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 struct node;
20
21 #ifndef CONFIG_ARCH_HAS_HUGEPD
22 typedef struct { unsigned long pd; } hugepd_t;
23 #define is_hugepd(hugepd) (0)
24 #define __hugepd(x) ((hugepd_t) { (x) })
25 #endif
26
27 #ifdef CONFIG_HUGETLB_PAGE
28
29 #include <linux/mempolicy.h>
30 #include <linux/shm.h>
31 #include <asm/tlbflush.h>
32
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 */
40 enum {
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,
46 #endif
47 #ifdef CONFIG_MEMORY_FAILURE
48 SUBPAGE_INDEX_HWPOISON,
49 #endif
50 __NR_USED_SUBPAGE,
51 };
52
53 struct hugepage_subpool {
54 spinlock_t lock;
55 long count;
56 long max_hpages; /* Maximum huge pages or -1 if no maximum. */
57 long used_hpages; /* Used count against maximum, includes */
58 /* both allocated and reserved pages. */
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 */
62 /* satisfy minimum size. */
63 };
64
65 struct resv_map {
66 struct kref refs;
67 spinlock_t lock;
68 struct list_head regions;
69 long adds_in_progress;
70 struct list_head region_cache;
71 long region_cache_count;
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
82 };
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
92 * indices into the associated mapping. from indicates the starting index
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 */
103 struct 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
118 struct hugetlb_vma_lock {
119 struct kref refs;
120 struct rw_semaphore rw_sema;
121 struct vm_area_struct *vma;
122 };
123
124 extern struct resv_map *resv_map_alloc(void);
125 void resv_map_release(struct kref *ref);
126
127 extern spinlock_t hugetlb_lock;
128 extern int hugetlb_max_hstate __read_mostly;
129 #define for_each_hstate(h) \
130 for ((h) = hstates; (h) < &hstates[hugetlb_max_hstate]; (h)++)
131
132 struct hugepage_subpool *hugepage_new_subpool(struct hstate *h, long max_hpages,
133 long min_hpages);
134 void hugepage_put_subpool(struct hugepage_subpool *spool);
135
136 void hugetlb_dup_vma_private(struct vm_area_struct *vma);
137 void clear_vma_resv_huge_pages(struct vm_area_struct *vma);
138 int hugetlb_sysctl_handler(struct ctl_table *, int, void *, size_t *, loff_t *);
139 int hugetlb_overcommit_handler(struct ctl_table *, int, void *, size_t *,
140 loff_t *);
141 int hugetlb_treat_movable_handler(struct ctl_table *, int, void *, size_t *,
142 loff_t *);
143 int hugetlb_mempolicy_sysctl_handler(struct ctl_table *, int, void *, size_t *,
144 loff_t *);
145
146 int 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);
150 int copy_hugetlb_page_range(struct mm_struct *, struct mm_struct *,
151 struct vm_area_struct *, struct vm_area_struct *);
152 long follow_hugetlb_page(struct mm_struct *, struct vm_area_struct *,
153 struct page **, struct vm_area_struct **,
154 unsigned long *, unsigned long *, long, unsigned int,
155 int *);
156 void unmap_hugepage_range(struct vm_area_struct *,
157 unsigned long, unsigned long, struct page *,
158 zap_flags_t);
159 void __unmap_hugepage_range_final(struct mmu_gather *tlb,
160 struct vm_area_struct *vma,
161 unsigned long start, unsigned long end,
162 struct page *ref_page, zap_flags_t zap_flags);
163 void hugetlb_report_meminfo(struct seq_file *);
164 int hugetlb_report_node_meminfo(char *buf, int len, int nid);
165 void hugetlb_show_meminfo_node(int nid);
166 unsigned long hugetlb_total_pages(void);
167 vm_fault_t hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
168 unsigned long address, unsigned int flags);
169 #ifdef CONFIG_USERFAULTFD
170 int hugetlb_mcopy_atomic_pte(struct mm_struct *dst_mm, pte_t *dst_pte,
171 struct vm_area_struct *dst_vma,
172 unsigned long dst_addr,
173 unsigned long src_addr,
174 enum mcopy_atomic_mode mode,
175 struct page **pagep,
176 bool wp_copy);
177 #endif /* CONFIG_USERFAULTFD */
178 bool hugetlb_reserve_pages(struct inode *inode, long from, long to,
179 struct vm_area_struct *vma,
180 vm_flags_t vm_flags);
181 long hugetlb_unreserve_pages(struct inode *inode, long start, long end,
182 long freed);
183 int isolate_hugetlb(struct page *page, struct list_head *list);
184 int get_hwpoison_huge_page(struct page *page, bool *hugetlb);
185 int get_huge_page_for_hwpoison(unsigned long pfn, int flags);
186 void putback_active_hugepage(struct page *page);
187 void move_hugetlb_state(struct page *oldpage, struct page *newpage, int reason);
188 void free_huge_page(struct page *page);
189 void hugetlb_fix_reserve_counts(struct inode *inode);
190 extern struct mutex *hugetlb_fault_mutex_table;
191 u32 hugetlb_fault_mutex_hash(struct address_space *mapping, pgoff_t idx);
192
193 pte_t *huge_pmd_share(struct mm_struct *mm, struct vm_area_struct *vma,
194 unsigned long addr, pud_t *pud);
195
196 struct address_space *hugetlb_page_mapping_lock_write(struct page *hpage);
197
198 extern int sysctl_hugetlb_shm_group;
199 extern struct list_head huge_boot_pages;
200
201 /* arch callbacks */
202
203 pte_t *huge_pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
204 unsigned long addr, unsigned long sz);
205 pte_t *huge_pte_offset(struct mm_struct *mm,
206 unsigned long addr, unsigned long sz);
