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