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 struct rw_semaphore rw_sema;
64 #ifdef CONFIG_CGROUP_HUGETLB
66 * On private mappings, the counter to uncharge reservations is stored
67 * here. If these fields are 0, then either the mapping is shared, or
68 * cgroup accounting is disabled for this resv_map.
70 struct page_counter *reservation_counter;
71 unsigned long pages_per_hpage;
72 struct cgroup_subsys_state *css;
77 * Region tracking -- allows tracking of reservations and instantiated pages
78 * across the pages in a mapping.
80 * The region data structures are embedded into a resv_map and protected
81 * by a resv_map's lock. The set of regions within the resv_map represent
82 * reservations for huge pages, or huge pages that have already been
83 * instantiated within the map. The from and to elements are huge page
84 * indices into the associated mapping. from indicates the starting index
85 * of the region. to represents the first index past the end of the region.
87 * For example, a file region structure with from == 0 and to == 4 represents
88 * four huge pages in a mapping. It is important to note that the to element
89 * represents the first element past the end of the region. This is used in
90 * arithmetic as 4(to) - 0(from) = 4 huge pages in the region.
92 * Interval notation of the form [from, to) will be used to indicate that
93 * the endpoint from is inclusive and to is exclusive.
96 struct list_head link;
99 #ifdef CONFIG_CGROUP_HUGETLB
101 * On shared mappings, each reserved region appears as a struct
102 * file_region in resv_map. These fields hold the info needed to
103 * uncharge each reservation.
105 struct page_counter *reservation_counter;
106 struct cgroup_subsys_state *css;
110 struct hugetlb_vma_lock {
112 struct rw_semaphore rw_sema;
113 struct vm_area_struct *vma;
116 extern struct resv_map *resv_map_alloc(void);
117 void resv_map_release(struct kref *ref);
119 extern spinlock_t hugetlb_lock;
120 extern int hugetlb_max_hstate __read_mostly;
121 #define for_each_hstate(h) \
122 for ((h) = hstates; (h) < &hstates[hugetlb_max_hstate]; (h)++)
124 struct hugepage_subpool *hugepage_new_subpool(struct hstate *h, long max_hpages,
126 void hugepage_put_subpool(struct hugepage_subpool *spool);
128 void hugetlb_dup_vma_private(struct vm_area_struct *vma);
129 void clear_vma_resv_huge_pages(struct vm_area_struct *vma);
130 int move_hugetlb_page_tables(struct vm_area_struct *vma,
131 struct vm_area_struct *new_vma,
132 unsigned long old_addr, unsigned long new_addr,
134 int copy_hugetlb_page_range(struct mm_struct *, struct mm_struct *,
135 struct vm_area_struct *, struct vm_area_struct *);
136 struct page *hugetlb_follow_page_mask(struct vm_area_struct *vma,
137 unsigned long address, unsigned int flags,
138 unsigned int *page_mask);
139 void unmap_hugepage_range(struct vm_area_struct *,
140 unsigned long, unsigned long, struct page *,
142 void __unmap_hugepage_range(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 hugetlb_fix_reserve_counts(struct inode *inode);
172 extern struct mutex *hugetlb_fault_mutex_table;
173 u32 hugetlb_fault_mutex_hash(struct address_space *mapping, pgoff_t idx);
175 pte_t *huge_pmd_share(struct mm_struct *mm, struct vm_area_struct *vma,
176 unsigned long addr, pud_t *pud);
178 struct address_space *hugetlb_page_mapping_lock_write(struct page *hpage);
180 extern int sysctl_hugetlb_shm_group;
181 extern struct list_head huge_boot_pages;
185 #ifndef CONFIG_HIGHPTE
187 * pte_offset_huge() and pte_alloc_huge() are helpers for those architectures
188 * which may go down to the lowest PTE level in their huge_pte_offset() and
189 * huge_pte_alloc(): to avoid reliance on pte_offset_map() without pte_unmap().
191 static inline pte_t *pte_offset_huge(pmd_t *pmd, unsigned long address)
193 return pte_offset_kernel(pmd, address);
195 static inline pte_t *pte_alloc_huge(struct mm_struct *mm, pmd_t *pmd,
196 unsigned long address)
198 return pte_alloc(mm, pmd) ? NULL : pte_offset_huge(pmd, address);
202 pte_t *huge_pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
203 unsigned long addr, unsigned long sz);
205 * huge_pte_offset(): Walk the hugetlb pgtable until the last level PTE.
