xref: /linux-6.15/include/linux/mm.h (revision 12bb14aa)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _LINUX_MM_H
3 #define _LINUX_MM_H
4 
5 #include <linux/errno.h>
6 
7 #ifdef __KERNEL__
8 
9 #include <linux/mmdebug.h>
10 #include <linux/gfp.h>
11 #include <linux/bug.h>
12 #include <linux/list.h>
13 #include <linux/mmzone.h>
14 #include <linux/rbtree.h>
15 #include <linux/atomic.h>
16 #include <linux/debug_locks.h>
17 #include <linux/mm_types.h>
18 #include <linux/range.h>
19 #include <linux/pfn.h>
20 #include <linux/percpu-refcount.h>
21 #include <linux/bit_spinlock.h>
22 #include <linux/shrinker.h>
23 #include <linux/resource.h>
24 #include <linux/page_ext.h>
25 #include <linux/err.h>
26 #include <linux/page_ref.h>
27 #include <linux/memremap.h>
28 
29 struct mempolicy;
30 struct anon_vma;
31 struct anon_vma_chain;
32 struct file_ra_state;
33 struct user_struct;
34 struct writeback_control;
35 struct bdi_writeback;
36 
37 void init_mm_internals(void);
38 
39 #ifndef CONFIG_NEED_MULTIPLE_NODES	/* Don't use mapnrs, do it properly */
40 extern unsigned long max_mapnr;
41 
42 static inline void set_max_mapnr(unsigned long limit)
43 {
44 	max_mapnr = limit;
45 }
46 #else
47 static inline void set_max_mapnr(unsigned long limit) { }
48 #endif
49 
50 extern unsigned long totalram_pages;
51 extern void * high_memory;
52 extern int page_cluster;
53 
54 #ifdef CONFIG_SYSCTL
55 extern int sysctl_legacy_va_layout;
56 #else
57 #define sysctl_legacy_va_layout 0
58 #endif
59 
60 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
61 extern const int mmap_rnd_bits_min;
62 extern const int mmap_rnd_bits_max;
63 extern int mmap_rnd_bits __read_mostly;
64 #endif
65 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
66 extern const int mmap_rnd_compat_bits_min;
67 extern const int mmap_rnd_compat_bits_max;
68 extern int mmap_rnd_compat_bits __read_mostly;
69 #endif
70 
71 #include <asm/page.h>
72 #include <asm/pgtable.h>
73 #include <asm/processor.h>
74 
75 #ifndef __pa_symbol
76 #define __pa_symbol(x)  __pa(RELOC_HIDE((unsigned long)(x), 0))
77 #endif
78 
79 #ifndef page_to_virt
80 #define page_to_virt(x)	__va(PFN_PHYS(page_to_pfn(x)))
81 #endif
82 
83 #ifndef lm_alias
84 #define lm_alias(x)	__va(__pa_symbol(x))
85 #endif
86 
87 /*
88  * To prevent common memory management code establishing
89  * a zero page mapping on a read fault.
90  * This macro should be defined within <asm/pgtable.h>.
91  * s390 does this to prevent multiplexing of hardware bits
92  * related to the physical page in case of virtualization.
93  */
94 #ifndef mm_forbids_zeropage
95 #define mm_forbids_zeropage(X)	(0)
96 #endif
97 
98 /*
99  * On some architectures it is expensive to call memset() for small sizes.
100  * Those architectures should provide their own implementation of "struct page"
101  * zeroing by defining this macro in <asm/pgtable.h>.
102  */
103 #ifndef mm_zero_struct_page
104 #define mm_zero_struct_page(pp)  ((void)memset((pp), 0, sizeof(struct page)))
105 #endif
106 
107 /*
108  * Default maximum number of active map areas, this limits the number of vmas
109  * per mm struct. Users can overwrite this number by sysctl but there is a
110  * problem.
111  *
112  * When a program's coredump is generated as ELF format, a section is created
113  * per a vma. In ELF, the number of sections is represented in unsigned short.
114  * This means the number of sections should be smaller than 65535 at coredump.
115  * Because the kernel adds some informative sections to a image of program at
116  * generating coredump, we need some margin. The number of extra sections is
117  * 1-3 now and depends on arch. We use "5" as safe margin, here.
118  *
119  * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
120  * not a hard limit any more. Although some userspace tools can be surprised by
121  * that.
122  */
123 #define MAPCOUNT_ELF_CORE_MARGIN	(5)
124 #define DEFAULT_MAX_MAP_COUNT	(USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
125 
126 extern int sysctl_max_map_count;
127 
128 extern unsigned long sysctl_user_reserve_kbytes;
129 extern unsigned long sysctl_admin_reserve_kbytes;
130 
131 extern int sysctl_overcommit_memory;
132 extern int sysctl_overcommit_ratio;
133 extern unsigned long sysctl_overcommit_kbytes;
134 
135 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
136 				    size_t *, loff_t *);
137 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
138 				    size_t *, loff_t *);
139 
140 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
141 
142 /* to align the pointer to the (next) page boundary */
143 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
144 
145 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
146 #define PAGE_ALIGNED(addr)	IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
147 
148 /*
149  * Linux kernel virtual memory manager primitives.
150  * The idea being to have a "virtual" mm in the same way
151  * we have a virtual fs - giving a cleaner interface to the
152  * mm details, and allowing different kinds of memory mappings
153  * (from shared memory to executable loading to arbitrary
154  * mmap() functions).
155  */
156 
157 extern struct kmem_cache *vm_area_cachep;
158 
159 #ifndef CONFIG_MMU
160 extern struct rb_root nommu_region_tree;
161 extern struct rw_semaphore nommu_region_sem;
162 
163 extern unsigned int kobjsize(const void *objp);
164 #endif
165 
166 /*
167  * vm_flags in vm_area_struct, see mm_types.h.
168  * When changing, update also include/trace/events/mmflags.h
169  */
170 #define VM_NONE		0x00000000
171 
172 #define VM_READ		0x00000001	/* currently active flags */
173 #define VM_WRITE	0x00000002
174 #define VM_EXEC		0x00000004
175 #define VM_SHARED	0x00000008
176 
177 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
178 #define VM_MAYREAD	0x00000010	/* limits for mprotect() etc */
179 #define VM_MAYWRITE	0x00000020
180 #define VM_MAYEXEC	0x00000040
181 #define VM_MAYSHARE	0x00000080
182 
183 #define VM_GROWSDOWN	0x00000100	/* general info on the segment */
184 #define VM_UFFD_MISSING	0x00000200	/* missing pages tracking */
185 #define VM_PFNMAP	0x00000400	/* Page-ranges managed without "struct page", just pure PFN */
186 #define VM_DENYWRITE	0x00000800	/* ETXTBSY on write attempts.. */
187 #define VM_UFFD_WP	0x00001000	/* wrprotect pages tracking */
188 
189 #define VM_LOCKED	0x00002000
190 #define VM_IO           0x00004000	/* Memory mapped I/O or similar */
191 
192 					/* Used by sys_madvise() */
193 #define VM_SEQ_READ	0x00008000	/* App will access data sequentially */
194 #define VM_RAND_READ	0x00010000	/* App will not benefit from clustered reads */
195 
196 #define VM_DONTCOPY	0x00020000      /* Do not copy this vma on fork */
197 #define VM_DONTEXPAND	0x00040000	/* Cannot expand with mremap() */
198 #define VM_LOCKONFAULT	0x00080000	/* Lock the pages covered when they are faulted in */
199 #define VM_ACCOUNT	0x00100000	/* Is a VM accounted object */
200 #define VM_NORESERVE	0x00200000	/* should the VM suppress accounting */
201 #define VM_HUGETLB	0x00400000	/* Huge TLB Page VM */
202 #define VM_SYNC		0x00800000	/* Synchronous page faults */
203 #define VM_ARCH_1	0x01000000	/* Architecture-specific flag */
204 #define VM_WIPEONFORK	0x02000000	/* Wipe VMA contents in child. */
205 #define VM_DONTDUMP	0x04000000	/* Do not include in the core dump */
206 
207 #ifdef CONFIG_MEM_SOFT_DIRTY
208 # define VM_SOFTDIRTY	0x08000000	/* Not soft dirty clean area */
209 #else
210 # define VM_SOFTDIRTY	0
211 #endif
212 
213 #define VM_MIXEDMAP	0x10000000	/* Can contain "struct page" and pure PFN pages */
214 #define VM_HUGEPAGE	0x20000000	/* MADV_HUGEPAGE marked this vma */
215 #define VM_NOHUGEPAGE	0x40000000	/* MADV_NOHUGEPAGE marked this vma */
216 #define VM_MERGEABLE	0x80000000	/* KSM may merge identical pages */
217 
218 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
219 #define VM_HIGH_ARCH_BIT_0	32	/* bit only usable on 64-bit architectures */
220 #define VM_HIGH_ARCH_BIT_1	33	/* bit only usable on 64-bit architectures */
221 #define VM_HIGH_ARCH_BIT_2	34	/* bit only usable on 64-bit architectures */
222 #define VM_HIGH_ARCH_BIT_3	35	/* bit only usable on 64-bit architectures */
223 #define VM_HIGH_ARCH_BIT_4	36	/* bit only usable on 64-bit architectures */
224 #define VM_HIGH_ARCH_0	BIT(VM_HIGH_ARCH_BIT_0)
225 #define VM_HIGH_ARCH_1	BIT(VM_HIGH_ARCH_BIT_1)
226 #define VM_HIGH_ARCH_2	BIT(VM_HIGH_ARCH_BIT_2)
227 #define VM_HIGH_ARCH_3	BIT(VM_HIGH_ARCH_BIT_3)
228 #define VM_HIGH_ARCH_4	BIT(VM_HIGH_ARCH_BIT_4)
229 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
230 
231 #if defined(CONFIG_X86)
232 # define VM_PAT		VM_ARCH_1	/* PAT reserves whole VMA at once (x86) */
233 #if defined (CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS)
234 # define VM_PKEY_SHIFT	VM_HIGH_ARCH_BIT_0
235 # define VM_PKEY_BIT0	VM_HIGH_ARCH_0	/* A protection key is a 4-bit value */
236 # define VM_PKEY_BIT1	VM_HIGH_ARCH_1
237 # define VM_PKEY_BIT2	VM_HIGH_ARCH_2
238 # define VM_PKEY_BIT3	VM_HIGH_ARCH_3
239 #endif
240 #elif defined(CONFIG_PPC)
241 # define VM_SAO		VM_ARCH_1	/* Strong Access Ordering (powerpc) */
242 #elif defined(CONFIG_PARISC)
243 # define VM_GROWSUP	VM_ARCH_1
244 #elif defined(CONFIG_METAG)
245 # define VM_GROWSUP	VM_ARCH_1
246 #elif defined(CONFIG_IA64)
247 # define VM_GROWSUP	VM_ARCH_1
248 #elif !defined(CONFIG_MMU)
249 # define VM_MAPPED_COPY	VM_ARCH_1	/* T if mapped copy of data (nommu mmap) */
250 #endif
251 
252 #if defined(CONFIG_X86_INTEL_MPX)
253 /* MPX specific bounds table or bounds directory */
254 # define VM_MPX		VM_HIGH_ARCH_4
255 #else
256 # define VM_MPX		VM_NONE
257 #endif
258 
259 #ifndef VM_GROWSUP
260 # define VM_GROWSUP	VM_NONE
261 #endif
262 
263 /* Bits set in the VMA until the stack is in its final location */
264 #define VM_STACK_INCOMPLETE_SETUP	(VM_RAND_READ | VM_SEQ_READ)
265 
266 #ifndef VM_STACK_DEFAULT_FLAGS		/* arch can override this */
267 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
268 #endif
269 
270 #ifdef CONFIG_STACK_GROWSUP
271 #define VM_STACK	VM_GROWSUP
272 #else
273 #define VM_STACK	VM_GROWSDOWN
274 #endif
275 
276 #define VM_STACK_FLAGS	(VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
277 
278 /*
279  * Special vmas that are non-mergable, non-mlock()able.
