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 void free_huge_folio(struct folio *folio);
31 #ifdef CONFIG_HUGETLB_PAGE
33 #include <linux/mempolicy.h>
34 #include <linux/shm.h>
35 #include <asm/tlbflush.h>
38 * For HugeTLB page, there are more metadata to save in the struct page. But
39 * the head struct page cannot meet our needs, so we have to abuse other tail
40 * struct page to store the metadata.
42 #define __NR_USED_SUBPAGE 3
44 struct hugepage_subpool {
47 long max_hpages; /* Maximum huge pages or -1 if no maximum. */
48 long used_hpages; /* Used count against maximum, includes */
49 /* both allocated and reserved pages. */
50 struct hstate *hstate;
51 long min_hpages; /* Minimum huge pages or -1 if no minimum. */
52 long rsv_hpages; /* Pages reserved against global pool to */
53 /* satisfy minimum size. */
59 struct list_head regions;
60 long adds_in_progress;
61 struct list_head region_cache;
62 long region_cache_count;
63 #ifdef CONFIG_CGROUP_HUGETLB
65 * On private mappings, the counter to uncharge reservations is stored
66 * here. If these fields are 0, then either the mapping is shared, or
67 * cgroup accounting is disabled for this resv_map.
69 struct page_counter *reservation_counter;
70 unsigned long pages_per_hpage;
71 struct cgroup_subsys_state *css;
76 * Region tracking -- allows tracking of reservations and instantiated pages
77 * across the pages in a mapping.
79 * The region data structures are embedded into a resv_map and protected
80 * by a resv_map's lock. The set of regions within the resv_map represent
81 * reservations for huge pages, or huge pages that have already been
82 * instantiated within the map. The from and to elements are huge page
83 * indices into the associated mapping. from indicates the starting index
84 * of the region. to represents the first index past the end of the region.
86 * For example, a file region structure with from == 0 and to == 4 represents
87 * four huge pages in a mapping. It is important to note that the to element
88 * represents the first element past the end of the region. This is used in
89 * arithmetic as 4(to) - 0(from) = 4 huge pages in the region.
91 * Interval notation of the form [from, to) will be used to indicate that
92 * the endpoint from is inclusive and to is exclusive.
95 struct list_head link;
98 #ifdef CONFIG_CGROUP_HUGETLB
100 * On shared mappings, each reserved region appears as a struct
101 * file_region in resv_map. These fields hold the info needed to
102 * uncharge each reservation.
104 struct page_counter *reservation_counter;
105 struct cgroup_subsys_state *css;
109 struct hugetlb_vma_lock {
111 struct rw_semaphore rw_sema;
112 struct vm_area_struct *vma;
115 extern struct resv_map *resv_map_alloc(void);
116 void resv_map_release(struct kref *ref);
118 extern spinlock_t hugetlb_lock;
119 extern int hugetlb_max_hstate __read_mostly;
120 #define for_each_hstate(h) \
121 for ((h) = hstates; (h) < &hstates[hugetlb_max_hstate]; (h)++)
123 struct hugepage_subpool *hugepage_new_subpool(struct hstate *h, long max_hpages,
125 void hugepage_put_subpool(struct hugepage_subpool *spool);
127 void hugetlb_dup_vma_private(struct vm_area_struct *vma);
128 void clear_vma_resv_huge_pages(struct vm_area_struct *vma);
129 int move_hugetlb_page_tables(struct vm_area_struct *vma,
130 struct vm_area_struct *new_vma,
131 unsigned long old_addr, unsigned long new_addr,
133 int copy_hugetlb_page_range(struct mm_struct *, struct mm_struct *,
134 struct vm_area_struct *, struct vm_area_struct *);
135 struct page *hugetlb_follow_page_mask(struct vm_area_struct *vma,
136 unsigned long address, unsigned int flags,
137 unsigned int *page_mask);
138 void unmap_hugepage_range(struct vm_area_struct *,
139 unsigned long, unsigned long, struct page *,
141 void __unmap_hugepage_range_final(struct mmu_gather *tlb,
142 struct vm_area_struct *vma,
