Merge tag 'vfio-v3.19-rc4' of git://github.com/awilliam/linux-vfio
[linux-2.6-block.git] / include / linux / hugetlb.h
CommitLineData
1da177e4
LT
1#ifndef _LINUX_HUGETLB_H
2#define _LINUX_HUGETLB_H
3
be93d8cf 4#include <linux/mm_types.h>
309381fe 5#include <linux/mmdebug.h>
4e950f6f 6#include <linux/fs.h>
8edf344c 7#include <linux/hugetlb_inline.h>
abb8206c 8#include <linux/cgroup.h>
9119a41e
JK
9#include <linux/list.h>
10#include <linux/kref.h>
4e950f6f 11
e9ea0e2d
AM
12struct ctl_table;
13struct user_struct;
24669e58 14struct mmu_gather;
e9ea0e2d 15
1da177e4
LT
16#ifdef CONFIG_HUGETLB_PAGE
17
18#include <linux/mempolicy.h>
516dffdc 19#include <linux/shm.h>
63551ae0 20#include <asm/tlbflush.h>
1da177e4 21
90481622
DG
22struct hugepage_subpool {
23 spinlock_t lock;
24 long count;
25 long max_hpages, used_hpages;
26};
27
9119a41e
JK
28struct resv_map {
29 struct kref refs;
7b24d861 30 spinlock_t lock;
9119a41e
JK
31 struct list_head regions;
32};
33extern struct resv_map *resv_map_alloc(void);
34void resv_map_release(struct kref *ref);
35
c3f38a38
AK
36extern spinlock_t hugetlb_lock;
37extern int hugetlb_max_hstate __read_mostly;
38#define for_each_hstate(h) \
39 for ((h) = hstates; (h) < &hstates[hugetlb_max_hstate]; (h)++)
40
90481622
DG
41struct hugepage_subpool *hugepage_new_subpool(long nr_blocks);
42void hugepage_put_subpool(struct hugepage_subpool *spool);
43
20a0307c
WF
44int PageHuge(struct page *page);
45
a1e78772 46void reset_vma_resv_huge_pages(struct vm_area_struct *vma);
8d65af78
AD
47int hugetlb_sysctl_handler(struct ctl_table *, int, void __user *, size_t *, loff_t *);
48int hugetlb_overcommit_handler(struct ctl_table *, int, void __user *, size_t *, loff_t *);
49int hugetlb_treat_movable_handler(struct ctl_table *, int, void __user *, size_t *, loff_t *);
06808b08
LS
50
51#ifdef CONFIG_NUMA
52int hugetlb_mempolicy_sysctl_handler(struct ctl_table *, int,
53 void __user *, size_t *, loff_t *);
54#endif
55
1da177e4 56int copy_hugetlb_page_range(struct mm_struct *, struct mm_struct *, struct vm_area_struct *);
28a35716
ML
57long follow_hugetlb_page(struct mm_struct *, struct vm_area_struct *,
58 struct page **, struct vm_area_struct **,
59 unsigned long *, unsigned long *, long, unsigned int);
04f2cbe3 60void unmap_hugepage_range(struct vm_area_struct *,
24669e58 61 unsigned long, unsigned long, struct page *);
d833352a
MG
62void __unmap_hugepage_range_final(struct mmu_gather *tlb,
63 struct vm_area_struct *vma,
64 unsigned long start, unsigned long end,
65 struct page *ref_page);
24669e58
AK
66void __unmap_hugepage_range(struct mmu_gather *tlb, struct vm_area_struct *vma,
67 unsigned long start, unsigned long end,
68 struct page *ref_page);
e1759c21 69void hugetlb_report_meminfo(struct seq_file *);
1da177e4 70int hugetlb_report_node_meminfo(int, char *);
949f7ec5 71void hugetlb_show_meminfo(void);
1da177e4 72unsigned long hugetlb_total_pages(void);
ac9b9c66 73int hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
788c7df4 74 unsigned long address, unsigned int flags);
a1e78772 75int hugetlb_reserve_pages(struct inode *inode, long from, long to,
5a6fe125 76 struct vm_area_struct *vma,
ca16d140 77 vm_flags_t vm_flags);
a43a8c39 78void hugetlb_unreserve_pages(struct inode *inode, long offset, long freed);
6de2b1aa 79int dequeue_hwpoisoned_huge_page(struct page *page);
31caf665
NH
80bool isolate_huge_page(struct page *page, struct list_head *list);
81void putback_active_hugepage(struct page *page);
c8721bbb 82bool is_hugepage_active(struct page *page);
8f1d26d0 83void free_huge_page(struct page *page);
1da177e4 84
3212b535
SC