280  * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
281  */
282 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
283 
284 /* This mask defines which mm->def_flags a process can inherit its parent */
285 #define VM_INIT_DEF_MASK	VM_NOHUGEPAGE
286 
287 /* This mask is used to clear all the VMA flags used by mlock */
288 #define VM_LOCKED_CLEAR_MASK	(~(VM_LOCKED | VM_LOCKONFAULT))
289 
290 /*
291  * mapping from the currently active vm_flags protection bits (the
292  * low four bits) to a page protection mask..
293  */
294 extern pgprot_t protection_map[16];
295 
296 #define FAULT_FLAG_WRITE	0x01	/* Fault was a write access */
297 #define FAULT_FLAG_MKWRITE	0x02	/* Fault was mkwrite of existing pte */
298 #define FAULT_FLAG_ALLOW_RETRY	0x04	/* Retry fault if blocking */
299 #define FAULT_FLAG_RETRY_NOWAIT	0x08	/* Don't drop mmap_sem and wait when retrying */
300 #define FAULT_FLAG_KILLABLE	0x10	/* The fault task is in SIGKILL killable region */
301 #define FAULT_FLAG_TRIED	0x20	/* Second try */
302 #define FAULT_FLAG_USER		0x40	/* The fault originated in userspace */
303 #define FAULT_FLAG_REMOTE	0x80	/* faulting for non current tsk/mm */
304 #define FAULT_FLAG_INSTRUCTION  0x100	/* The fault was during an instruction fetch */
305 
306 #define FAULT_FLAG_TRACE \
307 	{ FAULT_FLAG_WRITE,		"WRITE" }, \
308 	{ FAULT_FLAG_MKWRITE,		"MKWRITE" }, \
309 	{ FAULT_FLAG_ALLOW_RETRY,	"ALLOW_RETRY" }, \
310 	{ FAULT_FLAG_RETRY_NOWAIT,	"RETRY_NOWAIT" }, \
311 	{ FAULT_FLAG_KILLABLE,		"KILLABLE" }, \
312 	{ FAULT_FLAG_TRIED,		"TRIED" }, \
313 	{ FAULT_FLAG_USER,		"USER" }, \
314 	{ FAULT_FLAG_REMOTE,		"REMOTE" }, \
315 	{ FAULT_FLAG_INSTRUCTION,	"INSTRUCTION" }
316 
317 /*
318  * vm_fault is filled by the the pagefault handler and passed to the vma's
319  * ->fault function. The vma's ->fault is responsible for returning a bitmask
320  * of VM_FAULT_xxx flags that give details about how the fault was handled.
321  *
322  * MM layer fills up gfp_mask for page allocations but fault handler might
323  * alter it if its implementation requires a different allocation context.
324  *
325  * pgoff should be used in favour of virtual_address, if possible.
326  */
327 struct vm_fault {
328 	struct vm_area_struct *vma;	/* Target VMA */
329 	unsigned int flags;		/* FAULT_FLAG_xxx flags */
330 	gfp_t gfp_mask;			/* gfp mask to be used for allocations */
331 	pgoff_t pgoff;			/* Logical page offset based on vma */
332 	unsigned long address;		/* Faulting virtual address */
333 	pmd_t *pmd;			/* Pointer to pmd entry matching
334 					 * the 'address' */
335 	pud_t *pud;			/* Pointer to pud entry matching
336 					 * the 'address'
337 					 */
338 	pte_t orig_pte;			/* Value of PTE at the time of fault */
339 
340 	struct page *cow_page;		/* Page handler may use for COW fault */
341 	struct mem_cgroup *memcg;	/* Cgroup cow_page belongs to */
342 	struct page *page;		/* ->fault handlers should return a
343 					 * page here, unless VM_FAULT_NOPAGE
344 					 * is set (which is also implied by
345 					 * VM_FAULT_ERROR).
346 					 */
347 	/* These three entries are valid only while holding ptl lock */
348 	pte_t *pte;			/* Pointer to pte entry matching
349 					 * the 'address'. NULL if the page
350 					 * table hasn't been allocated.
351 					 */
352 	spinlock_t *ptl;		/* Page table lock.
353 					 * Protects pte page table if 'pte'
354 					 * is not NULL, otherwise pmd.
355 					 */
356 	pgtable_t prealloc_pte;		/* Pre-allocated pte page table.
357 					 * vm_ops->map_pages() calls
358 					 * alloc_set_pte() from atomic context.
359 					 * do_fault_around() pre-allocates
360 					 * page table to avoid allocation from
361 					 * atomic context.
362 					 */
363 };
364 
365 /* page entry size for vm->huge_fault() */
366 enum page_entry_size {
367 	PE_SIZE_PTE = 0,
368 	PE_SIZE_PMD,
369 	PE_SIZE_PUD,
370 };
371 
372 /*
373  * These are the virtual MM functions - opening of an area, closing and
374  * unmapping it (needed to keep files on disk up-to-date etc), pointer
375  * to the functions called when a no-page or a wp-page exception occurs.
376  */
377 struct vm_operations_struct {
378 	void (*open)(struct vm_area_struct * area);
379 	void (*close)(struct vm_area_struct * area);
380 	int (*mremap)(struct vm_area_struct * area);
381 	int (*fault)(struct vm_fault *vmf);
382 	int (*huge_fault)(struct vm_fault *vmf, enum page_entry_size pe_size);
383 	void (*map_pages)(struct vm_fault *vmf,
384 			pgoff_t start_pgoff, pgoff_t end_pgoff);
385 
386 	/* notification that a previously read-only page is about to become
387 	 * writable, if an error is returned it will cause a SIGBUS */
388 	int (*page_mkwrite)(struct vm_fault *vmf);
389 
390 	/* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
391 	int (*pfn_mkwrite)(struct vm_fault *vmf);
392 
393 	/* called by access_process_vm when get_user_pages() fails, typically
394 	 * for use by special VMAs that can switch between memory and hardware
395 	 */
396 	int (*access)(struct vm_area_struct *vma, unsigned long addr,
397 		      void *buf, int len, int write);
398 
399 	/* Called by the /proc/PID/maps code to ask the vma whether it
400 	 * has a special name.  Returning non-NULL will also cause this
401 	 * vma to be dumped unconditionally. */
402 	const char *(*name)(struct vm_area_struct *vma);
403 
404 #ifdef CONFIG_NUMA
405 	/*
406 	 * set_policy() op must add a reference to any non-NULL @new mempolicy
407 	 * to hold the policy upon return.  Caller should pass NULL @new to
408 	 * remove a policy and fall back to surrounding context--i.e. do not
409 	 * install a MPOL_DEFAULT policy, nor the task or system default
410 	 * mempolicy.
411 	 */
412 	int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
413 
414 	/*
415 	 * get_policy() op must add reference [mpol_get()] to any policy at
416 	 * (vma,addr) marked as MPOL_SHARED.  The shared policy infrastructure
417 	 * in mm/mempolicy.c will do this automatically.
418 	 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
419 	 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
420 	 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
421 	 * must return NULL--i.e., do not "fallback" to task or system default
422 	 * policy.
423 	 */
424 	struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
425 					unsigned long addr);
426 #endif
427 	/*
428 	 * Called by vm_normal_page() for special PTEs to find the
429 	 * page for @addr.  This is useful if the default behavior
430 	 * (using pte_page()) would not find the correct page.
431 	 */
432 	struct page *(*find_special_page)(struct vm_area_struct *vma,
433 					  unsigned long addr);
434 };
435 
436 struct mmu_gather;
437 struct inode;
438 
439 #define page_private(page)		((page)->private)
440 #define set_page_private(page, v)	((page)->private = (v))
441 
442 #if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
443 static inline int pmd_devmap(pmd_t pmd)
444 {
445 	return 0;
446 }
447 static inline int pud_devmap(pud_t pud)
448 {
449 	return 0;
450 }
451 static inline int pgd_devmap(pgd_t pgd)
452 {
453 	return 0;
454 }
455 #endif
456 
457 /*
458  * FIXME: take this include out, include page-flags.h in
459  * files which need it (119 of them)
460  */
461 #include <linux/page-flags.h>
462 #include <linux/huge_mm.h>
463 
464 /*
465  * Methods to modify the page usage count.
466  *
467  * What counts for a page usage:
468  * - cache mapping   (page->mapping)
469  * - private data    (page->private)
470  * - page mapped in a task's page tables, each mapping
471  *   is counted separately
472  *
473  * Also, many kernel routines increase the page count before a critical
474  * routine so they can be sure the page doesn't go away from under them.
475  */
476 
477 /*
478  * Drop a ref, return true if the refcount fell to zero (the page has no users)
479  */
480 static inline int put_page_testzero(struct page *page)
481 {
482 	VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
483 	return page_ref_dec_and_test(page);
484 }
485 
486 /*
487  * Try to grab a ref unless the page has a refcount of zero, return false if
488  * that is the case.
489  * This can be called when MMU is off so it must not access
490  * any of the virtual mappings.
491  */
492 static inline int get_page_unless_zero(struct page *page)
493 {
494 	return page_ref_add_unless(page, 1, 0);
495 }
496 
497 extern int page_is_ram(unsigned long pfn);
498 
499 enum {
500 	REGION_INTERSECTS,
501 	REGION_DISJOINT,
502 	REGION_MIXED,
503 };
504 
505 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
506 		      unsigned long desc);
507 
508 /* Support for virtually mapped pages */
509 struct page *vmalloc_to_page(const void *addr);
510 unsigned long vmalloc_to_pfn(const void *addr);
511 
512 /*
513  * Determine if an address is within the vmalloc range
514  *
515  * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
516  * is no special casing required.
517  */
518 static inline bool is_vmalloc_addr(const void *x)
519 {
520 #ifdef CONFIG_MMU
521 	unsigned long addr = (unsigned long)x;
522 
523 	return addr >= VMALLOC_START && addr < VMALLOC_END;
524 #else
525 	return false;
526 #endif
527 }
528 #ifdef CONFIG_MMU
529 extern int is_vmalloc_or_module_addr(const void *x);
530 #else
531 static inline int is_vmalloc_or_module_addr(const void *x)
532 {
533 	return 0;
534 }
535 #endif
536 
537 extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
538 static inline void *kvmalloc(size_t size, gfp_t flags)
539 {
540 	return kvmalloc_node(size, flags, NUMA_NO_NODE);
541 }
542 static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
543 {
544 	return kvmalloc_node(size, flags | __GFP_ZERO, node);
545 }
546 static inline void *kvzalloc(size_t size, gfp_t flags)
547 {
548 	return kvmalloc(size, flags | __GFP_ZERO);
549 }
550 
551 static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
552 {
553 	if (size != 0 && n > SIZE_MAX / size)
554 		return NULL;
555 
556 	return kvmalloc(n * size, flags);
557 }
558 
559 extern void kvfree(const void *addr);
560 
561 static inline atomic_t *compound_mapcount_ptr(struct page *page)
562 {
563 	return &page[1].compound_mapcount;
564 }
565 
566 static inline int compound_mapcount(struct page *page)
567 {
568 	VM_BUG_ON_PAGE(!PageCompound(page), page);
569 	page = compound_head(page);
570 	return atomic_read(compound_mapcount_ptr(page)) + 1;
571 }
572 
573 /*
574  * The atomic page->_mapcount, starts from -1: so that transitions
575  * both from it and to it can be tracked, using atomic_inc_and_test
576  * and atomic_add_negative(-1).
