]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - include/linux/mm.h
mm: introduce wrappers to access mm->nr_ptes
[mirror_ubuntu-bionic-kernel.git] / include / linux / mm.h
1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _LINUX_MM_H
3 #define _LINUX_MM_H
4
5 #include <linux/errno.h>
6
7 #ifdef __KERNEL__
8
9 #include <linux/mmdebug.h>
10 #include <linux/gfp.h>
11 #include <linux/bug.h>
12 #include <linux/list.h>
13 #include <linux/mmzone.h>
14 #include <linux/rbtree.h>
15 #include <linux/atomic.h>
16 #include <linux/debug_locks.h>
17 #include <linux/mm_types.h>
18 #include <linux/range.h>
19 #include <linux/pfn.h>
20 #include <linux/percpu-refcount.h>
21 #include <linux/bit_spinlock.h>
22 #include <linux/shrinker.h>
23 #include <linux/resource.h>
24 #include <linux/page_ext.h>
25 #include <linux/err.h>
26 #include <linux/page_ref.h>
27 #include <linux/memremap.h>
28
29 struct mempolicy;
30 struct anon_vma;
31 struct anon_vma_chain;
32 struct file_ra_state;
33 struct user_struct;
34 struct writeback_control;
35 struct bdi_writeback;
36
37 void init_mm_internals(void);
38
39 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
40 extern unsigned long max_mapnr;
41
42 static inline void set_max_mapnr(unsigned long limit)
43 {
44 max_mapnr = limit;
45 }
46 #else
47 static inline void set_max_mapnr(unsigned long limit) { }
48 #endif
49
50 extern unsigned long totalram_pages;
51 extern void * high_memory;
52 extern int page_cluster;
53
54 #ifdef CONFIG_SYSCTL
55 extern int sysctl_legacy_va_layout;
56 #else
57 #define sysctl_legacy_va_layout 0
58 #endif
59
60 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
61 extern const int mmap_rnd_bits_min;
62 extern const int mmap_rnd_bits_max;
63 extern int mmap_rnd_bits __read_mostly;
64 #endif
65 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
66 extern const int mmap_rnd_compat_bits_min;
67 extern const int mmap_rnd_compat_bits_max;
68 extern int mmap_rnd_compat_bits __read_mostly;
69 #endif
70
71 #include <asm/page.h>
72 #include <asm/pgtable.h>
73 #include <asm/processor.h>
74
75 #ifndef __pa_symbol
76 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
77 #endif
78
79 #ifndef page_to_virt
80 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
81 #endif
82
83 #ifndef lm_alias
84 #define lm_alias(x) __va(__pa_symbol(x))
85 #endif
86
87 /*
88 * To prevent common memory management code establishing
89 * a zero page mapping on a read fault.
90 * This macro should be defined within <asm/pgtable.h>.
91 * s390 does this to prevent multiplexing of hardware bits
92 * related to the physical page in case of virtualization.
93 */
94 #ifndef mm_forbids_zeropage
95 #define mm_forbids_zeropage(X) (0)
96 #endif
97
98 /*
99 * Default maximum number of active map areas, this limits the number of vmas
100 * per mm struct. Users can overwrite this number by sysctl but there is a
101 * problem.
102 *
103 * When a program's coredump is generated as ELF format, a section is created
104 * per a vma. In ELF, the number of sections is represented in unsigned short.
105 * This means the number of sections should be smaller than 65535 at coredump.
106 * Because the kernel adds some informative sections to a image of program at
107 * generating coredump, we need some margin. The number of extra sections is
108 * 1-3 now and depends on arch. We use "5" as safe margin, here.
109 *
110 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
111 * not a hard limit any more. Although some userspace tools can be surprised by
112 * that.
113 */
114 #define MAPCOUNT_ELF_CORE_MARGIN (5)
115 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
116
117 extern int sysctl_max_map_count;
118
119 extern unsigned long sysctl_user_reserve_kbytes;
120 extern unsigned long sysctl_admin_reserve_kbytes;
121
122 extern int sysctl_overcommit_memory;
123 extern int sysctl_overcommit_ratio;
124 extern unsigned long sysctl_overcommit_kbytes;
125
126 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
127 size_t *, loff_t *);
128 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
129 size_t *, loff_t *);
130
131 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
132
133 /* to align the pointer to the (next) page boundary */
134 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
135
136 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
137 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
138
139 /*
140 * Linux kernel virtual memory manager primitives.
141 * The idea being to have a "virtual" mm in the same way
142 * we have a virtual fs - giving a cleaner interface to the
143 * mm details, and allowing different kinds of memory mappings
144 * (from shared memory to executable loading to arbitrary
145 * mmap() functions).
146 */
147
148 extern struct kmem_cache *vm_area_cachep;
149
150 #ifndef CONFIG_MMU
151 extern struct rb_root nommu_region_tree;
152 extern struct rw_semaphore nommu_region_sem;
153
154 extern unsigned int kobjsize(const void *objp);
155 #endif
156
157 /*
158 * vm_flags in vm_area_struct, see mm_types.h.
159 * When changing, update also include/trace/events/mmflags.h
160 */
161 #define VM_NONE 0x00000000
162
163 #define VM_READ 0x00000001 /* currently active flags */
164 #define VM_WRITE 0x00000002
165 #define VM_EXEC 0x00000004
166 #define VM_SHARED 0x00000008
167
168 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
169 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
170 #define VM_MAYWRITE 0x00000020
171 #define VM_MAYEXEC 0x00000040
172 #define VM_MAYSHARE 0x00000080
173
174 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
175 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
176 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
177 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
178 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
179
180 #define VM_LOCKED 0x00002000
181 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
182
183 /* Used by sys_madvise() */
184 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
185 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
186
187 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
188 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
189 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
190 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
191 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
192 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
193 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
194 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
195 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
196
197 #ifdef CONFIG_MEM_SOFT_DIRTY
198 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
199 #else
200 # define VM_SOFTDIRTY 0
201 #endif
202
203 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
204 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
205 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
206 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
207
208 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
209 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
210 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
211 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
212 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
213 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
214 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
215 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
216 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
217 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
218 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
219 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
220
221 #if defined(CONFIG_X86)
222 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
223 #if defined (CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS)
224 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
225 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
226 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1
227 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
228 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
229 #endif
230 #elif defined(CONFIG_PPC)
231 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
232 #elif defined(CONFIG_PARISC)
233 # define VM_GROWSUP VM_ARCH_1
234 #elif defined(CONFIG_METAG)
235 # define VM_GROWSUP VM_ARCH_1
236 #elif defined(CONFIG_IA64)
237 # define VM_GROWSUP VM_ARCH_1
238 #elif !defined(CONFIG_MMU)
239 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
240 #endif
241
242 #if defined(CONFIG_X86_INTEL_MPX)
243 /* MPX specific bounds table or bounds directory */
244 # define VM_MPX VM_HIGH_ARCH_4
245 #else
246 # define VM_MPX VM_NONE
247 #endif
248
249 #ifndef VM_GROWSUP
250 # define VM_GROWSUP VM_NONE
251 #endif
252
253 /* Bits set in the VMA until the stack is in its final location */
254 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
255
256 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
257 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
258 #endif
259
260 #ifdef CONFIG_STACK_GROWSUP
261 #define VM_STACK VM_GROWSUP
262 #else
263 #define VM_STACK VM_GROWSDOWN
264 #endif
265
266 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
267
268 /*
269 * Special vmas that are non-mergable, non-mlock()able.
270 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
271 */
272 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
273
274 /* This mask defines which mm->def_flags a process can inherit its parent */
275 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
276
277 /* This mask is used to clear all the VMA flags used by mlock */
278 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
279
280 /*
281 * mapping from the currently active vm_flags protection bits (the
282 * low four bits) to a page protection mask..
283 */
284 extern pgprot_t protection_map[16];
285
286 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
287 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
288 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
289 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
290 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
291 #define FAULT_FLAG_TRIED 0x20 /* Second try */
292 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
293 #define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */
294 #define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */
295
296 #define FAULT_FLAG_TRACE \
297 { FAULT_FLAG_WRITE, "WRITE" }, \
298 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
299 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
300 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
301 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
302 { FAULT_FLAG_TRIED, "TRIED" }, \
303 { FAULT_FLAG_USER, "USER" }, \
304 { FAULT_FLAG_REMOTE, "REMOTE" }, \
305 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }
306
307 /*
308 * vm_fault is filled by the the pagefault handler and passed to the vma's
309 * ->fault function. The vma's ->fault is responsible for returning a bitmask
310 * of VM_FAULT_xxx flags that give details about how the fault was handled.
311 *
312 * MM layer fills up gfp_mask for page allocations but fault handler might
313 * alter it if its implementation requires a different allocation context.
314 *
315 * pgoff should be used in favour of virtual_address, if possible.
316 */
317 struct vm_fault {
318 struct vm_area_struct *vma; /* Target VMA */
319 unsigned int flags; /* FAULT_FLAG_xxx flags */
320 gfp_t gfp_mask; /* gfp mask to be used for allocations */
321 pgoff_t pgoff; /* Logical page offset based on vma */
322 unsigned long address; /* Faulting virtual address */
323 pmd_t *pmd; /* Pointer to pmd entry matching
324 * the 'address' */
325 pud_t *pud; /* Pointer to pud entry matching
326 * the 'address'
327 */
328 pte_t orig_pte; /* Value of PTE at the time of fault */
329
330 struct page *cow_page; /* Page handler may use for COW fault */
331 struct mem_cgroup *memcg; /* Cgroup cow_page belongs to */
332 struct page *page; /* ->fault handlers should return a
333 * page here, unless VM_FAULT_NOPAGE
334 * is set (which is also implied by
335 * VM_FAULT_ERROR).
336 */
337 /* These three entries are valid only while holding ptl lock */
338 pte_t *pte; /* Pointer to pte entry matching
339 * the 'address'. NULL if the page
340 * table hasn't been allocated.
341 */
342 spinlock_t *ptl; /* Page table lock.
343 * Protects pte page table if 'pte'
344 * is not NULL, otherwise pmd.