207 unsigned long hugetlb_mask_last_page(struct hstate *h);
208 int huge_pmd_unshare(struct mm_struct *mm, struct vm_area_struct *vma,
209 unsigned long addr, pte_t *ptep);
210 void adjust_range_if_pmd_sharing_possible(struct vm_area_struct *vma,
211 unsigned long *start, unsigned long *end);
212 struct page *follow_huge_addr(struct mm_struct *mm, unsigned long address,
213 int write);
214 struct page *follow_huge_pd(struct vm_area_struct *vma,
215 unsigned long address, hugepd_t hpd,
216 int flags, int pdshift);
217 struct page *follow_huge_pmd_pte(struct vm_area_struct *vma, unsigned long address,
218 int flags);
219 struct page *follow_huge_pud(struct mm_struct *mm, unsigned long address,
220 pud_t *pud, int flags);
221 struct page *follow_huge_pgd(struct mm_struct *mm, unsigned long address,
222 pgd_t *pgd, int flags);
223
224 void hugetlb_vma_lock_read(struct vm_area_struct *vma);
225 void hugetlb_vma_unlock_read(struct vm_area_struct *vma);
226 void hugetlb_vma_lock_write(struct vm_area_struct *vma);
227 void hugetlb_vma_unlock_write(struct vm_area_struct *vma);
228 int hugetlb_vma_trylock_write(struct vm_area_struct *vma);
229 void hugetlb_vma_assert_locked(struct vm_area_struct *vma);
230 void hugetlb_vma_lock_release(struct kref *kref);
231
232 int pmd_huge(pmd_t pmd);
233 int pud_huge(pud_t pud);
234 unsigned long hugetlb_change_protection(struct vm_area_struct *vma,
235 unsigned long address, unsigned long end, pgprot_t newprot,
236 unsigned long cp_flags);
237
238 bool is_hugetlb_entry_migration(pte_t pte);
239 void hugetlb_unshare_all_pmds(struct vm_area_struct *vma);
240
241 #else /* !CONFIG_HUGETLB_PAGE */
242
hugetlb_dup_vma_private(struct vm_area_struct * vma)243 static inline void hugetlb_dup_vma_private(struct vm_area_struct *vma)
244 {
245 }
246
clear_vma_resv_huge_pages(struct vm_area_struct * vma)247 static inline void clear_vma_resv_huge_pages(struct vm_area_struct *vma)
248 {
249 }
250
hugetlb_total_pages(void)251 static inline unsigned long hugetlb_total_pages(void)
252 {
253 return 0;
254 }
255
hugetlb_page_mapping_lock_write(struct page * hpage)256 static inline struct address_space *hugetlb_page_mapping_lock_write(
257 struct page *hpage)
258 {
259 return NULL;
260 }
261
huge_pmd_unshare(struct mm_struct * mm,struct vm_area_struct * vma,unsigned long addr,pte_t * ptep)262 static inline int huge_pmd_unshare(struct mm_struct *mm,
263 struct vm_area_struct *vma,
264 unsigned long addr, pte_t *ptep)
265 {
266 return 0;
267 }
268
adjust_range_if_pmd_sharing_possible(struct vm_area_struct * vma,unsigned long * start,unsigned long * end)269 static inline void adjust_range_if_pmd_sharing_possible(
270 struct vm_area_struct *vma,
271 unsigned long *start, unsigned long *end)
272 {
273 }
274
follow_hugetlb_page(struct mm_struct * mm,struct vm_area_struct * vma,struct page ** pages,struct vm_area_struct ** vmas,unsigned long * position,unsigned long * nr_pages,long i,unsigned int flags,int * nonblocking)275 static inline long follow_hugetlb_page(struct mm_struct *mm,
276 struct vm_area_struct *vma, struct page **pages,
277 struct vm_area_struct **vmas, unsigned long *position,
278 unsigned long *nr_pages, long i, unsigned int flags,
279 int *nonblocking)
280 {
281 BUG();
282 return 0;
283 }
284
follow_huge_addr(struct mm_struct * mm,unsigned long address,int write)285 static inline struct page *follow_huge_addr(struct mm_struct *mm,
286 unsigned long address, int write)
287 {
288 return ERR_PTR(-EINVAL);
289 }
290
copy_hugetlb_page_range(struct mm_struct * dst,struct mm_struct * src,struct vm_area_struct * dst_vma,struct vm_area_struct * src_vma)291 static inline int copy_hugetlb_page_range(struct mm_struct *dst,
292 struct mm_struct *src,
293 struct vm_area_struct *dst_vma,
294 struct vm_area_struct *src_vma)
295 {
296 BUG();
297 return 0;
298 }
299
move_hugetlb_page_tables(struct vm_area_struct * vma,struct vm_area_struct * new_vma,unsigned long old_addr,unsigned long new_addr,unsigned long len)300 static inline int move_hugetlb_page_tables(struct vm_area_struct *vma,
301 struct vm_area_struct *new_vma,
302 unsigned long old_addr,
303 unsigned long new_addr,
304 unsigned long len)
305 {
306 BUG();
307 return 0;
308 }
309
hugetlb_report_meminfo(struct seq_file * m)310 static inline void hugetlb_report_meminfo(struct seq_file *m)
311 {
312 }
313
hugetlb_report_node_meminfo(char * buf,int len,int nid)314 static inline int hugetlb_report_node_meminfo(char *buf, int len, int nid)
315 {
316 return 0;
317 }
318
hugetlb_show_meminfo_node(int nid)319 static inline void hugetlb_show_meminfo_node(int nid)
320 {
321 }
322
follow_huge_pd(struct vm_area_struct * vma,unsigned long address,hugepd_t hpd,int flags,int pdshift)323 static inline struct page *follow_huge_pd(struct vm_area_struct *vma,
324 unsigned long address, hugepd_t hpd, int flags,
325 int pdshift)
326 {
327 return NULL;
328 }
329
follow_huge_pmd_pte(struct vm_area_struct * vma,unsigned long address,int flags)330 static inline struct page *follow_huge_pmd_pte(struct vm_area_struct *vma,
331 unsigned long address, int flags)
332 {
333 return NULL;
334 }
335
