1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _LINUX_MM_H
3 #define _LINUX_MM_H
4 
5 #include <linux/errno.h>
6 
7 #ifdef __KERNEL__
8 
9 #include <linux/mmdebug.h>
10 #include <linux/gfp.h>
11 #include <linux/bug.h>
12 #include <linux/list.h>
13 #include <linux/mmzone.h>
14 #include <linux/rbtree.h>
15 #include <linux/atomic.h>
16 #include <linux/debug_locks.h>
17 #include <linux/mm_types.h>
18 #include <linux/range.h>
19 #include <linux/pfn.h>
20 #include <linux/percpu-refcount.h>
21 #include <linux/bit_spinlock.h>
22 #include <linux/shrinker.h>
23 #include <linux/resource.h>
24 #include <linux/page_ext.h>
25 #include <linux/err.h>
26 #include <linux/page_ref.h>
27 #include <linux/memremap.h>
28 #include <linux/overflow.h>
29 
30 struct mempolicy;
31 struct anon_vma;
32 struct anon_vma_chain;
33 struct file_ra_state;
34 struct user_struct;
35 struct writeback_control;
36 struct bdi_writeback;
37 
38 void init_mm_internals(void);
39 
40 #ifndef CONFIG_NEED_MULTIPLE_NODES	/* Don't use mapnrs, do it properly */
41 extern unsigned long max_mapnr;
42 
set_max_mapnr(unsigned long limit)43 static inline void set_max_mapnr(unsigned long limit)
44 {
45 	max_mapnr = limit;
46 }
47 #else
set_max_mapnr(unsigned long limit)48 static inline void set_max_mapnr(unsigned long limit) { }
49 #endif
50 
51 extern unsigned long totalram_pages;
52 extern void * high_memory;
53 extern int page_cluster;
54 
55 #ifdef CONFIG_SYSCTL
56 extern int sysctl_legacy_va_layout;
57 #else
58 #define sysctl_legacy_va_layout 0
59 #endif
60 
61 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
62 extern const int mmap_rnd_bits_min;
63 extern const int mmap_rnd_bits_max;
64 extern int mmap_rnd_bits __read_mostly;
65 #endif
66 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
67 extern const int mmap_rnd_compat_bits_min;
68 extern const int mmap_rnd_compat_bits_max;
69 extern int mmap_rnd_compat_bits __read_mostly;
70 #endif
71 
72 #include <asm/page.h>
73 #include <asm/pgtable.h>
74 #include <asm/processor.h>
75 
76 #ifndef __pa_symbol
77 #define __pa_symbol(x)  __pa(RELOC_HIDE((unsigned long)(x), 0))
78 #endif
79 
80 #ifndef page_to_virt
81 #define page_to_virt(x)	__va(PFN_PHYS(page_to_pfn(x)))
82 #endif
83 
84 #ifndef lm_alias
85 #define lm_alias(x)	__va(__pa_symbol(x))
86 #endif
87 
88 /*
89  * To prevent common memory management code establishing
90  * a zero page mapping on a read fault.
91  * This macro should be defined within <asm/pgtable.h>.
92  * s390 does this to prevent multiplexing of hardware bits
93  * related to the physical page in case of virtualization.
94  */
95 #ifndef mm_forbids_zeropage
96 #define mm_forbids_zeropage(X)	(0)
97 #endif
98 
99 /*
100  * On some architectures it is expensive to call memset() for small sizes.
101  * Those architectures should provide their own implementation of "struct page"
102  * zeroing by defining this macro in <asm/pgtable.h>.
103  */
104 #ifndef mm_zero_struct_page
105 #define mm_zero_struct_page(pp)  ((void)memset((pp), 0, sizeof(struct page)))
106 #endif
107 
108 /*
109  * Default maximum number of active map areas, this limits the number of vmas
110  * per mm struct. Users can overwrite this number by sysctl but there is a
111  * problem.
112  *
113  * When a program's coredump is generated as ELF format, a section is created
114  * per a vma. In ELF, the number of sections is represented in unsigned short.
115  * This means the number of sections should be smaller than 65535 at coredump.
116  * Because the kernel adds some informative sections to a image of program at
117  * generating coredump, we need some margin. The number of extra sections is
118  * 1-3 now and depends on arch. We use "5" as safe margin, here.
119  *
120  * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
121  * not a hard limit any more. Although some userspace tools can be surprised by
122  * that.
123  */
124 #define MAPCOUNT_ELF_CORE_MARGIN	(5)
125 #define DEFAULT_MAX_MAP_COUNT	(USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
126 
127 extern int sysctl_max_map_count;
128 
129 extern unsigned long sysctl_user_reserve_kbytes;
130 extern unsigned long sysctl_admin_reserve_kbytes;
131 
132 extern int sysctl_overcommit_memory;
133 extern int sysctl_overcommit_ratio;
134 extern unsigned long sysctl_overcommit_kbytes;
135 
136 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
137 				    size_t *, loff_t *);
138 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
139 				    size_t *, loff_t *);
140 
141 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
142 
143 /* to align the pointer to the (next) page boundary */
144 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
145 
146 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
147 #define PAGE_ALIGNED(addr)	IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
148 
149 /*
150  * Linux kernel virtual memory manager primitives.
151  * The idea being to have a "virtual" mm in the same way
152  * we have a virtual fs - giving a cleaner interface to the
153  * mm details, and allowing different kinds of memory mappings
154  * (from shared memory to executable loading to arbitrary
155  * mmap() functions).
156  */
157 
158 struct vm_area_struct *vm_area_alloc(struct mm_struct *);
159 struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
160 void vm_area_free(struct vm_area_struct *);
161 
162 #ifndef CONFIG_MMU
163 extern struct rb_root nommu_region_tree;
164 extern struct rw_semaphore nommu_region_sem;
165 
166 extern unsigned int kobjsize(const void *objp);
167 #endif
168 
169 /*
170  * vm_flags in vm_area_struct, see mm_types.h.
171  * When changing, update also include/trace/events/mmflags.h
172  */
173 #define VM_NONE		0x00000000
174 
175 #define VM_READ		0x00000001	/* currently active flags */
176 #define VM_WRITE	0x00000002
177 #define VM_EXEC		0x00000004
178 #define VM_SHARED	0x00000008
179 
180 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
181 #define VM_MAYREAD	0x00000010	/* limits for mprotect() etc */
182 #define VM_MAYWRITE	0x00000020
183 #define VM_MAYEXEC	0x00000040
184 #define VM_MAYSHARE	0x00000080
185 
186 #define VM_GROWSDOWN	0x00000100	/* general info on the segment */
187 #define VM_UFFD_MISSING	0x00000200	/* missing pages tracking */
188 #define VM_PFNMAP	0x00000400	/* Page-ranges managed without "struct page", just pure PFN */
189 #define VM_DENYWRITE	0x00000800	/* ETXTBSY on write attempts.. */
190 #define VM_UFFD_WP	0x00001000	/* wrprotect pages tracking */
191 
192 #define VM_LOCKED	0x00002000
193 #define VM_IO           0x00004000	/* Memory mapped I/O or similar */
194 
195 					/* Used by sys_madvise() */
196 #define VM_SEQ_READ	0x00008000	/* App will access data sequentially */
197 #define VM_RAND_READ	0x00010000	/* App will not benefit from clustered reads */
198 
199 #define VM_DONTCOPY	0x00020000      /* Do not copy this vma on fork */
200 #define VM_DONTEXPAND	0x00040000	/* Cannot expand with mremap() */
201 #define VM_LOCKONFAULT	0x00080000	/* Lock the pages covered when they are faulted in */
202 #define VM_ACCOUNT	0x00100000	/* Is a VM accounted object */
203 #define VM_NORESERVE	0x00200000	/* should the VM suppress accounting */
204 #define VM_HUGETLB	0x00400000	/* Huge TLB Page VM */
205 #define VM_SYNC		0x00800000	/* Synchronous page faults */
206 #define VM_ARCH_1	0x01000000	/* Architecture-specific flag */
207 #define VM_WIPEONFORK	0x02000000	/* Wipe VMA contents in child. */
208 #define VM_DONTDUMP	0x04000000	/* Do not include in the core dump */
209 
210 #ifdef CONFIG_MEM_SOFT_DIRTY
211 # define VM_SOFTDIRTY	0x08000000	/* Not soft dirty clean area */
212 #else
213 # define VM_SOFTDIRTY	0
214 #endif
215 
216 #define VM_MIXEDMAP	0x10000000	/* Can contain "struct page" and pure PFN pages */
217 #define VM_HUGEPAGE	0x20000000	/* MADV_HUGEPAGE marked this vma */
218 #define VM_NOHUGEPAGE	0x40000000	/* MADV_NOHUGEPAGE marked this vma */
219 #define VM_MERGEABLE	0x80000000	/* KSM may merge identical pages */
220 
221 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
222 #define VM_HIGH_ARCH_BIT_0	32	/* bit only usable on 64-bit architectures */
223 #define VM_HIGH_ARCH_BIT_1	33	/* bit only usable on 64-bit architectures */
224 #define VM_HIGH_ARCH_BIT_2	34	/* bit only usable on 64-bit architectures */
225 #define VM_HIGH_ARCH_BIT_3	35	/* bit only usable on 64-bit architectures */
226 #define VM_HIGH_ARCH_BIT_4	36	/* bit only usable on 64-bit architectures */
227 #define VM_HIGH_ARCH_0	BIT(VM_HIGH_ARCH_BIT_0)
228 #define VM_HIGH_ARCH_1	BIT(VM_HIGH_ARCH_BIT_1)
229 #define VM_HIGH_ARCH_2	BIT(VM_HIGH_ARCH_BIT_2)
230 #define VM_HIGH_ARCH_3	BIT(VM_HIGH_ARCH_BIT_3)
231 #define VM_HIGH_ARCH_4	BIT(VM_HIGH_ARCH_BIT_4)
232 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
233 
234 #ifdef CONFIG_ARCH_HAS_PKEYS
235 # define VM_PKEY_SHIFT	VM_HIGH_ARCH_BIT_0
236 # define VM_PKEY_BIT0	VM_HIGH_ARCH_0	/* A protection key is a 4-bit value */
237 # define VM_PKEY_BIT1	VM_HIGH_ARCH_1	/* on x86 and 5-bit value on ppc64   */
238 # define VM_PKEY_BIT2	VM_HIGH_ARCH_2
239 # define VM_PKEY_BIT3	VM_HIGH_ARCH_3
240 #ifdef CONFIG_PPC
241 # define VM_PKEY_BIT4  VM_HIGH_ARCH_4
242 #else
243 # define VM_PKEY_BIT4  0
244 #endif
245 #endif /* CONFIG_ARCH_HAS_PKEYS */
246 
247 #if defined(CONFIG_X86)
248 # define VM_PAT		VM_ARCH_1	/* PAT reserves whole VMA at once (x86) */
249 #elif defined(CONFIG_PPC)
250 # define VM_SAO		VM_ARCH_1	/* Strong Access Ordering (powerpc) */
251 #elif defined(CONFIG_PARISC)
252 # define VM_GROWSUP	VM_ARCH_1
253 #elif defined(CONFIG_IA64)
254 # define VM_GROWSUP	VM_ARCH_1
255 #elif defined(CONFIG_SPARC64)
256 # define VM_SPARC_ADI	VM_ARCH_1	/* Uses ADI tag for access control */
257 # define VM_ARCH_CLEAR	VM_SPARC_ADI
258 #elif !defined(CONFIG_MMU)
259 # define VM_MAPPED_COPY	VM_ARCH_1	/* T if mapped copy of data (nommu mmap) */
260 #endif
261 
262 #if defined(CONFIG_X86_INTEL_MPX)
263 /* MPX specific bounds table or bounds directory */
264 # define VM_MPX		VM_HIGH_ARCH_4
265 #else
266 # define VM_MPX		VM_NONE
267 #endif
268 
269 #ifndef VM_GROWSUP
270 # define VM_GROWSUP	VM_NONE
271 #endif
272 
273 /* Bits set in the VMA until the stack is in its final location */
274 #define VM_STACK_INCOMPLETE_SETUP	(VM_RAND_READ | VM_SEQ_READ)
275 
276 #ifndef VM_STACK_DEFAULT_FLAGS		/* arch can override this */
277 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
278 #endif
279 
280 #ifdef CONFIG_STACK_GROWSUP
281 #define VM_STACK	VM_GROWSUP
282 #else
283 #define VM_STACK	VM_GROWSDOWN
284 #endif
285 
286 #define VM_STACK_FLAGS	(VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
287 
288 /*
289  * Special vmas that are non-mergable, non-mlock()able.
290  * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
291  */
292 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
293 
294 /* This mask defines which mm->def_flags a process can inherit its parent */
295 #define VM_INIT_DEF_MASK	VM_NOHUGEPAGE
296 
297 /* This mask is used to clear all the VMA flags used by mlock */
298 #define VM_LOCKED_CLEAR_MASK	(~(VM_LOCKED | VM_LOCKONFAULT))
299 
300 /* Arch-specific flags to clear when updating VM flags on protection change */
301 #ifndef VM_ARCH_CLEAR
302 # define VM_ARCH_CLEAR	VM_NONE
303 #endif
304 #define VM_FLAGS_CLEAR	(ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
305 
306 /*
307  * mapping from the currently active vm_flags protection bits (the
308  * low four bits) to a page protection mask..