577  */
578 static inline void page_mapcount_reset(struct page *page)
579 {
580 	atomic_set(&(page)->_mapcount, -1);
581 }
582 
583 int __page_mapcount(struct page *page);
584 
585 static inline int page_mapcount(struct page *page)
586 {
587 	VM_BUG_ON_PAGE(PageSlab(page), page);
588 
589 	if (unlikely(PageCompound(page)))
590 		return __page_mapcount(page);
591 	return atomic_read(&page->_mapcount) + 1;
592 }
593 
594 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
595 int total_mapcount(struct page *page);
596 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
597 #else
598 static inline int total_mapcount(struct page *page)
599 {
600 	return page_mapcount(page);
601 }
602 static inline int page_trans_huge_mapcount(struct page *page,
603 					   int *total_mapcount)
604 {
605 	int mapcount = page_mapcount(page);
606 	if (total_mapcount)
607 		*total_mapcount = mapcount;
608 	return mapcount;
609 }
610 #endif
611 
612 static inline struct page *virt_to_head_page(const void *x)
613 {
614 	struct page *page = virt_to_page(x);
615 
616 	return compound_head(page);
617 }
618 
619 void __put_page(struct page *page);
620 
621 void put_pages_list(struct list_head *pages);
622 
623 void split_page(struct page *page, unsigned int order);
624 
625 /*
626  * Compound pages have a destructor function.  Provide a
627  * prototype for that function and accessor functions.
628  * These are _only_ valid on the head of a compound page.
629  */
630 typedef void compound_page_dtor(struct page *);
631 
632 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
633 enum compound_dtor_id {
634 	NULL_COMPOUND_DTOR,
635 	COMPOUND_PAGE_DTOR,
636 #ifdef CONFIG_HUGETLB_PAGE
637 	HUGETLB_PAGE_DTOR,
638 #endif
639 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
640 	TRANSHUGE_PAGE_DTOR,
641 #endif
642 	NR_COMPOUND_DTORS,
643 };
644 extern compound_page_dtor * const compound_page_dtors[];
645 
646 static inline void set_compound_page_dtor(struct page *page,
647 		enum compound_dtor_id compound_dtor)
648 {
649 	VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
650 	page[1].compound_dtor = compound_dtor;
651 }
652 
653 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
654 {
655 	VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
656 	return compound_page_dtors[page[1].compound_dtor];
657 }
658 
659 static inline unsigned int compound_order(struct page *page)
660 {
661 	if (!PageHead(page))
662 		return 0;
663 	return page[1].compound_order;
664 }
665 
666 static inline void set_compound_order(struct page *page, unsigned int order)
667 {
668 	page[1].compound_order = order;
669 }
670 
671 void free_compound_page(struct page *page);
672 
673 #ifdef CONFIG_MMU
674 /*
675  * Do pte_mkwrite, but only if the vma says VM_WRITE.  We do this when
676  * servicing faults for write access.  In the normal case, do always want
677  * pte_mkwrite.  But get_user_pages can cause write faults for mappings
678  * that do not have writing enabled, when used by access_process_vm.
679  */
680 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
681 {
682 	if (likely(vma->vm_flags & VM_WRITE))
683 		pte = pte_mkwrite(pte);
684 	return pte;
685 }
686 
687 int alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
688 		struct page *page);
689 int finish_fault(struct vm_fault *vmf);
690 int finish_mkwrite_fault(struct vm_fault *vmf);
691 #endif
692 
693 /*
694  * Multiple processes may "see" the same page. E.g. for untouched
695  * mappings of /dev/null, all processes see the same page full of
696  * zeroes, and text pages of executables and shared libraries have
697  * only one copy in memory, at most, normally.
698  *
699  * For the non-reserved pages, page_count(page) denotes a reference count.
700  *   page_count() == 0 means the page is free. page->lru is then used for
701  *   freelist management in the buddy allocator.
702  *   page_count() > 0  means the page has been allocated.
703  *
704  * Pages are allocated by the slab allocator in order to provide memory
705  * to kmalloc and kmem_cache_alloc. In this case, the management of the
706  * page, and the fields in 'struct page' are the responsibility of mm/slab.c
707  * unless a particular usage is carefully commented. (the responsibility of
708  * freeing the kmalloc memory is the caller's, of course).
709  *
710  * A page may be used by anyone else who does a __get_free_page().
711  * In this case, page_count still tracks the references, and should only
712  * be used through the normal accessor functions. The top bits of page->flags
713  * and page->virtual store page management information, but all other fields
714  * are unused and could be used privately, carefully. The management of this
715  * page is the responsibility of the one who allocated it, and those who have
716  * subsequently been given references to it.
717  *
718  * The other pages (we may call them "pagecache pages") are completely
719  * managed by the Linux memory manager: I/O, buffers, swapping etc.
720  * The following discussion applies only to them.
721  *
722  * A pagecache page contains an opaque `private' member, which belongs to the
723  * page's address_space. Usually, this is the address of a circular list of
724  * the page's disk buffers. PG_private must be set to tell the VM to call
725  * into the filesystem to release these pages.
726  *
727  * A page may belong to an inode's memory mapping. In this case, page->mapping
728  * is the pointer to the inode, and page->index is the file offset of the page,
729  * in units of PAGE_SIZE.
730  *
731  * If pagecache pages are not associated with an inode, they are said to be
732  * anonymous pages. These may become associated with the swapcache, and in that
733  * case PG_swapcache is set, and page->private is an offset into the swapcache.
734  *
735  * In either case (swapcache or inode backed), the pagecache itself holds one
736  * reference to the page. Setting PG_private should also increment the
737  * refcount. The each user mapping also has a reference to the page.
738  *
739  * The pagecache pages are stored in a per-mapping radix tree, which is
740  * rooted at mapping->page_tree, and indexed by offset.
741  * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
742  * lists, we instead now tag pages as dirty/writeback in the radix tree.
743  *
744  * All pagecache pages may be subject to I/O:
745  * - inode pages may need to be read from disk,
746  * - inode pages which have been modified and are MAP_SHARED may need
747  *   to be written back to the inode on disk,
748  * - anonymous pages (including MAP_PRIVATE file mappings) which have been
749  *   modified may need to be swapped out to swap space and (later) to be read
750  *   back into memory.
751  */
752 
753 /*
754  * The zone field is never updated after free_area_init_core()
755  * sets it, so none of the operations on it need to be atomic.
756  */
757 
758 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
759 #define SECTIONS_PGOFF		((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
760 #define NODES_PGOFF		(SECTIONS_PGOFF - NODES_WIDTH)
761 #define ZONES_PGOFF		(NODES_PGOFF - ZONES_WIDTH)
762 #define LAST_CPUPID_PGOFF	(ZONES_PGOFF - LAST_CPUPID_WIDTH)
763 
764 /*
765  * Define the bit shifts to access each section.  For non-existent
766  * sections we define the shift as 0; that plus a 0 mask ensures
767  * the compiler will optimise away reference to them.
768  */
769 #define SECTIONS_PGSHIFT	(SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
770 #define NODES_PGSHIFT		(NODES_PGOFF * (NODES_WIDTH != 0))
771 #define ZONES_PGSHIFT		(ZONES_PGOFF * (ZONES_WIDTH != 0))
772 #define LAST_CPUPID_PGSHIFT	(LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
773 
774 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
775 #ifdef NODE_NOT_IN_PAGE_FLAGS
776 #define ZONEID_SHIFT		(SECTIONS_SHIFT + ZONES_SHIFT)
777 #define ZONEID_PGOFF		((SECTIONS_PGOFF < ZONES_PGOFF)? \
778 						SECTIONS_PGOFF : ZONES_PGOFF)
779 #else
780 #define ZONEID_SHIFT		(NODES_SHIFT + ZONES_SHIFT)
781 #define ZONEID_PGOFF		((NODES_PGOFF < ZONES_PGOFF)? \
782 						NODES_PGOFF : ZONES_PGOFF)
783 #endif
784 
785 #define ZONEID_PGSHIFT		(ZONEID_PGOFF * (ZONEID_SHIFT != 0))
786 
787 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
788 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
789 #endif
790 
791 #define ZONES_MASK		((1UL << ZONES_WIDTH) - 1)
792 #define NODES_MASK		((1UL << NODES_WIDTH) - 1)
793 #define SECTIONS_MASK		((1UL << SECTIONS_WIDTH) - 1)
794 #define LAST_CPUPID_MASK	((1UL << LAST_CPUPID_SHIFT) - 1)
795 #define ZONEID_MASK		((1UL << ZONEID_SHIFT) - 1)
796 
797 static inline enum zone_type page_zonenum(const struct page *page)
798 {
799 	return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
800 }
801 
802 #ifdef CONFIG_ZONE_DEVICE
803 static inline bool is_zone_device_page(const struct page *page)
804 {
805 	return page_zonenum(page) == ZONE_DEVICE;
806 }
807 #else
808 static inline bool is_zone_device_page(const struct page *page)
809 {
810 	return false;
811 }
812 #endif
813 
814 #if defined(CONFIG_DEVICE_PRIVATE) || defined(CONFIG_DEVICE_PUBLIC)
815 void put_zone_device_private_or_public_page(struct page *page);
816 DECLARE_STATIC_KEY_FALSE(device_private_key);
817 #define IS_HMM_ENABLED static_branch_unlikely(&device_private_key)
818 static inline bool is_device_private_page(const struct page *page);
819 static inline bool is_device_public_page(const struct page *page);
820 #else /* CONFIG_DEVICE_PRIVATE || CONFIG_DEVICE_PUBLIC */
821 static inline void put_zone_device_private_or_public_page(struct page *page)
822 {
823 }
824 #define IS_HMM_ENABLED 0
825 static inline bool is_device_private_page(const struct page *page)
826 {
827 	return false;
828 }
829 static inline bool is_device_public_page(const struct page *page)
830 {
831 	return false;
832 }
833 #endif /* CONFIG_DEVICE_PRIVATE || CONFIG_DEVICE_PUBLIC */
834 
835 
836 static inline void get_page(struct page *page)
837 {
838 	page = compound_head(page);
839 	/*
840 	 * Getting a normal page or the head of a compound page
841 	 * requires to already have an elevated page->_refcount.
842 	 */
843 	VM_BUG_ON_PAGE(page_ref_count(page) <= 0, page);
844 	page_ref_inc(page);
845 }
846 
847 static inline void put_page(struct page *page)
848 {
849 	page = compound_head(page);
850 
851 	/*
852 	 * For private device pages we need to catch refcount transition from
853 	 * 2 to 1, when refcount reach one it means the private device page is
854 	 * free and we need to inform the device driver through callback. See
855 	 * include/linux/memremap.h and HMM for details.
856 	 */
857 	if (IS_HMM_ENABLED && unlikely(is_device_private_page(page) ||
858 	    unlikely(is_device_public_page(page)))) {
859 		put_zone_device_private_or_public_page(page);
860 		return;
861 	}
862 
863 	if (put_page_testzero(page))
864 		__put_page(page);
865 }
866 
867 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
868 #define SECTION_IN_PAGE_FLAGS
869 #endif
870 
871 /*
872  * The identification function is mainly used by the buddy allocator for
873  * determining if two pages could be buddies. We are not really identifying
874  * the zone since we could be using the section number id if we do not have
875  * node id available in page flags.
876  * We only guarantee that it will return the same value for two combinable
877  * pages in a zone.
