selftests/vm: drop mnt point for hugetlb in run_vmtests.sh
[linux-2.6-block.git] / include / linux / hugetlb.h
CommitLineData
b2441318 1/* SPDX-License-Identifier: GPL-2.0 */
1da177e4
LT
2#ifndef _LINUX_HUGETLB_H
3#define _LINUX_HUGETLB_H
4
be93d8cf 5#include <linux/mm_types.h>
309381fe 6#include <linux/mmdebug.h>
4e950f6f 7#include <linux/fs.h>
8edf344c 8#include <linux/hugetlb_inline.h>
abb8206c 9#include <linux/cgroup.h>
9119a41e
JK
10#include <linux/list.h>
11#include <linux/kref.h>
ca5999fd 12#include <linux/pgtable.h>
d92bbc27 13#include <linux/gfp.h>
f6191471 14#include <linux/userfaultfd_k.h>
4e950f6f 15
e9ea0e2d
AM
16struct ctl_table;
17struct user_struct;
24669e58 18struct mmu_gather;
a4a00b45 19struct node;
e9ea0e2d 20
691cdf01 21#ifndef CONFIG_ARCH_HAS_HUGEPD
e2299292
AK
22typedef struct { unsigned long pd; } hugepd_t;
23#define is_hugepd(hugepd) (0)
24#define __hugepd(x) ((hugepd_t) { (x) })
e2299292
AK
25#endif
26
1da177e4
LT
27#ifdef CONFIG_HUGETLB_PAGE
28
29#include <linux/mempolicy.h>
516dffdc 30#include <linux/shm.h>
63551ae0 31#include <asm/tlbflush.h>
1da177e4 32
cd39d4e9
MS
33/*
34 * For HugeTLB page, there are more metadata to save in the struct page. But
35 * the head struct page cannot meet our needs, so we have to abuse other tail
36 * struct page to store the metadata. In order to avoid conflicts caused by
37 * subsequent use of more tail struct pages, we gather these discrete indexes
38 * of tail struct page here.
39 */
40enum {
41 SUBPAGE_INDEX_SUBPOOL = 1, /* reuse page->private */
42#ifdef CONFIG_CGROUP_HUGETLB
43 SUBPAGE_INDEX_CGROUP, /* reuse page->private */
44 SUBPAGE_INDEX_CGROUP_RSVD, /* reuse page->private */
45 __MAX_CGROUP_SUBPAGE_INDEX = SUBPAGE_INDEX_CGROUP_RSVD,
161df60e
NH
46#endif
47#ifdef CONFIG_MEMORY_FAILURE
48 SUBPAGE_INDEX_HWPOISON,
cd39d4e9
MS
49#endif
50 __NR_USED_SUBPAGE,
51};
52
90481622
DG
53struct hugepage_subpool {
54 spinlock_t lock;
55 long count;
c6a91820
MK
56 long max_hpages; /* Maximum huge pages or -1 if no maximum. */
57 long used_hpages; /* Used count against maximum, includes */
06c88398 58 /* both allocated and reserved pages. */
c6a91820
MK
59 struct hstate *hstate;
60 long min_hpages; /* Minimum huge pages or -1 if no minimum. */
61 long rsv_hpages; /* Pages reserved against global pool to */
6c26d310 62 /* satisfy minimum size. */
90481622
DG
63};
64
9119a41e
JK
65struct resv_map {
66 struct kref refs;
7b24d861 67 spinlock_t lock;
9119a41e 68 struct list_head regions;
5e911373
MK
69 long adds_in_progress;
70 struct list_head region_cache;
71 long region_cache_count;
e9fe92ae
MA
72#ifdef CONFIG_CGROUP_HUGETLB
73 /*
74 * On private mappings, the counter to uncharge reservations is stored
75 * here. If these fields are 0, then either the mapping is shared, or
76 * cgroup accounting is disabled for this resv_map.
77 */
78 struct page_counter *reservation_counter;
79 unsigned long pages_per_hpage;
80 struct cgroup_subsys_state *css;
81#endif
9119a41e 82};
075a61d0
MA
83
84/*
85 * Region tracking -- allows tracking of reservations and instantiated pages
86 * across the pages in a mapping.
87 *
88 * The region data structures are embedded into a resv_map and protected
89 * by a resv_map's lock. The set of regions within the resv_map represent
90 * reservations for huge pages, or huge pages that have already been
91 * instantiated within the map. The from and to elements are huge page
06c88398 92 * indices into the associated mapping. from indicates the starting index
075a61d0
MA
93 * of the region. to represents the first index past the end of the region.
94 *
95 * For example, a file region structure with from == 0 and to == 4 represents
96 * four huge pages in a mapping. It is important to note that the to element
97 * represents the first element past the end of the region. This is used in
98 * arithmetic as 4(to) - 0(from) = 4 huge pages in the region.
99 *
100 * Interval notation of the form [from, to) will be used to indicate that
101 * the endpoint from is inclusive and to is exclusive.
102 */
103struct file_region {
104 struct list_head link;
105 long from;
106 long to;
107#ifdef CONFIG_CGROUP_HUGETLB
108 /*
109 * On shared mappings, each reserved region appears as a struct
110 * file_region in resv_map. These fields hold the info needed to
111 * uncharge each reservation.
