]> git.ipfire.org Git - thirdparty/kernel/stable.git/blob - include/linux/mm.h
541d99b86aea9efe8df91ed29eae76e9e6e23918
[thirdparty/kernel/stable.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 #include <linux/overflow.h>
29
30 struct mempolicy;
31 struct anon_vma;
32 struct anon_vma_chain;
33 struct file_ra_state;
34 struct user_struct;
35 struct writeback_control;
36 struct bdi_writeback;
37
38 void init_mm_internals(void);
39
40 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
41 extern unsigned long max_mapnr;
42
43 static inline void set_max_mapnr(unsigned long limit)
44 {
45 max_mapnr = limit;
46 }
47 #else
48 static inline void set_max_mapnr(unsigned long limit) { }
49 #endif
50
51 extern atomic_long_t _totalram_pages;
52 static inline unsigned long totalram_pages(void)
53 {
54 return (unsigned long)atomic_long_read(&_totalram_pages);
55 }
56
57 static inline void totalram_pages_inc(void)
58 {
59 atomic_long_inc(&_totalram_pages);
60 }
61
62 static inline void totalram_pages_dec(void)
63 {
64 atomic_long_dec(&_totalram_pages);
65 }
66
67 static inline void totalram_pages_add(long count)
68 {
69 atomic_long_add(count, &_totalram_pages);
70 }
71
72 static inline void totalram_pages_set(long val)
73 {
74 atomic_long_set(&_totalram_pages, val);
75 }
76
77 extern void * high_memory;
78 extern int page_cluster;
79
80 #ifdef CONFIG_SYSCTL
81 extern int sysctl_legacy_va_layout;
82 #else
83 #define sysctl_legacy_va_layout 0
84 #endif
85
86 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
87 extern const int mmap_rnd_bits_min;
88 extern const int mmap_rnd_bits_max;
89 extern int mmap_rnd_bits __read_mostly;
90 #endif
91 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
92 extern const int mmap_rnd_compat_bits_min;
93 extern const int mmap_rnd_compat_bits_max;
94 extern int mmap_rnd_compat_bits __read_mostly;
95 #endif
96
97 #include <asm/page.h>
98 #include <asm/pgtable.h>
99 #include <asm/processor.h>
100
101 #ifndef __pa_symbol
102 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
103 #endif
104
105 #ifndef page_to_virt
106 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
107 #endif
108
109 #ifndef lm_alias
110 #define lm_alias(x) __va(__pa_symbol(x))
111 #endif
112
113 /*
114 * To prevent common memory management code establishing
115 * a zero page mapping on a read fault.
116 * This macro should be defined within <asm/pgtable.h>.
117 * s390 does this to prevent multiplexing of hardware bits
118 * related to the physical page in case of virtualization.
119 */
120 #ifndef mm_forbids_zeropage
121 #define mm_forbids_zeropage(X) (0)
122 #endif
123
124 /*
125 * On some architectures it is expensive to call memset() for small sizes.
126 * Those architectures should provide their own implementation of "struct page"
127 * zeroing by defining this macro in <asm/pgtable.h>.
128 */
129 #ifndef mm_zero_struct_page
130 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
131 #endif
132
133 /*
134 * Default maximum number of active map areas, this limits the number of vmas
135 * per mm struct. Users can overwrite this number by sysctl but there is a
136 * problem.
137 *
138 * When a program's coredump is generated as ELF format, a section is created
139 * per a vma. In ELF, the number of sections is represented in unsigned short.
140 * This means the number of sections should be smaller than 65535 at coredump.
141 * Because the kernel adds some informative sections to a image of program at
142 * generating coredump, we need some margin. The number of extra sections is
143 * 1-3 now and depends on arch. We use "5" as safe margin, here.
144 *
145 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
146 * not a hard limit any more. Although some userspace tools can be surprised by
147 * that.
148 */
149 #define MAPCOUNT_ELF_CORE_MARGIN (5)
150 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
151
152 extern int sysctl_max_map_count;
153
154 extern unsigned long sysctl_user_reserve_kbytes;
155 extern unsigned long sysctl_admin_reserve_kbytes;
156
157 extern int sysctl_overcommit_memory;
158 extern int sysctl_overcommit_ratio;
159 extern unsigned long sysctl_overcommit_kbytes;
160
161 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
162 size_t *, loff_t *);
163 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
164 size_t *, loff_t *);
165
166 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
167
168 /* to align the pointer to the (next) page boundary */
169 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
170
171 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
172 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
173
174 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
175
176 /*
177 * Linux kernel virtual memory manager primitives.
178 * The idea being to have a "virtual" mm in the same way
179 * we have a virtual fs - giving a cleaner interface to the
180 * mm details, and allowing different kinds of memory mappings
181 * (from shared memory to executable loading to arbitrary
182 * mmap() functions).
183 */
184
185 struct vm_area_struct *vm_area_alloc(struct mm_struct *);
186 struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
187 void vm_area_free(struct vm_area_struct *);
188
189 #ifndef CONFIG_MMU
190 extern struct rb_root nommu_region_tree;
191 extern struct rw_semaphore nommu_region_sem;
192
193 extern unsigned int kobjsize(const void *objp);
194 #endif
195
196 /*
197 * vm_flags in vm_area_struct, see mm_types.h.
198 * When changing, update also include/trace/events/mmflags.h
199 */
200 #define VM_NONE 0x00000000
201
202 #define VM_READ 0x00000001 /* currently active flags */
203 #define VM_WRITE 0x00000002
204 #define VM_EXEC 0x00000004
205 #define VM_SHARED 0x00000008
206
207 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
208 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
209 #define VM_MAYWRITE 0x00000020
210 #define VM_MAYEXEC 0x00000040
211 #define VM_MAYSHARE 0x00000080
212
213 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
214 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
215 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
216 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
217 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
218
219 #define VM_LOCKED 0x00002000
220 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
221
222 /* Used by sys_madvise() */
223 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
224 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
225
226 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
227 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
228 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
229 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
230 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
231 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
232 #define VM_SYNC 0x00800000 /* Synchronous page faults */
233 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
234 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
235 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
236
237 #ifdef CONFIG_MEM_SOFT_DIRTY
238 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
239 #else
240 # define VM_SOFTDIRTY 0
241 #endif
242
243 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
244 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
245 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
246 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
247
248 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
249 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
250 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
251 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
252 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
253 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
254 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
255 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
256 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
257 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
258 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
259 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
260
261 #ifdef CONFIG_ARCH_HAS_PKEYS
262 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
263 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
264 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
265 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
266 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
267 #ifdef CONFIG_PPC
268 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
269 #else
270 # define VM_PKEY_BIT4 0
271 #endif
272 #endif /* CONFIG_ARCH_HAS_PKEYS */
273
274 #if defined(CONFIG_X86)
275 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
276 #elif defined(CONFIG_PPC)
277 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
278 #elif defined(CONFIG_PARISC)
279 # define VM_GROWSUP VM_ARCH_1
280 #elif defined(CONFIG_IA64)
281 # define VM_GROWSUP VM_ARCH_1
282 #elif defined(CONFIG_SPARC64)
283 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
284 # define VM_ARCH_CLEAR VM_SPARC_ADI
285 #elif !defined(CONFIG_MMU)
286 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
287 #endif
288
289 #if defined(CONFIG_X86_INTEL_MPX)
290 /* MPX specific bounds table or bounds directory */
291 # define VM_MPX VM_HIGH_ARCH_4
292 #else
293 # define VM_MPX VM_NONE
294 #endif
295
296 #ifndef VM_GROWSUP
297 # define VM_GROWSUP VM_NONE
298 #endif
299
300 /* Bits set in the VMA until the stack is in its final location */
301 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
302
303 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
304 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
305 #endif
306
307 #ifdef CONFIG_STACK_GROWSUP
308 #define VM_STACK VM_GROWSUP
309 #else
310 #define VM_STACK VM_GROWSDOWN
311 #endif
312
313 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
314
315 /*
316 * Special vmas that are non-mergable, non-mlock()able.
317 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
318 */
319 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
320
321 /* This mask defines which mm->def_flags a process can inherit its parent */
322 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
323
324 /* This mask is used to clear all the VMA flags used by mlock */
325 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
326
327 /* Arch-specific flags to clear when updating VM flags on protection change */
328 #ifndef VM_ARCH_CLEAR
329 # define VM_ARCH_CLEAR VM_NONE
330 #endif
331 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
332
333 /*
334 * mapping from the currently active vm_flags protection bits (the
335 * low four bits) to a page protection mask..
336 */
337 extern pgprot_t protection_map[16];
338
339 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
340 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
341 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
342 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
343 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
344 #define FAULT_FLAG_TRIED 0x20 /* Second try */
345 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
346 #define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */
347 #define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */
348
349 #define FAULT_FLAG_TRACE \
350 { FAULT_FLAG_WRITE, "WRITE" }, \
351 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
352 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
353 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
354 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
355 { FAULT_FLAG_TRIED, "TRIED" }, \
356 { FAULT_FLAG_USER, "USER" }, \
357 { FAULT_FLAG_REMOTE, "REMOTE" }, \
358 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }
359
360 /*
361 * vm_fault is filled by the the pagefault handler and passed to the vma's
362 * ->fault function. The vma's ->fault is responsible for returning a bitmask
363 * of VM_FAULT_xxx flags that give details about how the fault was handled.
364 *
365 * MM layer fills up gfp_mask for page allocations but fault handler might
366 * alter it if its implementation requires a different allocation context.
367 *
368 * pgoff should be used in favour of virtual_address, if possible.
369 */
370 struct vm_fault {
371 struct vm_area_struct *vma; /* Target VMA */
372 unsigned int flags; /* FAULT_FLAG_xxx flags */
373 gfp_t gfp_mask; /* gfp mask to be used for allocations */
374 pgoff_t pgoff; /* Logical page offset based on vma */
375 unsigned long address; /* Faulting virtual address */
376 pmd_t *pmd; /* Pointer to pmd entry matching
377 * the 'address' */
378 pud_t *pud; /* Pointer to pud entry matching
379 * the 'address'
380 */
381 pte_t orig_pte; /* Value of PTE at the time of fault */
382
383 struct page *cow_page; /* Page handler may use for COW fault */
384 struct mem_cgroup *memcg; /* Cgroup cow_page belongs to */
385 struct page *page; /* ->fault handlers should return a
386 * page here, unless VM_FAULT_NOPAGE
387 * is set (which is also implied by
388 * VM_FAULT_ERROR).
389 */
390 /* These three entries are valid only while holding ptl lock */
391 pte_t *pte; /* Pointer to pte entry matching
392 * the 'address'. NULL if the page
393 * table hasn't been allocated.