345 */
346 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
347 * vm_ops->map_pages() calls
348 * alloc_set_pte() from atomic context.
349 * do_fault_around() pre-allocates
350 * page table to avoid allocation from
351 * atomic context.
352 */
353 };
354
355 /* page entry size for vm->huge_fault() */
356 enum page_entry_size {
357 PE_SIZE_PTE = 0,
358 PE_SIZE_PMD,
359 PE_SIZE_PUD,
360 };
361
362 /*
363 * These are the virtual MM functions - opening of an area, closing and
364 * unmapping it (needed to keep files on disk up-to-date etc), pointer
365 * to the functions called when a no-page or a wp-page exception occurs.
366 */
367 struct vm_operations_struct {
368 void (*open)(struct vm_area_struct * area);
369 void (*close)(struct vm_area_struct * area);
370 int (*mremap)(struct vm_area_struct * area);
371 int (*fault)(struct vm_fault *vmf);
372 int (*huge_fault)(struct vm_fault *vmf, enum page_entry_size pe_size);
373 void (*map_pages)(struct vm_fault *vmf,
374 pgoff_t start_pgoff, pgoff_t end_pgoff);
375
376 /* notification that a previously read-only page is about to become
377 * writable, if an error is returned it will cause a SIGBUS */
378 int (*page_mkwrite)(struct vm_fault *vmf);
379
380 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
381 int (*pfn_mkwrite)(struct vm_fault *vmf);
382
383 /* called by access_process_vm when get_user_pages() fails, typically
384 * for use by special VMAs that can switch between memory and hardware
385 */
386 int (*access)(struct vm_area_struct *vma, unsigned long addr,
387 void *buf, int len, int write);
388
389 /* Called by the /proc/PID/maps code to ask the vma whether it
390 * has a special name. Returning non-NULL will also cause this
391 * vma to be dumped unconditionally. */
392 const char *(*name)(struct vm_area_struct *vma);
393
394 #ifdef CONFIG_NUMA
395 /*
396 * set_policy() op must add a reference to any non-NULL @new mempolicy
397 * to hold the policy upon return. Caller should pass NULL @new to
398 * remove a policy and fall back to surrounding context--i.e. do not
399 * install a MPOL_DEFAULT policy, nor the task or system default
400 * mempolicy.
401 */
402 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
403
404 /*
405 * get_policy() op must add reference [mpol_get()] to any policy at
406 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
407 * in mm/mempolicy.c will do this automatically.
408 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
409 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
410 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
411 * must return NULL--i.e., do not "fallback" to task or system default
412 * policy.
413 */
414 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
415 unsigned long addr);
416 #endif
417 /*
418 * Called by vm_normal_page() for special PTEs to find the
419 * page for @addr. This is useful if the default behavior
420 * (using pte_page()) would not find the correct page.
421 */
422 struct page *(*find_special_page)(struct vm_area_struct *vma,
423 unsigned long addr);
424 };
425
426 struct mmu_gather;
427 struct inode;
428
429 #define page_private(page) ((page)->private)
430 #define set_page_private(page, v) ((page)->private = (v))
431
432 #if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
433 static inline int pmd_devmap(pmd_t pmd)
434 {
435 return 0;
436 }
437 static inline int pud_devmap(pud_t pud)
438 {
439 return 0;
440 }
441 static inline int pgd_devmap(pgd_t pgd)
442 {
443 return 0;
444 }
445 #endif
446
447 /*
448 * FIXME: take this include out, include page-flags.h in
449 * files which need it (119 of them)
450 */
451 #include <linux/page-flags.h>
452 #include <linux/huge_mm.h>
453
454 /*
455 * Methods to modify the page usage count.
456 *
457 * What counts for a page usage:
458 * - cache mapping (page->mapping)
459 * - private data (page->private)
460 * - page mapped in a task's page tables, each mapping
461 * is counted separately
462 *
463 * Also, many kernel routines increase the page count before a critical
464 * routine so they can be sure the page doesn't go away from under them.
465 */
466
467 /*
468 * Drop a ref, return true if the refcount fell to zero (the page has no users)
469 */
470 static inline int put_page_testzero(struct page *page)
471 {
472 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
473 return page_ref_dec_and_test(page);
474 }
475
476 /*
477 * Try to grab a ref unless the page has a refcount of zero, return false if
478 * that is the case.
479 * This can be called when MMU is off so it must not access
480 * any of the virtual mappings.
481 */
482 static inline int get_page_unless_zero(struct page *page)
483 {
484 return page_ref_add_unless(page, 1, 0);
485 }
486
487 extern int page_is_ram(unsigned long pfn);
488
489 enum {
490 REGION_INTERSECTS,
491 REGION_DISJOINT,
492 REGION_MIXED,
493 };
494
495 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
496 unsigned long desc);
497
498 /* Support for virtually mapped pages */
499 struct page *vmalloc_to_page(const void *addr);
500 unsigned long vmalloc_to_pfn(const void *addr);
501
502 /*
503 * Determine if an address is within the vmalloc range
504 *
505 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
506 * is no special casing required.
507 */
508 static inline bool is_vmalloc_addr(const void *x)
509 {
510 #ifdef CONFIG_MMU
511 unsigned long addr = (unsigned long)x;
512
513 return addr >= VMALLOC_START && addr < VMALLOC_END;
514 #else
515 return false;
516 #endif
517 }
518 #ifdef CONFIG_MMU
519 extern int is_vmalloc_or_module_addr(const void *x);
520 #else
521 static inline int is_vmalloc_or_module_addr(const void *x)
522 {
523 return 0;
524 }
525 #endif
526
527 extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
528 static inline void *kvmalloc(size_t size, gfp_t flags)
529 {
530 return kvmalloc_node(size, flags, NUMA_NO_NODE);
531 }
532 static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
533 {
534 return kvmalloc_node(size, flags | __GFP_ZERO, node);
535 }
536 static inline void *kvzalloc(size_t size, gfp_t flags)
537 {
538 return kvmalloc(size, flags | __GFP_ZERO);
539 }
540
541 static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
542 {
543 if (size != 0 && n > SIZE_MAX / size)
544 return NULL;
545
546 return kvmalloc(n * size, flags);
547 }
548
549 extern void kvfree(const void *addr);
550
551 static inline atomic_t *compound_mapcount_ptr(struct page *page)
552 {
553 return &page[1].compound_mapcount;
554 }
555
556 static inline int compound_mapcount(struct page *page)
557 {
558 VM_BUG_ON_PAGE(!PageCompound(page), page);
559 page = compound_head(page);
560 return atomic_read(compound_mapcount_ptr(page)) + 1;
561 }
562
563 /*
564 * The atomic page->_mapcount, starts from -1: so that transitions
565 * both from it and to it can be tracked, using atomic_inc_and_test
566 * and atomic_add_negative(-1).
567 */
568 static inline void page_mapcount_reset(struct page *page)
569 {
570 atomic_set(&(page)->_mapcount, -1);
571 }
572
573 int __page_mapcount(struct page *page);
574
575 static inline int page_mapcount(struct page *page)
576 {
577 VM_BUG_ON_PAGE(PageSlab(page), page);
578
579 if (unlikely(PageCompound(page)))
580 return __page_mapcount(page);
581 return atomic_read(&page->_mapcount) + 1;
582 }
583
584 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
585 int total_mapcount(struct page *page);
586 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
587 #else
588 static inline int total_mapcount(struct page *page)
589 {
590 return page_mapcount(page);
591 }
592 static inline int page_trans_huge_mapcount(struct page *page,
593 int *total_mapcount)
594 {
595 int mapcount = page_mapcount(page);
596 if (total_mapcount)
597 *total_mapcount = mapcount;
598 return mapcount;
599 }
600 #endif
601
602 static inline struct page *virt_to_head_page(const void *x)
603 {
604 struct page *page = virt_to_page(x);
605
606 return compound_head(page);
607 }
608
609 void __put_page(struct page *page);
610
611 void put_pages_list(struct list_head *pages);
612
613 void split_page(struct page *page, unsigned int order);
614
615 /*
616 * Compound pages have a destructor function. Provide a
617 * prototype for that function and accessor functions.
618 * These are _only_ valid on the head of a compound page.
619 */
620 typedef void compound_page_dtor(struct page *);
621
622 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
623 enum compound_dtor_id {
624 NULL_COMPOUND_DTOR,
625 COMPOUND_PAGE_DTOR,
626 #ifdef CONFIG_HUGETLB_PAGE
627 HUGETLB_PAGE_DTOR,
628 #endif
629 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
630 TRANSHUGE_PAGE_DTOR,
631 #endif
632 NR_COMPOUND_DTORS,
633 };
634 extern compound_page_dtor * const compound_page_dtors[];
635
636 static inline void set_compound_page_dtor(struct page *page,
637 enum compound_dtor_id compound_dtor)
638 {
639 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
640 page[1].compound_dtor = compound_dtor;
641 }
642
643 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
644 {
645 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
646 return compound_page_dtors[page[1].compound_dtor];
647 }
648
649 static inline unsigned int compound_order(struct page *page)
650 {
651 if (!PageHead(page))
652 return 0;
653 return page[1].compound_order;
654 }
655
656 static inline void set_compound_order(struct page *page, unsigned int order)
657 {
658 page[1].compound_order = order;
659 }
660
661 void free_compound_page(struct page *page);
662
663 #ifdef CONFIG_MMU
664 /*
665 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
666 * servicing faults for write access. In the normal case, do always want
667 * pte_mkwrite. But get_user_pages can cause write faults for mappings
668 * that do not have writing enabled, when used by access_process_vm.
669 */
670 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
671 {
672 if (likely(vma->vm_flags & VM_WRITE))
673 pte = pte_mkwrite(pte);
674 return pte;
675 }
676
677 int alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
678 struct page *page);
679 int finish_fault(struct vm_fault *vmf);
680 int finish_mkwrite_fault(struct vm_fault *vmf);
681 #endif
682
683 /*
684 * Multiple processes may "see" the same page. E.g. for untouched
685 * mappings of /dev/null, all processes see the same page full of
686 * zeroes, and text pages of executables and shared libraries have
687 * only one copy in memory, at most, normally.