112 */
113 struct page_counter *reservation_counter;
114 struct cgroup_subsys_state *css;
115#endif
116};
117
8d9bfb26
MK
118struct hugetlb_vma_lock {
119 struct kref refs;
120 struct rw_semaphore rw_sema;
121 struct vm_area_struct *vma;
122};
123
9119a41e
JK
124extern struct resv_map *resv_map_alloc(void);
125void resv_map_release(struct kref *ref);
126
c3f38a38
AK
127extern spinlock_t hugetlb_lock;
128extern int hugetlb_max_hstate __read_mostly;
129#define for_each_hstate(h) \
130 for ((h) = hstates; (h) < &hstates[hugetlb_max_hstate]; (h)++)
131
7ca02d0a
MK
132struct hugepage_subpool *hugepage_new_subpool(struct hstate *h, long max_hpages,
133 long min_hpages);
90481622
DG
134void hugepage_put_subpool(struct hugepage_subpool *spool);
135
8d9bfb26 136void hugetlb_dup_vma_private(struct vm_area_struct *vma);
550a7d60 137void clear_vma_resv_huge_pages(struct vm_area_struct *vma);
32927393
CH
138int hugetlb_sysctl_handler(struct ctl_table *, int, void *, size_t *, loff_t *);
139int hugetlb_overcommit_handler(struct ctl_table *, int, void *, size_t *,
140 loff_t *);
141int hugetlb_treat_movable_handler(struct ctl_table *, int, void *, size_t *,
142 loff_t *);
143int hugetlb_mempolicy_sysctl_handler(struct ctl_table *, int, void *, size_t *,
144 loff_t *);
06808b08 145
550a7d60
MA
146int move_hugetlb_page_tables(struct vm_area_struct *vma,
147 struct vm_area_struct *new_vma,
148 unsigned long old_addr, unsigned long new_addr,
149 unsigned long len);
bc70fbf2
PX
150int copy_hugetlb_page_range(struct mm_struct *, struct mm_struct *,
151 struct vm_area_struct *, struct vm_area_struct *);
57a196a5
MK
152struct page *hugetlb_follow_page_mask(struct vm_area_struct *vma,
153 unsigned long address, unsigned int flags);
28a35716
ML
154long follow_hugetlb_page(struct mm_struct *, struct vm_area_struct *,
155 struct page **, struct vm_area_struct **,
87ffc118
AA
156 unsigned long *, unsigned long *, long, unsigned int,
157 int *);
04f2cbe3 158void unmap_hugepage_range(struct vm_area_struct *,
05e90bd0
PX
159 unsigned long, unsigned long, struct page *,
160 zap_flags_t);
d833352a
MG
161void __unmap_hugepage_range_final(struct mmu_gather *tlb,
162 struct vm_area_struct *vma,
163 unsigned long start, unsigned long end,
05e90bd0 164 struct page *ref_page, zap_flags_t zap_flags);
e1759c21 165void hugetlb_report_meminfo(struct seq_file *);
7981593b 166int hugetlb_report_node_meminfo(char *buf, int len, int nid);
dcadcf1c 167void hugetlb_show_meminfo_node(int nid);
1da177e4 168unsigned long hugetlb_total_pages(void);
2b740303 169vm_fault_t hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
788c7df4 170 unsigned long address, unsigned int flags);
714c1891 171#ifdef CONFIG_USERFAULTFD
8fb5debc
MK
172int hugetlb_mcopy_atomic_pte(struct mm_struct *dst_mm, pte_t *dst_pte,
173 struct vm_area_struct *dst_vma,
174 unsigned long dst_addr,
175 unsigned long src_addr,
f6191471 176 enum mcopy_atomic_mode mode,
6041c691
PX
177 struct page **pagep,
178 bool wp_copy);
714c1891 179#endif /* CONFIG_USERFAULTFD */
33b8f84a 180bool hugetlb_reserve_pages(struct inode *inode, long from, long to,
5a6fe125 181 struct vm_area_struct *vma,
ca16d140 182 vm_flags_t vm_flags);
b5cec28d
MK
183long hugetlb_unreserve_pages(struct inode *inode, long start, long end,
184 long freed);
7ce82f4c 185int isolate_hugetlb(struct page *page, struct list_head *list);
25182f05 186int get_hwpoison_huge_page(struct page *page, bool *hugetlb);
405ce051 187int get_huge_page_for_hwpoison(unsigned long pfn, int flags);
31caf665 188void putback_active_hugepage(struct page *page);
ab5ac90a 189void move_hugetlb_state(struct page *oldpage, struct page *newpage, int reason);
8f1d26d0 190void free_huge_page(struct page *page);
72e2936c 191void hugetlb_fix_reserve_counts(struct inode *inode);
c672c7f2 192extern struct mutex *hugetlb_fault_mutex_table;
188b04a7 193u32 hugetlb_fault_mutex_hash(struct address_space *mapping, pgoff_t idx);
1da177e4 194
aec44e0f
PX
195pte_t *huge_pmd_share(struct mm_struct *mm, struct vm_area_struct *vma,
196 unsigned long addr, pud_t *pud);
3212b535 197
c0d0381a
MK
198struct address_space *hugetlb_page_mapping_lock_write(struct page *hpage);
199
1da177e4 200extern int sysctl_hugetlb_shm_group;
53ba51d2 201extern struct list_head huge_boot_pages;
1da177e4 202
63551ae0
DG
203/* arch callbacks */
204
aec44e0f 205pte_t *huge_pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
a5516438 206 unsigned long addr, unsigned long sz);
7868a208
PA
207pte_t *huge_pte_offset(struct mm_struct *mm,
208 unsigned long addr, unsigned long sz);
e95a9851 209unsigned long hugetlb_mask_last_page(struct hstate *h);
34ae204f 210int huge_pmd_unshare(struct mm_struct *mm, struct vm_area_struct *vma,
4ddb4d91 211 unsigned long addr, pte_t *ptep);
017b1660
MK
212void adjust_range_if_pmd_sharing_possible(struct vm_area_struct *vma,
213 unsigned long *start, unsigned long *end);
faaa5b62 214
8d9bfb26
MK
215void hugetlb_vma_lock_read(struct vm_area_struct *vma);
216void hugetlb_vma_unlock_read(struct vm_area_struct *vma);
217void hugetlb_vma_lock_write(struct vm_area_struct *vma);
218void hugetlb_vma_unlock_write(struct vm_area_struct *vma);
219int hugetlb_vma_trylock_write(struct vm_area_struct *vma);
220void hugetlb_vma_assert_locked(struct vm_area_struct *vma);
221void hugetlb_vma_lock_release(struct kref *kref);
222
63551ae0 223int pmd_huge(pmd_t pmd);
c2febafc 224int pud_huge(pud_t pud);
7da4d641 225unsigned long hugetlb_change_protection(struct vm_area_struct *vma,
5a90d5a1
PX
226 unsigned long address, unsigned long end, pgprot_t newprot,
227 unsigned long cp_flags);
63551ae0 228
d5ed7444 229bool is_hugetlb_entry_migration(pte_t pte);