206 * Returns the pte_t* if found, or NULL if the address is not mapped.
208 * IMPORTANT: we should normally not directly call this function, instead
209 * this is only a common interface to implement arch-specific
210 * walker. Please use hugetlb_walk() instead, because that will attempt to
211 * verify the locking for you.
213 * Since this function will walk all the pgtable pages (including not only
214 * high-level pgtable page, but also PUD entry that can be unshared
215 * concurrently for VM_SHARED), the caller of this function should be
216 * responsible of its thread safety. One can follow this rule:
218 * (1) For private mappings: pmd unsharing is not possible, so holding the
219 * mmap_lock for either read or write is sufficient. Most callers
220 * already hold the mmap_lock, so normally, no special action is
223 * (2) For shared mappings: pmd unsharing is possible (so the PUD-ranged
224 * pgtable page can go away from under us! It can be done by a pmd
225 * unshare with a follow up munmap() on the other process), then we
228 * (2.1) hugetlb vma lock read or write held, to make sure pmd unshare
229 * won't happen upon the range (it also makes sure the pte_t we
230 * read is the right and stable one), or,
232 * (2.2) hugetlb mapping i_mmap_rwsem lock held read or write, to make
233 * sure even if unshare happened the racy unmap() will wait until
234 * i_mmap_rwsem is released.
236 * Option (2.1) is the safest, which guarantees pte stability from pmd
237 * sharing pov, until the vma lock released. Option (2.2) doesn't protect
238 * a concurrent pmd unshare, but it makes sure the pgtable page is safe to
241 pte_t *huge_pte_offset(struct mm_struct *mm,
242 unsigned long addr, unsigned long sz);
243 unsigned long hugetlb_mask_last_page(struct hstate *h);
244 int huge_pmd_unshare(struct mm_struct *mm, struct vm_area_struct *vma,
245 unsigned long addr, pte_t *ptep);
246 void adjust_range_if_pmd_sharing_possible(struct vm_area_struct *vma,
247 unsigned long *start, unsigned long *end);
249 extern void __hugetlb_zap_begin(struct vm_area_struct *vma,
250 unsigned long *begin, unsigned long *end);
251 extern void __hugetlb_zap_end(struct vm_area_struct *vma,
252 struct zap_details *details);
254 static inline void hugetlb_zap_begin(struct vm_area_struct *vma,
255 unsigned long *start, unsigned long *end)
257 if (is_vm_hugetlb_page(vma))
258 __hugetlb_zap_begin(vma, start, end);
261 static inline void hugetlb_zap_end(struct vm_area_struct *vma,
262 struct zap_details *details)
264 if (is_vm_hugetlb_page(vma))
265 __hugetlb_zap_end(vma, details);
268 void hugetlb_vma_lock_read(struct vm_area_struct *vma);
269 void hugetlb_vma_unlock_read(struct vm_area_struct *vma);
270 void hugetlb_vma_lock_write(struct vm_area_struct *vma);
271 void hugetlb_vma_unlock_write(struct vm_area_struct *vma);
272 int hugetlb_vma_trylock_write(struct vm_area_struct *vma);
273 void hugetlb_vma_assert_locked(struct vm_area_struct *vma);
274 void hugetlb_vma_lock_release(struct kref *kref);
276 int pmd_huge(pmd_t pmd);
277 int pud_huge(pud_t pud);
278 long hugetlb_change_protection(struct vm_area_struct *vma,
279 unsigned long address, unsigned long end, pgprot_t newprot,
280 unsigned long cp_flags);
282 bool is_hugetlb_entry_migration(pte_t pte);
283 void hugetlb_unshare_all_pmds(struct vm_area_struct *vma);
285 #else /* !CONFIG_HUGETLB_PAGE */
287 static inline void hugetlb_dup_vma_private(struct vm_area_struct *vma)
291 static inline void clear_vma_resv_huge_pages(struct vm_area_struct *vma)
295 static inline unsigned long hugetlb_total_pages(void)
300 static inline struct address_space *hugetlb_page_mapping_lock_write(
