Commit | Line | Data |
---|---|---|
b2441318 | 1 | /* SPDX-License-Identifier: GPL-2.0 */ |
1da177e4 LT |
2 | #ifndef _LINUX_HUGETLB_H |
3 | #define _LINUX_HUGETLB_H | |
4 | ||
9c67a207 | 5 | #include <linux/mm.h> |
be93d8cf | 6 | #include <linux/mm_types.h> |
309381fe | 7 | #include <linux/mmdebug.h> |
4e950f6f | 8 | #include <linux/fs.h> |
8edf344c | 9 | #include <linux/hugetlb_inline.h> |
abb8206c | 10 | #include <linux/cgroup.h> |
3489dbb6 | 11 | #include <linux/page_ref.h> |
9119a41e JK |
12 | #include <linux/list.h> |
13 | #include <linux/kref.h> | |
ca5999fd | 14 | #include <linux/pgtable.h> |
d92bbc27 | 15 | #include <linux/gfp.h> |
f6191471 | 16 | #include <linux/userfaultfd_k.h> |
4e950f6f | 17 | |
e9ea0e2d AM |
18 | struct ctl_table; |
19 | struct user_struct; | |
24669e58 | 20 | struct mmu_gather; |
a4a00b45 | 21 | struct node; |
e9ea0e2d | 22 | |
691cdf01 | 23 | #ifndef CONFIG_ARCH_HAS_HUGEPD |
e2299292 AK |
24 | typedef struct { unsigned long pd; } hugepd_t; |
25 | #define is_hugepd(hugepd) (0) | |
26 | #define __hugepd(x) ((hugepd_t) { (x) }) | |
e2299292 AK |
27 | #endif |
28 | ||
454a00c4 | 29 | void free_huge_folio(struct folio *folio); |
dd6fa0b6 | 30 | |
1da177e4 LT |
31 | #ifdef CONFIG_HUGETLB_PAGE |
32 | ||
10969b55 | 33 | #include <linux/pagemap.h> |
516dffdc | 34 | #include <linux/shm.h> |
63551ae0 | 35 | #include <asm/tlbflush.h> |
1da177e4 | 36 | |
cd39d4e9 MS |
37 | /* |
38 | * For HugeTLB page, there are more metadata to save in the struct page. But | |
39 | * the head struct page cannot meet our needs, so we have to abuse other tail | |
dad6a5eb | 40 | * struct page to store the metadata. |
cd39d4e9 | 41 | */ |
dad6a5eb | 42 | #define __NR_USED_SUBPAGE 3 |
cd39d4e9 | 43 | |
90481622 DG |
44 | struct hugepage_subpool { |
45 | spinlock_t lock; | |
46 | long count; | |
c6a91820 MK |
47 | long max_hpages; /* Maximum huge pages or -1 if no maximum. */ |
48 | long used_hpages; /* Used count against maximum, includes */ | |
06c88398 | 49 | /* both allocated and reserved pages. */ |
c6a91820 MK |
50 | struct hstate *hstate; |
51 | long min_hpages; /* Minimum huge pages or -1 if no minimum. */ | |
52 | long rsv_hpages; /* Pages reserved against global pool to */ | |
6c26d310 | 53 | /* satisfy minimum size. */ |
90481622 DG |
54 | }; |
55 | ||
9119a41e JK |
56 | struct resv_map { |
57 | struct kref refs; | |
7b24d861 | 58 | spinlock_t lock; |
9119a41e | 59 | struct list_head regions; |
5e911373 MK |
60 | long adds_in_progress; |
61 | struct list_head region_cache; | |
62 | long region_cache_count; | |
bf491692 | 63 | struct rw_semaphore rw_sema; |
e9fe92ae MA |
64 | #ifdef CONFIG_CGROUP_HUGETLB |
65 | /* | |
66 | * On private mappings, the counter to uncharge reservations is stored | |
67 | * here. If these fields are 0, then either the mapping is shared, or | |
68 | * cgroup accounting is disabled for this resv_map. | |
69 | */ | |
70 | struct page_counter *reservation_counter; | |
71 | unsigned long pages_per_hpage; | |
72 | struct cgroup_subsys_state *css; | |
73 | #endif | |
9119a41e | 74 | }; |
075a61d0 MA |
75 | |
76 | /* | |
77 | * Region tracking -- allows tracking of reservations and instantiated pages | |
78 | * across the pages in a mapping. | |
79 | * | |
80 | * The region data structures are embedded into a resv_map and protected | |
81 | * by a resv_map's lock. The set of regions within the resv_map represent | |
82 | * reservations for huge pages, or huge pages that have already been | |
83 | * instantiated within the map. The from and to elements are huge page | |
06c88398 | 84 | * indices into the associated mapping. from indicates the starting index |
075a61d0 MA |
85 | * of the region. to represents the first index past the end of the region. |
86 | * | |
87 | * For example, a file region structure with from == 0 and to == 4 represents | |
88 | * four huge pages in a mapping. It is important to note that the to element | |
89 | * represents the first element past the end of the region. This is used in | |
90 | * arithmetic as 4(to) - 0(from) = 4 huge pages in the region. | |
91 | * | |
92 | * Interval notation of the form [from, to) will be used to indicate that | |
93 | * the endpoint from is inclusive and to is exclusive. | |
94 | */ | |
95 | struct file_region { | |
96 | struct list_head link; | |
97 | long from; | |
98 | long to; | |
99 | #ifdef CONFIG_CGROUP_HUGETLB | |
100 | /* | |
101 | * On shared mappings, each reserved region appears as a struct | |
102 | * file_region in resv_map. These fields hold the info needed to | |
103 | * uncharge each reservation. | |
104 | */ | |
105 | struct page_counter *reservation_counter; | |
106 | struct cgroup_subsys_state *css; | |
107 | #endif | |
108 | }; | |
109 | ||
8d9bfb26 MK |
110 | struct hugetlb_vma_lock { |
111 | struct kref refs; | |
112 | struct rw_semaphore rw_sema; | |
113 | struct vm_area_struct *vma; | |
114 | }; | |
115 | ||
9119a41e JK |
116 | extern struct resv_map *resv_map_alloc(void); |
117 | void resv_map_release(struct kref *ref); | |
118 | ||
c3f38a38 AK |
119 | extern spinlock_t hugetlb_lock; |
120 | extern int hugetlb_max_hstate __read_mostly; | |
121 | #define for_each_hstate(h) \ | |
122 | for ((h) = hstates; (h) < &hstates[hugetlb_max_hstate]; (h)++) | |
123 | ||
7ca02d0a MK |
124 | struct hugepage_subpool *hugepage_new_subpool(struct hstate *h, long max_hpages, |
125 | long min_hpages); | |
90481622 DG |
126 | void hugepage_put_subpool(struct hugepage_subpool *spool); |
127 | ||
8d9bfb26 | 128 | void hugetlb_dup_vma_private(struct vm_area_struct *vma); |
550a7d60 | 129 | void clear_vma_resv_huge_pages(struct vm_area_struct *vma); |
550a7d60 MA |
130 | int move_hugetlb_page_tables(struct vm_area_struct *vma, |
131 | struct vm_area_struct *new_vma, | |
132 | unsigned long old_addr, unsigned long new_addr, | |
133 | unsigned long len); | |
bc70fbf2 PX |
134 | int copy_hugetlb_page_range(struct mm_struct *, struct mm_struct *, |
135 | struct vm_area_struct *, struct vm_area_struct *); | |
57a196a5 | 136 | struct page *hugetlb_follow_page_mask(struct vm_area_struct *vma, |
5502ea44 PX |
137 | unsigned long address, unsigned int flags, |
138 | unsigned int *page_mask); | |
04f2cbe3 | 139 | void unmap_hugepage_range(struct vm_area_struct *, |
05e90bd0 PX |
140 | unsigned long, unsigned long, struct page *, |
141 | zap_flags_t); | |
2820b0f0 | 142 | void __unmap_hugepage_range(struct mmu_gather *tlb, |
d833352a MG |
143 | struct vm_area_struct *vma, |
144 | unsigned long start, unsigned long end, | |
