1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _LINUX_MM_H
3 #define _LINUX_MM_H
4 
5 #include <linux/errno.h>
6 #include <linux/mmdebug.h>
7 #include <linux/gfp.h>
8 #include <linux/bug.h>
9 #include <linux/list.h>
10 #include <linux/mmzone.h>
11 #include <linux/rbtree.h>
12 #include <linux/atomic.h>
13 #include <linux/debug_locks.h>
14 #include <linux/mm_types.h>
15 #include <linux/mmap_lock.h>
16 #include <linux/range.h>
17 #include <linux/pfn.h>
18 #include <linux/percpu-refcount.h>
19 #include <linux/bit_spinlock.h>
20 #include <linux/shrinker.h>
21 #include <linux/resource.h>
22 #include <linux/page_ext.h>
23 #include <linux/err.h>
24 #include <linux/page-flags.h>
25 #include <linux/page_ref.h>
26 #include <linux/overflow.h>
27 #include <linux/sizes.h>
28 #include <linux/sched.h>
29 #include <linux/pgtable.h>
30 #include <linux/kasan.h>
31 #include <linux/memremap.h>
32 
33 struct mempolicy;
34 struct anon_vma;
35 struct anon_vma_chain;
36 struct user_struct;
37 struct pt_regs;
38 
39 extern int sysctl_page_lock_unfairness;
40 
41 void init_mm_internals(void);
42 
43 #ifndef CONFIG_NUMA		/* Don't use mapnrs, do it properly */
44 extern unsigned long max_mapnr;
45 
set_max_mapnr(unsigned long limit)46 static inline void set_max_mapnr(unsigned long limit)
47 {
48 	max_mapnr = limit;
49 }
50 #else
set_max_mapnr(unsigned long limit)51 static inline void set_max_mapnr(unsigned long limit) { }
52 #endif
53 
54 extern atomic_long_t _totalram_pages;
totalram_pages(void)55 static inline unsigned long totalram_pages(void)
56 {
57 	return (unsigned long)atomic_long_read(&_totalram_pages);
58 }
59 
totalram_pages_inc(void)60 static inline void totalram_pages_inc(void)
61 {
62 	atomic_long_inc(&_totalram_pages);
63 }
64 
totalram_pages_dec(void)65 static inline void totalram_pages_dec(void)
66 {
67 	atomic_long_dec(&_totalram_pages);
68 }
69 
totalram_pages_add(long count)70 static inline void totalram_pages_add(long count)
71 {
72 	atomic_long_add(count, &_totalram_pages);
73 }
74 
75 extern void * high_memory;
76 extern int page_cluster;
77 
78 #ifdef CONFIG_SYSCTL
79 extern int sysctl_legacy_va_layout;
80 #else
81 #define sysctl_legacy_va_layout 0
82 #endif
83 
84 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
85 extern const int mmap_rnd_bits_min;
86 extern const int mmap_rnd_bits_max;
87 extern int mmap_rnd_bits __read_mostly;
88 #endif
89 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
90 extern const int mmap_rnd_compat_bits_min;
91 extern const int mmap_rnd_compat_bits_max;
92 extern int mmap_rnd_compat_bits __read_mostly;
93 #endif
94 
95 #include <asm/page.h>
96 #include <asm/processor.h>
97 
98 /*
99  * Architectures that support memory tagging (assigning tags to memory regions,
100  * embedding these tags into addresses that point to these memory regions, and
101  * checking that the memory and the pointer tags match on memory accesses)
102  * redefine this macro to strip tags from pointers.
103  * It's defined as noop for architectures that don't support memory tagging.
104  */
105 #ifndef untagged_addr
106 #define untagged_addr(addr) (addr)
107 #endif
108 
109 #ifndef __pa_symbol
110 #define __pa_symbol(x)  __pa(RELOC_HIDE((unsigned long)(x), 0))
111 #endif
112 
113 #ifndef page_to_virt
114 #define page_to_virt(x)	__va(PFN_PHYS(page_to_pfn(x)))
115 #endif
116 
117 #ifndef lm_alias
118 #define lm_alias(x)	__va(__pa_symbol(x))
119 #endif
120 
121 /*
122  * To prevent common memory management code establishing
123  * a zero page mapping on a read fault.
124  * This macro should be defined within <asm/pgtable.h>.
125  * s390 does this to prevent multiplexing of hardware bits
126  * related to the physical page in case of virtualization.
127  */
128 #ifndef mm_forbids_zeropage
129 #define mm_forbids_zeropage(X)	(0)
130 #endif
131 
132 /*
133  * On some architectures it is expensive to call memset() for small sizes.
134  * If an architecture decides to implement their own version of
135  * mm_zero_struct_page they should wrap the defines below in a #ifndef and
136  * define their own version of this macro in <asm/pgtable.h>
137  */
138 #if BITS_PER_LONG == 64
139 /* This function must be updated when the size of struct page grows above 80
140  * or reduces below 56. The idea that compiler optimizes out switch()
141  * statement, and only leaves move/store instructions. Also the compiler can
142  * combine write statements if they are both assignments and can be reordered,
143  * this can result in several of the writes here being dropped.
144  */
145 #define	mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
__mm_zero_struct_page(struct page * page)146 static inline void __mm_zero_struct_page(struct page *page)
147 {
148 	unsigned long *_pp = (void *)page;
149 
150 	 /* Check that struct page is either 56, 64, 72, or 80 bytes */
151 	BUILD_BUG_ON(sizeof(struct page) & 7);
152 	BUILD_BUG_ON(sizeof(struct page) < 56);
153 	BUILD_BUG_ON(sizeof(struct page) > 80);
154 
155 	switch (sizeof(struct page)) {
156 	case 80:
157 		_pp[9] = 0;
158 		fallthrough;
159 	case 72:
160 		_pp[8] = 0;
161 		fallthrough;
162 	case 64:
163 		_pp[7] = 0;
164 		fallthrough;
165 	case 56:
166 		_pp[6] = 0;
167 		_pp[5] = 0;
168 		_pp[4] = 0;
169 		_pp[3] = 0;
170 		_pp[2] = 0;
171 		_pp[1] = 0;
172 		_pp[0] = 0;
173 	}
174 }
175 #else
176 #define mm_zero_struct_page(pp)  ((void)memset((pp), 0, sizeof(struct page)))
177 #endif
178 
179 /*
180  * Default maximum number of active map areas, this limits the number of vmas
181  * per mm struct. Users can overwrite this number by sysctl but there is a
182  * problem.
183  *
184  * When a program's coredump is generated as ELF format, a section is created
185  * per a vma. In ELF, the number of sections is represented in unsigned short.
186  * This means the number of sections should be smaller than 65535 at coredump.
187  * Because the kernel adds some informative sections to a image of program at
188  * generating coredump, we need some margin. The number of extra sections is
189  * 1-3 now and depends on arch. We use "5" as safe margin, here.
190  *
191  * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
192  * not a hard limit any more. Although some userspace tools can be surprised by
193  * that.
194  */
195 #define MAPCOUNT_ELF_CORE_MARGIN	(5)
196 #define DEFAULT_MAX_MAP_COUNT	(USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
197 
198 extern int sysctl_max_map_count;
199 
200 extern unsigned long sysctl_user_reserve_kbytes;
201 extern unsigned long sysctl_admin_reserve_kbytes;
202 
203 extern int sysctl_overcommit_memory;
204 extern int sysctl_overcommit_ratio;
205 extern unsigned long sysctl_overcommit_kbytes;
206 
207 int overcommit_ratio_handler(struct ctl_table *, int, void *, size_t *,
208 		loff_t *);
209 int overcommit_kbytes_handler(struct ctl_table *, int, void *, size_t *,
210 		loff_t *);
211 int overcommit_policy_handler(struct ctl_table *, int, void *, size_t *,
212 		loff_t *);
213 
214 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
215 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
216 #define folio_page_idx(folio, p)	(page_to_pfn(p) - folio_pfn(folio))
217 #else
218 #define nth_page(page,n) ((page) + (n))
219 #define folio_page_idx(folio, p)	((p) - &(folio)->page)
220 #endif
221 
222 /* to align the pointer to the (next) page boundary */
223 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
224 
225 /* to align the pointer to the (prev) page boundary */
226 #define PAGE_ALIGN_DOWN(addr) ALIGN_DOWN(addr, PAGE_SIZE)
227 
228 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
229 #define PAGE_ALIGNED(addr)	IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
230 
231 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
lru_to_folio(struct list_head * head)232 static inline struct folio *lru_to_folio(struct list_head *head)
233 {
234 	return list_entry((head)->prev, struct folio, lru);
235 }
236 
237 void setup_initial_init_mm(void *start_code, void *end_code,
238 			   void *end_data, void *brk);
239 
240 /*
241  * Linux kernel virtual memory manager primitives.
242  * The idea being to have a "virtual" mm in the same way
243  * we have a virtual fs - giving a cleaner interface to the
244  * mm details, and allowing different kinds of memory mappings
245  * (from shared memory to executable loading to arbitrary
246  * mmap() functions).
247  */
248 
249 struct vm_area_struct *vm_area_alloc(struct mm_struct *);
250 struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
251 void vm_area_free(struct vm_area_struct *);
252 
253 #ifndef CONFIG_MMU
254 extern struct rb_root nommu_region_tree;
255 extern struct rw_semaphore nommu_region_sem;
256 
257 extern unsigned int kobjsize(const void *objp);
258 #endif
259 
260 /*
261  * vm_flags in vm_area_struct, see mm_types.h.
262  * When changing, update also include/trace/events/mmflags.h
263  */
264 #define VM_NONE		0x00000000
265 
266 #define VM_READ		0x00000001	/* currently active flags */
267 #define VM_WRITE	0x00000002
268 #define VM_EXEC		0x00000004
269 #define VM_SHARED	0x00000008
270 
271 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
272 #define VM_MAYREAD	0x00000010	/* limits for mprotect() etc */
273 #define VM_MAYWRITE	0x00000020
274 #define VM_MAYEXEC	0x00000040
275 #define VM_MAYSHARE	0x00000080
276 
277 #define VM_GROWSDOWN	0x00000100	/* general info on the segment */
278 #define VM_UFFD_MISSING	0x00000200	/* missing pages tracking */
279 #define VM_PFNMAP	0x00000400	/* Page-ranges managed without "struct page", just pure PFN */
280 #define VM_UFFD_WP	0x00001000	/* wrprotect pages tracking */
281 
282 #define VM_LOCKED	0x00002000
283 #define VM_IO           0x00004000	/* Memory mapped I/O or similar */
284 
285 					/* Used by sys_madvise() */
286 #define VM_SEQ_READ	0x00008000	/* App will access data sequentially */
287 #define VM_RAND_READ	0x00010000	/* App will not benefit from clustered reads */
288 
289 #define VM_DONTCOPY	0x00020000      /* Do not copy this vma on fork */
290 #define VM_DONTEXPAND	0x00040000	/* Cannot expand with mremap() */
291 #define VM_LOCKONFAULT	0x00080000	/* Lock the pages covered when they are faulted in */
292 #define VM_ACCOUNT	0x00100000	/* Is a VM accounted object */
293 #define VM_NORESERVE	0x00200000	/* should the VM suppress accounting */
294 #define VM_HUGETLB	0x00400000	/* Huge TLB Page VM */
295 #define VM_SYNC		0x00800000	/* Synchronous page faults */
296 #define VM_ARCH_1	0x01000000	/* Architecture-specific flag */
297 #define VM_WIPEONFORK	0x02000000	/* Wipe VMA contents in child. */
298 #define VM_DONTDUMP	0x04000000	/* Do not include in the core dump */
299 
300 #ifdef CONFIG_MEM_SOFT_DIRTY
301 # define VM_SOFTDIRTY	0x08000000	/* Not soft dirty clean area */
302 #else
303 # define VM_SOFTDIRTY	0
304 #endif
305 
306 #define VM_MIXEDMAP	0x10000000	/* Can contain "struct page" and pure PFN pages */
307 #define VM_HUGEPAGE	0x20000000	/* MADV_HUGEPAGE marked this vma */
308 #define VM_NOHUGEPAGE	0x40000000	/* MADV_NOHUGEPAGE marked this vma */
309 #define VM_MERGEABLE	0x80000000	/* KSM may merge identical pages */
310 
311 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
312 #define VM_HIGH_ARCH_BIT_0	32	/* bit only usable on 64-bit architectures */
313 #define VM_HIGH_ARCH_BIT_1	33	/* bit only usable on 64-bit architectures */
314 #define VM_HIGH_ARCH_BIT_2	34	/* bit only usable on 64-bit architectures */
315 #define VM_HIGH_ARCH_BIT_3	35	/* bit only usable on 64-bit architectures */
316 #define VM_HIGH_ARCH_BIT_4	36	/* bit only usable on 64-bit architectures */
317 #define VM_HIGH_ARCH_0	BIT(VM_HIGH_ARCH_BIT_0)
318 #define VM_HIGH_ARCH_1	BIT(VM_HIGH_ARCH_BIT_1)
319 #define VM_HIGH_ARCH_2	BIT(VM_HIGH_ARCH_BIT_2)
320 #define VM_HIGH_ARCH_3	BIT(VM_HIGH_ARCH_BIT_3)
321 #define VM_HIGH_ARCH_4	BIT(VM_HIGH_ARCH_BIT_4)
322 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
323 
324 #ifdef CONFIG_ARCH_HAS_PKEYS
325 # define VM_PKEY_SHIFT	VM_HIGH_ARCH_BIT_0
326 # define VM_PKEY_BIT0	VM_HIGH_ARCH_0	/* A protection key is a 4-bit value */
327 # define VM_PKEY_BIT1	VM_HIGH_ARCH_1	/* on x86 and 5-bit value on ppc64   */
328 # define VM_PKEY_BIT2	VM_HIGH_ARCH_2
329 # define VM_PKEY_BIT3	VM_HIGH_ARCH_3
330 #ifdef CONFIG_PPC
331 # define VM_PKEY_BIT4  VM_HIGH_ARCH_4
332 #else
333 # define VM_PKEY_BIT4  0
334 #endif
335 #endif /* CONFIG_ARCH_HAS_PKEYS */
336 
337 #if defined(CONFIG_X86)
338 # define VM_PAT		VM_ARCH_1	/* PAT reserves whole VMA at once (x86) */
339 #elif defined(CONFIG_PPC)
340 # define VM_SAO		VM_ARCH_1	/* Strong Access Ordering (powerpc) */
341 #elif defined(CONFIG_PARISC)
342 # define VM_GROWSUP	VM_ARCH_1
343 #elif defined(CONFIG_IA64)
344 # define VM_GROWSUP	VM_ARCH_1
345 #elif defined(CONFIG_SPARC64)
346 # define VM_SPARC_ADI	VM_ARCH_1	/* Uses ADI tag for access control */
347 # define VM_ARCH_CLEAR	VM_SPARC_ADI
348 #elif defined(CONFIG_ARM64)
349 # define VM_ARM64_BTI	VM_ARCH_1	/* BTI guarded page, a.k.a. GP bit */
350 # define VM_ARCH_CLEAR	VM_ARM64_BTI
351 #elif !defined(CONFIG_MMU)
352 # define VM_MAPPED_COPY	VM_ARCH_1	/* T if mapped copy of data (nommu mmap) */
353 #endif
354 
355 #if defined(CONFIG_ARM64_MTE)
356 # define VM_MTE		VM_HIGH_ARCH_0	/* Use Tagged memory for access control */
357 # define VM_MTE_ALLOWED	VM_HIGH_ARCH_1	/* Tagged memory permitted */
358 #else
359 # define VM_MTE		VM_NONE
360 # define VM_MTE_ALLOWED	VM_NONE
361 #endif
362 
363 #ifndef VM_GROWSUP
364 # define VM_GROWSUP	VM_NONE
365 #endif
366 
367 #ifdef CONFIG_HAVE_ARCH_USERFAULTFD_MINOR
368 # define VM_UFFD_MINOR_BIT	37
369 # define VM_UFFD_MINOR		BIT(VM_UFFD_MINOR_BIT)	/* UFFD minor faults */
370 #else /* !CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
371 # define VM_UFFD_MINOR		VM_NONE
372 #endif /* CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
373 
374 /* Bits set in the VMA until the stack is in its final location */
375 #define VM_STACK_INCOMPLETE_SETUP	(VM_RAND_READ | VM_SEQ_READ)
376 
377 #define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0)
378 
379 /* Common data flag combinations */
380 #define VM_DATA_FLAGS_TSK_EXEC	(VM_READ | VM_WRITE | TASK_EXEC | \
381 				 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
382 #define VM_DATA_FLAGS_NON_EXEC	(VM_READ | VM_WRITE | VM_MAYREAD | \
383 				 VM_MAYWRITE | VM_MAYEXEC)
384 #define VM_DATA_FLAGS_EXEC	(VM_READ | VM_WRITE | VM_EXEC | \
385 				 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
386 
387 #ifndef VM_DATA_DEFAULT_FLAGS		/* arch can override this */
388 #define VM_DATA_DEFAULT_FLAGS  VM_DATA_FLAGS_EXEC
389 #endif
390 
391 #ifndef VM_STACK_DEFAULT_FLAGS		/* arch can override this */
392 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
393 #endif
394 
395 #ifdef CONFIG_STACK_GROWSUP
396 #define VM_STACK	VM_GROWSUP
397 #else
398 #define VM_STACK	VM_GROWSDOWN
399 #endif
400 
401 #define VM_STACK_FLAGS	(VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
402 
403 /* VMA basic access permission flags */
404 #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
405 
406 
407 /*
408  * Special vmas that are non-mergable, non-mlock()able.