follow_huge_pud(struct mm_struct * mm,unsigned long address,pud_t * pud,int flags)336 static inline struct page *follow_huge_pud(struct mm_struct *mm,
337 unsigned long address, pud_t *pud, int flags)
338 {
339 return NULL;
340 }
341
follow_huge_pgd(struct mm_struct * mm,unsigned long address,pgd_t * pgd,int flags)342 static inline struct page *follow_huge_pgd(struct mm_struct *mm,
343 unsigned long address, pgd_t *pgd, int flags)
344 {
345 return NULL;
346 }
347
prepare_hugepage_range(struct file * file,unsigned long addr,unsigned long len)348 static inline int prepare_hugepage_range(struct file *file,
349 unsigned long addr, unsigned long len)
350 {
351 return -EINVAL;
352 }
353
hugetlb_vma_lock_read(struct vm_area_struct * vma)354 static inline void hugetlb_vma_lock_read(struct vm_area_struct *vma)
355 {
356 }
357
hugetlb_vma_unlock_read(struct vm_area_struct * vma)358 static inline void hugetlb_vma_unlock_read(struct vm_area_struct *vma)
359 {
360 }
361
hugetlb_vma_lock_write(struct vm_area_struct * vma)362 static inline void hugetlb_vma_lock_write(struct vm_area_struct *vma)
363 {
364 }
365
hugetlb_vma_unlock_write(struct vm_area_struct * vma)366 static inline void hugetlb_vma_unlock_write(struct vm_area_struct *vma)
367 {
368 }
369
hugetlb_vma_trylock_write(struct vm_area_struct * vma)370 static inline int hugetlb_vma_trylock_write(struct vm_area_struct *vma)
371 {
372 return 1;
373 }
374
hugetlb_vma_assert_locked(struct vm_area_struct * vma)375 static inline void hugetlb_vma_assert_locked(struct vm_area_struct *vma)
376 {
377 }
378
pmd_huge(pmd_t pmd)379 static inline int pmd_huge(pmd_t pmd)
380 {
381 return 0;
382 }
383
pud_huge(pud_t pud)384 static inline int pud_huge(pud_t pud)
385 {
386 return 0;
387 }
388
is_hugepage_only_range(struct mm_struct * mm,unsigned long addr,unsigned long len)389 static inline int is_hugepage_only_range(struct mm_struct *mm,
390 unsigned long addr, unsigned long len)
391 {
392 return 0;
393 }
394
hugetlb_free_pgd_range(struct mmu_gather * tlb,unsigned long addr,unsigned long end,unsigned long floor,unsigned long ceiling)395 static inline void hugetlb_free_pgd_range(struct mmu_gather *tlb,
396 unsigned long addr, unsigned long end,
397 unsigned long floor, unsigned long ceiling)
398 {
399 BUG();
400 }
401
402 #ifdef CONFIG_USERFAULTFD
hugetlb_mcopy_atomic_pte(struct mm_struct * dst_mm,pte_t * dst_pte,struct vm_area_struct * dst_vma,unsigned long dst_addr,unsigned long src_addr,enum mcopy_atomic_mode mode,struct page ** pagep,bool wp_copy)403 static inline int hugetlb_mcopy_atomic_pte(struct mm_struct *dst_mm,
404 pte_t *dst_pte,
405 struct vm_area_struct *dst_vma,
406 unsigned long dst_addr,
407 unsigned long src_addr,
408 enum mcopy_atomic_mode mode,
409 struct page **pagep,
410 bool wp_copy)
411 {
412 BUG();
413 return 0;
414 }
415 #endif /* CONFIG_USERFAULTFD */
416
huge_pte_offset(struct mm_struct * mm,unsigned long addr,unsigned long sz)417 static inline pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr,
418 unsigned long sz)
419 {
420 return NULL;
421 }
422
isolate_hugetlb(struct page * page,struct list_head * list)423 static inline int isolate_hugetlb(struct page *page, struct list_head *list)
424 {
425 return -EBUSY;
426 }
427
get_hwpoison_huge_page(struct page * page,bool * hugetlb)428 static inline int get_hwpoison_huge_page(struct page *page, bool *hugetlb)
429 {
430 return 0;
431 }
432
get_huge_page_for_hwpoison(unsigned long pfn,int flags)433 static inline int get_huge_page_for_hwpoison(unsigned long pfn, int flags)
434 {
435 return 0;
436 }
437
putback_active_hugepage(struct page * page)438 static inline void putback_active_hugepage(struct page *page)
439 {
440 }
441
move_hugetlb_state(struct page * oldpage,struct page * newpage,int reason)442 static inline void move_hugetlb_state(struct page *oldpage,
443 struct page *newpage, int reason)
444 {
445 }
446
hugetlb_change_protection(struct vm_area_struct * vma,unsigned long address,unsigned long end,pgprot_t newprot,unsigned long cp_flags)447 static inline unsigned long hugetlb_change_protection(
448 struct vm_area_struct *vma, unsigned long address,
449 unsigned long end, pgprot_t newprot,
450 unsigned long cp_flags)
451 {
452 return 0;
453 }
454
__unmap_hugepage_range_final(struct mmu_gather * tlb,struct vm_area_struct * vma,unsigned long start,unsigned long end,struct page * ref_page,zap_flags_t zap_flags)455 static inline void __unmap_hugepage_range_final(struct mmu_gather *tlb,
456 struct vm_area_struct *vma, unsigned long start,
457 unsigned long end, struct page *ref_page,
458 zap_flags_t zap_flags)
459 {
460 BUG();
461 }
462
hugetlb_fault(struct mm_struct * mm,struct vm_area_struct * vma,unsigned long address,unsigned int flags)463 static inline vm_fault_t hugetlb_fault(struct mm_struct *mm,
464 struct vm_area_struct *vma, unsigned long address,
465 unsigned int flags)
466 {
467 BUG();
468 return 0;
469 }
470
hugetlb_unshare_all_pmds(struct vm_area_struct * vma)471 static inline void hugetlb_unshare_all_pmds(struct vm_area_struct *vma) { }
472
473 #endif /* !CONFIG_HUGETLB_PAGE */
474 /*
475 * hugepages at page global directory. If arch support
476 * hugepages at pgd level, they need to define this.
