struct *mm,
243 unsigned long addr, unsigned long sz);
244 unsigned long hugetlb_mask_last_page(struct hstate *h);
245 int huge_pmd_unshare(struct mm_struct *mm, struct vm_area_struct *vma,
246 unsigned long addr, pte_t *ptep);
247 void adjust_range_if_pmd_sharing_possible(struct vm_area_struct *vma,
248 unsigned long *start, unsigned long *end);
250 void hugetlb_vma_lock_read(struct vm_area_struct *vma);
251 void hugetlb_vma_unlock_read(struct vm_area_struct *vma);
252 void hugetlb_vma_lock_write(struct vm_area_struct *vma);
253 void hugetlb_vma_unlock_write(struct vm_area_struct *vma);
254 int hugetlb_vma_trylock_write(struct vm_area_struct *vma);
255 void hugetlb_vma_assert_locked(struct vm_area_struct *vma);
256 void hugetlb_vma_lock_release(struct kref *kref);
258 int pmd_huge(pmd_t pmd);
259 int pud_huge(pud_t pud);
260 long hugetlb_change_protection(struct vm_area_struct *vma,
261 unsigned long address, unsigned long end, pgprot_t newprot,
262 unsigned long cp_flags);
264 bool is_hugetlb_entry_migration(pte_t pte);
265 void hugetlb_unshare_all_pmds(struct vm_area_struct *vma);
267 #else /* !CONFIG_HUGETLB_PAGE */
269 static inline void hugetlb_dup_vma_private(struct vm_area_struct *vma)
273 static inline void clear_vma_resv_huge_pages(struct vm_area_struct *vma)
277 static inline unsigned long hugetlb_total_pages(void)
282 static inline struct address_space *hugetlb_page_mapping_lock_write(
288 static inline int huge_pmd_unshare(struct mm_struct *mm,
289 struct vm_area_struct *vma,
290 unsigned long addr, pte_t *ptep)
295 static inline void adjust_range_if_pmd_sharing_possible(
296 struct vm_area_struct *vma,
297 unsigned long *start, unsigned long *end)
301 static inline struct page *hugetlb_follow_page_mask(struct vm_area_struct *vma,
302 unsigned long address, unsigned int flags)
304 BUILD_BUG(); /* should never be compiled in if !CONFIG_HUGETLB_PAGE*/
307 static inline long follow_hugetlb_page(struct mm_struct *mm,
308 struct vm_area_struct *vma, struct page **pages,
309 struct vm_area_struct **vmas, unsigned long *position,
310 unsigned long *nr_pages, long i, unsigned int flags,
317 static inline int copy_hugetlb_page_range(struct mm_struct *dst,
318 struct mm_struct *src,
319 struct vm_area_struct *dst_vma,
320 struct vm_area_struct *src_vma)
326 static inline int move_hugetlb_page_tables(struct vm_area_struct *vma,
327 struct vm_area_struct *new_vma,
328 unsigned long old_addr,
329 unsigned long new_addr,
336 static inline void hugetlb_report_meminfo(struct seq_file *m)
340 static inline int hugetlb_report_node_meminfo(char *buf, int len, int nid)
345 static inline void hugetlb_show_meminfo_node(int nid)
349 static inline int prepare_hugepage_range(struct file *file,
350 unsigned long addr, unsigned long len)
355 static inline void hugetlb_vma_lock_read(struct vm_area_struct *vma)
359 static inline void hugetlb_vma_unlock_read(struct vm_area_struct *vma)
363 static inline void hugetlb_vma_lock_write(struct vm_area_struct *vma)
367 static inline void hugetlb_vma_unlock_write(struct vm_area_struct *vma)
371 static inline int hugetlb_vma_trylock_write(struct vm_area_struct *vma)
376 static inline void hugetlb_vma_assert_locked(struct vm_area_struct *vma)
380 static inline int pmd_huge(pmd_t pmd)
385 static inline int pud_huge(pud_t pud)
390 static inline int is_hugepage_only_range(struct mm_struct *mm,
391 unsigned long addr, unsigned long len)
396 static inline void hugetlb_free_pgd_range(struct mmu_gather *tlb,
397 unsigned long addr, unsigned long end,
398 unsigned long floor, unsigned long ceiling)
403 #ifdef CONFIG_USERFAULTFD
404 static inline int hugetlb_mfill_atomic_pte(pte_t *dst_pte,
405 struct vm_area_struct *dst_vma,
406 unsigned long dst_addr,
407 unsigned long src_addr,