309  */
310 extern pgprot_t protection_map[16];
311 
312 #define FAULT_FLAG_WRITE	0x01	/* Fault was a write access */
313 #define FAULT_FLAG_MKWRITE	0x02	/* Fault was mkwrite of existing pte */
314 #define FAULT_FLAG_ALLOW_RETRY	0x04	/* Retry fault if blocking */
315 #define FAULT_FLAG_RETRY_NOWAIT	0x08	/* Don't drop mmap_sem and wait when retrying */
316 #define FAULT_FLAG_KILLABLE	0x10	/* The fault task is in SIGKILL killable region */
317 #define FAULT_FLAG_TRIED	0x20	/* Second try */
318 #define FAULT_FLAG_USER		0x40	/* The fault originated in userspace */
319 #define FAULT_FLAG_REMOTE	0x80	/* faulting for non current tsk/mm */
320 #define FAULT_FLAG_INSTRUCTION  0x100	/* The fault was during an instruction fetch */
321 
322 #define FAULT_FLAG_TRACE \
323 	{ FAULT_FLAG_WRITE,		"WRITE" }, \
324 	{ FAULT_FLAG_MKWRITE,		"MKWRITE" }, \
325 	{ FAULT_FLAG_ALLOW_RETRY,	"ALLOW_RETRY" }, \
326 	{ FAULT_FLAG_RETRY_NOWAIT,	"RETRY_NOWAIT" }, \
327 	{ FAULT_FLAG_KILLABLE,		"KILLABLE" }, \
328 	{ FAULT_FLAG_TRIED,		"TRIED" }, \
329 	{ FAULT_FLAG_USER,		"USER" }, \
330 	{ FAULT_FLAG_REMOTE,		"REMOTE" }, \
331 	{ FAULT_FLAG_INSTRUCTION,	"INSTRUCTION" }
332 
333 /*
334  * vm_fault is filled by the the pagefault handler and passed to the vma's
335  * ->fault function. The vma's ->fault is responsible for returning a bitmask
336  * of VM_FAULT_xxx flags that give details about how the fault was handled.
337  *
338  * MM layer fills up gfp_mask for page allocations but fault handler might
339  * alter it if its implementation requires a different allocation context.
340  *
341  * pgoff should be used in favour of virtual_address, if possible.
342  */
343 struct vm_fault {
344 	struct vm_area_struct *vma;	/* Target VMA */
345 	unsigned int flags;		/* FAULT_FLAG_xxx flags */
346 	gfp_t gfp_mask;			/* gfp mask to be used for allocations */
347 	pgoff_t pgoff;			/* Logical page offset based on vma */
348 	unsigned long address;		/* Faulting virtual address */
349 	pmd_t *pmd;			/* Pointer to pmd entry matching
350 					 * the 'address' */
351 	pud_t *pud;			/* Pointer to pud entry matching
352 					 * the 'address'
353 					 */
354 	pte_t orig_pte;			/* Value of PTE at the time of fault */
355 
356 	struct page *cow_page;		/* Page handler may use for COW fault */
357 	struct mem_cgroup *memcg;	/* Cgroup cow_page belongs to */
358 	struct page *page;		/* ->fault handlers should return a
359 					 * page here, unless VM_FAULT_NOPAGE
360 					 * is set (which is also implied by
361 					 * VM_FAULT_ERROR).
362 					 */
363 	/* These three entries are valid only while holding ptl lock */
364 	pte_t *pte;			/* Pointer to pte entry matching
365 					 * the 'address'. NULL if the page
366 					 * table hasn't been allocated.
367 					 */
368 	spinlock_t *ptl;		/* Page table lock.
369 					 * Protects pte page table if 'pte'
370 					 * is not NULL, otherwise pmd.
371 					 */
372 	pgtable_t prealloc_pte;		/* Pre-allocated pte page table.
373 					 * vm_ops->map_pages() calls
374 					 * alloc_set_pte() from atomic context.
375 					 * do_fault_around() pre-allocates
376 					 * page table to avoid allocation from
377 					 * atomic context.
378 					 */
379 };
380 
381 /* page entry size for vm->huge_fault() */
382 enum page_entry_size {
383 	PE_SIZE_PTE = 0,
384 	PE_SIZE_PMD,
385 	PE_SIZE_PUD,
386 };
387 
388 /*
389  * These are the virtual MM functions - opening of an area, closing and
390  * unmapping it (needed to keep files on disk up-to-date etc), pointer
391  * to the functions called when a no-page or a wp-page exception occurs.
392  */
393 struct vm_operations_struct {
394 	void (*open)(struct vm_area_struct * area);
395 	void (*close)(struct vm_area_struct * area);
396 	int (*split)(struct vm_area_struct * area, unsigned long addr);
397 	int (*mremap)(struct vm_area_struct * area);
398 	vm_fault_t (*fault)(struct vm_fault *vmf);
399 	vm_fault_t (*huge_fault)(struct vm_fault *vmf,
400 			enum page_entry_size pe_size);
401 	void (*map_pages)(struct vm_fault *vmf,
402 			pgoff_t start_pgoff, pgoff_t end_pgoff);
403 	unsigned long (*pagesize)(struct vm_area_struct * area);
404 
405 	/* notification that a previously read-only page is about to become
406 	 * writable, if an error is returned it will cause a SIGBUS */
407 	vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
408 
409 	/* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
410 	vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
411 
412 	/* called by access_process_vm when get_user_pages() fails, typically
413 	 * for use by special VMAs that can switch between memory and hardware
414 	 */
415 	int (*access)(struct vm_area_struct *vma, unsigned long addr,
416 		      void *buf, int len, int write);
417 
418 	/* Called by the /proc/PID/maps code to ask the vma whether it
419 	 * has a special name.  Returning non-NULL will also cause this
420 	 * vma to be dumped unconditionally. */
421 	const char *(*name)(struct vm_area_struct *vma);
422 
423 #ifdef CONFIG_NUMA
424 	/*
425 	 * set_policy() op must add a reference to any non-NULL @new mempolicy
426 	 * to hold the policy upon return.  Caller should pass NULL @new to
427 	 * remove a policy and fall back to surrounding context--i.e. do not
428 	 * install a MPOL_DEFAULT policy, nor the task or system default
429 	 * mempolicy.
430 	 */
431 	int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
432 
433 	/*
434 	 * get_policy() op must add reference [mpol_get()] to any policy at
435 	 * (vma,addr) marked as MPOL_SHARED.  The shared policy infrastructure
436 	 * in mm/mempolicy.c will do this automatically.
437 	 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
438 	 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
439 	 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
440 	 * must return NULL--i.e., do not "fallback" to task or system default
441 	 * policy.
442 	 */
443 	struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
444 					unsigned long addr);
445 #endif
446 	/*
447 	 * Called by vm_normal_page() for special PTEs to find the
448 	 * page for @addr.  This is useful if the default behavior
449 	 * (using pte_page()) would not find the correct page.
450 	 */
451 	struct page *(*find_special_page)(struct vm_area_struct *vma,
452 					  unsigned long addr);
453 };
454 
vma_init(struct vm_area_struct * vma,struct mm_struct * mm)455 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
456 {
457 	static const struct vm_operations_struct dummy_vm_ops = {};
458 
459 	memset(vma, 0, sizeof(*vma));
460 	vma->vm_mm = mm;
461 	vma->vm_ops = &dummy_vm_ops;
462 	INIT_LIST_HEAD(&vma->anon_vma_chain);
463 }
464 
vma_set_anonymous(struct vm_area_struct * vma)465 static inline void vma_set_anonymous(struct vm_area_struct *vma)
466 {
467 	vma->vm_ops = NULL;
468 }
469 
470 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
471 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
472 
473 struct mmu_gather;
474 struct inode;
475 
476 #define page_private(page)		((page)->private)
477 #define set_page_private(page, v)	((page)->private = (v))
478 
479 #if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
pmd_devmap(pmd_t pmd)480 static inline int pmd_devmap(pmd_t pmd)
481 {
482 	return 0;
483 }
pud_devmap(pud_t pud)484 static inline int pud_devmap(pud_t pud)
485 {
486 	return 0;
487 }
pgd_devmap(pgd_t pgd)488 static inline int pgd_devmap(pgd_t pgd)
489 {
490 	return 0;
491 }
492 #endif
493 
494 /*
495  * FIXME: take this include out, include page-flags.h in
496  * files which need it (119 of them)
497  */
498 #include <linux/page-flags.h>
499 #include <linux/huge_mm.h>
500 
501 /*
502  * Methods to modify the page usage count.
503  *
504  * What counts for a page usage:
505  * - cache mapping   (page->mapping)
506  * - private data    (page->private)
507  * - page mapped in a task's page tables, each mapping
508  *   is counted separately
509  *
510  * Also, many kernel routines increase the page count before a critical
511  * routine so they can be sure the page doesn't go away from under them.
512  */
513 
514 /*
515  * Drop a ref, return true if the refcount fell to zero (the page has no users)
516  */
put_page_testzero(struct page * page)517 static inline int put_page_testzero(struct page *page)
518 {
519 	VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
520 	return page_ref_dec_and_test(page);
521 }
522 
523 /*
524  * Try to grab a ref unless the page has a refcount of zero, return false if
525  * that is the case.
526  * This can be called when MMU is off so it must not access
527  * any of the virtual mappings.
528  */
get_page_unless_zero(struct page * page)529 static inline int get_page_unless_zero(struct page *page)
530 {
531 	return page_ref_add_unless(page, 1, 0);
532 }
533 
534 extern int page_is_ram(unsigned long pfn);
535 
536 enum {
537 	REGION_INTERSECTS,
538 	REGION_DISJOINT,
539 	REGION_MIXED,
540 };
541 
542 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
543 		      unsigned long desc);
544 
545 /* Support for virtually mapped pages */
546 struct page *vmalloc_to_page(const void *addr);
547 unsigned long vmalloc_to_pfn(const void *addr);
548 
549 /*
550  * Determine if an address is within the vmalloc range
551  *
552  * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
553  * is no special casing required.
554  */
is_vmalloc_addr(const void * x)555 static inline bool is_vmalloc_addr(const void *x)
556 {
557 #ifdef CONFIG_MMU
558 	unsigned long addr = (unsigned long)x;
559 
560 	return addr >= VMALLOC_START && addr < VMALLOC_END;
561 #else
562 	return false;
563 #endif
564 }
565 #ifdef CONFIG_MMU
566 extern int is_vmalloc_or_module_addr(const void *x);
567 #else
is_vmalloc_or_module_addr(const void * x)568 static inline int is_vmalloc_or_module_addr(const void *x)
569 {
570 	return 0;
571 }
572 #endif
573 
574 extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
kvmalloc(size_t size,gfp_t flags)575 static inline void *kvmalloc(size_t size, gfp_t flags)
576 {
577 	return kvmalloc_node(size, flags, NUMA_NO_NODE);
578 }
kvzalloc_node(size_t size,gfp_t flags,int node)579 static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
580 {
581 	return kvmalloc_node(size, flags | __GFP_ZERO, node);
582 }
kvzalloc(size_t size,gfp_t flags)583 static inline void *kvzalloc(size_t size, gfp_t flags)
584 {
585 	return kvmalloc(size, flags | __GFP_ZERO);
586 }
587 
kvmalloc_array(size_t n,size_t size,gfp_t flags)588 static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
589 {
590 	size_t bytes;
591 
592 	if (unlikely(check_mul_overflow(n, size, &bytes)))
593 		return NULL;
594 
595 	return kvmalloc(bytes, flags);
596 }
597 
kvcalloc(size_t n,size_t size,gfp_t flags)598 static inline void *kvcalloc(size_t n, size_t size, gfp_t flags)
599 {
600 	return kvmalloc_array(n, size, flags | __GFP_ZERO);
601 }
602 
603 extern void kvfree(const void *addr);
604 extern void kvfree_sensitive(const void *addr, size_t len);
605 
606 /*
607  * Mapcount of compound page as a whole, does not include mapped sub-pages.
608  *
609  * Must be called only for compound pages or any their tail sub-pages.
610  */
compound_mapcount(struct page * page)611 static inline int compound_mapcount(struct page *page)
612 {
613 	VM_BUG_ON_PAGE(!PageCompound(page), page);
614 	page = compound_head(page);
615 	return atomic_read(compound_mapcount_ptr(page)) + 1;
616 }
617 
618 /*
619  * The atomic page->_mapcount, starts from -1: so that transitions
620  * both from it and to it can be tracked, using atomic_inc_and_test
621  * and atomic_add_negative(-1).
622  */
page_mapcount_reset(struct page * page)623 static inline void page_mapcount_reset(struct page *page)
624 {
625 	atomic_set(&(page)->_mapcount, -1);
626 }
627 
628 int __page_mapcount(struct page *page);
629 
630 /*
631  * Mapcount of 0-order page; when compound sub-page, includes
632  * compound_mapcount().
633  *
634  * Result is undefined for pages which cannot be mapped into userspace.
635  * For example SLAB or special types of pages. See function page_has_type().
636  * They use this place in struct page differently.
637  */
page_mapcount(struct page * page)638 static inline int page_mapcount(struct page *page)
639 {
640 	if (unlikely(PageCompound(page)))
641 		return __page_mapcount(page);
642 	return atomic_read(&page->_mapcount) + 1;
643 }
644 
645 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
646 int total_mapcount(struct page *page);
647 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
648 #else
total_mapcount(struct page * page)649 static inline int total_mapcount(struct page *page)
650 {
651 	return page_mapcount(page);
652 }
page_trans_huge_mapcount(struct page * page,int * total_mapcount)653 static inline int page_trans_huge_mapcount(struct page *page,
654 					   int *total_mapcount)
655 {
656 	int mapcount = page_mapcount(page);
657 	if (total_mapcount)
658 		*total_mapcount = mapcount;
659 	return mapcount;
660 }
661 #endif
662 
virt_to_head_page(const void * x)663 static inline struct page *virt_to_head_page(const void *x)
664 {
665 	struct page *page = virt_to_page(x);
666 
667 	return compound_head(page);
668 }
669 
670 void __put_page(struct page *page);
671 
672 void put_pages_list(struct list_head *pages);
673 
674 void split_page(struct page *page, unsigned int order);
675 
676 /*
677  * Compound pages have a destructor function.  Provide a
678  * prototype for that function and accessor functions.