143 unsigned long start, unsigned long end,
144 struct page *ref_page, zap_flags_t zap_flags);
145 void hugetlb_report_meminfo(struct seq_file *);
146 int hugetlb_report_node_meminfo(char *buf, int len, int nid);
147 void hugetlb_show_meminfo_node(int nid);
148 unsigned long hugetlb_total_pages(void);
149 vm_fault_t hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
150 unsigned long address, unsigned int flags);
151 #ifdef CONFIG_USERFAULTFD
152 int hugetlb_mfill_atomic_pte(pte_t *dst_pte,
153 struct vm_area_struct *dst_vma,
154 unsigned long dst_addr,
155 unsigned long src_addr,
157 struct folio **foliop);
158 #endif /* CONFIG_USERFAULTFD */
159 bool hugetlb_reserve_pages(struct inode *inode, long from, long to,
160 struct vm_area_struct *vma,
161 vm_flags_t vm_flags);
162 long hugetlb_unreserve_pages(struct inode *inode, long start, long end,
164 bool isolate_hugetlb(struct folio *folio, struct list_head *list);
165 int get_hwpoison_hugetlb_folio(struct folio *folio, bool *hugetlb, bool unpoison);
166 int get_huge_page_for_hwpoison(unsigned long pfn, int flags,
167 bool *migratable_cleared);
168 void folio_putback_active_hugetlb(struct folio *folio);
169 void move_hugetlb_state(struct folio *old_folio, struct folio *new_folio, int reason);
170 void hugetlb_fix_reserve_counts(struct inode *inode);
171 extern struct mutex *hugetlb_fault_mutex_table;
172 u32 hugetlb_fault_mutex_hash(struct address_space *mapping, pgoff_t idx);
174 pte_t *huge_pmd_share(struct mm_struct *mm, struct vm_area_struct *vma,
175 unsigned long addr, pud_t *pud);
177 struct address_space *hugetlb_page_mapping_lock_write(struct page *hpage);
179 extern int sysctl_hugetlb_shm_group;
180 extern struct list_head huge_boot_pages;
184 #ifndef CONFIG_HIGHPTE
186 * pte_offset_huge() and pte_alloc_huge() are helpers for those architectures
187 * which may go down to the lowest PTE level in their huge_pte_offset() and
188 * huge_pte_alloc(): to avoid reliance on pte_offset_map() without pte_unmap().
190 static inline pte_t *pte_offset_huge(pmd_t *pmd, unsigned long address)
192 return pte_offset_kernel(pmd, address);
194 static inline pte_t *pte_alloc_huge(struct mm_struct *mm, pmd_t *pmd,
195 unsigned long address)
197 return pte_alloc(mm, pmd) ? NULL : pte_offset_huge(pmd, address);
201 pte_t *huge_pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
202 unsigned long addr, unsigned long sz);
204 * huge_pte_offset(): Walk the hugetlb pgtable until the last level PTE.
205 * Returns the pte_t* if found, or NULL if the address is not mapped.
207 * IMPORTANT: we should normally not directly call this function, instead
208 * this is only a common interface to implement arch-specific
209 * walker. Please use hugetlb_walk() instead, because that will attempt to
210 * verify the locking for you.
212 * Since this function will walk all the pgtable pages (including not only
213 * high-level pgtable page, but also PUD entry that can be unshared
214 * concurrently for VM_SHARED), the caller of this function should be
215 * responsible of its thread safety. One can follow this rule:
217 * (1) For private mappings: pmd unsharing is not possible, so holding the
218 * mmap_lock for either read or write is sufficient. Most callers
219 * already hold the mmap_lock, so normally, no special action is
222 * (2) For shared mappings: pmd unsharing is possible (so the PUD-ranged
223 * pgtable page can go away from under us! It can be done by a pmd
224 * unshare with a follow up munmap() on the other process), then we
227 * (2.1) hugetlb vma lock read or write held, to make sure pmd unshare
228 * won't happen upon the range (it also makes sure the pte_t we
229 * read is the right and stable one), or,
231 * (2.2) hugetlb mapping i_mmap_rwsem lock held read or write, to make
232 * sure even if unshare happened the racy unmap() will wait until
233 * i_mmap_rwsem is released.