409 struct folio **foliop)
414 #endif /* CONFIG_USERFAULTFD */
416 static inline pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr,
422 static inline bool isolate_hugetlb(struct folio *folio, struct list_head *list)
427 static inline int get_hwpoison_hugetlb_folio(struct folio *folio, bool *hugetlb, bool unpoison)
432 static inline int get_huge_page_for_hwpoison(unsigned long pfn, int flags,
433 bool *migratable_cleared)
438 static inline void folio_putback_active_hugetlb(struct folio *folio)
442 static inline void move_hugetlb_state(struct folio *old_folio,
443 struct folio *new_folio, int reason)
447 static inline long hugetlb_change_protection(
448 struct vm_area_struct *vma, unsigned long address,
449 unsigned long end, pgprot_t newprot,
450 unsigned long cp_flags)
455 static inline void __unmap_hugepage_range_final(struct mmu_gather *tlb,
456 struct vm_area_struct *vma, unsigned long start,
457 unsigned long end, struct page *ref_page,
458 zap_flags_t zap_flags)
463 static inline vm_fault_t hugetlb_fault(struct mm_struct *mm,
464 struct vm_area_struct *vma, unsigned long address,
471 static inline void hugetlb_unshare_all_pmds(struct vm_area_struct *vma) { }
473 #endif /* !CONFIG_HUGETLB_PAGE */
475 * hugepages at page global directory. If arch support
476 * hugepages at pgd level, they need to define this.
479 #define pgd_huge(x) 0
482 #define p4d_huge(x) 0
486 static inline int pgd_write(pgd_t pgd)
493 #define HUGETLB_ANON_FILE "anon_hugepage"
497 * The file will be used as an shm file so shmfs accounting rules
500 HUGETLB_SHMFS_INODE = 1,
502 * The file is being created on the internal vfs mount and shmfs
503 * accounting rules do not apply
505 HUGETLB_ANONHUGE_INODE = 2,
508 #ifdef CONFIG_HUGETLBFS
509 struct hugetlbfs_sb_info {
510 long max_inodes; /* inodes allowed */
511 long free_inodes; /* inodes free */
512 spinlock_t stat_lock;
513 struct hstate *hstate;
514 struct hugepage_subpool *spool;
520 static inline struct hugetlbfs_sb_info *HUGETLBFS_SB(struct super_block *sb)
522 return sb->s_fs_info;
525 struct hugetlbfs_inode_info {
526 struct shared_policy policy;
527 struct inode vfs_inode;
531 static inline struct hugetlbfs_inode_info *HUGETLBFS_I(struct inode *inode)
533 return container_of(inode, struct hugetlbfs_inode_info, vfs_inode);
536 extern const struct file_operations hugetlbfs_file_operations;
537 extern const struct vm_operations_struct hugetlb_vm_ops;
538 struct file *hugetlb_file_setup(const char *name, size_t size, vm_flags_t acct,
539 int creat_flags, int page_size_log);
541 static inline bool is_file_hugepages(struct file *file)
543 if (file->f_op == &hugetlbfs_file_operations)
546 return is_file_shm_hugepages(file);
549 static inline struct hstate *hstate_inode(struct inode *i)
551 return HUGETLBFS_SB(i->i_sb)->hstate;
553 #else /* !CONFIG_HUGETLBFS */
555 #define is_file_hugepages(file) false
556 static inline struct file *
557 hugetlb_file_setup(const char *name, size_t size, vm_flags_t acctflag,
558 int creat_flags, int page_size_log)
560 return ERR_PTR(-ENOSYS);
563 static inline struct hstate *hstate_inode(struct inode *i)
567 #endif /* !CONFIG_HUGETLBFS */
569 #ifdef HAVE_ARCH_HUGETLB_UNMAPPED_AREA
570 unsigned long hugetlb_get_unmapped_area(struct file *file, unsigned long addr,
571 unsigned long len, unsigned long pgoff,
572 unsigned long flags);
573 #endif /* HAVE_ARCH_HUGETLB_UNMAPPED_AREA */
576 generic_hugetlb_get_unmapped_area(struct file *file, unsigned long addr,
577 unsigned long len, unsigned long pgoff,
578 unsigned long flags);
581 * huegtlb page specific state flags. These flags are located in page.private
582 * of the hugetlb head page. Functions created via the below macros should be
583 * used to manipulate these flags.