394 */
395 spinlock_t *ptl; /* Page table lock.
396 * Protects pte page table if 'pte'
397 * is not NULL, otherwise pmd.
398 */
399 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
400 * vm_ops->map_pages() calls
401 * alloc_set_pte() from atomic context.
402 * do_fault_around() pre-allocates
403 * page table to avoid allocation from
404 * atomic context.
405 */
406 };
407
408 /* page entry size for vm->huge_fault() */
409 enum page_entry_size {
410 PE_SIZE_PTE = 0,
411 PE_SIZE_PMD,
412 PE_SIZE_PUD,
413 };
414
415 /*
416 * These are the virtual MM functions - opening of an area, closing and
417 * unmapping it (needed to keep files on disk up-to-date etc), pointer
418 * to the functions called when a no-page or a wp-page exception occurs.
419 */
420 struct vm_operations_struct {
421 void (*open)(struct vm_area_struct * area);
422 void (*close)(struct vm_area_struct * area);
423 int (*split)(struct vm_area_struct * area, unsigned long addr);
424 int (*mremap)(struct vm_area_struct * area);
425 vm_fault_t (*fault)(struct vm_fault *vmf);
426 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
427 enum page_entry_size pe_size);
428 void (*map_pages)(struct vm_fault *vmf,
429 pgoff_t start_pgoff, pgoff_t end_pgoff);
430 unsigned long (*pagesize)(struct vm_area_struct * area);
431
432 /* notification that a previously read-only page is about to become
433 * writable, if an error is returned it will cause a SIGBUS */
434 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
435
436 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
437 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
438
439 /* called by access_process_vm when get_user_pages() fails, typically
440 * for use by special VMAs that can switch between memory and hardware
441 */
442 int (*access)(struct vm_area_struct *vma, unsigned long addr,
443 void *buf, int len, int write);
444
445 /* Called by the /proc/PID/maps code to ask the vma whether it
446 * has a special name. Returning non-NULL will also cause this
447 * vma to be dumped unconditionally. */
448 const char *(*name)(struct vm_area_struct *vma);
449
450 #ifdef CONFIG_NUMA
451 /*
452 * set_policy() op must add a reference to any non-NULL @new mempolicy
453 * to hold the policy upon return. Caller should pass NULL @new to
454 * remove a policy and fall back to surrounding context--i.e. do not
455 * install a MPOL_DEFAULT policy, nor the task or system default
456 * mempolicy.
457 */
458 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
459
460 /*
461 * get_policy() op must add reference [mpol_get()] to any policy at
462 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
463 * in mm/mempolicy.c will do this automatically.
464 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
465 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
466 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
467 * must return NULL--i.e., do not "fallback" to task or system default
468 * policy.
469 */
470 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
471 unsigned long addr);
472 #endif
473 /*
474 * Called by vm_normal_page() for special PTEs to find the
475 * page for @addr. This is useful if the default behavior
476 * (using pte_page()) would not find the correct page.
477 */
478 struct page *(*find_special_page)(struct vm_area_struct *vma,
479 unsigned long addr);
480 };
481
482 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
483 {
484 static const struct vm_operations_struct dummy_vm_ops = {};
485
486 memset(vma, 0, sizeof(*vma));
487 vma->vm_mm = mm;
488 vma->vm_ops = &dummy_vm_ops;
489 INIT_LIST_HEAD(&vma->anon_vma_chain);
490 }
491
492 static inline void vma_set_anonymous(struct vm_area_struct *vma)
493 {
494 vma->vm_ops = NULL;
495 }
496
497 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
498 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
499
500 struct mmu_gather;
501 struct inode;
502
503 #define page_private(page) ((page)->private)
504 #define set_page_private(page, v) ((page)->private = (v))
505
506 #if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
507 static inline int pmd_devmap(pmd_t pmd)
508 {
509 return 0;
510 }
511 static inline int pud_devmap(pud_t pud)
512 {
513 return 0;
514 }
515 static inline int pgd_devmap(pgd_t pgd)
516 {
517 return 0;
518 }
519 #endif
520
521 /*
522 * FIXME: take this include out, include page-flags.h in
523 * files which need it (119 of them)
524 */
525 #include <linux/page-flags.h>
526 #include <linux/huge_mm.h>
527
528 /*
529 * Methods to modify the page usage count.
530 *
531 * What counts for a page usage:
532 * - cache mapping (page->mapping)
533 * - private data (page->private)
534 * - page mapped in a task's page tables, each mapping
535 * is counted separately
536 *
537 * Also, many kernel routines increase the page count before a critical
538 * routine so they can be sure the page doesn't go away from under them.
539 */
540
541 /*
542 * Drop a ref, return true if the refcount fell to zero (the page has no users)
543 */
544 static inline int put_page_testzero(struct page *page)
545 {
546 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
547 return page_ref_dec_and_test(page);
548 }
549
550 /*
551 * Try to grab a ref unless the page has a refcount of zero, return false if
552 * that is the case.
553 * This can be called when MMU is off so it must not access
554 * any of the virtual mappings.
555 */
556 static inline int get_page_unless_zero(struct page *page)
557 {
558 return page_ref_add_unless(page, 1, 0);
559 }
560
561 extern int page_is_ram(unsigned long pfn);
562
563 enum {
564 REGION_INTERSECTS,
565 REGION_DISJOINT,
566 REGION_MIXED,
567 };
568
569 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
570 unsigned long desc);
571
572 /* Support for virtually mapped pages */
573 struct page *vmalloc_to_page(const void *addr);
574 unsigned long vmalloc_to_pfn(const void *addr);
575
576 /*
577 * Determine if an address is within the vmalloc range
578 *
579 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
580 * is no special casing required.
581 */
582 static inline bool is_vmalloc_addr(const void *x)
583 {
584 #ifdef CONFIG_MMU
585 unsigned long addr = (unsigned long)x;
586
587 return addr >= VMALLOC_START && addr < VMALLOC_END;
588 #else
589 return false;
590 #endif
591 }
592 #ifdef CONFIG_MMU
593 extern int is_vmalloc_or_module_addr(const void *x);
594 #else
595 static inline int is_vmalloc_or_module_addr(const void *x)
596 {
597 return 0;
598 }
599 #endif
600
601 extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
602 static inline void *kvmalloc(size_t size, gfp_t flags)
603 {
604 return kvmalloc_node(size, flags, NUMA_NO_NODE);
605 }
606 static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
607 {
608 return kvmalloc_node(size, flags | __GFP_ZERO, node);
609 }
610 static inline void *kvzalloc(size_t size, gfp_t flags)
611 {
612 return kvmalloc(size, flags | __GFP_ZERO);
613 }
614
615 static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
616 {
617 size_t bytes;
618
619 if (unlikely(check_mul_overflow(n, size, &bytes)))
620 return NULL;
621
622 return kvmalloc(bytes, flags);
623 }
624
625 static inline void *kvcalloc(size_t n, size_t size, gfp_t flags)
626 {
627 return kvmalloc_array(n, size, flags | __GFP_ZERO);
628 }
629
630 extern void kvfree(const void *addr);
631
632 static inline atomic_t *compound_mapcount_ptr(struct page *page)
633 {
634 return &page[1].compound_mapcount;
635 }
636
637 static inline int compound_mapcount(struct page *page)
638 {
639 VM_BUG_ON_PAGE(!PageCompound(page), page);
640 page = compound_head(page);
641 return atomic_read(compound_mapcount_ptr(page)) + 1;
642 }
643
644 /*
645 * The atomic page->_mapcount, starts from -1: so that transitions
646 * both from it and to it can be tracked, using atomic_inc_and_test
647 * and atomic_add_negative(-1).
648 */
649 static inline void page_mapcount_reset(struct page *page)
650 {
651 atomic_set(&(page)->_mapcount, -1);
652 }
653
654 int __page_mapcount(struct page *page);
655
656 static inline int page_mapcount(struct page *page)
657 {
658 VM_BUG_ON_PAGE(PageSlab(page), page);
659
660 if (unlikely(PageCompound(page)))
661 return __page_mapcount(page);
662 return atomic_read(&page->_mapcount) + 1;
663 }
664
665 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
666 int total_mapcount(struct page *page);
667 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
668 #else
669 static inline int total_mapcount(struct page *page)
670 {
671 return page_mapcount(page);
672 }
673 static inline int page_trans_huge_mapcount(struct page *page,
674 int *total_mapcount)
675 {
676 int mapcount = page_mapcount(page);
677 if (total_mapcount)
678 *total_mapcount = mapcount;
679 return mapcount;
680 }
681 #endif
682
683 static inline struct page *virt_to_head_page(const void *x)
684 {
685 struct page *page = virt_to_page(x);
686
687 return compound_head(page);
688 }
689
690 void __put_page(struct page *page);
691
692 void put_pages_list(struct list_head *pages);
693
694 void split_page(struct page *page, unsigned int order);
695
696 /*
697 * Compound pages have a destructor function. Provide a
698 * prototype for that function and accessor functions.
699 * These are _only_ valid on the head of a compound page.
700 */
701 typedef void compound_page_dtor(struct page *);
702
703 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
704 enum compound_dtor_id {
705 NULL_COMPOUND_DTOR,
706 COMPOUND_PAGE_DTOR,
707 #ifdef CONFIG_HUGETLB_PAGE
708 HUGETLB_PAGE_DTOR,
709 #endif
710 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
711 TRANSHUGE_PAGE_DTOR,
712 #endif
713 NR_COMPOUND_DTORS,
714 };
715 extern compound_page_dtor * const compound_page_dtors[];
716
717 static inline void set_compound_page_dtor(struct page *page,
718 enum compound_dtor_id compound_dtor)
719 {
720 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
721 page[1].compound_dtor = compound_dtor;
722 }
723
724 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
725 {
726 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
727 return compound_page_dtors[page[1].compound_dtor];
728 }
729
730 static inline unsigned int compound_order(struct page *page)
731 {
732 if (!PageHead(page))
733 return 0;
734 return page[1].compound_order;
735 }
736
737 static inline void set_compound_order(struct page *page, unsigned int order)
738 {
739 page[1].compound_order = order;
740 }
741
742 void free_compound_page(struct page *page);
743
744 #ifdef CONFIG_MMU
745 /*
746 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
747 * servicing faults for write access. In the normal case, do always want
748 * pte_mkwrite. But get_user_pages can cause write faults for mappings
749 * that do not have writing enabled, when used by access_process_vm.