878  */
879 static inline int page_zone_id(struct page *page)
880 {
881 	return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
882 }
883 
884 static inline int zone_to_nid(struct zone *zone)
885 {
886 #ifdef CONFIG_NUMA
887 	return zone->node;
888 #else
889 	return 0;
890 #endif
891 }
892 
893 #ifdef NODE_NOT_IN_PAGE_FLAGS
894 extern int page_to_nid(const struct page *page);
895 #else
896 static inline int page_to_nid(const struct page *page)
897 {
898 	return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
899 }
900 #endif
901 
902 #ifdef CONFIG_NUMA_BALANCING
903 static inline int cpu_pid_to_cpupid(int cpu, int pid)
904 {
905 	return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
906 }
907 
908 static inline int cpupid_to_pid(int cpupid)
909 {
910 	return cpupid & LAST__PID_MASK;
911 }
912 
913 static inline int cpupid_to_cpu(int cpupid)
914 {
915 	return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
916 }
917 
918 static inline int cpupid_to_nid(int cpupid)
919 {
920 	return cpu_to_node(cpupid_to_cpu(cpupid));
921 }
922 
923 static inline bool cpupid_pid_unset(int cpupid)
924 {
925 	return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
926 }
927 
928 static inline bool cpupid_cpu_unset(int cpupid)
929 {
930 	return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
931 }
932 
933 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
934 {
935 	return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
936 }
937 
938 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
939 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
940 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
941 {
942 	return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
943 }
944 
945 static inline int page_cpupid_last(struct page *page)
946 {
947 	return page->_last_cpupid;
948 }
949 static inline void page_cpupid_reset_last(struct page *page)
950 {
951 	page->_last_cpupid = -1 & LAST_CPUPID_MASK;
952 }
953 #else
954 static inline int page_cpupid_last(struct page *page)
955 {
956 	return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
957 }
958 
959 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
960 
961 static inline void page_cpupid_reset_last(struct page *page)
962 {
963 	page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
964 }
965 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
966 #else /* !CONFIG_NUMA_BALANCING */
967 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
968 {
969 	return page_to_nid(page); /* XXX */
970 }
971 
972 static inline int page_cpupid_last(struct page *page)
973 {
974 	return page_to_nid(page); /* XXX */
975 }
976 
977 static inline int cpupid_to_nid(int cpupid)
978 {
979 	return -1;
980 }
981 
982 static inline int cpupid_to_pid(int cpupid)
983 {
984 	return -1;
985 }
986 
987 static inline int cpupid_to_cpu(int cpupid)
988 {
989 	return -1;
990 }
991 
992 static inline int cpu_pid_to_cpupid(int nid, int pid)
993 {
994 	return -1;
995 }
996 
997 static inline bool cpupid_pid_unset(int cpupid)
998 {
999 	return 1;
1000 }
1001 
1002 static inline void page_cpupid_reset_last(struct page *page)
1003 {
1004 }
1005 
1006 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1007 {
1008 	return false;
1009 }
1010 #endif /* CONFIG_NUMA_BALANCING */
1011 
1012 static inline struct zone *page_zone(const struct page *page)
1013 {
1014 	return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1015 }
1016 
1017 static inline pg_data_t *page_pgdat(const struct page *page)
1018 {
1019 	return NODE_DATA(page_to_nid(page));
1020 }
1021 
1022 #ifdef SECTION_IN_PAGE_FLAGS
1023 static inline void set_page_section(struct page *page, unsigned long section)
1024 {
1025 	page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1026 	page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1027 }
1028 
1029 static inline unsigned long page_to_section(const struct page *page)
1030 {
1031 	return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1032 }
1033 #endif
1034 
1035 static inline void set_page_zone(struct page *page, enum zone_type zone)
1036 {
1037 	page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1038 	page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1039 }
1040 
1041 static inline void set_page_node(struct page *page, unsigned long node)
1042 {
1043 	page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1044 	page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1045 }
1046 
1047 static inline void set_page_links(struct page *page, enum zone_type zone,
1048 	unsigned long node, unsigned long pfn)
1049 {
1050 	set_page_zone(page, zone);
1051 	set_page_node(page, node);
1052 #ifdef SECTION_IN_PAGE_FLAGS
1053 	set_page_section(page, pfn_to_section_nr(pfn));
1054 #endif
1055 }
1056 
1057 #ifdef CONFIG_MEMCG
1058 static inline struct mem_cgroup *page_memcg(struct page *page)
1059 {
1060 	return page->mem_cgroup;
1061 }
1062 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1063 {
1064 	WARN_ON_ONCE(!rcu_read_lock_held());
1065 	return READ_ONCE(page->mem_cgroup);
1066 }
1067 #else
1068 static inline struct mem_cgroup *page_memcg(struct page *page)
1069 {
1070 	return NULL;
1071 }
1072 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1073 {
1074 	WARN_ON_ONCE(!rcu_read_lock_held());
1075 	return NULL;
1076 }
1077 #endif
1078 
1079 /*
1080  * Some inline functions in vmstat.h depend on page_zone()
1081  */
1082 #include <linux/vmstat.h>
1083 
1084 static __always_inline void *lowmem_page_address(const struct page *page)
1085 {
1086 	return page_to_virt(page);
1087 }
1088 
1089 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1090 #define HASHED_PAGE_VIRTUAL
1091 #endif
1092 
1093 #if defined(WANT_PAGE_VIRTUAL)
1094 static inline void *page_address(const struct page *page)
1095 {
1096 	return page->virtual;
1097 }
1098 static inline void set_page_address(struct page *page, void *address)
1099 {
1100 	page->virtual = address;
1101 }
1102 #define page_address_init()  do { } while(0)
1103 #endif
1104 
1105 #if defined(HASHED_PAGE_VIRTUAL)
1106 void *page_address(const struct page *page);
1107 void set_page_address(struct page *page, void *virtual);
1108 void page_address_init(void);
1109 #endif
1110 
1111 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1112 #define page_address(page) lowmem_page_address(page)
1113 #define set_page_address(page, address)  do { } while(0)
1114 #define page_address_init()  do { } while(0)
1115 #endif
1116 
1117 extern void *page_rmapping(struct page *page);
1118 extern struct anon_vma *page_anon_vma(struct page *page);
1119 extern struct address_space *page_mapping(struct page *page);
1120 
1121 extern struct address_space *__page_file_mapping(struct page *);
1122 
1123 static inline
1124 struct address_space *page_file_mapping(struct page *page)
1125 {
1126 	if (unlikely(PageSwapCache(page)))
1127 		return __page_file_mapping(page);
1128 
1129 	return page->mapping;
1130 }
1131 
1132 extern pgoff_t __page_file_index(struct page *page);
1133 
1134 /*
1135  * Return the pagecache index of the passed page.  Regular pagecache pages
1136  * use ->index whereas swapcache pages use swp_offset(->private)
1137  */
1138 static inline pgoff_t page_index(struct page *page)
1139 {
1140 	if (unlikely(PageSwapCache(page)))
1141 		return __page_file_index(page);
1142 	return page->index;
1143 }
1144 
1145 bool page_mapped(struct page *page);
1146 struct address_space *page_mapping(struct page *page);
1147 
1148 /*
1149  * Return true only if the page has been allocated with
1150  * ALLOC_NO_WATERMARKS and the low watermark was not
1151  * met implying that the system is under some pressure.
1152  */
1153 static inline bool page_is_pfmemalloc(struct page *page)
1154 {
1155 	/*
1156 	 * Page index cannot be this large so this must be
1157 	 * a pfmemalloc page.
1158 	 */
1159 	return page->index == -1UL;
1160 }
1161 
1162 /*
1163  * Only to be called by the page allocator on a freshly allocated
1164  * page.
1165  */
1166 static inline void set_page_pfmemalloc(struct page *page)
1167 {
1168 	page->index = -1UL;
1169 }
1170 
1171 static inline void clear_page_pfmemalloc(struct page *page)
1172 {
1173 	page->index = 0;
1174 }
1175 
1176 /*
1177  * Different kinds of faults, as returned by handle_mm_fault().
1178  * Used to decide whether a process gets delivered SIGBUS or
1179  * just gets major/minor fault counters bumped up.
1180  */
1181 
1182 #define VM_FAULT_OOM	0x0001
1183 #define VM_FAULT_SIGBUS	0x0002
1184 #define VM_FAULT_MAJOR	0x0004
1185 #define VM_FAULT_WRITE	0x0008	/* Special case for get_user_pages */
1186 #define VM_FAULT_HWPOISON 0x0010	/* Hit poisoned small page */
1187 #define VM_FAULT_HWPOISON_LARGE 0x0020  /* Hit poisoned large page. Index encoded in upper bits */
1188 #define VM_FAULT_SIGSEGV 0x0040
1189 
1190 #define VM_FAULT_NOPAGE	0x0100	/* ->fault installed the pte, not return page */
1191 #define VM_FAULT_LOCKED	0x0200	/* ->fault locked the returned page */
1192 #define VM_FAULT_RETRY	0x0400	/* ->fault blocked, must retry */
1193 #define VM_FAULT_FALLBACK 0x0800	/* huge page fault failed, fall back to small */
1194 #define VM_FAULT_DONE_COW   0x1000	/* ->fault has fully handled COW */
1195 #define VM_FAULT_NEEDDSYNC  0x2000	/* ->fault did not modify page tables
1196 					 * and needs fsync() to complete (for
1197 					 * synchronous page faults in DAX) */
1198 
1199 #define VM_FAULT_ERROR	(VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1200 			 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1201 			 VM_FAULT_FALLBACK)
1202 
1203 #define VM_FAULT_RESULT_TRACE \
1204 	{ VM_FAULT_OOM,			"OOM" }, \
1205 	{ VM_FAULT_SIGBUS,		"SIGBUS" }, \
1206 	{ VM_FAULT_MAJOR,		"MAJOR" }, \
1207 	{ VM_FAULT_WRITE,		"WRITE" }, \
1208 	{ VM_FAULT_HWPOISON,		"HWPOISON" }, \
1209 	{ VM_FAULT_HWPOISON_LARGE,	"HWPOISON_LARGE" }, \
1210 	{ VM_FAULT_SIGSEGV,		"SIGSEGV" }, \
1211 	{ VM_FAULT_NOPAGE,		"NOPAGE" }, \
1212 	{ VM_FAULT_LOCKED,		"LOCKED" }, \
1213 	{ VM_FAULT_RETRY,		"RETRY" }, \
1214 	{ VM_FAULT_FALLBACK,		"FALLBACK" }, \
1215 	{ VM_FAULT_DONE_COW,		"DONE_COW" }, \
1216 	{ VM_FAULT_NEEDDSYNC,		"NEEDDSYNC" }
1217 
1218 /* Encode hstate index for a hwpoisoned large page */
1219 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1220 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1221 
1222 /*
1223  * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1224  */
1225 extern void pagefault_out_of_memory(void);
1226 
1227 #define offset_in_page(p)	((unsigned long)(p) & ~PAGE_MASK)
1228 
1229 /*
1230  * Flags passed to show_mem() and show_free_areas() to suppress output in
1231  * various contexts.
1232  */
1233 #define SHOW_MEM_FILTER_NODES		(0x0001u)	/* disallowed nodes */
1234 
1235 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1236 
1237 extern bool can_do_mlock(void);
1238 extern int user_shm_lock(size_t, struct user_struct *);
1239 extern void user_shm_unlock(size_t, struct user_struct *);
1240 
1241 /*
1242  * Parameter block passed down to zap_pte_range in exceptional cases.