688 *
689 * For the non-reserved pages, page_count(page) denotes a reference count.
690 * page_count() == 0 means the page is free. page->lru is then used for
691 * freelist management in the buddy allocator.
692 * page_count() > 0 means the page has been allocated.
693 *
694 * Pages are allocated by the slab allocator in order to provide memory
695 * to kmalloc and kmem_cache_alloc. In this case, the management of the
696 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
697 * unless a particular usage is carefully commented. (the responsibility of
698 * freeing the kmalloc memory is the caller's, of course).
699 *
700 * A page may be used by anyone else who does a __get_free_page().
701 * In this case, page_count still tracks the references, and should only
702 * be used through the normal accessor functions. The top bits of page->flags
703 * and page->virtual store page management information, but all other fields
704 * are unused and could be used privately, carefully. The management of this
705 * page is the responsibility of the one who allocated it, and those who have
706 * subsequently been given references to it.
707 *
708 * The other pages (we may call them "pagecache pages") are completely
709 * managed by the Linux memory manager: I/O, buffers, swapping etc.
710 * The following discussion applies only to them.
711 *
712 * A pagecache page contains an opaque `private' member, which belongs to the
713 * page's address_space. Usually, this is the address of a circular list of
714 * the page's disk buffers. PG_private must be set to tell the VM to call
715 * into the filesystem to release these pages.
716 *
717 * A page may belong to an inode's memory mapping. In this case, page->mapping
718 * is the pointer to the inode, and page->index is the file offset of the page,
719 * in units of PAGE_SIZE.
720 *
721 * If pagecache pages are not associated with an inode, they are said to be
722 * anonymous pages. These may become associated with the swapcache, and in that
723 * case PG_swapcache is set, and page->private is an offset into the swapcache.
724 *
725 * In either case (swapcache or inode backed), the pagecache itself holds one
726 * reference to the page. Setting PG_private should also increment the
727 * refcount. The each user mapping also has a reference to the page.
728 *
729 * The pagecache pages are stored in a per-mapping radix tree, which is
730 * rooted at mapping->page_tree, and indexed by offset.
731 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
732 * lists, we instead now tag pages as dirty/writeback in the radix tree.
733 *
734 * All pagecache pages may be subject to I/O:
735 * - inode pages may need to be read from disk,
736 * - inode pages which have been modified and are MAP_SHARED may need
737 * to be written back to the inode on disk,
738 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
739 * modified may need to be swapped out to swap space and (later) to be read
740 * back into memory.
741 */
742
743 /*
744 * The zone field is never updated after free_area_init_core()
745 * sets it, so none of the operations on it need to be atomic.
746 */
747
748 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
749 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
750 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
751 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
752 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
753
754 /*
755 * Define the bit shifts to access each section. For non-existent
756 * sections we define the shift as 0; that plus a 0 mask ensures
757 * the compiler will optimise away reference to them.
758 */
759 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
760 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
761 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
762 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
763
764 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
765 #ifdef NODE_NOT_IN_PAGE_FLAGS
766 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
767 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
768 SECTIONS_PGOFF : ZONES_PGOFF)
769 #else
770 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
771 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
772 NODES_PGOFF : ZONES_PGOFF)
773 #endif
774
775 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
776
777 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
778 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
779 #endif
780
781 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
782 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
783 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
784 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
785 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
786
787 static inline enum zone_type page_zonenum(const struct page *page)
788 {
789 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
790 }
791
792 #ifdef CONFIG_ZONE_DEVICE
793 static inline bool is_zone_device_page(const struct page *page)
794 {
795 return page_zonenum(page) == ZONE_DEVICE;
796 }
797 #else
798 static inline bool is_zone_device_page(const struct page *page)
799 {
800 return false;
801 }
802 #endif
803
804 #if defined(CONFIG_DEVICE_PRIVATE) || defined(CONFIG_DEVICE_PUBLIC)
805 void put_zone_device_private_or_public_page(struct page *page);
806 DECLARE_STATIC_KEY_FALSE(device_private_key);
807 #define IS_HMM_ENABLED static_branch_unlikely(&device_private_key)
808 static inline bool is_device_private_page(const struct page *page);
809 static inline bool is_device_public_page(const struct page *page);
810 #else /* CONFIG_DEVICE_PRIVATE || CONFIG_DEVICE_PUBLIC */
811 static inline void put_zone_device_private_or_public_page(struct page *page)
812 {
813 }
814 #define IS_HMM_ENABLED 0
815 static inline bool is_device_private_page(const struct page *page)
816 {
817 return false;
818 }
819 static inline bool is_device_public_page(const struct page *page)
820 {
821 return false;
822 }
823 #endif /* CONFIG_DEVICE_PRIVATE || CONFIG_DEVICE_PUBLIC */
824
825
826 static inline void get_page(struct page *page)
827 {
828 page = compound_head(page);
829 /*
830 * Getting a normal page or the head of a compound page
831 * requires to already have an elevated page->_refcount.
832 */
833 VM_BUG_ON_PAGE(page_ref_count(page) <= 0, page);
834 page_ref_inc(page);
835 }
836
837 static inline void put_page(struct page *page)
838 {
839 page = compound_head(page);
840
841 /*
842 * For private device pages we need to catch refcount transition from
843 * 2 to 1, when refcount reach one it means the private device page is
844 * free and we need to inform the device driver through callback. See
845 * include/linux/memremap.h and HMM for details.
846 */
847 if (IS_HMM_ENABLED && unlikely(is_device_private_page(page) ||
848 unlikely(is_device_public_page(page)))) {
849 put_zone_device_private_or_public_page(page);
850 return;
851 }
852
853 if (put_page_testzero(page))
854 __put_page(page);
855 }
856
857 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
858 #define SECTION_IN_PAGE_FLAGS
859 #endif
860
861 /*
862 * The identification function is mainly used by the buddy allocator for
863 * determining if two pages could be buddies. We are not really identifying
864 * the zone since we could be using the section number id if we do not have
865 * node id available in page flags.
866 * We only guarantee that it will return the same value for two combinable
867 * pages in a zone.
868 */
869 static inline int page_zone_id(struct page *page)
870 {
871 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
872 }
873
874 static inline int zone_to_nid(struct zone *zone)
875 {
876 #ifdef CONFIG_NUMA
877 return zone->node;
878 #else
879 return 0;
880 #endif
881 }
882
883 #ifdef NODE_NOT_IN_PAGE_FLAGS
884 extern int page_to_nid(const struct page *page);
885 #else
886 static inline int page_to_nid(const struct page *page)
887 {
888 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
889 }
890 #endif
891
892 #ifdef CONFIG_NUMA_BALANCING
893 static inline int cpu_pid_to_cpupid(int cpu, int pid)
894 {
895 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
896 }
897
898 static inline int cpupid_to_pid(int cpupid)
899 {
900 return cpupid & LAST__PID_MASK;
901 }
902
903 static inline int cpupid_to_cpu(int cpupid)
904 {
905 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
906 }
907
908 static inline int cpupid_to_nid(int cpupid)
909 {
910 return cpu_to_node(cpupid_to_cpu(cpupid));
911 }
912
913 static inline bool cpupid_pid_unset(int cpupid)
914 {
915 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
916 }
917
918 static inline bool cpupid_cpu_unset(int cpupid)
919 {
920 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
921 }
922
923 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
924 {
925 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
926 }
927
928 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
929 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
930 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
931 {
932 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
933 }
934
935 static inline int page_cpupid_last(struct page *page)
936 {
937 return page->_last_cpupid;
938 }
939 static inline void page_cpupid_reset_last(struct page *page)
940 {
941 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
942 }
943 #else
944 static inline int page_cpupid_last(struct page *page)
945 {
946 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
947 }
948
949 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
950
951 static inline void page_cpupid_reset_last(struct page *page)
952 {
953 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
954 }
955 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
956 #else /* !CONFIG_NUMA_BALANCING */
957 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
958 {
959 return page_to_nid(page); /* XXX */
960 }
961
962 static inline int page_cpupid_last(struct page *page)
963 {
964 return page_to_nid(page); /* XXX */
965 }
966
967 static inline int cpupid_to_nid(int cpupid)
968 {
969 return -1;
970 }
971
972 static inline int cpupid_to_pid(int cpupid)
973 {
974 return -1;
975 }
976
977 static inline int cpupid_to_cpu(int cpupid)
978 {
979 return -1;
980 }
981
982 static inline int cpu_pid_to_cpupid(int nid, int pid)
983 {
984 return -1;
985 }
986
987 static inline bool cpupid_pid_unset(int cpupid)
988 {
989 return 1;
990 }
991
992 static inline void page_cpupid_reset_last(struct page *page)
993 {
994 }
995
996 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
997 {
998 return false;
999 }
1000 #endif /* CONFIG_NUMA_BALANCING */
1001
1002 static inline struct zone *page_zone(const struct page *page)
1003 {
1004 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1005 }
1006
1007 static inline pg_data_t *page_pgdat(const struct page *page)
1008 {
1009 return NODE_DATA(page_to_nid(page));
1010 }
1011
1012 #ifdef SECTION_IN_PAGE_FLAGS
1013 static inline void set_page_section(struct page *page, unsigned long section)
1014 {
1015 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1016 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1017 }
1018
1019 static inline unsigned long page_to_section(const struct page *page)
1020 {
1021 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1022 }
1023 #endif
1024
1025 static inline void set_page_zone(struct page *page, enum zone_type zone)
1026 {
1027 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1028 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1029 }
1030
1031 static inline void set_page_node(struct page *page, unsigned long node)
1032 {
1033 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1034 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1035 }
1036
1037 static inline void set_page_links(struct page *page, enum zone_type zone,
1038 unsigned long node, unsigned long pfn)
1039 {
1040 set_page_zone(page, zone);
1041 set_page_node(page, node);
1042 #ifdef SECTION_IN_PAGE_FLAGS
1043 set_page_section(page, pfn_to_section_nr(pfn));
1044 #endif
1045 }
1046
1047 #ifdef CONFIG_MEMCG
1048 static inline struct mem_cgroup *page_memcg(struct page *page)
1049 {
1050 return page->mem_cgroup;
1051 }
1052 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1053 {
1054 WARN_ON_ONCE(!rcu_read_lock_held());
1055 return READ_ONCE(page->mem_cgroup);
1056 }
1057 #else
1058 static inline struct mem_cgroup *page_memcg(struct page *page)
1059 {
1060 return NULL;
1061 }
1062 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1063 {
1064 WARN_ON_ONCE(!rcu_read_lock_held());
1065 return NULL;
1066 }
1067 #endif
1068
1069 /*
1070 * Some inline functions in vmstat.h depend on page_zone()
1071 */
1072 #include <linux/vmstat.h>
1073
1074 static __always_inline void *lowmem_page_address(const struct page *page)
1075 {
1076 return page_to_virt(page);
1077 }
1078
1079 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1080 #define HASHED_PAGE_VIRTUAL
1081 #endif
1082
1083 #if defined(WANT_PAGE_VIRTUAL)
1084 static inline void *page_address(const struct page *page)
1085 {
1086 return page->virtual;
1087 }
1088 static inline void set_page_address(struct page *page, void *address)
1089 {
1090 page->virtual = address;
1091 }
1092 #define page_address_init() do { } while(0)
1093 #endif
1094
1095 #if defined(HASHED_PAGE_VIRTUAL)
1096 void *page_address(const struct page *page);
1097 void set_page_address(struct page *page, void *virtual);
1098 void page_address_init(void);
1099 #endif
1100
1101 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1102 #define page_address(page) lowmem_page_address(page)
1103 #define set_page_address(page, address) do { } while(0)
1104 #define page_address_init() do { } while(0)
1105 #endif
1106
1107 extern void *page_rmapping(struct page *page);
1108 extern struct anon_vma *page_anon_vma(struct page *page);
1109 extern struct address_space *page_mapping(struct page *page);
1110
1111 extern struct address_space *__page_file_mapping(struct page *);
1112
1113 static inline
1114 struct address_space *page_file_mapping(struct page *page)
1115 {
1116 if (unlikely(PageSwapCache(page)))
1117 return __page_file_mapping(page);
1118
1119 return page->mapping;
1120 }
1121
1122 extern pgoff_t __page_file_index(struct page *page);
1123
1124 /*
1125 * Return the pagecache index of the passed page. Regular pagecache pages
1126 * use ->index whereas swapcache pages use swp_offset(->private)
1127 */
1128 static inline pgoff_t page_index(struct page *page)
1129 {
1130 if (unlikely(PageSwapCache(page)))
1131 return __page_file_index(page);
1132 return page->index;
1133 }
1134
1135 bool page_mapped(struct page *page);
1136 struct address_space *page_mapping(struct page *page);
1137
1138 /*
1139 * Return true only if the page has been allocated with
1140 * ALLOC_NO_WATERMARKS and the low watermark was not
1141 * met implying that the system is under some pressure.