750 */
751 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
752 {
753 if (likely(vma->vm_flags & VM_WRITE))
754 pte = pte_mkwrite(pte);
755 return pte;
756 }
757
758 vm_fault_t alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
759 struct page *page);
760 vm_fault_t finish_fault(struct vm_fault *vmf);
761 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
762 #endif
763
764 /*
765 * Multiple processes may "see" the same page. E.g. for untouched
766 * mappings of /dev/null, all processes see the same page full of
767 * zeroes, and text pages of executables and shared libraries have
768 * only one copy in memory, at most, normally.
769 *
770 * For the non-reserved pages, page_count(page) denotes a reference count.
771 * page_count() == 0 means the page is free. page->lru is then used for
772 * freelist management in the buddy allocator.
773 * page_count() > 0 means the page has been allocated.
774 *
775 * Pages are allocated by the slab allocator in order to provide memory
776 * to kmalloc and kmem_cache_alloc. In this case, the management of the
777 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
778 * unless a particular usage is carefully commented. (the responsibility of
779 * freeing the kmalloc memory is the caller's, of course).
780 *
781 * A page may be used by anyone else who does a __get_free_page().
782 * In this case, page_count still tracks the references, and should only
783 * be used through the normal accessor functions. The top bits of page->flags
784 * and page->virtual store page management information, but all other fields
785 * are unused and could be used privately, carefully. The management of this
786 * page is the responsibility of the one who allocated it, and those who have
787 * subsequently been given references to it.
788 *
789 * The other pages (we may call them "pagecache pages") are completely
790 * managed by the Linux memory manager: I/O, buffers, swapping etc.
791 * The following discussion applies only to them.
792 *
793 * A pagecache page contains an opaque `private' member, which belongs to the
794 * page's address_space. Usually, this is the address of a circular list of
795 * the page's disk buffers. PG_private must be set to tell the VM to call
796 * into the filesystem to release these pages.
797 *
798 * A page may belong to an inode's memory mapping. In this case, page->mapping
799 * is the pointer to the inode, and page->index is the file offset of the page,
800 * in units of PAGE_SIZE.
801 *
802 * If pagecache pages are not associated with an inode, they are said to be
803 * anonymous pages. These may become associated with the swapcache, and in that
804 * case PG_swapcache is set, and page->private is an offset into the swapcache.
805 *
806 * In either case (swapcache or inode backed), the pagecache itself holds one
807 * reference to the page. Setting PG_private should also increment the
808 * refcount. The each user mapping also has a reference to the page.
809 *
810 * The pagecache pages are stored in a per-mapping radix tree, which is
811 * rooted at mapping->i_pages, and indexed by offset.
812 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
813 * lists, we instead now tag pages as dirty/writeback in the radix tree.
814 *
815 * All pagecache pages may be subject to I/O:
816 * - inode pages may need to be read from disk,
817 * - inode pages which have been modified and are MAP_SHARED may need
818 * to be written back to the inode on disk,
819 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
820 * modified may need to be swapped out to swap space and (later) to be read
821 * back into memory.
822 */
823
824 /*
825 * The zone field is never updated after free_area_init_core()
826 * sets it, so none of the operations on it need to be atomic.
827 */
828
829 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
830 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
831 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
832 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
833 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
834 #define KASAN_TAG_PGOFF (LAST_CPUPID_PGOFF - KASAN_TAG_WIDTH)
835
836 /*
837 * Define the bit shifts to access each section. For non-existent
838 * sections we define the shift as 0; that plus a 0 mask ensures
839 * the compiler will optimise away reference to them.
840 */
841 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
842 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
843 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
844 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
845 #define KASAN_TAG_PGSHIFT (KASAN_TAG_PGOFF * (KASAN_TAG_WIDTH != 0))
846
847 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
848 #ifdef NODE_NOT_IN_PAGE_FLAGS
849 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
850 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
851 SECTIONS_PGOFF : ZONES_PGOFF)
852 #else
853 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
854 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
855 NODES_PGOFF : ZONES_PGOFF)
856 #endif
857
858 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
859
860 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
861 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
862 #endif
863
864 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
865 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
866 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
867 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
868 #define KASAN_TAG_MASK ((1UL << KASAN_TAG_WIDTH) - 1)
869 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
870
871 static inline enum zone_type page_zonenum(const struct page *page)
872 {
873 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
874 }
875
876 #ifdef CONFIG_ZONE_DEVICE
877 static inline bool is_zone_device_page(const struct page *page)
878 {
879 return page_zonenum(page) == ZONE_DEVICE;
880 }
881 extern void memmap_init_zone_device(struct zone *, unsigned long,
882 unsigned long, struct dev_pagemap *);
883 #else
884 static inline bool is_zone_device_page(const struct page *page)
885 {
886 return false;
887 }
888 #endif
889
890 #ifdef CONFIG_DEV_PAGEMAP_OPS
891 void dev_pagemap_get_ops(void);
892 void dev_pagemap_put_ops(void);
893 void __put_devmap_managed_page(struct page *page);
894 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
895 static inline bool put_devmap_managed_page(struct page *page)
896 {
897 if (!static_branch_unlikely(&devmap_managed_key))
898 return false;
899 if (!is_zone_device_page(page))
900 return false;
901 switch (page->pgmap->type) {
902 case MEMORY_DEVICE_PRIVATE:
903 case MEMORY_DEVICE_PUBLIC:
904 case MEMORY_DEVICE_FS_DAX:
905 __put_devmap_managed_page(page);
906 return true;
907 default:
908 break;
909 }
910 return false;
911 }
912
913 static inline bool is_device_private_page(const struct page *page)
914 {
915 return is_zone_device_page(page) &&
916 page->pgmap->type == MEMORY_DEVICE_PRIVATE;
917 }
918
919 static inline bool is_device_public_page(const struct page *page)
920 {
921 return is_zone_device_page(page) &&
922 page->pgmap->type == MEMORY_DEVICE_PUBLIC;
923 }
924
925 #ifdef CONFIG_PCI_P2PDMA
926 static inline bool is_pci_p2pdma_page(const struct page *page)
927 {
928 return is_zone_device_page(page) &&
929 page->pgmap->type == MEMORY_DEVICE_PCI_P2PDMA;
930 }
931 #else /* CONFIG_PCI_P2PDMA */
932 static inline bool is_pci_p2pdma_page(const struct page *page)
933 {
934 return false;
935 }
936 #endif /* CONFIG_PCI_P2PDMA */
937
938 #else /* CONFIG_DEV_PAGEMAP_OPS */
939 static inline void dev_pagemap_get_ops(void)
940 {
941 }
942
943 static inline void dev_pagemap_put_ops(void)
944 {
945 }
946
947 static inline bool put_devmap_managed_page(struct page *page)
948 {
949 return false;
950 }
951
952 static inline bool is_device_private_page(const struct page *page)
953 {
954 return false;
955 }
956
957 static inline bool is_device_public_page(const struct page *page)
958 {
959 return false;
960 }
961
962 static inline bool is_pci_p2pdma_page(const struct page *page)
963 {
964 return false;
965 }
966 #endif /* CONFIG_DEV_PAGEMAP_OPS */
967
968 /* 127: arbitrary random number, small enough to assemble well */
969 #define page_ref_zero_or_close_to_overflow(page) \
970 ((unsigned int) page_ref_count(page) + 127u <= 127u)
971
972 static inline void get_page(struct page *page)
973 {
974 page = compound_head(page);
975 /*
976 * Getting a normal page or the head of a compound page
977 * requires to already have an elevated page->_refcount.
978 */
979 VM_BUG_ON_PAGE(page_ref_zero_or_close_to_overflow(page), page);
980 page_ref_inc(page);
981 }
982
983 static inline void put_page(struct page *page)
984 {
985 page = compound_head(page);
986
987 /*
988 * For devmap managed pages we need to catch refcount transition from
989 * 2 to 1, when refcount reach one it means the page is free and we
990 * need to inform the device driver through callback. See
991 * include/linux/memremap.h and HMM for details.
992 */
993 if (put_devmap_managed_page(page))
994 return;
995
996 if (put_page_testzero(page))
997 __put_page(page);
998 }
999
1000 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1001 #define SECTION_IN_PAGE_FLAGS
1002 #endif
1003
1004 /*
1005 * The identification function is mainly used by the buddy allocator for
1006 * determining if two pages could be buddies. We are not really identifying
1007 * the zone since we could be using the section number id if we do not have
1008 * node id available in page flags.
1009 * We only guarantee that it will return the same value for two combinable
1010 * pages in a zone.