306 static inline int huge_pmd_unshare(struct mm_struct *mm,
307 struct vm_area_struct *vma,
308 unsigned long addr, pte_t *ptep)
313 static inline void adjust_range_if_pmd_sharing_possible(
314 struct vm_area_struct *vma,
315 unsigned long *start, unsigned long *end)
319 static inline void hugetlb_zap_begin(
320 struct vm_area_struct *vma,
321 unsigned long *start, unsigned long *end)
325 static inline void hugetlb_zap_end(
326 struct vm_area_struct *vma,
327 struct zap_details *details)
331 static inline struct page *hugetlb_follow_page_mask(
332 struct vm_area_struct *vma, unsigned long address, unsigned int flags,
333 unsigned int *page_mask)
335 BUILD_BUG(); /* should never be compiled in if !CONFIG_HUGETLB_PAGE*/
338 static inline int copy_hugetlb_page_range(struct mm_struct *dst,
339 struct mm_struct *src,
340 struct vm_area_struct *dst_vma,
341 struct vm_area_struct *src_vma)
347 static inline int move_hugetlb_page_tables(struct vm_area_struct *vma,
348 struct vm_area_struct *new_vma,
349 unsigned long old_addr,
350 unsigned long new_addr,
357 static inline void hugetlb_report_meminfo(struct seq_file *m)
361 static inline int hugetlb_report_node_meminfo(char *buf, int len, int nid)
366 static inline void hugetlb_show_meminfo_node(int nid)
370 static inline int prepare_hugepage_range(struct file *file,
371 unsigned long addr, unsigned long len)
376 static inline void hugetlb_vma_lock_read(struct vm_area_struct *vma)
380 static inline void hugetlb_vma_unlock_read(struct vm_area_struct *vma)
384 static inline void hugetlb_vma_lock_write(struct vm_area_struct *vma)
388 static inline void hugetlb_vma_unlock_write(struct vm_area_struct *vma)
392 static inline int hugetlb_vma_trylock_write(struct vm_area_struct *vma)
397 static inline void hugetlb_vma_assert_locked(struct vm_area_struct *vma)
401 static inline int pmd_huge(pmd_t pmd)
406 static inline int pud_huge(pud_t pud)
411 static inline int is_hugepage_only_range(struct mm_struct *mm,
412 unsigned long addr, unsigned long len)
417 static inline void hugetlb_free_pgd_range(struct mmu_gather *tlb,
418 unsigned long addr, unsigned long end,
419 unsigned long floor, unsigned long ceiling)
424 #ifdef CONFIG_USERFAULTFD
425 static inline int hugetlb_mfill_atomic_pte(pte_t *dst_pte,
426 struct vm_area_struct *dst_vma,
427 unsigned long dst_addr,
428 unsigned long src_addr,
430 struct folio **foliop)
435 #endif /* CONFIG_USERFAULTFD */
437 static inline pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr,
443 static inline bool isolate_hugetlb(struct folio *folio, struct list_head *list)
448 static inline int get_hwpoison_hugetlb_folio(struct folio *folio, bool *hugetlb, bool unpoison)
453 static inline int get_huge_page_for_hwpoison(unsigned long pfn, int flags,
454 bool *migratable_cleared)
459 static inline void folio_putback_active_hugetlb(struct folio *folio)
463 static inline void move_hugetlb_state(struct folio *old_folio,
464 struct folio *new_folio, int reason)
468 static inline long hugetlb_change_protection(
469 struct vm_area_struct *vma, unsigned long address,
470 unsigned long end, pgprot_t newprot,
471 unsigned long cp_flags)
476 static inline void __unmap_hugepage_range(struct mmu_gather *tlb,
477 struct vm_area_struct *vma, unsigned long start,
478 unsigned long end, struct page *ref_page,
479 zap_flags_t zap_flags)
484 static inline vm_fault_t hugetlb_fault(struct mm_struct *mm,
485 struct vm_area_struct *vma, unsigned long address,
492 static inline void hugetlb_unshare_all_pmds(struct vm_area_struct *vma) { }
494 #endif /* !CONFIG_HUGETLB_PAGE */
496 * hugepages at page global directory. If arch support
497 * hugepages at pgd level, they need to define this.
