xref: /linux-6.15/include/linux/hugetlb.h (revision 5ce9231c)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _LINUX_HUGETLB_H
3 #define _LINUX_HUGETLB_H
4 
5 #include <linux/mm_types.h>
6 #include <linux/mmdebug.h>
7 #include <linux/fs.h>
8 #include <linux/hugetlb_inline.h>
9 #include <linux/cgroup.h>
10 #include <linux/list.h>
11 #include <linux/kref.h>
12 #include <linux/pgtable.h>
13 #include <linux/gfp.h>
14 #include <linux/userfaultfd_k.h>
15 
16 struct ctl_table;
17 struct user_struct;
18 struct mmu_gather;
19 
20 #ifndef is_hugepd
21 typedef struct { unsigned long pd; } hugepd_t;
22 #define is_hugepd(hugepd) (0)
23 #define __hugepd(x) ((hugepd_t) { (x) })
24 #endif
25 
26 #ifdef CONFIG_HUGETLB_PAGE
27 
28 #include <linux/mempolicy.h>
29 #include <linux/shm.h>
30 #include <asm/tlbflush.h>
31 
32 /*
33  * For HugeTLB page, there are more metadata to save in the struct page. But
34  * the head struct page cannot meet our needs, so we have to abuse other tail
35  * struct page to store the metadata. In order to avoid conflicts caused by
36  * subsequent use of more tail struct pages, we gather these discrete indexes
37  * of tail struct page here.
38  */
39 enum {
40 	SUBPAGE_INDEX_SUBPOOL = 1,	/* reuse page->private */
41 #ifdef CONFIG_CGROUP_HUGETLB
42 	SUBPAGE_INDEX_CGROUP,		/* reuse page->private */
43 	SUBPAGE_INDEX_CGROUP_RSVD,	/* reuse page->private */
44 	__MAX_CGROUP_SUBPAGE_INDEX = SUBPAGE_INDEX_CGROUP_RSVD,
45 #endif
46 	__NR_USED_SUBPAGE,
47 };
48 
49 struct hugepage_subpool {
50 	spinlock_t lock;
51 	long count;
52 	long max_hpages;	/* Maximum huge pages or -1 if no maximum. */
53 	long used_hpages;	/* Used count against maximum, includes */
54 				/* both allocated and reserved pages. */
55 	struct hstate *hstate;
56 	long min_hpages;	/* Minimum huge pages or -1 if no minimum. */
57 	long rsv_hpages;	/* Pages reserved against global pool to */
58 				/* satisfy minimum size. */
59 };
60 
61 struct resv_map {
62 	struct kref refs;
63 	spinlock_t lock;
64 	struct list_head regions;
65 	long adds_in_progress;
66 	struct list_head region_cache;
67 	long region_cache_count;
68 #ifdef CONFIG_CGROUP_HUGETLB
69 	/*
70 	 * On private mappings, the counter to uncharge reservations is stored
71 	 * here. If these fields are 0, then either the mapping is shared, or
72 	 * cgroup accounting is disabled for this resv_map.
73 	 */
74 	struct page_counter *reservation_counter;
75 	unsigned long pages_per_hpage;
76 	struct cgroup_subsys_state *css;
77 #endif
78 };
79 
80 /*
81  * Region tracking -- allows tracking of reservations and instantiated pages
82  *                    across the pages in a mapping.
83  *
84  * The region data structures are embedded into a resv_map and protected
85  * by a resv_map's lock.  The set of regions within the resv_map represent
86  * reservations for huge pages, or huge pages that have already been
87  * instantiated within the map.  The from and to elements are huge page
88  * indices into the associated mapping.  from indicates the starting index
89  * of the region.  to represents the first index past the end of  the region.
90  *
91  * For example, a file region structure with from == 0 and to == 4 represents
92  * four huge pages in a mapping.  It is important to note that the to element
93  * represents the first element past the end of the region. This is used in
94  * arithmetic as 4(to) - 0(from) = 4 huge pages in the region.
95  *
96  * Interval notation of the form [from, to) will be used to indicate that
97  * the endpoint from is inclusive and to is exclusive.
98  */
99 struct file_region {
100 	struct list_head link;
101 	long from;
102 	long to;
103 #ifdef CONFIG_CGROUP_HUGETLB
104 	/*
105 	 * On shared mappings, each reserved region appears as a struct
106 	 * file_region in resv_map. These fields hold the info needed to
107 	 * uncharge each reservation.