1011 */
1012 static inline int page_zone_id(struct page *page)
1013 {
1014 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1015 }
1016
1017 #ifdef NODE_NOT_IN_PAGE_FLAGS
1018 extern int page_to_nid(const struct page *page);
1019 #else
1020 static inline int page_to_nid(const struct page *page)
1021 {
1022 struct page *p = (struct page *)page;
1023
1024 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1025 }
1026 #endif
1027
1028 #ifdef CONFIG_NUMA_BALANCING
1029 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1030 {
1031 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1032 }
1033
1034 static inline int cpupid_to_pid(int cpupid)
1035 {
1036 return cpupid & LAST__PID_MASK;
1037 }
1038
1039 static inline int cpupid_to_cpu(int cpupid)
1040 {
1041 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1042 }
1043
1044 static inline int cpupid_to_nid(int cpupid)
1045 {
1046 return cpu_to_node(cpupid_to_cpu(cpupid));
1047 }
1048
1049 static inline bool cpupid_pid_unset(int cpupid)
1050 {
1051 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1052 }
1053
1054 static inline bool cpupid_cpu_unset(int cpupid)
1055 {
1056 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1057 }
1058
1059 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1060 {
1061 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1062 }
1063
1064 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1065 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1066 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1067 {
1068 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1069 }
1070
1071 static inline int page_cpupid_last(struct page *page)
1072 {
1073 return page->_last_cpupid;
1074 }
1075 static inline void page_cpupid_reset_last(struct page *page)
1076 {
1077 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1078 }
1079 #else
1080 static inline int page_cpupid_last(struct page *page)
1081 {
1082 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1083 }
1084
1085 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1086
1087 static inline void page_cpupid_reset_last(struct page *page)
1088 {
1089 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1090 }
1091 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1092 #else /* !CONFIG_NUMA_BALANCING */
1093 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1094 {
1095 return page_to_nid(page); /* XXX */
1096 }
1097
1098 static inline int page_cpupid_last(struct page *page)
1099 {
1100 return page_to_nid(page); /* XXX */
1101 }
1102
1103 static inline int cpupid_to_nid(int cpupid)
1104 {
1105 return -1;
1106 }
1107
1108 static inline int cpupid_to_pid(int cpupid)
1109 {
1110 return -1;
1111 }
1112
1113 static inline int cpupid_to_cpu(int cpupid)
1114 {
1115 return -1;
1116 }
1117
1118 static inline int cpu_pid_to_cpupid(int nid, int pid)
1119 {
1120 return -1;
1121 }
1122
1123 static inline bool cpupid_pid_unset(int cpupid)
1124 {
1125 return 1;
1126 }
1127
1128 static inline void page_cpupid_reset_last(struct page *page)
1129 {
1130 }
1131
1132 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1133 {
1134 return false;
1135 }
1136 #endif /* CONFIG_NUMA_BALANCING */
1137
1138 #ifdef CONFIG_KASAN_SW_TAGS
1139 static inline u8 page_kasan_tag(const struct page *page)
1140 {
1141 return (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1142 }
1143
1144 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1145 {
1146 page->flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1147 page->flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1148 }
1149
1150 static inline void page_kasan_tag_reset(struct page *page)
1151 {
1152 page_kasan_tag_set(page, 0xff);
1153 }
1154 #else
1155 static inline u8 page_kasan_tag(const struct page *page)
1156 {
1157 return 0xff;
1158 }
1159
1160 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1161 static inline void page_kasan_tag_reset(struct page *page) { }
1162 #endif
1163
1164 static inline struct zone *page_zone(const struct page *page)
1165 {
1166 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1167 }
1168
1169 static inline pg_data_t *page_pgdat(const struct page *page)
1170 {
1171 return NODE_DATA(page_to_nid(page));
1172 }
1173
1174 #ifdef SECTION_IN_PAGE_FLAGS
1175 static inline void set_page_section(struct page *page, unsigned long section)
1176 {
1177 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1178 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1179 }
1180
1181 static inline unsigned long page_to_section(const struct page *page)
1182 {
1183 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1184 }
1185 #endif
1186
1187 static inline void set_page_zone(struct page *page, enum zone_type zone)
1188 {
1189 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1190 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1191 }
1192
1193 static inline void set_page_node(struct page *page, unsigned long node)
1194 {
1195 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1196 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1197 }
1198
1199 static inline void set_page_links(struct page *page, enum zone_type zone,
1200 unsigned long node, unsigned long pfn)
1201 {
1202 set_page_zone(page, zone);
1203 set_page_node(page, node);
1204 #ifdef SECTION_IN_PAGE_FLAGS
1205 set_page_section(page, pfn_to_section_nr(pfn));
1206 #endif
1207 }
1208
1209 #ifdef CONFIG_MEMCG
1210 static inline struct mem_cgroup *page_memcg(struct page *page)
1211 {
1212 return page->mem_cgroup;
1213 }
1214 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1215 {
1216 WARN_ON_ONCE(!rcu_read_lock_held());
1217 return READ_ONCE(page->mem_cgroup);
1218 }
1219 #else
1220 static inline struct mem_cgroup *page_memcg(struct page *page)
1221 {
1222 return NULL;
1223 }
1224 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1225 {
1226 WARN_ON_ONCE(!rcu_read_lock_held());
1227 return NULL;
1228 }
1229 #endif
1230
1231 /*
1232 * Some inline functions in vmstat.h depend on page_zone()
1233 */
1234 #include <linux/vmstat.h>
1235
1236 static __always_inline void *lowmem_page_address(const struct page *page)
1237 {
1238 return page_to_virt(page);
1239 }
1240
1241 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1242 #define HASHED_PAGE_VIRTUAL
1243 #endif
1244
1245 #if defined(WANT_PAGE_VIRTUAL)
1246 static inline void *page_address(const struct page *page)
1247 {
1248 return page->virtual;
1249 }
1250 static inline void set_page_address(struct page *page, void *address)
1251 {
1252 page->virtual = address;
1253 }
1254 #define page_address_init() do { } while(0)
1255 #endif
1256
1257 #if defined(HASHED_PAGE_VIRTUAL)
1258 void *page_address(const struct page *page);
1259 void set_page_address(struct page *page, void *virtual);
1260 void page_address_init(void);
1261 #endif
1262
1263 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1264 #define page_address(page) lowmem_page_address(page)
1265 #define set_page_address(page, address) do { } while(0)
1266 #define page_address_init() do { } while(0)
1267 #endif
1268
1269 extern void *page_rmapping(struct page *page);
1270 extern struct anon_vma *page_anon_vma(struct page *page);
1271 extern struct address_space *page_mapping(struct page *page);
1272
1273 extern struct address_space *__page_file_mapping(struct page *);
1274
1275 static inline
1276 struct address_space *page_file_mapping(struct page *page)
1277 {
1278 if (unlikely(PageSwapCache(page)))
1279 return __page_file_mapping(page);
1280
1281 return page->mapping;
1282 }
1283
1284 extern pgoff_t __page_file_index(struct page *page);
1285
1286 /*
1287 * Return the pagecache index of the passed page. Regular pagecache pages
1288 * use ->index whereas swapcache pages use swp_offset(->private)
1289 */
1290 static inline pgoff_t page_index(struct page *page)
1291 {
1292 if (unlikely(PageSwapCache(page)))
1293 return __page_file_index(page);
1294 return page->index;
1295 }
1296
1297 bool page_mapped(struct page *page);
1298 struct address_space *page_mapping(struct page *page);
1299 struct address_space *page_mapping_file(struct page *page);
1300
1301 /*
1302 * Return true only if the page has been allocated with
1303 * ALLOC_NO_WATERMARKS and the low watermark was not
1304 * met implying that the system is under some pressure.
1305 */
1306 static inline bool page_is_pfmemalloc(struct page *page)
1307 {
1308 /*
1309 * Page index cannot be this large so this must be
1310 * a pfmemalloc page.
1311 */
1312 return page->index == -1UL;
1313 }
1314
1315 /*
1316 * Only to be called by the page allocator on a freshly allocated
1317 * page.
1318 */
1319 static inline void set_page_pfmemalloc(struct page *page)
1320 {
1321 page->index = -1UL;
1322 }
1323
1324 static inline void clear_page_pfmemalloc(struct page *page)
1325 {
1326 page->index = 0;
1327 }
1328
1329 /*
1330 * Different kinds of faults, as returned by handle_mm_fault().
1331 * Used to decide whether a process gets delivered SIGBUS or
1332 * just gets major/minor fault counters bumped up.
1333 */
1334
1335 #define VM_FAULT_OOM 0x0001
1336 #define VM_FAULT_SIGBUS 0x0002
1337 #define VM_FAULT_MAJOR 0x0004
1338 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1339 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1340 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1341 #define VM_FAULT_SIGSEGV 0x0040
1342
1343 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1344 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1345 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1346 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1347 #define VM_FAULT_DONE_COW 0x1000 /* ->fault has fully handled COW */
1348 #define VM_FAULT_NEEDDSYNC 0x2000 /* ->fault did not modify page tables
1349 * and needs fsync() to complete (for
1350 * synchronous page faults in DAX) */
1351
1352 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1353 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1354 VM_FAULT_FALLBACK)
1355
1356 #define VM_FAULT_RESULT_TRACE \
1357 { VM_FAULT_OOM, "OOM" }, \
1358 { VM_FAULT_SIGBUS, "SIGBUS" }, \
1359 { VM_FAULT_MAJOR, "MAJOR" }, \
1360 { VM_FAULT_WRITE, "WRITE" }, \
1361 { VM_FAULT_HWPOISON, "HWPOISON" }, \
1362 { VM_FAULT_HWPOISON_LARGE, "HWPOISON_LARGE" }, \
1363 { VM_FAULT_SIGSEGV, "SIGSEGV" }, \
1364 { VM_FAULT_NOPAGE, "NOPAGE" }, \
1365 { VM_FAULT_LOCKED, "LOCKED" }, \
1366 { VM_FAULT_RETRY, "RETRY" }, \
1367 { VM_FAULT_FALLBACK, "FALLBACK" }, \
1368 { VM_FAULT_DONE_COW, "DONE_COW" }, \
1369 { VM_FAULT_NEEDDSYNC, "NEEDDSYNC" }
1370
1371 /* Encode hstate index for a hwpoisoned large page */
1372 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1373 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1374
1375 /*
1376 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1377 */
1378 extern void pagefault_out_of_memory(void);
1379
1380 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1381
1382 /*
1383 * Flags passed to show_mem() and show_free_areas() to suppress output in
1384 * various contexts.
1385 */
1386 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1387
1388 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1389
1390 extern bool can_do_mlock(void);
1391 extern int user_shm_lock(size_t, struct user_struct *);
1392 extern void user_shm_unlock(size_t, struct user_struct *);
1393
1394 /*
1395 * Parameter block passed down to zap_pte_range in exceptional cases.
1396 */
1397 struct zap_details {
1398 struct address_space *check_mapping; /* Check page->mapping if set */
1399 pgoff_t first_index; /* Lowest page->index to unmap */
1400 pgoff_t last_index; /* Highest page->index to unmap */
1401 };
1402
1403 struct page *_vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1404 pte_t pte, bool with_public_device);
1405 #define vm_normal_page(vma, addr, pte) _vm_normal_page(vma, addr, pte, false)
1406
1407 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1408 pmd_t pmd);
1409
1410 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1411 unsigned long size);
1412 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1413 unsigned long size);
1414 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1415 unsigned long start, unsigned long end);
1416
1417 /**
1418 * mm_walk - callbacks for walk_page_range
1419 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
1420 * this handler should only handle pud_trans_huge() puds.
1421 * the pmd_entry or pte_entry callbacks will be used for
1422 * regular PUDs.
1423 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1424 * this handler is required to be able to handle
1425 * pmd_trans_huge() pmds. They may simply choose to
1426 * split_huge_page() instead of handling it explicitly.
1427 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1428 * @pte_hole: if set, called for each hole at all levels
1429 * @hugetlb_entry: if set, called for each hugetlb entry
1430 * @test_walk: caller specific callback function to determine whether
1431 * we walk over the current vma or not. Returning 0
1432 * value means "do page table walk over the current vma,"
1433 * and a negative one means "abort current page table walk
1434 * right now." 1 means "skip the current vma."