409  */
410 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
411 
412 /* This mask prevents VMA from being scanned with khugepaged */
413 #define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB)
414 
415 /* This mask defines which mm->def_flags a process can inherit its parent */
416 #define VM_INIT_DEF_MASK	VM_NOHUGEPAGE
417 
418 /* This mask is used to clear all the VMA flags used by mlock */
419 #define VM_LOCKED_CLEAR_MASK	(~(VM_LOCKED | VM_LOCKONFAULT))
420 
421 /* Arch-specific flags to clear when updating VM flags on protection change */
422 #ifndef VM_ARCH_CLEAR
423 # define VM_ARCH_CLEAR	VM_NONE
424 #endif
425 #define VM_FLAGS_CLEAR	(ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
426 
427 /*
428  * mapping from the currently active vm_flags protection bits (the
429  * low four bits) to a page protection mask..
430  */
431 
432 /*
433  * The default fault flags that should be used by most of the
434  * arch-specific page fault handlers.
435  */
436 #define FAULT_FLAG_DEFAULT  (FAULT_FLAG_ALLOW_RETRY | \
437 			     FAULT_FLAG_KILLABLE | \
438 			     FAULT_FLAG_INTERRUPTIBLE)
439 
440 /**
441  * fault_flag_allow_retry_first - check ALLOW_RETRY the first time
442  * @flags: Fault flags.
443  *
444  * This is mostly used for places where we want to try to avoid taking
445  * the mmap_lock for too long a time when waiting for another condition
446  * to change, in which case we can try to be polite to release the
447  * mmap_lock in the first round to avoid potential starvation of other
448  * processes that would also want the mmap_lock.
449  *
450  * Return: true if the page fault allows retry and this is the first
451  * attempt of the fault handling; false otherwise.
452  */
fault_flag_allow_retry_first(enum fault_flag flags)453 static inline bool fault_flag_allow_retry_first(enum fault_flag flags)
454 {
455 	return (flags & FAULT_FLAG_ALLOW_RETRY) &&
456 	    (!(flags & FAULT_FLAG_TRIED));
457 }
458 
459 #define FAULT_FLAG_TRACE \
460 	{ FAULT_FLAG_WRITE,		"WRITE" }, \
461 	{ FAULT_FLAG_MKWRITE,		"MKWRITE" }, \
462 	{ FAULT_FLAG_ALLOW_RETRY,	"ALLOW_RETRY" }, \
463 	{ FAULT_FLAG_RETRY_NOWAIT,	"RETRY_NOWAIT" }, \
464 	{ FAULT_FLAG_KILLABLE,		"KILLABLE" }, \
465 	{ FAULT_FLAG_TRIED,		"TRIED" }, \
466 	{ FAULT_FLAG_USER,		"USER" }, \
467 	{ FAULT_FLAG_REMOTE,		"REMOTE" }, \
468 	{ FAULT_FLAG_INSTRUCTION,	"INSTRUCTION" }, \
469 	{ FAULT_FLAG_INTERRUPTIBLE,	"INTERRUPTIBLE" }
470 
471 /*
472  * vm_fault is filled by the pagefault handler and passed to the vma's
473  * ->fault function. The vma's ->fault is responsible for returning a bitmask
474  * of VM_FAULT_xxx flags that give details about how the fault was handled.
475  *
476  * MM layer fills up gfp_mask for page allocations but fault handler might
477  * alter it if its implementation requires a different allocation context.
478  *
479  * pgoff should be used in favour of virtual_address, if possible.
480  */
481 struct vm_fault {
482 	const struct {
483 		struct vm_area_struct *vma;	/* Target VMA */
484 		gfp_t gfp_mask;			/* gfp mask to be used for allocations */
485 		pgoff_t pgoff;			/* Logical page offset based on vma */
486 		unsigned long address;		/* Faulting virtual address - masked */
487 		unsigned long real_address;	/* Faulting virtual address - unmasked */
488 	};
489 	enum fault_flag flags;		/* FAULT_FLAG_xxx flags
490 					 * XXX: should really be 'const' */
491 	pmd_t *pmd;			/* Pointer to pmd entry matching
492 					 * the 'address' */
493 	pud_t *pud;			/* Pointer to pud entry matching
494 					 * the 'address'
495 					 */
496 	union {
497 		pte_t orig_pte;		/* Value of PTE at the time of fault */
498 		pmd_t orig_pmd;		/* Value of PMD at the time of fault,
499 					 * used by PMD fault only.
500 					 */
501 	};
502 
503 	struct page *cow_page;		/* Page handler may use for COW fault */
504 	struct page *page;		/* ->fault handlers should return a
505 					 * page here, unless VM_FAULT_NOPAGE
506 					 * is set (which is also implied by
507 					 * VM_FAULT_ERROR).
508 					 */
509 	/* These three entries are valid only while holding ptl lock */
510 	pte_t *pte;			/* Pointer to pte entry matching
511 					 * the 'address'. NULL if the page
512 					 * table hasn't been allocated.
513 					 */
514 	spinlock_t *ptl;		/* Page table lock.
515 					 * Protects pte page table if 'pte'
516 					 * is not NULL, otherwise pmd.
517 					 */
518 	pgtable_t prealloc_pte;		/* Pre-allocated pte page table.
519 					 * vm_ops->map_pages() sets up a page
520 					 * table from atomic context.
521 					 * do_fault_around() pre-allocates
522 					 * page table to avoid allocation from
523 					 * atomic context.
524 					 */
525 };
526 
527 /* page entry size for vm->huge_fault() */
528 enum page_entry_size {
529 	PE_SIZE_PTE = 0,
530 	PE_SIZE_PMD,
531 	PE_SIZE_PUD,
532 };
533 
534 /*
535  * These are the virtual MM functions - opening of an area, closing and
536  * unmapping it (needed to keep files on disk up-to-date etc), pointer
537  * to the functions called when a no-page or a wp-page exception occurs.
538  */
539 struct vm_operations_struct {
540 	void (*open)(struct vm_area_struct * area);
541 	/**
542 	 * @close: Called when the VMA is being removed from the MM.
543 	 * Context: User context.  May sleep.  Caller holds mmap_lock.
544 	 */
545 	void (*close)(struct vm_area_struct * area);
546 	/* Called any time before splitting to check if it's allowed */
547 	int (*may_split)(struct vm_area_struct *area, unsigned long addr);
548 	int (*mremap)(struct vm_area_struct *area);
549 	/*
550 	 * Called by mprotect() to make driver-specific permission
551 	 * checks before mprotect() is finalised.   The VMA must not
552 	 * be modified.  Returns 0 if eprotect() can proceed.
553 	 */
554 	int (*mprotect)(struct vm_area_struct *vma, unsigned long start,
555 			unsigned long end, unsigned long newflags);
556 	vm_fault_t (*fault)(struct vm_fault *vmf);
557 	vm_fault_t (*huge_fault)(struct vm_fault *vmf,
558 			enum page_entry_size pe_size);
559 	vm_fault_t (*map_pages)(struct vm_fault *vmf,
560 			pgoff_t start_pgoff, pgoff_t end_pgoff);
561 	unsigned long (*pagesize)(struct vm_area_struct * area);
562 
563 	/* notification that a previously read-only page is about to become
564 	 * writable, if an error is returned it will cause a SIGBUS */
565 	vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
566 
567 	/* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
568 	vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
569 
570 	/* called by access_process_vm when get_user_pages() fails, typically
571 	 * for use by special VMAs. See also generic_access_phys() for a generic
572 	 * implementation useful for any iomem mapping.
573 	 */
574 	int (*access)(struct vm_area_struct *vma, unsigned long addr,
575 		      void *buf, int len, int write);
576 
577 	/* Called by the /proc/PID/maps code to ask the vma whether it
578 	 * has a special name.  Returning non-NULL will also cause this
579 	 * vma to be dumped unconditionally. */
580 	const char *(*name)(struct vm_area_struct *vma);
581 
582 #ifdef CONFIG_NUMA
583 	/*
584 	 * set_policy() op must add a reference to any non-NULL @new mempolicy
585 	 * to hold the policy upon return.  Caller should pass NULL @new to
586 	 * remove a policy and fall back to surrounding context--i.e. do not
587 	 * install a MPOL_DEFAULT policy, nor the task or system default
588 	 * mempolicy.
589 	 */
590 	int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
591 
592 	/*
593 	 * get_policy() op must add reference [mpol_get()] to any policy at
594 	 * (vma,addr) marked as MPOL_SHARED.  The shared policy infrastructure
595 	 * in mm/mempolicy.c will do this automatically.
596 	 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
597 	 * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
598 	 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
599 	 * must return NULL--i.e., do not "fallback" to task or system default
600 	 * policy.
601 	 */
602 	struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
603 					unsigned long addr);
604 #endif
605 	/*
606 	 * Called by vm_normal_page() for special PTEs to find the
607 	 * page for @addr.  This is useful if the default behavior
608 	 * (using pte_page()) would not find the correct page.
609 	 */
610 	struct page *(*find_special_page)(struct vm_area_struct *vma,
611 					  unsigned long addr);
612 };
613 
vma_init(struct vm_area_struct * vma,struct mm_struct * mm)614 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
615 {
616 	static const struct vm_operations_struct dummy_vm_ops = {};
617 
618 	memset(vma, 0, sizeof(*vma));
619 	vma->vm_mm = mm;
620 	vma->vm_ops = &dummy_vm_ops;
621 	INIT_LIST_HEAD(&vma->anon_vma_chain);
622 }
623 
vma_set_anonymous(struct vm_area_struct * vma)624 static inline void vma_set_anonymous(struct vm_area_struct *vma)
625 {
626 	vma->vm_ops = NULL;
627 }
628 
vma_is_anonymous(struct vm_area_struct * vma)629 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
630 {
631 	return !vma->vm_ops;
632 }
633 
vma_is_temporary_stack(struct vm_area_struct * vma)634 static inline bool vma_is_temporary_stack(struct vm_area_struct *vma)
635 {
636 	int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP);
637 
638 	if (!maybe_stack)
639 		return false;
640 
641 	if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) ==
642 						VM_STACK_INCOMPLETE_SETUP)
643 		return true;
644 
645 	return false;
646 }
647 
vma_is_foreign(struct vm_area_struct * vma)648 static inline bool vma_is_foreign(struct vm_area_struct *vma)
649 {
650 	if (!current->mm)
651 		return true;
652 
653 	if (current->mm != vma->vm_mm)
654 		return true;
655 
656 	return false;
657 }
658 
vma_is_accessible(struct vm_area_struct * vma)659 static inline bool vma_is_accessible(struct vm_area_struct *vma)
660 {
661 	return vma->vm_flags & VM_ACCESS_FLAGS;
662 }
663 
664 static inline
vma_find(struct vma_iterator * vmi,unsigned long max)665 struct vm_area_struct *vma_find(struct vma_iterator *vmi, unsigned long max)
666 {
667 	return mas_find(&vmi->mas, max);
668 }
669 
vma_next(struct vma_iterator * vmi)670 static inline struct vm_area_struct *vma_next(struct vma_iterator *vmi)
671 {
672 	/*
673 	 * Uses vma_find() to get the first VMA when the iterator starts.
674 	 * Calling mas_next() could skip the first entry.
675 	 */
676 	return vma_find(vmi, ULONG_MAX);
677 }
678 
vma_prev(struct vma_iterator * vmi)679 static inline struct vm_area_struct *vma_prev(struct vma_iterator *vmi)
680 {
681 	return mas_prev(&vmi->mas, 0);
682 }
683 
vma_iter_addr(struct vma_iterator * vmi)684 static inline unsigned long vma_iter_addr(struct vma_iterator *vmi)
685 {
686 	return vmi->mas.index;
687 }
688 
689 #define for_each_vma(__vmi, __vma)					\
690 	while (((__vma) = vma_next(&(__vmi))) != NULL)
691 
692 /* The MM code likes to work with exclusive end addresses */
693 #define for_each_vma_range(__vmi, __vma, __end)				\
694 	while (((__vma) = vma_find(&(__vmi), (__end) - 1)) != NULL)
695 
696 #ifdef CONFIG_SHMEM
697 /*
698  * The vma_is_shmem is not inline because it is used only by slow
699  * paths in userfault.
700  */
701 bool vma_is_shmem(struct vm_area_struct *vma);
702 #else
vma_is_shmem(struct vm_area_struct * vma)703 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
704 #endif
705 
706 int vma_is_stack_for_current(struct vm_area_struct *vma);
707 
708 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
709 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
710 
711 struct mmu_gather;
712 struct inode;
713 
compound_order(struct page * page)714 static inline unsigned int compound_order(struct page *page)
715 {
716 	if (!PageHead(page))
717 		return 0;
718 	return page[1].compound_order;
719 }
720 
721 /**
722  * folio_order - The allocation order of a folio.
723  * @folio: The folio.
724  *
725  * A folio is composed of 2^order pages.  See get_order() for the definition
726  * of order.
727  *
728  * Return: The order of the folio.
729  */
folio_order(struct folio * folio)730 static inline unsigned int folio_order(struct folio *folio)
731 {
732 	if (!folio_test_large(folio))
733 		return 0;
734 	return folio->_folio_order;
735 }
736 
737 #include <linux/huge_mm.h>
738 
739 /*
740  * Methods to modify the page usage count.
741  *
742  * What counts for a page usage:
743  * - cache mapping   (page->mapping)
744  * - private data    (page->private)
745  * - page mapped in a task's page tables, each mapping
746  *   is counted separately
747  *
748  * Also, many kernel routines increase the page count before a critical
749  * routine so they can be sure the page doesn't go away from under them.
750  */
751 
752 /*
753  * Drop a ref, return true if the refcount fell to zero (the page has no users)
754  */
put_page_testzero(struct page * page)755 static inline int put_page_testzero(struct page *page)
756 {
757 	VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
758 	return page_ref_dec_and_test(page);
759 }
760 
folio_put_testzero(struct folio * folio)761 static inline int folio_put_testzero(struct folio *folio)
762 {
763 	return put_page_testzero(&folio->page);
764 }
765 
766 /*
767  * Try to grab a ref unless the page has a refcount of zero, return false if
768  * that is the case.
769  * This can be called when MMU is off so it must not access
770  * any of the virtual mappings.
771  */
get_page_unless_zero(struct page * page)772 static inline bool get_page_unless_zero(struct page *page)
773 {
774 	return page_ref_add_unless(page, 1, 0);
775 }
776 
777 extern int page_is_ram(unsigned long pfn);
778 
779 enum {
780 	REGION_INTERSECTS,
781 	REGION_DISJOINT,
782 	REGION_MIXED,
783 };
784 
785 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
786 		      unsigned long desc);
787 
788 /* Support for virtually mapped pages */
789 struct page *vmalloc_to_page(const void *addr);
790 unsigned long vmalloc_to_pfn(const void *addr);
791 
792 /*
793  * Determine if an address is within the vmalloc range
794  *
795  * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
796  * is no special casing required.