108 	 */
109 	struct page_counter *reservation_counter;
110 	struct cgroup_subsys_state *css;
111 #endif
112 };
113 
114 extern struct resv_map *resv_map_alloc(void);
115 void resv_map_release(struct kref *ref);
116 
117 extern spinlock_t hugetlb_lock;
118 extern int hugetlb_max_hstate __read_mostly;
119 #define for_each_hstate(h) \
120 	for ((h) = hstates; (h) < &hstates[hugetlb_max_hstate]; (h)++)
121 
122 struct hugepage_subpool *hugepage_new_subpool(struct hstate *h, long max_hpages,
123 						long min_hpages);
124 void hugepage_put_subpool(struct hugepage_subpool *spool);
125 
126 void reset_vma_resv_huge_pages(struct vm_area_struct *vma);
127 void clear_vma_resv_huge_pages(struct vm_area_struct *vma);
128 int hugetlb_sysctl_handler(struct ctl_table *, int, void *, size_t *, loff_t *);
129 int hugetlb_overcommit_handler(struct ctl_table *, int, void *, size_t *,
130 		loff_t *);
131 int hugetlb_treat_movable_handler(struct ctl_table *, int, void *, size_t *,
132 		loff_t *);
133 int hugetlb_mempolicy_sysctl_handler(struct ctl_table *, int, void *, size_t *,
134 		loff_t *);
135 
136 int move_hugetlb_page_tables(struct vm_area_struct *vma,
137 			     struct vm_area_struct *new_vma,
138 			     unsigned long old_addr, unsigned long new_addr,
139 			     unsigned long len);
140 int copy_hugetlb_page_range(struct mm_struct *, struct mm_struct *, struct vm_area_struct *);
141 long follow_hugetlb_page(struct mm_struct *, struct vm_area_struct *,
142 			 struct page **, struct vm_area_struct **,
143 			 unsigned long *, unsigned long *, long, unsigned int,
144 			 int *);
145 void unmap_hugepage_range(struct vm_area_struct *,
146 			  unsigned long, unsigned long, struct page *);
147 void __unmap_hugepage_range_final(struct mmu_gather *tlb,
148 			  struct vm_area_struct *vma,
149 			  unsigned long start, unsigned long end,
150 			  struct page *ref_page);
151 void hugetlb_report_meminfo(struct seq_file *);
152 int hugetlb_report_node_meminfo(char *buf, int len, int nid);
153 void hugetlb_show_meminfo(void);
154 unsigned long hugetlb_total_pages(void);
155 vm_fault_t hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
156 			unsigned long address, unsigned int flags);
157 #ifdef CONFIG_USERFAULTFD
158 int hugetlb_mcopy_atomic_pte(struct mm_struct *dst_mm, pte_t *dst_pte,
159 				struct vm_area_struct *dst_vma,
160 				unsigned long dst_addr,
161 				unsigned long src_addr,
162 				enum mcopy_atomic_mode mode,
163 				struct page **pagep);
164 #endif /* CONFIG_USERFAULTFD */
165 bool hugetlb_reserve_pages(struct inode *inode, long from, long to,
166 						struct vm_area_struct *vma,
167 						vm_flags_t vm_flags);
168 long hugetlb_unreserve_pages(struct inode *inode, long start, long end,
169 						long freed);
170 bool isolate_huge_page(struct page *page, struct list_head *list);
171 int get_hwpoison_huge_page(struct page *page, bool *hugetlb);
172 int get_huge_page_for_hwpoison(unsigned long pfn, int flags);
173 void putback_active_hugepage(struct page *page);
174 void move_hugetlb_state(struct page *oldpage, struct page *newpage, int reason);
175 void free_huge_page(struct page *page);
176 void hugetlb_fix_reserve_counts(struct inode *inode);
177 extern struct mutex *hugetlb_fault_mutex_table;
178 u32 hugetlb_fault_mutex_hash(struct address_space *mapping, pgoff_t idx);
179 
180 pte_t *huge_pmd_share(struct mm_struct *mm, struct vm_area_struct *vma,
181 		      unsigned long addr, pud_t *pud);
182 
183 struct address_space *hugetlb_page_mapping_lock_write(struct page *hpage);
184 
185 extern int sysctl_hugetlb_shm_group;
186 extern struct list_head huge_boot_pages;
187 
188 /* arch callbacks */
189 
190 pte_t *huge_pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
191 			unsigned long addr, unsigned long sz);
192 pte_t *huge_pte_offset(struct mm_struct *mm,
193 		       unsigned long addr, unsigned long sz);
194 int huge_pmd_unshare(struct mm_struct *mm, struct vm_area_struct *vma,
195 				unsigned long *addr, pte_t *ptep);
196 void adjust_range_if_pmd_sharing_possible(struct vm_area_struct *vma,
197 				unsigned long *start, unsigned long *end);
198 struct page *follow_huge_addr(struct mm_struct *mm, unsigned long address,
199 			      int write);
200 struct page *follow_huge_pd(struct vm_area_struct *vma,
201 			    unsigned long address, hugepd_t hpd,
202 			    int flags, int pdshift);
203 struct page *follow_huge_pmd(struct mm_struct *mm, unsigned long address,
204 				pmd_t *pmd, int flags);
205 struct page *follow_huge_pud(struct mm_struct *mm, unsigned long address,
206 				pud_t *pud, int flags);
207 struct page *follow_huge_pgd(struct mm_struct *mm, unsigned long address,
208 			     pgd_t *pgd, int flags);
209 
210 int pmd_huge(pmd_t pmd);
211 int pud_huge(pud_t pud);
212 unsigned long hugetlb_change_protection(struct vm_area_struct *vma,
213 		unsigned long address, unsigned long end, pgprot_t newprot);
214 
215 bool is_hugetlb_entry_migration(pte_t pte);
216 void hugetlb_unshare_all_pmds(struct vm_area_struct *vma);
217 
218 #else /* !CONFIG_HUGETLB_PAGE */
219 