1142 */
1143 static inline bool page_is_pfmemalloc(struct page *page)
1144 {
1145 /*
1146 * Page index cannot be this large so this must be
1147 * a pfmemalloc page.
1148 */
1149 return page->index == -1UL;
1150 }
1151
1152 /*
1153 * Only to be called by the page allocator on a freshly allocated
1154 * page.
1155 */
1156 static inline void set_page_pfmemalloc(struct page *page)
1157 {
1158 page->index = -1UL;
1159 }
1160
1161 static inline void clear_page_pfmemalloc(struct page *page)
1162 {
1163 page->index = 0;
1164 }
1165
1166 /*
1167 * Different kinds of faults, as returned by handle_mm_fault().
1168 * Used to decide whether a process gets delivered SIGBUS or
1169 * just gets major/minor fault counters bumped up.
1170 */
1171
1172 #define VM_FAULT_OOM 0x0001
1173 #define VM_FAULT_SIGBUS 0x0002
1174 #define VM_FAULT_MAJOR 0x0004
1175 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1176 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1177 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1178 #define VM_FAULT_SIGSEGV 0x0040
1179
1180 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1181 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1182 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1183 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1184 #define VM_FAULT_DONE_COW 0x1000 /* ->fault has fully handled COW */
1185
1186 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1187
1188 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1189 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1190 VM_FAULT_FALLBACK)
1191
1192 #define VM_FAULT_RESULT_TRACE \
1193 { VM_FAULT_OOM, "OOM" }, \
1194 { VM_FAULT_SIGBUS, "SIGBUS" }, \
1195 { VM_FAULT_MAJOR, "MAJOR" }, \
1196 { VM_FAULT_WRITE, "WRITE" }, \
1197 { VM_FAULT_HWPOISON, "HWPOISON" }, \
1198 { VM_FAULT_HWPOISON_LARGE, "HWPOISON_LARGE" }, \
1199 { VM_FAULT_SIGSEGV, "SIGSEGV" }, \
1200 { VM_FAULT_NOPAGE, "NOPAGE" }, \
1201 { VM_FAULT_LOCKED, "LOCKED" }, \
1202 { VM_FAULT_RETRY, "RETRY" }, \
1203 { VM_FAULT_FALLBACK, "FALLBACK" }, \
1204 { VM_FAULT_DONE_COW, "DONE_COW" }
1205
1206 /* Encode hstate index for a hwpoisoned large page */
1207 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1208 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1209
1210 /*
1211 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1212 */
1213 extern void pagefault_out_of_memory(void);
1214
1215 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1216
1217 /*
1218 * Flags passed to show_mem() and show_free_areas() to suppress output in
1219 * various contexts.
1220 */
1221 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1222
1223 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1224
1225 extern bool can_do_mlock(void);
1226 extern int user_shm_lock(size_t, struct user_struct *);
1227 extern void user_shm_unlock(size_t, struct user_struct *);
1228
1229 /*
1230 * Parameter block passed down to zap_pte_range in exceptional cases.
1231 */
1232 struct zap_details {
1233 struct address_space *check_mapping; /* Check page->mapping if set */
1234 pgoff_t first_index; /* Lowest page->index to unmap */
1235 pgoff_t last_index; /* Highest page->index to unmap */
1236 };
1237
1238 struct page *_vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1239 pte_t pte, bool with_public_device);
1240 #define vm_normal_page(vma, addr, pte) _vm_normal_page(vma, addr, pte, false)
1241
1242 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1243 pmd_t pmd);
1244
1245 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1246 unsigned long size);
1247 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1248 unsigned long size);
1249 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1250 unsigned long start, unsigned long end);
1251
1252 /**
1253 * mm_walk - callbacks for walk_page_range
1254 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
1255 * this handler should only handle pud_trans_huge() puds.
1256 * the pmd_entry or pte_entry callbacks will be used for
1257 * regular PUDs.
1258 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1259 * this handler is required to be able to handle
1260 * pmd_trans_huge() pmds. They may simply choose to
1261 * split_huge_page() instead of handling it explicitly.
1262 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1263 * @pte_hole: if set, called for each hole at all levels
1264 * @hugetlb_entry: if set, called for each hugetlb entry
1265 * @test_walk: caller specific callback function to determine whether
1266 * we walk over the current vma or not. Returning 0
1267 * value means "do page table walk over the current vma,"
1268 * and a negative one means "abort current page table walk
1269 * right now." 1 means "skip the current vma."