477 */
478 #ifndef pgd_huge
479 #define pgd_huge(x) 0
480 #endif
481 #ifndef p4d_huge
482 #define p4d_huge(x) 0
483 #endif
484
485 #ifndef pgd_write
pgd_write(pgd_t pgd)486 static inline int pgd_write(pgd_t pgd)
487 {
488 BUG();
489 return 0;
490 }
491 #endif
492
493 #define HUGETLB_ANON_FILE "anon_hugepage"
494
495 enum {
496 /*
497 * The file will be used as an shm file so shmfs accounting rules
498 * apply
499 */
500 HUGETLB_SHMFS_INODE = 1,
501 /*
502 * The file is being created on the internal vfs mount and shmfs
503 * accounting rules do not apply
504 */
505 HUGETLB_ANONHUGE_INODE = 2,
506 };
507
508 #ifdef CONFIG_HUGETLBFS
509 struct hugetlbfs_sb_info {
510 long max_inodes; /* inodes allowed */
511 long free_inodes; /* inodes free */
512 spinlock_t stat_lock;
513 struct hstate *hstate;
514 struct hugepage_subpool *spool;
515 kuid_t uid;
516 kgid_t gid;
517 umode_t mode;
518 };
519
HUGETLBFS_SB(struct super_block * sb)520 static inline struct hugetlbfs_sb_info *HUGETLBFS_SB(struct super_block *sb)
521 {
522 return sb->s_fs_info;
523 }
524
525 struct hugetlbfs_inode_info {
526 struct shared_policy policy;
527 struct inode vfs_inode;
528 unsigned int seals;
529 };
530
HUGETLBFS_I(struct inode * inode)531 static inline struct hugetlbfs_inode_info *HUGETLBFS_I(struct inode *inode)
532 {
533 return container_of(inode, struct hugetlbfs_inode_info, vfs_inode);
534 }
535
536 extern const struct file_operations hugetlbfs_file_operations;
537 extern const struct vm_operations_struct hugetlb_vm_ops;
538 struct file *hugetlb_file_setup(const char *name, size_t size, vm_flags_t acct,
539 int creat_flags, int page_size_log);
540
is_file_hugepages(struct file * file)541 static inline bool is_file_hugepages(struct file *file)
542 {
543 if (file->f_op == &hugetlbfs_file_operations)
544 return true;
545
546 return is_file_shm_hugepages(file);
547 }
548
hstate_inode(struct inode * i)549 static inline struct hstate *hstate_inode(struct inode *i)
550 {
551 return HUGETLBFS_SB(i->i_sb)->hstate;
552 }
553 #else /* !CONFIG_HUGETLBFS */
554
555 #define is_file_hugepages(file) false
556 static inline struct file *
hugetlb_file_setup(const char * name,size_t size,vm_flags_t acctflag,int creat_flags,int page_size_log)557 hugetlb_file_setup(const char *name, size_t size, vm_flags_t acctflag,
558 int creat_flags, int page_size_log)
559 {
560 return ERR_PTR(-ENOSYS);
561 }
562
hstate_inode(struct inode * i)563 static inline struct hstate *hstate_inode(struct inode *i)
564 {
565 return NULL;
566 }
567 #endif /* !CONFIG_HUGETLBFS */
568
569 #ifdef HAVE_ARCH_HUGETLB_UNMAPPED_AREA
570 unsigned long hugetlb_get_unmapped_area(struct file *file, unsigned long addr,
571 unsigned long len, unsigned long pgoff,
572 unsigned long flags);
573 #endif /* HAVE_ARCH_HUGETLB_UNMAPPED_AREA */
574
575 unsigned long
576 generic_hugetlb_get_unmapped_area(struct file *file, unsigned long addr,
577 unsigned long len, unsigned long pgoff,
578 unsigned long flags);
579
580 /*
581 * huegtlb page specific state flags. These flags are located in page.private
582 * of the hugetlb head page. Functions created via the below macros should be
583 * used to manipulate these flags.
584 *
585 * HPG_restore_reserve - Set when a hugetlb page consumes a reservation at
586 * allocation time. Cleared when page is fully instantiated. Free
587 * routine checks flag to restore a reservation on error paths.
588 * Synchronization: Examined or modified by code that knows it has
589 * the only reference to page. i.e. After allocation but before use
590 * or when the page is being freed.
591 * HPG_migratable - Set after a newly allocated page is added to the page
592 * cache and/or page tables. Indicates the page is a candidate for
593 * migration.
594 * Synchronization: Initially set after new page allocation with no
595 * locking. When examined and modified during migration processing
596 * (isolate, migrate, putback) the hugetlb_lock is held.