585 * HPG_restore_reserve - Set when a hugetlb page consumes a reservation at
586 * allocation time. Cleared when page is fully instantiated. Free
587 * routine checks flag to restore a reservation on error paths.
588 * Synchronization: Examined or modified by code that knows it has
589 * the only reference to page. i.e. After allocation but before use
590 * or when the page is being freed.
591 * HPG_migratable - Set after a newly allocated page is added to the page
592 * cache and/or page tables. Indicates the page is a candidate for
594 * Synchronization: Initially set after new page allocation with no
595 * locking. When examined and modified during migration processing
596 * (isolate, migrate, putback) the hugetlb_lock is held.
597 * HPG_temporary - Set on a page that is temporarily allocated from the buddy
598 * allocator. Typically used for migration target pages when no pages
599 * are available in the pool. The hugetlb free page path will
600 * immediately free pages with this flag set to the buddy allocator.
601 * Synchronization: Can be set after huge page allocation from buddy when
602 * code knows it has only reference. All other examinations and
603 * modifications require hugetlb_lock.
604 * HPG_freed - Set when page is on the free lists.
605 * Synchronization: hugetlb_lock held for examination and modification.
606 * HPG_vmemmap_optimized - Set when the vmemmap pages of the page are freed.
607 * HPG_raw_hwp_unreliable - Set when the hugetlb page has a hwpoison sub-page
608 * that is not tracked by raw_hwp_page list.
610 enum hugetlb_page_flags {
611 HPG_restore_reserve = 0,
615 HPG_vmemmap_optimized,
616 HPG_raw_hwp_unreliable,
621 * Macros to create test, set and clear function definitions for
622 * hugetlb specific page flags.
624 #ifdef CONFIG_HUGETLB_PAGE
625 #define TESTHPAGEFLAG(uname, flname) \
626 static __always_inline \
627 bool folio_test_hugetlb_##flname(struct folio *folio) \
628 { void *private = &folio->private; \
629 return test_bit(HPG_##flname, private); \
631 static inline int HPage##uname(struct page *page) \
632 { return test_bit(HPG_##flname, &(page->private)); }
634 #define SETHPAGEFLAG(uname, flname) \
635 static __always_inline \
636 void folio_set_hugetlb_##flname(struct folio *folio) \
637 { void *private = &folio->private; \
638 set_bit(HPG_##flname, private); \
640 static inline void SetHPage##uname(struct page *page) \
641 { set_bit(HPG_##flname, &(page->private)); }
643 #define CLEARHPAGEFLAG(uname, flname) \
644 static __always_inline \
645 void folio_clear_hugetlb_##flname(struct folio *folio) \
646 { void *private = &folio->private; \
647 clear_bit(HPG_##flname, private); \
649 static inline void ClearHPage##uname(struct page *page) \
650 { clear_bit(HPG_##flname, &(page->private)); }
652 #define TESTHPAGEFLAG(uname, flname) \
654 folio_test_hugetlb_##flname(struct folio *folio) \
656 static inline int HPage##uname(struct page *page) \
659 #define SETHPAGEFLAG(uname, flname) \
661 folio_set_hugetlb_##flname(struct folio *folio) \
663 static inline void SetHPage##uname(struct page *page) \
666 #define CLEARHPAGEFLAG(uname, flname) \
668 folio_clear_hugetlb_##flname(struct folio *folio) \
670 static inline void ClearHPage##uname(struct page *page) \
674 #define HPAGEFLAG(uname, flname) \
675 TESTHPAGEFLAG(uname, flname) \
676 SETHPAGEFLAG(uname, flname) \
677 CLEARHPAGEFLAG(uname, flname) \
680 * Create functions associated with hugetlb page flags