05e90bd0 | 145 | struct page *ref_page, zap_flags_t zap_flags); |
e1759c21 | 146 | void hugetlb_report_meminfo(struct seq_file *); |
7981593b | 147 | int hugetlb_report_node_meminfo(char *buf, int len, int nid); |
dcadcf1c | 148 | void hugetlb_show_meminfo_node(int nid); |
1da177e4 | 149 | unsigned long hugetlb_total_pages(void); |
2b740303 | 150 | vm_fault_t hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma, |
788c7df4 | 151 | unsigned long address, unsigned int flags); |
714c1891 | 152 | #ifdef CONFIG_USERFAULTFD |
61c50040 | 153 | int hugetlb_mfill_atomic_pte(pte_t *dst_pte, |
a734991c AR |
154 | struct vm_area_struct *dst_vma, |
155 | unsigned long dst_addr, | |
156 | unsigned long src_addr, | |
d9712937 | 157 | uffd_flags_t flags, |
0169fd51 | 158 | struct folio **foliop); |
714c1891 | 159 | #endif /* CONFIG_USERFAULTFD */ |
33b8f84a | 160 | bool hugetlb_reserve_pages(struct inode *inode, long from, long to, |
5a6fe125 | 161 | struct vm_area_struct *vma, |
ca16d140 | 162 | vm_flags_t vm_flags); |
b5cec28d MK |
163 | long hugetlb_unreserve_pages(struct inode *inode, long start, long end, |
164 | long freed); | |
9747b9e9 | 165 | bool isolate_hugetlb(struct folio *folio, struct list_head *list); |
04bac040 | 166 | int get_hwpoison_hugetlb_folio(struct folio *folio, bool *hugetlb, bool unpoison); |
e591ef7d NH |
167 | int get_huge_page_for_hwpoison(unsigned long pfn, int flags, |
168 | bool *migratable_cleared); | |
ea8e72f4 | 169 | void folio_putback_active_hugetlb(struct folio *folio); |
345c62d1 | 170 | void move_hugetlb_state(struct folio *old_folio, struct folio *new_folio, int reason); |
72e2936c | 171 | void hugetlb_fix_reserve_counts(struct inode *inode); |
c672c7f2 | 172 | extern struct mutex *hugetlb_fault_mutex_table; |
188b04a7 | 173 | u32 hugetlb_fault_mutex_hash(struct address_space *mapping, pgoff_t idx); |
1da177e4 | 174 | |
aec44e0f PX |
175 | pte_t *huge_pmd_share(struct mm_struct *mm, struct vm_area_struct *vma, |
176 | unsigned long addr, pud_t *pud); | |
24334e78 PX |
177 | bool hugetlbfs_pagecache_present(struct hstate *h, |
178 | struct vm_area_struct *vma, | |
179 | unsigned long address); | |
3212b535 | 180 | |
6e8cda4c | 181 | struct address_space *hugetlb_folio_mapping_lock_write(struct folio *folio); |
c0d0381a | 182 | |
1da177e4 | 183 | extern int sysctl_hugetlb_shm_group; |
b78b27d0 | 184 | extern struct list_head huge_boot_pages[MAX_NUMNODES]; |
1da177e4 | 185 | |
63551ae0 DG |
186 | /* arch callbacks */ |
187 | ||
f7243924 HD |
188 | #ifndef CONFIG_HIGHPTE |
189 | /* | |
190 | * pte_offset_huge() and pte_alloc_huge() are helpers for those architectures | |
191 | * which may go down to the lowest PTE level in their huge_pte_offset() and | |
192 | * huge_pte_alloc(): to avoid reliance on pte_offset_map() without pte_unmap(). | |
193 | */ | |
194 | static inline pte_t *pte_offset_huge(pmd_t *pmd, unsigned long address) | |
195 | { | |
196 | return pte_offset_kernel(pmd, address); | |
197 | } | |
198 | static inline pte_t *pte_alloc_huge(struct mm_struct *mm, pmd_t *pmd, | |
199 | unsigned long address) | |
200 | { | |
201 | return pte_alloc(mm, pmd) ? NULL : pte_offset_huge(pmd, address); | |
202 | } | |
203 | #endif | |
204 | ||
aec44e0f | 205 | pte_t *huge_pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma, |
a5516438 | 206 | unsigned long addr, unsigned long sz); |
fe7d4c6d PX |
207 | /* |
208 | * huge_pte_offset(): Walk the hugetlb pgtable until the last level PTE. | |
209 | * Returns the pte_t* if found, or NULL if the address is not mapped. | |
210 | * | |
9c67a207 PX |
211 | * IMPORTANT: we should normally not directly call this function, instead |
212 | * this is only a common interface to implement arch-specific | |
213 | * walker. Please use hugetlb_walk() instead, because that will attempt to | |
214 | * verify the locking for you. | |
215 | * | |
fe7d4c6d PX |
216 | * Since this function will walk all the pgtable pages (including not only |
217 | * high-level pgtable page, but also PUD entry that can be unshared | |
218 | * concurrently for VM_SHARED), the caller of this function should be | |
219 | * responsible of its thread safety. One can follow this rule: | |
220 | * | |
221 | * (1) For private mappings: pmd unsharing is not possible, so holding the | |
222 | * mmap_lock for either read or write is sufficient. Most callers | |
223 | * already hold the mmap_lock, so normally, no special action is | |
224 | * required. | |
225 | * | |
226 | * (2) For shared mappings: pmd unsharing is possible (so the PUD-ranged | |
227 | * pgtable page can go away from under us! It can be done by a pmd | |
228 | * unshare with a follow up munmap() on the other process), then we | |
229 | * need either: | |
230 | * | |
231 | * (2.1) hugetlb vma lock read or write held, to make sure pmd unshare | |
232 | * won't happen upon the range (it also makes sure the pte_t we | |
233 | * read is the right and stable one), or, | |
234 | * | |
235 | * (2.2) hugetlb mapping i_mmap_rwsem lock held read or write, to make | |
236 | * sure even if unshare happened the racy unmap() will wait until | |
237 | * i_mmap_rwsem is released. | |
238 | * | |
239 | * Option (2.1) is the safest, which guarantees pte stability from pmd | |
240 | * sharing pov, until the vma lock released. Option (2.2) doesn't protect | |
241 | * a concurrent pmd unshare, but it makes sure the pgtable page is safe to | |
242 | * access. | |
243 | */ | |
7868a208 PA |
244 | pte_t *huge_pte_offset(struct mm_struct *mm, |
245 | unsigned long addr, unsigned long sz); | |
e95a9851 | 246 | unsigned long hugetlb_mask_last_page(struct hstate *h); |
34ae204f | 247 | int huge_pmd_unshare(struct mm_struct *mm, struct vm_area_struct *vma, |
4ddb4d91 | 248 | unsigned long addr, pte_t *ptep); |
017b1660 MK |
249 | void adjust_range_if_pmd_sharing_possible(struct vm_area_struct *vma, |
250 | unsigned long *start, unsigned long *end); | |
faaa5b62 | 251 | |
2820b0f0 RR |
252 | extern void __hugetlb_zap_begin(struct vm_area_struct *vma, |
253 | unsigned long *begin, unsigned long *end); | |
254 | extern void __hugetlb_zap_end(struct vm_area_struct *vma, | |
255 | struct zap_details *details); | |
256 | ||
257 | static inline void hugetlb_zap_begin(struct vm_area_struct *vma, | |
258 | unsigned long *start, unsigned long *end) | |
259 | { | |
260 | if (is_vm_hugetlb_page(vma)) | |
261 | __hugetlb_zap_begin(vma, start, end); | |
262 | } | |
263 | ||
264 | static inline void hugetlb_zap_end(struct vm_area_struct *vma, | |