597 * HPG_temporary - Set on a page that is temporarily allocated from the buddy
598 * allocator. Typically used for migration target pages when no pages
599 * are available in the pool. The hugetlb free page path will
600 * immediately free pages with this flag set to the buddy allocator.
601 * Synchronization: Can be set after huge page allocation from buddy when
602 * code knows it has only reference. All other examinations and
603 * modifications require hugetlb_lock.
604 * HPG_freed - Set when page is on the free lists.
605 * Synchronization: hugetlb_lock held for examination and modification.
606 * HPG_vmemmap_optimized - Set when the vmemmap pages of the page are freed.
607 * HPG_raw_hwp_unreliable - Set when the hugetlb page has a hwpoison sub-page
608 * that is not tracked by raw_hwp_page list.
609 */
610 enum hugetlb_page_flags {
611 HPG_restore_reserve = 0,
612 HPG_migratable,
613 HPG_temporary,
614 HPG_freed,
615 HPG_vmemmap_optimized,
616 HPG_raw_hwp_unreliable,
617 __NR_HPAGEFLAGS,
618 };
619
620 /*
621 * Macros to create test, set and clear function definitions for
622 * hugetlb specific page flags.
623 */
624 #ifdef CONFIG_HUGETLB_PAGE
625 #define TESTHPAGEFLAG(uname, flname) \
626 static inline int HPage##uname(struct page *page) \
627 { return test_bit(HPG_##flname, &(page->private)); }
628
629 #define SETHPAGEFLAG(uname, flname) \
630 static inline void SetHPage##uname(struct page *page) \
631 { set_bit(HPG_##flname, &(page->private)); }
632
633 #define CLEARHPAGEFLAG(uname, flname) \
634 static inline void ClearHPage##uname(struct page *page) \
635 { clear_bit(HPG_##flname, &(page->private)); }
636 #else
637 #define TESTHPAGEFLAG(uname, flname) \
638 static inline int HPage##uname(struct page *page) \
639 { return 0; }
640
641 #define SETHPAGEFLAG(uname, flname) \
642 static inline void SetHPage##uname(struct page *page) \
643 { }
644
645 #define CLEARHPAGEFLAG(uname, flname) \
646 static inline void ClearHPage##uname(struct page *page) \
647 { }
648 #endif
649
650 #define HPAGEFLAG(uname, flname) \
651 TESTHPAGEFLAG(uname, flname) \
652 SETHPAGEFLAG(uname, flname) \
653 CLEARHPAGEFLAG(uname, flname) \
654
655 /*
656 * Create functions associated with hugetlb page flags
657 */
658 HPAGEFLAG(RestoreReserve, restore_reserve)
659 HPAGEFLAG(Migratable, migratable)
660 HPAGEFLAG(Temporary, temporary)
661 HPAGEFLAG(Freed, freed)
662 HPAGEFLAG(VmemmapOptimized, vmemmap_optimized)
663 HPAGEFLAG(RawHwpUnreliable, raw_hwp_unreliable)
664
665 #ifdef CONFIG_HUGETLB_PAGE
666
667 #define HSTATE_NAME_LEN 32
668 /* Defines one hugetlb page size */
669 struct hstate {
670 struct mutex resize_lock;
671 int next_nid_to_alloc;
672 int next_nid_to_free;
673 unsigned int order;
674 unsigned int demote_order;
675 unsigned long mask;
676 unsigned long max_huge_pages;
677 unsigned long nr_huge_pages;
678 unsigned long free_huge_pages;
679 unsigned long resv_huge_pages;
680 unsigned long surplus_huge_pages;
681 unsigned long nr_overcommit_huge_pages;
682 struct list_head hugepage_activelist;
683 struct list_head hugepage_freelists[MAX_NUMNODES];
684 unsigned int max_huge_pages_node[MAX_NUMNODES];
685 unsigned int nr_huge_pages_node[MAX_NUMNODES];
686 unsigned int free_huge_pages_node[MAX_NUMNODES];
687 unsigned int surplus_huge_pages_node[MAX_NUMNODES];
688 #ifdef CONFIG_CGROUP_HUGETLB
689 /* cgroup control files */
690 struct cftype cgroup_files_dfl[8];
691 struct cftype cgroup_files_legacy[10];
692 #endif
693 char name[HSTATE_NAME_LEN];
694 };
695
696 struct huge_bootmem_page {
697 struct list_head list;
698 struct hstate *hstate;
699 };
700
701 int isolate_or_dissolve_huge_page(struct page *page, struct list_head *list);
702 struct page *alloc_huge_page(struct vm_area_struct *vma,
703 unsigned long addr, int avoid_reserve);
704 struct page *alloc_huge_page_nodemask(struct hstate *h, int preferred_nid,
705 nodemask_t *nmask, gfp_t gfp_mask);
706 struct page *alloc_huge_page_vma(struct hstate *h, struct vm_area_struct *vma,
707 unsigned long address);
708 int hugetlb_add_to_page_cache(struct page *page, struct address_space *mapping,
709 pgoff_t idx);
710 void restore_reserve_on_error(struct hstate *h, struct vm_area_struct *vma,
711 unsigned long address, struct page *page);
712
713 /* arch callback */
714 int __init __alloc_bootmem_huge_page(struct hstate *h, int nid);
715 int __init alloc_bootmem_huge_page(struct hstate *h, int nid);
716 bool __init hugetlb_node_alloc_supported(void);
717
718 void __init hugetlb_add_hstate(unsigned order);
719 bool __init arch_hugetlb_valid_size(unsigned long size);
720 struct hstate *size_to_hstate(unsigned long size);
721
722 #ifndef HUGE_MAX_HSTATE
723 #define HUGE_MAX_HSTATE 1
724 #endif
725
726 extern struct hstate hstates[HUGE_MAX_HSTATE];
727 extern unsigned int default_hstate_idx;
728
729 #define default_hstate (hstates[default_hstate_idx])
730
731 /*