682 HPAGEFLAG(RestoreReserve, restore_reserve)
683 HPAGEFLAG(Migratable, migratable)
684 HPAGEFLAG(Temporary, temporary)
685 HPAGEFLAG(Freed, freed)
686 HPAGEFLAG(VmemmapOptimized, vmemmap_optimized)
687 HPAGEFLAG(RawHwpUnreliable, raw_hwp_unreliable)
689 #ifdef CONFIG_HUGETLB_PAGE
691 #define HSTATE_NAME_LEN 32
692 /* Defines one hugetlb page size */
694 struct mutex resize_lock;
695 int next_nid_to_alloc;
696 int next_nid_to_free;
698 unsigned int demote_order;
700 unsigned long max_huge_pages;
701 unsigned long nr_huge_pages;
702 unsigned long free_huge_pages;
703 unsigned long resv_huge_pages;
704 unsigned long surplus_huge_pages;
705 unsigned long nr_overcommit_huge_pages;
706 struct list_head hugepage_activelist;
707 struct list_head hugepage_freelists[MAX_NUMNODES];
708 unsigned int max_huge_pages_node[MAX_NUMNODES];
709 unsigned int nr_huge_pages_node[MAX_NUMNODES];
710 unsigned int free_huge_pages_node[MAX_NUMNODES];
711 unsigned int surplus_huge_pages_node[MAX_NUMNODES];
712 #ifdef CONFIG_CGROUP_HUGETLB
713 /* cgroup control files */
714 struct cftype cgroup_files_dfl[8];
715 struct cftype cgroup_files_legacy[10];
717 char name[HSTATE_NAME_LEN];
720 struct huge_bootmem_page {
721 struct list_head list;
722 struct hstate *hstate;
725 int isolate_or_dissolve_huge_page(struct page *page, struct list_head *list);
726 struct folio *alloc_hugetlb_folio(struct vm_area_struct *vma,
727 unsigned long addr, int avoid_reserve);
728 struct folio *alloc_hugetlb_folio_nodemask(struct hstate *h, int preferred_nid,
729 nodemask_t *nmask, gfp_t gfp_mask);
730 struct folio *alloc_hugetlb_folio_vma(struct hstate *h, struct vm_area_struct *vma,
731 unsigned long address);
732 int hugetlb_add_to_page_cache(struct folio *folio, struct address_space *mapping,
734 void restore_reserve_on_error(struct hstate *h, struct vm_area_struct *vma,
735 unsigned long address, struct folio *folio);
738 int __init __alloc_bootmem_huge_page(struct hstate *h, int nid);
739 int __init alloc_bootmem_huge_page(struct hstate *h, int nid);
740 bool __init hugetlb_node_alloc_supported(void);
742 void __init hugetlb_add_hstate(unsigned order);
743 bool __init arch_hugetlb_valid_size(unsigned long size);
744 struct hstate *size_to_hstate(unsigned long size);
746 #ifndef HUGE_MAX_HSTATE
747 #define HUGE_MAX_HSTATE 1
750 extern struct hstate hstates[HUGE_MAX_HSTATE];
751 extern unsigned int default_hstate_idx;
753 #define default_hstate (hstates[default_hstate_idx])
755 static inline struct hugepage_subpool *hugetlb_folio_subpool(struct folio *folio)
757 return folio->_hugetlb_subpool;
761 * hugetlb page subpool pointer located in hpage[2].hugetlb_subpool
763 static inline struct hugepage_subpool *hugetlb_page_subpool(struct page *hpage)
765 return hugetlb_folio_subpool(page_folio(hpage));
768 static inline void hugetlb_set_folio_subpool(struct folio *folio,
769 struct hugepage_subpool *subpool)
771 folio->_hugetlb_subpool = subpool;
774 static inline void hugetlb_set_page_subpool(struct page *hpage,
775 struct hugepage_subpool *subpool)
777 hugetlb_set_folio_subpool(page_folio(hpage), subpool);
780 static inline struct hstate *hstate_file(struct file *f)
782 return hstate_inode(file_inode(f));
785 static inline struct hstate *hstate_sizelog(int page_size_log)
788 return &default_hstate;
790 if (page_size_log < BITS_PER_LONG)
791 return size_to_hstate(1UL << page_size_log);
796 static inline struct hstate *hstate_vma(struct vm_area_struct *vma)