235 * Option (2.1) is the safest, which guarantees pte stability from pmd
236 * sharing pov, until the vma lock released. Option (2.2) doesn't protect
237 * a concurrent pmd unshare, but it makes sure the pgtable page is safe to
240 pte_t *huge_pte_offset(struct mm_struct *mm,
241 unsigned long addr, unsigned long sz);
242 unsigned long hugetlb_mask_last_page(struct hstate *h);
243 int huge_pmd_unshare(struct mm_struct *mm, struct vm_area_struct *vma,
244 unsigned long addr, pte_t *ptep);
245 void adjust_range_if_pmd_sharing_possible(struct vm_area_struct *vma,
246 unsigned long *start, unsigned long *end);
248 void hugetlb_vma_lock_read(struct vm_area_struct *vma);
249 void hugetlb_vma_unlock_read(struct vm_area_struct *vma);
250 void hugetlb_vma_lock_write(struct vm_area_struct *vma);
251 void hugetlb_vma_unlock_write(struct vm_area_struct *vma);
252 int hugetlb_vma_trylock_write(struct vm_area_struct *vma);
253 void hugetlb_vma_assert_locked(struct vm_area_struct *vma);
254 void hugetlb_vma_lock_release(struct kref *kref);
256 int pmd_huge(pmd_t pmd);
257 int pud_huge(pud_t pud);
258 long hugetlb_change_protection(struct vm_area_struct *vma,
259 unsigned long address, unsigned long end, pgprot_t newprot,
260 unsigned long cp_flags);
262 bool is_hugetlb_entry_migration(pte_t pte);
263 void hugetlb_unshare_all_pmds(struct vm_area_struct *vma);
265 #else /* !CONFIG_HUGETLB_PAGE */
267 static inline void hugetlb_dup_vma_private(struct vm_area_struct *vma)
271 static inline void clear_vma_resv_huge_pages(struct vm_area_struct *vma)
275 static inline unsigned long hugetlb_total_pages(void)
280 static inline struct address_space *hugetlb_page_mapping_lock_write(
286 static inline int huge_pmd_unshare(struct mm_struct *mm,
287 struct vm_area_struct *vma,
288 unsigned long addr, pte_t *ptep)
293 static inline void adjust_range_if_pmd_sharing_possible(
294 struct vm_area_struct *vma,
295 unsigned long *start, unsigned long *end)
299 static inline struct page *hugetlb_follow_page_mask(
300 struct vm_area_struct *vma, unsigned long address, unsigned int flags,
301 unsigned int *page_mask)
303 BUILD_BUG(); /* should never be compiled in if !CONFIG_HUGETLB_PAGE*/
306 static inline int copy_hugetlb_page_range(struct mm_struct *dst,
307 struct mm_struct *src,
308 struct vm_area_struct *dst_vma,
309 struct vm_area_struct *src_vma)
315 static inline int move_hugetlb_page_tables(struct vm_area_struct *vma,
316 struct vm_area_struct *new_vma,
317 unsigned long old_addr,
318 unsigned long new_addr,
325 static inline void hugetlb_report_meminfo(struct seq_file *m)
329 static inline int hugetlb_report_node_meminfo(char *buf, int len, int nid)
334 static inline void hugetlb_show_meminfo_node(int nid)
338 static inline int prepare_hugepage_range(struct file *file,
339 unsigned long addr, unsigned long len)
344 static inline void hugetlb_vma_lock_read(struct vm_area_struct *vma)
348 static inline void hugetlb_vma_unlock_read(struct vm_area_struct *vma)
352 static inline void hugetlb_vma_lock_write(struct vm_area_struct *vma)
356 static inline void hugetlb_vma_unlock_write(struct vm_area_struct *vma)
360 static inline int hugetlb_vma_trylock_write(struct vm_area_struct *vma)
365 static inline void hugetlb_vma_assert_locked(struct vm_area_struct *vma)
369 static inline int pmd_huge(pmd_t pmd)
374 static inline int pud_huge(pud_t pud)
379 static inline int is_hugepage_only_range(struct mm_struct *mm,
380 unsigned long addr, unsigned long len)
385 static inline void hugetlb_free_pgd_range(struct mmu_gather *tlb,
386 unsigned long addr, unsigned long end,
387 unsigned long floor, unsigned long ceiling)
392 #ifdef CONFIG_USERFAULTFD
393 static inline int hugetlb_mfill_atomic_pte(pte_t *dst_pte,
394 struct vm_area_struct *dst_vma,
395 unsigned long dst_addr,
396 unsigned long src_addr,
398 struct folio **foliop)
403 #endif /* CONFIG_USERFAULTFD */
405 static inline pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr,
411 static inline bool isolate_hugetlb(struct folio *folio, struct list_head *list)
416 static inline int get_hwpoison_hugetlb_folio(struct folio *folio, bool *hugetlb, bool unpoison)
421 static inline int get_huge_page_for_hwpoison(unsigned long pfn, int flags,
422 bool *migratable_cleared)
427 static inline void folio_putback_active_hugetlb(struct folio *folio)
431 static inline void move_hugetlb_state(struct folio *old_folio,
432 struct folio *new_folio, int reason)
436 static inline long hugetlb_change_protection(
437 struct vm_area_struct *vma, unsigned long address,
438 unsigned long end, pgprot_t newprot,
439 unsigned long cp_flags)
444 static inline void __unmap_hugepage_range_final(struct mmu_gather *tlb,
445 struct vm_area_struct *vma, unsigned long start,
446 unsigned long end, struct page *ref_page,
447 zap_flags_t zap_flags)
452 static inline vm_fault_t hugetlb_fault(struct mm_struct *mm,
453 struct vm_area_struct *vma, unsigned long address,
460 static inline void hugetlb_unshare_all_pmds(struct vm_area_struct *vma) { }
462 #endif /* !CONFIG_HUGETLB_PAGE */
464 * hugepages at page global directory. If arch support
465 * hugepages at pgd level, they need to define this.