1435 * @mm: mm_struct representing the target process of page table walk
1436 * @vma: vma currently walked (NULL if walking outside vmas)
1437 * @private: private data for callbacks' usage
1438 *
1439 * (see the comment on walk_page_range() for more details)
1440 */
1441 struct mm_walk {
1442 int (*pud_entry)(pud_t *pud, unsigned long addr,
1443 unsigned long next, struct mm_walk *walk);
1444 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1445 unsigned long next, struct mm_walk *walk);
1446 int (*pte_entry)(pte_t *pte, unsigned long addr,
1447 unsigned long next, struct mm_walk *walk);
1448 int (*pte_hole)(unsigned long addr, unsigned long next,
1449 struct mm_walk *walk);
1450 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1451 unsigned long addr, unsigned long next,
1452 struct mm_walk *walk);
1453 int (*test_walk)(unsigned long addr, unsigned long next,
1454 struct mm_walk *walk);
1455 struct mm_struct *mm;
1456 struct vm_area_struct *vma;
1457 void *private;
1458 };
1459
1460 struct mmu_notifier_range;
1461
1462 int walk_page_range(unsigned long addr, unsigned long end,
1463 struct mm_walk *walk);
1464 int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1465 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1466 unsigned long end, unsigned long floor, unsigned long ceiling);
1467 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1468 struct vm_area_struct *vma);
1469 int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
1470 struct mmu_notifier_range *range,
1471 pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp);
1472 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1473 unsigned long *pfn);
1474 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1475 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1476 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1477 void *buf, int len, int write);
1478
1479 extern void truncate_pagecache(struct inode *inode, loff_t new);
1480 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1481 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1482 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1483 int truncate_inode_page(struct address_space *mapping, struct page *page);
1484 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1485 int invalidate_inode_page(struct page *page);
1486
1487 #ifdef CONFIG_MMU
1488 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1489 unsigned long address, unsigned int flags);
1490 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1491 unsigned long address, unsigned int fault_flags,
1492 bool *unlocked);
1493 void unmap_mapping_pages(struct address_space *mapping,
1494 pgoff_t start, pgoff_t nr, bool even_cows);
1495 void unmap_mapping_range(struct address_space *mapping,
1496 loff_t const holebegin, loff_t const holelen, int even_cows);
1497 #else
1498 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1499 unsigned long address, unsigned int flags)
1500 {
1501 /* should never happen if there's no MMU */
1502 BUG();
1503 return VM_FAULT_SIGBUS;
1504 }
1505 static inline int fixup_user_fault(struct task_struct *tsk,
1506 struct mm_struct *mm, unsigned long address,
1507 unsigned int fault_flags, bool *unlocked)
1508 {
1509 /* should never happen if there's no MMU */
1510 BUG();
1511 return -EFAULT;
1512 }
1513 static inline void unmap_mapping_pages(struct address_space *mapping,
1514 pgoff_t start, pgoff_t nr, bool even_cows) { }
1515 static inline void unmap_mapping_range(struct address_space *mapping,
1516 loff_t const holebegin, loff_t const holelen, int even_cows) { }
1517 #endif
1518
1519 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1520 loff_t const holebegin, loff_t const holelen)
1521 {
1522 unmap_mapping_range(mapping, holebegin, holelen, 0);
1523 }
1524
1525 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1526 void *buf, int len, unsigned int gup_flags);
1527 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1528 void *buf, int len, unsigned int gup_flags);
1529 extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1530 unsigned long addr, void *buf, int len, unsigned int gup_flags);
1531
1532 long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1533 unsigned long start, unsigned long nr_pages,
1534 unsigned int gup_flags, struct page **pages,
1535 struct vm_area_struct **vmas, int *locked);
1536 long get_user_pages(unsigned long start, unsigned long nr_pages,
1537 unsigned int gup_flags, struct page **pages,
1538 struct vm_area_struct **vmas);
1539 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1540 unsigned int gup_flags, struct page **pages, int *locked);
1541 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1542 struct page **pages, unsigned int gup_flags);
1543 #ifdef CONFIG_FS_DAX
1544 long get_user_pages_longterm(unsigned long start, unsigned long nr_pages,
1545 unsigned int gup_flags, struct page **pages,
1546 struct vm_area_struct **vmas);
1547 #else
1548 static inline long get_user_pages_longterm(unsigned long start,
1549 unsigned long nr_pages, unsigned int gup_flags,
1550 struct page **pages, struct vm_area_struct **vmas)
1551 {
1552 return get_user_pages(start, nr_pages, gup_flags, pages, vmas);
1553 }
1554 #endif /* CONFIG_FS_DAX */
1555
1556 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1557 struct page **pages);
1558
1559 /* Container for pinned pfns / pages */
1560 struct frame_vector {
1561 unsigned int nr_allocated; /* Number of frames we have space for */
1562 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1563 bool got_ref; /* Did we pin pages by getting page ref? */
1564 bool is_pfns; /* Does array contain pages or pfns? */
1565 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1566 * pfns_vector_pages() or pfns_vector_pfns()
1567 * for access */
1568 };
1569
1570 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1571 void frame_vector_destroy(struct frame_vector *vec);
1572 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1573 unsigned int gup_flags, struct frame_vector *vec);
1574 void put_vaddr_frames(struct frame_vector *vec);
1575 int frame_vector_to_pages(struct frame_vector *vec);
1576 void frame_vector_to_pfns(struct frame_vector *vec);
1577
1578 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1579 {
1580 return vec->nr_frames;
1581 }
1582
1583 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1584 {
1585 if (vec->is_pfns) {
1586 int err = frame_vector_to_pages(vec);
1587
1588 if (err)
1589 return ERR_PTR(err);
1590 }
1591 return (struct page **)(vec->ptrs);
1592 }
1593
1594 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1595 {
1596 if (!vec->is_pfns)
1597 frame_vector_to_pfns(vec);
1598 return (unsigned long *)(vec->ptrs);
1599 }
1600
1601 struct kvec;
1602 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1603 struct page **pages);
1604 int get_kernel_page(unsigned long start, int write, struct page **pages);
1605 struct page *get_dump_page(unsigned long addr);
1606
1607 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1608 extern void do_invalidatepage(struct page *page, unsigned int offset,
1609 unsigned int length);
1610
1611 void __set_page_dirty(struct page *, struct address_space *, int warn);
1612 int __set_page_dirty_nobuffers(struct page *page);
1613 int __set_page_dirty_no_writeback(struct page *page);
1614 int redirty_page_for_writepage(struct writeback_control *wbc,
1615 struct page *page);
1616 void account_page_dirtied(struct page *page, struct address_space *mapping);
1617 void account_page_cleaned(struct page *page, struct address_space *mapping,
1618 struct bdi_writeback *wb);
1619 int set_page_dirty(struct page *page);
1620 int set_page_dirty_lock(struct page *page);
1621 void __cancel_dirty_page(struct page *page);
1622 static inline void cancel_dirty_page(struct page *page)
1623 {
1624 /* Avoid atomic ops, locking, etc. when not actually needed. */
1625 if (PageDirty(page))
1626 __cancel_dirty_page(page);
1627 }
1628 int clear_page_dirty_for_io(struct page *page);
1629
1630 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1631
1632 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1633 {
1634 return !vma->vm_ops;
1635 }
1636
1637 #ifdef CONFIG_SHMEM
1638 /*
1639 * The vma_is_shmem is not inline because it is used only by slow
1640 * paths in userfault.
1641 */
1642 bool vma_is_shmem(struct vm_area_struct *vma);
1643 #else
1644 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
1645 #endif
1646
1647 int vma_is_stack_for_current(struct vm_area_struct *vma);
1648
1649 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1650 unsigned long old_addr, struct vm_area_struct *new_vma,
1651 unsigned long new_addr, unsigned long len,
1652 bool need_rmap_locks);
1653 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1654 unsigned long end, pgprot_t newprot,
1655 int dirty_accountable, int prot_numa);
1656 extern int mprotect_fixup(struct vm_area_struct *vma,
1657 struct vm_area_struct **pprev, unsigned long start,
1658 unsigned long end, unsigned long newflags);
1659
1660 /*
1661 * doesn't attempt to fault and will return short.
1662 */
1663 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1664 struct page **pages);
1665 /*
1666 * per-process(per-mm_struct) statistics.
1667 */
1668 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1669 {
1670 long val = atomic_long_read(&mm->rss_stat.count[member]);
1671
1672 #ifdef SPLIT_RSS_COUNTING
1673 /*
1674 * counter is updated in asynchronous manner and may go to minus.
1675 * But it's never be expected number for users.
1676 */
1677 if (val < 0)
1678 val = 0;
1679 #endif
1680 return (unsigned long)val;
1681 }
1682
1683 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1684 {
1685 atomic_long_add(value, &mm->rss_stat.count[member]);
1686 }
1687
1688 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1689 {
1690 atomic_long_inc(&mm->rss_stat.count[member]);
1691 }
1692
1693 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1694 {
1695 atomic_long_dec(&mm->rss_stat.count[member]);
1696 }
1697
1698 /* Optimized variant when page is already known not to be PageAnon */
1699 static inline int mm_counter_file(struct page *page)
1700 {
1701 if (PageSwapBacked(page))
1702 return MM_SHMEMPAGES;
1703 return MM_FILEPAGES;
1704 }
1705
1706 static inline int mm_counter(struct page *page)
1707 {
1708 if (PageAnon(page))
1709 return MM_ANONPAGES;
1710 return mm_counter_file(page);
1711 }
1712
1713 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1714 {
1715 return get_mm_counter(mm, MM_FILEPAGES) +
1716 get_mm_counter(mm, MM_ANONPAGES) +
1717 get_mm_counter(mm, MM_SHMEMPAGES);
1718 }
1719
1720 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1721 {
1722 return max(mm->hiwater_rss, get_mm_rss(mm));
1723 }
1724
1725 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1726 {
1727 return max(mm->hiwater_vm, mm->total_vm);
1728 }
1729
1730 static inline void update_hiwater_rss(struct mm_struct *mm)
1731 {
1732 unsigned long _rss = get_mm_rss(mm);
1733
1734 if ((mm)->hiwater_rss < _rss)
1735 (mm)->hiwater_rss = _rss;
1736 }
1737
1738 static inline void update_hiwater_vm(struct mm_struct *mm)
1739 {
1740 if (mm->hiwater_vm < mm->total_vm)
1741 mm->hiwater_vm = mm->total_vm;
1742 }
1743
1744 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1745 {
1746 mm->hiwater_rss = get_mm_rss(mm);
1747 }
1748
1749 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1750 struct mm_struct *mm)
1751 {
1752 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1753
1754 if (*maxrss < hiwater_rss)
1755 *maxrss = hiwater_rss;
1756 }
1757
1758 #if defined(SPLIT_RSS_COUNTING)
1759 void sync_mm_rss(struct mm_struct *mm);
1760 #else
1761 static inline void sync_mm_rss(struct mm_struct *mm)
1762 {
1763 }
1764 #endif
1765
1766 #ifndef __HAVE_ARCH_PTE_DEVMAP
1767 static inline int pte_devmap(pte_t pte)
1768 {
1769 return 0;
1770 }
1771 #endif
1772
1773 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1774
1775 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1776 spinlock_t **ptl);
1777 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1778 spinlock_t **ptl)
1779 {
1780 pte_t *ptep;
1781 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1782 return ptep;
1783 }
1784
1785 #ifdef __PAGETABLE_P4D_FOLDED
1786 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1787 unsigned long address)
1788 {
1789 return 0;
1790 }
1791 #else
1792 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1793 #endif
1794
1795 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
1796 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1797 unsigned long address)
1798 {
1799 return 0;
1800 }
1801 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
1802 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
1803
1804 #else
1805 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
1806
1807 static inline void mm_inc_nr_puds(struct mm_struct *mm)
1808 {
1809 if (mm_pud_folded(mm))
1810 return;
1811 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1812 }
1813
1814 static inline void mm_dec_nr_puds(struct mm_struct *mm)
1815 {
1816 if (mm_pud_folded(mm))
1817 return;
1818 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1819 }
1820 #endif
1821
1822 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1823 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1824 unsigned long address)
1825 {
1826 return 0;
1827 }
1828
1829 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1830 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1831
1832 #else
1833 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1834
1835 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1836 {
1837 if (mm_pmd_folded(mm))
1838 return;
1839 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1840 }
1841
1842 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1843 {
1844 if (mm_pmd_folded(mm))
1845 return;
1846 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1847 }
1848 #endif
1849
1850 #ifdef CONFIG_MMU
1851 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
1852 {
1853 atomic_long_set(&mm->pgtables_bytes, 0);
1854 }
1855
1856 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1857 {
1858 return atomic_long_read(&mm->pgtables_bytes);
1859 }
1860
1861 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
1862 {
1863 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1864 }
1865
1866 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
1867 {
1868 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1869 }
1870 #else
1871
1872 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
1873 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1874 {
1875 return 0;
1876 }
1877
1878 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
1879 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
1880 #endif
1881
1882 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
1883 int __pte_alloc_kernel(pmd_t *pmd);
1884
1885 /*
1886 * The following ifdef needed to get the 4level-fixup.h header to work.