797  */
798 
799 #ifndef is_ioremap_addr
800 #define is_ioremap_addr(x) is_vmalloc_addr(x)
801 #endif
802 
803 #ifdef CONFIG_MMU
804 extern bool is_vmalloc_addr(const void *x);
805 extern int is_vmalloc_or_module_addr(const void *x);
806 #else
is_vmalloc_addr(const void * x)807 static inline bool is_vmalloc_addr(const void *x)
808 {
809 	return false;
810 }
is_vmalloc_or_module_addr(const void * x)811 static inline int is_vmalloc_or_module_addr(const void *x)
812 {
813 	return 0;
814 }
815 #endif
816 
817 /*
818  * How many times the entire folio is mapped as a single unit (eg by a
819  * PMD or PUD entry).  This is probably not what you want, except for
820  * debugging purposes; look at folio_mapcount() or page_mapcount()
821  * instead.
822  */
folio_entire_mapcount(struct folio * folio)823 static inline int folio_entire_mapcount(struct folio *folio)
824 {
825 	VM_BUG_ON_FOLIO(!folio_test_large(folio), folio);
826 	return atomic_read(folio_mapcount_ptr(folio)) + 1;
827 }
828 
829 /*
830  * Mapcount of compound page as a whole, does not include mapped sub-pages.
831  *
832  * Must be called only for compound pages.
833  */
compound_mapcount(struct page * page)834 static inline int compound_mapcount(struct page *page)
835 {
836 	return folio_entire_mapcount(page_folio(page));
837 }
838 
839 /*
840  * The atomic page->_mapcount, starts from -1: so that transitions
841  * both from it and to it can be tracked, using atomic_inc_and_test
842  * and atomic_add_negative(-1).
843  */
page_mapcount_reset(struct page * page)844 static inline void page_mapcount_reset(struct page *page)
845 {
846 	atomic_set(&(page)->_mapcount, -1);
847 }
848 
849 int __page_mapcount(struct page *page);
850 
851 /*
852  * Mapcount of 0-order page; when compound sub-page, includes
853  * compound_mapcount().
854  *
855  * Result is undefined for pages which cannot be mapped into userspace.
856  * For example SLAB or special types of pages. See function page_has_type().
857  * They use this place in struct page differently.
858  */
page_mapcount(struct page * page)859 static inline int page_mapcount(struct page *page)
860 {
861 	if (unlikely(PageCompound(page)))
862 		return __page_mapcount(page);
863 	return atomic_read(&page->_mapcount) + 1;
864 }
865 
866 int folio_mapcount(struct folio *folio);
867 
868 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
total_mapcount(struct page * page)869 static inline int total_mapcount(struct page *page)
870 {
871 	return folio_mapcount(page_folio(page));
872 }
873 
874 #else
total_mapcount(struct page * page)875 static inline int total_mapcount(struct page *page)
876 {
877 	return page_mapcount(page);
878 }
879 #endif
880 
virt_to_head_page(const void * x)881 static inline struct page *virt_to_head_page(const void *x)
882 {
883 	struct page *page = virt_to_page(x);
884 
885 	return compound_head(page);
886 }
887 
virt_to_folio(const void * x)888 static inline struct folio *virt_to_folio(const void *x)
889 {
890 	struct page *page = virt_to_page(x);
891 
892 	return page_folio(page);
893 }
894 
895 void __folio_put(struct folio *folio);
896 
897 void put_pages_list(struct list_head *pages);
898 
899 void split_page(struct page *page, unsigned int order);
900 void folio_copy(struct folio *dst, struct folio *src);
901 
902 unsigned long nr_free_buffer_pages(void);
903 
904 /*
905  * Compound pages have a destructor function.  Provide a
906  * prototype for that function and accessor functions.
907  * These are _only_ valid on the head of a compound page.
908  */
909 typedef void compound_page_dtor(struct page *);
910 
911 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
912 enum compound_dtor_id {
913 	NULL_COMPOUND_DTOR,
914 	COMPOUND_PAGE_DTOR,
915 #ifdef CONFIG_HUGETLB_PAGE
916 	HUGETLB_PAGE_DTOR,
917 #endif
918 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
919 	TRANSHUGE_PAGE_DTOR,
920 #endif
921 	NR_COMPOUND_DTORS,
922 };
923 extern compound_page_dtor * const compound_page_dtors[NR_COMPOUND_DTORS];
924 
set_compound_page_dtor(struct page * page,enum compound_dtor_id compound_dtor)925 static inline void set_compound_page_dtor(struct page *page,
926 		enum compound_dtor_id compound_dtor)
927 {
928 	VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
929 	page[1].compound_dtor = compound_dtor;
930 }
931 
932 void destroy_large_folio(struct folio *folio);
933 
head_compound_pincount(struct page * head)934 static inline int head_compound_pincount(struct page *head)
935 {
936 	return atomic_read(compound_pincount_ptr(head));
937 }
938 
set_compound_order(struct page * page,unsigned int order)939 static inline void set_compound_order(struct page *page, unsigned int order)
940 {
941 	page[1].compound_order = order;
942 #ifdef CONFIG_64BIT
943 	page[1].compound_nr = 1U << order;
944 #endif
945 }
946 
947 /* Returns the number of pages in this potentially compound page. */
compound_nr(struct page * page)948 static inline unsigned long compound_nr(struct page *page)
949 {
950 	if (!PageHead(page))
951 		return 1;
952 #ifdef CONFIG_64BIT
953 	return page[1].compound_nr;
954 #else
955 	return 1UL << compound_order(page);
956 #endif
957 }
958 
959 /* Returns the number of bytes in this potentially compound page. */
page_size(struct page * page)960 static inline unsigned long page_size(struct page *page)
961 {
962 	return PAGE_SIZE << compound_order(page);
963 }
964 
965 /* Returns the number of bits needed for the number of bytes in a page */
page_shift(struct page * page)966 static inline unsigned int page_shift(struct page *page)
967 {
968 	return PAGE_SHIFT + compound_order(page);
969 }
970 
971 /**
972  * thp_order - Order of a transparent huge page.
973  * @page: Head page of a transparent huge page.
974  */
thp_order(struct page * page)975 static inline unsigned int thp_order(struct page *page)
976 {
977 	VM_BUG_ON_PGFLAGS(PageTail(page), page);
978 	return compound_order(page);
979 }
980 
981 /**
982  * thp_nr_pages - The number of regular pages in this huge page.
983  * @page: The head page of a huge page.
984  */
thp_nr_pages(struct page * page)985 static inline int thp_nr_pages(struct page *page)
986 {
987 	VM_BUG_ON_PGFLAGS(PageTail(page), page);
988 	return compound_nr(page);
989 }
990 
991 /**
992  * thp_size - Size of a transparent huge page.
993  * @page: Head page of a transparent huge page.
994  *
995  * Return: Number of bytes in this page.
996  */
thp_size(struct page * page)997 static inline unsigned long thp_size(struct page *page)
998 {
999 	return PAGE_SIZE << thp_order(page);
1000 }
1001 
1002 void free_compound_page(struct page *page);
1003 
1004 #ifdef CONFIG_MMU
1005 /*
1006  * Do pte_mkwrite, but only if the vma says VM_WRITE.  We do this when
1007  * servicing faults for write access.  In the normal case, do always want
1008  * pte_mkwrite.  But get_user_pages can cause write faults for mappings
1009  * that do not have writing enabled, when used by access_process_vm.
1010  */
maybe_mkwrite(pte_t pte,struct vm_area_struct * vma)1011 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
1012 {
1013 	if (likely(vma->vm_flags & VM_WRITE))
1014 		pte = pte_mkwrite(pte);
1015 	return pte;
1016 }
1017 
1018 vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page);
1019 void do_set_pte(struct vm_fault *vmf, struct page *page, unsigned long addr);
1020 
1021 vm_fault_t finish_fault(struct vm_fault *vmf);
1022 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
1023 #endif
1024 
1025 /*
1026  * Multiple processes may "see" the same page. E.g. for untouched
1027  * mappings of /dev/null, all processes see the same page full of
1028  * zeroes, and text pages of executables and shared libraries have
1029  * only one copy in memory, at most, normally.
1030  *
1031  * For the non-reserved pages, page_count(page) denotes a reference count.
1032  *   page_count() == 0 means the page is free. page->lru is then used for
1033  *   freelist management in the buddy allocator.
1034  *   page_count() > 0  means the page has been allocated.
1035  *
1036  * Pages are allocated by the slab allocator in order to provide memory
1037  * to kmalloc and kmem_cache_alloc. In this case, the management of the
1038  * page, and the fields in 'struct page' are the responsibility of mm/slab.c
1039  * unless a particular usage is carefully commented. (the responsibility of
1040  * freeing the kmalloc memory is the caller's, of course).
1041  *
1042  * A page may be used by anyone else who does a __get_free_page().
1043  * In this case, page_count still tracks the references, and should only
1044  * be used through the normal accessor functions. The top bits of page->flags
1045  * and page->virtual store page management information, but all other fields
1046  * are unused and could be used privately, carefully. The management of this
1047  * page is the responsibility of the one who allocated it, and those who have
1048  * subsequently been given references to it.
1049  *
1050  * The other pages (we may call them "pagecache pages") are completely
1051  * managed by the Linux memory manager: I/O, buffers, swapping etc.
1052  * The following discussion applies only to them.
1053  *
1054  * A pagecache page contains an opaque `private' member, which belongs to the
1055  * page's address_space. Usually, this is the address of a circular list of
1056  * the page's disk buffers. PG_private must be set to tell the VM to call
1057  * into the filesystem to release these pages.
1058  *
1059  * A page may belong to an inode's memory mapping. In this case, page->mapping
1060  * is the pointer to the inode, and page->index is the file offset of the page,
1061  * in units of PAGE_SIZE.
1062  *
1063  * If pagecache pages are not associated with an inode, they are said to be
1064  * anonymous pages. These may become associated with the swapcache, and in that
1065  * case PG_swapcache is set, and page->private is an offset into the swapcache.
1066  *
1067  * In either case (swapcache or inode backed), the pagecache itself holds one
1068  * reference to the page. Setting PG_private should also increment the
1069  * refcount. The each user mapping also has a reference to the page.
1070  *
1071  * The pagecache pages are stored in a per-mapping radix tree, which is
1072  * rooted at mapping->i_pages, and indexed by offset.
1073  * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
1074  * lists, we instead now tag pages as dirty/writeback in the radix tree.
1075  *
1076  * All pagecache pages may be subject to I/O:
1077  * - inode pages may need to be read from disk,
1078  * - inode pages which have been modified and are MAP_SHARED may need
1079  *   to be written back to the inode on disk,
1080  * - anonymous pages (including MAP_PRIVATE file mappings) which have been
1081  *   modified may need to be swapped out to swap space and (later) to be read
1082  *   back into memory.
1083  */
1084 
1085 #if defined(CONFIG_ZONE_DEVICE) && defined(CONFIG_FS_DAX)
1086 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
1087 
1088 bool __put_devmap_managed_page_refs(struct page *page, int refs);
put_devmap_managed_page_refs(struct page * page,int refs)1089 static inline bool put_devmap_managed_page_refs(struct page *page, int refs)
1090 {
1091 	if (!static_branch_unlikely(&devmap_managed_key))
1092 		return false;
1093 	if (!is_zone_device_page(page))
1094 		return false;
1095 	return __put_devmap_managed_page_refs(page, refs);
1096 }
1097 #else /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
put_devmap_managed_page_refs(struct page * page,int refs)1098 static inline bool put_devmap_managed_page_refs(struct page *page, int refs)
1099 {
1100 	return false;
1101 }
1102 #endif /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1103 
put_devmap_managed_page(struct page * page)1104 static inline bool put_devmap_managed_page(struct page *page)
1105 {
1106 	return put_devmap_managed_page_refs(page, 1);
1107 }
1108 
1109 /* 127: arbitrary random number, small enough to assemble well */
1110 #define folio_ref_zero_or_close_to_overflow(folio) \
1111 	((unsigned int) folio_ref_count(folio) + 127u <= 127u)
1112 
1113 /**
1114  * folio_get - Increment the reference count on a folio.
1115  * @folio: The folio.
1116  *
1117  * Context: May be called in any context, as long as you know that
1118  * you have a refcount on the folio.  If you do not already have one,
1119  * folio_try_get() may be the right interface for you to use.
1120  */
folio_get(struct folio * folio)1121 static inline void folio_get(struct folio *folio)
1122 {
1123 	VM_BUG_ON_FOLIO(folio_ref_zero_or_close_to_overflow(folio), folio);
1124 	folio_ref_inc(folio);
1125 }
1126 
get_page(struct page * page)1127 static inline void get_page(struct page *page)
1128 {
1129 	folio_get(page_folio(page));
1130 }
1131 
1132 bool __must_check try_grab_page(struct page *page, unsigned int flags);
1133 
try_get_page(struct page * page)1134 static inline __must_check bool try_get_page(struct page *page)
1135 {
1136 	page = compound_head(page);
1137 	if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1138 		return false;
1139 	page_ref_inc(page);
1140 	return true;
1141 }
1142 
1143 /**
1144  * folio_put - Decrement the reference count on a folio.
1145  * @folio: The folio.
1146  *
1147  * If the folio's reference count reaches zero, the memory will be
1148  * released back to the page allocator and may be used by another
1149  * allocation immediately.  Do not access the memory or the struct folio
1150  * after calling folio_put() unless you can be sure that it wasn't the
1151  * last reference.
1152  *
1153  * Context: May be called in process or interrupt context, but not in NMI
1154  * context.  May be called while holding a spinlock.
1155  */
folio_put(struct folio * folio)1156 static inline void folio_put(struct folio *folio)
1157 {
1158 	if (folio_put_testzero(folio))
1159 		__folio_put(folio);
1160 }
1161 
1162 /**
1163  * folio_put_refs - Reduce the reference count on a folio.
1164  * @folio: The folio.
1165  * @refs: The amount to subtract from the folio's reference count.
1166  *
1167  * If the folio's reference count reaches zero, the memory will be
1168  * released back to the page allocator and may be used by another
1169  * allocation immediately.  Do not access the memory or the struct folio
1170  * after calling folio_put_refs() unless you can be sure that these weren't
1171  * the last references.
1172  *
1173  * Context: May be called in process or interrupt context, but not in NMI
1174  * context.  May be called while holding a spinlock.
1175  */
folio_put_refs(struct folio * folio,int refs)1176 static inline void folio_put_refs(struct folio *folio, int refs)
1177 {
1178 	if (folio_ref_sub_and_test(folio, refs))
1179 		__folio_put(folio);
1180 }
1181 
1182 void release_pages(struct page **pages, int nr);
1183 
1184 /**
1185  * folios_put - Decrement the reference count on an array of folios.
1186  * @folios: The folios.
1187  * @nr: How many folios there are.
1188  *
1189  * Like folio_put(), but for an array of folios.  This is more efficient
1190  * than writing the loop yourself as it will optimise the locks which
1191  * need to be taken if the folios are freed.
1192  *
1193  * Context: May be called in process or interrupt context, but not in NMI
1194  * context.  May be called while holding a spinlock.
1195  */
folios_put(struct folio ** folios,unsigned int nr)1196 static inline void folios_put(struct folio **folios, unsigned int nr)
1197 {
1198 	release_pages((struct page **)folios, nr);
1199 }
1200 
put_page(struct page * page)1201 static inline void put_page(struct page *page)
1202 {
1203 	struct folio *folio = page_folio(page);
1204 
1205 	/*
1206 	 * For some devmap managed pages we need to catch refcount transition
1207 	 * from 2 to 1:
1208 	 */
1209 	if (put_devmap_managed_page(&folio->page))
1210 		return;
1211 	folio_put(folio);
1212 }
1213 
1214 /*
1215  * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
1216  * the page's refcount so that two separate items are tracked: the original page
1217  * reference count, and also a new count of how many pin_user_pages() calls were
1218  * made against the page. ("gup-pinned" is another term for the latter).
1219  *
1220  * With this scheme, pin_user_pages() becomes special: such pages are marked as
1221  * distinct from normal pages. As such, the unpin_user_page() call (and its
1222  * variants) must be used in order to release gup-pinned pages.
1223  *
1224  * Choice of value:
1225  *
1226  * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
1227  * counts with respect to pin_user_pages() and unpin_user_page() becomes
1228  * simpler, due to the fact that adding an even power of two to the page
1229  * refcount has the effect of using only the upper N bits, for the code that
1230  * counts up using the bias value. This means that the lower bits are left for
1231  * the exclusive use of the original code that increments and decrements by one
1232  * (or at least, by much smaller values than the bias value).