220 static inline void reset_vma_resv_huge_pages(struct vm_area_struct *vma)
221 {
222 }
223 
224 static inline void clear_vma_resv_huge_pages(struct vm_area_struct *vma)
225 {
226 }
227 
228 static inline unsigned long hugetlb_total_pages(void)
229 {
230 	return 0;
231 }
232 
233 static inline struct address_space *hugetlb_page_mapping_lock_write(
234 							struct page *hpage)
235 {
236 	return NULL;
237 }
238 
239 static inline int huge_pmd_unshare(struct mm_struct *mm,
240 					struct vm_area_struct *vma,
241 					unsigned long *addr, pte_t *ptep)
242 {
243 	return 0;
244 }
245 
246 static inline void adjust_range_if_pmd_sharing_possible(
247 				struct vm_area_struct *vma,
248 				unsigned long *start, unsigned long *end)
249 {
250 }
251 
252 static inline long follow_hugetlb_page(struct mm_struct *mm,
253 			struct vm_area_struct *vma, struct page **pages,
254 			struct vm_area_struct **vmas, unsigned long *position,
255 			unsigned long *nr_pages, long i, unsigned int flags,
256 			int *nonblocking)
257 {
258 	BUG();
259 	return 0;
260 }
261 
262 static inline struct page *follow_huge_addr(struct mm_struct *mm,
263 					unsigned long address, int write)
264 {
265 	return ERR_PTR(-EINVAL);
266 }
267 
268 static inline int copy_hugetlb_page_range(struct mm_struct *dst,
269 			struct mm_struct *src, struct vm_area_struct *vma)
270 {
271 	BUG();
272 	return 0;
273 }
274 
275 static inline int move_hugetlb_page_tables(struct vm_area_struct *vma,
276 					   struct vm_area_struct *new_vma,
277 					   unsigned long old_addr,
278 					   unsigned long new_addr,
279 					   unsigned long len)
280 {
281 	BUG();
282 	return 0;
283 }
284 
285 static inline void hugetlb_report_meminfo(struct seq_file *m)
286 {
287 }
288 
289 static inline int hugetlb_report_node_meminfo(char *buf, int len, int nid)
290 {
291 	return 0;
292 }
293 
294 static inline void hugetlb_show_meminfo(void)
295 {
296 }
297 
298 static inline struct page *follow_huge_pd(struct vm_area_struct *vma,
299 				unsigned long address, hugepd_t hpd, int flags,
300 				int pdshift)
301 {
302 	return NULL;
303 }
304 
305 static inline struct page *follow_huge_pmd(struct mm_struct *mm,
306 				unsigned long address, pmd_t *pmd, int flags)
307 {
308 	return NULL;
309 }
310 
311 static inline struct page *follow_huge_pud(struct mm_struct *mm,
312 				unsigned long address, pud_t *pud, int flags)
313 {
314 	return NULL;
315 }
316 
317 static inline struct page *follow_huge_pgd(struct mm_struct *mm,
318 				unsigned long address, pgd_t *pgd, int flags)
319 {
320 	return NULL;
321 }
322 
323 static inline int prepare_hugepage_range(struct file *file,
324 				unsigned long addr, unsigned long len)
325 {
326 	return -EINVAL;
327 }
328 
329 static inline int pmd_huge(pmd_t pmd)
330 {
331 	return 0;
332 }
333 
334 static inline int pud_huge(pud_t pud)
335 {
336 	return 0;
337 }
338 
339 static inline int is_hugepage_only_range(struct mm_struct *mm,
340 					unsigned long addr, unsigned long len)
341 {
342 	return 0;
343 }
344 
345 static inline void hugetlb_free_pgd_range(struct mmu_gather *tlb,
346 				unsigned long addr, unsigned long end,
347 				unsigned long floor, unsigned long ceiling)
348 {
349 	BUG();
350 }
351 
352 #ifdef CONFIG_USERFAULTFD
353 static inline int hugetlb_mcopy_atomic_pte(struct mm_struct *dst_mm,
354 						pte_t *dst_pte,
355 						struct vm_area_struct *dst_vma,
356 						unsigned long dst_addr,
357 						unsigned long src_addr,
358 						enum mcopy_atomic_mode mode,
359 						struct page **pagep)
360 {
361 	BUG();
362 	return 0;
363 }
364 #endif /* CONFIG_USERFAULTFD */
365 
366 static inline pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr,
367 					unsigned long sz)
368 {
369 	return NULL;
370 }
371 
372 static inline bool isolate_huge_page(struct page *page, struct list_head *list)
373 {
374 	return false;
375 }
376 
377 static inline int get_hwpoison_huge_page(struct page *page, bool *hugetlb)
378 {
379 	return 0;
380 }
381 
382 static inline int get_huge_page_for_hwpoison(unsigned long pfn, int flags)
383 {
384 	return 0;
385 }
386 
387 static inline void putback_active_hugepage(struct page *page)
388 {
389 }
390 
391 static inline void move_hugetlb_state(struct page *oldpage,
392 					struct page *newpage, int reason)
393 {
394 }
395 
396 static inline unsigned long hugetlb_change_protection(
397 			struct vm_area_struct *vma, unsigned long address,
398 			unsigned long end, pgprot_t newprot)
399 {
400 	return 0;
401 }
402 
403 static inline void __unmap_hugepage_range_final(struct mmu_gather *tlb,
404 			struct vm_area_struct *vma, unsigned long start,
405 			unsigned long end, struct page *ref_page)
406 {
407 	BUG();
408 }
409 
410 static inline vm_fault_t hugetlb_fault(struct mm_struct *mm,
411 			struct vm_area_struct *vma, unsigned long address,
412 			unsigned int flags)
413 {
414 	BUG();
415 	return 0;
416 }
417 
418 static inline void hugetlb_unshare_all_pmds(struct vm_area_struct *vma) { }
419 
420 #endif /* !CONFIG_HUGETLB_PAGE */
421 /*
422  * hugepages at page global directory. If arch support
423  * hugepages at pgd level, they need to define this.