265 | struct zap_details *details) | |
266 | { | |
267 | if (is_vm_hugetlb_page(vma)) | |
268 | __hugetlb_zap_end(vma, details); | |
269 | } | |
270 | ||
8d9bfb26 MK |
271 | void hugetlb_vma_lock_read(struct vm_area_struct *vma); |
272 | void hugetlb_vma_unlock_read(struct vm_area_struct *vma); | |
273 | void hugetlb_vma_lock_write(struct vm_area_struct *vma); | |
274 | void hugetlb_vma_unlock_write(struct vm_area_struct *vma); | |
275 | int hugetlb_vma_trylock_write(struct vm_area_struct *vma); | |
276 | void hugetlb_vma_assert_locked(struct vm_area_struct *vma); | |
277 | void hugetlb_vma_lock_release(struct kref *kref); | |
a79390f5 | 278 | long hugetlb_change_protection(struct vm_area_struct *vma, |
5a90d5a1 PX |
279 | unsigned long address, unsigned long end, pgprot_t newprot, |
280 | unsigned long cp_flags); | |
d5ed7444 | 281 | bool is_hugetlb_entry_migration(pte_t pte); |
52526ca7 | 282 | bool is_hugetlb_entry_hwpoisoned(pte_t pte); |
6dfeaff9 | 283 | void hugetlb_unshare_all_pmds(struct vm_area_struct *vma); |
ab5ac90a | 284 | |
1da177e4 LT |
285 | #else /* !CONFIG_HUGETLB_PAGE */ |
286 | ||
8d9bfb26 | 287 | static inline void hugetlb_dup_vma_private(struct vm_area_struct *vma) |
a1e78772 MG |
288 | { |
289 | } | |
290 | ||
550a7d60 MA |
291 | static inline void clear_vma_resv_huge_pages(struct vm_area_struct *vma) |
292 | { | |
293 | } | |
294 | ||
1da177e4 LT |
295 | static inline unsigned long hugetlb_total_pages(void) |
296 | { | |
297 | return 0; | |
298 | } | |
299 | ||
6e8cda4c MWO |
300 | static inline struct address_space *hugetlb_folio_mapping_lock_write( |
301 | struct folio *folio) | |
c0d0381a MK |
302 | { |
303 | return NULL; | |
304 | } | |
305 | ||
34ae204f MK |
306 | static inline int huge_pmd_unshare(struct mm_struct *mm, |
307 | struct vm_area_struct *vma, | |
4ddb4d91 | 308 | unsigned long addr, pte_t *ptep) |
017b1660 MK |
309 | { |
310 | return 0; | |
311 | } | |
312 | ||
313 | static inline void adjust_range_if_pmd_sharing_possible( | |
314 | struct vm_area_struct *vma, | |
315 | unsigned long *start, unsigned long *end) | |
316 | { | |
317 | } | |
318 | ||
2820b0f0 RR |
319 | static inline void hugetlb_zap_begin( |
320 | struct vm_area_struct *vma, | |
321 | unsigned long *start, unsigned long *end) | |
322 | { | |
323 | } | |
324 | ||
325 | static inline void hugetlb_zap_end( | |
326 | struct vm_area_struct *vma, | |
327 | struct zap_details *details) | |
328 | { | |
329 | } | |
330 | ||
1f9dccb2 | 331 | static inline int copy_hugetlb_page_range(struct mm_struct *dst, |
bc70fbf2 PX |
332 | struct mm_struct *src, |
333 | struct vm_area_struct *dst_vma, | |
334 | struct vm_area_struct *src_vma) | |
1f9dccb2 MK |
335 | { |
336 | BUG(); | |
337 | return 0; | |
338 | } | |
339 | ||
550a7d60 MA |
340 | static inline int move_hugetlb_page_tables(struct vm_area_struct *vma, |
341 | struct vm_area_struct *new_vma, | |
342 | unsigned long old_addr, | |
343 | unsigned long new_addr, | |
344 | unsigned long len) | |
345 | { | |
346 | BUG(); | |
347 | return 0; | |
348 | } | |
349 | ||
e1759c21 AD |
350 | static inline void hugetlb_report_meminfo(struct seq_file *m) |
351 | { | |
352 | } | |
1f9dccb2 | 353 | |
7981593b | 354 | static inline int hugetlb_report_node_meminfo(char *buf, int len, int nid) |
1f9dccb2 MK |
355 | { |
356 | return 0; | |
357 | } | |
358 | ||
dcadcf1c | 359 | static inline void hugetlb_show_meminfo_node(int nid) |
949f7ec5 DR |
360 | { |
361 | } | |
1f9dccb2 | 362 | |
1f9dccb2 MK |
363 | static inline int prepare_hugepage_range(struct file *file, |
364 | unsigned long addr, unsigned long len) | |
365 | { | |
366 | return -EINVAL; | |
367 | } | |
368 | ||
8d9bfb26 MK |
369 | static inline void hugetlb_vma_lock_read(struct vm_area_struct *vma) |
370 | { | |
371 | } | |
372 | ||
373 | static inline void hugetlb_vma_unlock_read(struct vm_area_struct *vma) | |
374 | { | |
375 | } | |
376 | ||
377 | static inline void hugetlb_vma_lock_write(struct vm_area_struct *vma) | |
378 | { | |
379 | } | |
380 | ||
381 | static inline void hugetlb_vma_unlock_write(struct vm_area_struct *vma) | |
382 | { | |
383 | } | |
384 | ||
385 | static inline int hugetlb_vma_trylock_write(struct vm_area_struct *vma) | |
386 | { | |
387 | return 1; | |
388 | } | |
389 | ||
390 | static inline void hugetlb_vma_assert_locked(struct vm_area_struct *vma) | |
391 | { | |
392 | } | |
393 | ||
1f9dccb2 MK |
394 | static inline int is_hugepage_only_range(struct mm_struct *mm, |
395 | unsigned long addr, unsigned long len) | |
396 | { | |
397 | return 0; | |
398 | } | |
399 | ||
400 | static inline void hugetlb_free_pgd_range(struct mmu_gather *tlb, | |
401 | unsigned long addr, unsigned long end, | |
402 | unsigned long floor, unsigned long ceiling) | |
403 | { | |
404 | BUG(); | |
405 | } | |
406 | ||
714c1891 | 407 | #ifdef CONFIG_USERFAULTFD |
61c50040 | 408 | static inline int hugetlb_mfill_atomic_pte(pte_t *dst_pte, |
a734991c AR |
409 | struct vm_area_struct *dst_vma, |
410 | unsigned long dst_addr, | |
411 | unsigned long src_addr, | |
d9712937 | 412 | uffd_flags_t flags, |
0169fd51 | 413 | struct folio **foliop) |
1f9dccb2 MK |
414 | { |
415 | BUG(); | |
416 | return 0; | |
417 | } | |
714c1891 | 418 | #endif /* CONFIG_USERFAULTFD */ |
1f9dccb2 MK |
419 | |
420 | static inline pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr, | |
421 | unsigned long sz) | |
422 | { | |
423 | return NULL; | |
424 | } | |
24669e58 | 425 | |
9747b9e9 | 426 | static inline bool isolate_hugetlb(struct folio *folio, struct list_head *list) |
f40386a4 | 427 | { |
9747b9e9 | 428 | return false; |
f40386a4 | 429 | } |
1da177e4 | 430 | |
04bac040 | 431 | static inline int get_hwpoison_hugetlb_folio(struct folio *folio, bool *hugetlb, bool unpoison) |
25182f05 NH |
432 | { |
433 | return 0; | |
434 | } | |
435 | ||
e591ef7d NH |
436 | static inline int get_huge_page_for_hwpoison(unsigned long pfn, int flags, |
437 | bool *migratable_cleared) | |
405ce051 NH |
438 | { |
439 | return 0; | |
440 | } | |
441 | ||
ea8e72f4 | 442 | static inline void folio_putback_active_hugetlb(struct folio *folio) |
1f9dccb2 MK |
443 | { |
444 | } | |
445 | ||
345c62d1 SK |
446 | static inline void move_hugetlb_state(struct folio *old_folio, |
447 | struct folio *new_folio, int reason) | |
1f9dccb2 MK |
448 | { |
449 | } | |
450 | ||
a79390f5 | 451 | static inline long hugetlb_change_protection( |
1f9dccb2 | 452 | struct vm_area_struct *vma, unsigned long address, |
5a90d5a1 PX |