85#ifdef CONFIG_ARCH_WANT_HUGE_PMD_SHARE
86pte_t *huge_pmd_share(struct mm_struct *mm, unsigned long addr, pud_t *pud);
87#endif
88
396faf03 89extern unsigned long hugepages_treat_as_movable;
1da177e4 90extern int sysctl_hugetlb_shm_group;
53ba51d2 91extern struct list_head huge_boot_pages;
1da177e4 92
63551ae0
DG
93/* arch callbacks */
94
a5516438
AK
95pte_t *huge_pte_alloc(struct mm_struct *mm,
96 unsigned long addr, unsigned long sz);
63551ae0 97pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr);
39dde65c 98int huge_pmd_unshare(struct mm_struct *mm, unsigned long *addr, pte_t *ptep);
63551ae0
DG
99struct page *follow_huge_addr(struct mm_struct *mm, unsigned long address,
100 int write);
101struct page *follow_huge_pmd(struct mm_struct *mm, unsigned long address,
102 pmd_t *pmd, int write);
ceb86879
AK
103struct page *follow_huge_pud(struct mm_struct *mm, unsigned long address,
104 pud_t *pud, int write);
63551ae0 105int pmd_huge(pmd_t pmd);
ceb86879 106int pud_huge(pud_t pmd);
7da4d641 107unsigned long hugetlb_change_protection(struct vm_area_struct *vma,
8f860591 108 unsigned long address, unsigned long end, pgprot_t newprot);
63551ae0 109
1da177e4
LT
110#else /* !CONFIG_HUGETLB_PAGE */
111
20a0307c
WF
112static inline int PageHuge(struct page *page)
113{
114 return 0;
115}
116
a1e78772
MG
117static inline void reset_vma_resv_huge_pages(struct vm_area_struct *vma)
118{
119}
120
1da177e4
LT
121static inline unsigned long hugetlb_total_pages(void)
122{
123 return 0;
124}
125
5b23dbe8 126#define follow_hugetlb_page(m,v,p,vs,a,b,i,w) ({ BUG(); 0; })
1da177e4
LT
127#define follow_huge_addr(mm, addr, write) ERR_PTR(-EINVAL)
128#define copy_hugetlb_page_range(src, dst, vma) ({ BUG(); 0; })
e1759c21
AD
129static inline void hugetlb_report_meminfo(struct seq_file *m)
130{
131}
1da177e4 132#define hugetlb_report_node_meminfo(n, buf) 0
949f7ec5
DR
133static inline void hugetlb_show_meminfo(void)
134{
135}
1da177e4 136#define follow_huge_pmd(mm, addr, pmd, write) NULL
ceb86879 137#define follow_huge_pud(mm, addr, pud, write) NULL
a5516438 138#define prepare_hugepage_range(file, addr, len) (-EINVAL)
1da177e4 139#define pmd_huge(x) 0
ceb86879 140#define pud_huge(x) 0
1da177e4 141#define is_hugepage_only_range(mm, addr, len) 0
9da61aef 142#define hugetlb_free_pgd_range(tlb, addr, end, floor, ceiling) ({BUG(); 0; })
788c7df4 143#define hugetlb_fault(mm, vma, addr, flags) ({ BUG(); 0; })
0fe6e20b 144#define huge_pte_offset(mm, address) 0
24669e58
AK
145static inline int dequeue_hwpoisoned_huge_page(struct page *page)
146{
147 return 0;
148}
149
f40386a4
NH
150static inline bool isolate_huge_page(struct page *page, struct list_head *list)
151{
152 return false;
153}
31caf665 154#define putback_active_hugepage(p) do {} while (0)
c8721bbb 155#define is_hugepage_active(x) false
1da177e4 156
7da4d641
PZ
157static inline unsigned long hugetlb_change_protection(struct vm_area_struct *vma,
158 unsigned long address, unsigned long end, pgprot_t newprot)
159{
160 return 0;
161}
8f860591 162
d833352a
MG
163static inline void __unmap_hugepage_range_final(struct mmu_gather *tlb,
164 struct vm_area_struct *vma, unsigned long start,
165 unsigned long end, struct page *ref_page)
166{
167 BUG();
168}
169
24669e58
AK
170static inline void __unmap_hugepage_range(struct mmu_gather *tlb,
171 struct vm_area_struct *vma, unsigned long start,
172 unsigned long end, struct page *ref_page)
173{
174 BUG();
175}
176
1da177e4 177#endif /* !CONFIG_HUGETLB_PAGE */
f30c59e9
AK
178/*
179 * hugepages at page global directory. If arch support
180 * hugepages at pgd level, they need to define this.