424  */
425 #ifndef pgd_huge
426 #define pgd_huge(x)	0
427 #endif
428 #ifndef p4d_huge
429 #define p4d_huge(x)	0
430 #endif
431 
432 #ifndef pgd_write
433 static inline int pgd_write(pgd_t pgd)
434 {
435 	BUG();
436 	return 0;
437 }
438 #endif
439 
440 #define HUGETLB_ANON_FILE "anon_hugepage"
441 
442 enum {
443 	/*
444 	 * The file will be used as an shm file so shmfs accounting rules
445 	 * apply
446 	 */
447 	HUGETLB_SHMFS_INODE     = 1,
448 	/*
449 	 * The file is being created on the internal vfs mount and shmfs
450 	 * accounting rules do not apply
451 	 */
452 	HUGETLB_ANONHUGE_INODE  = 2,
453 };
454 
455 #ifdef CONFIG_HUGETLBFS
456 struct hugetlbfs_sb_info {
457 	long	max_inodes;   /* inodes allowed */
458 	long	free_inodes;  /* inodes free */
459 	spinlock_t	stat_lock;
460 	struct hstate *hstate;
461 	struct hugepage_subpool *spool;
462 	kuid_t	uid;
463 	kgid_t	gid;
464 	umode_t mode;
465 };
466 
467 static inline struct hugetlbfs_sb_info *HUGETLBFS_SB(struct super_block *sb)
468 {
469 	return sb->s_fs_info;
470 }
471 
472 struct hugetlbfs_inode_info {
473 	struct shared_policy policy;
474 	struct inode vfs_inode;
475 	unsigned int seals;
476 };
477 
478 static inline struct hugetlbfs_inode_info *HUGETLBFS_I(struct inode *inode)
479 {
480 	return container_of(inode, struct hugetlbfs_inode_info, vfs_inode);
481 }
482 
483 extern const struct file_operations hugetlbfs_file_operations;
484 extern const struct vm_operations_struct hugetlb_vm_ops;
485 struct file *hugetlb_file_setup(const char *name, size_t size, vm_flags_t acct,
486 				int creat_flags, int page_size_log);
487 
488 static inline bool is_file_hugepages(struct file *file)
489 {
490 	if (file->f_op == &hugetlbfs_file_operations)
491 		return true;
492 
493 	return is_file_shm_hugepages(file);
494 }
495 
496 static inline struct hstate *hstate_inode(struct inode *i)
497 {
498 	return HUGETLBFS_SB(i->i_sb)->hstate;
499 }
500 #else /* !CONFIG_HUGETLBFS */
501 
502 #define is_file_hugepages(file)			false
503 static inline struct file *
504 hugetlb_file_setup(const char *name, size_t size, vm_flags_t acctflag,
505 		int creat_flags, int page_size_log)
506 {
507 	return ERR_PTR(-ENOSYS);
508 }
509 
510 static inline struct hstate *hstate_inode(struct inode *i)
511 {
512 	return NULL;
513 }
514 #endif /* !CONFIG_HUGETLBFS */
515 
516 #ifdef HAVE_ARCH_HUGETLB_UNMAPPED_AREA
517 unsigned long hugetlb_get_unmapped_area(struct file *file, unsigned long addr,
518 					unsigned long len, unsigned long pgoff,
519 					unsigned long flags);
520 #endif /* HAVE_ARCH_HUGETLB_UNMAPPED_AREA */
521 
522 /*
523  * huegtlb page specific state flags.  These flags are located in page.private
524  * of the hugetlb head page.  Functions created via the below macros should be
525  * used to manipulate these flags.
526  *
527  * HPG_restore_reserve - Set when a hugetlb page consumes a reservation at
528  *	allocation time.  Cleared when page is fully instantiated.  Free
529  *	routine checks flag to restore a reservation on error paths.
530  *	Synchronization:  Examined or modified by code that knows it has
531  *	the only reference to page.  i.e. After allocation but before use
532  *	or when the page is being freed.
533  * HPG_migratable  - Set after a newly allocated page is added to the page
534  *	cache and/or page tables.  Indicates the page is a candidate for
535  *	migration.
536  *	Synchronization:  Initially set after new page allocation with no
537  *	locking.  When examined and modified during migration processing
538  *	(isolate, migrate, putback) the hugetlb_lock is held.