500 #define pgd_huge(x) 0
503 #define p4d_huge(x) 0
507 static inline int pgd_write(pgd_t pgd)
514 #define HUGETLB_ANON_FILE "anon_hugepage"
518 * The file will be used as an shm file so shmfs accounting rules
521 HUGETLB_SHMFS_INODE = 1,
523 * The file is being created on the internal vfs mount and shmfs
524 * accounting rules do not apply
526 HUGETLB_ANONHUGE_INODE = 2,
529 #ifdef CONFIG_HUGETLBFS
530 struct hugetlbfs_sb_info {
531 long max_inodes; /* inodes allowed */
532 long free_inodes; /* inodes free */
533 spinlock_t stat_lock;
534 struct hstate *hstate;
535 struct hugepage_subpool *spool;
541 static inline struct hugetlbfs_sb_info *HUGETLBFS_SB(struct super_block *sb)
543 return sb->s_fs_info;
546 struct hugetlbfs_inode_info {
547 struct shared_policy policy;
548 struct inode vfs_inode;
552 static inline struct hugetlbfs_inode_info *HUGETLBFS_I(struct inode *inode)
554 return container_of(inode, struct hugetlbfs_inode_info, vfs_inode);
557 extern const struct file_operations hugetlbfs_file_operations;
558 extern const struct vm_operations_struct hugetlb_vm_ops;
559 struct file *hugetlb_file_setup(const char *name, size_t size, vm_flags_t acct,
560 int creat_flags, int page_size_log);
562 static inline bool is_file_hugepages(struct file *file)
564 if (file->f_op == &hugetlbfs_file_operations)
567 return is_file_shm_hugepages(file);
570 static inline struct hstate *hstate_inode(struct inode *i)
572 return HUGETLBFS_SB(i->i_sb)->hstate;
574 #else /* !CONFIG_HUGETLBFS */
576 #define is_file_hugepages(file) false
577 static inline struct file *
578 hugetlb_file_setup(const char *name, size_t size, vm_flags_t acctflag,
579 int creat_flags, int page_size_log)
581 return ERR_PTR(-ENOSYS);
584 static inline struct hstate *hstate_inode(struct inode *i)
588 #endif /* !CONFIG_HUGETLBFS */
590 #ifdef HAVE_ARCH_HUGETLB_UNMAPPED_AREA
591 unsigned long hugetlb_get_unmapped_area(struct file *file, unsigned long addr,
592 unsigned long len, unsigned long pgoff,
593 unsigned long flags);
594 #endif /* HAVE_ARCH_HUGETLB_UNMAPPED_AREA */
597 generic_hugetlb_get_unmapped_area(struct file *file, unsigned long addr,
598 unsigned long len, unsigned long pgoff,
599 unsigned long flags);
602 * huegtlb page specific state flags. These flags are located in page.private
603 * of the hugetlb head page. Functions created via the below macros should be
604 * used to manipulate these flags.
606 * HPG_restore_reserve - Set when a hugetlb page consumes a reservation at
607 * allocation time. Cleared when page is fully instantiated. Free
608 * routine checks flag to restore a reservation on error paths.
609 * Synchronization: Examined or modified by code that knows it has
610 * the only reference to page. i.e. After allocation but before use
611 * or when the page is being freed.
612 * HPG_migratable - Set after a newly allocated page is added to the page
613 * cache and/or page tables. Indicates the page is a candidate for
615 * Synchronization: Initially set after new page allocation with no
616 * locking. When examined and modified during migration processing
617 * (isolate, migrate, putback) the hugetlb_lock is held.
618 * HPG_temporary - Set on a page that is temporarily allocated from the buddy
619 * allocator. Typically used for migration target pages when no pages
620 * are available in the pool. The hugetlb free page path will
621 * immediately free pages with this flag set to the buddy allocator.
622 * Synchronization: Can be set after huge page allocation from buddy when
623 * code knows it has only reference. All other examinations and
624 * modifications require hugetlb_lock.
625 * HPG_freed - Set when page is on the free lists.
626 * Synchronization: hugetlb_lock held for examination and modification.
627 * HPG_vmemmap_optimized - Set when the vmemmap pages of the page are freed.
628 * HPG_raw_hwp_unreliable - Set when the hugetlb page has a hwpoison sub-page
629 * that is not tracked by raw_hwp_page list.
631 enum hugetlb_page_flags {
632 HPG_restore_reserve = 0,
636 HPG_vmemmap_optimized,
637 HPG_raw_hwp_unreliable,
642 * Macros to create test, set and clear function definitions for
643 * hugetlb specific page flags.