181 */
182#ifndef pgd_huge
183#define pgd_huge(x) 0
184#endif
185
186#ifndef pgd_write
187static inline int pgd_write(pgd_t pgd)
188{
189 BUG();
190 return 0;
191}
192#endif
193
194#ifndef pud_write
195static inline int pud_write(pud_t pud)
196{
197 BUG();
198 return 0;
199}
200#endif
201
202#ifndef is_hugepd
203/*
204 * Some architectures requires a hugepage directory format that is
205 * required to support multiple hugepage sizes. For example
206 * a4fe3ce76 "powerpc/mm: Allow more flexible layouts for hugepage pagetables"
207 * introduced the same on powerpc. This allows for a more flexible hugepage
208 * pagetable layout.
209 */
210typedef struct { unsigned long pd; } hugepd_t;
211#define is_hugepd(hugepd) (0)
212#define __hugepd(x) ((hugepd_t) { (x) })
213static inline int gup_huge_pd(hugepd_t hugepd, unsigned long addr,
214 unsigned pdshift, unsigned long end,
215 int write, struct page **pages, int *nr)
216{
217 return 0;
218}
219#else
220extern int gup_huge_pd(hugepd_t hugepd, unsigned long addr,
221 unsigned pdshift, unsigned long end,
222 int write, struct page **pages, int *nr);
223#endif
1da177e4 224
4e52780d
EM
225#define HUGETLB_ANON_FILE "anon_hugepage"
226
6bfde05b
EM
227enum {
228 /*
229 * The file will be used as an shm file so shmfs accounting rules
230 * apply
231 */
232 HUGETLB_SHMFS_INODE = 1,
4e52780d
EM
233 /*
234 * The file is being created on the internal vfs mount and shmfs
235 * accounting rules do not apply
236 */
237 HUGETLB_ANONHUGE_INODE = 2,
6bfde05b
EM
238};
239
1da177e4 240#ifdef CONFIG_HUGETLBFS
1da177e4 241struct hugetlbfs_sb_info {
1da177e4
LT
242 long max_inodes; /* inodes allowed */
243 long free_inodes; /* inodes free */
244 spinlock_t stat_lock;
a137e1cc 245 struct hstate *hstate;
90481622 246 struct hugepage_subpool *spool;
1da177e4
LT
247};
248
1da177e4
LT
249static inline struct hugetlbfs_sb_info *HUGETLBFS_SB(struct super_block *sb)
250{
251 return sb->s_fs_info;
252}
253
4b6f5d20 254extern const struct file_operations hugetlbfs_file_operations;
f0f37e2f 255extern const struct vm_operations_struct hugetlb_vm_ops;
af73e4d9 256struct file *hugetlb_file_setup(const char *name, size_t size, vm_flags_t acct,
42d7395f
AK
257 struct user_struct **user, int creat_flags,
258 int page_size_log);
1da177e4
LT
259
260static inline int is_file_hugepages(struct file *file)
261{
516dffdc
AL
262 if (file->f_op == &hugetlbfs_file_operations)
263 return 1;
264 if (is_file_shm_hugepages(file))
265 return 1;
266
267 return 0;
1da177e4
LT
268}
269
42d7395f 270
1da177e4
LT
271#else /* !CONFIG_HUGETLBFS */
272
1db8508c 273#define is_file_hugepages(file) 0
40716e29 274static inline struct file *
af73e4d9
NH
275hugetlb_file_setup(const char *name, size_t size, vm_flags_t acctflag,
276 struct user_struct **user, int creat_flags,
42d7395f 277 int page_size_log)
e9ea0e2d
AM
278{
279 return ERR_PTR(-ENOSYS);
280}
1da177e4
LT
281
282#endif /* !CONFIG_HUGETLBFS */
283
d2ba27e8
AB
284#ifdef HAVE_ARCH_HUGETLB_UNMAPPED_AREA
285unsigned long hugetlb_get_unmapped_area(struct file *file, unsigned long addr,
286 unsigned long len, unsigned long pgoff,
287 unsigned long flags);
288#endif /* HAVE_ARCH_HUGETLB_UNMAPPED_AREA */
289
a5516438
AK
290#ifdef CONFIG_HUGETLB_PAGE
291
a3437870 292#define HSTATE_NAME_LEN 32
a5516438
AK
293/* Defines one hugetlb page size */
294struct hstate {
e8c5c824
LS
295 int next_nid_to_alloc;
296 int next_nid_to_free;