539  * HPG_temporary - - Set on a page that is temporarily allocated from the buddy
540  *	allocator.  Typically used for migration target pages when no pages
541  *	are available in the pool.  The hugetlb free page path will
542  *	immediately free pages with this flag set to the buddy allocator.
543  *	Synchronization: Can be set after huge page allocation from buddy when
544  *	code knows it has only reference.  All other examinations and
545  *	modifications require hugetlb_lock.
546  * HPG_freed - Set when page is on the free lists.
547  *	Synchronization: hugetlb_lock held for examination and modification.
548  * HPG_vmemmap_optimized - Set when the vmemmap pages of the page are freed.
549  */
550 enum hugetlb_page_flags {
551 	HPG_restore_reserve = 0,
552 	HPG_migratable,
553 	HPG_temporary,
554 	HPG_freed,
555 	HPG_vmemmap_optimized,
556 	__NR_HPAGEFLAGS,
557 };
558 
559 /*
560  * Macros to create test, set and clear function definitions for
561  * hugetlb specific page flags.
562  */
563 #ifdef CONFIG_HUGETLB_PAGE
564 #define TESTHPAGEFLAG(uname, flname)				\
565 static inline int HPage##uname(struct page *page)		\
566 	{ return test_bit(HPG_##flname, &(page->private)); }
567 
568 #define SETHPAGEFLAG(uname, flname)				\
569 static inline void SetHPage##uname(struct page *page)		\
570 	{ set_bit(HPG_##flname, &(page->private)); }
571 
572 #define CLEARHPAGEFLAG(uname, flname)				\
573 static inline void ClearHPage##uname(struct page *page)		\
574 	{ clear_bit(HPG_##flname, &(page->private)); }
575 #else
576 #define TESTHPAGEFLAG(uname, flname)				\
577 static inline int HPage##uname(struct page *page)		\
578 	{ return 0; }
579 
580 #define SETHPAGEFLAG(uname, flname)				\
581 static inline void SetHPage##uname(struct page *page)		\
582 	{ }
583 
584 #define CLEARHPAGEFLAG(uname, flname)				\
585 static inline void ClearHPage##uname(struct page *page)		\
586 	{ }
587 #endif
588 
589 #define HPAGEFLAG(uname, flname)				\
590 	TESTHPAGEFLAG(uname, flname)				\
591 	SETHPAGEFLAG(uname, flname)				\
592 	CLEARHPAGEFLAG(uname, flname)				\
593 
594 /*
595  * Create functions associated with hugetlb page flags
596  */
597 HPAGEFLAG(RestoreReserve, restore_reserve)
598 HPAGEFLAG(Migratable, migratable)
599 HPAGEFLAG(Temporary, temporary)
600 HPAGEFLAG(Freed, freed)
601 HPAGEFLAG(VmemmapOptimized, vmemmap_optimized)
602 
603 #ifdef CONFIG_HUGETLB_PAGE
604 
605 #define HSTATE_NAME_LEN 32
606 /* Defines one hugetlb page size */
607 struct hstate {
608 	struct mutex resize_lock;
609 	int next_nid_to_alloc;
610 	int next_nid_to_free;
611 	unsigned int order;
612 	unsigned int demote_order;
613 	unsigned long mask;
614 	unsigned long max_huge_pages;
615 	unsigned long nr_huge_pages;
616 	unsigned long free_huge_pages;
617 	unsigned long resv_huge_pages;
618 	unsigned long surplus_huge_pages;
619 	unsigned long nr_overcommit_huge_pages;
620 	struct list_head hugepage_activelist;
621 	struct list_head hugepage_freelists[MAX_NUMNODES];
622 	unsigned int max_huge_pages_node[MAX_NUMNODES];
623 	unsigned int nr_huge_pages_node[MAX_NUMNODES];
624 	unsigned int free_huge_pages_node[MAX_NUMNODES];
625 	unsigned int surplus_huge_pages_node[MAX_NUMNODES];
626 #ifdef CONFIG_HUGETLB_PAGE_FREE_VMEMMAP
627 	unsigned int nr_free_vmemmap_pages;
628 #endif
629 #ifdef CONFIG_CGROUP_HUGETLB
630 	/* cgroup control files */
631 	struct cftype cgroup_files_dfl[8];
632 	struct cftype cgroup_files_legacy[10];
633 #endif
634 	char name[HSTATE_NAME_LEN];
635 };
636 
637 struct huge_bootmem_page {
638 	struct list_head list;
639 	struct hstate *hstate;
640 };
641 
642 int isolate_or_dissolve_huge_page(struct page *page, struct list_head *list);
643 struct page *alloc_huge_page(struct vm_area_struct *vma,
644 				unsigned long addr, int avoid_reserve);
645 struct page *alloc_huge_page_nodemask(struct hstate *h, int preferred_nid,
646 				nodemask_t *nmask, gfp_t gfp_mask);
647 struct page *alloc_huge_page_vma(struct hstate *h, struct vm_area_struct *vma,
648 				unsigned long address);
649 int huge_add_to_page_cache(struct page *page, struct address_space *mapping,
650 			pgoff_t idx);
651 void restore_reserve_on_error(struct hstate *h, struct vm_area_struct *vma,
652 				unsigned long address, struct page *page);
653 
654 /* arch callback */
655 int __init __alloc_bootmem_huge_page(struct hstate *h, int nid);