468 #define pgd_huge(x) 0
471 #define p4d_huge(x) 0
475 static inline int pgd_write(pgd_t pgd)
482 #define HUGETLB_ANON_FILE "anon_hugepage"
486 * The file will be used as an shm file so shmfs accounting rules
489 HUGETLB_SHMFS_INODE = 1,
491 * The file is being created on the internal vfs mount and shmfs
492 * accounting rules do not apply
494 HUGETLB_ANONHUGE_INODE = 2,
497 #ifdef CONFIG_HUGETLBFS
498 struct hugetlbfs_sb_info {
499 long max_inodes; /* inodes allowed */
500 long free_inodes; /* inodes free */
501 spinlock_t stat_lock;
502 struct hstate *hstate;
503 struct hugepage_subpool *spool;
509 static inline struct hugetlbfs_sb_info *HUGETLBFS_SB(struct super_block *sb)
511 return sb->s_fs_info;
514 struct hugetlbfs_inode_info {
515 struct shared_policy policy;
516 struct inode vfs_inode;
520 static inline struct hugetlbfs_inode_info *HUGETLBFS_I(struct inode *inode)
522 return container_of(inode, struct hugetlbfs_inode_info, vfs_inode);
525 extern const struct file_operations hugetlbfs_file_operations;
526 extern const struct vm_operations_struct hugetlb_vm_ops;
527 struct file *hugetlb_file_setup(const char *name, size_t size, vm_flags_t acct,
528 int creat_flags, int page_size_log);
530 static inline bool is_file_hugepages(struct file *file)
532 if (file->f_op == &hugetlbfs_file_operations)
535 return is_file_shm_hugepages(file);
538 static inline struct hstate *hstate_inode(struct inode *i)
540 return HUGETLBFS_SB(i->i_sb)->hstate;
542 #else /* !CONFIG_HUGETLBFS */
544 #define is_file_hugepages(file) false
545 static inline struct file *
546 hugetlb_file_setup(const char *name, size_t size, vm_flags_t acctflag,
547 int creat_flags, int page_size_log)
549 return ERR_PTR(-ENOSYS);
552 static inline struct hstate *hstate_inode(struct inode *i)
556 #endif /* !CONFIG_HUGETLBFS */
558 #ifdef HAVE_ARCH_HUGETLB_UNMAPPED_AREA
559 unsigned long hugetlb_get_unmapped_area(struct file *file, unsigned long addr,
560 unsigned long len, unsigned long pgoff,
561 unsigned long flags);
562 #endif /* HAVE_ARCH_HUGETLB_UNMAPPED_AREA */
565 generic_hugetlb_get_unmapped_area(struct file *file, unsigned long addr,
566 unsigned long len, unsigned long pgoff,
567 unsigned long flags);
570 * huegtlb page specific state flags. These flags are located in page.private
571 * of the hugetlb head page. Functions created via the below macros should be
572 * used to manipulate these flags.
574 * HPG_restore_reserve - Set when a hugetlb page consumes a reservation at
575 * allocation time. Cleared when page is fully instantiated. Free
576 * routine checks flag to restore a reservation on error paths.
577 * Synchronization: Examined or modified by code that knows it has
578 * the only reference to page. i.e. After allocation but before use
579 * or when the page is being freed.
580 * HPG_migratable - Set after a newly allocated page is added to the page
581 * cache and/or page tables. Indicates the page is a candidate for
583 * Synchronization: Initially set after new page allocation with no
584 * locking. When examined and modified during migration processing
585 * (isolate, migrate, putback) the hugetlb_lock is held.
586 * HPG_temporary - Set on a page that is temporarily allocated from the buddy
587 * allocator. Typically used for migration target pages when no pages
588 * are available in the pool. The hugetlb free page path will
589 * immediately free pages with this flag set to the buddy allocator.
590 * Synchronization: Can be set after huge page allocation from buddy when
591 * code knows it has only reference. All other examinations and
592 * modifications require hugetlb_lock.
593 * HPG_freed - Set when page is on the free lists.
594 * Synchronization: hugetlb_lock held for examination and modification.
595 * HPG_vmemmap_optimized - Set when the vmemmap pages of the page are freed.
596 * HPG_raw_hwp_unreliable - Set when the hugetlb page has a hwpoison sub-page
597 * that is not tracked by raw_hwp_page list.