1270 * @mm: mm_struct representing the target process of page table walk
1271 * @vma: vma currently walked (NULL if walking outside vmas)
1272 * @private: private data for callbacks' usage
1273 *
1274 * (see the comment on walk_page_range() for more details)
1275 */
1276 struct mm_walk {
1277 int (*pud_entry)(pud_t *pud, unsigned long addr,
1278 unsigned long next, struct mm_walk *walk);
1279 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1280 unsigned long next, struct mm_walk *walk);
1281 int (*pte_entry)(pte_t *pte, unsigned long addr,
1282 unsigned long next, struct mm_walk *walk);
1283 int (*pte_hole)(unsigned long addr, unsigned long next,
1284 struct mm_walk *walk);
1285 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1286 unsigned long addr, unsigned long next,
1287 struct mm_walk *walk);
1288 int (*test_walk)(unsigned long addr, unsigned long next,
1289 struct mm_walk *walk);
1290 struct mm_struct *mm;
1291 struct vm_area_struct *vma;
1292 void *private;
1293 };
1294
1295 int walk_page_range(unsigned long addr, unsigned long end,
1296 struct mm_walk *walk);
1297 int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1298 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1299 unsigned long end, unsigned long floor, unsigned long ceiling);
1300 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1301 struct vm_area_struct *vma);
1302 void unmap_mapping_range(struct address_space *mapping,
1303 loff_t const holebegin, loff_t const holelen, int even_cows);
1304 int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
1305 unsigned long *start, unsigned long *end,
1306 pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp);
1307 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1308 unsigned long *pfn);
1309 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1310 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1311 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1312 void *buf, int len, int write);
1313
1314 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1315 loff_t const holebegin, loff_t const holelen)
1316 {
1317 unmap_mapping_range(mapping, holebegin, holelen, 0);
1318 }
1319
1320 extern void truncate_pagecache(struct inode *inode, loff_t new);
1321 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1322 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1323 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1324 int truncate_inode_page(struct address_space *mapping, struct page *page);
1325 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1326 int invalidate_inode_page(struct page *page);
1327
1328 #ifdef CONFIG_MMU
1329 extern int handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
1330 unsigned int flags);
1331 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1332 unsigned long address, unsigned int fault_flags,
1333 bool *unlocked);
1334 #else
1335 static inline int handle_mm_fault(struct vm_area_struct *vma,
1336 unsigned long address, unsigned int flags)
1337 {
1338 /* should never happen if there's no MMU */
1339 BUG();
1340 return VM_FAULT_SIGBUS;
1341 }
1342 static inline int fixup_user_fault(struct task_struct *tsk,
1343 struct mm_struct *mm, unsigned long address,
1344 unsigned int fault_flags, bool *unlocked)
1345 {
1346 /* should never happen if there's no MMU */
1347 BUG();
1348 return -EFAULT;
1349 }
1350 #endif
1351
1352 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len,
1353 unsigned int gup_flags);
1354 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1355 void *buf, int len, unsigned int gup_flags);
1356 extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1357 unsigned long addr, void *buf, int len, unsigned int gup_flags);
1358
1359 long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1360 unsigned long start, unsigned long nr_pages,
1361 unsigned int gup_flags, struct page **pages,
1362 struct vm_area_struct **vmas, int *locked);
1363 long get_user_pages(unsigned long start, unsigned long nr_pages,
1364 unsigned int gup_flags, struct page **pages,
1365 struct vm_area_struct **vmas);
1366 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1367 unsigned int gup_flags, struct page **pages, int *locked);
1368 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1369 struct page **pages, unsigned int gup_flags);
1370 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1371 struct page **pages);
1372
1373 /* Container for pinned pfns / pages */
1374 struct frame_vector {
1375 unsigned int nr_allocated; /* Number of frames we have space for */
1376 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1377 bool got_ref; /* Did we pin pages by getting page ref? */
1378 bool is_pfns; /* Does array contain pages or pfns? */
1379 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1380 * pfns_vector_pages() or pfns_vector_pfns()
1381 * for access */
1382 };
1383
1384 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1385 void frame_vector_destroy(struct frame_vector *vec);
1386 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1387 unsigned int gup_flags, struct frame_vector *vec);
1388 void put_vaddr_frames(struct frame_vector *vec);
1389 int frame_vector_to_pages(struct frame_vector *vec);
1390 void frame_vector_to_pfns(struct frame_vector *vec);
1391
1392 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1393 {
1394 return vec->nr_frames;
1395 }
1396
1397 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1398 {
1399 if (vec->is_pfns) {
1400 int err = frame_vector_to_pages(vec);
1401
1402 if (err)
1403 return ERR_PTR(err);
1404 }
1405 return (struct page **)(vec->ptrs);
1406 }
1407
1408 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1409 {
1410 if (!vec->is_pfns)
1411 frame_vector_to_pfns(vec);
1412 return (unsigned long *)(vec->ptrs);
1413 }
1414
1415 struct kvec;
1416 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1417 struct page **pages);
1418 int get_kernel_page(unsigned long start, int write, struct page **pages);
1419 struct page *get_dump_page(unsigned long addr);
1420
1421 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1422 extern void do_invalidatepage(struct page *page, unsigned int offset,
1423 unsigned int length);
1424
1425 int __set_page_dirty_nobuffers(struct page *page);
1426 int __set_page_dirty_no_writeback(struct page *page);
1427 int redirty_page_for_writepage(struct writeback_control *wbc,
1428 struct page *page);
1429 void account_page_dirtied(struct page *page, struct address_space *mapping);
1430 void account_page_cleaned(struct page *page, struct address_space *mapping,
1431 struct bdi_writeback *wb);
1432 int set_page_dirty(struct page *page);
1433 int set_page_dirty_lock(struct page *page);
1434 void cancel_dirty_page(struct page *page);
1435 int clear_page_dirty_for_io(struct page *page);
1436
1437 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1438
1439 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1440 {
1441 return !vma->vm_ops;
1442 }
1443
1444 #ifdef CONFIG_SHMEM
1445 /*
1446 * The vma_is_shmem is not inline because it is used only by slow
1447 * paths in userfault.
1448 */
1449 bool vma_is_shmem(struct vm_area_struct *vma);
1450 #else
1451 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
1452 #endif
1453
1454 int vma_is_stack_for_current(struct vm_area_struct *vma);
1455
1456 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1457 unsigned long old_addr, struct vm_area_struct *new_vma,
1458 unsigned long new_addr, unsigned long len,
1459 bool need_rmap_locks);
1460 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1461 unsigned long end, pgprot_t newprot,
1462 int dirty_accountable, int prot_numa);
1463 extern int mprotect_fixup(struct vm_area_struct *vma,
1464 struct vm_area_struct **pprev, unsigned long start,
1465 unsigned long end, unsigned long newflags);
1466
1467 /*
1468 * doesn't attempt to fault and will return short.
1469 */
1470 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1471 struct page **pages);
1472 /*
1473 * per-process(per-mm_struct) statistics.
1474 */
1475 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1476 {
1477 long val = atomic_long_read(&mm->rss_stat.count[member]);
1478
1479 #ifdef SPLIT_RSS_COUNTING
1480 /*
1481 * counter is updated in asynchronous manner and may go to minus.
1482 * But it's never be expected number for users.
1483 */
1484 if (val < 0)
1485 val = 0;
1486 #endif
1487 return (unsigned long)val;
1488 }
1489
1490 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1491 {
1492 atomic_long_add(value, &mm->rss_stat.count[member]);
1493 }
1494
1495 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1496 {
1497 atomic_long_inc(&mm->rss_stat.count[member]);
1498 }
1499
1500 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1501 {
1502 atomic_long_dec(&mm->rss_stat.count[member]);
1503 }
1504
1505 /* Optimized variant when page is already known not to be PageAnon */
1506 static inline int mm_counter_file(struct page *page)
1507 {
1508 if (PageSwapBacked(page))
1509 return MM_SHMEMPAGES;
1510 return MM_FILEPAGES;
1511 }
1512
1513 static inline int mm_counter(struct page *page)
1514 {
1515 if (PageAnon(page))
1516 return MM_ANONPAGES;
1517 return mm_counter_file(page);
1518 }
1519
1520 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1521 {
1522 return get_mm_counter(mm, MM_FILEPAGES) +
1523 get_mm_counter(mm, MM_ANONPAGES) +
1524 get_mm_counter(mm, MM_SHMEMPAGES);
1525 }
1526
1527 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1528 {
1529 return max(mm->hiwater_rss, get_mm_rss(mm));
1530 }
1531
1532 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1533 {
1534 return max(mm->hiwater_vm, mm->total_vm);
1535 }
1536
1537 static inline void update_hiwater_rss(struct mm_struct *mm)
1538 {
1539 unsigned long _rss = get_mm_rss(mm);
1540
1541 if ((mm)->hiwater_rss < _rss)
1542 (mm)->hiwater_rss = _rss;
1543 }
1544
1545 static inline void update_hiwater_vm(struct mm_struct *mm)
1546 {
1547 if (mm->hiwater_vm < mm->total_vm)
1548 mm->hiwater_vm = mm->total_vm;
1549 }
1550
1551 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1552 {
1553 mm->hiwater_rss = get_mm_rss(mm);
1554 }
1555
1556 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1557 struct mm_struct *mm)
1558 {
1559 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1560
1561 if (*maxrss < hiwater_rss)
1562 *maxrss = hiwater_rss;
1563 }
1564
1565 #if defined(SPLIT_RSS_COUNTING)
1566 void sync_mm_rss(struct mm_struct *mm);
1567 #else
1568 static inline void sync_mm_rss(struct mm_struct *mm)
1569 {
1570 }
1571 #endif
1572
1573 #ifndef __HAVE_ARCH_PTE_DEVMAP
1574 static inline int pte_devmap(pte_t pte)
1575 {
1576 return 0;
1577 }
1578 #endif
1579
1580 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1581
1582 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1583 spinlock_t **ptl);
1584 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1585 spinlock_t **ptl)
1586 {
1587 pte_t *ptep;
1588 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1589 return ptep;
1590 }
1591
1592 #ifdef __PAGETABLE_P4D_FOLDED
1593 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1594 unsigned long address)
1595 {
1596 return 0;
1597 }
1598 #else
1599 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1600 #endif
1601
1602 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
1603 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1604 unsigned long address)
1605 {
1606 return 0;
1607 }
1608
1609 static inline unsigned long mm_nr_puds(const struct mm_struct *mm)
1610 {
1611 return 0;
1612 }
1613
1614 static inline void mm_nr_puds_init(struct mm_struct *mm) {}
1615 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
1616 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
1617
1618 #else
1619 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
1620
1621 static inline void mm_nr_puds_init(struct mm_struct *mm)
1622 {
1623 atomic_long_set(&mm->nr_puds, 0);
1624 }
1625
1626 static inline unsigned long mm_nr_puds(const struct mm_struct *mm)
1627 {
1628 return atomic_long_read(&mm->nr_puds);
1629 }
1630
1631 static inline void mm_inc_nr_puds(struct mm_struct *mm)
1632 {
1633 atomic_long_inc(&mm->nr_puds);
1634 }
1635
1636 static inline void mm_dec_nr_puds(struct mm_struct *mm)
1637 {
1638 atomic_long_dec(&mm->nr_puds);
1639 }
1640 #endif
1641
1642 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1643 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1644 unsigned long address)
1645 {
1646 return 0;
1647 }
1648
1649 static inline void mm_nr_pmds_init(struct mm_struct *mm) {}
1650
1651 static inline unsigned long mm_nr_pmds(const struct mm_struct *mm)
1652 {
1653 return 0;
1654 }
1655
1656 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1657 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1658
1659 #else
1660 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1661
1662 static inline void mm_nr_pmds_init(struct mm_struct *mm)
1663 {
1664 atomic_long_set(&mm->nr_pmds, 0);
1665 }
1666
1667 static inline unsigned long mm_nr_pmds(const struct mm_struct *mm)
1668 {
1669 return atomic_long_read(&mm->nr_pmds);
1670 }
1671
1672 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1673 {
1674 atomic_long_inc(&mm->nr_pmds);
1675 }
1676
1677 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1678 {
1679 atomic_long_dec(&mm->nr_pmds);
1680 }
1681 #endif
1682
1683 #ifdef CONFIG_MMU
1684 static inline void mm_nr_ptes_init(struct mm_struct *mm)
1685 {
1686 atomic_long_set(&mm->nr_ptes, 0);
1687 }
1688
1689 static inline unsigned long mm_nr_ptes(const struct mm_struct *mm)
1690 {
1691 return atomic_long_read(&mm->nr_ptes);
1692 }
1693
1694 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
1695 {
1696 atomic_long_inc(&mm->nr_ptes);
1697 }
1698
1699 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
1700 {
1701 atomic_long_dec(&mm->nr_ptes);
1702 }
1703 #else
1704 static inline void mm_nr_ptes_init(struct mm_struct *mm) {}
1705
1706 static inline unsigned long mm_nr_ptes(const struct mm_struct *mm)
1707 {
1708 return 0;
1709 }
1710
1711 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
1712 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
1713 #endif
1714
1715 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
1716 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1717
1718 /*
1719 * The following ifdef needed to get the 4level-fixup.h header to work.