656 int __init alloc_bootmem_huge_page(struct hstate *h, int nid);
657 bool __init hugetlb_node_alloc_supported(void);
658 
659 void __init hugetlb_add_hstate(unsigned order);
660 bool __init arch_hugetlb_valid_size(unsigned long size);
661 struct hstate *size_to_hstate(unsigned long size);
662 
663 #ifndef HUGE_MAX_HSTATE
664 #define HUGE_MAX_HSTATE 1
665 #endif
666 
667 extern struct hstate hstates[HUGE_MAX_HSTATE];
668 extern unsigned int default_hstate_idx;
669 
670 #define default_hstate (hstates[default_hstate_idx])
671 
672 /*
673  * hugetlb page subpool pointer located in hpage[1].private
674  */
675 static inline struct hugepage_subpool *hugetlb_page_subpool(struct page *hpage)
676 {
677 	return (void *)page_private(hpage + SUBPAGE_INDEX_SUBPOOL);
678 }
679 
680 static inline void hugetlb_set_page_subpool(struct page *hpage,
681 					struct hugepage_subpool *subpool)
682 {
683 	set_page_private(hpage + SUBPAGE_INDEX_SUBPOOL, (unsigned long)subpool);
684 }
685 
686 static inline struct hstate *hstate_file(struct file *f)
687 {
688 	return hstate_inode(file_inode(f));
689 }
690 
691 static inline struct hstate *hstate_sizelog(int page_size_log)
692 {
693 	if (!page_size_log)
694 		return &default_hstate;
695 
696 	return size_to_hstate(1UL << page_size_log);
697 }
698 
699 static inline struct hstate *hstate_vma(struct vm_area_struct *vma)
700 {
701 	return hstate_file(vma->vm_file);
702 }
703 
704 static inline unsigned long huge_page_size(struct hstate *h)
705 {
706 	return (unsigned long)PAGE_SIZE << h->order;
707 }
708 
709 extern unsigned long vma_kernel_pagesize(struct vm_area_struct *vma);
710 
711 extern unsigned long vma_mmu_pagesize(struct vm_area_struct *vma);
712 
713 static inline unsigned long huge_page_mask(struct hstate *h)
714 {
715 	return h->mask;
716 }
717 
718 static inline unsigned int huge_page_order(struct hstate *h)
719 {
720 	return h->order;
721 }
722 
723 static inline unsigned huge_page_shift(struct hstate *h)
724 {
725 	return h->order + PAGE_SHIFT;
726 }
727 
728 static inline bool hstate_is_gigantic(struct hstate *h)
729 {
730 	return huge_page_order(h) >= MAX_ORDER;
731 }
732 
733 static inline unsigned int pages_per_huge_page(struct hstate *h)
734 {
735 	return 1 << h->order;
736 }
737 
738 static inline unsigned int blocks_per_huge_page(struct hstate *h)
739 {
740 	return huge_page_size(h) / 512;
741 }
742 
743 #include <asm/hugetlb.h>
744 
745 #ifndef is_hugepage_only_range
746 static inline int is_hugepage_only_range(struct mm_struct *mm,
747 					unsigned long addr, unsigned long len)
748 {
749 	return 0;
750 }
751 #define is_hugepage_only_range is_hugepage_only_range
752 #endif
753 
754 #ifndef arch_clear_hugepage_flags
755 static inline void arch_clear_hugepage_flags(struct page *page) { }
756 #define arch_clear_hugepage_flags arch_clear_hugepage_flags
757 #endif
758 
759 #ifndef arch_make_huge_pte
760 static inline pte_t arch_make_huge_pte(pte_t entry, unsigned int shift,
761 				       vm_flags_t flags)
762 {
763 	return pte_mkhuge(entry);
764 }
765 #endif
766 
767 static inline struct hstate *page_hstate(struct page *page)
768 {
769 	VM_BUG_ON_PAGE(!PageHuge(page), page);
770 	return size_to_hstate(page_size(page));
771 }
772 
773 static inline unsigned hstate_index_to_shift(unsigned index)
774 {
775 	return hstates[index].order + PAGE_SHIFT;
776 }
777 
778 static inline int hstate_index(struct hstate *h)
779 {
780 	return h - hstates;
781 }
782 
783 extern int dissolve_free_huge_page(struct page *page);
784 extern int dissolve_free_huge_pages(unsigned long start_pfn,
785 				    unsigned long end_pfn);
786 
787 #ifdef CONFIG_ARCH_ENABLE_HUGEPAGE_MIGRATION
788 #ifndef arch_hugetlb_migration_supported
789 static inline bool arch_hugetlb_migration_supported(struct hstate *h)
790 {
791 	if ((huge_page_shift(h) == PMD_SHIFT) ||
792 		(huge_page_shift(h) == PUD_SHIFT) ||
793 			(huge_page_shift(h) == PGDIR_SHIFT))
794 		return true;
795 	else
796 		return false;
797 }
798 #endif
799 #else
800 static inline bool arch_hugetlb_migration_supported(struct hstate *h)
801 {
802 	return false;
803 }
804 #endif
805 
806 static inline bool hugepage_migration_supported(struct hstate *h)
807 {
808 	return arch_hugetlb_migration_supported(h);
809 }
810 
811 /*
812  * Movability check is different as compared to migration check.