679  * These are _only_ valid on the head of a compound page.
680  */
681 typedef void compound_page_dtor(struct page *);
682 
683 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
684 enum compound_dtor_id {
685 	NULL_COMPOUND_DTOR,
686 	COMPOUND_PAGE_DTOR,
687 #ifdef CONFIG_HUGETLB_PAGE
688 	HUGETLB_PAGE_DTOR,
689 #endif
690 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
691 	TRANSHUGE_PAGE_DTOR,
692 #endif
693 	NR_COMPOUND_DTORS,
694 };
695 extern compound_page_dtor * const compound_page_dtors[];
696 
set_compound_page_dtor(struct page * page,enum compound_dtor_id compound_dtor)697 static inline void set_compound_page_dtor(struct page *page,
698 		enum compound_dtor_id compound_dtor)
699 {
700 	VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
701 	page[1].compound_dtor = compound_dtor;
702 }
703 
get_compound_page_dtor(struct page * page)704 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
705 {
706 	VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
707 	return compound_page_dtors[page[1].compound_dtor];
708 }
709 
compound_order(struct page * page)710 static inline unsigned int compound_order(struct page *page)
711 {
712 	if (!PageHead(page))
713 		return 0;
714 	return page[1].compound_order;
715 }
716 
set_compound_order(struct page * page,unsigned int order)717 static inline void set_compound_order(struct page *page, unsigned int order)
718 {
719 	page[1].compound_order = order;
720 }
721 
722 void free_compound_page(struct page *page);
723 
724 #ifdef CONFIG_MMU
725 /*
726  * Do pte_mkwrite, but only if the vma says VM_WRITE.  We do this when
727  * servicing faults for write access.  In the normal case, do always want
728  * pte_mkwrite.  But get_user_pages can cause write faults for mappings
729  * that do not have writing enabled, when used by access_process_vm.
730  */
maybe_mkwrite(pte_t pte,struct vm_area_struct * vma)731 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
732 {
733 	if (likely(vma->vm_flags & VM_WRITE))
734 		pte = pte_mkwrite(pte);
735 	return pte;
736 }
737 
738 vm_fault_t alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
739 		struct page *page);
740 vm_fault_t finish_fault(struct vm_fault *vmf);
741 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
742 #endif
743 
744 /*
745  * Multiple processes may "see" the same page. E.g. for untouched
746  * mappings of /dev/null, all processes see the same page full of
747  * zeroes, and text pages of executables and shared libraries have
748  * only one copy in memory, at most, normally.
749  *
750  * For the non-reserved pages, page_count(page) denotes a reference count.
751  *   page_count() == 0 means the page is free. page->lru is then used for
752  *   freelist management in the buddy allocator.
753  *   page_count() > 0  means the page has been allocated.
754  *
755  * Pages are allocated by the slab allocator in order to provide memory
756  * to kmalloc and kmem_cache_alloc. In this case, the management of the
757  * page, and the fields in 'struct page' are the responsibility of mm/slab.c
758  * unless a particular usage is carefully commented. (the responsibility of
759  * freeing the kmalloc memory is the caller's, of course).
760  *
761  * A page may be used by anyone else who does a __get_free_page().
762  * In this case, page_count still tracks the references, and should only
763  * be used through the normal accessor functions. The top bits of page->flags
764  * and page->virtual store page management information, but all other fields
765  * are unused and could be used privately, carefully. The management of this
766  * page is the responsibility of the one who allocated it, and those who have
767  * subsequently been given references to it.
768  *
769  * The other pages (we may call them "pagecache pages") are completely
770  * managed by the Linux memory manager: I/O, buffers, swapping etc.
771  * The following discussion applies only to them.
772  *
773  * A pagecache page contains an opaque `private' member, which belongs to the
774  * page's address_space. Usually, this is the address of a circular list of
775  * the page's disk buffers. PG_private must be set to tell the VM to call
776  * into the filesystem to release these pages.
777  *
778  * A page may belong to an inode's memory mapping. In this case, page->mapping
779  * is the pointer to the inode, and page->index is the file offset of the page,
780  * in units of PAGE_SIZE.
781  *
782  * If pagecache pages are not associated with an inode, they are said to be
783  * anonymous pages. These may become associated with the swapcache, and in that
784  * case PG_swapcache is set, and page->private is an offset into the swapcache.
785  *
786  * In either case (swapcache or inode backed), the pagecache itself holds one
787  * reference to the page. Setting PG_private should also increment the
788  * refcount. The each user mapping also has a reference to the page.
789  *
790  * The pagecache pages are stored in a per-mapping radix tree, which is
791  * rooted at mapping->i_pages, and indexed by offset.
792  * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
793  * lists, we instead now tag pages as dirty/writeback in the radix tree.
794  *
795  * All pagecache pages may be subject to I/O:
796  * - inode pages may need to be read from disk,
797  * - inode pages which have been modified and are MAP_SHARED may need
798  *   to be written back to the inode on disk,
799  * - anonymous pages (including MAP_PRIVATE file mappings) which have been
800  *   modified may need to be swapped out to swap space and (later) to be read
801  *   back into memory.
802  */
803 
804 /*
805  * The zone field is never updated after free_area_init_core()
806  * sets it, so none of the operations on it need to be atomic.
807  */
808 
809 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
810 #define SECTIONS_PGOFF		((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
811 #define NODES_PGOFF		(SECTIONS_PGOFF - NODES_WIDTH)
812 #define ZONES_PGOFF		(NODES_PGOFF - ZONES_WIDTH)
813 #define LAST_CPUPID_PGOFF	(ZONES_PGOFF - LAST_CPUPID_WIDTH)
814 
815 /*
816  * Define the bit shifts to access each section.  For non-existent
817  * sections we define the shift as 0; that plus a 0 mask ensures
818  * the compiler will optimise away reference to them.
819  */
820 #define SECTIONS_PGSHIFT	(SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
821 #define NODES_PGSHIFT		(NODES_PGOFF * (NODES_WIDTH != 0))
822 #define ZONES_PGSHIFT		(ZONES_PGOFF * (ZONES_WIDTH != 0))
823 #define LAST_CPUPID_PGSHIFT	(LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
824 
825 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
826 #ifdef NODE_NOT_IN_PAGE_FLAGS
827 #define ZONEID_SHIFT		(SECTIONS_SHIFT + ZONES_SHIFT)
828 #define ZONEID_PGOFF		((SECTIONS_PGOFF < ZONES_PGOFF)? \
829 						SECTIONS_PGOFF : ZONES_PGOFF)
830 #else
831 #define ZONEID_SHIFT		(NODES_SHIFT + ZONES_SHIFT)
832 #define ZONEID_PGOFF		((NODES_PGOFF < ZONES_PGOFF)? \
833 						NODES_PGOFF : ZONES_PGOFF)
834 #endif
835 
836 #define ZONEID_PGSHIFT		(ZONEID_PGOFF * (ZONEID_SHIFT != 0))
837 
838 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
839 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
840 #endif
841 
842 #define ZONES_MASK		((1UL << ZONES_WIDTH) - 1)
843 #define NODES_MASK		((1UL << NODES_WIDTH) - 1)
844 #define SECTIONS_MASK		((1UL << SECTIONS_WIDTH) - 1)
845 #define LAST_CPUPID_MASK	((1UL << LAST_CPUPID_SHIFT) - 1)
846 #define ZONEID_MASK		((1UL << ZONEID_SHIFT) - 1)
847 
page_zonenum(const struct page * page)848 static inline enum zone_type page_zonenum(const struct page *page)
849 {
850 	return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
851 }
852 
853 #ifdef CONFIG_ZONE_DEVICE
is_zone_device_page(const struct page * page)854 static inline bool is_zone_device_page(const struct page *page)
855 {
856 	return page_zonenum(page) == ZONE_DEVICE;
857 }
858 #else
is_zone_device_page(const struct page * page)859 static inline bool is_zone_device_page(const struct page *page)
860 {
861 	return false;
862 }
863 #endif
864 
865 #ifdef CONFIG_DEV_PAGEMAP_OPS
866 void dev_pagemap_get_ops(void);
867 void dev_pagemap_put_ops(void);
868 void __put_devmap_managed_page(struct page *page);
869 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
put_devmap_managed_page(struct page * page)870 static inline bool put_devmap_managed_page(struct page *page)
871 {
872 	if (!static_branch_unlikely(&devmap_managed_key))
873 		return false;
874 	if (!is_zone_device_page(page))
875 		return false;
876 	switch (page->pgmap->type) {
877 	case MEMORY_DEVICE_PRIVATE:
878 	case MEMORY_DEVICE_PUBLIC:
879 	case MEMORY_DEVICE_FS_DAX:
880 		__put_devmap_managed_page(page);
881 		return true;
882 	default:
883 		break;
884 	}
885 	return false;
886 }
887 
is_device_private_page(const struct page * page)888 static inline bool is_device_private_page(const struct page *page)
889 {
890 	return is_zone_device_page(page) &&
891 		page->pgmap->type == MEMORY_DEVICE_PRIVATE;
892 }
893 
is_device_public_page(const struct page * page)894 static inline bool is_device_public_page(const struct page *page)
895 {
896 	return is_zone_device_page(page) &&
897 		page->pgmap->type == MEMORY_DEVICE_PUBLIC;
898 }
899 
900 #else /* CONFIG_DEV_PAGEMAP_OPS */
dev_pagemap_get_ops(void)901 static inline void dev_pagemap_get_ops(void)
902 {
903 }
904 
dev_pagemap_put_ops(void)905 static inline void dev_pagemap_put_ops(void)
906 {
907 }
908 
put_devmap_managed_page(struct page * page)909 static inline bool put_devmap_managed_page(struct page *page)
910 {
911 	return false;
912 }
913 
is_device_private_page(const struct page * page)914 static inline bool is_device_private_page(const struct page *page)
915 {
916 	return false;
917 }
918 
is_device_public_page(const struct page * page)919 static inline bool is_device_public_page(const struct page *page)
920 {
921 	return false;
922 }
923 #endif /* CONFIG_DEV_PAGEMAP_OPS */
924 
925 /* 127: arbitrary random number, small enough to assemble well */
926 #define page_ref_zero_or_close_to_overflow(page) \
927 	((unsigned int) page_ref_count(page) + 127u <= 127u)
928 
get_page(struct page * page)929 static inline void get_page(struct page *page)
930 {
931 	page = compound_head(page);
932 	/*
933 	 * Getting a normal page or the head of a compound page
934 	 * requires to already have an elevated page->_refcount.
935 	 */
936 	VM_BUG_ON_PAGE(page_ref_zero_or_close_to_overflow(page), page);
937 	page_ref_inc(page);
938 }
939 
try_get_page(struct page * page)940 static inline __must_check bool try_get_page(struct page *page)
941 {
942 	page = compound_head(page);
943 	if (WARN_ON_ONCE(page_ref_count(page) <= 0))
944 		return false;
945 	page_ref_inc(page);
946 	return true;
947 }
948 
put_page(struct page * page)949 static inline void put_page(struct page *page)
950 {
951 	page = compound_head(page);
952 
953 	/*
954 	 * For devmap managed pages we need to catch refcount transition from
955 	 * 2 to 1, when refcount reach one it means the page is free and we
956 	 * need to inform the device driver through callback. See
957 	 * include/linux/memremap.h and HMM for details.
958 	 */
959 	if (put_devmap_managed_page(page))
960 		return;
961 
962 	if (put_page_testzero(page))
963 		__put_page(page);
964 }
965 
966 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
967 #define SECTION_IN_PAGE_FLAGS
968 #endif
969 
970 /*
971  * The identification function is mainly used by the buddy allocator for
972  * determining if two pages could be buddies. We are not really identifying
973  * the zone since we could be using the section number id if we do not have
974  * node id available in page flags.
975  * We only guarantee that it will return the same value for two combinable
976  * pages in a zone.