1243  */
1244 struct zap_details {
1245 	struct address_space *check_mapping;	/* Check page->mapping if set */
1246 	pgoff_t	first_index;			/* Lowest page->index to unmap */
1247 	pgoff_t last_index;			/* Highest page->index to unmap */
1248 };
1249 
1250 struct page *_vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1251 			     pte_t pte, bool with_public_device);
1252 #define vm_normal_page(vma, addr, pte) _vm_normal_page(vma, addr, pte, false)
1253 
1254 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1255 				pmd_t pmd);
1256 
1257 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1258 		unsigned long size);
1259 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1260 		unsigned long size);
1261 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1262 		unsigned long start, unsigned long end);
1263 
1264 /**
1265  * mm_walk - callbacks for walk_page_range
1266  * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
1267  *	       this handler should only handle pud_trans_huge() puds.
1268  *	       the pmd_entry or pte_entry callbacks will be used for
1269  *	       regular PUDs.
1270  * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1271  *	       this handler is required to be able to handle
1272  *	       pmd_trans_huge() pmds.  They may simply choose to
1273  *	       split_huge_page() instead of handling it explicitly.
1274  * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1275  * @pte_hole: if set, called for each hole at all levels
1276  * @hugetlb_entry: if set, called for each hugetlb entry
1277  * @test_walk: caller specific callback function to determine whether
1278  *             we walk over the current vma or not. Returning 0
1279  *             value means "do page table walk over the current vma,"
1280  *             and a negative one means "abort current page table walk
1281  *             right now." 1 means "skip the current vma."
1282  * @mm:        mm_struct representing the target process of page table walk
1283  * @vma:       vma currently walked (NULL if walking outside vmas)
1284  * @private:   private data for callbacks' usage
1285  *
1286  * (see the comment on walk_page_range() for more details)
1287  */
1288 struct mm_walk {
1289 	int (*pud_entry)(pud_t *pud, unsigned long addr,
1290 			 unsigned long next, struct mm_walk *walk);
1291 	int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1292 			 unsigned long next, struct mm_walk *walk);
1293 	int (*pte_entry)(pte_t *pte, unsigned long addr,
1294 			 unsigned long next, struct mm_walk *walk);
1295 	int (*pte_hole)(unsigned long addr, unsigned long next,
1296 			struct mm_walk *walk);
1297 	int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1298 			     unsigned long addr, unsigned long next,
1299 			     struct mm_walk *walk);
1300 	int (*test_walk)(unsigned long addr, unsigned long next,
1301 			struct mm_walk *walk);
1302 	struct mm_struct *mm;
1303 	struct vm_area_struct *vma;
1304 	void *private;
1305 };
1306 
1307 int walk_page_range(unsigned long addr, unsigned long end,
1308 		struct mm_walk *walk);
1309 int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1310 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1311 		unsigned long end, unsigned long floor, unsigned long ceiling);
1312 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1313 			struct vm_area_struct *vma);
1314 void unmap_mapping_range(struct address_space *mapping,
1315 		loff_t const holebegin, loff_t const holelen, int even_cows);
1316 int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
1317 			     unsigned long *start, unsigned long *end,
1318 			     pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp);
1319 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1320 	unsigned long *pfn);
1321 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1322 		unsigned int flags, unsigned long *prot, resource_size_t *phys);
1323 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1324 			void *buf, int len, int write);
1325 
1326 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1327 		loff_t const holebegin, loff_t const holelen)
1328 {
1329 	unmap_mapping_range(mapping, holebegin, holelen, 0);
1330 }
1331 
1332 extern void truncate_pagecache(struct inode *inode, loff_t new);
1333 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1334 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1335 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1336 int truncate_inode_page(struct address_space *mapping, struct page *page);
1337 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1338 int invalidate_inode_page(struct page *page);
1339 
1340 #ifdef CONFIG_MMU
1341 extern int handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
1342 		unsigned int flags);
1343 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1344 			    unsigned long address, unsigned int fault_flags,
1345 			    bool *unlocked);
1346 #else
1347 static inline int handle_mm_fault(struct vm_area_struct *vma,
1348 		unsigned long address, unsigned int flags)
1349 {
1350 	/* should never happen if there's no MMU */
1351 	BUG();
1352 	return VM_FAULT_SIGBUS;
1353 }
1354 static inline int fixup_user_fault(struct task_struct *tsk,
1355 		struct mm_struct *mm, unsigned long address,
1356 		unsigned int fault_flags, bool *unlocked)
1357 {
1358 	/* should never happen if there's no MMU */
1359 	BUG();
1360 	return -EFAULT;
1361 }
1362 #endif
1363 
1364 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len,
1365 		unsigned int gup_flags);
1366 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1367 		void *buf, int len, unsigned int gup_flags);
1368 extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1369 		unsigned long addr, void *buf, int len, unsigned int gup_flags);
1370 
1371 long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1372 			    unsigned long start, unsigned long nr_pages,
1373 			    unsigned int gup_flags, struct page **pages,
1374 			    struct vm_area_struct **vmas, int *locked);
1375 long get_user_pages(unsigned long start, unsigned long nr_pages,
1376 			    unsigned int gup_flags, struct page **pages,
1377 			    struct vm_area_struct **vmas);
1378 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1379 		    unsigned int gup_flags, struct page **pages, int *locked);
1380 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1381 		    struct page **pages, unsigned int gup_flags);
1382 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1383 			struct page **pages);
1384 
1385 /* Container for pinned pfns / pages */
1386 struct frame_vector {
1387 	unsigned int nr_allocated;	/* Number of frames we have space for */
1388 	unsigned int nr_frames;	/* Number of frames stored in ptrs array */
1389 	bool got_ref;		/* Did we pin pages by getting page ref? */
1390 	bool is_pfns;		/* Does array contain pages or pfns? */
1391 	void *ptrs[0];		/* Array of pinned pfns / pages. Use
1392 				 * pfns_vector_pages() or pfns_vector_pfns()
1393 				 * for access */
1394 };
1395 
1396 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1397 void frame_vector_destroy(struct frame_vector *vec);
1398 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1399 		     unsigned int gup_flags, struct frame_vector *vec);
1400 void put_vaddr_frames(struct frame_vector *vec);
1401 int frame_vector_to_pages(struct frame_vector *vec);
1402 void frame_vector_to_pfns(struct frame_vector *vec);
1403 
1404 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1405 {
1406 	return vec->nr_frames;
1407 }
1408 
1409 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1410 {
1411 	if (vec->is_pfns) {
1412 		int err = frame_vector_to_pages(vec);
1413 
1414 		if (err)
1415 			return ERR_PTR(err);
1416 	}
1417 	return (struct page **)(vec->ptrs);
1418 }
1419 
1420 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1421 {
1422 	if (!vec->is_pfns)
1423 		frame_vector_to_pfns(vec);
1424 	return (unsigned long *)(vec->ptrs);
1425 }
1426 
1427 struct kvec;
1428 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1429 			struct page **pages);
1430 int get_kernel_page(unsigned long start, int write, struct page **pages);
1431 struct page *get_dump_page(unsigned long addr);
1432 
1433 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1434 extern void do_invalidatepage(struct page *page, unsigned int offset,
1435 			      unsigned int length);
1436 
1437 int __set_page_dirty_nobuffers(struct page *page);
1438 int __set_page_dirty_no_writeback(struct page *page);
1439 int redirty_page_for_writepage(struct writeback_control *wbc,
1440 				struct page *page);
1441 void account_page_dirtied(struct page *page, struct address_space *mapping);
1442 void account_page_cleaned(struct page *page, struct address_space *mapping,
1443 			  struct bdi_writeback *wb);
1444 int set_page_dirty(struct page *page);
1445 int set_page_dirty_lock(struct page *page);
1446 void __cancel_dirty_page(struct page *page);
1447 static inline void cancel_dirty_page(struct page *page)
1448 {
1449 	/* Avoid atomic ops, locking, etc. when not actually needed. */
1450 	if (PageDirty(page))
1451 		__cancel_dirty_page(page);
1452 }
1453 int clear_page_dirty_for_io(struct page *page);
1454 
1455 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1456 
1457 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1458 {
1459 	return !vma->vm_ops;
1460 }
1461 
1462 #ifdef CONFIG_SHMEM
1463 /*
1464  * The vma_is_shmem is not inline because it is used only by slow
1465  * paths in userfault.
1466  */
1467 bool vma_is_shmem(struct vm_area_struct *vma);
1468 #else
1469 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
1470 #endif
1471 
1472 int vma_is_stack_for_current(struct vm_area_struct *vma);
1473 
1474 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1475 		unsigned long old_addr, struct vm_area_struct *new_vma,
1476 		unsigned long new_addr, unsigned long len,
1477 		bool need_rmap_locks);
1478 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1479 			      unsigned long end, pgprot_t newprot,
1480 			      int dirty_accountable, int prot_numa);
1481 extern int mprotect_fixup(struct vm_area_struct *vma,
1482 			  struct vm_area_struct **pprev, unsigned long start,
1483 			  unsigned long end, unsigned long newflags);
1484 
1485 /*
1486  * doesn't attempt to fault and will return short.
1487  */
1488 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1489 			  struct page **pages);
1490 /*
1491  * per-process(per-mm_struct) statistics.
1492  */
1493 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1494 {
1495 	long val = atomic_long_read(&mm->rss_stat.count[member]);
1496 
1497 #ifdef SPLIT_RSS_COUNTING
1498 	/*
1499 	 * counter is updated in asynchronous manner and may go to minus.
1500 	 * But it's never be expected number for users.