599 enum hugetlb_page_flags {
600 HPG_restore_reserve = 0,
604 HPG_vmemmap_optimized,
605 HPG_raw_hwp_unreliable,
610 * Macros to create test, set and clear function definitions for
611 * hugetlb specific page flags.
613 #ifdef CONFIG_HUGETLB_PAGE
614 #define TESTHPAGEFLAG(uname, flname) \
615 static __always_inline \
616 bool folio_test_hugetlb_##flname(struct folio *folio) \
617 { void *private = &folio->private; \
618 return test_bit(HPG_##flname, private); \
620 static inline int HPage##uname(struct page *page) \
621 { return test_bit(HPG_##flname, &(page->private)); }
623 #define SETHPAGEFLAG(uname, flname) \
624 static __always_inline \
625 void folio_set_hugetlb_##flname(struct folio *folio) \
626 { void *private = &folio->private; \
627 set_bit(HPG_##flname, private); \
629 static inline void SetHPage##uname(struct page *page) \
630 { set_bit(HPG_##flname, &(page->private)); }
632 #define CLEARHPAGEFLAG(uname, flname) \
633 static __always_inline \
634 void folio_clear_hugetlb_##flname(struct folio *folio) \
635 { void *private = &folio->private; \
636 clear_bit(HPG_##flname, private); \
638 static inline void ClearHPage##uname(struct page *page) \
639 { clear_bit(HPG_##flname, &(page->private)); }
641 #define TESTHPAGEFLAG(uname, flname) \
643 folio_test_hugetlb_##flname(struct folio *folio) \
645 static inline int HPage##uname(struct page *page) \
648 #define SETHPAGEFLAG(uname, flname) \
650 folio_set_hugetlb_##flname(struct folio *folio) \
652 static inline void SetHPage##uname(struct page *page) \
655 #define CLEARHPAGEFLAG(uname, flname) \
657 folio_clear_hugetlb_##flname(struct folio *folio) \
659 static inline void ClearHPage##uname(struct page *page) \
663 #define HPAGEFLAG(uname, flname) \
664 TESTHPAGEFLAG(uname, flname) \
665 SETHPAGEFLAG(uname, flname) \
666 CLEARHPAGEFLAG(uname, flname) \
669 * Create functions associated with hugetlb page flags
671 HPAGEFLAG(RestoreReserve, restore_reserve)
672 HPAGEFLAG(Migratable, migratable)
673 HPAGEFLAG(Temporary, temporary)
674 HPAGEFLAG(Freed, freed)
675 HPAGEFLAG(VmemmapOptimized, vmemmap_optimized)
676 HPAGEFLAG(RawHwpUnreliable, raw_hwp_unreliable)
678 #ifdef CONFIG_HUGETLB_PAGE
680 #define HSTATE_NAME_LEN 32
681 /* Defines one hugetlb page size */
683 struct mutex resize_lock;
684 int next_nid_to_alloc;
685 int next_nid_to_free;
687 unsigned int demote_order;
689 unsigned long max_huge_pages;
690 unsigned long nr_huge_pages;
691 unsigned long free_huge_pages;
692 unsigned long resv_huge_pages;
693 unsigned long surplus_huge_pages;
694 unsigned long nr_overcommit_huge_pages;
695 struct list_head hugepage_activelist;
696 struct list_head hugepage_freelists[MAX_NUMNODES];
697 unsigned int max_huge_pages_node[MAX_NUMNODES];
698 unsigned int nr_huge_pages_node[MAX_NUMNODES];
699 unsigned int free_huge_pages_node[MAX_NUMNODES];
700 unsigned int surplus_huge_pages_node[MAX_NUMNODES];
701 #ifdef CONFIG_CGROUP_HUGETLB
702 /* cgroup control files */
703 struct cftype cgroup_files_dfl[8];
704 struct cftype cgroup_files_legacy[10];
706 char name[HSTATE_NAME_LEN];
709 struct huge_bootmem_page {
710 struct list_head list;
711 struct hstate *hstate;
714 int isolate_or_dissolve_huge_page(struct page *page, struct list_head *list);
715 struct folio *alloc_hugetlb_folio(struct vm_area_struct *vma,
716 unsigned long addr, int avoid_reserve);
717 struct folio *alloc_hugetlb_folio_nodemask(struct hstate *h, int preferred_nid,
718 nodemask_t *nmask, gfp_t gfp_mask);
719 struct folio *alloc_hugetlb_folio_vma(struct hstate *h, struct vm_area_struct *vma,
720 unsigned long address);
721 int hugetlb_add_to_page_cache(struct folio *folio, struct address_space *mapping,
723 void restore_reserve_on_error(struct hstate *h, struct vm_area_struct *vma,
724 unsigned long address, struct folio *folio);
727 int __init __alloc_bootmem_huge_page(struct hstate *h, int nid);