977  */
page_zone_id(struct page * page)978 static inline int page_zone_id(struct page *page)
979 {
980 	return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
981 }
982 
983 #ifdef NODE_NOT_IN_PAGE_FLAGS
984 extern int page_to_nid(const struct page *page);
985 #else
page_to_nid(const struct page * page)986 static inline int page_to_nid(const struct page *page)
987 {
988 	struct page *p = (struct page *)page;
989 
990 	return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
991 }
992 #endif
993 
994 #ifdef CONFIG_NUMA_BALANCING
cpu_pid_to_cpupid(int cpu,int pid)995 static inline int cpu_pid_to_cpupid(int cpu, int pid)
996 {
997 	return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
998 }
999 
cpupid_to_pid(int cpupid)1000 static inline int cpupid_to_pid(int cpupid)
1001 {
1002 	return cpupid & LAST__PID_MASK;
1003 }
1004 
cpupid_to_cpu(int cpupid)1005 static inline int cpupid_to_cpu(int cpupid)
1006 {
1007 	return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1008 }
1009 
cpupid_to_nid(int cpupid)1010 static inline int cpupid_to_nid(int cpupid)
1011 {
1012 	return cpu_to_node(cpupid_to_cpu(cpupid));
1013 }
1014 
cpupid_pid_unset(int cpupid)1015 static inline bool cpupid_pid_unset(int cpupid)
1016 {
1017 	return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1018 }
1019 
cpupid_cpu_unset(int cpupid)1020 static inline bool cpupid_cpu_unset(int cpupid)
1021 {
1022 	return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1023 }
1024 
__cpupid_match_pid(pid_t task_pid,int cpupid)1025 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1026 {
1027 	return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1028 }
1029 
1030 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1031 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
page_cpupid_xchg_last(struct page * page,int cpupid)1032 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1033 {
1034 	return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1035 }
1036 
page_cpupid_last(struct page * page)1037 static inline int page_cpupid_last(struct page *page)
1038 {
1039 	return page->_last_cpupid;
1040 }
page_cpupid_reset_last(struct page * page)1041 static inline void page_cpupid_reset_last(struct page *page)
1042 {
1043 	page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1044 }
1045 #else
page_cpupid_last(struct page * page)1046 static inline int page_cpupid_last(struct page *page)
1047 {
1048 	return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1049 }
1050 
1051 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1052 
page_cpupid_reset_last(struct page * page)1053 static inline void page_cpupid_reset_last(struct page *page)
1054 {
1055 	page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1056 }
1057 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1058 #else /* !CONFIG_NUMA_BALANCING */
page_cpupid_xchg_last(struct page * page,int cpupid)1059 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1060 {
1061 	return page_to_nid(page); /* XXX */
1062 }
1063 
page_cpupid_last(struct page * page)1064 static inline int page_cpupid_last(struct page *page)
1065 {
1066 	return page_to_nid(page); /* XXX */
1067 }
1068 
cpupid_to_nid(int cpupid)1069 static inline int cpupid_to_nid(int cpupid)
1070 {
1071 	return -1;
1072 }
1073 
cpupid_to_pid(int cpupid)1074 static inline int cpupid_to_pid(int cpupid)
1075 {
1076 	return -1;
1077 }
1078 
cpupid_to_cpu(int cpupid)1079 static inline int cpupid_to_cpu(int cpupid)
1080 {
1081 	return -1;
1082 }
1083 
cpu_pid_to_cpupid(int nid,int pid)1084 static inline int cpu_pid_to_cpupid(int nid, int pid)
1085 {
1086 	return -1;
1087 }
1088 
cpupid_pid_unset(int cpupid)1089 static inline bool cpupid_pid_unset(int cpupid)
1090 {
1091 	return 1;
1092 }
1093 
page_cpupid_reset_last(struct page * page)1094 static inline void page_cpupid_reset_last(struct page *page)
1095 {
1096 }
1097 
cpupid_match_pid(struct task_struct * task,int cpupid)1098 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1099 {
1100 	return false;
1101 }
1102 #endif /* CONFIG_NUMA_BALANCING */
1103 
page_zone(const struct page * page)1104 static inline struct zone *page_zone(const struct page *page)
1105 {
1106 	return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1107 }
1108 
page_pgdat(const struct page * page)1109 static inline pg_data_t *page_pgdat(const struct page *page)
1110 {
1111 	return NODE_DATA(page_to_nid(page));
1112 }
1113 
1114 #ifdef SECTION_IN_PAGE_FLAGS
set_page_section(struct page * page,unsigned long section)1115 static inline void set_page_section(struct page *page, unsigned long section)
1116 {
1117 	page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1118 	page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1119 }
1120 
page_to_section(const struct page * page)1121 static inline unsigned long page_to_section(const struct page *page)
1122 {
1123 	return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1124 }
1125 #endif
1126 
set_page_zone(struct page * page,enum zone_type zone)1127 static inline void set_page_zone(struct page *page, enum zone_type zone)
1128 {
1129 	page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1130 	page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1131 }
1132 
set_page_node(struct page * page,unsigned long node)1133 static inline void set_page_node(struct page *page, unsigned long node)
1134 {
1135 	page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1136 	page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1137 }
1138 
set_page_links(struct page * page,enum zone_type zone,unsigned long node,unsigned long pfn)1139 static inline void set_page_links(struct page *page, enum zone_type zone,
1140 	unsigned long node, unsigned long pfn)
1141 {
1142 	set_page_zone(page, zone);
1143 	set_page_node(page, node);
1144 #ifdef SECTION_IN_PAGE_FLAGS
1145 	set_page_section(page, pfn_to_section_nr(pfn));
1146 #endif
1147 }
1148 
1149 #ifdef CONFIG_MEMCG
page_memcg(struct page * page)1150 static inline struct mem_cgroup *page_memcg(struct page *page)
1151 {
1152 	return page->mem_cgroup;
1153 }
page_memcg_rcu(struct page * page)1154 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1155 {
1156 	WARN_ON_ONCE(!rcu_read_lock_held());
1157 	return READ_ONCE(page->mem_cgroup);
1158 }
1159 #else
page_memcg(struct page * page)1160 static inline struct mem_cgroup *page_memcg(struct page *page)
1161 {
1162 	return NULL;
1163 }
page_memcg_rcu(struct page * page)1164 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1165 {
1166 	WARN_ON_ONCE(!rcu_read_lock_held());
1167 	return NULL;
1168 }
1169 #endif
1170 
1171 /*
1172  * Some inline functions in vmstat.h depend on page_zone()
1173  */
1174 #include <linux/vmstat.h>
1175 
lowmem_page_address(const struct page * page)1176 static __always_inline void *lowmem_page_address(const struct page *page)
1177 {
1178 	return page_to_virt(page);
1179 }
1180 
1181 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1182 #define HASHED_PAGE_VIRTUAL
1183 #endif
1184 
1185 #if defined(WANT_PAGE_VIRTUAL)
page_address(const struct page * page)1186 static inline void *page_address(const struct page *page)
1187 {
1188 	return page->virtual;
1189 }
set_page_address(struct page * page,void * address)1190 static inline void set_page_address(struct page *page, void *address)
1191 {
1192 	page->virtual = address;
1193 }
1194 #define page_address_init()  do { } while(0)
1195 #endif
1196 
1197 #if defined(HASHED_PAGE_VIRTUAL)
1198 void *page_address(const struct page *page);
1199 void set_page_address(struct page *page, void *virtual);
1200 void page_address_init(void);
1201 #endif
1202 
1203 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1204 #define page_address(page) lowmem_page_address(page)
1205 #define set_page_address(page, address)  do { } while(0)
1206 #define page_address_init()  do { } while(0)
1207 #endif
1208 
1209 extern void *page_rmapping(struct page *page);
1210 extern struct anon_vma *page_anon_vma(struct page *page);
1211 extern struct address_space *page_mapping(struct page *page);
1212 
1213 extern struct address_space *__page_file_mapping(struct page *);
1214 
1215 static inline
page_file_mapping(struct page * page)1216 struct address_space *page_file_mapping(struct page *page)
1217 {
1218 	if (unlikely(PageSwapCache(page)))
1219 		return __page_file_mapping(page);
1220 
1221 	return page->mapping;
1222 }
1223 
1224 extern pgoff_t __page_file_index(struct page *page);
1225 
1226 /*
1227  * Return the pagecache index of the passed page.  Regular pagecache pages
1228  * use ->index whereas swapcache pages use swp_offset(->private)
1229  */
page_index(struct page * page)1230 static inline pgoff_t page_index(struct page *page)
1231 {
1232 	if (unlikely(PageSwapCache(page)))
1233 		return __page_file_index(page);
1234 	return page->index;
1235 }
1236 
1237 bool page_mapped(struct page *page);
1238 struct address_space *page_mapping(struct page *page);
1239 struct address_space *page_mapping_file(struct page *page);
1240 
1241 /*
1242  * Return true only if the page has been allocated with
1243  * ALLOC_NO_WATERMARKS and the low watermark was not
1244  * met implying that the system is under some pressure.
1245  */
page_is_pfmemalloc(struct page * page)1246 static inline bool page_is_pfmemalloc(struct page *page)
1247 {
1248 	/*
1249 	 * Page index cannot be this large so this must be
1250 	 * a pfmemalloc page.
1251 	 */
1252 	return page->index == -1UL;
1253 }
1254 
1255 /*
1256  * Only to be called by the page allocator on a freshly allocated
1257  * page.
1258  */
set_page_pfmemalloc(struct page * page)1259 static inline void set_page_pfmemalloc(struct page *page)
1260 {
1261 	page->index = -1UL;
1262 }
1263 
clear_page_pfmemalloc(struct page * page)1264 static inline void clear_page_pfmemalloc(struct page *page)
1265 {
1266 	page->index = 0;
1267 }
1268 
1269 /*
1270  * Different kinds of faults, as returned by handle_mm_fault().
1271  * Used to decide whether a process gets delivered SIGBUS or
1272  * just gets major/minor fault counters bumped up.
1273  */
1274 
1275 #define VM_FAULT_OOM	0x0001
1276 #define VM_FAULT_SIGBUS	0x0002
1277 #define VM_FAULT_MAJOR	0x0004
1278 #define VM_FAULT_WRITE	0x0008	/* Special case for get_user_pages */
1279 #define VM_FAULT_HWPOISON 0x0010	/* Hit poisoned small page */
1280 #define VM_FAULT_HWPOISON_LARGE 0x0020  /* Hit poisoned large page. Index encoded in upper bits */
1281 #define VM_FAULT_SIGSEGV 0x0040
1282 
1283 #define VM_FAULT_NOPAGE	0x0100	/* ->fault installed the pte, not return page */
1284 #define VM_FAULT_LOCKED	0x0200	/* ->fault locked the returned page */
1285 #define VM_FAULT_RETRY	0x0400	/* ->fault blocked, must retry */
1286 #define VM_FAULT_FALLBACK 0x0800	/* huge page fault failed, fall back to small */
1287 #define VM_FAULT_DONE_COW   0x1000	/* ->fault has fully handled COW */
1288 #define VM_FAULT_NEEDDSYNC  0x2000	/* ->fault did not modify page tables
1289 					 * and needs fsync() to complete (for
1290 					 * synchronous page faults in DAX) */
1291 
1292 #define VM_FAULT_ERROR	(VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1293 			 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1294 			 VM_FAULT_FALLBACK)
1295 
1296 #define VM_FAULT_RESULT_TRACE \
1297 	{ VM_FAULT_OOM,			"OOM" }, \
1298 	{ VM_FAULT_SIGBUS,		"SIGBUS" }, \
1299 	{ VM_FAULT_MAJOR,		"MAJOR" }, \
1300 	{ VM_FAULT_WRITE,		"WRITE" }, \
1301 	{ VM_FAULT_HWPOISON,		"HWPOISON" }, \
1302 	{ VM_FAULT_HWPOISON_LARGE,	"HWPOISON_LARGE" }, \
1303 	{ VM_FAULT_SIGSEGV,		"SIGSEGV" }, \
1304 	{ VM_FAULT_NOPAGE,		"NOPAGE" }, \
1305 	{ VM_FAULT_LOCKED,		"LOCKED" }, \
1306 	{ VM_FAULT_RETRY,		"RETRY" }, \
1307 	{ VM_FAULT_FALLBACK,		"FALLBACK" }, \
1308 	{ VM_FAULT_DONE_COW,		"DONE_COW" }, \
1309 	{ VM_FAULT_NEEDDSYNC,		"NEEDDSYNC" }
1310 
1311 /* Encode hstate index for a hwpoisoned large page */
1312 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1313 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1314 
1315 /*
1316  * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1317  */
1318 extern void pagefault_out_of_memory(void);
1319 
1320 #define offset_in_page(p)	((unsigned long)(p) & ~PAGE_MASK)
1321 
1322 /*
1323  * Flags passed to show_mem() and show_free_areas() to suppress output in
1324  * various contexts.
1325  */
1326 #define SHOW_MEM_FILTER_NODES		(0x0001u)	/* disallowed nodes */
1327 
1328 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1329 
1330 extern bool can_do_mlock(void);
1331 extern int user_shm_lock(size_t, struct user_struct *);
1332 extern void user_shm_unlock(size_t, struct user_struct *);
1333 
1334 /*
1335  * Parameter block passed down to zap_pte_range in exceptional cases.
1336  */
1337 struct zap_details {
1338 	struct address_space *check_mapping;	/* Check page->mapping if set */
1339 	pgoff_t	first_index;			/* Lowest page->index to unmap */
1340 	pgoff_t last_index;			/* Highest page->index to unmap */
1341 	struct page *single_page;		/* Locked page to be unmapped */
1342 };
1343 
1344 struct page *_vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1345 			     pte_t pte, bool with_public_device);
1346 #define vm_normal_page(vma, addr, pte) _vm_normal_page(vma, addr, pte, false)
1347 
1348 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1349 				pmd_t pmd);
1350 
1351 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1352 		  unsigned long size);
1353 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1354 		    unsigned long size);
1355 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1356 		unsigned long start, unsigned long end);
1357 
1358 /**
1359  * mm_walk - callbacks for walk_page_range
1360  * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
1361  *	       this handler should only handle pud_trans_huge() puds.
1362  *	       the pmd_entry or pte_entry callbacks will be used for
1363  *	       regular PUDs.
1364  * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1365  *	       this handler is required to be able to handle
1366  *	       pmd_trans_huge() pmds.  They may simply choose to
1367  *	       split_huge_page() instead of handling it explicitly.
1368  * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1369  * @pte_hole: if set, called for each hole at all levels
1370  * @hugetlb_entry: if set, called for each hugetlb entry
1371  * @test_walk: caller specific callback function to determine whether
1372  *             we walk over the current vma or not. Returning 0
1373  *             value means "do page table walk over the current vma,"
1374  *             and a negative one means "abort current page table walk
1375  *             right now." 1 means "skip the current vma."