732 * hugetlb page subpool pointer located in hpage[1].private
733 */
hugetlb_page_subpool(struct page * hpage)734 static inline struct hugepage_subpool *hugetlb_page_subpool(struct page *hpage)
735 {
736 return (void *)page_private(hpage + SUBPAGE_INDEX_SUBPOOL);
737 }
738
hugetlb_set_page_subpool(struct page * hpage,struct hugepage_subpool * subpool)739 static inline void hugetlb_set_page_subpool(struct page *hpage,
740 struct hugepage_subpool *subpool)
741 {
742 set_page_private(hpage + SUBPAGE_INDEX_SUBPOOL, (unsigned long)subpool);
743 }
744
hstate_file(struct file * f)745 static inline struct hstate *hstate_file(struct file *f)
746 {
747 return hstate_inode(file_inode(f));
748 }
749
hstate_sizelog(int page_size_log)750 static inline struct hstate *hstate_sizelog(int page_size_log)
751 {
752 if (!page_size_log)
753 return &default_hstate;
754
755 return size_to_hstate(1UL << page_size_log);
756 }
757
hstate_vma(struct vm_area_struct * vma)758 static inline struct hstate *hstate_vma(struct vm_area_struct *vma)
759 {
760 return hstate_file(vma->vm_file);
761 }
762
huge_page_size(const struct hstate * h)763 static inline unsigned long huge_page_size(const struct hstate *h)
764 {
765 return (unsigned long)PAGE_SIZE << h->order;
766 }
767
768 extern unsigned long vma_kernel_pagesize(struct vm_area_struct *vma);
769
770 extern unsigned long vma_mmu_pagesize(struct vm_area_struct *vma);
771
huge_page_mask(struct hstate * h)772 static inline unsigned long huge_page_mask(struct hstate *h)
773 {
774 return h->mask;
775 }
776
huge_page_order(struct hstate * h)777 static inline unsigned int huge_page_order(struct hstate *h)
778 {
779 return h->order;
780 }
781
huge_page_shift(struct hstate * h)782 static inline unsigned huge_page_shift(struct hstate *h)
783 {
784 return h->order + PAGE_SHIFT;
785 }
786
hstate_is_gigantic(struct hstate * h)787 static inline bool hstate_is_gigantic(struct hstate *h)
788 {
789 return huge_page_order(h) >= MAX_ORDER;
790 }
791
pages_per_huge_page(const struct hstate * h)792 static inline unsigned int pages_per_huge_page(const struct hstate *h)
793 {
794 return 1 << h->order;
795 }
796
blocks_per_huge_page(struct hstate * h)797 static inline unsigned int blocks_per_huge_page(struct hstate *h)
798 {
799 return huge_page_size(h) / 512;
800 }
801
802 #include <asm/hugetlb.h>
803
804 #ifndef is_hugepage_only_range
is_hugepage_only_range(struct mm_struct * mm,unsigned long addr,unsigned long len)805 static inline int is_hugepage_only_range(struct mm_struct *mm,
806 unsigned long addr, unsigned long len)
807 {
808 return 0;
809 }
810 #define is_hugepage_only_range is_hugepage_only_range
811 #endif
812
813 #ifndef arch_clear_hugepage_flags
arch_clear_hugepage_flags(struct page * page)814 static inline void arch_clear_hugepage_flags(struct page *page) { }
815 #define arch_clear_hugepage_flags arch_clear_hugepage_flags
816 #endif
817
818 #ifndef arch_make_huge_pte
arch_make_huge_pte(pte_t entry,unsigned int shift,vm_flags_t flags)819 static inline pte_t arch_make_huge_pte(pte_t entry, unsigned int shift,
820 vm_flags_t flags)
821 {
822 return pte_mkhuge(entry);
823 }
824 #endif
825
page_hstate(struct page * page)826 static inline struct hstate *page_hstate(struct page *page)
827 {
828 VM_BUG_ON_PAGE(!PageHuge(page), page);
829 return size_to_hstate(page_size(page));
830 }
831
hstate_index_to_shift(unsigned index)832 static inline unsigned hstate_index_to_shift(unsigned index)
833 {
834 return hstates[index].order + PAGE_SHIFT;
835 }
836
hstate_index(struct hstate * h)837 static inline int hstate_index(struct hstate *h)
838 {
839 return h - hstates;
840 }
841
842 extern int dissolve_free_huge_page(struct page *page);
843 extern int dissolve_free_huge_pages(unsigned long start_pfn,
844 unsigned long end_pfn);
845
846 #ifdef CONFIG_MEMORY_FAILURE
847 extern void hugetlb_clear_page_hwpoison(struct page *hpage);
848 #else
hugetlb_clear_page_hwpoison(struct page * hpage)849 static inline void hugetlb_clear_page_hwpoison(struct page *hpage)
850 {
851 }
852 #endif
853
854 #ifdef CONFIG_ARCH_ENABLE_HUGEPAGE_MIGRATION
855 #ifndef arch_hugetlb_migration_supported
arch_hugetlb_migration_supported(struct hstate * h)856 static inline bool arch_hugetlb_migration_supported(struct hstate *h)
857 {
858 if ((huge_page_shift(h) == PMD_SHIFT) ||
859 (huge_page_shift(h) == PUD_SHIFT) ||
860 (huge_page_shift(h) == PGDIR_SHIFT))
861 return true;
862 else
863 return false;
864 }
865 #endif
866 #else
arch_hugetlb_migration_supported(struct hstate * h)867 static inline bool arch_hugetlb_migration_supported(struct hstate *h)
868 {
869 return false;
870 }
871 #endif
872
hugepage_migration_supported(struct hstate * h)873 static inline bool hugepage_migration_supported(struct hstate *h)
874 {
875 return arch_hugetlb_migration_supported(h);
876 }
877
878 /*
879 * Movability check is different as compared to migration check.