6dfeaff9 230void hugetlb_unshare_all_pmds(struct vm_area_struct *vma);
ab5ac90a 231
1da177e4
LT
232#else /* !CONFIG_HUGETLB_PAGE */
233
8d9bfb26 234static inline void hugetlb_dup_vma_private(struct vm_area_struct *vma)
a1e78772
MG
235{
236}
237
550a7d60
MA
238static inline void clear_vma_resv_huge_pages(struct vm_area_struct *vma)
239{
240}
241
1da177e4
LT
242static inline unsigned long hugetlb_total_pages(void)
243{
244 return 0;
245}
246
c0d0381a
MK
247static inline struct address_space *hugetlb_page_mapping_lock_write(
248 struct page *hpage)
249{
250 return NULL;
251}
252
34ae204f
MK
253static inline int huge_pmd_unshare(struct mm_struct *mm,
254 struct vm_area_struct *vma,
4ddb4d91 255 unsigned long addr, pte_t *ptep)
017b1660
MK
256{
257 return 0;
258}
259
260static inline void adjust_range_if_pmd_sharing_possible(
261 struct vm_area_struct *vma,
262 unsigned long *start, unsigned long *end)
263{
264}
265
57a196a5
MK
266static inline struct page *hugetlb_follow_page_mask(struct vm_area_struct *vma,
267 unsigned long address, unsigned int flags)
268{
269 BUILD_BUG(); /* should never be compiled in if !CONFIG_HUGETLB_PAGE*/
270}
271
1f9dccb2
MK
272static inline long follow_hugetlb_page(struct mm_struct *mm,
273 struct vm_area_struct *vma, struct page **pages,
274 struct vm_area_struct **vmas, unsigned long *position,
275 unsigned long *nr_pages, long i, unsigned int flags,
276 int *nonblocking)
277{
278 BUG();
279 return 0;
280}
281
1f9dccb2 282static inline int copy_hugetlb_page_range(struct mm_struct *dst,
bc70fbf2
PX
283 struct mm_struct *src,
284 struct vm_area_struct *dst_vma,
285 struct vm_area_struct *src_vma)
1f9dccb2
MK
286{
287 BUG();
288 return 0;
289}
290
550a7d60
MA
291static inline int move_hugetlb_page_tables(struct vm_area_struct *vma,
292 struct vm_area_struct *new_vma,
293 unsigned long old_addr,
294 unsigned long new_addr,
295 unsigned long len)
296{
297 BUG();
298 return 0;
299}
300
e1759c21
AD
301static inline void hugetlb_report_meminfo(struct seq_file *m)
302{
303}
1f9dccb2 304
7981593b 305static inline int hugetlb_report_node_meminfo(char *buf, int len, int nid)
1f9dccb2
MK
306{
307 return 0;
308}
309
dcadcf1c 310static inline void hugetlb_show_meminfo_node(int nid)
949f7ec5
DR
311{
312}
1f9dccb2 313
1f9dccb2
MK
314static inline int prepare_hugepage_range(struct file *file,
315 unsigned long addr, unsigned long len)
316{
317 return -EINVAL;
318}
319
8d9bfb26
MK
320static inline void hugetlb_vma_lock_read(struct vm_area_struct *vma)
321{
322}
323
324static inline void hugetlb_vma_unlock_read(struct vm_area_struct *vma)
325{
326}
327
328static inline void hugetlb_vma_lock_write(struct vm_area_struct *vma)
329{
330}
331
332static inline void hugetlb_vma_unlock_write(struct vm_area_struct *vma)
333{
334}
335
336static inline int hugetlb_vma_trylock_write(struct vm_area_struct *vma)
337{
338 return 1;
339}
340
341static inline void hugetlb_vma_assert_locked(struct vm_area_struct *vma)
342{
343}
344
1f9dccb2
MK
345static inline int pmd_huge(pmd_t pmd)
346{
347 return 0;
348}
349
350static inline int pud_huge(pud_t pud)
351{
352 return 0;
353}
354
355static inline int is_hugepage_only_range(struct mm_struct *mm,
356 unsigned long addr, unsigned long len)
357{
358 return 0;
359}
360
361static inline void hugetlb_free_pgd_range(struct mmu_gather *tlb,
362 unsigned long addr, unsigned long end,
363 unsigned long floor, unsigned long ceiling)
364{
365 BUG();
366}
367
714c1891 368#ifdef CONFIG_USERFAULTFD
1f9dccb2
MK
369static inline int hugetlb_mcopy_atomic_pte(struct mm_struct *dst_mm,
370 pte_t *dst_pte,
371 struct vm_area_struct *dst_vma,
372 unsigned long dst_addr,
373 unsigned long src_addr,
f6191471 374 enum mcopy_atomic_mode mode,
6041c691
PX
375 struct page **pagep,
376 bool wp_copy)
1f9dccb2
MK
377{
378 BUG();
379 return 0;
380}
714c1891 381#endif /* CONFIG_USERFAULTFD */
1f9dccb2
MK
382
383static inline pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr,
384 unsigned long sz)
385{
386 return NULL;
387}
24669e58 388
7ce82f4c 389static inline int isolate_hugetlb(struct page *page, struct list_head *list)
f40386a4 390{
7ce82f4c 391 return -EBUSY;
f40386a4 392}
1da177e4 393
25182f05
NH
394static inline int get_hwpoison_huge_page(struct page *page, bool *hugetlb)
395{
396 return 0;
397}
398
405ce051
NH
399static inline int get_huge_page_for_hwpoison(unsigned long pfn, int flags)
400{
401 return 0;
402}
403
1f9dccb2
MK
404static inline void putback_active_hugepage(struct page *page)
405{
406}
407
408static inline void move_hugetlb_state(struct page *oldpage,
409 struct page *newpage, int reason)
410{
411}
412
413static inline unsigned long hugetlb_change_protection(
414 struct vm_area_struct *vma, unsigned long address,
5a90d5a1
PX
415 unsigned long end, pgprot_t newprot,
416 unsigned long cp_flags)
7da4d641
PZ
417{
418 return 0;
419}
8f860591 420
d833352a
MG
421static inline void __unmap_hugepage_range_final(struct mmu_gather *tlb,
422 struct vm_area_struct *vma, unsigned long start,
05e90bd0
PX
423 unsigned long end, struct page *ref_page,
424 zap_flags_t zap_flags)
d833352a
MG
425{
426 BUG();
427}
428
a953e772 429static inline vm_fault_t hugetlb_fault(struct mm_struct *mm,
1f9dccb2
MK
430 struct vm_area_struct *vma, unsigned long address,
431 unsigned int flags)
a953e772
SJ
432{
433 BUG();
434 return 0;
435}
24669e58 436
6dfeaff9
PX
437static inline void hugetlb_unshare_all_pmds(struct vm_area_struct *vma) { }
438
1da177e4 439#endif /* !CONFIG_HUGETLB_PAGE */
f30c59e9
AK
440/*
441 * hugepages at page global directory. If arch support
442 * hugepages at pgd level, they need to define this.