1233  *
1234  * Of course, once the lower bits overflow into the upper bits (and this is
1235  * OK, because subtraction recovers the original values), then visual inspection
1236  * no longer suffices to directly view the separate counts. However, for normal
1237  * applications that don't have huge page reference counts, this won't be an
1238  * issue.
1239  *
1240  * Locking: the lockless algorithm described in folio_try_get_rcu()
1241  * provides safe operation for get_user_pages(), page_mkclean() and
1242  * other calls that race to set up page table entries.
1243  */
1244 #define GUP_PIN_COUNTING_BIAS (1U << 10)
1245 
1246 void unpin_user_page(struct page *page);
1247 void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1248 				 bool make_dirty);
1249 void unpin_user_page_range_dirty_lock(struct page *page, unsigned long npages,
1250 				      bool make_dirty);
1251 void unpin_user_pages(struct page **pages, unsigned long npages);
1252 
is_cow_mapping(vm_flags_t flags)1253 static inline bool is_cow_mapping(vm_flags_t flags)
1254 {
1255 	return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
1256 }
1257 
1258 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1259 #define SECTION_IN_PAGE_FLAGS
1260 #endif
1261 
1262 /*
1263  * The identification function is mainly used by the buddy allocator for
1264  * determining if two pages could be buddies. We are not really identifying
1265  * the zone since we could be using the section number id if we do not have
1266  * node id available in page flags.
1267  * We only guarantee that it will return the same value for two combinable
1268  * pages in a zone.
1269  */
page_zone_id(struct page * page)1270 static inline int page_zone_id(struct page *page)
1271 {
1272 	return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1273 }
1274 
1275 #ifdef NODE_NOT_IN_PAGE_FLAGS
1276 extern int page_to_nid(const struct page *page);
1277 #else
page_to_nid(const struct page * page)1278 static inline int page_to_nid(const struct page *page)
1279 {
1280 	struct page *p = (struct page *)page;
1281 
1282 	return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1283 }
1284 #endif
1285 
folio_nid(const struct folio * folio)1286 static inline int folio_nid(const struct folio *folio)
1287 {
1288 	return page_to_nid(&folio->page);
1289 }
1290 
1291 #ifdef CONFIG_NUMA_BALANCING
1292 /* page access time bits needs to hold at least 4 seconds */
1293 #define PAGE_ACCESS_TIME_MIN_BITS	12
1294 #if LAST_CPUPID_SHIFT < PAGE_ACCESS_TIME_MIN_BITS
1295 #define PAGE_ACCESS_TIME_BUCKETS				\
1296 	(PAGE_ACCESS_TIME_MIN_BITS - LAST_CPUPID_SHIFT)
1297 #else
1298 #define PAGE_ACCESS_TIME_BUCKETS	0
1299 #endif
1300 
1301 #define PAGE_ACCESS_TIME_MASK				\
1302 	(LAST_CPUPID_MASK << PAGE_ACCESS_TIME_BUCKETS)
1303 
cpu_pid_to_cpupid(int cpu,int pid)1304 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1305 {
1306 	return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1307 }
1308 
cpupid_to_pid(int cpupid)1309 static inline int cpupid_to_pid(int cpupid)
1310 {
1311 	return cpupid & LAST__PID_MASK;
1312 }
1313 
cpupid_to_cpu(int cpupid)1314 static inline int cpupid_to_cpu(int cpupid)
1315 {
1316 	return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1317 }
1318 
cpupid_to_nid(int cpupid)1319 static inline int cpupid_to_nid(int cpupid)
1320 {
1321 	return cpu_to_node(cpupid_to_cpu(cpupid));
1322 }
1323 
cpupid_pid_unset(int cpupid)1324 static inline bool cpupid_pid_unset(int cpupid)
1325 {
1326 	return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1327 }
1328 
cpupid_cpu_unset(int cpupid)1329 static inline bool cpupid_cpu_unset(int cpupid)
1330 {
1331 	return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1332 }
1333 
__cpupid_match_pid(pid_t task_pid,int cpupid)1334 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1335 {
1336 	return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1337 }
1338 
1339 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1340 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
page_cpupid_xchg_last(struct page * page,int cpupid)1341 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1342 {
1343 	return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1344 }
1345 
page_cpupid_last(struct page * page)1346 static inline int page_cpupid_last(struct page *page)
1347 {
1348 	return page->_last_cpupid;
1349 }
page_cpupid_reset_last(struct page * page)1350 static inline void page_cpupid_reset_last(struct page *page)
1351 {
1352 	page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1353 }
1354 #else
page_cpupid_last(struct page * page)1355 static inline int page_cpupid_last(struct page *page)
1356 {
1357 	return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1358 }
1359 
1360 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1361 
page_cpupid_reset_last(struct page * page)1362 static inline void page_cpupid_reset_last(struct page *page)
1363 {
1364 	page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1365 }
1366 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1367 
xchg_page_access_time(struct page * page,int time)1368 static inline int xchg_page_access_time(struct page *page, int time)
1369 {
1370 	int last_time;
1371 
1372 	last_time = page_cpupid_xchg_last(page, time >> PAGE_ACCESS_TIME_BUCKETS);
1373 	return last_time << PAGE_ACCESS_TIME_BUCKETS;
1374 }
1375 #else /* !CONFIG_NUMA_BALANCING */
page_cpupid_xchg_last(struct page * page,int cpupid)1376 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1377 {
1378 	return page_to_nid(page); /* XXX */
1379 }
1380 
xchg_page_access_time(struct page * page,int time)1381 static inline int xchg_page_access_time(struct page *page, int time)
1382 {
1383 	return 0;
1384 }
1385 
page_cpupid_last(struct page * page)1386 static inline int page_cpupid_last(struct page *page)
1387 {
1388 	return page_to_nid(page); /* XXX */
1389 }
1390 
cpupid_to_nid(int cpupid)1391 static inline int cpupid_to_nid(int cpupid)
1392 {
1393 	return -1;
1394 }
1395 
cpupid_to_pid(int cpupid)1396 static inline int cpupid_to_pid(int cpupid)
1397 {
1398 	return -1;
1399 }
1400 
cpupid_to_cpu(int cpupid)1401 static inline int cpupid_to_cpu(int cpupid)
1402 {
1403 	return -1;
1404 }
1405 
cpu_pid_to_cpupid(int nid,int pid)1406 static inline int cpu_pid_to_cpupid(int nid, int pid)
1407 {
1408 	return -1;
1409 }
1410 
cpupid_pid_unset(int cpupid)1411 static inline bool cpupid_pid_unset(int cpupid)
1412 {
1413 	return true;
1414 }
1415 
page_cpupid_reset_last(struct page * page)1416 static inline void page_cpupid_reset_last(struct page *page)
1417 {
1418 }
1419 
cpupid_match_pid(struct task_struct * task,int cpupid)1420 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1421 {
1422 	return false;
1423 }
1424 #endif /* CONFIG_NUMA_BALANCING */
1425 
1426 #if defined(CONFIG_KASAN_SW_TAGS) || defined(CONFIG_KASAN_HW_TAGS)
1427 
1428 /*
1429  * KASAN per-page tags are stored xor'ed with 0xff. This allows to avoid
1430  * setting tags for all pages to native kernel tag value 0xff, as the default
1431  * value 0x00 maps to 0xff.
1432  */
1433 
page_kasan_tag(const struct page * page)1434 static inline u8 page_kasan_tag(const struct page *page)
1435 {
1436 	u8 tag = 0xff;
1437 
1438 	if (kasan_enabled()) {
1439 		tag = (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1440 		tag ^= 0xff;
1441 	}
1442 
1443 	return tag;
1444 }
1445 
page_kasan_tag_set(struct page * page,u8 tag)1446 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1447 {
1448 	unsigned long old_flags, flags;
1449 
1450 	if (!kasan_enabled())
1451 		return;
1452 
1453 	tag ^= 0xff;
1454 	old_flags = READ_ONCE(page->flags);
1455 	do {
1456 		flags = old_flags;
1457 		flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1458 		flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1459 	} while (unlikely(!try_cmpxchg(&page->flags, &old_flags, flags)));
1460 }
1461 
page_kasan_tag_reset(struct page * page)1462 static inline void page_kasan_tag_reset(struct page *page)
1463 {
1464 	if (kasan_enabled())
1465 		page_kasan_tag_set(page, 0xff);
1466 }
1467 
1468 #else /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1469 
page_kasan_tag(const struct page * page)1470 static inline u8 page_kasan_tag(const struct page *page)
1471 {
1472 	return 0xff;
1473 }
1474 
page_kasan_tag_set(struct page * page,u8 tag)1475 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
page_kasan_tag_reset(struct page * page)1476 static inline void page_kasan_tag_reset(struct page *page) { }
1477 
1478 #endif /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1479 
page_zone(const struct page * page)1480 static inline struct zone *page_zone(const struct page *page)
1481 {
1482 	return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1483 }
1484 
page_pgdat(const struct page * page)1485 static inline pg_data_t *page_pgdat(const struct page *page)
1486 {
1487 	return NODE_DATA(page_to_nid(page));
1488 }
1489 
folio_zone(const struct folio * folio)1490 static inline struct zone *folio_zone(const struct folio *folio)
1491 {
1492 	return page_zone(&folio->page);
1493 }
1494 
folio_pgdat(const struct folio * folio)1495 static inline pg_data_t *folio_pgdat(const struct folio *folio)
1496 {
1497 	return page_pgdat(&folio->page);
1498 }
1499 
1500 #ifdef SECTION_IN_PAGE_FLAGS
set_page_section(struct page * page,unsigned long section)1501 static inline void set_page_section(struct page *page, unsigned long section)
1502 {
1503 	page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1504 	page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1505 }
1506 
page_to_section(const struct page * page)1507 static inline unsigned long page_to_section(const struct page *page)
1508 {
1509 	return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1510 }
1511 #endif
1512 
1513 /**
1514  * folio_pfn - Return the Page Frame Number of a folio.
1515  * @folio: The folio.
1516  *
1517  * A folio may contain multiple pages.  The pages have consecutive
1518  * Page Frame Numbers.
1519  *
1520  * Return: The Page Frame Number of the first page in the folio.
1521  */
folio_pfn(struct folio * folio)1522 static inline unsigned long folio_pfn(struct folio *folio)
1523 {
1524 	return page_to_pfn(&folio->page);
1525 }
1526 
pfn_folio(unsigned long pfn)1527 static inline struct folio *pfn_folio(unsigned long pfn)
1528 {
1529 	return page_folio(pfn_to_page(pfn));
1530 }
1531 
folio_pincount_ptr(struct folio * folio)1532 static inline atomic_t *folio_pincount_ptr(struct folio *folio)
1533 {
1534 	return &folio_page(folio, 1)->compound_pincount;
1535 }
1536 
1537 /**
1538  * folio_maybe_dma_pinned - Report if a folio may be pinned for DMA.
1539  * @folio: The folio.
1540  *
1541  * This function checks if a folio has been pinned via a call to
1542  * a function in the pin_user_pages() family.
1543  *
1544  * For small folios, the return value is partially fuzzy: false is not fuzzy,
1545  * because it means "definitely not pinned for DMA", but true means "probably
1546  * pinned for DMA, but possibly a false positive due to having at least
1547  * GUP_PIN_COUNTING_BIAS worth of normal folio references".
1548  *
1549  * False positives are OK, because: a) it's unlikely for a folio to
1550  * get that many refcounts, and b) all the callers of this routine are
1551  * expected to be able to deal gracefully with a false positive.
1552  *
1553  * For large folios, the result will be exactly correct. That's because
1554  * we have more tracking data available: the compound_pincount is used
1555  * instead of the GUP_PIN_COUNTING_BIAS scheme.
1556  *
1557  * For more information, please see Documentation/core-api/pin_user_pages.rst.
1558  *
1559  * Return: True, if it is likely that the page has been "dma-pinned".
1560  * False, if the page is definitely not dma-pinned.
1561  */
folio_maybe_dma_pinned(struct folio * folio)1562 static inline bool folio_maybe_dma_pinned(struct folio *folio)
1563 {
1564 	if (folio_test_large(folio))
1565 		return atomic_read(folio_pincount_ptr(folio)) > 0;
1566 
1567 	/*
1568 	 * folio_ref_count() is signed. If that refcount overflows, then
1569 	 * folio_ref_count() returns a negative value, and callers will avoid
1570 	 * further incrementing the refcount.
1571 	 *
1572 	 * Here, for that overflow case, use the sign bit to count a little
1573 	 * bit higher via unsigned math, and thus still get an accurate result.
1574 	 */
1575 	return ((unsigned int)folio_ref_count(folio)) >=
1576 		GUP_PIN_COUNTING_BIAS;
1577 }
1578 
page_maybe_dma_pinned(struct page * page)1579 static inline bool page_maybe_dma_pinned(struct page *page)
1580 {
1581 	return folio_maybe_dma_pinned(page_folio(page));
1582 }
1583 
1584 /*
1585  * This should most likely only be called during fork() to see whether we
1586  * should break the cow immediately for an anon page on the src mm.
1587  *
1588  * The caller has to hold the PT lock and the vma->vm_mm->->write_protect_seq.
1589  */
page_needs_cow_for_dma(struct vm_area_struct * vma,struct page * page)1590 static inline bool page_needs_cow_for_dma(struct vm_area_struct *vma,
1591 					  struct page *page)
1592 {
1593 	VM_BUG_ON(!(raw_read_seqcount(&vma->vm_mm->write_protect_seq) & 1));
1594 
1595 	if (!test_bit(MMF_HAS_PINNED, &vma->vm_mm->flags))
1596 		return false;
1597 
1598 	return page_maybe_dma_pinned(page);
1599 }
1600 
1601 /* MIGRATE_CMA and ZONE_MOVABLE do not allow pin pages */
1602 #ifdef CONFIG_MIGRATION
is_longterm_pinnable_page(struct page * page)1603 static inline bool is_longterm_pinnable_page(struct page *page)
1604 {
1605 #ifdef CONFIG_CMA
1606 	int mt = get_pageblock_migratetype(page);
1607 
1608 	if (mt == MIGRATE_CMA || mt == MIGRATE_ISOLATE)
1609 		return false;
1610 #endif
1611 	/* The zero page may always be pinned */
1612 	if (is_zero_pfn(page_to_pfn(page)))
1613 		return true;
1614 
1615 	/* Coherent device memory must always allow eviction. */
1616 	if (is_device_coherent_page(page))
1617 		return false;
1618 
1619 	/* Otherwise, non-movable zone pages can be pinned. */
1620 	return !is_zone_movable_page(page);
1621 }
1622 #else
is_longterm_pinnable_page(struct page * page)1623 static inline bool is_longterm_pinnable_page(struct page *page)
1624 {
1625 	return true;
1626 }
1627 #endif
1628 
folio_is_longterm_pinnable(struct folio * folio)1629 static inline bool folio_is_longterm_pinnable(struct folio *folio)
1630 {
1631 	return is_longterm_pinnable_page(&folio->page);
1632 }
1633 
set_page_zone(struct page * page,enum zone_type zone)1634 static inline void set_page_zone(struct page *page, enum zone_type zone)
1635 {
1636 	page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1637 	page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1638 }
1639 
set_page_node(struct page * page,unsigned long node)1640 static inline void set_page_node(struct page *page, unsigned long node)
1641 {
1642 	page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1643 	page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1644 }
1645 
set_page_links(struct page * page,enum zone_type zone,unsigned long node,unsigned long pfn)1646 static inline void set_page_links(struct page *page, enum zone_type zone,
1647 	unsigned long node, unsigned long pfn)
1648 {
1649 	set_page_zone(page, zone);
1650 	set_page_node(page, node);
1651 #ifdef SECTION_IN_PAGE_FLAGS
1652 	set_page_section(page, pfn_to_section_nr(pfn));
1653 #endif
1654 }
1655 
1656 /**
1657  * folio_nr_pages - The number of pages in the folio.
1658  * @folio: The folio.
1659  *
1660  * Return: A positive power of two.
1661  */
folio_nr_pages(struct folio * folio)1662 static inline long folio_nr_pages(struct folio *folio)
1663 {
1664 	if (!folio_test_large(folio))
1665 		return 1;
1666 #ifdef CONFIG_64BIT
1667 	return folio->_folio_nr_pages;
1668 #else
1669 	return 1L << folio->_folio_order;
1670 #endif
1671 }
1672 
1673 /**
1674  * folio_next - Move to the next physical folio.