a5516438
AK
297 unsigned int order;
298 unsigned long mask;
299 unsigned long max_huge_pages;
300 unsigned long nr_huge_pages;
301 unsigned long free_huge_pages;
302 unsigned long resv_huge_pages;
303 unsigned long surplus_huge_pages;
304 unsigned long nr_overcommit_huge_pages;
0edaecfa 305 struct list_head hugepage_activelist;
a5516438
AK
306 struct list_head hugepage_freelists[MAX_NUMNODES];
307 unsigned int nr_huge_pages_node[MAX_NUMNODES];
308 unsigned int free_huge_pages_node[MAX_NUMNODES];
309 unsigned int surplus_huge_pages_node[MAX_NUMNODES];
abb8206c
AK
310#ifdef CONFIG_CGROUP_HUGETLB
311 /* cgroup control files */
312 struct cftype cgroup_files[5];
313#endif
a3437870 314 char name[HSTATE_NAME_LEN];
a5516438
AK
315};
316
53ba51d2
JT
317struct huge_bootmem_page {
318 struct list_head list;
319 struct hstate *hstate;
ee8f248d
BB
320#ifdef CONFIG_HIGHMEM
321 phys_addr_t phys;
322#endif
53ba51d2
JT
323};
324
bf50bab2 325struct page *alloc_huge_page_node(struct hstate *h, int nid);
74060e4d
NH
326struct page *alloc_huge_page_noerr(struct vm_area_struct *vma,
327 unsigned long addr, int avoid_reserve);
bf50bab2 328
53ba51d2
JT
329/* arch callback */
330int __init alloc_bootmem_huge_page(struct hstate *h);
331
e5ff2159
AK
332void __init hugetlb_add_hstate(unsigned order);
333struct hstate *size_to_hstate(unsigned long size);
334
335#ifndef HUGE_MAX_HSTATE
336#define HUGE_MAX_HSTATE 1
337#endif
338
339extern struct hstate hstates[HUGE_MAX_HSTATE];
340extern unsigned int default_hstate_idx;
341
342#define default_hstate (hstates[default_hstate_idx])
a5516438 343
a137e1cc 344static inline struct hstate *hstate_inode(struct inode *i)
a5516438 345{
a137e1cc
AK
346 struct hugetlbfs_sb_info *hsb;
347 hsb = HUGETLBFS_SB(i->i_sb);
348 return hsb->hstate;
a5516438
AK
349}
350
351static inline struct hstate *hstate_file(struct file *f)
352{
496ad9aa 353 return hstate_inode(file_inode(f));
a5516438
AK
354}
355
af73e4d9
NH
356static inline struct hstate *hstate_sizelog(int page_size_log)
357{
358 if (!page_size_log)
359 return &default_hstate;
97ad2be1
SL
360
361 return size_to_hstate(1UL << page_size_log);
af73e4d9
NH
362}
363
a137e1cc 364static inline struct hstate *hstate_vma(struct vm_area_struct *vma)
a5516438 365{
a137e1cc 366 return hstate_file(vma->vm_file);
a5516438
AK
367}
368
369static inline unsigned long huge_page_size(struct hstate *h)
370{
371 return (unsigned long)PAGE_SIZE << h->order;
372}
373
08fba699
MG
374extern unsigned long vma_kernel_pagesize(struct vm_area_struct *vma);
375
3340289d
MG
376extern unsigned long vma_mmu_pagesize(struct vm_area_struct *vma);
377
a5516438
AK
378static inline unsigned long huge_page_mask(struct hstate *h)
379{
380 return h->mask;
381}
382
383static inline unsigned int huge_page_order(struct hstate *h)
384{
385 return h->order;
386}
387
388static inline unsigned huge_page_shift(struct hstate *h)
389{
390 return h->order + PAGE_SHIFT;
391}
392
bae7f4ae
LC
393static inline bool hstate_is_gigantic(struct hstate *h)
394{
395 return huge_page_order(h) >= MAX_ORDER;
396}
397
a5516438
AK
398static inline unsigned int pages_per_huge_page(struct hstate *h)
399{
400 return 1 << h->order;
401}
402
403static inline unsigned int blocks_per_huge_page(struct hstate *h)
404{
405 return huge_page_size(h) / 512;
406}
407
408#include <asm/hugetlb.h>
409
d9ed9faa
CM
410#ifndef arch_make_huge_pte