443 */
444#ifndef pgd_huge
445#define pgd_huge(x) 0
446#endif
c2febafc
KS
447#ifndef p4d_huge
448#define p4d_huge(x) 0
449#endif
f30c59e9
AK
450
451#ifndef pgd_write
452static inline int pgd_write(pgd_t pgd)
453{
454 BUG();
455 return 0;
456}
457#endif
458
4e52780d
EM
459#define HUGETLB_ANON_FILE "anon_hugepage"
460
6bfde05b
EM
461enum {
462 /*
463 * The file will be used as an shm file so shmfs accounting rules
464 * apply
465 */
466 HUGETLB_SHMFS_INODE = 1,
4e52780d
EM
467 /*
468 * The file is being created on the internal vfs mount and shmfs
469 * accounting rules do not apply
470 */
471 HUGETLB_ANONHUGE_INODE = 2,
6bfde05b
EM
472};
473
1da177e4 474#ifdef CONFIG_HUGETLBFS
1da177e4 475struct hugetlbfs_sb_info {
1da177e4
LT
476 long max_inodes; /* inodes allowed */
477 long free_inodes; /* inodes free */
478 spinlock_t stat_lock;
a137e1cc 479 struct hstate *hstate;
90481622 480 struct hugepage_subpool *spool;
4a25220d
DH
481 kuid_t uid;
482 kgid_t gid;
483 umode_t mode;
1da177e4
LT
484};
485
1da177e4
LT
486static inline struct hugetlbfs_sb_info *HUGETLBFS_SB(struct super_block *sb)
487{
488 return sb->s_fs_info;
489}
490
da14c1e5
MAL
491struct hugetlbfs_inode_info {
492 struct shared_policy policy;
493 struct inode vfs_inode;
ff62a342 494 unsigned int seals;
da14c1e5
MAL
495};
496
497static inline struct hugetlbfs_inode_info *HUGETLBFS_I(struct inode *inode)
498{
499 return container_of(inode, struct hugetlbfs_inode_info, vfs_inode);
500}
501
4b6f5d20 502extern const struct file_operations hugetlbfs_file_operations;
f0f37e2f 503extern const struct vm_operations_struct hugetlb_vm_ops;
af73e4d9 504struct file *hugetlb_file_setup(const char *name, size_t size, vm_flags_t acct,
83c1fd76 505 int creat_flags, int page_size_log);
1da177e4 506
719ff321 507static inline bool is_file_hugepages(struct file *file)
1da177e4 508{
516dffdc 509 if (file->f_op == &hugetlbfs_file_operations)
719ff321 510 return true;
516dffdc 511
719ff321 512 return is_file_shm_hugepages(file);
1da177e4
LT
513}
514
bb297bb2
CL
515static inline struct hstate *hstate_inode(struct inode *i)
516{
517 return HUGETLBFS_SB(i->i_sb)->hstate;
518}
1da177e4
LT
519#else /* !CONFIG_HUGETLBFS */
520
719ff321 521#define is_file_hugepages(file) false
40716e29 522static inline struct file *
af73e4d9 523hugetlb_file_setup(const char *name, size_t size, vm_flags_t acctflag,
83c1fd76 524 int creat_flags, int page_size_log)
e9ea0e2d
AM
525{
526 return ERR_PTR(-ENOSYS);
527}
1da177e4 528
bb297bb2
CL
529static inline struct hstate *hstate_inode(struct inode *i)
530{
531 return NULL;
532}
1da177e4
LT
533#endif /* !CONFIG_HUGETLBFS */
534
d2ba27e8
AB
535#ifdef HAVE_ARCH_HUGETLB_UNMAPPED_AREA
536unsigned long hugetlb_get_unmapped_area(struct file *file, unsigned long addr,
537 unsigned long len, unsigned long pgoff,
538 unsigned long flags);
539#endif /* HAVE_ARCH_HUGETLB_UNMAPPED_AREA */
540
4b439e25
CL
541unsigned long
542generic_hugetlb_get_unmapped_area(struct file *file, unsigned long addr,
543 unsigned long len, unsigned long pgoff,
544 unsigned long flags);
545
d6995da3
MK
546/*
547 * huegtlb page specific state flags. These flags are located in page.private
548 * of the hugetlb head page. Functions created via the below macros should be
549 * used to manipulate these flags.