1501 	 */
1502 	if (val < 0)
1503 		val = 0;
1504 #endif
1505 	return (unsigned long)val;
1506 }
1507 
1508 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1509 {
1510 	atomic_long_add(value, &mm->rss_stat.count[member]);
1511 }
1512 
1513 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1514 {
1515 	atomic_long_inc(&mm->rss_stat.count[member]);
1516 }
1517 
1518 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1519 {
1520 	atomic_long_dec(&mm->rss_stat.count[member]);
1521 }
1522 
1523 /* Optimized variant when page is already known not to be PageAnon */
1524 static inline int mm_counter_file(struct page *page)
1525 {
1526 	if (PageSwapBacked(page))
1527 		return MM_SHMEMPAGES;
1528 	return MM_FILEPAGES;
1529 }
1530 
1531 static inline int mm_counter(struct page *page)
1532 {
1533 	if (PageAnon(page))
1534 		return MM_ANONPAGES;
1535 	return mm_counter_file(page);
1536 }
1537 
1538 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1539 {
1540 	return get_mm_counter(mm, MM_FILEPAGES) +
1541 		get_mm_counter(mm, MM_ANONPAGES) +
1542 		get_mm_counter(mm, MM_SHMEMPAGES);
1543 }
1544 
1545 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1546 {
1547 	return max(mm->hiwater_rss, get_mm_rss(mm));
1548 }
1549 
1550 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1551 {
1552 	return max(mm->hiwater_vm, mm->total_vm);
1553 }
1554 
1555 static inline void update_hiwater_rss(struct mm_struct *mm)
1556 {
1557 	unsigned long _rss = get_mm_rss(mm);
1558 
1559 	if ((mm)->hiwater_rss < _rss)
1560 		(mm)->hiwater_rss = _rss;
1561 }
1562 
1563 static inline void update_hiwater_vm(struct mm_struct *mm)
1564 {
1565 	if (mm->hiwater_vm < mm->total_vm)
1566 		mm->hiwater_vm = mm->total_vm;
1567 }
1568 
1569 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1570 {
1571 	mm->hiwater_rss = get_mm_rss(mm);
1572 }
1573 
1574 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1575 					 struct mm_struct *mm)
1576 {
1577 	unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1578 
1579 	if (*maxrss < hiwater_rss)
1580 		*maxrss = hiwater_rss;
1581 }
1582 
1583 #if defined(SPLIT_RSS_COUNTING)
1584 void sync_mm_rss(struct mm_struct *mm);
1585 #else
1586 static inline void sync_mm_rss(struct mm_struct *mm)
1587 {
1588 }
1589 #endif
1590 
1591 #ifndef __HAVE_ARCH_PTE_DEVMAP
1592 static inline int pte_devmap(pte_t pte)
1593 {
1594 	return 0;
1595 }
1596 #endif
1597 
1598 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1599 
1600 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1601 			       spinlock_t **ptl);
1602 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1603 				    spinlock_t **ptl)
1604 {
1605 	pte_t *ptep;
1606 	__cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1607 	return ptep;
1608 }
1609 
1610 #ifdef __PAGETABLE_P4D_FOLDED
1611 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1612 						unsigned long address)
1613 {
1614 	return 0;
1615 }
1616 #else
1617 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1618 #endif
1619 
1620 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
1621 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1622 						unsigned long address)
1623 {
1624 	return 0;
1625 }
1626 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
1627 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
1628 
1629 #else
1630 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
1631 
1632 static inline void mm_inc_nr_puds(struct mm_struct *mm)
1633 {
1634 	atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1635 }
1636 
1637 static inline void mm_dec_nr_puds(struct mm_struct *mm)
1638 {
1639 	atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1640 }
1641 #endif
1642 
1643 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1644 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1645 						unsigned long address)
1646 {
1647 	return 0;
1648 }
1649 
1650 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1651 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1652 
1653 #else
1654 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1655 
1656 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1657 {
1658 	atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1659 }
1660 
1661 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1662 {
1663 	atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1664 }
1665 #endif
1666 
1667 #ifdef CONFIG_MMU
1668 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
1669 {
1670 	atomic_long_set(&mm->pgtables_bytes, 0);
1671 }
1672 
1673 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1674 {
1675 	return atomic_long_read(&mm->pgtables_bytes);
1676 }
1677 
1678 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
1679 {
1680 	atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1681 }
1682 
1683 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
1684 {
1685 	atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1686 }
1687 #else
1688 
1689 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
1690 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1691 {
1692 	return 0;
1693 }
1694 
1695 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
1696 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
1697 #endif
1698 
1699 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
1700 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1701 
1702 /*
1703  * The following ifdef needed to get the 4level-fixup.h header to work.
1704  * Remove it when 4level-fixup.h has been removed.
1705  */
1706 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1707 
1708 #ifndef __ARCH_HAS_5LEVEL_HACK
1709 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1710 		unsigned long address)
1711 {
1712 	return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
1713 		NULL : p4d_offset(pgd, address);
1714 }
1715 
1716 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1717 		unsigned long address)
1718 {
1719 	return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
1720 		NULL : pud_offset(p4d, address);
1721 }
1722 #endif /* !__ARCH_HAS_5LEVEL_HACK */
1723 
1724 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1725 {
1726 	return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1727 		NULL: pmd_offset(pud, address);
1728 }
1729 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1730 
1731 #if USE_SPLIT_PTE_PTLOCKS
1732 #if ALLOC_SPLIT_PTLOCKS
1733 void __init ptlock_cache_init(void);
1734 extern bool ptlock_alloc(struct page *page);
1735 extern void ptlock_free(struct page *page);
1736 
1737 static inline spinlock_t *ptlock_ptr(struct page *page)
1738 {
1739 	return page->ptl;
1740 }
1741 #else /* ALLOC_SPLIT_PTLOCKS */
1742 static inline void ptlock_cache_init(void)
1743 {
1744 }
1745 
1746 static inline bool ptlock_alloc(struct page *page)
1747 {
1748 	return true;
1749 }
1750 
1751 static inline void ptlock_free(struct page *page)
1752 {
1753 }
1754 
1755 static inline spinlock_t *ptlock_ptr(struct page *page)
1756 {
1757 	return &page->ptl;
1758 }
1759 #endif /* ALLOC_SPLIT_PTLOCKS */
1760 
1761 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1762 {
1763 	return ptlock_ptr(pmd_page(*pmd));
1764 }
1765 
1766 static inline bool ptlock_init(struct page *page)
1767 {
1768 	/*
1769 	 * prep_new_page() initialize page->private (and therefore page->ptl)
1770 	 * with 0. Make sure nobody took it in use in between.
1771 	 *
1772 	 * It can happen if arch try to use slab for page table allocation:
1773 	 * slab code uses page->slab_cache, which share storage with page->ptl.
1774 	 */
1775 	VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1776 	if (!ptlock_alloc(page))
1777 		return false;
1778 	spin_lock_init(ptlock_ptr(page));
1779 	return true;
1780 }
1781 
1782 /* Reset page->mapping so free_pages_check won't complain. */
1783 static inline void pte_lock_deinit(struct page *page)
1784 {
1785 	page->mapping = NULL;
1786 	ptlock_free(page);
1787 }
1788 
1789 #else	/* !USE_SPLIT_PTE_PTLOCKS */
1790 /*
1791  * We use mm->page_table_lock to guard all pagetable pages of the mm.
1792  */
1793 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1794 {
1795 	return &mm->page_table_lock;
1796 }
1797 static inline void ptlock_cache_init(void) {}
1798 static inline bool ptlock_init(struct page *page) { return true; }
1799 static inline void pte_lock_deinit(struct page *page) {}
1800 #endif /* USE_SPLIT_PTE_PTLOCKS */
1801 
1802 static inline void pgtable_init(void)
1803 {
1804 	ptlock_cache_init();
1805 	pgtable_cache_init();
1806 }
1807 
1808 static inline bool pgtable_page_ctor(struct page *page)
1809 {
1810 	if (!ptlock_init(page))
1811 		return false;
1812 	inc_zone_page_state(page, NR_PAGETABLE);
1813 	return true;
1814 }
1815 
1816 static inline void pgtable_page_dtor(struct page *page)
1817 {
1818 	pte_lock_deinit(page);
1819 	dec_zone_page_state(page, NR_PAGETABLE);
1820 }
1821 
1822 #define pte_offset_map_lock(mm, pmd, address, ptlp)	\
1823 ({							\
1824 	spinlock_t *__ptl = pte_lockptr(mm, pmd);	\
1825 	pte_t *__pte = pte_offset_map(pmd, address);	\
1826 	*(ptlp) = __ptl;				\
1827 	spin_lock(__ptl);				\
1828 	__pte;						\
1829 })
1830 
1831 #define pte_unmap_unlock(pte, ptl)	do {		\
1832 	spin_unlock(ptl);				\
1833 	pte_unmap(pte);					\
1834 } while (0)
1835 
1836 #define pte_alloc(mm, pmd, address)			\
1837 	(unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd, address))
1838 
1839 #define pte_alloc_map(mm, pmd, address)			\
1840 	(pte_alloc(mm, pmd, address) ? NULL : pte_offset_map(pmd, address))
1841 
1842 #define pte_alloc_map_lock(mm, pmd, address, ptlp)	\
1843 	(pte_alloc(mm, pmd, address) ?			\
1844 		 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
1845 
1846 #define pte_alloc_kernel(pmd, address)			\
1847 	((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1848 		NULL: pte_offset_kernel(pmd, address))
1849 
1850 #if USE_SPLIT_PMD_PTLOCKS
1851 
1852 static struct page *pmd_to_page(pmd_t *pmd)
1853 {
1854 	unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1855 	return virt_to_page((void *)((unsigned long) pmd & mask));
1856 }
1857 
1858 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1859 {
1860 	return ptlock_ptr(pmd_to_page(pmd));
1861 }
1862 
1863 static inline bool pgtable_pmd_page_ctor(struct page *page)
1864 {
1865 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1866 	page->pmd_huge_pte = NULL;
1867 #endif
1868 	return ptlock_init(page);
1869 }
1870 
1871 static inline void pgtable_pmd_page_dtor(struct page *page)
1872 {
1873 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1874 	VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1875 #endif
1876 	ptlock_free(page);
1877 }
1878 
1879 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1880 
1881 #else
1882 
1883 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1884 {
1885 	return &mm->page_table_lock;
1886 }
1887 
1888 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1889 static inline void pgtable_pmd_page_dtor(struct page *page) {}
1890 
1891 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1892 
1893 #endif
1894 
1895 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1896 {
1897 	spinlock_t *ptl = pmd_lockptr(mm, pmd);
1898 	spin_lock(ptl);
1899 	return ptl;
1900 }
1901 
1902 /*
1903  * No scalability reason to split PUD locks yet, but follow the same pattern
1904  * as the PMD locks to make it easier if we decide to.  The VM should not be
1905  * considered ready to switch to split PUD locks yet; there may be places
1906  * which need to be converted from page_table_lock.
1907  */
1908 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
1909 {
1910 	return &mm->page_table_lock;
1911 }
1912 
1913 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
1914 {
1915 	spinlock_t *ptl = pud_lockptr(mm, pud);
1916 
1917 	spin_lock(ptl);
1918 	return ptl;
1919 }
1920 
1921 extern void __init pagecache_init(void);
1922 extern void free_area_init(unsigned long * zones_size);
1923 extern void free_area_init_node(int nid, unsigned long * zones_size,
1924 		unsigned long zone_start_pfn, unsigned long *zholes_size);
1925 extern void free_initmem(void);
1926 
1927 /*
1928  * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1929  * into the buddy system. The freed pages will be poisoned with pattern
1930  * "poison" if it's within range [0, UCHAR_MAX].
1931  * Return pages freed into the buddy system.
1932  */
1933 extern unsigned long free_reserved_area(void *start, void *end,
1934 					int poison, char *s);
1935 
1936 #ifdef	CONFIG_HIGHMEM
1937 /*
1938  * Free a highmem page into the buddy system, adjusting totalhigh_pages
1939  * and totalram_pages.
1940  */
1941 extern void free_highmem_page(struct page *page);
1942 #endif
1943 
1944 extern void adjust_managed_page_count(struct page *page, long count);
1945 extern void mem_init_print_info(const char *str);
1946 
1947 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
1948 
1949 /* Free the reserved page into the buddy system, so it gets managed. */
1950 static inline void __free_reserved_page(struct page *page)
1951 {
1952 	ClearPageReserved(page);
1953 	init_page_count(page);
1954 	__free_page(page);
1955 }
1956 
1957 static inline void free_reserved_page(struct page *page)
1958 {
1959 	__free_reserved_page(page);
1960 	adjust_managed_page_count(page, 1);
1961 }
1962 
1963 static inline void mark_page_reserved(struct page *page)
1964 {
1965 	SetPageReserved(page);
1966 	adjust_managed_page_count(page, -1);
1967 }
1968 
1969 /*
1970  * Default method to free all the __init memory into the buddy system.
1971  * The freed pages will be poisoned with pattern "poison" if it's within
1972  * range [0, UCHAR_MAX].
1973  * Return pages freed into the buddy system.
1974  */
1975 static inline unsigned long free_initmem_default(int poison)
1976 {
1977 	extern char __init_begin[], __init_end[];
1978 
1979 	return free_reserved_area(&__init_begin, &__init_end,
1980 				  poison, "unused kernel");
1981 }
1982 
1983 static inline unsigned long get_num_physpages(void)
1984 {
1985 	int nid;
1986 	unsigned long phys_pages = 0;
1987 
1988 	for_each_online_node(nid)
1989 		phys_pages += node_present_pages(nid);
1990 
1991 	return phys_pages;
1992 }
1993 
1994 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1995 /*
1996  * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1997  * zones, allocate the backing mem_map and account for memory holes in a more
1998  * architecture independent manner. This is a substitute for creating the
1999  * zone_sizes[] and zholes_size[] arrays and passing them to
2000  * free_area_init_node()
2001  *
2002  * An architecture is expected to register range of page frames backed by
2003  * physical memory with memblock_add[_node]() before calling
2004  * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
2005  * usage, an architecture is expected to do something like
2006  *
2007  * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2008  * 							 max_highmem_pfn};
2009  * for_each_valid_physical_page_range()
2010  * 	memblock_add_node(base, size, nid)
2011  * free_area_init_nodes(max_zone_pfns);
2012  *
2013  * free_bootmem_with_active_regions() calls free_bootmem_node() for each
2014  * registered physical page range.  Similarly
2015  * sparse_memory_present_with_active_regions() calls memory_present() for
2016  * each range when SPARSEMEM is enabled.