453 | unsigned long end, pgprot_t newprot, |
454 | unsigned long cp_flags) | |
7da4d641 PZ |
455 | { |
456 | return 0; | |
457 | } | |
8f860591 | 458 | |
2820b0f0 | 459 | static inline void __unmap_hugepage_range(struct mmu_gather *tlb, |
d833352a | 460 | struct vm_area_struct *vma, unsigned long start, |
05e90bd0 PX |
461 | unsigned long end, struct page *ref_page, |
462 | zap_flags_t zap_flags) | |
d833352a MG |
463 | { |
464 | BUG(); | |
465 | } | |
466 | ||
a953e772 | 467 | static inline vm_fault_t hugetlb_fault(struct mm_struct *mm, |
1f9dccb2 MK |
468 | struct vm_area_struct *vma, unsigned long address, |
469 | unsigned int flags) | |
a953e772 SJ |
470 | { |
471 | BUG(); | |
472 | return 0; | |
473 | } | |
24669e58 | 474 | |
6dfeaff9 PX |
475 | static inline void hugetlb_unshare_all_pmds(struct vm_area_struct *vma) { } |
476 | ||
1da177e4 | 477 | #endif /* !CONFIG_HUGETLB_PAGE */ |
f30c59e9 AK |
478 | |
479 | #ifndef pgd_write | |
480 | static inline int pgd_write(pgd_t pgd) | |
481 | { | |
482 | BUG(); | |
483 | return 0; | |
484 | } | |
485 | #endif | |
486 | ||
4e52780d EM |
487 | #define HUGETLB_ANON_FILE "anon_hugepage" |
488 | ||
6bfde05b EM |
489 | enum { |
490 | /* | |
491 | * The file will be used as an shm file so shmfs accounting rules | |
492 | * apply | |
493 | */ | |
494 | HUGETLB_SHMFS_INODE = 1, | |
4e52780d EM |
495 | /* |
496 | * The file is being created on the internal vfs mount and shmfs | |
497 | * accounting rules do not apply | |
498 | */ | |
499 | HUGETLB_ANONHUGE_INODE = 2, | |
6bfde05b EM |
500 | }; |
501 | ||
1da177e4 | 502 | #ifdef CONFIG_HUGETLBFS |
1da177e4 | 503 | struct hugetlbfs_sb_info { |
1da177e4 LT |
504 | long max_inodes; /* inodes allowed */ |
505 | long free_inodes; /* inodes free */ | |
506 | spinlock_t stat_lock; | |
a137e1cc | 507 | struct hstate *hstate; |
90481622 | 508 | struct hugepage_subpool *spool; |
4a25220d DH |
509 | kuid_t uid; |
510 | kgid_t gid; | |
511 | umode_t mode; | |
1da177e4 LT |
512 | }; |
513 | ||
1da177e4 LT |
514 | static inline struct hugetlbfs_sb_info *HUGETLBFS_SB(struct super_block *sb) |
515 | { | |
516 | return sb->s_fs_info; | |
517 | } | |
518 | ||
da14c1e5 | 519 | struct hugetlbfs_inode_info { |
da14c1e5 | 520 | struct inode vfs_inode; |
ff62a342 | 521 | unsigned int seals; |
da14c1e5 MAL |
522 | }; |
523 | ||
524 | static inline struct hugetlbfs_inode_info *HUGETLBFS_I(struct inode *inode) | |
525 | { | |
526 | return container_of(inode, struct hugetlbfs_inode_info, vfs_inode); | |
527 | } | |
528 | ||
f0f37e2f | 529 | extern const struct vm_operations_struct hugetlb_vm_ops; |
af73e4d9 | 530 | struct file *hugetlb_file_setup(const char *name, size_t size, vm_flags_t acct, |
83c1fd76 | 531 | int creat_flags, int page_size_log); |
1da177e4 | 532 | |
886b94d2 | 533 | static inline bool is_file_hugepages(const struct file *file) |
1da177e4 | 534 | { |
886b94d2 | 535 | return file->f_op->fop_flags & FOP_HUGE_PAGES; |
1da177e4 LT |
536 | } |
537 | ||
bb297bb2 CL |
538 | static inline struct hstate *hstate_inode(struct inode *i) |
539 | { | |
540 | return HUGETLBFS_SB(i->i_sb)->hstate; | |
541 | } | |
1da177e4 LT |
542 | #else /* !CONFIG_HUGETLBFS */ |
543 | ||
719ff321 | 544 | #define is_file_hugepages(file) false |
40716e29 | 545 | static inline struct file * |
af73e4d9 | 546 | hugetlb_file_setup(const char *name, size_t size, vm_flags_t acctflag, |
83c1fd76 | 547 | int creat_flags, int page_size_log) |
e9ea0e2d AM |
548 | { |
549 | return ERR_PTR(-ENOSYS); | |
550 | } | |
1da177e4 | 551 | |
bb297bb2 CL |
552 | static inline struct hstate *hstate_inode(struct inode *i) |
553 | { | |
554 | return NULL; | |
555 | } | |
1da177e4 LT |
556 | #endif /* !CONFIG_HUGETLBFS */ |
557 | ||
d2ba27e8 AB |
558 | #ifdef HAVE_ARCH_HUGETLB_UNMAPPED_AREA |
559 | unsigned long hugetlb_get_unmapped_area(struct file *file, unsigned long addr, | |
560 | unsigned long len, unsigned long pgoff, | |
561 | unsigned long flags); | |
562 | #endif /* HAVE_ARCH_HUGETLB_UNMAPPED_AREA */ | |
563 | ||
4b439e25 CL |
564 | unsigned long |
565 | generic_hugetlb_get_unmapped_area(struct file *file, unsigned long addr, | |
566 | unsigned long len, unsigned long pgoff, | |
567 | unsigned long flags); | |
568 | ||
d6995da3 MK |
569 | /* |
570 | * huegtlb page specific state flags. These flags are located in page.private | |
571 | * of the hugetlb head page. Functions created via the below macros should be | |
572 | * used to manipulate these flags. | |
573 | * | |
574 | * HPG_restore_reserve - Set when a hugetlb page consumes a reservation at | |
575 | * allocation time. Cleared when page is fully instantiated. Free | |
576 | * routine checks flag to restore a reservation on error paths. | |
d95c0337 MK |
577 | * Synchronization: Examined or modified by code that knows it has |
578 | * the only reference to page. i.e. After allocation but before use | |
579 | * or when the page is being freed. | |
8f251a3d MK |
580 | * HPG_migratable - Set after a newly allocated page is added to the page |
581 | * cache and/or page tables. Indicates the page is a candidate for | |
582 | * migration. | |
d95c0337 MK |
583 | * Synchronization: Initially set after new page allocation with no |
584 | * locking. When examined and modified during migration processing | |
585 | * (isolate, migrate, putback) the hugetlb_lock is held. | |
161df60e | 586 | * HPG_temporary - Set on a page that is temporarily allocated from the buddy |
9157c311 MK |
587 | * allocator. Typically used for migration target pages when no pages |
588 | * are available in the pool. The hugetlb free page path will | |
589 | * immediately free pages with this flag set to the buddy allocator. | |
d95c0337 MK |
590 | * Synchronization: Can be set after huge page allocation from buddy when |
591 | * code knows it has only reference. All other examinations and | |
592 | * modifications require hugetlb_lock. | |
6c037149 | 593 | * HPG_freed - Set when page is on the free lists. |
d95c0337 | 594 | * Synchronization: hugetlb_lock held for examination and modification. |
ad2fa371 | 595 | * HPG_vmemmap_optimized - Set when the vmemmap pages of the page are freed. |
161df60e NH |
596 | * HPG_raw_hwp_unreliable - Set when the hugetlb page has a hwpoison sub-page |
597 | * that is not tracked by raw_hwp_page list. | |
d6995da3 MK |
598 | */ |
599 | enum hugetlb_page_flags { | |
600 | HPG_restore_reserve = 0, | |
8f251a3d | 601 | HPG_migratable, |
9157c311 | 602 | HPG_temporary, |
6c037149 | 603 | HPG_freed, |