550 *
551 * HPG_restore_reserve - Set when a hugetlb page consumes a reservation at
552 * allocation time. Cleared when page is fully instantiated. Free
553 * routine checks flag to restore a reservation on error paths.
d95c0337
MK
554 * Synchronization: Examined or modified by code that knows it has
555 * the only reference to page. i.e. After allocation but before use
556 * or when the page is being freed.
8f251a3d
MK
557 * HPG_migratable - Set after a newly allocated page is added to the page
558 * cache and/or page tables. Indicates the page is a candidate for
559 * migration.
d95c0337
MK
560 * Synchronization: Initially set after new page allocation with no
561 * locking. When examined and modified during migration processing
562 * (isolate, migrate, putback) the hugetlb_lock is held.
161df60e 563 * HPG_temporary - Set on a page that is temporarily allocated from the buddy
9157c311
MK
564 * allocator. Typically used for migration target pages when no pages
565 * are available in the pool. The hugetlb free page path will
566 * immediately free pages with this flag set to the buddy allocator.
d95c0337
MK
567 * Synchronization: Can be set after huge page allocation from buddy when
568 * code knows it has only reference. All other examinations and
569 * modifications require hugetlb_lock.
6c037149 570 * HPG_freed - Set when page is on the free lists.
d95c0337 571 * Synchronization: hugetlb_lock held for examination and modification.
ad2fa371 572 * HPG_vmemmap_optimized - Set when the vmemmap pages of the page are freed.
161df60e
NH
573 * HPG_raw_hwp_unreliable - Set when the hugetlb page has a hwpoison sub-page
574 * that is not tracked by raw_hwp_page list.
d6995da3
MK
575 */
576enum hugetlb_page_flags {
577 HPG_restore_reserve = 0,
8f251a3d 578 HPG_migratable,
9157c311 579 HPG_temporary,
6c037149 580 HPG_freed,
ad2fa371 581 HPG_vmemmap_optimized,
161df60e 582 HPG_raw_hwp_unreliable,
d6995da3
MK
583 __NR_HPAGEFLAGS,
584};
585
586/*
587 * Macros to create test, set and clear function definitions for
588 * hugetlb specific page flags.
589 */
590#ifdef CONFIG_HUGETLB_PAGE
591#define TESTHPAGEFLAG(uname, flname) \
592static inline int HPage##uname(struct page *page) \
593 { return test_bit(HPG_##flname, &(page->private)); }
594
595#define SETHPAGEFLAG(uname, flname) \
596static inline void SetHPage##uname(struct page *page) \
597 { set_bit(HPG_##flname, &(page->private)); }
598
599#define CLEARHPAGEFLAG(uname, flname) \
600static inline void ClearHPage##uname(struct page *page) \
601 { clear_bit(HPG_##flname, &(page->private)); }
602#else
603#define TESTHPAGEFLAG(uname, flname) \
604static inline int HPage##uname(struct page *page) \
605 { return 0; }
606
607#define SETHPAGEFLAG(uname, flname) \
608static inline void SetHPage##uname(struct page *page) \
609 { }
610
611#define CLEARHPAGEFLAG(uname, flname) \
612static inline void ClearHPage##uname(struct page *page) \
613 { }
614#endif
615
616#define HPAGEFLAG(uname, flname) \
617 TESTHPAGEFLAG(uname, flname) \
618 SETHPAGEFLAG(uname, flname) \
619 CLEARHPAGEFLAG(uname, flname) \
620
621/*
622 * Create functions associated with hugetlb page flags
623 */
624HPAGEFLAG(RestoreReserve, restore_reserve)
8f251a3d 625HPAGEFLAG(Migratable, migratable)
9157c311 626HPAGEFLAG(Temporary, temporary)
6c037149 627HPAGEFLAG(Freed, freed)
ad2fa371 628HPAGEFLAG(VmemmapOptimized, vmemmap_optimized)
161df60e 629HPAGEFLAG(RawHwpUnreliable, raw_hwp_unreliable)
d6995da3 630
a5516438
AK
631#ifdef CONFIG_HUGETLB_PAGE
632
a3437870 633#define HSTATE_NAME_LEN 32
a5516438
AK
634/* Defines one hugetlb page size */
635struct hstate {
29383967 636 struct mutex resize_lock;
e8c5c824
LS
637 int next_nid_to_alloc;
638 int next_nid_to_free;
a5516438 639 unsigned int order;
79dfc695 640 unsigned int demote_order;
a5516438
AK
641 unsigned long mask;
642 unsigned long max_huge_pages;
643 unsigned long nr_huge_pages;
644 unsigned long free_huge_pages;
645 unsigned long resv_huge_pages;
646 unsigned long surplus_huge_pages;
647 unsigned long nr_overcommit_huge_pages;
0edaecfa 648 struct list_head hugepage_activelist;
a5516438 649 struct list_head hugepage_freelists[MAX_NUMNODES];
b5389086 650 unsigned int max_huge_pages_node[MAX_NUMNODES];
a5516438
AK
651 unsigned int nr_huge_pages_node[MAX_NUMNODES];
652 unsigned int free_huge_pages_node[MAX_NUMNODES];
653 unsigned int surplus_huge_pages_node[MAX_NUMNODES];
abb8206c
AK
654#ifdef CONFIG_CGROUP_HUGETLB
655 /* cgroup control files */
f4776199
MA
656 struct cftype cgroup_files_dfl[8];
657 struct cftype cgroup_files_legacy[10];
abb8206c 658#endif
a3437870 659 char name[HSTATE_NAME_LEN];
a5516438
AK
660};
661
53ba51d2
JT
662struct huge_bootmem_page {
663 struct list_head list;
664 struct hstate *hstate;
665};
666
ae37c7ff 667int isolate_or_dissolve_huge_page(struct page *page, struct list_head *list);
70c3547e
MK
668struct page *alloc_huge_page(struct vm_area_struct *vma,
669 unsigned long addr, int avoid_reserve);
3e59fcb0 670struct page *alloc_huge_page_nodemask(struct hstate *h, int preferred_nid,
d92bbc27 671 nodemask_t *nmask, gfp_t gfp_mask);
389c8178
MH
672struct page *alloc_huge_page_vma(struct hstate *h, struct vm_area_struct *vma,
673 unsigned long address);
7e1813d4 674int hugetlb_add_to_page_cache(struct page *page, struct address_space *mapping,