1675  * @folio: The folio we're currently operating on.
1676  *
1677  * If you have physically contiguous memory which may span more than
1678  * one folio (eg a &struct bio_vec), use this function to move from one
1679  * folio to the next.  Do not use it if the memory is only virtually
1680  * contiguous as the folios are almost certainly not adjacent to each
1681  * other.  This is the folio equivalent to writing ``page++``.
1682  *
1683  * Context: We assume that the folios are refcounted and/or locked at a
1684  * higher level and do not adjust the reference counts.
1685  * Return: The next struct folio.
1686  */
folio_next(struct folio * folio)1687 static inline struct folio *folio_next(struct folio *folio)
1688 {
1689 	return (struct folio *)folio_page(folio, folio_nr_pages(folio));
1690 }
1691 
1692 /**
1693  * folio_shift - The size of the memory described by this folio.
1694  * @folio: The folio.
1695  *
1696  * A folio represents a number of bytes which is a power-of-two in size.
1697  * This function tells you which power-of-two the folio is.  See also
1698  * folio_size() and folio_order().
1699  *
1700  * Context: The caller should have a reference on the folio to prevent
1701  * it from being split.  It is not necessary for the folio to be locked.
1702  * Return: The base-2 logarithm of the size of this folio.
1703  */
folio_shift(struct folio * folio)1704 static inline unsigned int folio_shift(struct folio *folio)
1705 {
1706 	return PAGE_SHIFT + folio_order(folio);
1707 }
1708 
1709 /**
1710  * folio_size - The number of bytes in a folio.
1711  * @folio: The folio.
1712  *
1713  * Context: The caller should have a reference on the folio to prevent
1714  * it from being split.  It is not necessary for the folio to be locked.
1715  * Return: The number of bytes in this folio.
1716  */
folio_size(struct folio * folio)1717 static inline size_t folio_size(struct folio *folio)
1718 {
1719 	return PAGE_SIZE << folio_order(folio);
1720 }
1721 
1722 #ifndef HAVE_ARCH_MAKE_PAGE_ACCESSIBLE
arch_make_page_accessible(struct page * page)1723 static inline int arch_make_page_accessible(struct page *page)
1724 {
1725 	return 0;
1726 }
1727 #endif
1728 
1729 #ifndef HAVE_ARCH_MAKE_FOLIO_ACCESSIBLE
arch_make_folio_accessible(struct folio * folio)1730 static inline int arch_make_folio_accessible(struct folio *folio)
1731 {
1732 	int ret;
1733 	long i, nr = folio_nr_pages(folio);
1734 
1735 	for (i = 0; i < nr; i++) {
1736 		ret = arch_make_page_accessible(folio_page(folio, i));
1737 		if (ret)
1738 			break;
1739 	}
1740 
1741 	return ret;
1742 }
1743 #endif
1744 
1745 /*
1746  * Some inline functions in vmstat.h depend on page_zone()
1747  */
1748 #include <linux/vmstat.h>
1749 
lowmem_page_address(const struct page * page)1750 static __always_inline void *lowmem_page_address(const struct page *page)
1751 {
1752 	return page_to_virt(page);
1753 }
1754 
1755 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1756 #define HASHED_PAGE_VIRTUAL
1757 #endif
1758 
1759 #if defined(WANT_PAGE_VIRTUAL)
page_address(const struct page * page)1760 static inline void *page_address(const struct page *page)
1761 {
1762 	return page->virtual;
1763 }
set_page_address(struct page * page,void * address)1764 static inline void set_page_address(struct page *page, void *address)
1765 {
1766 	page->virtual = address;
1767 }
1768 #define page_address_init()  do { } while(0)
1769 #endif
1770 
1771 #if defined(HASHED_PAGE_VIRTUAL)
1772 void *page_address(const struct page *page);
1773 void set_page_address(struct page *page, void *virtual);
1774 void page_address_init(void);
1775 #endif
1776 
1777 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1778 #define page_address(page) lowmem_page_address(page)
1779 #define set_page_address(page, address)  do { } while(0)
1780 #define page_address_init()  do { } while(0)
1781 #endif
1782 
folio_address(const struct folio * folio)1783 static inline void *folio_address(const struct folio *folio)
1784 {
1785 	return page_address(&folio->page);
1786 }
1787 
1788 extern void *page_rmapping(struct page *page);
1789 extern pgoff_t __page_file_index(struct page *page);
1790 
1791 /*
1792  * Return the pagecache index of the passed page.  Regular pagecache pages
1793  * use ->index whereas swapcache pages use swp_offset(->private)
1794  */
page_index(struct page * page)1795 static inline pgoff_t page_index(struct page *page)
1796 {
1797 	if (unlikely(PageSwapCache(page)))
1798 		return __page_file_index(page);
1799 	return page->index;
1800 }
1801 
1802 bool page_mapped(struct page *page);
1803 bool folio_mapped(struct folio *folio);
1804 
1805 /*
1806  * Return true only if the page has been allocated with
1807  * ALLOC_NO_WATERMARKS and the low watermark was not
1808  * met implying that the system is under some pressure.
1809  */
page_is_pfmemalloc(const struct page * page)1810 static inline bool page_is_pfmemalloc(const struct page *page)
1811 {
1812 	/*
1813 	 * lru.next has bit 1 set if the page is allocated from the
1814 	 * pfmemalloc reserves.  Callers may simply overwrite it if
1815 	 * they do not need to preserve that information.
1816 	 */
1817 	return (uintptr_t)page->lru.next & BIT(1);
1818 }
1819 
1820 /*
1821  * Only to be called by the page allocator on a freshly allocated
1822  * page.
1823  */
set_page_pfmemalloc(struct page * page)1824 static inline void set_page_pfmemalloc(struct page *page)
1825 {
1826 	page->lru.next = (void *)BIT(1);
1827 }
1828 
clear_page_pfmemalloc(struct page * page)1829 static inline void clear_page_pfmemalloc(struct page *page)
1830 {
1831 	page->lru.next = NULL;
1832 }
1833 
1834 /*
1835  * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1836  */
1837 extern void pagefault_out_of_memory(void);
1838 
1839 #define offset_in_page(p)	((unsigned long)(p) & ~PAGE_MASK)
1840 #define offset_in_thp(page, p)	((unsigned long)(p) & (thp_size(page) - 1))
1841 #define offset_in_folio(folio, p) ((unsigned long)(p) & (folio_size(folio) - 1))
1842 
1843 /*
1844  * Flags passed to show_mem() and show_free_areas() to suppress output in
1845  * various contexts.
1846  */
1847 #define SHOW_MEM_FILTER_NODES		(0x0001u)	/* disallowed nodes */
1848 
1849 extern void __show_free_areas(unsigned int flags, nodemask_t *nodemask, int max_zone_idx);
show_free_areas(unsigned int flags,nodemask_t * nodemask)1850 static void __maybe_unused show_free_areas(unsigned int flags, nodemask_t *nodemask)
1851 {
1852 	__show_free_areas(flags, nodemask, MAX_NR_ZONES - 1);
1853 }
1854 
1855 /*
1856  * Parameter block passed down to zap_pte_range in exceptional cases.
1857  */
1858 struct zap_details {
1859 	struct folio *single_folio;	/* Locked folio to be unmapped */
1860 	bool even_cows;			/* Zap COWed private pages too? */
1861 	zap_flags_t zap_flags;		/* Extra flags for zapping */
1862 };
1863 
1864 /*
1865  * Whether to drop the pte markers, for example, the uffd-wp information for
1866  * file-backed memory.  This should only be specified when we will completely
1867  * drop the page in the mm, either by truncation or unmapping of the vma.  By
1868  * default, the flag is not set.
1869  */
1870 #define  ZAP_FLAG_DROP_MARKER        ((__force zap_flags_t) BIT(0))
1871 /* Set in unmap_vmas() to indicate a final unmap call.  Only used by hugetlb */
1872 #define  ZAP_FLAG_UNMAP              ((__force zap_flags_t) BIT(1))
1873 
1874 #ifdef CONFIG_MMU
1875 extern bool can_do_mlock(void);
1876 #else
can_do_mlock(void)1877 static inline bool can_do_mlock(void) { return false; }
1878 #endif
1879 extern int user_shm_lock(size_t, struct ucounts *);
1880 extern void user_shm_unlock(size_t, struct ucounts *);
1881 
1882 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1883 			     pte_t pte);
1884 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1885 				pmd_t pmd);
1886 
1887 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1888 		  unsigned long size);
1889 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1890 		    unsigned long size);
1891 void zap_page_range_single(struct vm_area_struct *vma, unsigned long address,
1892 			   unsigned long size, struct zap_details *details);
1893 void unmap_vmas(struct mmu_gather *tlb, struct maple_tree *mt,
1894 		struct vm_area_struct *start_vma, unsigned long start,
1895 		unsigned long end);
1896 
1897 struct mmu_notifier_range;
1898 
1899 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1900 		unsigned long end, unsigned long floor, unsigned long ceiling);
1901 int
1902 copy_page_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma);
1903 int follow_pte(struct mm_struct *mm, unsigned long address,
1904 	       pte_t **ptepp, spinlock_t **ptlp);
1905 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1906 	unsigned long *pfn);
1907 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1908 		unsigned int flags, unsigned long *prot, resource_size_t *phys);
1909 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1910 			void *buf, int len, int write);
1911 
1912 extern void truncate_pagecache(struct inode *inode, loff_t new);
1913 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1914 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1915 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1916 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1917 
1918 #ifdef CONFIG_MMU
1919 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1920 				  unsigned long address, unsigned int flags,
1921 				  struct pt_regs *regs);
1922 extern int fixup_user_fault(struct mm_struct *mm,
1923 			    unsigned long address, unsigned int fault_flags,
1924 			    bool *unlocked);
1925 void unmap_mapping_pages(struct address_space *mapping,
1926 		pgoff_t start, pgoff_t nr, bool even_cows);
1927 void unmap_mapping_range(struct address_space *mapping,
1928 		loff_t const holebegin, loff_t const holelen, int even_cows);
1929 #else
handle_mm_fault(struct vm_area_struct * vma,unsigned long address,unsigned int flags,struct pt_regs * regs)1930 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1931 					 unsigned long address, unsigned int flags,
1932 					 struct pt_regs *regs)
1933 {
1934 	/* should never happen if there's no MMU */
1935 	BUG();
1936 	return VM_FAULT_SIGBUS;
1937 }
fixup_user_fault(struct mm_struct * mm,unsigned long address,unsigned int fault_flags,bool * unlocked)1938 static inline int fixup_user_fault(struct mm_struct *mm, unsigned long address,
1939 		unsigned int fault_flags, bool *unlocked)
1940 {
1941 	/* should never happen if there's no MMU */
1942 	BUG();
1943 	return -EFAULT;
1944 }
unmap_mapping_pages(struct address_space * mapping,pgoff_t start,pgoff_t nr,bool even_cows)1945 static inline void unmap_mapping_pages(struct address_space *mapping,
1946 		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)1947 static inline void unmap_mapping_range(struct address_space *mapping,
1948 		loff_t const holebegin, loff_t const holelen, int even_cows) { }
1949 #endif
1950 
unmap_shared_mapping_range(struct address_space * mapping,loff_t const holebegin,loff_t const holelen)1951 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1952 		loff_t const holebegin, loff_t const holelen)
1953 {
1954 	unmap_mapping_range(mapping, holebegin, holelen, 0);
1955 }
1956 
1957 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1958 		void *buf, int len, unsigned int gup_flags);
1959 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1960 		void *buf, int len, unsigned int gup_flags);
1961 extern int __access_remote_vm(struct mm_struct *mm, unsigned long addr,
1962 			      void *buf, int len, unsigned int gup_flags);
1963 
1964 long get_user_pages_remote(struct mm_struct *mm,
1965 			    unsigned long start, unsigned long nr_pages,
1966 			    unsigned int gup_flags, struct page **pages,
1967 			    struct vm_area_struct **vmas, int *locked);
1968 long pin_user_pages_remote(struct mm_struct *mm,
1969 			   unsigned long start, unsigned long nr_pages,
1970 			   unsigned int gup_flags, struct page **pages,
1971 			   struct vm_area_struct **vmas, int *locked);
1972 long get_user_pages(unsigned long start, unsigned long nr_pages,
1973 			    unsigned int gup_flags, struct page **pages,
1974 			    struct vm_area_struct **vmas);
1975 long pin_user_pages(unsigned long start, unsigned long nr_pages,
1976 		    unsigned int gup_flags, struct page **pages,
1977 		    struct vm_area_struct **vmas);
1978 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1979 		    struct page **pages, unsigned int gup_flags);
1980 long pin_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1981 		    struct page **pages, unsigned int gup_flags);
1982 
1983 int get_user_pages_fast(unsigned long start, int nr_pages,
1984 			unsigned int gup_flags, struct page **pages);
1985 int pin_user_pages_fast(unsigned long start, int nr_pages,
1986 			unsigned int gup_flags, struct page **pages);
1987 
1988 int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
1989 int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
1990 			struct task_struct *task, bool bypass_rlim);
1991 
1992 struct kvec;
1993 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1994 			struct page **pages);
1995 struct page *get_dump_page(unsigned long addr);
1996 
1997 bool folio_mark_dirty(struct folio *folio);
1998 bool set_page_dirty(struct page *page);
1999 int set_page_dirty_lock(struct page *page);
2000 
2001 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
2002 
2003 extern unsigned long move_page_tables(struct vm_area_struct *vma,
2004 		unsigned long old_addr, struct vm_area_struct *new_vma,
2005 		unsigned long new_addr, unsigned long len,
2006 		bool need_rmap_locks);
2007 
2008 /*
2009  * Flags used by change_protection().  For now we make it a bitmap so
2010  * that we can pass in multiple flags just like parameters.  However
2011  * for now all the callers are only use one of the flags at the same
2012  * time.
2013  */
2014 /*
2015  * Whether we should manually check if we can map individual PTEs writable,
2016  * because something (e.g., COW, uffd-wp) blocks that from happening for all
2017  * PTEs automatically in a writable mapping.
2018  */
2019 #define  MM_CP_TRY_CHANGE_WRITABLE	   (1UL << 0)
2020 /* Whether this protection change is for NUMA hints */
2021 #define  MM_CP_PROT_NUMA                   (1UL << 1)
2022 /* Whether this change is for write protecting */
2023 #define  MM_CP_UFFD_WP                     (1UL << 2) /* do wp */
2024 #define  MM_CP_UFFD_WP_RESOLVE             (1UL << 3) /* Resolve wp */
2025 #define  MM_CP_UFFD_WP_ALL                 (MM_CP_UFFD_WP | \
2026 					    MM_CP_UFFD_WP_RESOLVE)
2027 
2028 extern unsigned long change_protection(struct mmu_gather *tlb,
2029 			      struct vm_area_struct *vma, unsigned long start,
2030 			      unsigned long end, pgprot_t newprot,
2031 			      unsigned long cp_flags);
2032 extern int mprotect_fixup(struct mmu_gather *tlb, struct vm_area_struct *vma,
2033 			  struct vm_area_struct **pprev, unsigned long start,
2034 			  unsigned long end, unsigned long newflags);
2035 
2036 /*
2037  * doesn't attempt to fault and will return short.
2038  */
2039 int get_user_pages_fast_only(unsigned long start, int nr_pages,
2040 			     unsigned int gup_flags, struct page **pages);
2041 int pin_user_pages_fast_only(unsigned long start, int nr_pages,
2042 			     unsigned int gup_flags, struct page **pages);
2043 
get_user_page_fast_only(unsigned long addr,unsigned int gup_flags,struct page ** pagep)2044 static inline bool get_user_page_fast_only(unsigned long addr,
2045 			unsigned int gup_flags, struct page **pagep)
2046 {
2047 	return get_user_pages_fast_only(addr, 1, gup_flags, pagep) == 1;
2048 }
2049 /*
2050  * per-process(per-mm_struct) statistics.
2051  */
get_mm_counter(struct mm_struct * mm,int member)2052 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
2053 {
2054 	long val = atomic_long_read(&mm->rss_stat.count[member]);
2055 
2056 #ifdef SPLIT_RSS_COUNTING
2057 	/*
2058 	 * counter is updated in asynchronous manner and may go to minus.
2059 	 * But it's never be expected number for users.