1376  * @mm:        mm_struct representing the target process of page table walk
1377  * @vma:       vma currently walked (NULL if walking outside vmas)
1378  * @private:   private data for callbacks' usage
1379  *
1380  * (see the comment on walk_page_range() for more details)
1381  */
1382 struct mm_walk {
1383 	int (*pud_entry)(pud_t *pud, unsigned long addr,
1384 			 unsigned long next, struct mm_walk *walk);
1385 	int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1386 			 unsigned long next, struct mm_walk *walk);
1387 	int (*pte_entry)(pte_t *pte, unsigned long addr,
1388 			 unsigned long next, struct mm_walk *walk);
1389 	int (*pte_hole)(unsigned long addr, unsigned long next,
1390 			struct mm_walk *walk);
1391 	int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1392 			     unsigned long addr, unsigned long next,
1393 			     struct mm_walk *walk);
1394 	int (*test_walk)(unsigned long addr, unsigned long next,
1395 			struct mm_walk *walk);
1396 	struct mm_struct *mm;
1397 	struct vm_area_struct *vma;
1398 	void *private;
1399 };
1400 
1401 int walk_page_range(unsigned long addr, unsigned long end,
1402 		struct mm_walk *walk);
1403 int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1404 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1405 		unsigned long end, unsigned long floor, unsigned long ceiling);
1406 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1407 			struct vm_area_struct *vma);
1408 int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
1409 			     unsigned long *start, unsigned long *end,
1410 			     pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp);
1411 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1412 	unsigned long *pfn);
1413 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1414 		unsigned int flags, unsigned long *prot, resource_size_t *phys);
1415 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1416 			void *buf, int len, int write);
1417 
1418 extern void truncate_pagecache(struct inode *inode, loff_t new);
1419 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1420 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1421 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1422 int truncate_inode_page(struct address_space *mapping, struct page *page);
1423 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1424 int invalidate_inode_page(struct page *page);
1425 
1426 #ifdef CONFIG_MMU
1427 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1428 			unsigned long address, unsigned int flags);
1429 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1430 			    unsigned long address, unsigned int fault_flags,
1431 			    bool *unlocked);
1432 void unmap_mapping_page(struct page *page);
1433 void unmap_mapping_pages(struct address_space *mapping,
1434 		pgoff_t start, pgoff_t nr, bool even_cows);
1435 void unmap_mapping_range(struct address_space *mapping,
1436 		loff_t const holebegin, loff_t const holelen, int even_cows);
1437 #else
handle_mm_fault(struct vm_area_struct * vma,unsigned long address,unsigned int flags)1438 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1439 		unsigned long address, unsigned int flags)
1440 {
1441 	/* should never happen if there's no MMU */
1442 	BUG();
1443 	return VM_FAULT_SIGBUS;
1444 }
fixup_user_fault(struct task_struct * tsk,struct mm_struct * mm,unsigned long address,unsigned int fault_flags,bool * unlocked)1445 static inline int fixup_user_fault(struct task_struct *tsk,
1446 		struct mm_struct *mm, unsigned long address,
1447 		unsigned int fault_flags, bool *unlocked)
1448 {
1449 	/* should never happen if there's no MMU */
1450 	BUG();
1451 	return -EFAULT;
1452 }
unmap_mapping_page(struct page * page)1453 static inline void unmap_mapping_page(struct page *page) { }
unmap_mapping_pages(struct address_space * mapping,pgoff_t start,pgoff_t nr,bool even_cows)1454 static inline void unmap_mapping_pages(struct address_space *mapping,
1455 		pgoff_t start, pgoff_t nr, bool even_cows) { }
unmap_mapping_range(struct address_space * mapping,loff_t const holebegin,loff_t const holelen,int even_cows)1456 static inline void unmap_mapping_range(struct address_space *mapping,
1457 		loff_t const holebegin, loff_t const holelen, int even_cows) { }
1458 #endif
1459 
unmap_shared_mapping_range(struct address_space * mapping,loff_t const holebegin,loff_t const holelen)1460 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1461 		loff_t const holebegin, loff_t const holelen)
1462 {
1463 	unmap_mapping_range(mapping, holebegin, holelen, 0);
1464 }
1465 
1466 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1467 		void *buf, int len, unsigned int gup_flags);
1468 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1469 		void *buf, int len, unsigned int gup_flags);
1470 extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1471 		unsigned long addr, void *buf, int len, unsigned int gup_flags);
1472 
1473 long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1474 			    unsigned long start, unsigned long nr_pages,
1475 			    unsigned int gup_flags, struct page **pages,
1476 			    struct vm_area_struct **vmas, int *locked);
1477 long get_user_pages(unsigned long start, unsigned long nr_pages,
1478 			    unsigned int gup_flags, struct page **pages,
1479 			    struct vm_area_struct **vmas);
1480 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1481 		    unsigned int gup_flags, struct page **pages, int *locked);
1482 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1483 		    struct page **pages, unsigned int gup_flags);
1484 #ifdef CONFIG_FS_DAX
1485 long get_user_pages_longterm(unsigned long start, unsigned long nr_pages,
1486 			    unsigned int gup_flags, struct page **pages,
1487 			    struct vm_area_struct **vmas);
1488 #else
get_user_pages_longterm(unsigned long start,unsigned long nr_pages,unsigned int gup_flags,struct page ** pages,struct vm_area_struct ** vmas)1489 static inline long get_user_pages_longterm(unsigned long start,
1490 		unsigned long nr_pages, unsigned int gup_flags,
1491 		struct page **pages, struct vm_area_struct **vmas)
1492 {
1493 	return get_user_pages(start, nr_pages, gup_flags, pages, vmas);
1494 }
1495 #endif /* CONFIG_FS_DAX */
1496 
1497 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1498 			struct page **pages);
1499 
1500 /* Container for pinned pfns / pages */
1501 struct frame_vector {
1502 	unsigned int nr_allocated;	/* Number of frames we have space for */
1503 	unsigned int nr_frames;	/* Number of frames stored in ptrs array */
1504 	bool got_ref;		/* Did we pin pages by getting page ref? */
1505 	bool is_pfns;		/* Does array contain pages or pfns? */
1506 	void *ptrs[0];		/* Array of pinned pfns / pages. Use
1507 				 * pfns_vector_pages() or pfns_vector_pfns()
1508 				 * for access */
1509 };
1510 
1511 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1512 void frame_vector_destroy(struct frame_vector *vec);
1513 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1514 		     unsigned int gup_flags, struct frame_vector *vec);
1515 void put_vaddr_frames(struct frame_vector *vec);
1516 int frame_vector_to_pages(struct frame_vector *vec);
1517 void frame_vector_to_pfns(struct frame_vector *vec);
1518 
frame_vector_count(struct frame_vector * vec)1519 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1520 {
1521 	return vec->nr_frames;
1522 }
1523 
frame_vector_pages(struct frame_vector * vec)1524 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1525 {
1526 	if (vec->is_pfns) {
1527 		int err = frame_vector_to_pages(vec);
1528 
1529 		if (err)
1530 			return ERR_PTR(err);
1531 	}
1532 	return (struct page **)(vec->ptrs);
1533 }
1534 
frame_vector_pfns(struct frame_vector * vec)1535 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1536 {
1537 	if (!vec->is_pfns)
1538 		frame_vector_to_pfns(vec);
1539 	return (unsigned long *)(vec->ptrs);
1540 }
1541 
1542 struct kvec;
1543 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1544 			struct page **pages);
1545 int get_kernel_page(unsigned long start, int write, struct page **pages);
1546 struct page *get_dump_page(unsigned long addr);
1547 
1548 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1549 extern void do_invalidatepage(struct page *page, unsigned int offset,
1550 			      unsigned int length);
1551 
1552 void __set_page_dirty(struct page *, struct address_space *, int warn);
1553 int __set_page_dirty_nobuffers(struct page *page);
1554 int __set_page_dirty_no_writeback(struct page *page);
1555 int redirty_page_for_writepage(struct writeback_control *wbc,
1556 				struct page *page);
1557 void account_page_dirtied(struct page *page, struct address_space *mapping);
1558 void account_page_cleaned(struct page *page, struct address_space *mapping,
1559 			  struct bdi_writeback *wb);
1560 int set_page_dirty(struct page *page);
1561 int set_page_dirty_lock(struct page *page);
1562 void __cancel_dirty_page(struct page *page);
cancel_dirty_page(struct page * page)1563 static inline void cancel_dirty_page(struct page *page)
1564 {
1565 	/* Avoid atomic ops, locking, etc. when not actually needed. */
1566 	if (PageDirty(page))
1567 		__cancel_dirty_page(page);
1568 }
1569 int clear_page_dirty_for_io(struct page *page);
1570 
1571 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1572 
vma_is_anonymous(struct vm_area_struct * vma)1573 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1574 {
1575 	return !vma->vm_ops;
1576 }
1577 
1578 #ifdef CONFIG_SHMEM
1579 /*
1580  * The vma_is_shmem is not inline because it is used only by slow
1581  * paths in userfault.
1582  */
1583 bool vma_is_shmem(struct vm_area_struct *vma);
1584 #else
vma_is_shmem(struct vm_area_struct * vma)1585 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
1586 #endif
1587 
1588 int vma_is_stack_for_current(struct vm_area_struct *vma);
1589 
1590 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1591 		unsigned long old_addr, struct vm_area_struct *new_vma,
1592 		unsigned long new_addr, unsigned long len,
1593 		bool need_rmap_locks);
1594 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1595 			      unsigned long end, pgprot_t newprot,
1596 			      int dirty_accountable, int prot_numa);
1597 extern int mprotect_fixup(struct vm_area_struct *vma,
1598 			  struct vm_area_struct **pprev, unsigned long start,
1599 			  unsigned long end, unsigned long newflags);
1600 
1601 /*
1602  * doesn't attempt to fault and will return short.
1603  */
1604 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1605 			  struct page **pages);
1606 /*
1607  * per-process(per-mm_struct) statistics.
1608  */
get_mm_counter(struct mm_struct * mm,int member)1609 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1610 {
1611 	long val = atomic_long_read(&mm->rss_stat.count[member]);
1612 
1613 #ifdef SPLIT_RSS_COUNTING
1614 	/*
1615 	 * counter is updated in asynchronous manner and may go to minus.
1616 	 * But it's never be expected number for users.
1617 	 */
1618 	if (val < 0)
1619 		val = 0;
1620 #endif
1621 	return (unsigned long)val;
1622 }
1623 
add_mm_counter(struct mm_struct * mm,int member,long value)1624 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1625 {
1626 	atomic_long_add(value, &mm->rss_stat.count[member]);
1627 }
1628 
inc_mm_counter(struct mm_struct * mm,int member)1629 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1630 {
1631 	atomic_long_inc(&mm->rss_stat.count[member]);
1632 }
1633 
dec_mm_counter(struct mm_struct * mm,int member)1634 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1635 {
1636 	atomic_long_dec(&mm->rss_stat.count[member]);
1637 }
1638 
1639 /* Optimized variant when page is already known not to be PageAnon */
mm_counter_file(struct page * page)1640 static inline int mm_counter_file(struct page *page)
1641 {
1642 	if (PageSwapBacked(page))
1643 		return MM_SHMEMPAGES;
1644 	return MM_FILEPAGES;
1645 }
1646 
mm_counter(struct page * page)1647 static inline int mm_counter(struct page *page)
1648 {
1649 	if (PageAnon(page))
1650 		return MM_ANONPAGES;
1651 	return mm_counter_file(page);
1652 }
1653 
get_mm_rss(struct mm_struct * mm)1654 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1655 {
1656 	return get_mm_counter(mm, MM_FILEPAGES) +
1657 		get_mm_counter(mm, MM_ANONPAGES) +
1658 		get_mm_counter(mm, MM_SHMEMPAGES);
1659 }
1660 
get_mm_hiwater_rss(struct mm_struct * mm)1661 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1662 {
1663 	return max(mm->hiwater_rss, get_mm_rss(mm));
1664 }
1665 
get_mm_hiwater_vm(struct mm_struct * mm)1666 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1667 {
1668 	return max(mm->hiwater_vm, mm->total_vm);
1669 }
1670 
update_hiwater_rss(struct mm_struct * mm)1671 static inline void update_hiwater_rss(struct mm_struct *mm)
1672 {
1673 	unsigned long _rss = get_mm_rss(mm);
1674 
1675 	if ((mm)->hiwater_rss < _rss)
1676 		(mm)->hiwater_rss = _rss;
1677 }
1678 
update_hiwater_vm(struct mm_struct * mm)1679 static inline void update_hiwater_vm(struct mm_struct *mm)
1680 {
1681 	if (mm->hiwater_vm < mm->total_vm)
1682 		mm->hiwater_vm = mm->total_vm;
1683 }
1684 
reset_mm_hiwater_rss(struct mm_struct * mm)1685 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1686 {
1687 	mm->hiwater_rss = get_mm_rss(mm);
1688 }
1689 
setmax_mm_hiwater_rss(unsigned long * maxrss,struct mm_struct * mm)1690 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1691 					 struct mm_struct *mm)
1692 {
1693 	unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1694 
1695 	if (*maxrss < hiwater_rss)
1696 		*maxrss = hiwater_rss;
1697 }
1698 
1699 #if defined(SPLIT_RSS_COUNTING)
1700 void sync_mm_rss(struct mm_struct *mm);
1701 #else
sync_mm_rss(struct mm_struct * mm)1702 static inline void sync_mm_rss(struct mm_struct *mm)
1703 {
1704 }
1705 #endif
1706 
1707 #ifndef __HAVE_ARCH_PTE_DEVMAP
pte_devmap(pte_t pte)1708 static inline int pte_devmap(pte_t pte)
1709 {
1710 	return 0;
1711 }
1712 #endif
1713 
1714 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1715 
1716 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1717 			       spinlock_t **ptl);
get_locked_pte(struct mm_struct * mm,unsigned long addr,spinlock_t ** ptl)1718 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1719 				    spinlock_t **ptl)
1720 {
1721 	pte_t *ptep;
1722 	__cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1723 	return ptep;
1724 }
1725 
1726 #ifdef __PAGETABLE_P4D_FOLDED
__p4d_alloc(struct mm_struct * mm,pgd_t * pgd,unsigned long address)1727 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1728 						unsigned long address)
1729 {
1730 	return 0;
1731 }
1732 #else
1733 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1734 #endif
1735 
1736 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
__pud_alloc(struct mm_struct * mm,p4d_t * p4d,unsigned long address)1737 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1738 						unsigned long address)
1739 {
1740 	return 0;
1741 }
mm_inc_nr_puds(struct mm_struct * mm)1742 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
mm_dec_nr_puds(struct mm_struct * mm)1743 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
1744 
1745 #else
1746 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
1747 
mm_inc_nr_puds(struct mm_struct * mm)1748 static inline void mm_inc_nr_puds(struct mm_struct *mm)
1749 {
1750 	if (mm_pud_folded(mm))
1751 		return;
1752 	atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1753 }
1754 
mm_dec_nr_puds(struct mm_struct * mm)1755 static inline void mm_dec_nr_puds(struct mm_struct *mm)
1756 {
1757 	if (mm_pud_folded(mm))
1758 		return;
1759 	atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1760 }
1761 #endif
1762 
1763 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
__pmd_alloc(struct mm_struct * mm,pud_t * pud,unsigned long address)1764 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1765 						unsigned long address)
1766 {
1767 	return 0;
1768 }
1769 
mm_inc_nr_pmds(struct mm_struct * mm)1770 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
mm_dec_nr_pmds(struct mm_struct * mm)1771 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1772 
1773 #else
1774 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1775 
mm_inc_nr_pmds(struct mm_struct * mm)1776 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1777 {
1778 	if (mm_pmd_folded(mm))
1779 		return;
1780 	atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1781 }
1782 
mm_dec_nr_pmds(struct mm_struct * mm)1783 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1784 {
1785 	if (mm_pmd_folded(mm))
1786 		return;
1787 	atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1788 }
1789 #endif
1790 
1791 #ifdef CONFIG_MMU
mm_pgtables_bytes_init(struct mm_struct * mm)1792 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
1793 {
1794 	atomic_long_set(&mm->pgtables_bytes, 0);
1795 }
1796 
mm_pgtables_bytes(const struct mm_struct * mm)1797 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1798 {
1799 	return atomic_long_read(&mm->pgtables_bytes);
1800 }
1801 
mm_inc_nr_ptes(struct mm_struct * mm)1802 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
1803 {
1804 	atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1805 }
1806 
mm_dec_nr_ptes(struct mm_struct * mm)1807 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
1808 {
1809 	atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1810 }
1811 #else
1812 
mm_pgtables_bytes_init(struct mm_struct * mm)1813 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
mm_pgtables_bytes(const struct mm_struct * mm)1814 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1815 {
1816 	return 0;
1817 }
1818 
mm_inc_nr_ptes(struct mm_struct * mm)1819 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
mm_dec_nr_ptes(struct mm_struct * mm)1820 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
1821 #endif
1822 
1823 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
1824 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1825 
1826 /*
1827  * The following ifdef needed to get the 4level-fixup.h header to work.