ad2fa371 | 604 | HPG_vmemmap_optimized, |
161df60e | 605 | HPG_raw_hwp_unreliable, |
d6995da3 MK |
606 | __NR_HPAGEFLAGS, |
607 | }; | |
608 | ||
609 | /* | |
610 | * Macros to create test, set and clear function definitions for | |
611 | * hugetlb specific page flags. | |
612 | */ | |
613 | #ifdef CONFIG_HUGETLB_PAGE | |
614 | #define TESTHPAGEFLAG(uname, flname) \ | |
d03c376d SK |
615 | static __always_inline \ |
616 | bool folio_test_hugetlb_##flname(struct folio *folio) \ | |
617 | { void *private = &folio->private; \ | |
618 | return test_bit(HPG_##flname, private); \ | |
619 | } \ | |
d6995da3 MK |
620 | static inline int HPage##uname(struct page *page) \ |
621 | { return test_bit(HPG_##flname, &(page->private)); } | |
622 | ||
623 | #define SETHPAGEFLAG(uname, flname) \ | |
d03c376d SK |
624 | static __always_inline \ |
625 | void folio_set_hugetlb_##flname(struct folio *folio) \ | |
626 | { void *private = &folio->private; \ | |
627 | set_bit(HPG_##flname, private); \ | |
628 | } \ | |
d6995da3 MK |
629 | static inline void SetHPage##uname(struct page *page) \ |
630 | { set_bit(HPG_##flname, &(page->private)); } | |
631 | ||
632 | #define CLEARHPAGEFLAG(uname, flname) \ | |
d03c376d SK |
633 | static __always_inline \ |
634 | void folio_clear_hugetlb_##flname(struct folio *folio) \ | |
635 | { void *private = &folio->private; \ | |
636 | clear_bit(HPG_##flname, private); \ | |
637 | } \ | |
d6995da3 MK |
638 | static inline void ClearHPage##uname(struct page *page) \ |
639 | { clear_bit(HPG_##flname, &(page->private)); } | |
640 | #else | |
641 | #define TESTHPAGEFLAG(uname, flname) \ | |
d03c376d SK |
642 | static inline bool \ |
643 | folio_test_hugetlb_##flname(struct folio *folio) \ | |
644 | { return 0; } \ | |
d6995da3 MK |
645 | static inline int HPage##uname(struct page *page) \ |
646 | { return 0; } | |
647 | ||
648 | #define SETHPAGEFLAG(uname, flname) \ | |
d03c376d SK |
649 | static inline void \ |
650 | folio_set_hugetlb_##flname(struct folio *folio) \ | |
651 | { } \ | |
d6995da3 MK |
652 | static inline void SetHPage##uname(struct page *page) \ |
653 | { } | |
654 | ||
655 | #define CLEARHPAGEFLAG(uname, flname) \ | |
d03c376d SK |
656 | static inline void \ |
657 | folio_clear_hugetlb_##flname(struct folio *folio) \ | |
658 | { } \ | |
d6995da3 MK |
659 | static inline void ClearHPage##uname(struct page *page) \ |
660 | { } | |
661 | #endif | |
662 | ||
663 | #define HPAGEFLAG(uname, flname) \ | |
664 | TESTHPAGEFLAG(uname, flname) \ | |
665 | SETHPAGEFLAG(uname, flname) \ | |
666 | CLEARHPAGEFLAG(uname, flname) \ | |
667 | ||
668 | /* | |
669 | * Create functions associated with hugetlb page flags | |
670 | */ | |
671 | HPAGEFLAG(RestoreReserve, restore_reserve) | |
8f251a3d | 672 | HPAGEFLAG(Migratable, migratable) |
9157c311 | 673 | HPAGEFLAG(Temporary, temporary) |
6c037149 | 674 | HPAGEFLAG(Freed, freed) |
ad2fa371 | 675 | HPAGEFLAG(VmemmapOptimized, vmemmap_optimized) |
161df60e | 676 | HPAGEFLAG(RawHwpUnreliable, raw_hwp_unreliable) |
d6995da3 | 677 | |
a5516438 AK |
678 | #ifdef CONFIG_HUGETLB_PAGE |
679 | ||
a3437870 | 680 | #define HSTATE_NAME_LEN 32 |
a5516438 AK |
681 | /* Defines one hugetlb page size */ |
682 | struct hstate { | |
29383967 | 683 | struct mutex resize_lock; |
e8c5c824 LS |
684 | int next_nid_to_alloc; |
685 | int next_nid_to_free; | |
a5516438 | 686 | unsigned int order; |
79dfc695 | 687 | unsigned int demote_order; |
a5516438 AK |
688 | unsigned long mask; |
689 | unsigned long max_huge_pages; | |
690 | unsigned long nr_huge_pages; | |
691 | unsigned long free_huge_pages; | |
692 | unsigned long resv_huge_pages; | |
693 | unsigned long surplus_huge_pages; | |
694 | unsigned long nr_overcommit_huge_pages; | |
0edaecfa | 695 | struct list_head hugepage_activelist; |
a5516438 | 696 | struct list_head hugepage_freelists[MAX_NUMNODES]; |
b5389086 | 697 | unsigned int max_huge_pages_node[MAX_NUMNODES]; |
a5516438 AK |
698 | unsigned int nr_huge_pages_node[MAX_NUMNODES]; |
699 | unsigned int free_huge_pages_node[MAX_NUMNODES]; | |
700 | unsigned int surplus_huge_pages_node[MAX_NUMNODES]; | |
abb8206c AK |
701 | #ifdef CONFIG_CGROUP_HUGETLB |
702 | /* cgroup control files */ | |
f4776199 MA |
703 | struct cftype cgroup_files_dfl[8]; |
704 | struct cftype cgroup_files_legacy[10]; | |
abb8206c | 705 | #endif |
a3437870 | 706 | char name[HSTATE_NAME_LEN]; |
a5516438 AK |
707 | }; |
708 | ||
53ba51d2 JT |
709 | struct huge_bootmem_page { |
710 | struct list_head list; | |
711 | struct hstate *hstate; | |
712 | }; | |
713 | ||
ae37c7ff | 714 | int isolate_or_dissolve_huge_page(struct page *page, struct list_head *list); |
d0ce0e47 | 715 | struct folio *alloc_hugetlb_folio(struct vm_area_struct *vma, |
70c3547e | 716 | unsigned long addr, int avoid_reserve); |
e37d3e83 | 717 | struct folio *alloc_hugetlb_folio_nodemask(struct hstate *h, int preferred_nid, |
42d0c3fb BW |
718 | nodemask_t *nmask, gfp_t gfp_mask, |
719 | bool allow_alloc_fallback); | |
9b91c0e2 | 720 | int hugetlb_add_to_page_cache(struct folio *folio, struct address_space *mapping, |
ab76ad54 | 721 | pgoff_t idx); |
846be085 | 722 | void restore_reserve_on_error(struct hstate *h, struct vm_area_struct *vma, |
d2d7bb44 | 723 | unsigned long address, struct folio *folio); |
bf50bab2 | 724 | |
53ba51d2 | 725 | /* arch callback */ |
b5389086 ZY |
726 | int __init __alloc_bootmem_huge_page(struct hstate *h, int nid); |
727 | int __init alloc_bootmem_huge_page(struct hstate *h, int nid); | |
728 | bool __init hugetlb_node_alloc_supported(void); | |
53ba51d2 | 729 | |
e5ff2159 | 730 | void __init hugetlb_add_hstate(unsigned order); |
ae94da89 | 731 | bool __init arch_hugetlb_valid_size(unsigned long size); |
e5ff2159 AK |
732 | struct hstate *size_to_hstate(unsigned long size); |
733 | ||
734 | #ifndef HUGE_MAX_HSTATE | |
735 | #define HUGE_MAX_HSTATE 1 | |
736 | #endif | |
737 | ||
738 | extern struct hstate hstates[HUGE_MAX_HSTATE]; | |
739 | extern unsigned int default_hstate_idx; | |
740 | ||
741 | #define default_hstate (hstates[default_hstate_idx]) | |
a5516438 | 742 | |
149562f7 SK |
743 | static inline struct hugepage_subpool *hugetlb_folio_subpool(struct folio *folio) |
744 | { | |
dad6a5eb | 745 | return folio->_hugetlb_subpool; |
149562f7 SK |
746 | } |
747 | ||
149562f7 SK |
748 | static inline void hugetlb_set_folio_subpool(struct folio *folio, |
749 | struct hugepage_subpool *subpool) | |
750 | { | |
dad6a5eb | 751 | folio->_hugetlb_subpool = subpool; |
d6995da3 MK |