2060 	 */
2061 	if (val < 0)
2062 		val = 0;
2063 #endif
2064 	return (unsigned long)val;
2065 }
2066 
2067 void mm_trace_rss_stat(struct mm_struct *mm, int member, long count);
2068 
add_mm_counter(struct mm_struct * mm,int member,long value)2069 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
2070 {
2071 	long count = atomic_long_add_return(value, &mm->rss_stat.count[member]);
2072 
2073 	mm_trace_rss_stat(mm, member, count);
2074 }
2075 
inc_mm_counter(struct mm_struct * mm,int member)2076 static inline void inc_mm_counter(struct mm_struct *mm, int member)
2077 {
2078 	long count = atomic_long_inc_return(&mm->rss_stat.count[member]);
2079 
2080 	mm_trace_rss_stat(mm, member, count);
2081 }
2082 
dec_mm_counter(struct mm_struct * mm,int member)2083 static inline void dec_mm_counter(struct mm_struct *mm, int member)
2084 {
2085 	long count = atomic_long_dec_return(&mm->rss_stat.count[member]);
2086 
2087 	mm_trace_rss_stat(mm, member, count);
2088 }
2089 
2090 /* Optimized variant when page is already known not to be PageAnon */
mm_counter_file(struct page * page)2091 static inline int mm_counter_file(struct page *page)
2092 {
2093 	if (PageSwapBacked(page))
2094 		return MM_SHMEMPAGES;
2095 	return MM_FILEPAGES;
2096 }
2097 
mm_counter(struct page * page)2098 static inline int mm_counter(struct page *page)
2099 {
2100 	if (PageAnon(page))
2101 		return MM_ANONPAGES;
2102 	return mm_counter_file(page);
2103 }
2104 
get_mm_rss(struct mm_struct * mm)2105 static inline unsigned long get_mm_rss(struct mm_struct *mm)
2106 {
2107 	return get_mm_counter(mm, MM_FILEPAGES) +
2108 		get_mm_counter(mm, MM_ANONPAGES) +
2109 		get_mm_counter(mm, MM_SHMEMPAGES);
2110 }
2111 
get_mm_hiwater_rss(struct mm_struct * mm)2112 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
2113 {
2114 	return max(mm->hiwater_rss, get_mm_rss(mm));
2115 }
2116 
get_mm_hiwater_vm(struct mm_struct * mm)2117 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
2118 {
2119 	return max(mm->hiwater_vm, mm->total_vm);
2120 }
2121 
update_hiwater_rss(struct mm_struct * mm)2122 static inline void update_hiwater_rss(struct mm_struct *mm)
2123 {
2124 	unsigned long _rss = get_mm_rss(mm);
2125 
2126 	if ((mm)->hiwater_rss < _rss)
2127 		(mm)->hiwater_rss = _rss;
2128 }
2129 
update_hiwater_vm(struct mm_struct * mm)2130 static inline void update_hiwater_vm(struct mm_struct *mm)
2131 {
2132 	if (mm->hiwater_vm < mm->total_vm)
2133 		mm->hiwater_vm = mm->total_vm;
2134 }
2135 
reset_mm_hiwater_rss(struct mm_struct * mm)2136 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
2137 {
2138 	mm->hiwater_rss = get_mm_rss(mm);
2139 }
2140 
setmax_mm_hiwater_rss(unsigned long * maxrss,struct mm_struct * mm)2141 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
2142 					 struct mm_struct *mm)
2143 {
2144 	unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
2145 
2146 	if (*maxrss < hiwater_rss)
2147 		*maxrss = hiwater_rss;
2148 }
2149 
2150 #if defined(SPLIT_RSS_COUNTING)
2151 void sync_mm_rss(struct mm_struct *mm);
2152 #else
sync_mm_rss(struct mm_struct * mm)2153 static inline void sync_mm_rss(struct mm_struct *mm)
2154 {
2155 }
2156 #endif
2157 
2158 #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
pte_special(pte_t pte)2159 static inline int pte_special(pte_t pte)
2160 {
2161 	return 0;
2162 }
2163 
pte_mkspecial(pte_t pte)2164 static inline pte_t pte_mkspecial(pte_t pte)
2165 {
2166 	return pte;
2167 }
2168 #endif
2169 
2170 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
pte_devmap(pte_t pte)2171 static inline int pte_devmap(pte_t pte)
2172 {
2173 	return 0;
2174 }
2175 #endif
2176 
2177 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
2178 
2179 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
2180 			       spinlock_t **ptl);
get_locked_pte(struct mm_struct * mm,unsigned long addr,spinlock_t ** ptl)2181 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
2182 				    spinlock_t **ptl)
2183 {
2184 	pte_t *ptep;
2185 	__cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
2186 	return ptep;
2187 }
2188 
2189 #ifdef __PAGETABLE_P4D_FOLDED
__p4d_alloc(struct mm_struct * mm,pgd_t * pgd,unsigned long address)2190 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2191 						unsigned long address)
2192 {
2193 	return 0;
2194 }
2195 #else
2196 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
2197 #endif
2198 
2199 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
__pud_alloc(struct mm_struct * mm,p4d_t * p4d,unsigned long address)2200 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2201 						unsigned long address)
2202 {
2203 	return 0;
2204 }
mm_inc_nr_puds(struct mm_struct * mm)2205 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
mm_dec_nr_puds(struct mm_struct * mm)2206 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
2207 
2208 #else
2209 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
2210 
mm_inc_nr_puds(struct mm_struct * mm)2211 static inline void mm_inc_nr_puds(struct mm_struct *mm)
2212 {
2213 	if (mm_pud_folded(mm))
2214 		return;
2215 	atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2216 }
2217 
mm_dec_nr_puds(struct mm_struct * mm)2218 static inline void mm_dec_nr_puds(struct mm_struct *mm)
2219 {
2220 	if (mm_pud_folded(mm))
2221 		return;
2222 	atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2223 }
2224 #endif
2225 
2226 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
__pmd_alloc(struct mm_struct * mm,pud_t * pud,unsigned long address)2227 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
2228 						unsigned long address)
2229 {
2230 	return 0;
2231 }
2232 
mm_inc_nr_pmds(struct mm_struct * mm)2233 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
mm_dec_nr_pmds(struct mm_struct * mm)2234 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
2235 
2236 #else
2237 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
2238 
mm_inc_nr_pmds(struct mm_struct * mm)2239 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
2240 {
2241 	if (mm_pmd_folded(mm))
2242 		return;
2243 	atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2244 }
2245 
mm_dec_nr_pmds(struct mm_struct * mm)2246 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
2247 {
2248 	if (mm_pmd_folded(mm))
2249 		return;
2250 	atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2251 }
2252 #endif
2253 
2254 #ifdef CONFIG_MMU
mm_pgtables_bytes_init(struct mm_struct * mm)2255 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
2256 {
2257 	atomic_long_set(&mm->pgtables_bytes, 0);
2258 }
2259 
mm_pgtables_bytes(const struct mm_struct * mm)2260 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2261 {
2262 	return atomic_long_read(&mm->pgtables_bytes);
2263 }
2264 
mm_inc_nr_ptes(struct mm_struct * mm)2265 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
2266 {
2267 	atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2268 }
2269 
mm_dec_nr_ptes(struct mm_struct * mm)2270 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
2271 {
2272 	atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2273 }
2274 #else
2275 
mm_pgtables_bytes_init(struct mm_struct * mm)2276 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
mm_pgtables_bytes(const struct mm_struct * mm)2277 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2278 {
2279 	return 0;
2280 }
2281 
mm_inc_nr_ptes(struct mm_struct * mm)2282 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
mm_dec_nr_ptes(struct mm_struct * mm)2283 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
2284 #endif
2285 
2286 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
2287 int __pte_alloc_kernel(pmd_t *pmd);
2288 
2289 #if defined(CONFIG_MMU)
2290 
p4d_alloc(struct mm_struct * mm,pgd_t * pgd,unsigned long address)2291 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2292 		unsigned long address)
2293 {
2294 	return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
2295 		NULL : p4d_offset(pgd, address);
2296 }
2297 
pud_alloc(struct mm_struct * mm,p4d_t * p4d,unsigned long address)2298 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2299 		unsigned long address)
2300 {
2301 	return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
2302 		NULL : pud_offset(p4d, address);
2303 }
2304 
pmd_alloc(struct mm_struct * mm,pud_t * pud,unsigned long address)2305 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
2306 {
2307 	return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
2308 		NULL: pmd_offset(pud, address);
2309 }
2310 #endif /* CONFIG_MMU */
2311 
2312 #if USE_SPLIT_PTE_PTLOCKS
2313 #if ALLOC_SPLIT_PTLOCKS
2314 void __init ptlock_cache_init(void);
2315 extern bool ptlock_alloc(struct page *page);
2316 extern void ptlock_free(struct page *page);
2317 
ptlock_ptr(struct page * page)2318 static inline spinlock_t *ptlock_ptr(struct page *page)
2319 {
2320 	return page->ptl;
2321 }
2322 #else /* ALLOC_SPLIT_PTLOCKS */
ptlock_cache_init(void)2323 static inline void ptlock_cache_init(void)
2324 {
2325 }
2326 
ptlock_alloc(struct page * page)2327 static inline bool ptlock_alloc(struct page *page)
2328 {
2329 	return true;
2330 }
2331 
ptlock_free(struct page * page)2332 static inline void ptlock_free(struct page *page)
2333 {
2334 }
2335 
ptlock_ptr(struct page * page)2336 static inline spinlock_t *ptlock_ptr(struct page *page)
2337 {
2338 	return &page->ptl;
2339 }
2340 #endif /* ALLOC_SPLIT_PTLOCKS */
2341 
pte_lockptr(struct mm_struct * mm,pmd_t * pmd)2342 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2343 {
2344 	return ptlock_ptr(pmd_page(*pmd));
2345 }
2346 
ptlock_init(struct page * page)2347 static inline bool ptlock_init(struct page *page)
2348 {
2349 	/*
2350 	 * prep_new_page() initialize page->private (and therefore page->ptl)
2351 	 * with 0. Make sure nobody took it in use in between.
2352 	 *
2353 	 * It can happen if arch try to use slab for page table allocation:
2354 	 * slab code uses page->slab_cache, which share storage with page->ptl.
2355 	 */
2356 	VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
2357 	if (!ptlock_alloc(page))
2358 		return false;
2359 	spin_lock_init(ptlock_ptr(page));
2360 	return true;
2361 }
2362 
2363 #else	/* !USE_SPLIT_PTE_PTLOCKS */
2364 /*
2365  * We use mm->page_table_lock to guard all pagetable pages of the mm.
2366  */
pte_lockptr(struct mm_struct * mm,pmd_t * pmd)2367 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2368 {
2369 	return &mm->page_table_lock;
2370 }
ptlock_cache_init(void)2371 static inline void ptlock_cache_init(void) {}
ptlock_init(struct page * page)2372 static inline bool ptlock_init(struct page *page) { return true; }
ptlock_free(struct page * page)2373 static inline void ptlock_free(struct page *page) {}
2374 #endif /* USE_SPLIT_PTE_PTLOCKS */
2375 
pgtable_init(void)2376 static inline void pgtable_init(void)
2377 {
2378 	ptlock_cache_init();
2379 	pgtable_cache_init();
2380 }
2381 
pgtable_pte_page_ctor(struct page * page)2382 static inline bool pgtable_pte_page_ctor(struct page *page)
2383 {
2384 	if (!ptlock_init(page))
2385 		return false;
2386 	__SetPageTable(page);
2387 	inc_lruvec_page_state(page, NR_PAGETABLE);
2388 	return true;
2389 }
2390 
pgtable_pte_page_dtor(struct page * page)2391 static inline void pgtable_pte_page_dtor(struct page *page)
2392 {
2393 	ptlock_free(page);
2394 	__ClearPageTable(page);
2395 	dec_lruvec_page_state(page, NR_PAGETABLE);
2396 }
2397 
2398 #define pte_offset_map_lock(mm, pmd, address, ptlp)	\
2399 ({							\
2400 	spinlock_t *__ptl = pte_lockptr(mm, pmd);	\
2401 	pte_t *__pte = pte_offset_map(pmd, address);	\
2402 	*(ptlp) = __ptl;				\
2403 	spin_lock(__ptl);				\
2404 	__pte;						\
2405 })
2406 
2407 #define pte_unmap_unlock(pte, ptl)	do {		\
2408 	spin_unlock(ptl);				\
2409 	pte_unmap(pte);					\
2410 } while (0)
2411 
2412 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2413 
2414 #define pte_alloc_map(mm, pmd, address)			\
2415 	(pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2416 
2417 #define pte_alloc_map_lock(mm, pmd, address, ptlp)	\
2418 	(pte_alloc(mm, pmd) ?			\
2419 		 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2420 
2421 #define pte_alloc_kernel(pmd, address)			\
2422 	((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2423 		NULL: pte_offset_kernel(pmd, address))
2424 
2425 #if USE_SPLIT_PMD_PTLOCKS
2426 
pmd_to_page(pmd_t * pmd)2427 static struct page *pmd_to_page(pmd_t *pmd)
2428 {
2429 	unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2430 	return virt_to_page((void *)((unsigned long) pmd & mask));
2431 }
2432 
pmd_lockptr(struct mm_struct * mm,pmd_t * pmd)2433 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2434 {
2435 	return ptlock_ptr(pmd_to_page(pmd));
2436 }
2437 
pmd_ptlock_init(struct page * page)2438 static inline bool pmd_ptlock_init(struct page *page)
2439 {
2440 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2441 	page->pmd_huge_pte = NULL;
2442 #endif
2443 	return ptlock_init(page);
2444 }
2445 
pmd_ptlock_free(struct page * page)2446 static inline void pmd_ptlock_free(struct page *page)
2447 {
2448 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2449 	VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2450 #endif
2451 	ptlock_free(page);
2452 }
2453 
2454 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2455 
2456 #else
2457 
pmd_lockptr(struct mm_struct * mm,pmd_t * pmd)2458 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2459 {
2460 	return &mm->page_table_lock;
2461 }
2462 
pmd_ptlock_init(struct page * page)2463 static inline bool pmd_ptlock_init(struct page *page) { return true; }
pmd_ptlock_free(struct page * page)2464 static inline void pmd_ptlock_free(struct page *page) {}
2465 
2466 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2467 
2468 #endif
2469 
pmd_lock(struct mm_struct * mm,pmd_t * pmd)2470 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2471 {
2472 	spinlock_t *ptl = pmd_lockptr(mm, pmd);
2473 	spin_lock(ptl);
2474 	return ptl;
2475 }
2476 
pgtable_pmd_page_ctor(struct page * page)2477 static inline bool pgtable_pmd_page_ctor(struct page *page)
2478 {
2479 	if (!pmd_ptlock_init(page))
2480 		return false;
2481 	__SetPageTable(page);
2482 	inc_lruvec_page_state(page, NR_PAGETABLE);
2483 	return true;
2484 }
2485 
pgtable_pmd_page_dtor(struct page * page)2486 static inline void pgtable_pmd_page_dtor(struct page *page)
2487 {
2488 	pmd_ptlock_free(page);
2489 	__ClearPageTable(page);
2490 	dec_lruvec_page_state(page, NR_PAGETABLE);
2491 }
2492 
2493 /*
2494  * No scalability reason to split PUD locks yet, but follow the same pattern
2495  * as the PMD locks to make it easier if we decide to.  The VM should not be
2496  * considered ready to switch to split PUD locks yet; there may be places
2497  * which need to be converted from page_table_lock.
2498  */
pud_lockptr(struct mm_struct * mm,pud_t * pud)2499 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2500 {
2501 	return &mm->page_table_lock;
2502 }
2503 
pud_lock(struct mm_struct * mm,pud_t * pud)2504 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2505 {
2506 	spinlock_t *ptl = pud_lockptr(mm, pud);
2507 
2508 	spin_lock(ptl);
2509 	return ptl;
2510 }
2511 
2512 extern void __init pagecache_init(void);
2513 extern void free_initmem(void);
2514 
2515 /*
2516  * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2517  * into the buddy system. The freed pages will be poisoned with pattern
2518  * "poison" if it's within range [0, UCHAR_MAX].
2519  * Return pages freed into the buddy system.