1720 * Remove it when 4level-fixup.h has been removed.
1721 */
1722 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1723
1724 #ifndef __ARCH_HAS_5LEVEL_HACK
1725 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1726 unsigned long address)
1727 {
1728 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
1729 NULL : p4d_offset(pgd, address);
1730 }
1731
1732 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1733 unsigned long address)
1734 {
1735 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
1736 NULL : pud_offset(p4d, address);
1737 }
1738 #endif /* !__ARCH_HAS_5LEVEL_HACK */
1739
1740 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1741 {
1742 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1743 NULL: pmd_offset(pud, address);
1744 }
1745 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1746
1747 #if USE_SPLIT_PTE_PTLOCKS
1748 #if ALLOC_SPLIT_PTLOCKS
1749 void __init ptlock_cache_init(void);
1750 extern bool ptlock_alloc(struct page *page);
1751 extern void ptlock_free(struct page *page);
1752
1753 static inline spinlock_t *ptlock_ptr(struct page *page)
1754 {
1755 return page->ptl;
1756 }
1757 #else /* ALLOC_SPLIT_PTLOCKS */
1758 static inline void ptlock_cache_init(void)
1759 {
1760 }
1761
1762 static inline bool ptlock_alloc(struct page *page)
1763 {
1764 return true;
1765 }
1766
1767 static inline void ptlock_free(struct page *page)
1768 {
1769 }
1770
1771 static inline spinlock_t *ptlock_ptr(struct page *page)
1772 {
1773 return &page->ptl;
1774 }
1775 #endif /* ALLOC_SPLIT_PTLOCKS */
1776
1777 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1778 {
1779 return ptlock_ptr(pmd_page(*pmd));
1780 }
1781
1782 static inline bool ptlock_init(struct page *page)
1783 {
1784 /*
1785 * prep_new_page() initialize page->private (and therefore page->ptl)
1786 * with 0. Make sure nobody took it in use in between.
1787 *
1788 * It can happen if arch try to use slab for page table allocation:
1789 * slab code uses page->slab_cache, which share storage with page->ptl.
1790 */
1791 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1792 if (!ptlock_alloc(page))
1793 return false;
1794 spin_lock_init(ptlock_ptr(page));
1795 return true;
1796 }
1797
1798 /* Reset page->mapping so free_pages_check won't complain. */
1799 static inline void pte_lock_deinit(struct page *page)
1800 {
1801 page->mapping = NULL;
1802 ptlock_free(page);
1803 }
1804
1805 #else /* !USE_SPLIT_PTE_PTLOCKS */
1806 /*
1807 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1808 */
1809 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1810 {
1811 return &mm->page_table_lock;
1812 }
1813 static inline void ptlock_cache_init(void) {}
1814 static inline bool ptlock_init(struct page *page) { return true; }
1815 static inline void pte_lock_deinit(struct page *page) {}
1816 #endif /* USE_SPLIT_PTE_PTLOCKS */
1817
1818 static inline void pgtable_init(void)
1819 {
1820 ptlock_cache_init();
1821 pgtable_cache_init();
1822 }
1823
1824 static inline bool pgtable_page_ctor(struct page *page)
1825 {
1826 if (!ptlock_init(page))
1827 return false;
1828 inc_zone_page_state(page, NR_PAGETABLE);
1829 return true;
1830 }
1831
1832 static inline void pgtable_page_dtor(struct page *page)
1833 {
1834 pte_lock_deinit(page);
1835 dec_zone_page_state(page, NR_PAGETABLE);
1836 }
1837
1838 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1839 ({ \
1840 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1841 pte_t *__pte = pte_offset_map(pmd, address); \
1842 *(ptlp) = __ptl; \
1843 spin_lock(__ptl); \
1844 __pte; \
1845 })
1846
1847 #define pte_unmap_unlock(pte, ptl) do { \
1848 spin_unlock(ptl); \
1849 pte_unmap(pte); \
1850 } while (0)
1851
1852 #define pte_alloc(mm, pmd, address) \
1853 (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd, address))
1854
1855 #define pte_alloc_map(mm, pmd, address) \
1856 (pte_alloc(mm, pmd, address) ? NULL : pte_offset_map(pmd, address))
1857
1858 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1859 (pte_alloc(mm, pmd, address) ? \
1860 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
1861
1862 #define pte_alloc_kernel(pmd, address) \
1863 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1864 NULL: pte_offset_kernel(pmd, address))
1865
1866 #if USE_SPLIT_PMD_PTLOCKS
1867
1868 static struct page *pmd_to_page(pmd_t *pmd)
1869 {
1870 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1871 return virt_to_page((void *)((unsigned long) pmd & mask));
1872 }
1873
1874 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1875 {
1876 return ptlock_ptr(pmd_to_page(pmd));
1877 }
1878
1879 static inline bool pgtable_pmd_page_ctor(struct page *page)
1880 {
1881 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1882 page->pmd_huge_pte = NULL;
1883 #endif
1884 return ptlock_init(page);
1885 }
1886
1887 static inline void pgtable_pmd_page_dtor(struct page *page)
1888 {
1889 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1890 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1891 #endif
1892 ptlock_free(page);
1893 }
1894
1895 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1896
1897 #else
1898
1899 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1900 {
1901 return &mm->page_table_lock;
1902 }
1903
1904 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1905 static inline void pgtable_pmd_page_dtor(struct page *page) {}
1906
1907 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1908
1909 #endif
1910
1911 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1912 {
1913 spinlock_t *ptl = pmd_lockptr(mm, pmd);
1914 spin_lock(ptl);
1915 return ptl;
1916 }
1917
1918 /*
1919 * No scalability reason to split PUD locks yet, but follow the same pattern
1920 * as the PMD locks to make it easier if we decide to. The VM should not be
1921 * considered ready to switch to split PUD locks yet; there may be places
1922 * which need to be converted from page_table_lock.
1923 */
1924 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
1925 {
1926 return &mm->page_table_lock;
1927 }
1928
1929 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
1930 {
1931 spinlock_t *ptl = pud_lockptr(mm, pud);
1932
1933 spin_lock(ptl);
1934 return ptl;
1935 }
1936
1937 extern void __init pagecache_init(void);
1938 extern void free_area_init(unsigned long * zones_size);
1939 extern void free_area_init_node(int nid, unsigned long * zones_size,
1940 unsigned long zone_start_pfn, unsigned long *zholes_size);
1941 extern void free_initmem(void);
1942
1943 /*
1944 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1945 * into the buddy system. The freed pages will be poisoned with pattern
1946 * "poison" if it's within range [0, UCHAR_MAX].
1947 * Return pages freed into the buddy system.
1948 */
1949 extern unsigned long free_reserved_area(void *start, void *end,
1950 int poison, char *s);
1951
1952 #ifdef CONFIG_HIGHMEM
1953 /*
1954 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1955 * and totalram_pages.
1956 */
1957 extern void free_highmem_page(struct page *page);
1958 #endif
1959
1960 extern void adjust_managed_page_count(struct page *page, long count);
1961 extern void mem_init_print_info(const char *str);
1962
1963 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
1964
1965 /* Free the reserved page into the buddy system, so it gets managed. */
1966 static inline void __free_reserved_page(struct page *page)
1967 {
1968 ClearPageReserved(page);
1969 init_page_count(page);
1970 __free_page(page);
1971 }
1972
1973 static inline void free_reserved_page(struct page *page)
1974 {
1975 __free_reserved_page(page);
1976 adjust_managed_page_count(page, 1);
1977 }
1978
1979 static inline void mark_page_reserved(struct page *page)
1980 {
1981 SetPageReserved(page);
1982 adjust_managed_page_count(page, -1);
1983 }
1984
1985 /*
1986 * Default method to free all the __init memory into the buddy system.
1987 * The freed pages will be poisoned with pattern "poison" if it's within
1988 * range [0, UCHAR_MAX].
1989 * Return pages freed into the buddy system.
1990 */
1991 static inline unsigned long free_initmem_default(int poison)
1992 {
1993 extern char __init_begin[], __init_end[];
1994
1995 return free_reserved_area(&__init_begin, &__init_end,
1996 poison, "unused kernel");
1997 }
1998
1999 static inline unsigned long get_num_physpages(void)
2000 {
2001 int nid;
2002 unsigned long phys_pages = 0;
2003
2004 for_each_online_node(nid)
2005 phys_pages += node_present_pages(nid);
2006
2007 return phys_pages;
2008 }
2009
2010 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
2011 /*
2012 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
2013 * zones, allocate the backing mem_map and account for memory holes in a more
2014 * architecture independent manner. This is a substitute for creating the
2015 * zone_sizes[] and zholes_size[] arrays and passing them to
2016 * free_area_init_node()
2017 *
2018 * An architecture is expected to register range of page frames backed by
2019 * physical memory with memblock_add[_node]() before calling
2020 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
2021 * usage, an architecture is expected to do something like
2022 *
2023 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2024 * max_highmem_pfn};
2025 * for_each_valid_physical_page_range()
2026 * memblock_add_node(base, size, nid)
2027 * free_area_init_nodes(max_zone_pfns);
2028 *
2029 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
2030 * registered physical page range. Similarly
2031 * sparse_memory_present_with_active_regions() calls memory_present() for
2032 * each range when SPARSEMEM is enabled.