1887 * Remove it when 4level-fixup.h has been removed.
1888 */
1889 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1890
1891 #ifndef __ARCH_HAS_5LEVEL_HACK
1892 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1893 unsigned long address)
1894 {
1895 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
1896 NULL : p4d_offset(pgd, address);
1897 }
1898
1899 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1900 unsigned long address)
1901 {
1902 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
1903 NULL : pud_offset(p4d, address);
1904 }
1905 #endif /* !__ARCH_HAS_5LEVEL_HACK */
1906
1907 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1908 {
1909 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1910 NULL: pmd_offset(pud, address);
1911 }
1912 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1913
1914 #if USE_SPLIT_PTE_PTLOCKS
1915 #if ALLOC_SPLIT_PTLOCKS
1916 void __init ptlock_cache_init(void);
1917 extern bool ptlock_alloc(struct page *page);
1918 extern void ptlock_free(struct page *page);
1919
1920 static inline spinlock_t *ptlock_ptr(struct page *page)
1921 {
1922 return page->ptl;
1923 }
1924 #else /* ALLOC_SPLIT_PTLOCKS */
1925 static inline void ptlock_cache_init(void)
1926 {
1927 }
1928
1929 static inline bool ptlock_alloc(struct page *page)
1930 {
1931 return true;
1932 }
1933
1934 static inline void ptlock_free(struct page *page)
1935 {
1936 }
1937
1938 static inline spinlock_t *ptlock_ptr(struct page *page)
1939 {
1940 return &page->ptl;
1941 }
1942 #endif /* ALLOC_SPLIT_PTLOCKS */
1943
1944 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1945 {
1946 return ptlock_ptr(pmd_page(*pmd));
1947 }
1948
1949 static inline bool ptlock_init(struct page *page)
1950 {
1951 /*
1952 * prep_new_page() initialize page->private (and therefore page->ptl)
1953 * with 0. Make sure nobody took it in use in between.
1954 *
1955 * It can happen if arch try to use slab for page table allocation:
1956 * slab code uses page->slab_cache, which share storage with page->ptl.
1957 */
1958 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1959 if (!ptlock_alloc(page))
1960 return false;
1961 spin_lock_init(ptlock_ptr(page));
1962 return true;
1963 }
1964
1965 #else /* !USE_SPLIT_PTE_PTLOCKS */
1966 /*
1967 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1968 */
1969 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1970 {
1971 return &mm->page_table_lock;
1972 }
1973 static inline void ptlock_cache_init(void) {}
1974 static inline bool ptlock_init(struct page *page) { return true; }
1975 static inline void ptlock_free(struct page *page) {}
1976 #endif /* USE_SPLIT_PTE_PTLOCKS */
1977
1978 static inline void pgtable_init(void)
1979 {
1980 ptlock_cache_init();
1981 pgtable_cache_init();
1982 }
1983
1984 static inline bool pgtable_page_ctor(struct page *page)
1985 {
1986 if (!ptlock_init(page))
1987 return false;
1988 __SetPageTable(page);
1989 inc_zone_page_state(page, NR_PAGETABLE);
1990 return true;
1991 }
1992
1993 static inline void pgtable_page_dtor(struct page *page)
1994 {
1995 ptlock_free(page);
1996 __ClearPageTable(page);
1997 dec_zone_page_state(page, NR_PAGETABLE);
1998 }
1999
2000 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
2001 ({ \
2002 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
2003 pte_t *__pte = pte_offset_map(pmd, address); \
2004 *(ptlp) = __ptl; \
2005 spin_lock(__ptl); \
2006 __pte; \
2007 })
2008
2009 #define pte_unmap_unlock(pte, ptl) do { \
2010 spin_unlock(ptl); \
2011 pte_unmap(pte); \
2012 } while (0)
2013
2014 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2015
2016 #define pte_alloc_map(mm, pmd, address) \
2017 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2018
2019 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2020 (pte_alloc(mm, pmd) ? \
2021 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2022
2023 #define pte_alloc_kernel(pmd, address) \
2024 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2025 NULL: pte_offset_kernel(pmd, address))
2026
2027 #if USE_SPLIT_PMD_PTLOCKS
2028
2029 static struct page *pmd_to_page(pmd_t *pmd)
2030 {
2031 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2032 return virt_to_page((void *)((unsigned long) pmd & mask));
2033 }
2034
2035 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2036 {
2037 return ptlock_ptr(pmd_to_page(pmd));
2038 }
2039
2040 static inline bool pgtable_pmd_page_ctor(struct page *page)
2041 {
2042 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2043 page->pmd_huge_pte = NULL;
2044 #endif
2045 return ptlock_init(page);
2046 }
2047
2048 static inline void pgtable_pmd_page_dtor(struct page *page)
2049 {
2050 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2051 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2052 #endif
2053 ptlock_free(page);
2054 }
2055
2056 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2057
2058 #else
2059
2060 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2061 {
2062 return &mm->page_table_lock;
2063 }
2064
2065 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
2066 static inline void pgtable_pmd_page_dtor(struct page *page) {}
2067
2068 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2069
2070 #endif
2071
2072 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2073 {
2074 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2075 spin_lock(ptl);
2076 return ptl;
2077 }
2078
2079 /*
2080 * No scalability reason to split PUD locks yet, but follow the same pattern
2081 * as the PMD locks to make it easier if we decide to. The VM should not be
2082 * considered ready to switch to split PUD locks yet; there may be places
2083 * which need to be converted from page_table_lock.
2084 */
2085 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2086 {
2087 return &mm->page_table_lock;
2088 }
2089
2090 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2091 {
2092 spinlock_t *ptl = pud_lockptr(mm, pud);
2093
2094 spin_lock(ptl);
2095 return ptl;
2096 }
2097
2098 extern void __init pagecache_init(void);
2099 extern void free_area_init(unsigned long * zones_size);
2100 extern void __init free_area_init_node(int nid, unsigned long * zones_size,
2101 unsigned long zone_start_pfn, unsigned long *zholes_size);
2102 extern void free_initmem(void);
2103
2104 /*
2105 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2106 * into the buddy system. The freed pages will be poisoned with pattern
2107 * "poison" if it's within range [0, UCHAR_MAX].
2108 * Return pages freed into the buddy system.
2109 */
2110 extern unsigned long free_reserved_area(void *start, void *end,
2111 int poison, const char *s);
2112
2113 #ifdef CONFIG_HIGHMEM
2114 /*
2115 * Free a highmem page into the buddy system, adjusting totalhigh_pages
2116 * and totalram_pages.
2117 */
2118 extern void free_highmem_page(struct page *page);
2119 #endif
2120
2121 extern void adjust_managed_page_count(struct page *page, long count);
2122 extern void mem_init_print_info(const char *str);
2123
2124 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2125
2126 /* Free the reserved page into the buddy system, so it gets managed. */
2127 static inline void __free_reserved_page(struct page *page)
2128 {
2129 ClearPageReserved(page);
2130 init_page_count(page);
2131 __free_page(page);
2132 }
2133
2134 static inline void free_reserved_page(struct page *page)
2135 {
2136 __free_reserved_page(page);
2137 adjust_managed_page_count(page, 1);
2138 }
2139
2140 static inline void mark_page_reserved(struct page *page)
2141 {
2142 SetPageReserved(page);
2143 adjust_managed_page_count(page, -1);
2144 }
2145
2146 /*
2147 * Default method to free all the __init memory into the buddy system.
2148 * The freed pages will be poisoned with pattern "poison" if it's within
2149 * range [0, UCHAR_MAX].
2150 * Return pages freed into the buddy system.
2151 */
2152 static inline unsigned long free_initmem_default(int poison)
2153 {
2154 extern char __init_begin[], __init_end[];
2155
2156 return free_reserved_area(&__init_begin, &__init_end,
2157 poison, "unused kernel");
2158 }
2159
2160 static inline unsigned long get_num_physpages(void)
2161 {
2162 int nid;
2163 unsigned long phys_pages = 0;
2164
2165 for_each_online_node(nid)
2166 phys_pages += node_present_pages(nid);
2167
2168 return phys_pages;
2169 }
2170
2171 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
2172 /*
2173 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
2174 * zones, allocate the backing mem_map and account for memory holes in a more
2175 * architecture independent manner. This is a substitute for creating the
2176 * zone_sizes[] and zholes_size[] arrays and passing them to
2177 * free_area_init_node()
2178 *
2179 * An architecture is expected to register range of page frames backed by
2180 * physical memory with memblock_add[_node]() before calling
2181 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
2182 * usage, an architecture is expected to do something like
2183 *
2184 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2185 * max_highmem_pfn};
2186 * for_each_valid_physical_page_range()
2187 * memblock_add_node(base, size, nid)
2188 * free_area_init_nodes(max_zone_pfns);
2189 *
2190 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
2191 * registered physical page range. Similarly
2192 * sparse_memory_present_with_active_regions() calls memory_present() for
2193 * each range when SPARSEMEM is enabled.