411static inline pte_t arch_make_huge_pte(pte_t entry, struct vm_area_struct *vma,
412 struct page *page, int writable)
413{
414 return entry;
415}
416#endif
417
e5ff2159
AK
418static inline struct hstate *page_hstate(struct page *page)
419{
309381fe 420 VM_BUG_ON_PAGE(!PageHuge(page), page);
e5ff2159
AK
421 return size_to_hstate(PAGE_SIZE << compound_order(page));
422}
423
aa50d3a7
AK
424static inline unsigned hstate_index_to_shift(unsigned index)
425{
426 return hstates[index].order + PAGE_SHIFT;
427}
428
972dc4de
AK
429static inline int hstate_index(struct hstate *h)
430{
431 return h - hstates;
432}
433
13d60f4b
ZY
434pgoff_t __basepage_index(struct page *page);
435
436/* Return page->index in PAGE_SIZE units */
437static inline pgoff_t basepage_index(struct page *page)
438{
439 if (!PageCompound(page))
440 return page->index;
441
442 return __basepage_index(page);
443}
444
c8721bbb
NH
445extern void dissolve_free_huge_pages(unsigned long start_pfn,
446 unsigned long end_pfn);
100873d7 447static inline int hugepage_migration_supported(struct hstate *h)
83467efb 448{
c177c81e
NH
449#ifdef CONFIG_ARCH_ENABLE_HUGEPAGE_MIGRATION
450 return huge_page_shift(h) == PMD_SHIFT;
451#else
452 return 0;
453#endif
83467efb 454}
c8721bbb 455
cb900f41
KS
456static inline spinlock_t *huge_pte_lockptr(struct hstate *h,
457 struct mm_struct *mm, pte_t *pte)
458{
459 if (huge_page_size(h) == PMD_SIZE)
460 return pmd_lockptr(mm, (pmd_t *) pte);
461 VM_BUG_ON(huge_page_size(h) == PAGE_SIZE);
462 return &mm->page_table_lock;
463}
464
457c1b27
NA
465static inline bool hugepages_supported(void)
466{
467 /*
468 * Some platform decide whether they support huge pages at boot
469 * time. On these, such as powerpc, HPAGE_SHIFT is set to 0 when
470 * there is no such support
471 */
472 return HPAGE_SHIFT != 0;
473}
474
af73e4d9 475#else /* CONFIG_HUGETLB_PAGE */
a5516438 476struct hstate {};
bf50bab2 477#define alloc_huge_page_node(h, nid) NULL
74060e4d 478#define alloc_huge_page_noerr(v, a, r) NULL
53ba51d2 479#define alloc_bootmem_huge_page(h) NULL
a5516438 480#define hstate_file(f) NULL
af73e4d9 481#define hstate_sizelog(s) NULL
a5516438
AK
482#define hstate_vma(v) NULL
483#define hstate_inode(i) NULL
cb900f41 484#define page_hstate(page) NULL
a5516438
AK
485#define huge_page_size(h) PAGE_SIZE
486#define huge_page_mask(h) PAGE_MASK
08fba699 487#define vma_kernel_pagesize(v) PAGE_SIZE
3340289d 488#define vma_mmu_pagesize(v) PAGE_SIZE
a5516438
AK
489#define huge_page_order(h) 0
490#define huge_page_shift(h) PAGE_SHIFT
510a35d4
AR
491static inline unsigned int pages_per_huge_page(struct hstate *h)
492{
493 return 1;
494}
aa50d3a7 495#define hstate_index_to_shift(index) 0
972dc4de 496#define hstate_index(h) 0
13d60f4b
ZY
497
498static inline pgoff_t basepage_index(struct page *page)
499{
500 return page->index;
501}
c8721bbb 502#define dissolve_free_huge_pages(s, e) do {} while (0)
100873d7 503#define hugepage_migration_supported(h) 0
cb900f41
KS
504
505static inline spinlock_t *huge_pte_lockptr(struct hstate *h,
506 struct mm_struct *mm, pte_t *pte)
507{
508 return &mm->page_table_lock;
509}
af73e4d9 510#endif /* CONFIG_HUGETLB_PAGE */
a5516438 511
cb900f41
KS
512static inline spinlock_t *huge_pte_lock(struct hstate *h,
513 struct mm_struct *mm, pte_t *pte)
514{
515 spinlock_t *ptl;
516
517 ptl = huge_pte_lockptr(h, mm, pte);
518 spin_lock(ptl);
519 return ptl;
520}
521
1da177e4 522#endif /* _LINUX_HUGETLB_H */