645 #ifdef CONFIG_HUGETLB_PAGE
646 #define TESTHPAGEFLAG(uname, flname) \
647 static __always_inline \
648 bool folio_test_hugetlb_##flname(struct folio *folio) \
649 { void *private = &folio->private; \
650 return test_bit(HPG_##flname, private); \
652 static inline int HPage##uname(struct page *page) \
653 { return test_bit(HPG_##flname, &(page->private)); }
655 #define SETHPAGEFLAG(uname, flname) \
656 static __always_inline \
657 void folio_set_hugetlb_##flname(struct folio *folio) \
658 { void *private = &folio->private; \
659 set_bit(HPG_##flname, private); \
661 static inline void SetHPage##uname(struct page *page) \
662 { set_bit(HPG_##flname, &(page->private)); }
664 #define CLEARHPAGEFLAG(uname, flname) \
665 static __always_inline \
666 void folio_clear_hugetlb_##flname(struct folio *folio) \
667 { void *private = &folio->private; \
668 clear_bit(HPG_##flname, private); \
670 static inline void ClearHPage##uname(struct page *page) \
671 { clear_bit(HPG_##flname, &(page->private)); }
673 #define TESTHPAGEFLAG(uname, flname) \
675 folio_test_hugetlb_##flname(struct folio *folio) \
677 static inline int HPage##uname(struct page *page) \
680 #define SETHPAGEFLAG(uname, flname) \
682 folio_set_hugetlb_##flname(struct folio *folio) \
684 static inline void SetHPage##uname(struct page *page) \
687 #define CLEARHPAGEFLAG(uname, flname) \
689 folio_clear_hugetlb_##flname(struct folio *folio) \
691 static inline void ClearHPage##uname(struct page *page) \
695 #define HPAGEFLAG(uname, flname) \
696 TESTHPAGEFLAG(uname, flname) \
697 SETHPAGEFLAG(uname, flname) \
698 CLEARHPAGEFLAG(uname, flname) \
701 * Create functions associated with hugetlb page flags
703 HPAGEFLAG(RestoreReserve, restore_reserve)
704 HPAGEFLAG(Migratable, migratable)
705 HPAGEFLAG(Temporary, temporary)
706 HPAGEFLAG(Freed, freed)
707 HPAGEFLAG(VmemmapOptimized, vmemmap_optimized)
708 HPAGEFLAG(RawHwpUnreliable, raw_hwp_unreliable)
710 #ifdef CONFIG_HUGETLB_PAGE
712 #define HSTATE_NAME_LEN 32
713 /* Defines one hugetlb page size */
715 struct mutex resize_lock;
716 int next_nid_to_alloc;
717 int next_nid_to_free;
719 unsigned int demote_order;
721 unsigned long max_huge_pages;
722 unsigned long nr_huge_pages;
723 unsigned long free_huge_pages;
724 unsigned long resv_huge_pages;
725 unsigned long surplus_huge_pages;
726 unsigned long nr_overcommit_huge_pages;
727 struct list_head hugepage_activelist;
728 struct list_head hugepage_freelists[MAX_NUMNODES];
729 unsigned int max_huge_pages_node[MAX_NUMNODES];
730 unsigned int nr_huge_pages_node[MAX_NUMNODES];
731 unsigned int free_huge_pages_node[MAX_NUMNODES];
732 unsigned int surplus_huge_pages_node[MAX_NUMNODES];
733 #ifdef CONFIG_CGROUP_HUGETLB
734 /* cgroup control files */
735 struct cftype cgroup_files_dfl[8];
736 struct cftype cgroup_files_legacy[10];
738 char name[HSTATE_NAME_LEN];
741 struct huge_bootmem_page {
742 struct list_head list;
743 struct hstate *hstate;
746 int isolate_or_dissolve_huge_page(struct page *page, struct list_head *list);
747 struct folio *alloc_hugetlb_folio(struct vm_area_struct *vma,
748 unsigned long addr, int avoid_reserve);
749 struct folio *alloc_hugetlb_folio_nodemask(struct hstate *h, int preferred_nid,
750 nodemask_t *nmask, gfp_t gfp_mask);
751 struct folio *alloc_hugetlb_folio_vma(struct hstate *h, struct vm_area_struct *vma,
752 unsigned long address);
753 int hugetlb_add_to_page_cache(struct folio *folio, struct address_space *mapping,
755 void restore_reserve_on_error(struct hstate *h, struct vm_area_struct *vma,
756 unsigned long address, struct folio *folio);
759 int __init __alloc_bootmem_huge_page(struct hstate *h, int nid);
760 int __init alloc_bootmem_huge_page(struct hstate *h, int nid);
761 bool __init hugetlb_node_alloc_supported(void);
763 void __init hugetlb_add_hstate(unsigned order);
764 bool __init arch_hugetlb_valid_size(unsigned long size);
765 struct hstate *size_to_hstate(unsigned long size);
767 #ifndef HUGE_MAX_HSTATE
768 #define HUGE_MAX_HSTATE 1
771 extern struct hstate hstates[HUGE_MAX_HSTATE];