880 * It determines whether or not a huge page should be placed on
881 * movable zone or not. Movability of any huge page should be
882 * required only if huge page size is supported for migration.
883 * There won't be any reason for the huge page to be movable if
884 * it is not migratable to start with. Also the size of the huge
885 * page should be large enough to be placed under a movable zone
886 * and still feasible enough to be migratable. Just the presence
887 * in movable zone does not make the migration feasible.
888 *
889 * So even though large huge page sizes like the gigantic ones
890 * are migratable they should not be movable because its not
891 * feasible to migrate them from movable zone.
892 */
hugepage_movable_supported(struct hstate * h)893 static inline bool hugepage_movable_supported(struct hstate *h)
894 {
895 if (!hugepage_migration_supported(h))
896 return false;
897
898 if (hstate_is_gigantic(h))
899 return false;
900 return true;
901 }
902
903 /* Movability of hugepages depends on migration support. */
htlb_alloc_mask(struct hstate * h)904 static inline gfp_t htlb_alloc_mask(struct hstate *h)
905 {
906 if (hugepage_movable_supported(h))
907 return GFP_HIGHUSER_MOVABLE;
908 else
909 return GFP_HIGHUSER;
910 }
911
htlb_modify_alloc_mask(struct hstate * h,gfp_t gfp_mask)912 static inline gfp_t htlb_modify_alloc_mask(struct hstate *h, gfp_t gfp_mask)
913 {
914 gfp_t modified_mask = htlb_alloc_mask(h);
915
916 /* Some callers might want to enforce node */
917 modified_mask |= (gfp_mask & __GFP_THISNODE);
918
919 modified_mask |= (gfp_mask & __GFP_NOWARN);
920
921 return modified_mask;
922 }
923
huge_pte_lockptr(struct hstate * h,struct mm_struct * mm,pte_t * pte)924 static inline spinlock_t *huge_pte_lockptr(struct hstate *h,
925 struct mm_struct *mm, pte_t *pte)
926 {
927 if (huge_page_size(h) == PMD_SIZE)
928 return pmd_lockptr(mm, (pmd_t *) pte);
929 VM_BUG_ON(huge_page_size(h) == PAGE_SIZE);
930 return &mm->page_table_lock;
931 }
932
933 #ifndef hugepages_supported
934 /*
935 * Some platform decide whether they support huge pages at boot
936 * time. Some of them, such as powerpc, set HPAGE_SHIFT to 0
937 * when there is no such support
938 */
939 #define hugepages_supported() (HPAGE_SHIFT != 0)
940 #endif
941
942 void hugetlb_report_usage(struct seq_file *m, struct mm_struct *mm);
943
hugetlb_count_init(struct mm_struct * mm)944 static inline void hugetlb_count_init(struct mm_struct *mm)
945 {
946 atomic_long_set(&mm->hugetlb_usage, 0);
947 }
948
hugetlb_count_add(long l,struct mm_struct * mm)949 static inline void hugetlb_count_add(long l, struct mm_struct *mm)
950 {
951 atomic_long_add(l, &mm->hugetlb_usage);
952 }
953
hugetlb_count_sub(long l,struct mm_struct * mm)954 static inline void hugetlb_count_sub(long l, struct mm_struct *mm)
955 {
956 atomic_long_sub(l, &mm->hugetlb_usage);
957 }
958
959 #ifndef huge_ptep_modify_prot_start
960 #define huge_ptep_modify_prot_start huge_ptep_modify_prot_start
huge_ptep_modify_prot_start(struct vm_area_struct * vma,unsigned long addr,pte_t * ptep)961 static inline pte_t huge_ptep_modify_prot_start(struct vm_area_struct *vma,
962 unsigned long addr, pte_t *ptep)
963 {
964 return huge_ptep_get_and_clear(vma->vm_mm, addr, ptep);
965 }
966 #endif
967
968 #ifndef huge_ptep_modify_prot_commit
969 #define huge_ptep_modify_prot_commit huge_ptep_modify_prot_commit
huge_ptep_modify_prot_commit(struct vm_area_struct * vma,unsigned long addr,pte_t * ptep,pte_t old_pte,pte_t pte)970 static inline void huge_ptep_modify_prot_commit(struct vm_area_struct *vma,
971 unsigned long addr, pte_t *ptep,
972 pte_t old_pte, pte_t pte)
973 {
974 set_huge_pte_at(vma->vm_mm, addr, ptep, pte);
975 }
976 #endif
977
978 #ifdef CONFIG_NUMA
979 void hugetlb_register_node(struct node *node);
980 void hugetlb_unregister_node(struct node *node);
981 #endif
982
983 #else /* CONFIG_HUGETLB_PAGE */
984 struct hstate {};
985
986 static inline struct hugepage_subpool *hugetlb_page_subpool(struct page *hpage)
987 {
988 return NULL;
989 }
990
991 static inline int isolate_or_dissolve_huge_page(struct page *page,
992 struct list_head *list)
993 {
994 return -ENOMEM;
995 }
996
997 static inline struct page *alloc_huge_page(struct vm_area_struct *vma,
998 unsigned long addr,
999 int avoid_reserve)
1000 {
1001 return NULL;
1002 }
1003
1004 static inline struct page *
1005 alloc_huge_page_nodemask(struct hstate *h, int preferred_nid,