2520  */
2521 extern unsigned long free_reserved_area(void *start, void *end,
2522 					int poison, const char *s);
2523 
2524 extern void adjust_managed_page_count(struct page *page, long count);
2525 extern void mem_init_print_info(void);
2526 
2527 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2528 
2529 /* Free the reserved page into the buddy system, so it gets managed. */
free_reserved_page(struct page * page)2530 static inline void free_reserved_page(struct page *page)
2531 {
2532 	ClearPageReserved(page);
2533 	init_page_count(page);
2534 	__free_page(page);
2535 	adjust_managed_page_count(page, 1);
2536 }
2537 #define free_highmem_page(page) free_reserved_page(page)
2538 
mark_page_reserved(struct page * page)2539 static inline void mark_page_reserved(struct page *page)
2540 {
2541 	SetPageReserved(page);
2542 	adjust_managed_page_count(page, -1);
2543 }
2544 
2545 /*
2546  * Default method to free all the __init memory into the buddy system.
2547  * The freed pages will be poisoned with pattern "poison" if it's within
2548  * range [0, UCHAR_MAX].
2549  * Return pages freed into the buddy system.
2550  */
free_initmem_default(int poison)2551 static inline unsigned long free_initmem_default(int poison)
2552 {
2553 	extern char __init_begin[], __init_end[];
2554 
2555 	return free_reserved_area(&__init_begin, &__init_end,
2556 				  poison, "unused kernel image (initmem)");
2557 }
2558 
get_num_physpages(void)2559 static inline unsigned long get_num_physpages(void)
2560 {
2561 	int nid;
2562 	unsigned long phys_pages = 0;
2563 
2564 	for_each_online_node(nid)
2565 		phys_pages += node_present_pages(nid);
2566 
2567 	return phys_pages;
2568 }
2569 
2570 /*
2571  * Using memblock node mappings, an architecture may initialise its
2572  * zones, allocate the backing mem_map and account for memory holes in an
2573  * architecture independent manner.
2574  *
2575  * An architecture is expected to register range of page frames backed by
2576  * physical memory with memblock_add[_node]() before calling
2577  * free_area_init() passing in the PFN each zone ends at. At a basic
2578  * usage, an architecture is expected to do something like
2579  *
2580  * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2581  * 							 max_highmem_pfn};
2582  * for_each_valid_physical_page_range()
2583  *	memblock_add_node(base, size, nid, MEMBLOCK_NONE)
2584  * free_area_init(max_zone_pfns);
2585  */
2586 void free_area_init(unsigned long *max_zone_pfn);
2587 unsigned long node_map_pfn_alignment(void);
2588 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2589 						unsigned long end_pfn);
2590 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2591 						unsigned long end_pfn);
2592 extern void get_pfn_range_for_nid(unsigned int nid,
2593 			unsigned long *start_pfn, unsigned long *end_pfn);
2594 
2595 #ifndef CONFIG_NUMA
early_pfn_to_nid(unsigned long pfn)2596 static inline int early_pfn_to_nid(unsigned long pfn)
2597 {
2598 	return 0;
2599 }
2600 #else
2601 /* please see mm/page_alloc.c */
2602 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2603 #endif
2604 
2605 extern void set_dma_reserve(unsigned long new_dma_reserve);
2606 extern void memmap_init_range(unsigned long, int, unsigned long,
2607 		unsigned long, unsigned long, enum meminit_context,
2608 		struct vmem_altmap *, int migratetype);
2609 extern void setup_per_zone_wmarks(void);
2610 extern void calculate_min_free_kbytes(void);
2611 extern int __meminit init_per_zone_wmark_min(void);
2612 extern void mem_init(void);
2613 extern void __init mmap_init(void);
2614 
2615 extern void __show_mem(unsigned int flags, nodemask_t *nodemask, int max_zone_idx);
show_mem(unsigned int flags,nodemask_t * nodemask)2616 static inline void show_mem(unsigned int flags, nodemask_t *nodemask)
2617 {
2618 	__show_mem(flags, nodemask, MAX_NR_ZONES - 1);
2619 }
2620 extern long si_mem_available(void);
2621 extern void si_meminfo(struct sysinfo * val);
2622 extern void si_meminfo_node(struct sysinfo *val, int nid);
2623 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2624 extern unsigned long arch_reserved_kernel_pages(void);
2625 #endif
2626 
2627 extern __printf(3, 4)
2628 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2629 
2630 extern void setup_per_cpu_pageset(void);
2631 
2632 /* page_alloc.c */
2633 extern int min_free_kbytes;
2634 extern int watermark_boost_factor;
2635 extern int watermark_scale_factor;
2636 extern bool arch_has_descending_max_zone_pfns(void);
2637 
2638 /* nommu.c */
2639 extern atomic_long_t mmap_pages_allocated;
2640 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2641 
2642 /* interval_tree.c */
2643 void vma_interval_tree_insert(struct vm_area_struct *node,
2644 			      struct rb_root_cached *root);
2645 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2646 				    struct vm_area_struct *prev,
2647 				    struct rb_root_cached *root);
2648 void vma_interval_tree_remove(struct vm_area_struct *node,
2649 			      struct rb_root_cached *root);
2650 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2651 				unsigned long start, unsigned long last);
2652 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2653 				unsigned long start, unsigned long last);
2654 
2655 #define vma_interval_tree_foreach(vma, root, start, last)		\
2656 	for (vma = vma_interval_tree_iter_first(root, start, last);	\
2657 	     vma; vma = vma_interval_tree_iter_next(vma, start, last))
2658 
2659 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2660 				   struct rb_root_cached *root);
2661 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2662 				   struct rb_root_cached *root);
2663 struct anon_vma_chain *
2664 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2665 				  unsigned long start, unsigned long last);
2666 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2667 	struct anon_vma_chain *node, unsigned long start, unsigned long last);
2668 #ifdef CONFIG_DEBUG_VM_RB
2669 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2670 #endif
2671 
2672 #define anon_vma_interval_tree_foreach(avc, root, start, last)		 \
2673 	for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2674 	     avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2675 
2676 /* mmap.c */
2677 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2678 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2679 	unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2680 	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)2681 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2682 	unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2683 {
2684 	return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2685 }
2686 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2687 	struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2688 	unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2689 	struct mempolicy *, struct vm_userfaultfd_ctx, struct anon_vma_name *);
2690 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2691 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2692 	unsigned long addr, int new_below);
2693 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2694 	unsigned long addr, int new_below);
2695 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2696 extern void unlink_file_vma(struct vm_area_struct *);
2697 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2698 	unsigned long addr, unsigned long len, pgoff_t pgoff,
2699 	bool *need_rmap_locks);
2700 extern void exit_mmap(struct mm_struct *);
2701 
2702 void vma_mas_store(struct vm_area_struct *vma, struct ma_state *mas);
2703 void vma_mas_remove(struct vm_area_struct *vma, struct ma_state *mas);
2704 
check_data_rlimit(unsigned long rlim,unsigned long new,unsigned long start,unsigned long end_data,unsigned long start_data)2705 static inline int check_data_rlimit(unsigned long rlim,
2706 				    unsigned long new,
2707 				    unsigned long start,
2708 				    unsigned long end_data,
2709 				    unsigned long start_data)
2710 {
2711 	if (rlim < RLIM_INFINITY) {
2712 		if (((new - start) + (end_data - start_data)) > rlim)
2713 			return -ENOSPC;
2714 	}
2715 
2716 	return 0;
2717 }
2718 
2719 extern int mm_take_all_locks(struct mm_struct *mm);
2720 extern void mm_drop_all_locks(struct mm_struct *mm);
2721 
2722 extern int set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2723 extern int replace_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2724 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2725 extern struct file *get_task_exe_file(struct task_struct *task);
2726 
2727 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2728 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2729 
2730 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2731 				   const struct vm_special_mapping *sm);
2732 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2733 				   unsigned long addr, unsigned long len,
2734 				   unsigned long flags,
2735 				   const struct vm_special_mapping *spec);
2736 /* This is an obsolete alternative to _install_special_mapping. */
2737 extern int install_special_mapping(struct mm_struct *mm,
2738 				   unsigned long addr, unsigned long len,
2739 				   unsigned long flags, struct page **pages);
2740 
2741 unsigned long randomize_stack_top(unsigned long stack_top);
2742 unsigned long randomize_page(unsigned long start, unsigned long range);
2743 
2744 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2745 
2746 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2747 	unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2748 	struct list_head *uf);
2749 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2750 	unsigned long len, unsigned long prot, unsigned long flags,
2751 	unsigned long pgoff, unsigned long *populate, struct list_head *uf);
2752 extern int do_mas_munmap(struct ma_state *mas, struct mm_struct *mm,
2753 			 unsigned long start, size_t len, struct list_head *uf,
2754 			 bool downgrade);
2755 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2756 		     struct list_head *uf);
2757 extern int do_madvise(struct mm_struct *mm, unsigned long start, size_t len_in, int behavior);
2758 
2759 #ifdef CONFIG_MMU
2760 extern int __mm_populate(unsigned long addr, unsigned long len,
2761 			 int ignore_errors);
mm_populate(unsigned long addr,unsigned long len)2762 static inline void mm_populate(unsigned long addr, unsigned long len)
2763 {
2764 	/* Ignore errors */
2765 	(void) __mm_populate(addr, len, 1);
2766 }
2767 #else
mm_populate(unsigned long addr,unsigned long len)2768 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2769 #endif
2770 
2771 /* These take the mm semaphore themselves */
2772 extern int __must_check vm_brk(unsigned long, unsigned long);
2773 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2774 extern int vm_munmap(unsigned long, size_t);
2775 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2776         unsigned long, unsigned long,
2777         unsigned long, unsigned long);
2778 
2779 struct vm_unmapped_area_info {
2780 #define VM_UNMAPPED_AREA_TOPDOWN 1
2781 	unsigned long flags;
2782 	unsigned long length;
2783 	unsigned long low_limit;
2784 	unsigned long high_limit;
2785 	unsigned long align_mask;
2786 	unsigned long align_offset;
2787 };
2788 
2789 extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info);
2790 
2791 /* truncate.c */
2792 extern void truncate_inode_pages(struct address_space *, loff_t);
2793 extern void truncate_inode_pages_range(struct address_space *,
2794 				       loff_t lstart, loff_t lend);
2795 extern void truncate_inode_pages_final(struct address_space *);
2796 
2797 /* generic vm_area_ops exported for stackable file systems */
2798 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2799 extern vm_fault_t filemap_map_pages(struct vm_fault *vmf,
2800 		pgoff_t start_pgoff, pgoff_t end_pgoff);
2801 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2802 
2803 extern unsigned long stack_guard_gap;
2804 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2805 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2806 
2807 /* CONFIG_STACK_GROWSUP still needs to grow downwards at some places */
2808 extern int expand_downwards(struct vm_area_struct *vma,
2809 		unsigned long address);
2810 #if VM_GROWSUP
2811 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2812 #else
2813   #define expand_upwards(vma, address) (0)
2814 #endif
2815 
2816 /* Look up the first VMA which satisfies  addr < vm_end,  NULL if none. */
2817 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2818 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2819 					     struct vm_area_struct **pprev);
2820 
2821 /*
2822  * Look up the first VMA which intersects the interval [start_addr, end_addr)
2823  * NULL if none.  Assume start_addr < end_addr.
2824  */
2825 struct vm_area_struct *find_vma_intersection(struct mm_struct *mm,
2826 			unsigned long start_addr, unsigned long end_addr);
2827 
2828 /**
2829  * vma_lookup() - Find a VMA at a specific address
2830  * @mm: The process address space.
2831  * @addr: The user address.
2832  *
2833  * Return: The vm_area_struct at the given address, %NULL otherwise.
2834  */
2835 static inline
vma_lookup(struct mm_struct * mm,unsigned long addr)2836 struct vm_area_struct *vma_lookup(struct mm_struct *mm, unsigned long addr)
2837 {
2838 	return mtree_load(&mm->mm_mt, addr);
2839 }
2840 
vm_start_gap(struct vm_area_struct * vma)2841 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2842 {
2843 	unsigned long vm_start = vma->vm_start;
2844 
2845 	if (vma->vm_flags & VM_GROWSDOWN) {
2846 		vm_start -= stack_guard_gap;
2847 		if (vm_start > vma->vm_start)
2848 			vm_start = 0;
2849 	}
2850 	return vm_start;
2851 }
2852 
vm_end_gap(struct vm_area_struct * vma)2853 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2854 {
2855 	unsigned long vm_end = vma->vm_end;
2856 
2857 	if (vma->vm_flags & VM_GROWSUP) {
2858 		vm_end += stack_guard_gap;
2859 		if (vm_end < vma->vm_end)
2860 			vm_end = -PAGE_SIZE;
2861 	}
2862 	return vm_end;
2863 }
2864 
vma_pages(struct vm_area_struct * vma)2865 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2866 {
2867 	return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2868 }
2869 
2870 /* 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)2871 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2872 				unsigned long vm_start, unsigned long vm_end)
2873 {
2874 	struct vm_area_struct *vma = vma_lookup(mm, vm_start);
2875 
2876 	if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2877 		vma = NULL;
2878 
2879 	return vma;
2880 }
2881 
range_in_vma(struct vm_area_struct * vma,unsigned long start,unsigned long end)2882 static inline bool range_in_vma(struct vm_area_struct *vma,
2883 				unsigned long start, unsigned long end)
2884 {
2885 	return (vma && vma->vm_start <= start && end <= vma->vm_end);
2886 }
2887 
2888 #ifdef CONFIG_MMU
2889 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2890 void vma_set_page_prot(struct vm_area_struct *vma);
2891 #else
vm_get_page_prot(unsigned long vm_flags)2892 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2893 {
2894 	return __pgprot(0);
2895 }
vma_set_page_prot(struct vm_area_struct * vma)2896 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2897 {
2898 	vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2899 }
2900 #endif
2901 
2902 void vma_set_file(struct vm_area_struct *vma, struct file *file);
2903 
2904 #ifdef CONFIG_NUMA_BALANCING
2905 unsigned long change_prot_numa(struct vm_area_struct *vma,
2906 			unsigned long start, unsigned long end);
2907 #endif
2908 
2909 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2910 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2911 			unsigned long pfn, unsigned long size, pgprot_t);
2912 int remap_pfn_range_notrack(struct vm_area_struct *vma, unsigned long addr,
2913 		unsigned long pfn, unsigned long size, pgprot_t prot);
2914 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2915 int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
2916 			struct page **pages, unsigned long *num);
2917 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
2918 				unsigned long num);
2919 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
2920 				unsigned long num);
2921 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2922 			unsigned long pfn);
2923 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2924 			unsigned long pfn, pgprot_t pgprot);
2925 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2926 			pfn_t pfn);
2927 vm_fault_t vmf_insert_mixed_prot(struct vm_area_struct *vma, unsigned long addr,
2928 			pfn_t pfn, pgprot_t pgprot);
2929 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
2930 		unsigned long addr, pfn_t pfn);
2931 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2932 
vmf_insert_page(struct vm_area_struct * vma,unsigned long addr,struct page * page)2933 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2934 				unsigned long addr, struct page *page)
2935 {
2936 	int err = vm_insert_page(vma, addr, page);
2937 
2938 	if (err == -ENOMEM)
2939 		return VM_FAULT_OOM;
2940 	if (err < 0 && err != -EBUSY)
2941 		return VM_FAULT_SIGBUS;
2942 
2943 	return VM_FAULT_NOPAGE;
2944 }
2945 
2946 #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)2947 static inline int io_remap_pfn_range(struct vm_area_struct *vma,
2948 				     unsigned long addr, unsigned long pfn,
2949 				     unsigned long size, pgprot_t prot)
2950 {
2951 	return remap_pfn_range(vma, addr, pfn, size, pgprot_decrypted(prot));
2952 }
2953 #endif
2954 
vmf_error(int err)2955 static inline vm_fault_t vmf_error(int err)
2956 {
2957 	if (err == -ENOMEM)
2958 		return VM_FAULT_OOM;
2959 	return VM_FAULT_SIGBUS;
2960 }
2961 
2962 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
2963 			 unsigned int foll_flags);
2964 
2965 #define FOLL_WRITE	0x01	/* check pte is writable */
2966 #define FOLL_TOUCH	0x02	/* mark page accessed */
2967 #define FOLL_GET	0x04	/* do get_page on page */
2968 #define FOLL_DUMP	0x08	/* give error on hole if it would be zero */
2969 #define FOLL_FORCE	0x10	/* get_user_pages read/write w/o permission */
2970 #define FOLL_NOWAIT	0x20	/* if a disk transfer is needed, start the IO
2971 				 * and return without waiting upon it */
2972 #define FOLL_NOFAULT	0x80	/* do not fault in pages */
2973 #define FOLL_HWPOISON	0x100	/* check page is hwpoisoned */
2974 #define FOLL_MIGRATION	0x400	/* wait for page to replace migration entry */
2975 #define FOLL_TRIED	0x800	/* a retry, previous pass started an IO */
2976 #define FOLL_REMOTE	0x2000	/* we are working on non-current tsk/mm */
2977 #define FOLL_ANON	0x8000	/* don't do file mappings */
2978 #define FOLL_LONGTERM	0x10000	/* mapping lifetime is indefinite: see below */
2979 #define FOLL_SPLIT_PMD	0x20000	/* split huge pmd before returning */
2980 #define FOLL_PIN	0x40000	/* pages must be released via unpin_user_page */
2981 #define FOLL_FAST_ONLY	0x80000	/* gup_fast: prevent fall-back to slow gup */
2982 
2983 /*
2984  * FOLL_PIN and FOLL_LONGTERM may be used in various combinations with each
2985  * other. Here is what they mean, and how to use them:
2986  *
2987  * FOLL_LONGTERM indicates that the page will be held for an indefinite time
2988  * period _often_ under userspace control.  This is in contrast to
2989  * iov_iter_get_pages(), whose usages are transient.