728 int __init alloc_bootmem_huge_page(struct hstate *h, int nid);
729 bool __init hugetlb_node_alloc_supported(void);
731 void __init hugetlb_add_hstate(unsigned order);
732 bool __init arch_hugetlb_valid_size(unsigned long size);
733 struct hstate *size_to_hstate(unsigned long size);
735 #ifndef HUGE_MAX_HSTATE
736 #define HUGE_MAX_HSTATE 1
739 extern struct hstate hstates[HUGE_MAX_HSTATE];
740 extern unsigned int default_hstate_idx;
742 #define default_hstate (hstates[default_hstate_idx])
744 static inline struct hugepage_subpool *hugetlb_folio_subpool(struct folio *folio)
746 return folio->_hugetlb_subpool;
749 static inline void hugetlb_set_folio_subpool(struct folio *folio,
750 struct hugepage_subpool *subpool)
752 folio->_hugetlb_subpool = subpool;
755 static inline struct hstate *hstate_file(struct file *f)
757 return hstate_inode(file_inode(f));
760 static inline struct hstate *hstate_sizelog(int page_size_log)
763 return &default_hstate;
765 if (page_size_log < BITS_PER_LONG)
766 return size_to_hstate(1UL << page_size_log);
771 static inline struct hstate *hstate_vma(struct vm_area_struct *vma)
773 return hstate_file(vma->vm_file);
776 static inline unsigned long huge_page_size(const struct hstate *h)
778 return (unsigned long)PAGE_SIZE << h->order;
781 extern unsigned long vma_kernel_pagesize(struct vm_area_struct *vma);
783 extern unsigned long vma_mmu_pagesize(struct vm_area_struct *vma);
785 static inline unsigned long huge_page_mask(struct hstate *h)
790 static inline unsigned int huge_page_order(struct hstate *h)
795 static inline unsigned huge_page_shift(struct hstate *h)
797 return h->order + PAGE_SHIFT;
800 static inline bool hstate_is_gigantic(struct hstate *h)
802 return huge_page_order(h) > MAX_ORDER;
805 static inline unsigned int pages_per_huge_page(const struct hstate *h)
807 return 1 << h->order;
810 static inline unsigned int blocks_per_huge_page(struct hstate *h)
812 return huge_page_size(h) / 512;
815 #include <asm/hugetlb.h>
817 #ifndef is_hugepage_only_range
818 static inline int is_hugepage_only_range(struct mm_struct *mm,
819 unsigned long addr, unsigned long len)
823 #define is_hugepage_only_range is_hugepage_only_range
826 #ifndef arch_clear_hugepage_flags
827 static inline void arch_clear_hugepage_flags(struct page *page) { }
828 #define arch_clear_hugepage_flags arch_clear_hugepage_flags
831 #ifndef arch_make_huge_pte
832 static inline pte_t arch_make_huge_pte(pte_t entry, unsigned int shift,
835 return pte_mkhuge(entry);
839 static inline struct hstate *folio_hstate(struct folio *folio)
841 VM_BUG_ON_FOLIO(!folio_test_hugetlb(folio), folio);
842 return size_to_hstate(folio_size(folio));
845 static inline unsigned hstate_index_to_shift(unsigned index)
847 return hstates[index].order + PAGE_SHIFT;
850 static inline int hstate_index(struct hstate *h)
855 extern int dissolve_free_huge_page(struct page *page);
856 extern int dissolve_free_huge_pages(unsigned long start_pfn,
857 unsigned long end_pfn);
859 #ifdef CONFIG_MEMORY_FAILURE
860 extern void folio_clear_hugetlb_hwpoison(struct folio *folio);
862 static inline void folio_clear_hugetlb_hwpoison(struct folio *folio)
867 #ifdef CONFIG_ARCH_ENABLE_HUGEPAGE_MIGRATION
868 #ifndef arch_hugetlb_migration_supported
869 static inline bool arch_hugetlb_migration_supported(struct hstate *h)
871 if ((huge_page_shift(h) == PMD_SHIFT) ||
872 (huge_page_shift(h) == PUD_SHIFT) ||
873 (huge_page_shift(h) == PGDIR_SHIFT))
880 static inline bool arch_hugetlb_migration_supported(struct hstate *h)
886 static inline bool hugepage_migration_supported(struct hstate *h)
888 return arch_hugetlb_migration_supported(h);
892 * Movability check is different as compared to migration check.
893 * It determines whether or not a huge page should be placed on
894 * movable zone or not. Movability of any huge page should be
895 * required only if huge page size is supported for migration.