772 extern unsigned int default_hstate_idx;
774 #define default_hstate (hstates[default_hstate_idx])
776 static inline struct hugepage_subpool *hugetlb_folio_subpool(struct folio *folio)
778 return folio->_hugetlb_subpool;
781 static inline void hugetlb_set_folio_subpool(struct folio *folio,
782 struct hugepage_subpool *subpool)
784 folio->_hugetlb_subpool = subpool;
787 static inline struct hstate *hstate_file(struct file *f)
789 return hstate_inode(file_inode(f));
792 static inline struct hstate *hstate_sizelog(int page_size_log)
795 return &default_hstate;
797 if (page_size_log < BITS_PER_LONG)
798 return size_to_hstate(1UL << page_size_log);
803 static inline struct hstate *hstate_vma(struct vm_area_struct *vma)
805 return hstate_file(vma->vm_file);
808 static inline unsigned long huge_page_size(const struct hstate *h)
810 return (unsigned long)PAGE_SIZE << h->order;
813 extern unsigned long vma_kernel_pagesize(struct vm_area_struct *vma);
815 extern unsigned long vma_mmu_pagesize(struct vm_area_struct *vma);
817 static inline unsigned long huge_page_mask(struct hstate *h)
822 static inline unsigned int huge_page_order(struct hstate *h)
827 static inline unsigned huge_page_shift(struct hstate *h)
829 return h->order + PAGE_SHIFT;
832 static inline bool hstate_is_gigantic(struct hstate *h)
834 return huge_page_order(h) > MAX_ORDER;
837 static inline unsigned int pages_per_huge_page(const struct hstate *h)
839 return 1 << h->order;
842 static inline unsigned int blocks_per_huge_page(struct hstate *h)
844 return huge_page_size(h) / 512;
847 #include <asm/hugetlb.h>
849 #ifndef is_hugepage_only_range
850 static inline int is_hugepage_only_range(struct mm_struct *mm,
851 unsigned long addr, unsigned long len)
855 #define is_hugepage_only_range is_hugepage_only_range
858 #ifndef arch_clear_hugepage_flags
859 static inline void arch_clear_hugepage_flags(struct page *page) { }
860 #define arch_clear_hugepage_flags arch_clear_hugepage_flags
863 #ifndef arch_make_huge_pte
864 static inline pte_t arch_make_huge_pte(pte_t entry, unsigned int shift,
867 return pte_mkhuge(entry);
871 static inline struct hstate *folio_hstate(struct folio *folio)
873 VM_BUG_ON_FOLIO(!folio_test_hugetlb(folio), folio);
874 return size_to_hstate(folio_size(folio));
877 static inline unsigned hstate_index_to_shift(unsigned index)
879 return hstates[index].order + PAGE_SHIFT;
882 static inline int hstate_index(struct hstate *h)
887 extern int dissolve_free_huge_page(struct page *page);
888 extern int dissolve_free_huge_pages(unsigned long start_pfn,
889 unsigned long end_pfn);
891 #ifdef CONFIG_MEMORY_FAILURE
892 extern void folio_clear_hugetlb_hwpoison(struct folio *folio);
894 static inline void folio_clear_hugetlb_hwpoison(struct folio *folio)
899 #ifdef CONFIG_ARCH_ENABLE_HUGEPAGE_MIGRATION
900 #ifndef arch_hugetlb_migration_supported
901 static inline bool arch_hugetlb_migration_supported(struct hstate *h)
903 if ((huge_page_shift(h) == PMD_SHIFT) ||
904 (huge_page_shift(h) == PUD_SHIFT) ||
905 (huge_page_shift(h) == PGDIR_SHIFT))
912 static inline bool arch_hugetlb_migration_supported(struct hstate *h)
918 static inline bool hugepage_migration_supported(struct hstate *h)
920 return arch_hugetlb_migration_supported(h);
924 * Movability check is different as compared to migration check.
925 * It determines whether or not a huge page should be placed on
926 * movable zone or not. Movability of any huge page should be
927 * required only if huge page size is supported for migration.
928 * There won't be any reason for the huge page to be movable if
929 * it is not migratable to start with. Also the size of the huge
930 * page should be large enough to be placed under a movable zone
931 * and still feasible enough to be migratable. Just the presence
932 * in movable zone does not make the migration feasible.
934 * So even though large huge page sizes like the gigantic ones
935 * are migratable they should not be movable because its not
936 * feasible to migrate them from movable zone.