813  * It determines whether or not a huge page should be placed on
814  * movable zone or not. Movability of any huge page should be
815  * required only if huge page size is supported for migration.
816  * There won't be any reason for the huge page to be movable if
817  * it is not migratable to start with. Also the size of the huge
818  * page should be large enough to be placed under a movable zone
819  * and still feasible enough to be migratable. Just the presence
820  * in movable zone does not make the migration feasible.
821  *
822  * So even though large huge page sizes like the gigantic ones
823  * are migratable they should not be movable because its not
824  * feasible to migrate them from movable zone.
825  */
826 static inline bool hugepage_movable_supported(struct hstate *h)
827 {
828 	if (!hugepage_migration_supported(h))
829 		return false;
830 
831 	if (hstate_is_gigantic(h))
832 		return false;
833 	return true;
834 }
835 
836 /* Movability of hugepages depends on migration support. */
837 static inline gfp_t htlb_alloc_mask(struct hstate *h)
838 {
839 	if (hugepage_movable_supported(h))
840 		return GFP_HIGHUSER_MOVABLE;
841 	else
842 		return GFP_HIGHUSER;
843 }
844 
845 static inline gfp_t htlb_modify_alloc_mask(struct hstate *h, gfp_t gfp_mask)
846 {
847 	gfp_t modified_mask = htlb_alloc_mask(h);
848 
849 	/* Some callers might want to enforce node */
850 	modified_mask |= (gfp_mask & __GFP_THISNODE);
851 
852 	modified_mask |= (gfp_mask & __GFP_NOWARN);
853 
854 	return modified_mask;
855 }
856 
857 static inline spinlock_t *huge_pte_lockptr(struct hstate *h,
858 					   struct mm_struct *mm, pte_t *pte)
859 {
860 	if (huge_page_size(h) == PMD_SIZE)
861 		return pmd_lockptr(mm, (pmd_t *) pte);
862 	VM_BUG_ON(huge_page_size(h) == PAGE_SIZE);
863 	return &mm->page_table_lock;
864 }
865 
866 #ifndef hugepages_supported
867 /*
868  * Some platform decide whether they support huge pages at boot
869  * time. Some of them, such as powerpc, set HPAGE_SHIFT to 0
870  * when there is no such support
871  */
872 #define hugepages_supported() (HPAGE_SHIFT != 0)
873 #endif
874 
875 void hugetlb_report_usage(struct seq_file *m, struct mm_struct *mm);
876 
877 static inline void hugetlb_count_init(struct mm_struct *mm)
878 {
879 	atomic_long_set(&mm->hugetlb_usage, 0);
880 }
881 
882 static inline void hugetlb_count_add(long l, struct mm_struct *mm)
883 {
884 	atomic_long_add(l, &mm->hugetlb_usage);
885 }
886 
887 static inline void hugetlb_count_sub(long l, struct mm_struct *mm)
888 {
889 	atomic_long_sub(l, &mm->hugetlb_usage);
890 }
891 
892 #ifndef set_huge_swap_pte_at
893 static inline void set_huge_swap_pte_at(struct mm_struct *mm, unsigned long addr,
894 					pte_t *ptep, pte_t pte, unsigned long sz)
895 {
896 	set_huge_pte_at(mm, addr, ptep, pte);
897 }
898 #endif
899 
900 #ifndef huge_ptep_modify_prot_start
901 #define huge_ptep_modify_prot_start huge_ptep_modify_prot_start
902 static inline pte_t huge_ptep_modify_prot_start(struct vm_area_struct *vma,
903 						unsigned long addr, pte_t *ptep)
904 {
905 	return huge_ptep_get_and_clear(vma->vm_mm, addr, ptep);
906 }
907 #endif
908 
909 #ifndef huge_ptep_modify_prot_commit
910 #define huge_ptep_modify_prot_commit huge_ptep_modify_prot_commit
911 static inline void huge_ptep_modify_prot_commit(struct vm_area_struct *vma,
912 						unsigned long addr, pte_t *ptep,
913 						pte_t old_pte, pte_t pte)
914 {
915 	set_huge_pte_at(vma->vm_mm, addr, ptep, pte);
916 }
917 #endif
918 
919 #else	/* CONFIG_HUGETLB_PAGE */
920 struct hstate {};
921 
922 static inline struct hugepage_subpool *hugetlb_page_subpool(struct page *hpage)
923 {
924 	return NULL;
925 }
926 
927 static inline int isolate_or_dissolve_huge_page(struct page *page,
928 						struct list_head *list)
929 {
930 	return -ENOMEM;
931 }
932 
933 static inline struct page *alloc_huge_page(struct vm_area_struct *vma,
934 					   unsigned long addr,
935 					   int avoid_reserve)
936 {
937 	return NULL;
938 }
939 
940 static inline struct page *
941 alloc_huge_page_nodemask(struct hstate *h, int preferred_nid,
942 			nodemask_t *nmask, gfp_t gfp_mask)
943 {
944 	return NULL;