ab76ad54 675 pgoff_t idx);
846be085
MK
676void restore_reserve_on_error(struct hstate *h, struct vm_area_struct *vma,
677 unsigned long address, struct page *page);
bf50bab2 678
53ba51d2 679/* arch callback */
b5389086
ZY
680int __init __alloc_bootmem_huge_page(struct hstate *h, int nid);
681int __init alloc_bootmem_huge_page(struct hstate *h, int nid);
682bool __init hugetlb_node_alloc_supported(void);
53ba51d2 683
e5ff2159 684void __init hugetlb_add_hstate(unsigned order);
ae94da89 685bool __init arch_hugetlb_valid_size(unsigned long size);
e5ff2159
AK
686struct hstate *size_to_hstate(unsigned long size);
687
688#ifndef HUGE_MAX_HSTATE
689#define HUGE_MAX_HSTATE 1
690#endif
691
692extern struct hstate hstates[HUGE_MAX_HSTATE];
693extern unsigned int default_hstate_idx;
694
695#define default_hstate (hstates[default_hstate_idx])
a5516438 696
d6995da3
MK
697/*
698 * hugetlb page subpool pointer located in hpage[1].private
699 */
700static inline struct hugepage_subpool *hugetlb_page_subpool(struct page *hpage)
701{
cd39d4e9 702 return (void *)page_private(hpage + SUBPAGE_INDEX_SUBPOOL);
d6995da3
MK
703}
704
705static inline void hugetlb_set_page_subpool(struct page *hpage,
706 struct hugepage_subpool *subpool)
707{
cd39d4e9 708 set_page_private(hpage + SUBPAGE_INDEX_SUBPOOL, (unsigned long)subpool);
d6995da3
MK
709}
710
a5516438
AK
711static inline struct hstate *hstate_file(struct file *f)
712{
496ad9aa 713 return hstate_inode(file_inode(f));
a5516438
AK
714}
715
af73e4d9
NH
716static inline struct hstate *hstate_sizelog(int page_size_log)
717{
718 if (!page_size_log)
719 return &default_hstate;
97ad2be1
SL
720
721 return size_to_hstate(1UL << page_size_log);
af73e4d9
NH
722}
723
a137e1cc 724static inline struct hstate *hstate_vma(struct vm_area_struct *vma)
a5516438 725{
a137e1cc 726 return hstate_file(vma->vm_file);
a5516438
AK
727}
728
6213834c 729static inline unsigned long huge_page_size(const struct hstate *h)
a5516438
AK
730{
731 return (unsigned long)PAGE_SIZE << h->order;
732}
733
08fba699
MG
734extern unsigned long vma_kernel_pagesize(struct vm_area_struct *vma);
735
3340289d
MG
736extern unsigned long vma_mmu_pagesize(struct vm_area_struct *vma);
737
a5516438
AK
738static inline unsigned long huge_page_mask(struct hstate *h)
739{
740 return h->mask;
741}
742
743static inline unsigned int huge_page_order(struct hstate *h)
744{
745 return h->order;
746}
747
748static inline unsigned huge_page_shift(struct hstate *h)
749{
750 return h->order + PAGE_SHIFT;
751}
752
bae7f4ae
LC
753static inline bool hstate_is_gigantic(struct hstate *h)
754{
755 return huge_page_order(h) >= MAX_ORDER;
756}
757
6213834c 758static inline unsigned int pages_per_huge_page(const struct hstate *h)
a5516438
AK
759{
760 return 1 << h->order;
761}
762
763static inline unsigned int blocks_per_huge_page(struct hstate *h)
764{
765 return huge_page_size(h) / 512;
766}
767
768#include <asm/hugetlb.h>
769
b0eae98c
AK
770#ifndef is_hugepage_only_range
771static inline int is_hugepage_only_range(struct mm_struct *mm,
772 unsigned long addr, unsigned long len)
773{
774 return 0;
775}
776#define is_hugepage_only_range is_hugepage_only_range
777#endif
778
5be99343
AK
779#ifndef arch_clear_hugepage_flags
780static inline void arch_clear_hugepage_flags(struct page *page) { }
781#define arch_clear_hugepage_flags arch_clear_hugepage_flags
782#endif
783
d9ed9faa 784#ifndef arch_make_huge_pte
79c1c594
CL
785static inline pte_t arch_make_huge_pte(pte_t entry, unsigned int shift,
786 vm_flags_t flags)
d9ed9faa 787{
16785bd7 788 return pte_mkhuge(entry);
d9ed9faa
CM
789}
790#endif
791
e5ff2159
AK
792static inline struct hstate *page_hstate(struct page *page)
793{
309381fe 794 VM_BUG_ON_PAGE(!PageHuge(page), page);
a50b854e 795 return size_to_hstate(page_size(page));
e5ff2159
AK
796}
797
aa50d3a7
AK
798static inline unsigned hstate_index_to_shift(unsigned index)
799{
800 return hstates[index].order + PAGE_SHIFT;
801}
802
972dc4de
AK
803static inline int hstate_index(struct hstate *h)
804{
805 return h - hstates;
806}
807
c3114a84 808extern int dissolve_free_huge_page(struct page *page);
082d5b6b
GS
809extern int dissolve_free_huge_pages(unsigned long start_pfn,
810 unsigned long end_pfn);
e693de18 811
161df60e
NH
812#ifdef CONFIG_MEMORY_FAILURE
813extern void hugetlb_clear_page_hwpoison(struct page *hpage);
814#else
815static inline void hugetlb_clear_page_hwpoison(struct page *hpage)
816{
817}
818#endif
819
c177c81e 820#ifdef CONFIG_ARCH_ENABLE_HUGEPAGE_MIGRATION
e693de18
AK
821#ifndef arch_hugetlb_migration_supported
822static inline bool arch_hugetlb_migration_supported(struct hstate *h)
823{
94310cbc 824 if ((huge_page_shift(h) == PMD_SHIFT) ||
9b553bf5
AK
825 (huge_page_shift(h) == PUD_SHIFT) ||
826 (huge_page_shift(h) == PGDIR_SHIFT))
94310cbc
AK
827 return true;
828 else
829 return false;
e693de18
AK
830}
831#endif
c177c81e 832#else
e693de18
AK
833static inline bool arch_hugetlb_migration_supported(struct hstate *h)
834{
d70c17d4 835 return false;
e693de18 836}
c177c81e 837#endif
e693de18
AK
838
839static inline bool hugepage_migration_supported(struct hstate *h)
840{
841 return arch_hugetlb_migration_supported(h);
83467efb 842}
c8721bbb 843
7ed2c31d
AK
844/*
845 * Movability check is different as compared to migration check.