1828  * Remove it when 4level-fixup.h has been removed.
1829  */
1830 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1831 
1832 #ifndef __ARCH_HAS_5LEVEL_HACK
p4d_alloc(struct mm_struct * mm,pgd_t * pgd,unsigned long address)1833 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1834 		unsigned long address)
1835 {
1836 	return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
1837 		NULL : p4d_offset(pgd, address);
1838 }
1839 
pud_alloc(struct mm_struct * mm,p4d_t * p4d,unsigned long address)1840 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1841 		unsigned long address)
1842 {
1843 	return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
1844 		NULL : pud_offset(p4d, address);
1845 }
1846 #endif /* !__ARCH_HAS_5LEVEL_HACK */
1847 
pmd_alloc(struct mm_struct * mm,pud_t * pud,unsigned long address)1848 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1849 {
1850 	return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1851 		NULL: pmd_offset(pud, address);
1852 }
1853 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1854 
1855 #if USE_SPLIT_PTE_PTLOCKS
1856 #if ALLOC_SPLIT_PTLOCKS
1857 void __init ptlock_cache_init(void);
1858 extern bool ptlock_alloc(struct page *page);
1859 extern void ptlock_free(struct page *page);
1860 
ptlock_ptr(struct page * page)1861 static inline spinlock_t *ptlock_ptr(struct page *page)
1862 {
1863 	return page->ptl;
1864 }
1865 #else /* ALLOC_SPLIT_PTLOCKS */
ptlock_cache_init(void)1866 static inline void ptlock_cache_init(void)
1867 {
1868 }
1869 
ptlock_alloc(struct page * page)1870 static inline bool ptlock_alloc(struct page *page)
1871 {
1872 	return true;
1873 }
1874 
ptlock_free(struct page * page)1875 static inline void ptlock_free(struct page *page)
1876 {
1877 }
1878 
ptlock_ptr(struct page * page)1879 static inline spinlock_t *ptlock_ptr(struct page *page)
1880 {
1881 	return &page->ptl;
1882 }
1883 #endif /* ALLOC_SPLIT_PTLOCKS */
1884 
pte_lockptr(struct mm_struct * mm,pmd_t * pmd)1885 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1886 {
1887 	return ptlock_ptr(pmd_page(*pmd));
1888 }
1889 
ptlock_init(struct page * page)1890 static inline bool ptlock_init(struct page *page)
1891 {
1892 	/*
1893 	 * prep_new_page() initialize page->private (and therefore page->ptl)
1894 	 * with 0. Make sure nobody took it in use in between.
1895 	 *
1896 	 * It can happen if arch try to use slab for page table allocation:
1897 	 * slab code uses page->slab_cache, which share storage with page->ptl.
1898 	 */
1899 	VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1900 	if (!ptlock_alloc(page))
1901 		return false;
1902 	spin_lock_init(ptlock_ptr(page));
1903 	return true;
1904 }
1905 
1906 /* Reset page->mapping so free_pages_check won't complain. */
pte_lock_deinit(struct page * page)1907 static inline void pte_lock_deinit(struct page *page)
1908 {
1909 	page->mapping = NULL;
1910 	ptlock_free(page);
1911 }
1912 
1913 #else	/* !USE_SPLIT_PTE_PTLOCKS */
1914 /*
1915  * We use mm->page_table_lock to guard all pagetable pages of the mm.
1916  */
pte_lockptr(struct mm_struct * mm,pmd_t * pmd)1917 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1918 {
1919 	return &mm->page_table_lock;
1920 }
ptlock_cache_init(void)1921 static inline void ptlock_cache_init(void) {}
ptlock_init(struct page * page)1922 static inline bool ptlock_init(struct page *page) { return true; }
pte_lock_deinit(struct page * page)1923 static inline void pte_lock_deinit(struct page *page) {}
1924 #endif /* USE_SPLIT_PTE_PTLOCKS */
1925 
pgtable_init(void)1926 static inline void pgtable_init(void)
1927 {
1928 	ptlock_cache_init();
1929 	pgtable_cache_init();
1930 }
1931 
pgtable_page_ctor(struct page * page)1932 static inline bool pgtable_page_ctor(struct page *page)
1933 {
1934 	if (!ptlock_init(page))
1935 		return false;
1936 	__SetPageTable(page);
1937 	inc_zone_page_state(page, NR_PAGETABLE);
1938 	return true;
1939 }
1940 
pgtable_page_dtor(struct page * page)1941 static inline void pgtable_page_dtor(struct page *page)
1942 {
1943 	pte_lock_deinit(page);
1944 	__ClearPageTable(page);
1945 	dec_zone_page_state(page, NR_PAGETABLE);
1946 }
1947 
1948 #define pte_offset_map_lock(mm, pmd, address, ptlp)	\
1949 ({							\
1950 	spinlock_t *__ptl = pte_lockptr(mm, pmd);	\
1951 	pte_t *__pte = pte_offset_map(pmd, address);	\
1952 	*(ptlp) = __ptl;				\
1953 	spin_lock(__ptl);				\
1954 	__pte;						\
1955 })
1956 
1957 #define pte_unmap_unlock(pte, ptl)	do {		\
1958 	spin_unlock(ptl);				\
1959 	pte_unmap(pte);					\
1960 } while (0)
1961 
1962 #define pte_alloc(mm, pmd, address)			\
1963 	(unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd, address))
1964 
1965 #define pte_alloc_map(mm, pmd, address)			\
1966 	(pte_alloc(mm, pmd, address) ? NULL : pte_offset_map(pmd, address))
1967 
1968 #define pte_alloc_map_lock(mm, pmd, address, ptlp)	\
1969 	(pte_alloc(mm, pmd, address) ?			\
1970 		 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
1971 
1972 #define pte_alloc_kernel(pmd, address)			\
1973 	((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1974 		NULL: pte_offset_kernel(pmd, address))
1975 
1976 #if USE_SPLIT_PMD_PTLOCKS
1977 
pmd_to_page(pmd_t * pmd)1978 static struct page *pmd_to_page(pmd_t *pmd)
1979 {
1980 	unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1981 	return virt_to_page((void *)((unsigned long) pmd & mask));
1982 }
1983 
pmd_lockptr(struct mm_struct * mm,pmd_t * pmd)1984 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1985 {
1986 	return ptlock_ptr(pmd_to_page(pmd));
1987 }
1988 
pgtable_pmd_page_ctor(struct page * page)1989 static inline bool pgtable_pmd_page_ctor(struct page *page)
1990 {
1991 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1992 	page->pmd_huge_pte = NULL;
1993 #endif
1994 	return ptlock_init(page);
1995 }
1996 
pgtable_pmd_page_dtor(struct page * page)1997 static inline void pgtable_pmd_page_dtor(struct page *page)
1998 {
1999 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2000 	VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2001 #endif
2002 	ptlock_free(page);
2003 }
2004 
2005 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2006 
2007 #else
2008 
pmd_lockptr(struct mm_struct * mm,pmd_t * pmd)2009 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2010 {
2011 	return &mm->page_table_lock;
2012 }
2013 
pgtable_pmd_page_ctor(struct page * page)2014 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
pgtable_pmd_page_dtor(struct page * page)2015 static inline void pgtable_pmd_page_dtor(struct page *page) {}
2016 
2017 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2018 
2019 #endif
2020 
pmd_lock(struct mm_struct * mm,pmd_t * pmd)2021 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2022 {
2023 	spinlock_t *ptl = pmd_lockptr(mm, pmd);
2024 	spin_lock(ptl);
2025 	return ptl;
2026 }
2027 
2028 /*
2029  * No scalability reason to split PUD locks yet, but follow the same pattern
2030  * as the PMD locks to make it easier if we decide to.  The VM should not be
2031  * considered ready to switch to split PUD locks yet; there may be places
2032  * which need to be converted from page_table_lock.
2033  */
pud_lockptr(struct mm_struct * mm,pud_t * pud)2034 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2035 {
2036 	return &mm->page_table_lock;
2037 }
2038 
pud_lock(struct mm_struct * mm,pud_t * pud)2039 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2040 {
2041 	spinlock_t *ptl = pud_lockptr(mm, pud);
2042 
2043 	spin_lock(ptl);
2044 	return ptl;
2045 }
2046 
2047 extern void __init pagecache_init(void);
2048 extern void free_area_init(unsigned long * zones_size);
2049 extern void __init free_area_init_node(int nid, unsigned long * zones_size,
2050 		unsigned long zone_start_pfn, unsigned long *zholes_size);
2051 extern void free_initmem(void);
2052 
2053 /*
2054  * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2055  * into the buddy system. The freed pages will be poisoned with pattern
2056  * "poison" if it's within range [0, UCHAR_MAX].
2057  * Return pages freed into the buddy system.
2058  */
2059 extern unsigned long free_reserved_area(void *start, void *end,
2060 					int poison, char *s);
2061 
2062 #ifdef	CONFIG_HIGHMEM
2063 /*
2064  * Free a highmem page into the buddy system, adjusting totalhigh_pages
2065  * and totalram_pages.
2066  */
2067 extern void free_highmem_page(struct page *page);
2068 #endif
2069 
2070 extern void adjust_managed_page_count(struct page *page, long count);
2071 extern void mem_init_print_info(const char *str);
2072 
2073 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2074 
2075 /* Free the reserved page into the buddy system, so it gets managed. */
__free_reserved_page(struct page * page)2076 static inline void __free_reserved_page(struct page *page)
2077 {
2078 	ClearPageReserved(page);
2079 	init_page_count(page);
2080 	__free_page(page);
2081 }
2082 
free_reserved_page(struct page * page)2083 static inline void free_reserved_page(struct page *page)
2084 {
2085 	__free_reserved_page(page);
2086 	adjust_managed_page_count(page, 1);
2087 }
2088 
mark_page_reserved(struct page * page)2089 static inline void mark_page_reserved(struct page *page)
2090 {
2091 	SetPageReserved(page);
2092 	adjust_managed_page_count(page, -1);
2093 }
2094 
2095 /*
2096  * Default method to free all the __init memory into the buddy system.
2097  * The freed pages will be poisoned with pattern "poison" if it's within
2098  * range [0, UCHAR_MAX].
2099  * Return pages freed into the buddy system.
2100  */
free_initmem_default(int poison)2101 static inline unsigned long free_initmem_default(int poison)
2102 {
2103 	extern char __init_begin[], __init_end[];
2104 
2105 	return free_reserved_area(&__init_begin, &__init_end,
2106 				  poison, "unused kernel");
2107 }
2108 
get_num_physpages(void)2109 static inline unsigned long get_num_physpages(void)
2110 {
2111 	int nid;
2112 	unsigned long phys_pages = 0;
2113 
2114 	for_each_online_node(nid)
2115 		phys_pages += node_present_pages(nid);
2116 
2117 	return phys_pages;
2118 }
2119 
2120 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
2121 /*
2122  * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
2123  * zones, allocate the backing mem_map and account for memory holes in a more
2124  * architecture independent manner. This is a substitute for creating the
2125  * zone_sizes[] and zholes_size[] arrays and passing them to
2126  * free_area_init_node()
2127  *
2128  * An architecture is expected to register range of page frames backed by
2129  * physical memory with memblock_add[_node]() before calling
2130  * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
2131  * usage, an architecture is expected to do something like
2132  *
2133  * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2134  * 							 max_highmem_pfn};
2135  * for_each_valid_physical_page_range()
2136  * 	memblock_add_node(base, size, nid)
2137  * free_area_init_nodes(max_zone_pfns);
2138  *
2139  * free_bootmem_with_active_regions() calls free_bootmem_node() for each
2140  * registered physical page range.  Similarly
2141  * sparse_memory_present_with_active_regions() calls memory_present() for
2142  * each range when SPARSEMEM is enabled.