2033 *
2034 * See mm/page_alloc.c for more information on each function exposed by
2035 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
2036 */
2037 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
2038 unsigned long node_map_pfn_alignment(void);
2039 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2040 unsigned long end_pfn);
2041 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2042 unsigned long end_pfn);
2043 extern void get_pfn_range_for_nid(unsigned int nid,
2044 unsigned long *start_pfn, unsigned long *end_pfn);
2045 extern unsigned long find_min_pfn_with_active_regions(void);
2046 extern void free_bootmem_with_active_regions(int nid,
2047 unsigned long max_low_pfn);
2048 extern void sparse_memory_present_with_active_regions(int nid);
2049
2050 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
2051
2052 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
2053 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
2054 static inline int __early_pfn_to_nid(unsigned long pfn,
2055 struct mminit_pfnnid_cache *state)
2056 {
2057 return 0;
2058 }
2059 #else
2060 /* please see mm/page_alloc.c */
2061 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2062 /* there is a per-arch backend function. */
2063 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
2064 struct mminit_pfnnid_cache *state);
2065 #endif
2066
2067 extern void set_dma_reserve(unsigned long new_dma_reserve);
2068 extern void memmap_init_zone(unsigned long, int, unsigned long,
2069 unsigned long, enum memmap_context);
2070 extern void setup_per_zone_wmarks(void);
2071 extern int __meminit init_per_zone_wmark_min(void);
2072 extern void mem_init(void);
2073 extern void __init mmap_init(void);
2074 extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2075 extern long si_mem_available(void);
2076 extern void si_meminfo(struct sysinfo * val);
2077 extern void si_meminfo_node(struct sysinfo *val, int nid);
2078 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2079 extern unsigned long arch_reserved_kernel_pages(void);
2080 #endif
2081
2082 extern __printf(3, 4)
2083 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2084
2085 extern void setup_per_cpu_pageset(void);
2086
2087 extern void zone_pcp_update(struct zone *zone);
2088 extern void zone_pcp_reset(struct zone *zone);
2089
2090 /* page_alloc.c */
2091 extern int min_free_kbytes;
2092 extern int watermark_scale_factor;
2093
2094 /* nommu.c */
2095 extern atomic_long_t mmap_pages_allocated;
2096 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2097
2098 /* interval_tree.c */
2099 void vma_interval_tree_insert(struct vm_area_struct *node,
2100 struct rb_root_cached *root);
2101 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2102 struct vm_area_struct *prev,
2103 struct rb_root_cached *root);
2104 void vma_interval_tree_remove(struct vm_area_struct *node,
2105 struct rb_root_cached *root);
2106 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2107 unsigned long start, unsigned long last);
2108 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2109 unsigned long start, unsigned long last);
2110
2111 #define vma_interval_tree_foreach(vma, root, start, last) \
2112 for (vma = vma_interval_tree_iter_first(root, start, last); \
2113 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2114
2115 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2116 struct rb_root_cached *root);
2117 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2118 struct rb_root_cached *root);
2119 struct anon_vma_chain *
2120 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2121 unsigned long start, unsigned long last);
2122 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2123 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2124 #ifdef CONFIG_DEBUG_VM_RB
2125 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2126 #endif
2127
2128 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2129 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2130 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2131
2132 /* mmap.c */
2133 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2134 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2135 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2136 struct vm_area_struct *expand);
2137 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2138 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2139 {
2140 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2141 }
2142 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2143 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2144 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2145 struct mempolicy *, struct vm_userfaultfd_ctx);
2146 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2147 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2148 unsigned long addr, int new_below);
2149 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2150 unsigned long addr, int new_below);
2151 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2152 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2153 struct rb_node **, struct rb_node *);
2154 extern void unlink_file_vma(struct vm_area_struct *);
2155 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2156 unsigned long addr, unsigned long len, pgoff_t pgoff,
2157 bool *need_rmap_locks);
2158 extern void exit_mmap(struct mm_struct *);
2159
2160 static inline int check_data_rlimit(unsigned long rlim,
2161 unsigned long new,
2162 unsigned long start,
2163 unsigned long end_data,
2164 unsigned long start_data)
2165 {
2166 if (rlim < RLIM_INFINITY) {
2167 if (((new - start) + (end_data - start_data)) > rlim)
2168 return -ENOSPC;
2169 }
2170
2171 return 0;
2172 }
2173
2174 extern int mm_take_all_locks(struct mm_struct *mm);
2175 extern void mm_drop_all_locks(struct mm_struct *mm);
2176
2177 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2178 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2179 extern struct file *get_task_exe_file(struct task_struct *task);
2180
2181 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2182 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2183
2184 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2185 const struct vm_special_mapping *sm);
2186 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2187 unsigned long addr, unsigned long len,
2188 unsigned long flags,
2189 const struct vm_special_mapping *spec);
2190 /* This is an obsolete alternative to _install_special_mapping. */
2191 extern int install_special_mapping(struct mm_struct *mm,
2192 unsigned long addr, unsigned long len,
2193 unsigned long flags, struct page **pages);
2194
2195 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2196
2197 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2198 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2199 struct list_head *uf);
2200 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2201 unsigned long len, unsigned long prot, unsigned long flags,
2202 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate,
2203 struct list_head *uf);
2204 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2205 struct list_head *uf);
2206
2207 static inline unsigned long
2208 do_mmap_pgoff(struct file *file, unsigned long addr,
2209 unsigned long len, unsigned long prot, unsigned long flags,
2210 unsigned long pgoff, unsigned long *populate,
2211 struct list_head *uf)
2212 {
2213 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate, uf);
2214 }
2215
2216 #ifdef CONFIG_MMU
2217 extern int __mm_populate(unsigned long addr, unsigned long len,
2218 int ignore_errors);
2219 static inline void mm_populate(unsigned long addr, unsigned long len)
2220 {
2221 /* Ignore errors */
2222 (void) __mm_populate(addr, len, 1);
2223 }
2224 #else
2225 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2226 #endif
2227
2228 /* These take the mm semaphore themselves */
2229 extern int __must_check vm_brk(unsigned long, unsigned long);
2230 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2231 extern int vm_munmap(unsigned long, size_t);
2232 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2233 unsigned long, unsigned long,
2234 unsigned long, unsigned long);
2235
2236 struct vm_unmapped_area_info {
2237 #define VM_UNMAPPED_AREA_TOPDOWN 1
2238 unsigned long flags;
2239 unsigned long length;
2240 unsigned long low_limit;
2241 unsigned long high_limit;
2242 unsigned long align_mask;
2243 unsigned long align_offset;
2244 };
2245
2246 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2247 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2248
2249 /*
2250 * Search for an unmapped address range.
2251 *
2252 * We are looking for a range that:
2253 * - does not intersect with any VMA;
2254 * - is contained within the [low_limit, high_limit) interval;
2255 * - is at least the desired size.
2256 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2257 */
2258 static inline unsigned long
2259 vm_unmapped_area(struct vm_unmapped_area_info *info)
2260 {
2261 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2262 return unmapped_area_topdown(info);
2263 else
2264 return unmapped_area(info);
2265 }
2266
2267 /* truncate.c */
2268 extern void truncate_inode_pages(struct address_space *, loff_t);
2269 extern void truncate_inode_pages_range(struct address_space *,
2270 loff_t lstart, loff_t lend);
2271 extern void truncate_inode_pages_final(struct address_space *);
2272
2273 /* generic vm_area_ops exported for stackable file systems */
2274 extern int filemap_fault(struct vm_fault *vmf);
2275 extern void filemap_map_pages(struct vm_fault *vmf,
2276 pgoff_t start_pgoff, pgoff_t end_pgoff);
2277 extern int filemap_page_mkwrite(struct vm_fault *vmf);
2278
2279 /* mm/page-writeback.c */
2280 int __must_check write_one_page(struct page *page);
2281 void task_dirty_inc(struct task_struct *tsk);
2282
2283 /* readahead.c */
2284 #define VM_MAX_READAHEAD 128 /* kbytes */
2285 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
2286
2287 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2288 pgoff_t offset, unsigned long nr_to_read);
2289
2290 void page_cache_sync_readahead(struct address_space *mapping,
2291 struct file_ra_state *ra,
2292 struct file *filp,
2293 pgoff_t offset,
2294 unsigned long size);
2295
2296 void page_cache_async_readahead(struct address_space *mapping,
2297 struct file_ra_state *ra,
2298 struct file *filp,
2299 struct page *pg,
2300 pgoff_t offset,
2301 unsigned long size);
2302
2303 extern unsigned long stack_guard_gap;
2304 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2305 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2306
2307 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2308 extern int expand_downwards(struct vm_area_struct *vma,
2309 unsigned long address);
2310 #if VM_GROWSUP
2311 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2312 #else
2313 #define expand_upwards(vma, address) (0)
2314 #endif
2315
2316 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2317 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2318 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2319 struct vm_area_struct **pprev);
2320
2321 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2322 NULL if none. Assume start_addr < end_addr. */
2323 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2324 {
2325 struct vm_area_struct * vma = find_vma(mm,start_addr);
2326
2327 if (vma && end_addr <= vma->vm_start)