945 }
946 
947 static inline struct page *alloc_huge_page_vma(struct hstate *h,
948 					       struct vm_area_struct *vma,
949 					       unsigned long address)
950 {
951 	return NULL;
952 }
953 
954 static inline int __alloc_bootmem_huge_page(struct hstate *h)
955 {
956 	return 0;
957 }
958 
959 static inline struct hstate *hstate_file(struct file *f)
960 {
961 	return NULL;
962 }
963 
964 static inline struct hstate *hstate_sizelog(int page_size_log)
965 {
966 	return NULL;
967 }
968 
969 static inline struct hstate *hstate_vma(struct vm_area_struct *vma)
970 {
971 	return NULL;
972 }
973 
974 static inline struct hstate *page_hstate(struct page *page)
975 {
976 	return NULL;
977 }
978 
979 static inline struct hstate *size_to_hstate(unsigned long size)
980 {
981 	return NULL;
982 }
983 
984 static inline unsigned long huge_page_size(struct hstate *h)
985 {
986 	return PAGE_SIZE;
987 }
988 
989 static inline unsigned long huge_page_mask(struct hstate *h)
990 {
991 	return PAGE_MASK;
992 }
993 
994 static inline unsigned long vma_kernel_pagesize(struct vm_area_struct *vma)
995 {
996 	return PAGE_SIZE;
997 }
998 
999 static inline unsigned long vma_mmu_pagesize(struct vm_area_struct *vma)
1000 {
1001 	return PAGE_SIZE;
1002 }
1003 
1004 static inline unsigned int huge_page_order(struct hstate *h)
1005 {
1006 	return 0;
1007 }
1008 
1009 static inline unsigned int huge_page_shift(struct hstate *h)
1010 {
1011 	return PAGE_SHIFT;
1012 }
1013 
1014 static inline bool hstate_is_gigantic(struct hstate *h)
1015 {
1016 	return false;
1017 }
1018 
1019 static inline unsigned int pages_per_huge_page(struct hstate *h)
1020 {
1021 	return 1;
1022 }
1023 
1024 static inline unsigned hstate_index_to_shift(unsigned index)
1025 {
1026 	return 0;
1027 }
1028 
1029 static inline int hstate_index(struct hstate *h)
1030 {
1031 	return 0;
1032 }
1033 
1034 static inline int dissolve_free_huge_page(struct page *page)
1035 {
1036 	return 0;
1037 }
1038 
1039 static inline int dissolve_free_huge_pages(unsigned long start_pfn,
1040 					   unsigned long end_pfn)
1041 {
1042 	return 0;
1043 }
1044 
1045 static inline bool hugepage_migration_supported(struct hstate *h)
1046 {
1047 	return false;
1048 }
1049 
1050 static inline bool hugepage_movable_supported(struct hstate *h)
1051 {
1052 	return false;
1053 }
1054 
1055 static inline gfp_t htlb_alloc_mask(struct hstate *h)
1056 {
1057 	return 0;
1058 }
1059 
1060 static inline gfp_t htlb_modify_alloc_mask(struct hstate *h, gfp_t gfp_mask)
1061 {
1062 	return 0;
1063 }
1064 
1065 static inline spinlock_t *huge_pte_lockptr(struct hstate *h,
1066 					   struct mm_struct *mm, pte_t *pte)
1067 {
1068 	return &mm->page_table_lock;
1069 }
1070 
1071 static inline void hugetlb_count_init(struct mm_struct *mm)
1072 {
1073 }
1074 
1075 static inline void hugetlb_report_usage(struct seq_file *f, struct mm_struct *m)
1076 {
1077 }
1078 
1079 static inline void hugetlb_count_sub(long l, struct mm_struct *mm)
1080 {
1081 }
1082 
1083 static inline void set_huge_swap_pte_at(struct mm_struct *mm, unsigned long addr,
1084 					pte_t *ptep, pte_t pte, unsigned long sz)
1085 {
1086 }
1087 #endif	/* CONFIG_HUGETLB_PAGE */
1088 
1089 static inline spinlock_t *huge_pte_lock(struct hstate *h,
1090 					struct mm_struct *mm, pte_t *pte)
1091 {
1092 	spinlock_t *ptl;
1093 
1094 	ptl = huge_pte_lockptr(h, mm, pte);
1095 	spin_lock(ptl);
1096 	return ptl;
1097 }
1098 
1099 #if defined(CONFIG_HUGETLB_PAGE) && defined(CONFIG_CMA)
1100 extern void __init hugetlb_cma_reserve(int order);
1101 extern void __init hugetlb_cma_check(void);
1102 #else
1103 static inline __init void hugetlb_cma_reserve(int order)
1104 {
1105 }
1106 static inline __init void hugetlb_cma_check(void)
1107 {
1108 }
1109 #endif
1110 
1111 bool want_pmd_share(struct vm_area_struct *vma, unsigned long addr);
1112 
1113 #ifndef __HAVE_ARCH_FLUSH_HUGETLB_TLB_RANGE
1114 /*
1115  * ARCHes with special requirements for evicting HUGETLB backing TLB entries can
1116  * implement this.
1117  */
1118 #define flush_hugetlb_tlb_range(vma, addr, end)	flush_tlb_range(vma, addr, end)
1119 #endif
1120 
1121 #endif /* _LINUX_HUGETLB_H */
1122