896 * There won't be any reason for the huge page to be movable if
897 * it is not migratable to start with. Also the size of the huge
898 * page should be large enough to be placed under a movable zone
899 * and still feasible enough to be migratable. Just the presence
900 * in movable zone does not make the migration feasible.
902 * So even though large huge page sizes like the gigantic ones
903 * are migratable they should not be movable because its not
904 * feasible to migrate them from movable zone.
906 static inline bool hugepage_movable_supported(struct hstate *h)
908 if (!hugepage_migration_supported(h))
911 if (hstate_is_gigantic(h))
916 /* Movability of hugepages depends on migration support. */
917 static inline gfp_t htlb_alloc_mask(struct hstate *h)
919 if (hugepage_movable_supported(h))
920 return GFP_HIGHUSER_MOVABLE;
925 static inline gfp_t htlb_modify_alloc_mask(struct hstate *h, gfp_t gfp_mask)
927 gfp_t modified_mask = htlb_alloc_mask(h);
929 /* Some callers might want to enforce node */
930 modified_mask |= (gfp_mask & __GFP_THISNODE);
932 modified_mask |= (gfp_mask & __GFP_NOWARN);
934 return modified_mask;
937 static inline spinlock_t *huge_pte_lockptr(struct hstate *h,
938 struct mm_struct *mm, pte_t *pte)
940 if (huge_page_size(h) == PMD_SIZE)
941 return pmd_lockptr(mm, (pmd_t *) pte);
942 VM_BUG_ON(huge_page_size(h) == PAGE_SIZE);
943 return &mm->page_table_lock;
946 #ifndef hugepages_supported
948 * Some platform decide whether they support huge pages at boot
949 * time. Some of them, such as powerpc, set HPAGE_SHIFT to 0
950 * when there is no such support
952 #define hugepages_supported() (HPAGE_SHIFT != 0)
955 void hugetlb_report_usage(struct seq_file *m, struct mm_struct *mm);
957 static inline void hugetlb_count_init(struct mm_struct *mm)
959 atomic_long_set(&mm->hugetlb_usage, 0);
962 static inline void hugetlb_count_add(long l, struct mm_struct *mm)
964 atomic_long_add(l, &mm->hugetlb_usage);
967 static inline void hugetlb_count_sub(long l, struct mm_struct *mm)
969 atomic_long_sub(l, &mm->hugetlb_usage);
972 #ifndef huge_ptep_modify_prot_start
973 #define huge_ptep_modify_prot_start huge_ptep_modify_prot_start
974 static inline pte_t huge_ptep_modify_prot_start(struct vm_area_struct *vma,
975 unsigned long addr, pte_t *ptep)
977 return huge_ptep_get_and_clear(vma->vm_mm, addr, ptep);
981 #ifndef huge_ptep_modify_prot_commit
982 #define huge_ptep_modify_prot_commit huge_ptep_modify_prot_commit
983 static inline void huge_ptep_modify_prot_commit(struct vm_area_struct *vma,
984 unsigned long addr, pte_t *ptep,
985 pte_t old_pte, pte_t pte)
987 unsigned long psize = huge_page_size(hstate_vma(vma));
989 set_huge_pte_at(vma->vm_mm, addr, ptep, pte, psize);
994 void hugetlb_register_node(struct node *node);
995 void hugetlb_unregister_node(struct node *node);
999 * Check if a given raw @page in a hugepage is HWPOISON.