846 * It determines whether or not a huge page should be placed on
847 * movable zone or not. Movability of any huge page should be
848 * required only if huge page size is supported for migration.
06c88398 849 * There won't be any reason for the huge page to be movable if
7ed2c31d
AK
850 * it is not migratable to start with. Also the size of the huge
851 * page should be large enough to be placed under a movable zone
852 * and still feasible enough to be migratable. Just the presence
853 * in movable zone does not make the migration feasible.
854 *
855 * So even though large huge page sizes like the gigantic ones
856 * are migratable they should not be movable because its not
857 * feasible to migrate them from movable zone.
858 */
859static inline bool hugepage_movable_supported(struct hstate *h)
860{
861 if (!hugepage_migration_supported(h))
862 return false;
863
864 if (hstate_is_gigantic(h))
865 return false;
866 return true;
867}
868
d92bbc27
JK
869/* Movability of hugepages depends on migration support. */
870static inline gfp_t htlb_alloc_mask(struct hstate *h)
871{
872 if (hugepage_movable_supported(h))
873 return GFP_HIGHUSER_MOVABLE;
874 else
875 return GFP_HIGHUSER;
876}
877
19fc7bed
JK
878static inline gfp_t htlb_modify_alloc_mask(struct hstate *h, gfp_t gfp_mask)
879{
880 gfp_t modified_mask = htlb_alloc_mask(h);
881
882 /* Some callers might want to enforce node */
883 modified_mask |= (gfp_mask & __GFP_THISNODE);
884
41b4dc14
JK
885 modified_mask |= (gfp_mask & __GFP_NOWARN);
886
19fc7bed
JK
887 return modified_mask;
888}
889
cb900f41
KS
890static inline spinlock_t *huge_pte_lockptr(struct hstate *h,
891 struct mm_struct *mm, pte_t *pte)
892{
893 if (huge_page_size(h) == PMD_SIZE)
894 return pmd_lockptr(mm, (pmd_t *) pte);
895 VM_BUG_ON(huge_page_size(h) == PAGE_SIZE);
896 return &mm->page_table_lock;
897}
898
2531c8cf
DD
899#ifndef hugepages_supported
900/*
901 * Some platform decide whether they support huge pages at boot
902 * time. Some of them, such as powerpc, set HPAGE_SHIFT to 0
903 * when there is no such support
904 */
905#define hugepages_supported() (HPAGE_SHIFT != 0)
906#endif
457c1b27 907
5d317b2b
NH
908void hugetlb_report_usage(struct seq_file *m, struct mm_struct *mm);
909
13db8c50
LZ
910static inline void hugetlb_count_init(struct mm_struct *mm)
911{
912 atomic_long_set(&mm->hugetlb_usage, 0);
913}
914
5d317b2b
NH
915static inline void hugetlb_count_add(long l, struct mm_struct *mm)
916{
917 atomic_long_add(l, &mm->hugetlb_usage);
918}
919
920static inline void hugetlb_count_sub(long l, struct mm_struct *mm)
921{
922 atomic_long_sub(l, &mm->hugetlb_usage);
923}
e5251fd4 924
023bdd00
AK
925#ifndef huge_ptep_modify_prot_start
926#define huge_ptep_modify_prot_start huge_ptep_modify_prot_start
927static inline pte_t huge_ptep_modify_prot_start(struct vm_area_struct *vma,
928 unsigned long addr, pte_t *ptep)
929{
930 return huge_ptep_get_and_clear(vma->vm_mm, addr, ptep);
931}
932#endif
933
934#ifndef huge_ptep_modify_prot_commit
935#define huge_ptep_modify_prot_commit huge_ptep_modify_prot_commit
936static inline void huge_ptep_modify_prot_commit(struct vm_area_struct *vma,
937 unsigned long addr, pte_t *ptep,
938 pte_t old_pte, pte_t pte)
939{
940 set_huge_pte_at(vma->vm_mm, addr, ptep, pte);
941}
942#endif
943
a4a00b45
MS
944#ifdef CONFIG_NUMA
945void hugetlb_register_node(struct node *node);
946void hugetlb_unregister_node(struct node *node);
947#endif
948
af73e4d9 949#else /* CONFIG_HUGETLB_PAGE */
a5516438 950struct hstate {};
442a5a9a 951
6acfb5ba
MS
952static inline struct hugepage_subpool *hugetlb_page_subpool(struct page *hpage)
953{
954 return NULL;
955}
956
ae37c7ff
OS
957static inline int isolate_or_dissolve_huge_page(struct page *page,
958 struct list_head *list)
369fa227
OS
959{
960 return -ENOMEM;
961}
962
442a5a9a
JG
963static inline struct page *alloc_huge_page(struct vm_area_struct *vma,
964 unsigned long addr,
965 int avoid_reserve)
966{
967 return NULL;
968}
969
442a5a9a 970static inline struct page *
d92bbc27
JK
971alloc_huge_page_nodemask(struct hstate *h, int preferred_nid,
972 nodemask_t *nmask, gfp_t gfp_mask)
442a5a9a
JG
973{
974 return NULL;
975}
976
977static inline struct page *alloc_huge_page_vma(struct hstate *h,
978 struct vm_area_struct *vma,
979 unsigned long address)
980{