2194 *
2195 * See mm/page_alloc.c for more information on each function exposed by
2196 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
2197 */
2198 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
2199 unsigned long node_map_pfn_alignment(void);
2200 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2201 unsigned long end_pfn);
2202 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2203 unsigned long end_pfn);
2204 extern void get_pfn_range_for_nid(unsigned int nid,
2205 unsigned long *start_pfn, unsigned long *end_pfn);
2206 extern unsigned long find_min_pfn_with_active_regions(void);
2207 extern void free_bootmem_with_active_regions(int nid,
2208 unsigned long max_low_pfn);
2209 extern void sparse_memory_present_with_active_regions(int nid);
2210
2211 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
2212
2213 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
2214 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
2215 static inline int __early_pfn_to_nid(unsigned long pfn,
2216 struct mminit_pfnnid_cache *state)
2217 {
2218 return 0;
2219 }
2220 #else
2221 /* please see mm/page_alloc.c */
2222 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2223 /* there is a per-arch backend function. */
2224 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
2225 struct mminit_pfnnid_cache *state);
2226 #endif
2227
2228 #if !defined(CONFIG_FLAT_NODE_MEM_MAP)
2229 void zero_resv_unavail(void);
2230 #else
2231 static inline void zero_resv_unavail(void) {}
2232 #endif
2233
2234 extern void set_dma_reserve(unsigned long new_dma_reserve);
2235 extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long,
2236 enum memmap_context, struct vmem_altmap *);
2237 extern void setup_per_zone_wmarks(void);
2238 extern int __meminit init_per_zone_wmark_min(void);
2239 extern void mem_init(void);
2240 extern void __init mmap_init(void);
2241 extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2242 extern long si_mem_available(void);
2243 extern void si_meminfo(struct sysinfo * val);
2244 extern void si_meminfo_node(struct sysinfo *val, int nid);
2245 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2246 extern unsigned long arch_reserved_kernel_pages(void);
2247 #endif
2248
2249 extern __printf(3, 4)
2250 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2251
2252 extern void setup_per_cpu_pageset(void);
2253
2254 extern void zone_pcp_update(struct zone *zone);
2255 extern void zone_pcp_reset(struct zone *zone);
2256
2257 /* page_alloc.c */
2258 extern int min_free_kbytes;
2259 extern int watermark_boost_factor;
2260 extern int watermark_scale_factor;
2261
2262 /* nommu.c */
2263 extern atomic_long_t mmap_pages_allocated;
2264 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2265
2266 /* interval_tree.c */
2267 void vma_interval_tree_insert(struct vm_area_struct *node,
2268 struct rb_root_cached *root);
2269 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2270 struct vm_area_struct *prev,
2271 struct rb_root_cached *root);
2272 void vma_interval_tree_remove(struct vm_area_struct *node,
2273 struct rb_root_cached *root);
2274 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2275 unsigned long start, unsigned long last);
2276 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2277 unsigned long start, unsigned long last);
2278
2279 #define vma_interval_tree_foreach(vma, root, start, last) \
2280 for (vma = vma_interval_tree_iter_first(root, start, last); \
2281 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2282
2283 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2284 struct rb_root_cached *root);
2285 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2286 struct rb_root_cached *root);
2287 struct anon_vma_chain *
2288 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2289 unsigned long start, unsigned long last);
2290 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2291 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2292 #ifdef CONFIG_DEBUG_VM_RB
2293 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2294 #endif
2295
2296 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2297 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2298 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2299
2300 /* mmap.c */
2301 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2302 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2303 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2304 struct vm_area_struct *expand);
2305 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2306 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2307 {
2308 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2309 }
2310 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2311 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2312 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2313 struct mempolicy *, struct vm_userfaultfd_ctx);
2314 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2315 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2316 unsigned long addr, int new_below);
2317 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2318 unsigned long addr, int new_below);
2319 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2320 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2321 struct rb_node **, struct rb_node *);
2322 extern void unlink_file_vma(struct vm_area_struct *);
2323 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2324 unsigned long addr, unsigned long len, pgoff_t pgoff,
2325 bool *need_rmap_locks);
2326 extern void exit_mmap(struct mm_struct *);
2327
2328 static inline int check_data_rlimit(unsigned long rlim,
2329 unsigned long new,
2330 unsigned long start,
2331 unsigned long end_data,
2332 unsigned long start_data)
2333 {
2334 if (rlim < RLIM_INFINITY) {
2335 if (((new - start) + (end_data - start_data)) > rlim)
2336 return -ENOSPC;
2337 }
2338
2339 return 0;
2340 }
2341
2342 extern int mm_take_all_locks(struct mm_struct *mm);
2343 extern void mm_drop_all_locks(struct mm_struct *mm);
2344
2345 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2346 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2347 extern struct file *get_task_exe_file(struct task_struct *task);
2348
2349 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2350 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2351
2352 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2353 const struct vm_special_mapping *sm);
2354 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2355 unsigned long addr, unsigned long len,
2356 unsigned long flags,
2357 const struct vm_special_mapping *spec);
2358 /* This is an obsolete alternative to _install_special_mapping. */
2359 extern int install_special_mapping(struct mm_struct *mm,
2360 unsigned long addr, unsigned long len,
2361 unsigned long flags, struct page **pages);
2362
2363 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2364
2365 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2366 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2367 struct list_head *uf);
2368 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2369 unsigned long len, unsigned long prot, unsigned long flags,
2370 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate,
2371 struct list_head *uf);
2372 extern int __do_munmap(struct mm_struct *, unsigned long, size_t,
2373 struct list_head *uf, bool downgrade);
2374 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2375 struct list_head *uf);
2376
2377 static inline unsigned long
2378 do_mmap_pgoff(struct file *file, unsigned long addr,
2379 unsigned long len, unsigned long prot, unsigned long flags,
2380 unsigned long pgoff, unsigned long *populate,
2381 struct list_head *uf)
2382 {
2383 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate, uf);
2384 }
2385
2386 #ifdef CONFIG_MMU
2387 extern int __mm_populate(unsigned long addr, unsigned long len,
2388 int ignore_errors);
2389 static inline void mm_populate(unsigned long addr, unsigned long len)
2390 {
2391 /* Ignore errors */
2392 (void) __mm_populate(addr, len, 1);
2393 }
2394 #else
2395 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2396 #endif
2397
2398 /* These take the mm semaphore themselves */
2399 extern int __must_check vm_brk(unsigned long, unsigned long);
2400 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2401 extern int vm_munmap(unsigned long, size_t);
2402 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2403 unsigned long, unsigned long,
2404 unsigned long, unsigned long);
2405
2406 struct vm_unmapped_area_info {
2407 #define VM_UNMAPPED_AREA_TOPDOWN 1
2408 unsigned long flags;
2409 unsigned long length;
2410 unsigned long low_limit;
2411 unsigned long high_limit;
2412 unsigned long align_mask;
2413 unsigned long align_offset;
2414 };
2415
2416 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2417 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2418
2419 /*
2420 * Search for an unmapped address range.
2421 *
2422 * We are looking for a range that:
2423 * - does not intersect with any VMA;
2424 * - is contained within the [low_limit, high_limit) interval;
2425 * - is at least the desired size.