2143  *
2144  * See mm/page_alloc.c for more information on each function exposed by
2145  * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
2146  */
2147 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
2148 unsigned long node_map_pfn_alignment(void);
2149 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2150 						unsigned long end_pfn);
2151 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2152 						unsigned long end_pfn);
2153 extern void get_pfn_range_for_nid(unsigned int nid,
2154 			unsigned long *start_pfn, unsigned long *end_pfn);
2155 extern unsigned long find_min_pfn_with_active_regions(void);
2156 extern void free_bootmem_with_active_regions(int nid,
2157 						unsigned long max_low_pfn);
2158 extern void sparse_memory_present_with_active_regions(int nid);
2159 
2160 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
2161 
2162 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
2163     !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
__early_pfn_to_nid(unsigned long pfn,struct mminit_pfnnid_cache * state)2164 static inline int __early_pfn_to_nid(unsigned long pfn,
2165 					struct mminit_pfnnid_cache *state)
2166 {
2167 	return 0;
2168 }
2169 #else
2170 /* please see mm/page_alloc.c */
2171 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2172 /* there is a per-arch backend function. */
2173 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
2174 					struct mminit_pfnnid_cache *state);
2175 #endif
2176 
2177 #if defined(CONFIG_HAVE_MEMBLOCK) && !defined(CONFIG_FLAT_NODE_MEM_MAP)
2178 void zero_resv_unavail(void);
2179 #else
zero_resv_unavail(void)2180 static inline void zero_resv_unavail(void) {}
2181 #endif
2182 
2183 extern void set_dma_reserve(unsigned long new_dma_reserve);
2184 extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long,
2185 		enum meminit_context, struct vmem_altmap *);
2186 extern void setup_per_zone_wmarks(void);
2187 extern int __meminit init_per_zone_wmark_min(void);
2188 extern void mem_init(void);
2189 extern void __init mmap_init(void);
2190 extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2191 extern long si_mem_available(void);
2192 extern void si_meminfo(struct sysinfo * val);
2193 extern void si_meminfo_node(struct sysinfo *val, int nid);
2194 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2195 extern unsigned long arch_reserved_kernel_pages(void);
2196 #endif
2197 
2198 extern __printf(3, 4)
2199 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2200 
2201 extern void setup_per_cpu_pageset(void);
2202 
2203 extern void zone_pcp_update(struct zone *zone);
2204 extern void zone_pcp_reset(struct zone *zone);
2205 
2206 /* page_alloc.c */
2207 extern int min_free_kbytes;
2208 extern int watermark_scale_factor;
2209 
2210 /* nommu.c */
2211 extern atomic_long_t mmap_pages_allocated;
2212 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2213 
2214 /* interval_tree.c */
2215 void vma_interval_tree_insert(struct vm_area_struct *node,
2216 			      struct rb_root_cached *root);
2217 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2218 				    struct vm_area_struct *prev,
2219 				    struct rb_root_cached *root);
2220 void vma_interval_tree_remove(struct vm_area_struct *node,
2221 			      struct rb_root_cached *root);
2222 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2223 				unsigned long start, unsigned long last);
2224 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2225 				unsigned long start, unsigned long last);
2226 
2227 #define vma_interval_tree_foreach(vma, root, start, last)		\
2228 	for (vma = vma_interval_tree_iter_first(root, start, last);	\
2229 	     vma; vma = vma_interval_tree_iter_next(vma, start, last))
2230 
2231 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2232 				   struct rb_root_cached *root);
2233 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2234 				   struct rb_root_cached *root);
2235 struct anon_vma_chain *
2236 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2237 				  unsigned long start, unsigned long last);
2238 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2239 	struct anon_vma_chain *node, unsigned long start, unsigned long last);
2240 #ifdef CONFIG_DEBUG_VM_RB
2241 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2242 #endif
2243 
2244 #define anon_vma_interval_tree_foreach(avc, root, start, last)		 \
2245 	for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2246 	     avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2247 
2248 /* mmap.c */
2249 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2250 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2251 	unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2252 	struct vm_area_struct *expand);
vma_adjust(struct vm_area_struct * vma,unsigned long start,unsigned long end,pgoff_t pgoff,struct vm_area_struct * insert)2253 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2254 	unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2255 {
2256 	return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2257 }
2258 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2259 	struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2260 	unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2261 	struct mempolicy *, struct vm_userfaultfd_ctx);
2262 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2263 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2264 	unsigned long addr, int new_below);
2265 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2266 	unsigned long addr, int new_below);
2267 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2268 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2269 	struct rb_node **, struct rb_node *);
2270 extern void unlink_file_vma(struct vm_area_struct *);
2271 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2272 	unsigned long addr, unsigned long len, pgoff_t pgoff,
2273 	bool *need_rmap_locks);
2274 extern void exit_mmap(struct mm_struct *);
2275 
check_data_rlimit(unsigned long rlim,unsigned long new,unsigned long start,unsigned long end_data,unsigned long start_data)2276 static inline int check_data_rlimit(unsigned long rlim,
2277 				    unsigned long new,
2278 				    unsigned long start,
2279 				    unsigned long end_data,
2280 				    unsigned long start_data)
2281 {
2282 	if (rlim < RLIM_INFINITY) {
2283 		if (((new - start) + (end_data - start_data)) > rlim)
2284 			return -ENOSPC;
2285 	}
2286 
2287 	return 0;
2288 }
2289 
2290 extern int mm_take_all_locks(struct mm_struct *mm);
2291 extern void mm_drop_all_locks(struct mm_struct *mm);
2292 
2293 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2294 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2295 extern struct file *get_task_exe_file(struct task_struct *task);
2296 
2297 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2298 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2299 
2300 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2301 				   const struct vm_special_mapping *sm);
2302 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2303 				   unsigned long addr, unsigned long len,
2304 				   unsigned long flags,
2305 				   const struct vm_special_mapping *spec);
2306 /* This is an obsolete alternative to _install_special_mapping. */
2307 extern int install_special_mapping(struct mm_struct *mm,
2308 				   unsigned long addr, unsigned long len,
2309 				   unsigned long flags, struct page **pages);
2310 
2311 unsigned long randomize_page(unsigned long start, unsigned long range);
2312 
2313 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2314 
2315 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2316 	unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2317 	struct list_head *uf);
2318 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2319 	unsigned long len, unsigned long prot, unsigned long flags,
2320 	vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate,
2321 	struct list_head *uf);
2322 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2323 		     struct list_head *uf);
2324 
2325 static inline unsigned long
do_mmap_pgoff(struct file * file,unsigned long addr,unsigned long len,unsigned long prot,unsigned long flags,unsigned long pgoff,unsigned long * populate,struct list_head * uf)2326 do_mmap_pgoff(struct file *file, unsigned long addr,
2327 	unsigned long len, unsigned long prot, unsigned long flags,
2328 	unsigned long pgoff, unsigned long *populate,
2329 	struct list_head *uf)
2330 {
2331 	return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate, uf);
2332 }
2333 
2334 #ifdef CONFIG_MMU
2335 extern int __mm_populate(unsigned long addr, unsigned long len,
2336 			 int ignore_errors);
mm_populate(unsigned long addr,unsigned long len)2337 static inline void mm_populate(unsigned long addr, unsigned long len)
2338 {
2339 	/* Ignore errors */
2340 	(void) __mm_populate(addr, len, 1);
2341 }
2342 #else
mm_populate(unsigned long addr,unsigned long len)2343 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2344 #endif
2345 
2346 /* These take the mm semaphore themselves */
2347 extern int __must_check vm_brk(unsigned long, unsigned long);
2348 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2349 extern int vm_munmap(unsigned long, size_t);
2350 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2351         unsigned long, unsigned long,
2352         unsigned long, unsigned long);
2353 
2354 struct vm_unmapped_area_info {
2355 #define VM_UNMAPPED_AREA_TOPDOWN 1
2356 	unsigned long flags;
2357 	unsigned long length;
2358 	unsigned long low_limit;
2359 	unsigned long high_limit;
2360 	unsigned long align_mask;
2361 	unsigned long align_offset;
2362 };
2363 
2364 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2365 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2366 
2367 /*
2368  * Search for an unmapped address range.
2369  *
2370  * We are looking for a range that:
2371  * - does not intersect with any VMA;
2372  * - is contained within the [low_limit, high_limit) interval;
2373  * - is at least the desired size.
2374  * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2375  */
2376 static inline unsigned long
vm_unmapped_area(struct vm_unmapped_area_info * info)2377 vm_unmapped_area(struct vm_unmapped_area_info *info)
2378 {
2379 	if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2380 		return unmapped_area_topdown(info);
2381 	else
2382 		return unmapped_area(info);
2383 }
2384 
2385 /* truncate.c */
2386 extern void truncate_inode_pages(struct address_space *, loff_t);
2387 extern void truncate_inode_pages_range(struct address_space *,
2388 				       loff_t lstart, loff_t lend);
2389 extern void truncate_inode_pages_final(struct address_space *);
2390 
2391 /* generic vm_area_ops exported for stackable file systems */
2392 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2393 extern void filemap_map_pages(struct vm_fault *vmf,
2394 		pgoff_t start_pgoff, pgoff_t end_pgoff);
2395 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2396 
2397 /* mm/page-writeback.c */
2398 int __must_check write_one_page(struct page *page);
2399 void task_dirty_inc(struct task_struct *tsk);
2400 
2401 /* readahead.c */
2402 #define VM_MAX_READAHEAD	128	/* kbytes */
2403 #define VM_MIN_READAHEAD	16	/* kbytes (includes current page) */
2404 
2405 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2406 			pgoff_t offset, unsigned long nr_to_read);
2407 
2408 void page_cache_sync_readahead(struct address_space *mapping,
2409 			       struct file_ra_state *ra,
2410 			       struct file *filp,
2411 			       pgoff_t offset,
2412 			       unsigned long size);
2413 
2414 void page_cache_async_readahead(struct address_space *mapping,
2415 				struct file_ra_state *ra,
2416 				struct file *filp,
2417 				struct page *pg,
2418 				pgoff_t offset,
2419 				unsigned long size);
2420 
2421 extern unsigned long stack_guard_gap;
2422 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2423 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2424 
2425 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2426 extern int expand_downwards(struct vm_area_struct *vma,
2427 		unsigned long address);
2428 #if VM_GROWSUP
2429 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2430 #else
2431   #define expand_upwards(vma, address) (0)
2432 #endif
2433 
2434 /* Look up the first VMA which satisfies  addr < vm_end,  NULL if none. */
2435 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2436 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2437 					     struct vm_area_struct **pprev);
2438 
2439 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2440    NULL if none.  Assume start_addr < end_addr. */
find_vma_intersection(struct mm_struct * mm,unsigned long start_addr,unsigned long end_addr)2441 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2442 {
2443 	struct vm_area_struct * vma = find_vma(mm,start_addr);
2444 
2445 	if (vma && end_addr <= vma->vm_start)
2446 		vma = NULL;
2447 	return vma;
2448 }
2449 
vm_start_gap(struct vm_area_struct * vma)2450 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2451 {
2452 	unsigned long vm_start = vma->vm_start;
2453 
2454 	if (vma->vm_flags & VM_GROWSDOWN) {
2455 		vm_start -= stack_guard_gap;
2456 		if (vm_start > vma->vm_start)
2457 			vm_start = 0;
2458 	}
2459 	return vm_start;
2460 }
2461 
vm_end_gap(struct vm_area_struct * vma)2462 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2463 {
2464 	unsigned long vm_end = vma->vm_end;
2465 
2466 	if (vma->vm_flags & VM_GROWSUP) {
2467 		vm_end += stack_guard_gap;
2468 		if (vm_end < vma->vm_end)
2469 			vm_end = -PAGE_SIZE;
2470 	}
2471 	return vm_end;
2472 }
2473 
vma_pages(struct vm_area_struct * vma)2474 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2475 {
2476 	return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2477 }
2478 
2479 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