752 | } |
753 | ||
a5516438 AK |
754 | static inline struct hstate *hstate_file(struct file *f) |
755 | { | |
496ad9aa | 756 | return hstate_inode(file_inode(f)); |
a5516438 AK |
757 | } |
758 | ||
af73e4d9 NH |
759 | static inline struct hstate *hstate_sizelog(int page_size_log) |
760 | { | |
761 | if (!page_size_log) | |
762 | return &default_hstate; | |
97ad2be1 | 763 | |
ec4288fe MK |
764 | if (page_size_log < BITS_PER_LONG) |
765 | return size_to_hstate(1UL << page_size_log); | |
766 | ||
767 | return NULL; | |
af73e4d9 NH |
768 | } |
769 | ||
a137e1cc | 770 | static inline struct hstate *hstate_vma(struct vm_area_struct *vma) |
a5516438 | 771 | { |
a137e1cc | 772 | return hstate_file(vma->vm_file); |
a5516438 AK |
773 | } |
774 | ||
6213834c | 775 | static inline unsigned long huge_page_size(const struct hstate *h) |
a5516438 AK |
776 | { |
777 | return (unsigned long)PAGE_SIZE << h->order; | |
778 | } | |
779 | ||
08fba699 MG |
780 | extern unsigned long vma_kernel_pagesize(struct vm_area_struct *vma); |
781 | ||
3340289d MG |
782 | extern unsigned long vma_mmu_pagesize(struct vm_area_struct *vma); |
783 | ||
a5516438 AK |
784 | static inline unsigned long huge_page_mask(struct hstate *h) |
785 | { | |
786 | return h->mask; | |
787 | } | |
788 | ||
789 | static inline unsigned int huge_page_order(struct hstate *h) | |
790 | { | |
791 | return h->order; | |
792 | } | |
793 | ||
794 | static inline unsigned huge_page_shift(struct hstate *h) | |
795 | { | |
796 | return h->order + PAGE_SHIFT; | |
797 | } | |
798 | ||
bae7f4ae LC |
799 | static inline bool hstate_is_gigantic(struct hstate *h) |
800 | { | |
5e0a760b | 801 | return huge_page_order(h) > MAX_PAGE_ORDER; |
bae7f4ae LC |
802 | } |
803 | ||
6213834c | 804 | static inline unsigned int pages_per_huge_page(const struct hstate *h) |
a5516438 AK |
805 | { |
806 | return 1 << h->order; | |
807 | } | |
808 | ||
809 | static inline unsigned int blocks_per_huge_page(struct hstate *h) | |
810 | { | |
811 | return huge_page_size(h) / 512; | |
812 | } | |
813 | ||
a08c7193 SK |
814 | static inline struct folio *filemap_lock_hugetlb_folio(struct hstate *h, |
815 | struct address_space *mapping, pgoff_t idx) | |
816 | { | |
817 | return filemap_lock_folio(mapping, idx << huge_page_order(h)); | |
818 | } | |
819 | ||
a5516438 AK |
820 | #include <asm/hugetlb.h> |
821 | ||
b0eae98c AK |
822 | #ifndef is_hugepage_only_range |
823 | static inline int is_hugepage_only_range(struct mm_struct *mm, | |
824 | unsigned long addr, unsigned long len) | |
825 | { | |
826 | return 0; | |
827 | } | |
828 | #define is_hugepage_only_range is_hugepage_only_range | |
829 | #endif | |
830 | ||
51718e25 MWO |
831 | #ifndef arch_clear_hugetlb_flags |
832 | static inline void arch_clear_hugetlb_flags(struct folio *folio) { } | |
833 | #define arch_clear_hugetlb_flags arch_clear_hugetlb_flags | |
5be99343 AK |
834 | #endif |
835 | ||
d9ed9faa | 836 | #ifndef arch_make_huge_pte |
79c1c594 CL |
837 | static inline pte_t arch_make_huge_pte(pte_t entry, unsigned int shift, |
838 | vm_flags_t flags) | |
d9ed9faa | 839 | { |
16785bd7 | 840 | return pte_mkhuge(entry); |
d9ed9faa CM |
841 | } |
842 | #endif | |
843 | ||
e51da3a9 SK |
844 | static inline struct hstate *folio_hstate(struct folio *folio) |
845 | { | |
846 | VM_BUG_ON_FOLIO(!folio_test_hugetlb(folio), folio); | |
847 | return size_to_hstate(folio_size(folio)); | |
848 | } | |
849 | ||
aa50d3a7 AK |
850 | static inline unsigned hstate_index_to_shift(unsigned index) |
851 | { | |
852 | return hstates[index].order + PAGE_SHIFT; | |
853 | } | |
854 | ||
972dc4de AK |
855 | static inline int hstate_index(struct hstate *h) |
856 | { | |
857 | return h - hstates; | |
858 | } | |
859 | ||
54fa49b2 | 860 | int dissolve_free_hugetlb_folio(struct folio *folio); |
d199483c | 861 | int dissolve_free_hugetlb_folios(unsigned long start_pfn, |
082d5b6b | 862 | unsigned long end_pfn); |
e693de18 | 863 | |
161df60e | 864 | #ifdef CONFIG_MEMORY_FAILURE |
2ff6cece | 865 | extern void folio_clear_hugetlb_hwpoison(struct folio *folio); |
161df60e | 866 | #else |
2ff6cece | 867 | static inline void folio_clear_hugetlb_hwpoison(struct folio *folio) |
161df60e NH |
868 | { |
869 | } | |
870 | #endif | |
871 | ||
c177c81e | 872 | #ifdef CONFIG_ARCH_ENABLE_HUGEPAGE_MIGRATION |
e693de18 AK |
873 | #ifndef arch_hugetlb_migration_supported |
874 | static inline bool arch_hugetlb_migration_supported(struct hstate *h) | |
875 | { | |
94310cbc | 876 | if ((huge_page_shift(h) == PMD_SHIFT) || |
9b553bf5 AK |
877 | (huge_page_shift(h) == PUD_SHIFT) || |
878 | (huge_page_shift(h) == PGDIR_SHIFT)) | |
94310cbc AK |
879 | return true; |
880 | else | |
881 | return false; | |
e693de18 AK |
882 | } |
883 | #endif | |
c177c81e | 884 | #else |
e693de18 AK |
885 | static inline bool arch_hugetlb_migration_supported(struct hstate *h) |
886 | { | |
d70c17d4 | 887 | return false; |
e693de18 | 888 | } |
c177c81e | 889 | #endif |
e693de18 AK |
890 | |
891 | static inline bool hugepage_migration_supported(struct hstate *h) | |
892 | { | |
893 | return arch_hugetlb_migration_supported(h); | |
83467efb | 894 | } |
c8721bbb | 895 | |
7ed2c31d AK |
896 | /* |
897 | * Movability check is different as compared to migration check. | |
898 | * It determines whether or not a huge page should be placed on | |
899 | * movable zone or not. Movability of any huge page should be | |
900 | * required only if huge page size is supported for migration. | |
06c88398 | 901 | * There won't be any reason for the huge page to be movable if |
7ed2c31d AK |
902 | * it is not migratable to start with. Also the size of the huge |
903 | * page should be large enough to be placed under a movable zone | |
904 | * and still feasible enough to be migratable. Just the presence | |
905 | * in movable zone does not make the migration feasible. | |
906 | * | |
907 | * So even though large huge page sizes like the gigantic ones | |
908 | * are migratable they should not be movable because its not | |
909 | * feasible to migrate them from movable zone. | |
910 | */ | |
911 | static inline bool hugepage_movable_supported(struct hstate *h) | |
912 | { | |
913 | if (!hugepage_migration_supported(h)) | |
914 | return false; | |
915 | ||
916 | if (hstate_is_gigantic(h)) | |
917 | return false; | |
918 | return true; | |
919 | } | |
920 | ||
d92bbc27 JK |
921 | /* Movability of hugepages depends on migration support. */ |
922 | static inline gfp_t htlb_alloc_mask(struct hstate *h) | |
923 | { | |