1001 bool is_raw_hwpoison_page_in_hugepage(struct page *page);
1003 #else /* CONFIG_HUGETLB_PAGE */
1006 static inline struct hugepage_subpool *hugetlb_folio_subpool(struct folio *folio)
1011 static inline int isolate_or_dissolve_huge_page(struct page *page,
1012 struct list_head *list)
1017 static inline struct folio *alloc_hugetlb_folio(struct vm_area_struct *vma,
1024 static inline struct folio *
1025 alloc_hugetlb_folio_nodemask(struct hstate *h, int preferred_nid,
1026 nodemask_t *nmask, gfp_t gfp_mask)
1031 static inline struct folio *alloc_hugetlb_folio_vma(struct hstate *h,
1032 struct vm_area_struct *vma,
1033 unsigned long address)
1038 static inline int __alloc_bootmem_huge_page(struct hstate *h)
1043 static inline struct hstate *hstate_file(struct file *f)
1048 static inline struct hstate *hstate_sizelog(int page_size_log)
1053 static inline struct hstate *hstate_vma(struct vm_area_struct *vma)
1058 static inline struct hstate *folio_hstate(struct folio *folio)
1063 static inline struct hstate *size_to_hstate(unsigned long size)
1068 static inline unsigned long huge_page_size(struct hstate *h)
1073 static inline unsigned long huge_page_mask(struct hstate *h)
1078 static inline unsigned long vma_kernel_pagesize(struct vm_area_struct *vma)
1083 static inline unsigned long vma_mmu_pagesize(struct vm_area_struct *vma)
1088 static inline unsigned int huge_page_order(struct hstate *h)
1093 static inline unsigned int huge_page_shift(struct hstate *h)
1098 static inline bool hstate_is_gigantic(struct hstate *h)
1103 static inline unsigned int pages_per_huge_page(struct hstate *h)
1108 static inline unsigned hstate_index_to_shift(unsigned index)
1113 static inline int hstate_index(struct hstate *h)
1118 static inline int dissolve_free_huge_page(struct page *page)
1123 static inline int dissolve_free_huge_pages(unsigned long start_pfn,
1124 unsigned long end_pfn)
1129 static inline bool hugepage_migration_supported(struct hstate *h)
1134 static inline bool hugepage_movable_supported(struct hstate *h)
1139 static inline gfp_t htlb_alloc_mask(struct hstate *h)
1144 static inline gfp_t htlb_modify_alloc_mask(struct hstate *h, gfp_t gfp_mask)
1149 static inline spinlock_t *huge_pte_lockptr(struct hstate *h,
1150 struct mm_struct *mm, pte_t *pte)
1152 return &mm->page_table_lock;
1155 static inline void hugetlb_count_init(struct mm_struct *mm)
1159 static inline void hugetlb_report_usage(struct seq_file *f, struct mm_struct *m)
1163 static inline void hugetlb_count_sub(long l, struct mm_struct *mm)
1167 static inline pte_t huge_ptep_clear_flush(struct vm_area_struct *vma,
1168 unsigned long addr, pte_t *ptep)
1171 return ptep_get(ptep);
1177 static inline void set_huge_pte_at(struct mm_struct *mm, unsigned long addr,
1178 pte_t *ptep, pte_t pte, unsigned long sz)
1182 static inline void hugetlb_register_node(struct node *node)
1186 static inline void hugetlb_unregister_node(struct node *node)
1189 #endif /* CONFIG_HUGETLB_PAGE */
1191 static inline spinlock_t *huge_pte_lock(struct hstate *h,
1192 struct mm_struct *mm, pte_t *pte)
1196 ptl = huge_pte_lockptr(h, mm, pte);
1201 #if defined(CONFIG_HUGETLB_PAGE) && defined(CONFIG_CMA)
1202 extern void __init hugetlb_cma_reserve(int order);
1204 static inline __init void hugetlb_cma_reserve(int order)
1209 #ifdef CONFIG_ARCH_WANT_HUGE_PMD_SHARE
1210 static inline bool hugetlb_pmd_shared(pte_t *pte)
1212 return page_count(virt_to_page(pte)) > 1;
1215 static inline bool hugetlb_pmd_shared(pte_t *pte)
1221 bool want_pmd_share(struct vm_area_struct *vma, unsigned long addr);
1223 #ifndef __HAVE_ARCH_FLUSH_HUGETLB_TLB_RANGE
1225 * ARCHes with special requirements for evicting HUGETLB backing TLB entries can
1228 #define flush_hugetlb_tlb_range(vma, addr, end) flush_tlb_range(vma, addr, end)
1231 static inline bool __vma_shareable_lock(struct vm_area_struct *vma)
1233 return (vma->vm_flags & VM_MAYSHARE) && vma->vm_private_data;
1237 * Safe version of huge_pte_offset() to check the locks. See comments
1238 * above huge_pte_offset().
1240 static inline pte_t *
1241 hugetlb_walk(struct vm_area_struct *vma, unsigned long addr, unsigned long sz)
1243 #if defined(CONFIG_HUGETLB_PAGE) && \
1244 defined(CONFIG_ARCH_WANT_HUGE_PMD_SHARE) && defined(CONFIG_LOCKDEP)
1245 struct hugetlb_vma_lock *vma_lock = vma->vm_private_data;
1248 * If pmd sharing possible, locking needed to safely walk the
1249 * hugetlb pgtables. More information can be found at the comment
1250 * above huge_pte_offset() in the same file.
1252 * NOTE: lockdep_is_held() is only defined with CONFIG_LOCKDEP.
1254 if (__vma_shareable_lock(vma))
1255 WARN_ON_ONCE(!lockdep_is_held(&vma_lock->rw_sema) &&
1257 &vma->vm_file->f_mapping->i_mmap_rwsem));
1259 return huge_pte_offset(vma->vm_mm, addr, sz);
1262 #endif /* _LINUX_HUGETLB_H */