1006 nodemask_t *nmask, gfp_t gfp_mask)
1007 {
1008 return NULL;
1009 }
1010
1011 static inline struct page *alloc_huge_page_vma(struct hstate *h,
1012 struct vm_area_struct *vma,
1013 unsigned long address)
1014 {
1015 return NULL;
1016 }
1017
1018 static inline int __alloc_bootmem_huge_page(struct hstate *h)
1019 {
1020 return 0;
1021 }
1022
1023 static inline struct hstate *hstate_file(struct file *f)
1024 {
1025 return NULL;
1026 }
1027
1028 static inline struct hstate *hstate_sizelog(int page_size_log)
1029 {
1030 return NULL;
1031 }
1032
1033 static inline struct hstate *hstate_vma(struct vm_area_struct *vma)
1034 {
1035 return NULL;
1036 }
1037
1038 static inline struct hstate *page_hstate(struct page *page)
1039 {
1040 return NULL;
1041 }
1042
1043 static inline struct hstate *size_to_hstate(unsigned long size)
1044 {
1045 return NULL;
1046 }
1047
1048 static inline unsigned long huge_page_size(struct hstate *h)
1049 {
1050 return PAGE_SIZE;
1051 }
1052
1053 static inline unsigned long huge_page_mask(struct hstate *h)
1054 {
1055 return PAGE_MASK;
1056 }
1057
1058 static inline unsigned long vma_kernel_pagesize(struct vm_area_struct *vma)
1059 {
1060 return PAGE_SIZE;
1061 }
1062
1063 static inline unsigned long vma_mmu_pagesize(struct vm_area_struct *vma)
1064 {
1065 return PAGE_SIZE;
1066 }
1067
1068 static inline unsigned int huge_page_order(struct hstate *h)
1069 {
1070 return 0;
1071 }
1072
1073 static inline unsigned int huge_page_shift(struct hstate *h)
1074 {
1075 return PAGE_SHIFT;
1076 }
1077
1078 static inline bool hstate_is_gigantic(struct hstate *h)
1079 {
1080 return false;
1081 }
1082
1083 static inline unsigned int pages_per_huge_page(struct hstate *h)
1084 {
1085 return 1;
1086 }
1087
1088 static inline unsigned hstate_index_to_shift(unsigned index)
1089 {
1090 return 0;
1091 }
1092
1093 static inline int hstate_index(struct hstate *h)
1094 {
1095 return 0;
1096 }
1097
1098 static inline int dissolve_free_huge_page(struct page *page)
1099 {
1100 return 0;
1101 }
1102
1103 static inline int dissolve_free_huge_pages(unsigned long start_pfn,
1104 unsigned long end_pfn)
1105 {
1106 return 0;
1107 }
1108
1109 static inline bool hugepage_migration_supported(struct hstate *h)
1110 {
1111 return false;
1112 }
1113
1114 static inline bool hugepage_movable_supported(struct hstate *h)
1115 {
1116 return false;
1117 }
1118
1119 static inline gfp_t htlb_alloc_mask(struct hstate *h)
1120 {
1121 return 0;
1122 }
1123
1124 static inline gfp_t htlb_modify_alloc_mask(struct hstate *h, gfp_t gfp_mask)
1125 {
1126 return 0;
1127 }
1128
1129 static inline spinlock_t *huge_pte_lockptr(struct hstate *h,
1130 struct mm_struct *mm, pte_t *pte)
1131 {
1132 return &mm->page_table_lock;
1133 }
1134
1135 static inline void hugetlb_count_init(struct mm_struct *mm)
1136 {
1137 }
1138
1139 static inline void hugetlb_report_usage(struct seq_file *f, struct mm_struct *m)
1140 {
1141 }
1142
1143 static inline void hugetlb_count_sub(long l, struct mm_struct *mm)
1144 {
1145 }
1146
1147 static inline pte_t huge_ptep_clear_flush(struct vm_area_struct *vma,
1148 unsigned long addr, pte_t *ptep)
1149 {
1150 return *ptep;
1151 }
1152
1153 static inline void set_huge_pte_at(struct mm_struct *mm, unsigned long addr,
1154 pte_t *ptep, pte_t pte)
1155 {
1156 }
1157
1158 static inline void hugetlb_register_node(struct node *node)
1159 {
1160 }
1161
1162 static inline void hugetlb_unregister_node(struct node *node)
1163 {
1164 }
1165 #endif /* CONFIG_HUGETLB_PAGE */
1166
huge_pte_lock(struct hstate * h,struct mm_struct * mm,pte_t * pte)1167 static inline spinlock_t *huge_pte_lock(struct hstate *h,
1168 struct mm_struct *mm, pte_t *pte)
1169 {
1170 spinlock_t *ptl;
1171
1172 ptl = huge_pte_lockptr(h, mm, pte);
1173 spin_lock(ptl);
1174 return ptl;
1175 }
1176
1177 #if defined(CONFIG_HUGETLB_PAGE) && defined(CONFIG_CMA)
1178 extern void __init hugetlb_cma_reserve(int order);
1179 #else
hugetlb_cma_reserve(int order)1180 static inline __init void hugetlb_cma_reserve(int order)
1181 {
1182 }
1183 #endif
1184
1185 bool want_pmd_share(struct vm_area_struct *vma, unsigned long addr);
1186
1187 #ifndef __HAVE_ARCH_FLUSH_HUGETLB_TLB_RANGE
1188 /*
1189 * ARCHes with special requirements for evicting HUGETLB backing TLB entries can
1190 * implement this.
1191 */
1192 #define flush_hugetlb_tlb_range(vma, addr, end) flush_tlb_range(vma, addr, end)
1193 #endif
1194
1195 #endif /* _LINUX_HUGETLB_H */
1196