938 static inline bool hugepage_movable_supported(struct hstate *h)
940 if (!hugepage_migration_supported(h))
943 if (hstate_is_gigantic(h))
948 /* Movability of hugepages depends on migration support. */
949 static inline gfp_t htlb_alloc_mask(struct hstate *h)
951 if (hugepage_movable_supported(h))
952 return GFP_HIGHUSER_MOVABLE;
957 static inline gfp_t htlb_modify_alloc_mask(struct hstate *h, gfp_t gfp_mask)
959 gfp_t modified_mask = htlb_alloc_mask(h);
961 /* Some callers might want to enforce node */
962 modified_mask |= (gfp_mask & __GFP_THISNODE);
964 modified_mask |= (gfp_mask & __GFP_NOWARN);
966 return modified_mask;
969 static inline spinlock_t *huge_pte_lockptr(struct hstate *h,
970 struct mm_struct *mm, pte_t *pte)
972 if (huge_page_size(h) == PMD_SIZE)
973 return pmd_lockptr(mm, (pmd_t *) pte);
974 VM_BUG_ON(huge_page_size(h) == PAGE_SIZE);
975 return &mm->page_table_lock;
978 #ifndef hugepages_supported
980 * Some platform decide whether they support huge pages at boot
981 * time. Some of them, such as powerpc, set HPAGE_SHIFT to 0
982 * when there is no such support
984 #define hugepages_supported() (HPAGE_SHIFT != 0)
987 void hugetlb_report_usage(struct seq_file *m, struct mm_struct *mm);
989 static inline void hugetlb_count_init(struct mm_struct *mm)
991 atomic_long_set(&mm->hugetlb_usage, 0);
994 static inline void hugetlb_count_add(long l, struct mm_struct *mm)
996 atomic_long_add(l, &mm->hugetlb_usage);
999 static inline void hugetlb_count_sub(long l, struct mm_struct *mm)
1001 atomic_long_sub(l, &mm->hugetlb_usage);
1004 #ifndef huge_ptep_modify_prot_start
1005 #define huge_ptep_modify_prot_start huge_ptep_modify_prot_start
1006 static inline pte_t huge_ptep_modify_prot_start(struct vm_area_struct *vma,
1007 unsigned long addr, pte_t *ptep)
1009 return huge_ptep_get_and_clear(vma->vm_mm, addr, ptep);
1013 #ifndef huge_ptep_modify_prot_commit
1014 #define huge_ptep_modify_prot_commit huge_ptep_modify_prot_commit
1015 static inline void huge_ptep_modify_prot_commit(struct vm_area_struct *vma,
1016 unsigned long addr, pte_t *ptep,
1017 pte_t old_pte, pte_t pte)
1019 unsigned long psize = huge_page_size(hstate_vma(vma));
1021 set_huge_pte_at(vma->vm_mm, addr, ptep, pte, psize);
1026 void hugetlb_register_node(struct node *node);
1027 void hugetlb_unregister_node(struct node *node);
1031 * Check if a given raw @page in a hugepage is HWPOISON.
1033 bool is_raw_hwpoison_page_in_hugepage(struct page *page);
1035 #else /* CONFIG_HUGETLB_PAGE */
1038 static inline struct hugepage_subpool *hugetlb_folio_subpool(struct folio *folio)
1043 static inline int isolate_or_dissolve_huge_page(struct page *page,
1044 struct list_head *list)
1049 static inline struct folio *alloc_hugetlb_folio(struct vm_area_struct *vma,
1056 static inline struct folio *
1057 alloc_hugetlb_folio_nodemask(struct hstate *h, int preferred_nid,
1058 nodemask_t *nmask, gfp_t gfp_mask)
1063 static inline struct folio *alloc_hugetlb_folio_vma(struct hstate *h,
1064 struct vm_area_struct *vma,
1065 unsigned long address)
1070 static inline int __alloc_bootmem_huge_page(struct hstate *h)
1075 static inline struct hstate *hstate_file(struct file *f)
1080 static inline struct hstate *hstate_sizelog(int page_size_log)
1085 static inline struct hstate *hstate_vma(struct vm_area_struct *vma)
1090 static inline struct hstate *folio_hstate(struct folio *folio)
1095 static inline struct hstate *size_to_hstate(unsigned long size)
1100 static inline unsigned long huge_page_size(struct hstate *h)
1105 static inline unsigned long huge_page_mask(struct hstate *h)
1110 static inline unsigned long vma_kernel_pagesize(struct vm_area_struct *vma)