2017  *
2018  * See mm/page_alloc.c for more information on each function exposed by
2019  * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
2020  */
2021 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
2022 unsigned long node_map_pfn_alignment(void);
2023 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2024 						unsigned long end_pfn);
2025 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2026 						unsigned long end_pfn);
2027 extern void get_pfn_range_for_nid(unsigned int nid,
2028 			unsigned long *start_pfn, unsigned long *end_pfn);
2029 extern unsigned long find_min_pfn_with_active_regions(void);
2030 extern void free_bootmem_with_active_regions(int nid,
2031 						unsigned long max_low_pfn);
2032 extern void sparse_memory_present_with_active_regions(int nid);
2033 
2034 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
2035 
2036 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
2037     !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
2038 static inline int __early_pfn_to_nid(unsigned long pfn,
2039 					struct mminit_pfnnid_cache *state)
2040 {
2041 	return 0;
2042 }
2043 #else
2044 /* please see mm/page_alloc.c */
2045 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2046 /* there is a per-arch backend function. */
2047 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
2048 					struct mminit_pfnnid_cache *state);
2049 #endif
2050 
2051 #ifdef CONFIG_HAVE_MEMBLOCK
2052 void zero_resv_unavail(void);
2053 #else
2054 static inline void zero_resv_unavail(void) {}
2055 #endif
2056 
2057 extern void set_dma_reserve(unsigned long new_dma_reserve);
2058 extern void memmap_init_zone(unsigned long, int, unsigned long,
2059 				unsigned long, enum memmap_context);
2060 extern void setup_per_zone_wmarks(void);
2061 extern int __meminit init_per_zone_wmark_min(void);
2062 extern void mem_init(void);
2063 extern void __init mmap_init(void);
2064 extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2065 extern long si_mem_available(void);
2066 extern void si_meminfo(struct sysinfo * val);
2067 extern void si_meminfo_node(struct sysinfo *val, int nid);
2068 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2069 extern unsigned long arch_reserved_kernel_pages(void);
2070 #endif
2071 
2072 extern __printf(3, 4)
2073 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2074 
2075 extern void setup_per_cpu_pageset(void);
2076 
2077 extern void zone_pcp_update(struct zone *zone);
2078 extern void zone_pcp_reset(struct zone *zone);
2079 
2080 /* page_alloc.c */
2081 extern int min_free_kbytes;
2082 extern int watermark_scale_factor;
2083 
2084 /* nommu.c */
2085 extern atomic_long_t mmap_pages_allocated;
2086 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2087 
2088 /* interval_tree.c */
2089 void vma_interval_tree_insert(struct vm_area_struct *node,
2090 			      struct rb_root_cached *root);
2091 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2092 				    struct vm_area_struct *prev,
2093 				    struct rb_root_cached *root);
2094 void vma_interval_tree_remove(struct vm_area_struct *node,
2095 			      struct rb_root_cached *root);
2096 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2097 				unsigned long start, unsigned long last);
2098 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2099 				unsigned long start, unsigned long last);
2100 
2101 #define vma_interval_tree_foreach(vma, root, start, last)		\
2102 	for (vma = vma_interval_tree_iter_first(root, start, last);	\
2103 	     vma; vma = vma_interval_tree_iter_next(vma, start, last))
2104 
2105 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2106 				   struct rb_root_cached *root);
2107 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2108 				   struct rb_root_cached *root);
2109 struct anon_vma_chain *
2110 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2111 				  unsigned long start, unsigned long last);
2112 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2113 	struct anon_vma_chain *node, unsigned long start, unsigned long last);
2114 #ifdef CONFIG_DEBUG_VM_RB
2115 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2116 #endif
2117 
2118 #define anon_vma_interval_tree_foreach(avc, root, start, last)		 \
2119 	for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2120 	     avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2121 
2122 /* mmap.c */
2123 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2124 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2125 	unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2126 	struct vm_area_struct *expand);
2127 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2128 	unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2129 {
2130 	return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2131 }
2132 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2133 	struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2134 	unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2135 	struct mempolicy *, struct vm_userfaultfd_ctx);
2136 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2137 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2138 	unsigned long addr, int new_below);
2139 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2140 	unsigned long addr, int new_below);
2141 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2142 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2143 	struct rb_node **, struct rb_node *);
2144 extern void unlink_file_vma(struct vm_area_struct *);
2145 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2146 	unsigned long addr, unsigned long len, pgoff_t pgoff,
2147 	bool *need_rmap_locks);
2148 extern void exit_mmap(struct mm_struct *);
2149 
2150 static inline int check_data_rlimit(unsigned long rlim,
2151 				    unsigned long new,
2152 				    unsigned long start,
2153 				    unsigned long end_data,
2154 				    unsigned long start_data)
2155 {
2156 	if (rlim < RLIM_INFINITY) {
2157 		if (((new - start) + (end_data - start_data)) > rlim)
2158 			return -ENOSPC;
2159 	}
2160 
2161 	return 0;
2162 }
2163 
2164 extern int mm_take_all_locks(struct mm_struct *mm);
2165 extern void mm_drop_all_locks(struct mm_struct *mm);
2166 
2167 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2168 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2169 extern struct file *get_task_exe_file(struct task_struct *task);
2170 
2171 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2172 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2173 
2174 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2175 				   const struct vm_special_mapping *sm);
2176 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2177 				   unsigned long addr, unsigned long len,
2178 				   unsigned long flags,
2179 				   const struct vm_special_mapping *spec);
2180 /* This is an obsolete alternative to _install_special_mapping. */
2181 extern int install_special_mapping(struct mm_struct *mm,
2182 				   unsigned long addr, unsigned long len,
2183 				   unsigned long flags, struct page **pages);
2184 
2185 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2186 
2187 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2188 	unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2189 	struct list_head *uf);
2190 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2191 	unsigned long len, unsigned long prot, unsigned long flags,
2192 	vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate,
2193 	struct list_head *uf);
2194 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2195 		     struct list_head *uf);
2196 
2197 static inline unsigned long
2198 do_mmap_pgoff(struct file *file, unsigned long addr,
2199 	unsigned long len, unsigned long prot, unsigned long flags,
2200 	unsigned long pgoff, unsigned long *populate,
2201 	struct list_head *uf)
2202 {
2203 	return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate, uf);
2204 }
2205 
2206 #ifdef CONFIG_MMU
2207 extern int __mm_populate(unsigned long addr, unsigned long len,
2208 			 int ignore_errors);
2209 static inline void mm_populate(unsigned long addr, unsigned long len)
2210 {
2211 	/* Ignore errors */
2212 	(void) __mm_populate(addr, len, 1);
2213 }
2214 #else
2215 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2216 #endif
2217 
2218 /* These take the mm semaphore themselves */
2219 extern int __must_check vm_brk(unsigned long, unsigned long);
2220 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2221 extern int vm_munmap(unsigned long, size_t);
2222 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2223         unsigned long, unsigned long,
2224         unsigned long, unsigned long);
2225 
2226 struct vm_unmapped_area_info {
2227 #define VM_UNMAPPED_AREA_TOPDOWN 1
2228 	unsigned long flags;
2229 	unsigned long length;
2230 	unsigned long low_limit;
2231 	unsigned long high_limit;
2232 	unsigned long align_mask;
2233 	unsigned long align_offset;
2234 };
2235 
2236 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2237 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2238 
2239 /*
2240  * Search for an unmapped address range.
2241  *
2242  * We are looking for a range that:
2243  * - does not intersect with any VMA;
2244  * - is contained within the [low_limit, high_limit) interval;
2245  * - is at least the desired size.