find_exact_vma(struct mm_struct * mm,unsigned long vm_start,unsigned long vm_end)2480 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2481 				unsigned long vm_start, unsigned long vm_end)
2482 {
2483 	struct vm_area_struct *vma = find_vma(mm, vm_start);
2484 
2485 	if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2486 		vma = NULL;
2487 
2488 	return vma;
2489 }
2490 
range_in_vma(struct vm_area_struct * vma,unsigned long start,unsigned long end)2491 static inline bool range_in_vma(struct vm_area_struct *vma,
2492 				unsigned long start, unsigned long end)
2493 {
2494 	return (vma && vma->vm_start <= start && end <= vma->vm_end);
2495 }
2496 
2497 #ifdef CONFIG_MMU
2498 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2499 void vma_set_page_prot(struct vm_area_struct *vma);
2500 #else
vm_get_page_prot(unsigned long vm_flags)2501 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2502 {
2503 	return __pgprot(0);
2504 }
vma_set_page_prot(struct vm_area_struct * vma)2505 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2506 {
2507 	vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2508 }
2509 #endif
2510 
2511 #ifdef CONFIG_NUMA_BALANCING
2512 unsigned long change_prot_numa(struct vm_area_struct *vma,
2513 			unsigned long start, unsigned long end);
2514 #endif
2515 
2516 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2517 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2518 			unsigned long pfn, unsigned long size, pgprot_t);
2519 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2520 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2521 			unsigned long pfn);
2522 int vm_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2523 			unsigned long pfn, pgprot_t pgprot);
2524 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2525 			pfn_t pfn);
2526 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
2527 		unsigned long addr, pfn_t pfn);
2528 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2529 
vmf_insert_page(struct vm_area_struct * vma,unsigned long addr,struct page * page)2530 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2531 				unsigned long addr, struct page *page)
2532 {
2533 	int err = vm_insert_page(vma, addr, page);
2534 
2535 	if (err == -ENOMEM)
2536 		return VM_FAULT_OOM;
2537 	if (err < 0 && err != -EBUSY)
2538 		return VM_FAULT_SIGBUS;
2539 
2540 	return VM_FAULT_NOPAGE;
2541 }
2542 
vmf_insert_mixed(struct vm_area_struct * vma,unsigned long addr,pfn_t pfn)2543 static inline vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma,
2544 				unsigned long addr, pfn_t pfn)
2545 {
2546 	int err = vm_insert_mixed(vma, addr, pfn);
2547 
2548 	if (err == -ENOMEM)
2549 		return VM_FAULT_OOM;
2550 	if (err < 0 && err != -EBUSY)
2551 		return VM_FAULT_SIGBUS;
2552 
2553 	return VM_FAULT_NOPAGE;
2554 }
2555 
vmf_insert_pfn(struct vm_area_struct * vma,unsigned long addr,unsigned long pfn)2556 static inline vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma,
2557 			unsigned long addr, unsigned long pfn)
2558 {
2559 	int err = vm_insert_pfn(vma, addr, pfn);
2560 
2561 	if (err == -ENOMEM)
2562 		return VM_FAULT_OOM;
2563 	if (err < 0 && err != -EBUSY)
2564 		return VM_FAULT_SIGBUS;
2565 
2566 	return VM_FAULT_NOPAGE;
2567 }
2568 
2569 #ifndef io_remap_pfn_range
io_remap_pfn_range(struct vm_area_struct * vma,unsigned long addr,unsigned long pfn,unsigned long size,pgprot_t prot)2570 static inline int io_remap_pfn_range(struct vm_area_struct *vma,
2571 				     unsigned long addr, unsigned long pfn,
2572 				     unsigned long size, pgprot_t prot)
2573 {
2574 	return remap_pfn_range(vma, addr, pfn, size, pgprot_decrypted(prot));
2575 }
2576 #endif
2577 
vmf_error(int err)2578 static inline vm_fault_t vmf_error(int err)
2579 {
2580 	if (err == -ENOMEM)
2581 		return VM_FAULT_OOM;
2582 	return VM_FAULT_SIGBUS;
2583 }
2584 
2585 struct page *follow_page_mask(struct vm_area_struct *vma,
2586 			      unsigned long address, unsigned int foll_flags,
2587 			      unsigned int *page_mask);
2588 
follow_page(struct vm_area_struct * vma,unsigned long address,unsigned int foll_flags)2589 static inline struct page *follow_page(struct vm_area_struct *vma,
2590 		unsigned long address, unsigned int foll_flags)
2591 {
2592 	unsigned int unused_page_mask;
2593 	return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2594 }
2595 
2596 #define FOLL_WRITE	0x01	/* check pte is writable */
2597 #define FOLL_TOUCH	0x02	/* mark page accessed */
2598 #define FOLL_GET	0x04	/* do get_page on page */
2599 #define FOLL_DUMP	0x08	/* give error on hole if it would be zero */
2600 #define FOLL_FORCE	0x10	/* get_user_pages read/write w/o permission */
2601 #define FOLL_NOWAIT	0x20	/* if a disk transfer is needed, start the IO
2602 				 * and return without waiting upon it */
2603 #define FOLL_POPULATE	0x40	/* fault in page */
2604 #define FOLL_SPLIT	0x80	/* don't return transhuge pages, split them */
2605 #define FOLL_HWPOISON	0x100	/* check page is hwpoisoned */
2606 #define FOLL_NUMA	0x200	/* force NUMA hinting page fault */
2607 #define FOLL_MIGRATION	0x400	/* wait for page to replace migration entry */
2608 #define FOLL_TRIED	0x800	/* a retry, previous pass started an IO */
2609 #define FOLL_MLOCK	0x1000	/* lock present pages */
2610 #define FOLL_REMOTE	0x2000	/* we are working on non-current tsk/mm */
2611 #define FOLL_COW	0x4000	/* internal GUP flag */
2612 #define FOLL_ANON	0x8000	/* don't do file mappings */
2613 
vm_fault_to_errno(vm_fault_t vm_fault,int foll_flags)2614 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
2615 {
2616 	if (vm_fault & VM_FAULT_OOM)
2617 		return -ENOMEM;
2618 	if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2619 		return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2620 	if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2621 		return -EFAULT;
2622 	return 0;
2623 }
2624 
2625 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2626 			void *data);
2627 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2628 			       unsigned long size, pte_fn_t fn, void *data);
2629 
2630 
2631 #ifdef CONFIG_PAGE_POISONING
2632 extern bool page_poisoning_enabled(void);
2633 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2634 #else
page_poisoning_enabled(void)2635 static inline bool page_poisoning_enabled(void) { return false; }
kernel_poison_pages(struct page * page,int numpages,int enable)2636 static inline void kernel_poison_pages(struct page *page, int numpages,
2637 					int enable) { }
2638 #endif
2639 
2640 #ifdef CONFIG_DEBUG_PAGEALLOC
2641 extern bool _debug_pagealloc_enabled;
2642 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2643 
debug_pagealloc_enabled(void)2644 static inline bool debug_pagealloc_enabled(void)
2645 {
2646 	return _debug_pagealloc_enabled;
2647 }
2648 
2649 static inline void
kernel_map_pages(struct page * page,int numpages,int enable)2650 kernel_map_pages(struct page *page, int numpages, int enable)
2651 {
2652 	if (!debug_pagealloc_enabled())
2653 		return;
2654 
2655 	__kernel_map_pages(page, numpages, enable);
2656 }
2657 #ifdef CONFIG_HIBERNATION
2658 extern bool kernel_page_present(struct page *page);
2659 #endif	/* CONFIG_HIBERNATION */
2660 #else	/* CONFIG_DEBUG_PAGEALLOC */
2661 static inline void
kernel_map_pages(struct page * page,int numpages,int enable)2662 kernel_map_pages(struct page *page, int numpages, int enable) {}
2663 #ifdef CONFIG_HIBERNATION
kernel_page_present(struct page * page)2664 static inline bool kernel_page_present(struct page *page) { return true; }
2665 #endif	/* CONFIG_HIBERNATION */
debug_pagealloc_enabled(void)2666 static inline bool debug_pagealloc_enabled(void)
2667 {
2668 	return false;
2669 }
2670 #endif	/* CONFIG_DEBUG_PAGEALLOC */
2671 
2672 #ifdef __HAVE_ARCH_GATE_AREA
2673 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2674 extern int in_gate_area_no_mm(unsigned long addr);
2675 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2676 #else
get_gate_vma(struct mm_struct * mm)2677 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2678 {
2679 	return NULL;
2680 }
in_gate_area_no_mm(unsigned long addr)2681 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
in_gate_area(struct mm_struct * mm,unsigned long addr)2682 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2683 {
2684 	return 0;
2685 }
2686 #endif	/* __HAVE_ARCH_GATE_AREA */
2687 
2688 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2689 
2690 #ifdef CONFIG_SYSCTL
2691 extern int sysctl_drop_caches;
2692 int drop_caches_sysctl_handler(struct ctl_table *, int,
2693 					void __user *, size_t *, loff_t *);
2694 #endif
2695 
2696 void drop_slab(void);
2697 void drop_slab_node(int nid);
2698 
2699 #ifndef CONFIG_MMU
2700 #define randomize_va_space 0
2701 #else
2702 extern int randomize_va_space;
2703 #endif
2704 
2705 const char * arch_vma_name(struct vm_area_struct *vma);
2706 void print_vma_addr(char *prefix, unsigned long rip);
2707 
2708 void *sparse_buffer_alloc(unsigned long size);
2709 struct page *sparse_mem_map_populate(unsigned long pnum, int nid,
2710 		struct vmem_altmap *altmap);
2711 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2712 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
2713 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
2714 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2715 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2716 void *vmemmap_alloc_block(unsigned long size, int node);
2717 struct vmem_altmap;
2718 void *vmemmap_alloc_block_buf(unsigned long size, int node);
2719 void *altmap_alloc_block_buf(unsigned long size, struct vmem_altmap *altmap);
2720 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2721 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2722 			       int node);
2723 int vmemmap_populate(unsigned long start, unsigned long end, int node,
2724 		struct vmem_altmap *altmap);
2725 void vmemmap_populate_print_last(void);
2726 #ifdef CONFIG_MEMORY_HOTPLUG
2727 void vmemmap_free(unsigned long start, unsigned long end,
2728 		struct vmem_altmap *altmap);
2729 #endif
2730 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2731 				  unsigned long nr_pages);
2732 
2733 enum mf_flags {
2734 	MF_COUNT_INCREASED = 1 << 0,
2735 	MF_ACTION_REQUIRED = 1 << 1,
2736 	MF_MUST_KILL = 1 << 2,
2737 	MF_SOFT_OFFLINE = 1 << 3,
2738 };
2739 extern int memory_failure(unsigned long pfn, int flags);
2740 extern void memory_failure_queue(unsigned long pfn, int flags);
2741 extern int unpoison_memory(unsigned long pfn);
2742 extern int get_hwpoison_page(struct page *page);
2743 #define put_hwpoison_page(page)	put_page(page)
2744 extern int sysctl_memory_failure_early_kill;
2745 extern int sysctl_memory_failure_recovery;
2746 extern void shake_page(struct page *p, int access);
2747 extern atomic_long_t num_poisoned_pages __read_mostly;
2748 extern int soft_offline_page(struct page *page, int flags);
2749 
2750 
2751 /*
2752  * Error handlers for various types of pages.
2753  */
2754 enum mf_result {
2755 	MF_IGNORED,	/* Error: cannot be handled */
2756 	MF_FAILED,	/* Error: handling failed */
2757 	MF_DELAYED,	/* Will be handled later */
2758 	MF_RECOVERED,	/* Successfully recovered */
2759 };
2760 
2761 enum mf_action_page_type {
2762 	MF_MSG_KERNEL,
2763 	MF_MSG_KERNEL_HIGH_ORDER,
2764 	MF_MSG_SLAB,
2765 	MF_MSG_DIFFERENT_COMPOUND,
2766 	MF_MSG_POISONED_HUGE,
2767 	MF_MSG_HUGE,
2768 	MF_MSG_FREE_HUGE,
2769 	MF_MSG_NON_PMD_HUGE,
2770 	MF_MSG_UNMAP_FAILED,
2771 	MF_MSG_DIRTY_SWAPCACHE,
2772 	MF_MSG_CLEAN_SWAPCACHE,
2773 	MF_MSG_DIRTY_MLOCKED_LRU,
2774 	MF_MSG_CLEAN_MLOCKED_LRU,
2775 	MF_MSG_DIRTY_UNEVICTABLE_LRU,
2776 	MF_MSG_CLEAN_UNEVICTABLE_LRU,
2777 	MF_MSG_DIRTY_LRU,
2778 	MF_MSG_CLEAN_LRU,
2779 	MF_MSG_TRUNCATED_LRU,
2780 	MF_MSG_BUDDY,
2781 	MF_MSG_BUDDY_2ND,
2782 	MF_MSG_DAX,
2783 	MF_MSG_UNKNOWN,
2784 };
2785 
2786 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2787 extern void clear_huge_page(struct page *page,
2788 			    unsigned long addr_hint,
2789 			    unsigned int pages_per_huge_page);
2790 extern void copy_user_huge_page(struct page *dst, struct page *src,
2791 				unsigned long addr_hint,
2792 				struct vm_area_struct *vma,
2793 				unsigned int pages_per_huge_page);
2794 extern long copy_huge_page_from_user(struct page *dst_page,
2795 				const void __user *usr_src,
2796 				unsigned int pages_per_huge_page,
2797 				bool allow_pagefault);
2798 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2799 
2800 extern struct page_ext_operations debug_guardpage_ops;
2801 
2802 #ifdef CONFIG_DEBUG_PAGEALLOC
2803 extern unsigned int _debug_guardpage_minorder;
2804 extern bool _debug_guardpage_enabled;
2805 
debug_guardpage_minorder(void)2806 static inline unsigned int debug_guardpage_minorder(void)
2807 {
2808 	return _debug_guardpage_minorder;
2809 }
2810 
debug_guardpage_enabled(void)2811 static inline bool debug_guardpage_enabled(void)
2812 {
2813 	return _debug_guardpage_enabled;
2814 }
2815 
page_is_guard(struct page * page)2816 static inline bool page_is_guard(struct page *page)
2817 {
2818 	struct page_ext *page_ext;
2819 
2820 	if (!debug_guardpage_enabled())
2821 		return false;
2822 
2823 	page_ext = lookup_page_ext(page);
2824 	if (unlikely(!page_ext))
2825 		return false;
2826 
2827 	return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2828 }
2829 #else
debug_guardpage_minorder(void)2830 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
debug_guardpage_enabled(void)2831 static inline bool debug_guardpage_enabled(void) { return false; }
page_is_guard(struct page * page)2832 static inline bool page_is_guard(struct page *page) { return false; }
2833 #endif /* CONFIG_DEBUG_PAGEALLOC */
2834 
2835 #if MAX_NUMNODES > 1
2836 void __init setup_nr_node_ids(void);
2837 #else
setup_nr_node_ids(void)2838 static inline void setup_nr_node_ids(void) {}
2839 #endif
2840 
2841 #endif /* __KERNEL__ */
2842 #endif /* _LINUX_MM_H */
2843