924 | if (hugepage_movable_supported(h)) | |
925 | return GFP_HIGHUSER_MOVABLE; | |
926 | else | |
927 | return GFP_HIGHUSER; | |
928 | } | |
929 | ||
19fc7bed JK |
930 | static inline gfp_t htlb_modify_alloc_mask(struct hstate *h, gfp_t gfp_mask) |
931 | { | |
932 | gfp_t modified_mask = htlb_alloc_mask(h); | |
933 | ||
934 | /* Some callers might want to enforce node */ | |
935 | modified_mask |= (gfp_mask & __GFP_THISNODE); | |
936 | ||
41b4dc14 JK |
937 | modified_mask |= (gfp_mask & __GFP_NOWARN); |
938 | ||
19fc7bed JK |
939 | return modified_mask; |
940 | } | |
941 | ||
42d0c3fb BW |
942 | static inline bool htlb_allow_alloc_fallback(int reason) |
943 | { | |
944 | bool allowed_fallback = false; | |
945 | ||
946 | /* | |
947 | * Note: the memory offline, memory failure and migration syscalls will | |
948 | * be allowed to fallback to other nodes due to lack of a better chioce, | |
949 | * that might break the per-node hugetlb pool. While other cases will | |
950 | * set the __GFP_THISNODE to avoid breaking the per-node hugetlb pool. | |
951 | */ | |
952 | switch (reason) { | |
953 | case MR_MEMORY_HOTPLUG: | |
954 | case MR_MEMORY_FAILURE: | |
955 | case MR_SYSCALL: | |
956 | case MR_MEMPOLICY_MBIND: | |
957 | allowed_fallback = true; | |
958 | break; | |
959 | default: | |
960 | break; | |
961 | } | |
962 | ||
963 | return allowed_fallback; | |
964 | } | |
965 | ||
cb900f41 KS |
966 | static inline spinlock_t *huge_pte_lockptr(struct hstate *h, |
967 | struct mm_struct *mm, pte_t *pte) | |
968 | { | |
969 | if (huge_page_size(h) == PMD_SIZE) | |
970 | return pmd_lockptr(mm, (pmd_t *) pte); | |
971 | VM_BUG_ON(huge_page_size(h) == PAGE_SIZE); | |
972 | return &mm->page_table_lock; | |
973 | } | |
974 | ||
2531c8cf DD |
975 | #ifndef hugepages_supported |
976 | /* | |
977 | * Some platform decide whether they support huge pages at boot | |
978 | * time. Some of them, such as powerpc, set HPAGE_SHIFT to 0 | |
979 | * when there is no such support | |
980 | */ | |
981 | #define hugepages_supported() (HPAGE_SHIFT != 0) | |
982 | #endif | |
457c1b27 | 983 | |
5d317b2b NH |
984 | void hugetlb_report_usage(struct seq_file *m, struct mm_struct *mm); |
985 | ||
13db8c50 LZ |
986 | static inline void hugetlb_count_init(struct mm_struct *mm) |
987 | { | |
988 | atomic_long_set(&mm->hugetlb_usage, 0); | |
989 | } | |
990 | ||
5d317b2b NH |
991 | static inline void hugetlb_count_add(long l, struct mm_struct *mm) |
992 | { | |
993 | atomic_long_add(l, &mm->hugetlb_usage); | |
994 | } | |
995 | ||
996 | static inline void hugetlb_count_sub(long l, struct mm_struct *mm) | |
997 | { | |
998 | atomic_long_sub(l, &mm->hugetlb_usage); | |
999 | } | |
e5251fd4 | 1000 | |
023bdd00 AK |
1001 | #ifndef huge_ptep_modify_prot_start |
1002 | #define huge_ptep_modify_prot_start huge_ptep_modify_prot_start | |
1003 | static inline pte_t huge_ptep_modify_prot_start(struct vm_area_struct *vma, | |
1004 | unsigned long addr, pte_t *ptep) | |
1005 | { | |
1006 | return huge_ptep_get_and_clear(vma->vm_mm, addr, ptep); | |
1007 | } | |
1008 | #endif | |
1009 | ||
1010 | #ifndef huge_ptep_modify_prot_commit | |
1011 | #define huge_ptep_modify_prot_commit huge_ptep_modify_prot_commit | |
1012 | static inline void huge_ptep_modify_prot_commit(struct vm_area_struct *vma, | |
1013 | unsigned long addr, pte_t *ptep, | |
1014 | pte_t old_pte, pte_t pte) | |
1015 | { | |
935d4f0c RR |
1016 | unsigned long psize = huge_page_size(hstate_vma(vma)); |
1017 | ||
1018 | set_huge_pte_at(vma->vm_mm, addr, ptep, pte, psize); | |
023bdd00 AK |
1019 | } |
1020 | #endif | |
1021 | ||
a4a00b45 MS |
1022 | #ifdef CONFIG_NUMA |
1023 | void hugetlb_register_node(struct node *node); | |
1024 | void hugetlb_unregister_node(struct node *node); | |
1025 | #endif | |
1026 | ||
b79f8eb4 JY |
1027 | /* |
1028 | * Check if a given raw @page in a hugepage is HWPOISON. | |
1029 | */ | |
1030 | bool is_raw_hwpoison_page_in_hugepage(struct page *page); | |
1031 | ||
af73e4d9 | 1032 | #else /* CONFIG_HUGETLB_PAGE */ |
a5516438 | 1033 | struct hstate {}; |
442a5a9a | 1034 | |
345c62d1 SK |
1035 | static inline struct hugepage_subpool *hugetlb_folio_subpool(struct folio *folio) |
1036 | { | |
1037 | return NULL; | |
1038 | } | |
1039 | ||
a08c7193 SK |
1040 | static inline struct folio *filemap_lock_hugetlb_folio(struct hstate *h, |
1041 | struct address_space *mapping, pgoff_t idx) | |
1042 | { | |
1043 | return NULL; | |
1044 | } | |
1045 | ||
ae37c7ff OS |
1046 | static inline int isolate_or_dissolve_huge_page(struct page *page, |
1047 | struct list_head *list) | |
369fa227 OS |
1048 | { |
1049 | return -ENOMEM; | |
1050 | } | |
1051 | ||
d0ce0e47 | 1052 | static inline struct folio *alloc_hugetlb_folio(struct vm_area_struct *vma, |
442a5a9a JG |
1053 | unsigned long addr, |
1054 | int avoid_reserve) | |
1055 | { | |
1056 | return NULL; | |
1057 | } | |
1058 | ||
e37d3e83 SK |
1059 | static inline struct folio * |
1060 | alloc_hugetlb_folio_nodemask(struct hstate *h, int preferred_nid, | |
42d0c3fb BW |
1061 | nodemask_t *nmask, gfp_t gfp_mask, |
1062 | bool allow_alloc_fallback) | |
442a5a9a JG |
1063 | { |
1064 | return NULL; | |
1065 | } | |
1066 | ||
442a5a9a JG |
1067 | static inline int __alloc_bootmem_huge_page(struct hstate *h) |
1068 | { | |
1069 | return 0; | |
1070 | } | |
1071 | ||
1072 | static inline struct hstate *hstate_file(struct file *f) | |
1073 | { | |
1074 | return NULL; | |
1075 | } | |
1076 | ||
1077 | static inline struct hstate *hstate_sizelog(int page_size_log) | |
1078 | { | |
1079 | return NULL; | |
1080 | } | |
1081 | ||
1082 | static inline struct hstate *hstate_vma(struct vm_area_struct *vma) | |
1083 | { | |
1084 | return NULL; | |
442a5a9a JG |
1085 | } |
1086 | ||
e51da3a9 SK |
1087 | static inline struct hstate *folio_hstate(struct folio *folio) |
1088 | { | |
1089 | return NULL; | |
1090 | } | |
1091 | ||
2aff7a47 MWO |
1092 | static inline struct hstate *size_to_hstate(unsigned long size) |
1093 | { | |
1094 | return NULL; | |
1095 | } | |
1096 | ||
442a5a9a JG |
1097 | static inline unsigned long huge_page_size(struct hstate *h) |
1098 | { | |
1099 | return PAGE_SIZE; | |
1100 | } | |
1101 | ||
1102 | static inline unsigned long huge_page_mask(struct hstate *h) | |
1103 | { | |
1104 | return PAGE_MASK; | |
1105 | } | |
1106 | ||
1107 | static inline unsigned long vma_kernel_pagesize(struct vm_area_struct *vma) | |
1108 | { | |
1109 | return PAGE_SIZE; | |
1110 | } | |
1111 | ||