798 return hstate_file(vma->vm_file);
801 static inline unsigned long huge_page_size(const struct hstate *h)
803 return (unsigned long)PAGE_SIZE << h->order;
806 extern unsigned long vma_kernel_pagesize(struct vm_area_struct *vma);
808 extern unsigned long vma_mmu_pagesize(struct vm_area_struct *vma);
810 static inline unsigned long huge_page_mask(struct hstate *h)
815 static inline unsigned int huge_page_order(struct hstate *h)
820 static inline unsigned huge_page_shift(struct hstate *h)
822 return h->order + PAGE_SHIFT;
825 static inline bool hstate_is_gigantic(struct hstate *h)
827 return huge_page_order(h) > MAX_ORDER;
830 static inline unsigned int pages_per_huge_page(const struct hstate *h)
832 return 1 << h->order;
835 static inline unsigned int blocks_per_huge_page(struct hstate *h)
837 return huge_page_size(h) / 512;
840 #include <asm/hugetlb.h>
842 #ifndef is_hugepage_only_range
843 static inline int is_hugepage_only_range(struct mm_struct *mm,
844 unsigned long addr, unsigned long len)
848 #define is_hugepage_only_range is_hugepage_only_range
851 #ifndef arch_clear_hugepage_flags
852 static inline void arch_clear_hugepage_flags(struct page *page) { }
853 #define arch_clear_hugepage_flags arch_clear_hugepage_flags
856 #ifndef arch_make_huge_pte
857 static inline pte_t arch_make_huge_pte(pte_t entry, unsigned int shift,
860 return pte_mkhuge(entry);
864 static inline struct hstate *folio_hstate(struct folio *folio)
866 VM_BUG_ON_FOLIO(!folio_test_hugetlb(folio), folio);
867 return size_to_hstate(folio_size(folio));
870 static inline struct hstate *page_hstate(struct page *page)
872 return folio_hstate(page_folio(page));
875 static inline unsigned hstate_index_to_shift(unsigned index)
877 return hstates[index].order + PAGE_SHIFT;
880 static inline int hstate_index(struct hstate *h)
885 extern int dissolve_free_huge_page(struct page *page);
886 extern int dissolve_free_huge_pages(unsigned long start_pfn,
887 unsigned long end_pfn);
889 #ifdef CONFIG_MEMORY_FAILURE
890 extern void folio_clear_hugetlb_hwpoison(struct folio *folio);
892 static inline void folio_clear_hugetlb_hwpoison(struct folio *folio)
897 #ifdef CONFIG_ARCH_ENABLE_HUGEPAGE_MIGRATION
898 #ifndef arch_hugetlb_migration_supported
899 static inline bool arch_hugetlb_migration_supported(struct hstate *h)
901 if ((huge_page_shift(h) == PMD_SHIFT) ||
902 (huge_page_shift(h) == PUD_SHIFT) ||
903 (huge_page_shift(h) == PGDIR_SHIFT))
910 static inline bool arch_hugetlb_migration_supported(struct hstate *h)
916 static inline bool hugepage_migration_supported(struct hstate *h)
918 return arch_hugetlb_migration_supported(h);
922 * Movability check is different as compared to migration check.
923 * It determines whether or not a huge page should be placed on
924 * movable zone or not. Movability of any huge page should be
925 * required only if huge page size is supported for migration.
926 * There won't be any reason for the huge page to be movable if
927 * it is not migratable to start with. Also the size of the huge
928 * page should be large enough to be placed under a movable zone
929 * and still feasible enough to be migratable. Just the presence
930 * in movable zone does not make the migration feasible.
932 * So even though large huge page sizes like the gigantic ones
933 * are migratable they should not be movable because its not
934 * feasible to migrate them from movable zone.