2426 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2427 */
2428 static inline unsigned long
2429 vm_unmapped_area(struct vm_unmapped_area_info *info)
2430 {
2431 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2432 return unmapped_area_topdown(info);
2433 else
2434 return unmapped_area(info);
2435 }
2436
2437 /* truncate.c */
2438 extern void truncate_inode_pages(struct address_space *, loff_t);
2439 extern void truncate_inode_pages_range(struct address_space *,
2440 loff_t lstart, loff_t lend);
2441 extern void truncate_inode_pages_final(struct address_space *);
2442
2443 /* generic vm_area_ops exported for stackable file systems */
2444 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2445 extern void filemap_map_pages(struct vm_fault *vmf,
2446 pgoff_t start_pgoff, pgoff_t end_pgoff);
2447 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2448
2449 /* mm/page-writeback.c */
2450 int __must_check write_one_page(struct page *page);
2451 void task_dirty_inc(struct task_struct *tsk);
2452
2453 /* readahead.c */
2454 #define VM_MAX_READAHEAD 128 /* kbytes */
2455 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
2456
2457 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2458 pgoff_t offset, unsigned long nr_to_read);
2459
2460 void page_cache_sync_readahead(struct address_space *mapping,
2461 struct file_ra_state *ra,
2462 struct file *filp,
2463 pgoff_t offset,
2464 unsigned long size);
2465
2466 void page_cache_async_readahead(struct address_space *mapping,
2467 struct file_ra_state *ra,
2468 struct file *filp,
2469 struct page *pg,
2470 pgoff_t offset,
2471 unsigned long size);
2472
2473 extern unsigned long stack_guard_gap;
2474 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2475 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2476
2477 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2478 extern int expand_downwards(struct vm_area_struct *vma,
2479 unsigned long address);
2480 #if VM_GROWSUP
2481 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2482 #else
2483 #define expand_upwards(vma, address) (0)
2484 #endif
2485
2486 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2487 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2488 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2489 struct vm_area_struct **pprev);
2490
2491 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2492 NULL if none. Assume start_addr < end_addr. */
2493 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2494 {
2495 struct vm_area_struct * vma = find_vma(mm,start_addr);
2496
2497 if (vma && end_addr <= vma->vm_start)
2498 vma = NULL;
2499 return vma;
2500 }
2501
2502 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2503 {
2504 unsigned long vm_start = vma->vm_start;
2505
2506 if (vma->vm_flags & VM_GROWSDOWN) {
2507 vm_start -= stack_guard_gap;
2508 if (vm_start > vma->vm_start)
2509 vm_start = 0;
2510 }
2511 return vm_start;
2512 }
2513
2514 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2515 {
2516 unsigned long vm_end = vma->vm_end;
2517
2518 if (vma->vm_flags & VM_GROWSUP) {
2519 vm_end += stack_guard_gap;
2520 if (vm_end < vma->vm_end)
2521 vm_end = -PAGE_SIZE;
2522 }
2523 return vm_end;
2524 }
2525
2526 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2527 {
2528 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2529 }
2530
2531 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2532 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2533 unsigned long vm_start, unsigned long vm_end)
2534 {
2535 struct vm_area_struct *vma = find_vma(mm, vm_start);
2536
2537 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2538 vma = NULL;
2539
2540 return vma;
2541 }
2542
2543 static inline bool range_in_vma(struct vm_area_struct *vma,
2544 unsigned long start, unsigned long end)
2545 {
2546 return (vma && vma->vm_start <= start && end <= vma->vm_end);
2547 }
2548
2549 #ifdef CONFIG_MMU
2550 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2551 void vma_set_page_prot(struct vm_area_struct *vma);
2552 #else
2553 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2554 {
2555 return __pgprot(0);
2556 }
2557 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2558 {
2559 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2560 }
2561 #endif
2562
2563 #ifdef CONFIG_NUMA_BALANCING
2564 unsigned long change_prot_numa(struct vm_area_struct *vma,
2565 unsigned long start, unsigned long end);
2566 #endif
2567
2568 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2569 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2570 unsigned long pfn, unsigned long size, pgprot_t);
2571 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2572 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2573 unsigned long pfn);
2574 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2575 unsigned long pfn, pgprot_t pgprot);
2576 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2577 pfn_t pfn);
2578 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
2579 unsigned long addr, pfn_t pfn);
2580 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2581
2582 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2583 unsigned long addr, struct page *page)
2584 {
2585 int err = vm_insert_page(vma, addr, page);
2586
2587 if (err == -ENOMEM)
2588 return VM_FAULT_OOM;
2589 if (err < 0 && err != -EBUSY)
2590 return VM_FAULT_SIGBUS;
2591
2592 return VM_FAULT_NOPAGE;
2593 }
2594
2595 static inline vm_fault_t vmf_error(int err)
2596 {
2597 if (err == -ENOMEM)
2598 return VM_FAULT_OOM;
2599 return VM_FAULT_SIGBUS;
2600 }
2601
2602 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
2603 unsigned int foll_flags);
2604
2605 #define FOLL_WRITE 0x01 /* check pte is writable */
2606 #define FOLL_TOUCH 0x02 /* mark page accessed */
2607 #define FOLL_GET 0x04 /* do get_page on page */
2608 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2609 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2610 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2611 * and return without waiting upon it */
2612 #define FOLL_POPULATE 0x40 /* fault in page */
2613 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2614 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2615 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2616 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2617 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2618 #define FOLL_MLOCK 0x1000 /* lock present pages */
2619 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2620 #define FOLL_COW 0x4000 /* internal GUP flag */
2621 #define FOLL_ANON 0x8000 /* don't do file mappings */
2622
2623 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
2624 {
2625 if (vm_fault & VM_FAULT_OOM)
2626 return -ENOMEM;
2627 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2628 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2629 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2630 return -EFAULT;
2631 return 0;
2632 }
2633
2634 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2635 void *data);
2636 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2637 unsigned long size, pte_fn_t fn, void *data);
2638
2639
2640 #ifdef CONFIG_PAGE_POISONING
2641 extern bool page_poisoning_enabled(void);
2642 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2643 #else
2644 static inline bool page_poisoning_enabled(void) { return false; }
2645 static inline void kernel_poison_pages(struct page *page, int numpages,
2646 int enable) { }
2647 #endif
2648
2649 #ifdef CONFIG_DEBUG_PAGEALLOC
2650 extern bool _debug_pagealloc_enabled;
2651 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2652
2653 static inline bool debug_pagealloc_enabled(void)
2654 {
2655 return _debug_pagealloc_enabled;
2656 }
2657
2658 static inline void
2659 kernel_map_pages(struct page *page, int numpages, int enable)
2660 {
2661 if (!debug_pagealloc_enabled())
2662 return;
2663
2664 __kernel_map_pages(page, numpages, enable);
2665 }
2666 #ifdef CONFIG_HIBERNATION
2667 extern bool kernel_page_present(struct page *page);
2668 #endif /* CONFIG_HIBERNATION */
2669 #else /* CONFIG_DEBUG_PAGEALLOC */
2670 static inline void
2671 kernel_map_pages(struct page *page, int numpages, int enable) {}
2672 #ifdef CONFIG_HIBERNATION
2673 static inline bool kernel_page_present(struct page *page) { return true; }
2674 #endif /* CONFIG_HIBERNATION */
2675 static inline bool debug_pagealloc_enabled(void)
2676 {
2677 return false;
2678 }
2679 #endif /* CONFIG_DEBUG_PAGEALLOC */
2680
2681 #ifdef __HAVE_ARCH_GATE_AREA
2682 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2683 extern int in_gate_area_no_mm(unsigned long addr);
2684 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2685 #else
2686 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2687 {
2688 return NULL;
2689 }
2690 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2691 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2692 {
2693 return 0;
2694 }
2695 #endif /* __HAVE_ARCH_GATE_AREA */
2696
2697 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2698
2699 #ifdef CONFIG_SYSCTL
2700 extern int sysctl_drop_caches;
2701 int drop_caches_sysctl_handler(struct ctl_table *, int,
2702 void __user *, size_t *, loff_t *);
2703 #endif
2704
2705 void drop_slab(void);
2706 void drop_slab_node(int nid);
2707
2708 #ifndef CONFIG_MMU
2709 #define randomize_va_space 0
2710 #else
2711 extern int randomize_va_space;
2712 #endif
2713
2714 const char * arch_vma_name(struct vm_area_struct *vma);
2715 void print_vma_addr(char *prefix, unsigned long rip);
2716
2717 void *sparse_buffer_alloc(unsigned long size);
2718 struct page *sparse_mem_map_populate(unsigned long pnum, int nid,
2719 struct vmem_altmap *altmap);
2720 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2721 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
2722 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
2723 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2724 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2725 void *vmemmap_alloc_block(unsigned long size, int node);
2726 struct vmem_altmap;
2727 void *vmemmap_alloc_block_buf(unsigned long size, int node);
2728 void *altmap_alloc_block_buf(unsigned long size, struct vmem_altmap *altmap);
2729 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2730 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2731 int node);
2732 int vmemmap_populate(unsigned long start, unsigned long end, int node,
2733 struct vmem_altmap *altmap);
2734 void vmemmap_populate_print_last(void);
2735 #ifdef CONFIG_MEMORY_HOTPLUG
2736 void vmemmap_free(unsigned long start, unsigned long end,
2737 struct vmem_altmap *altmap);
2738 #endif
2739 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2740 unsigned long nr_pages);
2741
2742 enum mf_flags {
2743 MF_COUNT_INCREASED = 1 << 0,
2744 MF_ACTION_REQUIRED = 1 << 1,
2745 MF_MUST_KILL = 1 << 2,
2746 MF_SOFT_OFFLINE = 1 << 3,
2747 };
2748 extern int memory_failure(unsigned long pfn, int flags);
2749 extern void memory_failure_queue(unsigned long pfn, int flags);
2750 extern int unpoison_memory(unsigned long pfn);
2751 extern int get_hwpoison_page(struct page *page);
2752 #define put_hwpoison_page(page) put_page(page)
2753 extern int sysctl_memory_failure_early_kill;
2754 extern int sysctl_memory_failure_recovery;
2755 extern void shake_page(struct page *p, int access);
2756 extern atomic_long_t num_poisoned_pages __read_mostly;
2757 extern int soft_offline_page(struct page *page, int flags);
2758
2759
2760 /*
2761 * Error handlers for various types of pages.
2762 */
2763 enum mf_result {
2764 MF_IGNORED, /* Error: cannot be handled */
2765 MF_FAILED, /* Error: handling failed */
2766 MF_DELAYED, /* Will be handled later */
2767 MF_RECOVERED, /* Successfully recovered */
2768 };
2769
2770 enum mf_action_page_type {
2771 MF_MSG_KERNEL,
2772 MF_MSG_KERNEL_HIGH_ORDER,
2773 MF_MSG_SLAB,
2774 MF_MSG_DIFFERENT_COMPOUND,
2775 MF_MSG_POISONED_HUGE,
2776 MF_MSG_HUGE,
2777 MF_MSG_FREE_HUGE,
2778 MF_MSG_NON_PMD_HUGE,
2779 MF_MSG_UNMAP_FAILED,
2780 MF_MSG_DIRTY_SWAPCACHE,
2781 MF_MSG_CLEAN_SWAPCACHE,
2782 MF_MSG_DIRTY_MLOCKED_LRU,
2783 MF_MSG_CLEAN_MLOCKED_LRU,
2784 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2785 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2786 MF_MSG_DIRTY_LRU,
2787 MF_MSG_CLEAN_LRU,
2788 MF_MSG_TRUNCATED_LRU,
2789 MF_MSG_BUDDY,
2790 MF_MSG_BUDDY_2ND,
2791 MF_MSG_DAX,
2792 MF_MSG_UNKNOWN,
2793 };
2794
2795 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2796 extern void clear_huge_page(struct page *page,
2797 unsigned long addr_hint,
2798 unsigned int pages_per_huge_page);
2799 extern void copy_user_huge_page(struct page *dst, struct page *src,
2800 unsigned long addr_hint,
2801 struct vm_area_struct *vma,
2802 unsigned int pages_per_huge_page);
2803 extern long copy_huge_page_from_user(struct page *dst_page,
2804 const void __user *usr_src,
2805 unsigned int pages_per_huge_page,
2806 bool allow_pagefault);
2807 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2808
2809 extern struct page_ext_operations debug_guardpage_ops;
2810
2811 #ifdef CONFIG_DEBUG_PAGEALLOC
2812 extern unsigned int _debug_guardpage_minorder;
2813 extern bool _debug_guardpage_enabled;
2814
2815 static inline unsigned int debug_guardpage_minorder(void)
2816 {
2817 return _debug_guardpage_minorder;
2818 }
2819
2820 static inline bool debug_guardpage_enabled(void)
2821 {
2822 return _debug_guardpage_enabled;
2823 }
2824
2825 static inline bool page_is_guard(struct page *page)
2826 {
2827 struct page_ext *page_ext;
2828
2829 if (!debug_guardpage_enabled())
2830 return false;
2831
2832 page_ext = lookup_page_ext(page);
2833 if (unlikely(!page_ext))
2834 return false;
2835
2836 return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2837 }
2838 #else
2839 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2840 static inline bool debug_guardpage_enabled(void) { return false; }
2841 static inline bool page_is_guard(struct page *page) { return false; }
2842 #endif /* CONFIG_DEBUG_PAGEALLOC */
2843
2844 #if MAX_NUMNODES > 1
2845 void __init setup_nr_node_ids(void);
2846 #else
2847 static inline void setup_nr_node_ids(void) {}
2848 #endif
2849
2850 #endif /* __KERNEL__ */
2851 #endif /* _LINUX_MM_H */