981 return NULL;
982}
983
984static inline int __alloc_bootmem_huge_page(struct hstate *h)
985{
986 return 0;
987}
988
989static inline struct hstate *hstate_file(struct file *f)
990{
991 return NULL;
992}
993
994static inline struct hstate *hstate_sizelog(int page_size_log)
995{
996 return NULL;
997}
998
999static inline struct hstate *hstate_vma(struct vm_area_struct *vma)
1000{
1001 return NULL;
442a5a9a
JG
1002}
1003
1004static inline struct hstate *page_hstate(struct page *page)
1005{
1006 return NULL;
1007}
1008
2aff7a47
MWO
1009static inline struct hstate *size_to_hstate(unsigned long size)
1010{
1011 return NULL;
1012}
1013
442a5a9a
JG
1014static inline unsigned long huge_page_size(struct hstate *h)
1015{
1016 return PAGE_SIZE;
1017}
1018
1019static inline unsigned long huge_page_mask(struct hstate *h)
1020{
1021 return PAGE_MASK;
1022}
1023
1024static inline unsigned long vma_kernel_pagesize(struct vm_area_struct *vma)
1025{
1026 return PAGE_SIZE;
1027}
1028
1029static inline unsigned long vma_mmu_pagesize(struct vm_area_struct *vma)
1030{
1031 return PAGE_SIZE;
1032}
1033
1034static inline unsigned int huge_page_order(struct hstate *h)
1035{
1036 return 0;
1037}
1038
1039static inline unsigned int huge_page_shift(struct hstate *h)
1040{
1041 return PAGE_SHIFT;
1042}
1043
94310cbc
AK
1044static inline bool hstate_is_gigantic(struct hstate *h)
1045{
1046 return false;
1047}
1048
510a35d4
AR
1049static inline unsigned int pages_per_huge_page(struct hstate *h)
1050{
1051 return 1;
1052}
c3114a84
AK
1053
1054static inline unsigned hstate_index_to_shift(unsigned index)
1055{
1056 return 0;
1057}
1058
1059static inline int hstate_index(struct hstate *h)
1060{
1061 return 0;
1062}
13d60f4b 1063
c3114a84
AK
1064static inline int dissolve_free_huge_page(struct page *page)
1065{
1066 return 0;
1067}
1068
1069static inline int dissolve_free_huge_pages(unsigned long start_pfn,
1070 unsigned long end_pfn)
1071{
1072 return 0;
1073}
1074
1075static inline bool hugepage_migration_supported(struct hstate *h)
1076{
1077 return false;
1078}
cb900f41 1079
7ed2c31d
AK
1080static inline bool hugepage_movable_supported(struct hstate *h)
1081{
1082 return false;
1083}
1084
d92bbc27
JK
1085static inline gfp_t htlb_alloc_mask(struct hstate *h)
1086{
1087 return 0;
1088}
1089
19fc7bed
JK
1090static inline gfp_t htlb_modify_alloc_mask(struct hstate *h, gfp_t gfp_mask)
1091{
1092 return 0;
1093}
1094
cb900f41
KS
1095static inline spinlock_t *huge_pte_lockptr(struct hstate *h,
1096 struct mm_struct *mm, pte_t *pte)
1097{
1098 return &mm->page_table_lock;
1099}
5d317b2b 1100
13db8c50
LZ
1101static inline void hugetlb_count_init(struct mm_struct *mm)
1102{
1103}
1104
5d317b2b
NH
1105static inline void hugetlb_report_usage(struct seq_file *f, struct mm_struct *m)
1106{
1107}
1108
1109static inline void hugetlb_count_sub(long l, struct mm_struct *mm)
1110{
1111}
e5251fd4 1112
5d4af619
BW
1113static inline pte_t huge_ptep_clear_flush(struct vm_area_struct *vma,
1114 unsigned long addr, pte_t *ptep)
1115{
1116 return *ptep;
1117}
1118
1119static inline void set_huge_pte_at(struct mm_struct *mm, unsigned long addr,
1120 pte_t *ptep, pte_t pte)
1121{
1122}
a4a00b45
MS
1123
1124static inline void hugetlb_register_node(struct node *node)
1125{
1126}
1127
1128static inline void hugetlb_unregister_node(struct node *node)
1129{
1130}
af73e4d9 1131#endif /* CONFIG_HUGETLB_PAGE */
a5516438 1132
cb900f41
KS
1133static inline spinlock_t *huge_pte_lock(struct hstate *h,
1134 struct mm_struct *mm, pte_t *pte)
1135{
1136 spinlock_t *ptl;
1137
1138 ptl = huge_pte_lockptr(h, mm, pte);
1139 spin_lock(ptl);
1140 return ptl;
1141}
1142
cf11e85f
RG
1143#if defined(CONFIG_HUGETLB_PAGE) && defined(CONFIG_CMA)
1144extern void __init hugetlb_cma_reserve(int order);
cf11e85f
RG
1145#else
1146static inline __init void hugetlb_cma_reserve(int order)
1147{
1148}
cf11e85f
RG
1149#endif
1150
c1991e07
PX
1151bool want_pmd_share(struct vm_area_struct *vma, unsigned long addr);
1152
537cf30b
PX
1153#ifndef __HAVE_ARCH_FLUSH_HUGETLB_TLB_RANGE
1154/*
1155 * ARCHes with special requirements for evicting HUGETLB backing TLB entries can
1156 * implement this.
1157 */
1158#define flush_hugetlb_tlb_range(vma, addr, end) flush_tlb_range(vma, addr, end)
1159#endif
1160
1da177e4 1161#endif /* _LINUX_HUGETLB_H */