936 static inline bool hugepage_movable_supported(struct hstate *h)
938 if (!hugepage_migration_supported(h))
941 if (hstate_is_gigantic(h))
946 /* Movability of hugepages depends on migration support. */
947 static inline gfp_t htlb_alloc_mask(struct hstate *h)
949 if (hugepage_movable_supported(h))
950 return GFP_HIGHUSER_MOVABLE;
955 static inline gfp_t htlb_modify_alloc_mask(struct hstate *h, gfp_t gfp_mask)
957 gfp_t modified_mask = htlb_alloc_mask(h);
959 /* Some callers might want to enforce node */
960 modified_mask |= (gfp_mask & __GFP_THISNODE);
962 modified_mask |= (gfp_mask & __GFP_NOWARN);
964 return modified_mask;
967 static inline spinlock_t *huge_pte_lockptr(struct hstate *h,
968 struct mm_struct *mm, pte_t *pte)
970 if (huge_page_size(h) == PMD_SIZE)
971 return pmd_lockptr(mm, (pmd_t *) pte);
972 VM_BUG_ON(huge_page_size(h) == PAGE_SIZE);
973 return &mm->page_table_lock;
976 #ifndef hugepages_supported
978 * Some platform decide whether they support huge pages at boot
979 * time. Some of them, such as powerpc, set HPAGE_SHIFT to 0
980 * when there is no such support
982 #define hugepages_supported() (HPAGE_SHIFT != 0)
985 void hugetlb_report_usage(struct seq_file *m, struct mm_struct *mm);
987 static inline void hugetlb_count_init(struct mm_struct *mm)
989 atomic_long_set(&mm->hugetlb_usage, 0);
992 static inline void hugetlb_count_add(long l, struct mm_struct *mm)
994 atomic_long_add(l, &mm->hugetlb_usage);
997 static inline void hugetlb_count_sub(long l, struct mm_struct *mm)
999 atomic_long_sub(l, &mm->hugetlb_usage);
1002 #ifndef huge_ptep_modify_prot_start
1003 #define huge_ptep_modify_prot_start huge_ptep_modify_prot_start
1004 static inline pte_t huge_ptep_modify_prot_start(struct vm_area_struct *vma,
1005 unsigned long addr, pte_t *ptep)
1007 return huge_ptep_get_and_clear(vma->vm_mm, addr, ptep);
1011 #ifndef huge_ptep_modify_prot_commit
1012 #define huge_ptep_modify_prot_commit huge_ptep_modify_prot_commit
1013 static inline void huge_ptep_modify_prot_commit(struct vm_area_struct *vma,
1014 unsigned long addr, pte_t *ptep,
1015 pte_t old_pte, pte_t pte)
1017 set_huge_pte_at(vma->vm_mm, addr, ptep, pte);
1022 void hugetlb_register_node(struct node *node);
1023 void hugetlb_unregister_node(struct node *node);
1026 #else /* CONFIG_HUGETLB_PAGE */
1029 static inline struct hugepage_subpool *hugetlb_folio_subpool(struct folio *folio)
1034 static inline struct hugepage_subpool *hugetlb_page_subpool(struct page *hpage)
1039 static inline int isolate_or_dissolve_huge_page(struct page *page,
1040 struct list_head *list)
1045 static inline struct folio *alloc_hugetlb_folio(struct vm_area_struct *vma,
1052 static inline struct folio *
1053 alloc_hugetlb_folio_nodemask(struct hstate *h, int preferred_nid,
1054 nodemask_t *nmask, gfp_t gfp_mask)
1059 static inline struct folio *alloc_hugetlb_folio_vma(struct hstate *h,
1060 struct vm_area_struct *vma,
1061 unsigned long address)
1066 static inline int __alloc_bootmem_huge_page(struct hstate *h)
1071 static inline struct hstate *hstate_file(struct file *f)
1076 static inline struct hstate *hstate_sizelog(int page_size_log)
1081 static inline struct hstate *hstate_vma(struct vm_area_struct *vma)
1086 static inline struct hstate *folio_hstate(struct folio *folio)
1091 static inline struct hstate *page_hstate(struct page *page)
1096 static inline struct hstate *size_to_hstate(unsigned long size)
1101 static inline unsigned long huge_page_size(struct hstate *h)
1106 static inline unsigned long huge_page_mask(struct hstate *h)