2328 vma = NULL;
2329 return vma;
2330 }
2331
2332 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2333 {
2334 unsigned long vm_start = vma->vm_start;
2335
2336 if (vma->vm_flags & VM_GROWSDOWN) {
2337 vm_start -= stack_guard_gap;
2338 if (vm_start > vma->vm_start)
2339 vm_start = 0;
2340 }
2341 return vm_start;
2342 }
2343
2344 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2345 {
2346 unsigned long vm_end = vma->vm_end;
2347
2348 if (vma->vm_flags & VM_GROWSUP) {
2349 vm_end += stack_guard_gap;
2350 if (vm_end < vma->vm_end)
2351 vm_end = -PAGE_SIZE;
2352 }
2353 return vm_end;
2354 }
2355
2356 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2357 {
2358 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2359 }
2360
2361 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2362 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2363 unsigned long vm_start, unsigned long vm_end)
2364 {
2365 struct vm_area_struct *vma = find_vma(mm, vm_start);
2366
2367 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2368 vma = NULL;
2369
2370 return vma;
2371 }
2372
2373 #ifdef CONFIG_MMU
2374 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2375 void vma_set_page_prot(struct vm_area_struct *vma);
2376 #else
2377 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2378 {
2379 return __pgprot(0);
2380 }
2381 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2382 {
2383 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2384 }
2385 #endif
2386
2387 #ifdef CONFIG_NUMA_BALANCING
2388 unsigned long change_prot_numa(struct vm_area_struct *vma,
2389 unsigned long start, unsigned long end);
2390 #endif
2391
2392 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2393 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2394 unsigned long pfn, unsigned long size, pgprot_t);
2395 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2396 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2397 unsigned long pfn);
2398 int vm_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2399 unsigned long pfn, pgprot_t pgprot);
2400 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2401 pfn_t pfn);
2402 int vm_insert_mixed_mkwrite(struct vm_area_struct *vma, unsigned long addr,
2403 pfn_t pfn);
2404 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2405
2406
2407 struct page *follow_page_mask(struct vm_area_struct *vma,
2408 unsigned long address, unsigned int foll_flags,
2409 unsigned int *page_mask);
2410
2411 static inline struct page *follow_page(struct vm_area_struct *vma,
2412 unsigned long address, unsigned int foll_flags)
2413 {
2414 unsigned int unused_page_mask;
2415 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2416 }
2417
2418 #define FOLL_WRITE 0x01 /* check pte is writable */
2419 #define FOLL_TOUCH 0x02 /* mark page accessed */
2420 #define FOLL_GET 0x04 /* do get_page on page */
2421 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2422 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2423 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2424 * and return without waiting upon it */
2425 #define FOLL_POPULATE 0x40 /* fault in page */
2426 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2427 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2428 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2429 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2430 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2431 #define FOLL_MLOCK 0x1000 /* lock present pages */
2432 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2433 #define FOLL_COW 0x4000 /* internal GUP flag */
2434
2435 static inline int vm_fault_to_errno(int vm_fault, int foll_flags)
2436 {
2437 if (vm_fault & VM_FAULT_OOM)
2438 return -ENOMEM;
2439 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2440 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2441 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2442 return -EFAULT;
2443 return 0;
2444 }
2445
2446 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2447 void *data);
2448 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2449 unsigned long size, pte_fn_t fn, void *data);
2450
2451
2452 #ifdef CONFIG_PAGE_POISONING
2453 extern bool page_poisoning_enabled(void);
2454 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2455 extern bool page_is_poisoned(struct page *page);
2456 #else
2457 static inline bool page_poisoning_enabled(void) { return false; }
2458 static inline void kernel_poison_pages(struct page *page, int numpages,
2459 int enable) { }
2460 static inline bool page_is_poisoned(struct page *page) { return false; }
2461 #endif
2462
2463 #ifdef CONFIG_DEBUG_PAGEALLOC
2464 extern bool _debug_pagealloc_enabled;
2465 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2466
2467 static inline bool debug_pagealloc_enabled(void)
2468 {
2469 return _debug_pagealloc_enabled;
2470 }
2471
2472 static inline void
2473 kernel_map_pages(struct page *page, int numpages, int enable)
2474 {
2475 if (!debug_pagealloc_enabled())
2476 return;
2477
2478 __kernel_map_pages(page, numpages, enable);
2479 }
2480 #ifdef CONFIG_HIBERNATION
2481 extern bool kernel_page_present(struct page *page);
2482 #endif /* CONFIG_HIBERNATION */
2483 #else /* CONFIG_DEBUG_PAGEALLOC */
2484 static inline void
2485 kernel_map_pages(struct page *page, int numpages, int enable) {}
2486 #ifdef CONFIG_HIBERNATION
2487 static inline bool kernel_page_present(struct page *page) { return true; }
2488 #endif /* CONFIG_HIBERNATION */
2489 static inline bool debug_pagealloc_enabled(void)
2490 {
2491 return false;
2492 }
2493 #endif /* CONFIG_DEBUG_PAGEALLOC */
2494
2495 #ifdef __HAVE_ARCH_GATE_AREA
2496 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2497 extern int in_gate_area_no_mm(unsigned long addr);
2498 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2499 #else
2500 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2501 {
2502 return NULL;
2503 }
2504 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2505 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2506 {
2507 return 0;
2508 }
2509 #endif /* __HAVE_ARCH_GATE_AREA */
2510
2511 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2512
2513 #ifdef CONFIG_SYSCTL
2514 extern int sysctl_drop_caches;
2515 int drop_caches_sysctl_handler(struct ctl_table *, int,
2516 void __user *, size_t *, loff_t *);
2517 #endif
2518
2519 void drop_slab(void);
2520 void drop_slab_node(int nid);
2521
2522 #ifndef CONFIG_MMU
2523 #define randomize_va_space 0
2524 #else
2525 extern int randomize_va_space;
2526 #endif
2527
2528 const char * arch_vma_name(struct vm_area_struct *vma);
2529 void print_vma_addr(char *prefix, unsigned long rip);
2530
2531 void sparse_mem_maps_populate_node(struct page **map_map,
2532 unsigned long pnum_begin,
2533 unsigned long pnum_end,
2534 unsigned long map_count,
2535 int nodeid);
2536
2537 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
2538 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2539 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
2540 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
2541 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2542 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2543 void *vmemmap_alloc_block(unsigned long size, int node);
2544 struct vmem_altmap;
2545 void *__vmemmap_alloc_block_buf(unsigned long size, int node,
2546 struct vmem_altmap *altmap);
2547 static inline void *vmemmap_alloc_block_buf(unsigned long size, int node)
2548 {
2549 return __vmemmap_alloc_block_buf(size, node, NULL);
2550 }
2551
2552 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2553 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2554 int node);
2555 int vmemmap_populate(unsigned long start, unsigned long end, int node);
2556 void vmemmap_populate_print_last(void);
2557 #ifdef CONFIG_MEMORY_HOTPLUG
2558 void vmemmap_free(unsigned long start, unsigned long end);
2559 #endif
2560 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2561 unsigned long nr_pages);
2562
2563 enum mf_flags {
2564 MF_COUNT_INCREASED = 1 << 0,
2565 MF_ACTION_REQUIRED = 1 << 1,
2566 MF_MUST_KILL = 1 << 2,
2567 MF_SOFT_OFFLINE = 1 << 3,
2568 };
2569 extern int memory_failure(unsigned long pfn, int trapno, int flags);
2570 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
2571 extern int unpoison_memory(unsigned long pfn);
2572 extern int get_hwpoison_page(struct page *page);
2573 #define put_hwpoison_page(page) put_page(page)
2574 extern int sysctl_memory_failure_early_kill;
2575 extern int sysctl_memory_failure_recovery;
2576 extern void shake_page(struct page *p, int access);
2577 extern atomic_long_t num_poisoned_pages;
2578 extern int soft_offline_page(struct page *page, int flags);
2579
2580
2581 /*
2582 * Error handlers for various types of pages.
2583 */
2584 enum mf_result {
2585 MF_IGNORED, /* Error: cannot be handled */
2586 MF_FAILED, /* Error: handling failed */
2587 MF_DELAYED, /* Will be handled later */
2588 MF_RECOVERED, /* Successfully recovered */
2589 };
2590
2591 enum mf_action_page_type {
2592 MF_MSG_KERNEL,
2593 MF_MSG_KERNEL_HIGH_ORDER,
2594 MF_MSG_SLAB,
2595 MF_MSG_DIFFERENT_COMPOUND,
2596 MF_MSG_POISONED_HUGE,
2597 MF_MSG_HUGE,
2598 MF_MSG_FREE_HUGE,
2599 MF_MSG_UNMAP_FAILED,
2600 MF_MSG_DIRTY_SWAPCACHE,
2601 MF_MSG_CLEAN_SWAPCACHE,
2602 MF_MSG_DIRTY_MLOCKED_LRU,
2603 MF_MSG_CLEAN_MLOCKED_LRU,
2604 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2605 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2606 MF_MSG_DIRTY_LRU,
2607 MF_MSG_CLEAN_LRU,
2608 MF_MSG_TRUNCATED_LRU,
2609 MF_MSG_BUDDY,
2610 MF_MSG_BUDDY_2ND,
2611 MF_MSG_UNKNOWN,
2612 };
2613
2614 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2615 extern void clear_huge_page(struct page *page,
2616 unsigned long addr_hint,
2617 unsigned int pages_per_huge_page);
2618 extern void copy_user_huge_page(struct page *dst, struct page *src,
2619 unsigned long addr, struct vm_area_struct *vma,
2620 unsigned int pages_per_huge_page);
2621 extern long copy_huge_page_from_user(struct page *dst_page,
2622 const void __user *usr_src,
2623 unsigned int pages_per_huge_page,
2624 bool allow_pagefault);
2625 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2626
2627 extern struct page_ext_operations debug_guardpage_ops;
2628
2629 #ifdef CONFIG_DEBUG_PAGEALLOC
2630 extern unsigned int _debug_guardpage_minorder;
2631 extern bool _debug_guardpage_enabled;
2632
2633 static inline unsigned int debug_guardpage_minorder(void)
2634 {
2635 return _debug_guardpage_minorder;
2636 }
2637
2638 static inline bool debug_guardpage_enabled(void)
2639 {
2640 return _debug_guardpage_enabled;
2641 }
2642
2643 static inline bool page_is_guard(struct page *page)
2644 {
2645 struct page_ext *page_ext;
2646
2647 if (!debug_guardpage_enabled())
2648 return false;
2649
2650 page_ext = lookup_page_ext(page);
2651 if (unlikely(!page_ext))
2652 return false;
2653
2654 return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2655 }
2656 #else
2657 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2658 static inline bool debug_guardpage_enabled(void) { return false; }
2659 static inline bool page_is_guard(struct page *page) { return false; }
2660 #endif /* CONFIG_DEBUG_PAGEALLOC */
2661
2662 #if MAX_NUMNODES > 1
2663 void __init setup_nr_node_ids(void);
2664 #else
2665 static inline void setup_nr_node_ids(void) {}
2666 #endif
2667
2668 #endif /* __KERNEL__ */
2669 #endif /* _LINUX_MM_H */