1112 | static inline unsigned long vma_mmu_pagesize(struct vm_area_struct *vma) | |
1113 | { | |
1114 | return PAGE_SIZE; | |
1115 | } | |
1116 | ||
1117 | static inline unsigned int huge_page_order(struct hstate *h) | |
1118 | { | |
1119 | return 0; | |
1120 | } | |
1121 | ||
1122 | static inline unsigned int huge_page_shift(struct hstate *h) | |
1123 | { | |
1124 | return PAGE_SHIFT; | |
1125 | } | |
1126 | ||
94310cbc AK |
1127 | static inline bool hstate_is_gigantic(struct hstate *h) |
1128 | { | |
1129 | return false; | |
1130 | } | |
1131 | ||
510a35d4 AR |
1132 | static inline unsigned int pages_per_huge_page(struct hstate *h) |
1133 | { | |
1134 | return 1; | |
1135 | } | |
c3114a84 AK |
1136 | |
1137 | static inline unsigned hstate_index_to_shift(unsigned index) | |
1138 | { | |
1139 | return 0; | |
1140 | } | |
1141 | ||
1142 | static inline int hstate_index(struct hstate *h) | |
1143 | { | |
1144 | return 0; | |
1145 | } | |
13d60f4b | 1146 | |
54fa49b2 | 1147 | static inline int dissolve_free_hugetlb_folio(struct folio *folio) |
c3114a84 AK |
1148 | { |
1149 | return 0; | |
1150 | } | |
1151 | ||
d199483c | 1152 | static inline int dissolve_free_hugetlb_folios(unsigned long start_pfn, |
c3114a84 AK |
1153 | unsigned long end_pfn) |
1154 | { | |
1155 | return 0; | |
1156 | } | |
1157 | ||
1158 | static inline bool hugepage_migration_supported(struct hstate *h) | |
1159 | { | |
1160 | return false; | |
1161 | } | |
cb900f41 | 1162 | |
7ed2c31d AK |
1163 | static inline bool hugepage_movable_supported(struct hstate *h) |
1164 | { | |
1165 | return false; | |
1166 | } | |
1167 | ||
d92bbc27 JK |
1168 | static inline gfp_t htlb_alloc_mask(struct hstate *h) |
1169 | { | |
1170 | return 0; | |
1171 | } | |
1172 | ||
19fc7bed JK |
1173 | static inline gfp_t htlb_modify_alloc_mask(struct hstate *h, gfp_t gfp_mask) |
1174 | { | |
1175 | return 0; | |
1176 | } | |
1177 | ||
42d0c3fb BW |
1178 | static inline bool htlb_allow_alloc_fallback(int reason) |
1179 | { | |
1180 | return false; | |
1181 | } | |
1182 | ||
cb900f41 KS |
1183 | static inline spinlock_t *huge_pte_lockptr(struct hstate *h, |
1184 | struct mm_struct *mm, pte_t *pte) | |
1185 | { | |
1186 | return &mm->page_table_lock; | |
1187 | } | |
5d317b2b | 1188 | |
13db8c50 LZ |
1189 | static inline void hugetlb_count_init(struct mm_struct *mm) |
1190 | { | |
1191 | } | |
1192 | ||
5d317b2b NH |
1193 | static inline void hugetlb_report_usage(struct seq_file *f, struct mm_struct *m) |
1194 | { | |
1195 | } | |
1196 | ||
1197 | static inline void hugetlb_count_sub(long l, struct mm_struct *mm) | |
1198 | { | |
1199 | } | |
e5251fd4 | 1200 | |
5d4af619 BW |
1201 | static inline pte_t huge_ptep_clear_flush(struct vm_area_struct *vma, |
1202 | unsigned long addr, pte_t *ptep) | |
1203 | { | |
c33c7948 RR |
1204 | #ifdef CONFIG_MMU |
1205 | return ptep_get(ptep); | |
1206 | #else | |
5d4af619 | 1207 | return *ptep; |
c33c7948 | 1208 | #endif |
5d4af619 BW |
1209 | } |
1210 | ||
1211 | static inline void set_huge_pte_at(struct mm_struct *mm, unsigned long addr, | |
935d4f0c | 1212 | pte_t *ptep, pte_t pte, unsigned long sz) |
5d4af619 BW |
1213 | { |
1214 | } | |
a4a00b45 MS |
1215 | |
1216 | static inline void hugetlb_register_node(struct node *node) | |
1217 | { | |
1218 | } | |
1219 | ||
1220 | static inline void hugetlb_unregister_node(struct node *node) | |
1221 | { | |
1222 | } | |
24334e78 PX |
1223 | |
1224 | static inline bool hugetlbfs_pagecache_present( | |
1225 | struct hstate *h, struct vm_area_struct *vma, unsigned long address) | |
1226 | { | |
1227 | return false; | |
1228 | } | |
af73e4d9 | 1229 | #endif /* CONFIG_HUGETLB_PAGE */ |
a5516438 | 1230 | |
cb900f41 KS |
1231 | static inline spinlock_t *huge_pte_lock(struct hstate *h, |
1232 | struct mm_struct *mm, pte_t *pte) | |
1233 | { | |
1234 | spinlock_t *ptl; | |
1235 | ||
1236 | ptl = huge_pte_lockptr(h, mm, pte); | |
1237 | spin_lock(ptl); | |
1238 | return ptl; | |
1239 | } | |
1240 | ||
cf11e85f RG |
1241 | #if defined(CONFIG_HUGETLB_PAGE) && defined(CONFIG_CMA) |
1242 | extern void __init hugetlb_cma_reserve(int order); | |
cf11e85f RG |
1243 | #else |
1244 | static inline __init void hugetlb_cma_reserve(int order) | |
1245 | { | |
1246 | } | |
cf11e85f RG |
1247 | #endif |
1248 | ||
3489dbb6 MK |
1249 | #ifdef CONFIG_ARCH_WANT_HUGE_PMD_SHARE |
1250 | static inline bool hugetlb_pmd_shared(pte_t *pte) | |
1251 | { | |
1252 | return page_count(virt_to_page(pte)) > 1; | |
1253 | } | |
1254 | #else | |
1255 | static inline bool hugetlb_pmd_shared(pte_t *pte) | |
1256 | { | |
1257 | return false; | |
1258 | } | |
1259 | #endif | |
1260 | ||
c1991e07 PX |
1261 | bool want_pmd_share(struct vm_area_struct *vma, unsigned long addr); |
1262 | ||
537cf30b PX |
1263 | #ifndef __HAVE_ARCH_FLUSH_HUGETLB_TLB_RANGE |
1264 | /* | |
1265 | * ARCHes with special requirements for evicting HUGETLB backing TLB entries can | |
1266 | * implement this. | |
1267 | */ | |
1268 | #define flush_hugetlb_tlb_range(vma, addr, end) flush_tlb_range(vma, addr, end) | |
1269 | #endif | |
1270 | ||
9c67a207 PX |
1271 | static inline bool __vma_shareable_lock(struct vm_area_struct *vma) |
1272 | { | |
1273 | return (vma->vm_flags & VM_MAYSHARE) && vma->vm_private_data; | |
1274 | } | |
1275 | ||
187da0f8 | 1276 | bool __vma_private_lock(struct vm_area_struct *vma); |
bf491692 | 1277 | |
9c67a207 PX |
1278 | /* |
1279 | * Safe version of huge_pte_offset() to check the locks. See comments | |
1280 | * above huge_pte_offset(). | |
1281 | */ | |
1282 | static inline pte_t * | |
1283 | hugetlb_walk(struct vm_area_struct *vma, unsigned long addr, unsigned long sz) | |
1284 | { | |
1285 | #if defined(CONFIG_HUGETLB_PAGE) && \ | |
1286 | defined(CONFIG_ARCH_WANT_HUGE_PMD_SHARE) && defined(CONFIG_LOCKDEP) | |
1287 | struct hugetlb_vma_lock *vma_lock = vma->vm_private_data; | |
1288 | ||
1289 | /* | |
1290 | * If pmd sharing possible, locking needed to safely walk the | |
1291 | * hugetlb pgtables. More information can be found at the comment | |
1292 | * above huge_pte_offset() in the same file. | |
1293 | * | |
1294 | * NOTE: lockdep_is_held() is only defined with CONFIG_LOCKDEP. | |
1295 | */ | |
1296 | if (__vma_shareable_lock(vma)) | |
1297 | WARN_ON_ONCE(!lockdep_is_held(&vma_lock->rw_sema) && | |
1298 | !lockdep_is_held( | |
1299 | &vma->vm_file->f_mapping->i_mmap_rwsem)); | |
1300 | #endif | |
1301 | return huge_pte_offset(vma->vm_mm, addr, sz); | |
1302 | } | |
1303 | ||
1da177e4 | 1304 | #endif /* _LINUX_HUGETLB_H */ |