1111 static inline unsigned long vma_kernel_pagesize(struct vm_area_struct *vma)
1116 static inline unsigned long vma_mmu_pagesize(struct vm_area_struct *vma)
1121 static inline unsigned int huge_page_order(struct hstate *h)
1126 static inline unsigned int huge_page_shift(struct hstate *h)
1131 static inline bool hstate_is_gigantic(struct hstate *h)
1136 static inline unsigned int pages_per_huge_page(struct hstate *h)
1141 static inline unsigned hstate_index_to_shift(unsigned index)
1146 static inline int hstate_index(struct hstate *h)
1151 static inline int dissolve_free_huge_page(struct page *page)
1156 static inline int dissolve_free_huge_pages(unsigned long start_pfn,
1157 unsigned long end_pfn)
1162 static inline bool hugepage_migration_supported(struct hstate *h)
1167 static inline bool hugepage_movable_supported(struct hstate *h)
1172 static inline gfp_t htlb_alloc_mask(struct hstate *h)
1177 static inline gfp_t htlb_modify_alloc_mask(struct hstate *h, gfp_t gfp_mask)
1182 static inline spinlock_t *huge_pte_lockptr(struct hstate *h,
1183 struct mm_struct *mm, pte_t *pte)
1185 return &mm->page_table_lock;
1188 static inline void hugetlb_count_init(struct mm_struct *mm)
1192 static inline void hugetlb_report_usage(struct seq_file *f, struct mm_struct *m)
1196 static inline void hugetlb_count_sub(long l, struct mm_struct *mm)
1200 static inline pte_t huge_ptep_clear_flush(struct vm_area_struct *vma,
1201 unsigned long addr, pte_t *ptep)
1206 static inline void set_huge_pte_at(struct mm_struct *mm, unsigned long addr,
1207 pte_t *ptep, pte_t pte)
1211 static inline void hugetlb_register_node(struct node *node)
1215 static inline void hugetlb_unregister_node(struct node *node)
1218 #endif /* CONFIG_HUGETLB_PAGE */
1220 static inline spinlock_t *huge_pte_lock(struct hstate *h,
1221 struct mm_struct *mm, pte_t *pte)
1225 ptl = huge_pte_lockptr(h, mm, pte);
1230 #if defined(CONFIG_HUGETLB_PAGE) && defined(CONFIG_CMA)
1231 extern void __init hugetlb_cma_reserve(int order);
1233 static inline __init void hugetlb_cma_reserve(int order)
1238 #ifdef CONFIG_ARCH_WANT_HUGE_PMD_SHARE
1239 static inline bool hugetlb_pmd_shared(pte_t *pte)
1241 return page_count(virt_to_page(pte)) > 1;
1244 static inline bool hugetlb_pmd_shared(pte_t *pte)
1250 bool want_pmd_share(struct vm_area_struct *vma, unsigned long addr);
1252 #ifndef __HAVE_ARCH_FLUSH_HUGETLB_TLB_RANGE
1254 * ARCHes with special requirements for evicting HUGETLB backing TLB entries can
1257 #define flush_hugetlb_tlb_range(vma, addr, end) flush_tlb_range(vma, addr, end)
1260 static inline bool __vma_shareable_lock(struct vm_area_struct *vma)
1262 return (vma->vm_flags & VM_MAYSHARE) && vma->vm_private_data;
1266 * Safe version of huge_pte_offset() to check the locks. See comments
1267 * above huge_pte_offset().
1269 static inline pte_t *
1270 hugetlb_walk(struct vm_area_struct *vma, unsigned long addr, unsigned long sz)
1272 #if defined(CONFIG_HUGETLB_PAGE) && \
1273 defined(CONFIG_ARCH_WANT_HUGE_PMD_SHARE) && defined(CONFIG_LOCKDEP)
1274 struct hugetlb_vma_lock *vma_lock = vma->vm_private_data;
1277 * If pmd sharing possible, locking needed to safely walk the
1278 * hugetlb pgtables. More information can be found at the comment
1279 * above huge_pte_offset() in the same file.
1281 * NOTE: lockdep_is_held() is only defined with CONFIG_LOCKDEP.
1283 if (__vma_shareable_lock(vma))
1284 WARN_ON_ONCE(!lockdep_is_held(&vma_lock->rw_sema) &&
1286 &vma->vm_file->f_mapping->i_mmap_rwsem));
1288 return huge_pte_offset(vma->vm_mm, addr, sz);
1291 #endif /* _LINUX_HUGETLB_H */