2246  * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2247  */
2248 static inline unsigned long
2249 vm_unmapped_area(struct vm_unmapped_area_info *info)
2250 {
2251 	if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2252 		return unmapped_area_topdown(info);
2253 	else
2254 		return unmapped_area(info);
2255 }
2256 
2257 /* truncate.c */
2258 extern void truncate_inode_pages(struct address_space *, loff_t);
2259 extern void truncate_inode_pages_range(struct address_space *,
2260 				       loff_t lstart, loff_t lend);
2261 extern void truncate_inode_pages_final(struct address_space *);
2262 
2263 /* generic vm_area_ops exported for stackable file systems */
2264 extern int filemap_fault(struct vm_fault *vmf);
2265 extern void filemap_map_pages(struct vm_fault *vmf,
2266 		pgoff_t start_pgoff, pgoff_t end_pgoff);
2267 extern int filemap_page_mkwrite(struct vm_fault *vmf);
2268 
2269 /* mm/page-writeback.c */
2270 int __must_check write_one_page(struct page *page);
2271 void task_dirty_inc(struct task_struct *tsk);
2272 
2273 /* readahead.c */
2274 #define VM_MAX_READAHEAD	128	/* kbytes */
2275 #define VM_MIN_READAHEAD	16	/* kbytes (includes current page) */
2276 
2277 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2278 			pgoff_t offset, unsigned long nr_to_read);
2279 
2280 void page_cache_sync_readahead(struct address_space *mapping,
2281 			       struct file_ra_state *ra,
2282 			       struct file *filp,
2283 			       pgoff_t offset,
2284 			       unsigned long size);
2285 
2286 void page_cache_async_readahead(struct address_space *mapping,
2287 				struct file_ra_state *ra,
2288 				struct file *filp,
2289 				struct page *pg,
2290 				pgoff_t offset,
2291 				unsigned long size);
2292 
2293 extern unsigned long stack_guard_gap;
2294 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2295 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2296 
2297 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2298 extern int expand_downwards(struct vm_area_struct *vma,
2299 		unsigned long address);
2300 #if VM_GROWSUP
2301 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2302 #else
2303   #define expand_upwards(vma, address) (0)
2304 #endif
2305 
2306 /* Look up the first VMA which satisfies  addr < vm_end,  NULL if none. */
2307 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2308 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2309 					     struct vm_area_struct **pprev);
2310 
2311 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2312    NULL if none.  Assume start_addr < end_addr. */
2313 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2314 {
2315 	struct vm_area_struct * vma = find_vma(mm,start_addr);
2316 
2317 	if (vma && end_addr <= vma->vm_start)
2318 		vma = NULL;
2319 	return vma;
2320 }
2321 
2322 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2323 {
2324 	unsigned long vm_start = vma->vm_start;
2325 
2326 	if (vma->vm_flags & VM_GROWSDOWN) {
2327 		vm_start -= stack_guard_gap;
2328 		if (vm_start > vma->vm_start)
2329 			vm_start = 0;
2330 	}
2331 	return vm_start;
2332 }
2333 
2334 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2335 {
2336 	unsigned long vm_end = vma->vm_end;
2337 
2338 	if (vma->vm_flags & VM_GROWSUP) {
2339 		vm_end += stack_guard_gap;
2340 		if (vm_end < vma->vm_end)
2341 			vm_end = -PAGE_SIZE;
2342 	}
2343 	return vm_end;
2344 }
2345 
2346 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2347 {
2348 	return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2349 }
2350 
2351 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2352 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2353 				unsigned long vm_start, unsigned long vm_end)
2354 {
2355 	struct vm_area_struct *vma = find_vma(mm, vm_start);
2356 
2357 	if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2358 		vma = NULL;
2359 
2360 	return vma;
2361 }
2362 
2363 #ifdef CONFIG_MMU
2364 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2365 void vma_set_page_prot(struct vm_area_struct *vma);
2366 #else
2367 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2368 {
2369 	return __pgprot(0);
2370 }
2371 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2372 {
2373 	vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2374 }
2375 #endif
2376 
2377 #ifdef CONFIG_NUMA_BALANCING
2378 unsigned long change_prot_numa(struct vm_area_struct *vma,
2379 			unsigned long start, unsigned long end);
2380 #endif
2381 
2382 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2383 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2384 			unsigned long pfn, unsigned long size, pgprot_t);
2385 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2386 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2387 			unsigned long pfn);
2388 int vm_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2389 			unsigned long pfn, pgprot_t pgprot);
2390 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2391 			pfn_t pfn);
2392 int vm_insert_mixed_mkwrite(struct vm_area_struct *vma, unsigned long addr,
2393 			pfn_t pfn);
2394 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2395 
2396 
2397 struct page *follow_page_mask(struct vm_area_struct *vma,
2398 			      unsigned long address, unsigned int foll_flags,
2399 			      unsigned int *page_mask);
2400 
2401 static inline struct page *follow_page(struct vm_area_struct *vma,
2402 		unsigned long address, unsigned int foll_flags)
2403 {
2404 	unsigned int unused_page_mask;
2405 	return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2406 }
2407 
2408 #define FOLL_WRITE	0x01	/* check pte is writable */
2409 #define FOLL_TOUCH	0x02	/* mark page accessed */
2410 #define FOLL_GET	0x04	/* do get_page on page */
2411 #define FOLL_DUMP	0x08	/* give error on hole if it would be zero */
2412 #define FOLL_FORCE	0x10	/* get_user_pages read/write w/o permission */
2413 #define FOLL_NOWAIT	0x20	/* if a disk transfer is needed, start the IO
2414 				 * and return without waiting upon it */
2415 #define FOLL_POPULATE	0x40	/* fault in page */
2416 #define FOLL_SPLIT	0x80	/* don't return transhuge pages, split them */
2417 #define FOLL_HWPOISON	0x100	/* check page is hwpoisoned */
2418 #define FOLL_NUMA	0x200	/* force NUMA hinting page fault */
2419 #define FOLL_MIGRATION	0x400	/* wait for page to replace migration entry */
2420 #define FOLL_TRIED	0x800	/* a retry, previous pass started an IO */
2421 #define FOLL_MLOCK	0x1000	/* lock present pages */
2422 #define FOLL_REMOTE	0x2000	/* we are working on non-current tsk/mm */
2423 #define FOLL_COW	0x4000	/* internal GUP flag */
2424 
2425 static inline int vm_fault_to_errno(int vm_fault, int foll_flags)
2426 {
2427 	if (vm_fault & VM_FAULT_OOM)
2428 		return -ENOMEM;
2429 	if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2430 		return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2431 	if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2432 		return -EFAULT;
2433 	return 0;
2434 }
2435 
2436 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2437 			void *data);
2438 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2439 			       unsigned long size, pte_fn_t fn, void *data);
2440 
2441 
2442 #ifdef CONFIG_PAGE_POISONING
2443 extern bool page_poisoning_enabled(void);
2444 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2445 extern bool page_is_poisoned(struct page *page);
2446 #else
2447 static inline bool page_poisoning_enabled(void) { return false; }
2448 static inline void kernel_poison_pages(struct page *page, int numpages,
2449 					int enable) { }
2450 static inline bool page_is_poisoned(struct page *page) { return false; }
2451 #endif
2452 
2453 #ifdef CONFIG_DEBUG_PAGEALLOC
2454 extern bool _debug_pagealloc_enabled;
2455 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2456 
2457 static inline bool debug_pagealloc_enabled(void)
2458 {
2459 	return _debug_pagealloc_enabled;
2460 }
2461 
2462 static inline void
2463 kernel_map_pages(struct page *page, int numpages, int enable)
2464 {
2465 	if (!debug_pagealloc_enabled())
2466 		return;
2467 
2468 	__kernel_map_pages(page, numpages, enable);
2469 }
2470 #ifdef CONFIG_HIBERNATION
2471 extern bool kernel_page_present(struct page *page);
2472 #endif	/* CONFIG_HIBERNATION */
2473 #else	/* CONFIG_DEBUG_PAGEALLOC */
2474 static inline void
2475 kernel_map_pages(struct page *page, int numpages, int enable) {}
2476 #ifdef CONFIG_HIBERNATION
2477 static inline bool kernel_page_present(struct page *page) { return true; }
2478 #endif	/* CONFIG_HIBERNATION */
2479 static inline bool debug_pagealloc_enabled(void)
2480 {
2481 	return false;
2482 }
2483 #endif	/* CONFIG_DEBUG_PAGEALLOC */
2484 
2485 #ifdef __HAVE_ARCH_GATE_AREA
2486 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2487 extern int in_gate_area_no_mm(unsigned long addr);
2488 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2489 #else
2490 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2491 {
2492 	return NULL;
2493 }
2494 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2495 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2496 {
2497 	return 0;
2498 }
2499 #endif	/* __HAVE_ARCH_GATE_AREA */
2500 
2501 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2502 
2503 #ifdef CONFIG_SYSCTL
2504 extern int sysctl_drop_caches;
2505 int drop_caches_sysctl_handler(struct ctl_table *, int,
2506 					void __user *, size_t *, loff_t *);
2507 #endif
2508 
2509 void drop_slab(void);
2510 void drop_slab_node(int nid);
2511 
2512 #ifndef CONFIG_MMU
2513 #define randomize_va_space 0
2514 #else
2515 extern int randomize_va_space;
2516 #endif
2517 
2518 const char * arch_vma_name(struct vm_area_struct *vma);
2519 void print_vma_addr(char *prefix, unsigned long rip);
2520 
2521 void sparse_mem_maps_populate_node(struct page **map_map,
2522 				   unsigned long pnum_begin,
2523 				   unsigned long pnum_end,
2524 				   unsigned long map_count,
2525 				   int nodeid);
2526 
2527 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
2528 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2529 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
2530 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
2531 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2532 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2533 void *vmemmap_alloc_block(unsigned long size, int node);
2534 struct vmem_altmap;
2535 void *__vmemmap_alloc_block_buf(unsigned long size, int node,
2536 		struct vmem_altmap *altmap);
2537 static inline void *vmemmap_alloc_block_buf(unsigned long size, int node)
2538 {
2539 	return __vmemmap_alloc_block_buf(size, node, NULL);
2540 }
2541 
2542 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2543 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2544 			       int node);
2545 int vmemmap_populate(unsigned long start, unsigned long end, int node);
2546 void vmemmap_populate_print_last(void);
2547 #ifdef CONFIG_MEMORY_HOTPLUG
2548 void vmemmap_free(unsigned long start, unsigned long end);
2549 #endif
2550 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2551 				  unsigned long nr_pages);
2552 
2553 enum mf_flags {
2554 	MF_COUNT_INCREASED = 1 << 0,
2555 	MF_ACTION_REQUIRED = 1 << 1,
2556 	MF_MUST_KILL = 1 << 2,
2557 	MF_SOFT_OFFLINE = 1 << 3,
2558 };
2559 extern int memory_failure(unsigned long pfn, int trapno, int flags);
2560 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
2561 extern int unpoison_memory(unsigned long pfn);
2562 extern int get_hwpoison_page(struct page *page);
2563 #define put_hwpoison_page(page)	put_page(page)
2564 extern int sysctl_memory_failure_early_kill;
2565 extern int sysctl_memory_failure_recovery;
2566 extern void shake_page(struct page *p, int access);
2567 extern atomic_long_t num_poisoned_pages;
2568 extern int soft_offline_page(struct page *page, int flags);
2569 
2570 
2571 /*
2572  * Error handlers for various types of pages.
2573  */
2574 enum mf_result {
2575 	MF_IGNORED,	/* Error: cannot be handled */
2576 	MF_FAILED,	/* Error: handling failed */
2577 	MF_DELAYED,	/* Will be handled later */
2578 	MF_RECOVERED,	/* Successfully recovered */
2579 };
2580 
2581 enum mf_action_page_type {
2582 	MF_MSG_KERNEL,
2583 	MF_MSG_KERNEL_HIGH_ORDER,
2584 	MF_MSG_SLAB,
2585 	MF_MSG_DIFFERENT_COMPOUND,
2586 	MF_MSG_POISONED_HUGE,
2587 	MF_MSG_HUGE,
2588 	MF_MSG_FREE_HUGE,
2589 	MF_MSG_UNMAP_FAILED,
2590 	MF_MSG_DIRTY_SWAPCACHE,
2591 	MF_MSG_CLEAN_SWAPCACHE,
2592 	MF_MSG_DIRTY_MLOCKED_LRU,
2593 	MF_MSG_CLEAN_MLOCKED_LRU,
2594 	MF_MSG_DIRTY_UNEVICTABLE_LRU,
2595 	MF_MSG_CLEAN_UNEVICTABLE_LRU,
2596 	MF_MSG_DIRTY_LRU,
2597 	MF_MSG_CLEAN_LRU,
2598 	MF_MSG_TRUNCATED_LRU,
2599 	MF_MSG_BUDDY,
2600 	MF_MSG_BUDDY_2ND,
2601 	MF_MSG_UNKNOWN,
2602 };
2603 
2604 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2605 extern void clear_huge_page(struct page *page,
2606 			    unsigned long addr_hint,
2607 			    unsigned int pages_per_huge_page);
2608 extern void copy_user_huge_page(struct page *dst, struct page *src,
2609 				unsigned long addr, struct vm_area_struct *vma,
2610 				unsigned int pages_per_huge_page);
2611 extern long copy_huge_page_from_user(struct page *dst_page,
2612 				const void __user *usr_src,
2613 				unsigned int pages_per_huge_page,
2614 				bool allow_pagefault);
2615 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2616 
2617 extern struct page_ext_operations debug_guardpage_ops;
2618 
2619 #ifdef CONFIG_DEBUG_PAGEALLOC
2620 extern unsigned int _debug_guardpage_minorder;
2621 extern bool _debug_guardpage_enabled;
2622 
2623 static inline unsigned int debug_guardpage_minorder(void)
2624 {
2625 	return _debug_guardpage_minorder;
2626 }
2627 
2628 static inline bool debug_guardpage_enabled(void)
2629 {
2630 	return _debug_guardpage_enabled;
2631 }
2632 
2633 static inline bool page_is_guard(struct page *page)
2634 {
2635 	struct page_ext *page_ext;
2636 
2637 	if (!debug_guardpage_enabled())
2638 		return false;
2639 
2640 	page_ext = lookup_page_ext(page);
2641 	if (unlikely(!page_ext))
2642 		return false;
2643 
2644 	return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2645 }
2646 #else
2647 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2648 static inline bool debug_guardpage_enabled(void) { return false; }
2649 static inline bool page_is_guard(struct page *page) { return false; }
2650 #endif /* CONFIG_DEBUG_PAGEALLOC */
2651 
2652 #if MAX_NUMNODES > 1
2653 void __init setup_nr_node_ids(void);
2654 #else
2655 static inline void setup_nr_node_ids(void) {}
2656 #endif
2657 
2658 #endif /* __KERNEL__ */
2659 #endif /* _LINUX_MM_H */
2660