1115 static inline unsigned long vma_mmu_pagesize(struct vm_area_struct *vma)
1120 static inline unsigned int huge_page_order(struct hstate *h)
1125 static inline unsigned int huge_page_shift(struct hstate *h)
1130 static inline bool hstate_is_gigantic(struct hstate *h)
1135 static inline unsigned int pages_per_huge_page(struct hstate *h)
1140 static inline unsigned hstate_index_to_shift(unsigned index)
1145 static inline int hstate_index(struct hstate *h)
1150 static inline int dissolve_free_huge_page(struct page *page)
1155 static inline int dissolve_free_huge_pages(unsigned long start_pfn,
1156 unsigned long end_pfn)
1161 static inline bool hugepage_migration_supported(struct hstate *h)
1166 static inline bool hugepage_movable_supported(struct hstate *h)
1171 static inline gfp_t htlb_alloc_mask(struct hstate *h)
1176 static inline gfp_t htlb_modify_alloc_mask(struct hstate *h, gfp_t gfp_mask)
1181 static inline spinlock_t *huge_pte_lockptr(struct hstate *h,
1182 struct mm_struct *mm, pte_t *pte)
1184 return &mm->page_table_lock;
1187 static inline void hugetlb_count_init(struct mm_struct *mm)
1191 static inline void hugetlb_report_usage(struct seq_file *f, struct mm_struct *m)
1195 static inline void hugetlb_count_sub(long l, struct mm_struct *mm)
1199 static inline pte_t huge_ptep_clear_flush(struct vm_area_struct *vma,
1200 unsigned long addr, pte_t *ptep)
1203 return ptep_get(ptep);
1209 static inline void set_huge_pte_at(struct mm_struct *mm, unsigned long addr,
1210 pte_t *ptep, pte_t pte, unsigned long sz)
1214 static inline void hugetlb_register_node(struct node *node)
1218 static inline void hugetlb_unregister_node(struct node *node)
1221 #endif /* CONFIG_HUGETLB_PAGE */
1223 static inline spinlock_t *huge_pte_lock(struct hstate *h,
1224 struct mm_struct *mm, pte_t *pte)
1228 ptl = huge_pte_lockptr(h, mm, pte);
1233 #if defined(CONFIG_HUGETLB_PAGE) && defined(CONFIG_CMA)
1234 extern void __init hugetlb_cma_reserve(int order);
1236 static inline __init void hugetlb_cma_reserve(int order)
1241 #ifdef CONFIG_ARCH_WANT_HUGE_PMD_SHARE
1242 static inline bool hugetlb_pmd_shared(pte_t *pte)
1244 return page_count(virt_to_page(pte)) > 1;
1247 static inline bool hugetlb_pmd_shared(pte_t *pte)
1253 bool want_pmd_share(struct vm_area_struct *vma, unsigned long addr);
1255 #ifndef __HAVE_ARCH_FLUSH_HUGETLB_TLB_RANGE
1257 * ARCHes with special requirements for evicting HUGETLB backing TLB entries can
1260 #define flush_hugetlb_tlb_range(vma, addr, end) flush_tlb_range(vma, addr, end)
1263 static inline bool __vma_shareable_lock(struct vm_area_struct *vma)
1265 return (vma->vm_flags & VM_MAYSHARE) && vma->vm_private_data;
1268 bool __vma_private_lock(struct vm_area_struct *vma);
1271 * Safe version of huge_pte_offset() to check the locks. See comments
1272 * above huge_pte_offset().
1274 static inline pte_t *
1275 hugetlb_walk(struct vm_area_struct *vma, unsigned long addr, unsigned long sz)
1277 #if defined(CONFIG_HUGETLB_PAGE) && \
1278 defined(CONFIG_ARCH_WANT_HUGE_PMD_SHARE) && defined(CONFIG_LOCKDEP)
1279 struct hugetlb_vma_lock *vma_lock = vma->vm_private_data;
1282 * If pmd sharing possible, locking needed to safely walk the
1283 * hugetlb pgtables. More information can be found at the comment
1284 * above huge_pte_offset() in the same file.
1286 * NOTE: lockdep_is_held() is only defined with CONFIG_LOCKDEP.
1288 if (__vma_shareable_lock(vma))
1289 WARN_ON_ONCE(!lockdep_is_held(&vma_lock->rw_sema) &&
1291 &vma->vm_file->f_mapping->i_mmap_rwsem));
1293 return huge_pte_offset(vma->vm_mm, addr, sz);
1296 #endif /* _LINUX_HUGETLB_H */