2990  *
2991  * FIXME: For pages which are part of a filesystem, mappings are subject to the
2992  * lifetime enforced by the filesystem and we need guarantees that longterm
2993  * users like RDMA and V4L2 only establish mappings which coordinate usage with
2994  * the filesystem.  Ideas for this coordination include revoking the longterm
2995  * pin, delaying writeback, bounce buffer page writeback, etc.  As FS DAX was
2996  * added after the problem with filesystems was found FS DAX VMAs are
2997  * specifically failed.  Filesystem pages are still subject to bugs and use of
2998  * FOLL_LONGTERM should be avoided on those pages.
2999  *
3000  * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call.
3001  * Currently only get_user_pages() and get_user_pages_fast() support this flag
3002  * and calls to get_user_pages_[un]locked are specifically not allowed.  This
3003  * is due to an incompatibility with the FS DAX check and
3004  * FAULT_FLAG_ALLOW_RETRY.
3005  *
3006  * In the CMA case: long term pins in a CMA region would unnecessarily fragment
3007  * that region.  And so, CMA attempts to migrate the page before pinning, when
3008  * FOLL_LONGTERM is specified.
3009  *
3010  * FOLL_PIN indicates that a special kind of tracking (not just page->_refcount,
3011  * but an additional pin counting system) will be invoked. This is intended for
3012  * anything that gets a page reference and then touches page data (for example,
3013  * Direct IO). This lets the filesystem know that some non-file-system entity is
3014  * potentially changing the pages' data. In contrast to FOLL_GET (whose pages
3015  * are released via put_page()), FOLL_PIN pages must be released, ultimately, by
3016  * a call to unpin_user_page().
3017  *
3018  * FOLL_PIN is similar to FOLL_GET: both of these pin pages. They use different
3019  * and separate refcounting mechanisms, however, and that means that each has
3020  * its own acquire and release mechanisms:
3021  *
3022  *     FOLL_GET: get_user_pages*() to acquire, and put_page() to release.
3023  *
3024  *     FOLL_PIN: pin_user_pages*() to acquire, and unpin_user_pages to release.
3025  *
3026  * FOLL_PIN and FOLL_GET are mutually exclusive for a given function call.
3027  * (The underlying pages may experience both FOLL_GET-based and FOLL_PIN-based
3028  * calls applied to them, and that's perfectly OK. This is a constraint on the
3029  * callers, not on the pages.)
3030  *
3031  * FOLL_PIN should be set internally by the pin_user_pages*() APIs, never
3032  * directly by the caller. That's in order to help avoid mismatches when
3033  * releasing pages: get_user_pages*() pages must be released via put_page(),
3034  * while pin_user_pages*() pages must be released via unpin_user_page().
3035  *
3036  * Please see Documentation/core-api/pin_user_pages.rst for more information.
3037  */
3038 
vm_fault_to_errno(vm_fault_t vm_fault,int foll_flags)3039 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
3040 {
3041 	if (vm_fault & VM_FAULT_OOM)
3042 		return -ENOMEM;
3043 	if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
3044 		return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
3045 	if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
3046 		return -EFAULT;
3047 	return 0;
3048 }
3049 
3050 /*
3051  * Indicates for which pages that are write-protected in the page table,
3052  * whether GUP has to trigger unsharing via FAULT_FLAG_UNSHARE such that the
3053  * GUP pin will remain consistent with the pages mapped into the page tables
3054  * of the MM.
3055  *
3056  * Temporary unmapping of PageAnonExclusive() pages or clearing of
3057  * PageAnonExclusive() has to protect against concurrent GUP:
3058  * * Ordinary GUP: Using the PT lock
3059  * * GUP-fast and fork(): mm->write_protect_seq
3060  * * GUP-fast and KSM or temporary unmapping (swap, migration): see
3061  *    page_try_share_anon_rmap()
3062  *
3063  * Must be called with the (sub)page that's actually referenced via the
3064  * page table entry, which might not necessarily be the head page for a
3065  * PTE-mapped THP.
3066  */
gup_must_unshare(unsigned int flags,struct page * page)3067 static inline bool gup_must_unshare(unsigned int flags, struct page *page)
3068 {
3069 	/*
3070 	 * FOLL_WRITE is implicitly handled correctly as the page table entry
3071 	 * has to be writable -- and if it references (part of) an anonymous
3072 	 * folio, that part is required to be marked exclusive.
3073 	 */
3074 	if ((flags & (FOLL_WRITE | FOLL_PIN)) != FOLL_PIN)
3075 		return false;
3076 	/*
3077 	 * Note: PageAnon(page) is stable until the page is actually getting
3078 	 * freed.
3079 	 */
3080 	if (!PageAnon(page))
3081 		return false;
3082 
3083 	/* Paired with a memory barrier in page_try_share_anon_rmap(). */
3084 	if (IS_ENABLED(CONFIG_HAVE_FAST_GUP))
3085 		smp_rmb();
3086 
3087 	/*
3088 	 * Note that PageKsm() pages cannot be exclusive, and consequently,
3089 	 * cannot get pinned.
3090 	 */
3091 	return !PageAnonExclusive(page);
3092 }
3093 
3094 /*
3095  * Indicates whether GUP can follow a PROT_NONE mapped page, or whether
3096  * a (NUMA hinting) fault is required.
3097  */
gup_can_follow_protnone(unsigned int flags)3098 static inline bool gup_can_follow_protnone(unsigned int flags)
3099 {
3100 	/*
3101 	 * FOLL_FORCE has to be able to make progress even if the VMA is
3102 	 * inaccessible. Further, FOLL_FORCE access usually does not represent
3103 	 * application behaviour and we should avoid triggering NUMA hinting
3104 	 * faults.
3105 	 */
3106 	return flags & FOLL_FORCE;
3107 }
3108 
3109 typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
3110 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
3111 			       unsigned long size, pte_fn_t fn, void *data);
3112 extern int apply_to_existing_page_range(struct mm_struct *mm,
3113 				   unsigned long address, unsigned long size,
3114 				   pte_fn_t fn, void *data);
3115 
3116 extern void __init init_mem_debugging_and_hardening(void);
3117 #ifdef CONFIG_PAGE_POISONING
3118 extern void __kernel_poison_pages(struct page *page, int numpages);
3119 extern void __kernel_unpoison_pages(struct page *page, int numpages);
3120 extern bool _page_poisoning_enabled_early;
3121 DECLARE_STATIC_KEY_FALSE(_page_poisoning_enabled);
page_poisoning_enabled(void)3122 static inline bool page_poisoning_enabled(void)
3123 {
3124 	return _page_poisoning_enabled_early;
3125 }
3126 /*
3127  * For use in fast paths after init_mem_debugging() has run, or when a
3128  * false negative result is not harmful when called too early.
3129  */
page_poisoning_enabled_static(void)3130 static inline bool page_poisoning_enabled_static(void)
3131 {
3132 	return static_branch_unlikely(&_page_poisoning_enabled);
3133 }
kernel_poison_pages(struct page * page,int numpages)3134 static inline void kernel_poison_pages(struct page *page, int numpages)
3135 {
3136 	if (page_poisoning_enabled_static())
3137 		__kernel_poison_pages(page, numpages);
3138 }
kernel_unpoison_pages(struct page * page,int numpages)3139 static inline void kernel_unpoison_pages(struct page *page, int numpages)
3140 {
3141 	if (page_poisoning_enabled_static())
3142 		__kernel_unpoison_pages(page, numpages);
3143 }
3144 #else
page_poisoning_enabled(void)3145 static inline bool page_poisoning_enabled(void) { return false; }
page_poisoning_enabled_static(void)3146 static inline bool page_poisoning_enabled_static(void) { return false; }
__kernel_poison_pages(struct page * page,int nunmpages)3147 static inline void __kernel_poison_pages(struct page *page, int nunmpages) { }
kernel_poison_pages(struct page * page,int numpages)3148 static inline void kernel_poison_pages(struct page *page, int numpages) { }
kernel_unpoison_pages(struct page * page,int numpages)3149 static inline void kernel_unpoison_pages(struct page *page, int numpages) { }
3150 #endif
3151 
3152 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_ALLOC_DEFAULT_ON, init_on_alloc);
want_init_on_alloc(gfp_t flags)3153 static inline bool want_init_on_alloc(gfp_t flags)
3154 {
3155 	if (static_branch_maybe(CONFIG_INIT_ON_ALLOC_DEFAULT_ON,
3156 				&init_on_alloc))
3157 		return true;
3158 	return flags & __GFP_ZERO;
3159 }
3160 
3161 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_FREE_DEFAULT_ON, init_on_free);
want_init_on_free(void)3162 static inline bool want_init_on_free(void)
3163 {
3164 	return static_branch_maybe(CONFIG_INIT_ON_FREE_DEFAULT_ON,
3165 				   &init_on_free);
3166 }
3167 
3168 extern bool _debug_pagealloc_enabled_early;
3169 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
3170 
debug_pagealloc_enabled(void)3171 static inline bool debug_pagealloc_enabled(void)
3172 {
3173 	return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
3174 		_debug_pagealloc_enabled_early;
3175 }
3176 
3177 /*
3178  * For use in fast paths after init_debug_pagealloc() has run, or when a
3179  * false negative result is not harmful when called too early.
3180  */
debug_pagealloc_enabled_static(void)3181 static inline bool debug_pagealloc_enabled_static(void)
3182 {
3183 	if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
3184 		return false;
3185 
3186 	return static_branch_unlikely(&_debug_pagealloc_enabled);
3187 }
3188 
3189 #ifdef CONFIG_DEBUG_PAGEALLOC
3190 /*
3191  * To support DEBUG_PAGEALLOC architecture must ensure that
3192  * __kernel_map_pages() never fails
3193  */
3194 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
3195 
debug_pagealloc_map_pages(struct page * page,int numpages)3196 static inline void debug_pagealloc_map_pages(struct page *page, int numpages)
3197 {
3198 	if (debug_pagealloc_enabled_static())
3199 		__kernel_map_pages(page, numpages, 1);
3200 }
3201 
debug_pagealloc_unmap_pages(struct page * page,int numpages)3202 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages)
3203 {
3204 	if (debug_pagealloc_enabled_static())
3205 		__kernel_map_pages(page, numpages, 0);
3206 }
3207 #else	/* CONFIG_DEBUG_PAGEALLOC */
debug_pagealloc_map_pages(struct page * page,int numpages)3208 static inline void debug_pagealloc_map_pages(struct page *page, int numpages) {}
debug_pagealloc_unmap_pages(struct page * page,int numpages)3209 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages) {}
3210 #endif	/* CONFIG_DEBUG_PAGEALLOC */
3211 
3212 #ifdef __HAVE_ARCH_GATE_AREA
3213 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
3214 extern int in_gate_area_no_mm(unsigned long addr);
3215 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
3216 #else
get_gate_vma(struct mm_struct * mm)3217 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
3218 {
3219 	return NULL;
3220 }
in_gate_area_no_mm(unsigned long addr)3221 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
in_gate_area(struct mm_struct * mm,unsigned long addr)3222 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
3223 {
3224 	return 0;
3225 }
3226 #endif	/* __HAVE_ARCH_GATE_AREA */
3227 
3228 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
3229 
3230 #ifdef CONFIG_SYSCTL
3231 extern int sysctl_drop_caches;
3232 int drop_caches_sysctl_handler(struct ctl_table *, int, void *, size_t *,
3233 		loff_t *);
3234 #endif
3235 
3236 void drop_slab(void);
3237 
3238 #ifndef CONFIG_MMU
3239 #define randomize_va_space 0
3240 #else
3241 extern int randomize_va_space;
3242 #endif
3243 
3244 const char * arch_vma_name(struct vm_area_struct *vma);
3245 #ifdef CONFIG_MMU
3246 void print_vma_addr(char *prefix, unsigned long rip);
3247 #else
print_vma_addr(char * prefix,unsigned long rip)3248 static inline void print_vma_addr(char *prefix, unsigned long rip)
3249 {
3250 }
3251 #endif
3252 
3253 void *sparse_buffer_alloc(unsigned long size);
3254 struct page * __populate_section_memmap(unsigned long pfn,
3255 		unsigned long nr_pages, int nid, struct vmem_altmap *altmap,
3256 		struct dev_pagemap *pgmap);
3257 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
3258 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
3259 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
3260 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
3261 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node,
3262 			    struct vmem_altmap *altmap, struct page *reuse);
3263 void *vmemmap_alloc_block(unsigned long size, int node);
3264 struct vmem_altmap;
3265 void *vmemmap_alloc_block_buf(unsigned long size, int node,
3266 			      struct vmem_altmap *altmap);
3267 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
3268 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
3269 			       int node, struct vmem_altmap *altmap);
3270 int vmemmap_populate(unsigned long start, unsigned long end, int node,
3271 		struct vmem_altmap *altmap);
3272 void vmemmap_populate_print_last(void);
3273 #ifdef CONFIG_MEMORY_HOTPLUG
3274 void vmemmap_free(unsigned long start, unsigned long end,
3275 		struct vmem_altmap *altmap);
3276 #endif
3277 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
3278 				  unsigned long nr_pages);
3279 
3280 enum mf_flags {
3281 	MF_COUNT_INCREASED = 1 << 0,
3282 	MF_ACTION_REQUIRED = 1 << 1,
3283 	MF_MUST_KILL = 1 << 2,
3284 	MF_SOFT_OFFLINE = 1 << 3,
3285 	MF_UNPOISON = 1 << 4,
3286 	MF_SW_SIMULATED = 1 << 5,
3287 	MF_NO_RETRY = 1 << 6,
3288 };
3289 int mf_dax_kill_procs(struct address_space *mapping, pgoff_t index,
3290 		      unsigned long count, int mf_flags);
3291 extern int memory_failure(unsigned long pfn, int flags);
3292 extern void memory_failure_queue(unsigned long pfn, int flags);
3293 extern void memory_failure_queue_kick(int cpu);
3294 extern int unpoison_memory(unsigned long pfn);
3295 extern int sysctl_memory_failure_early_kill;
3296 extern int sysctl_memory_failure_recovery;
3297 extern void shake_page(struct page *p);
3298 extern atomic_long_t num_poisoned_pages __read_mostly;
3299 extern int soft_offline_page(unsigned long pfn, int flags);
3300 #ifdef CONFIG_MEMORY_FAILURE
3301 extern int __get_huge_page_for_hwpoison(unsigned long pfn, int flags);
3302 #else
__get_huge_page_for_hwpoison(unsigned long pfn,int flags)3303 static inline int __get_huge_page_for_hwpoison(unsigned long pfn, int flags)
3304 {
3305 	return 0;
3306 }
3307 #endif
3308 
3309 #ifndef arch_memory_failure
arch_memory_failure(unsigned long pfn,int flags)3310 static inline int arch_memory_failure(unsigned long pfn, int flags)
3311 {
3312 	return -ENXIO;
3313 }
3314 #endif
3315 
3316 #ifndef arch_is_platform_page
arch_is_platform_page(u64 paddr)3317 static inline bool arch_is_platform_page(u64 paddr)
3318 {
3319 	return false;
3320 }
3321 #endif
3322 
3323 /*
3324  * Error handlers for various types of pages.
3325  */
3326 enum mf_result {
3327 	MF_IGNORED,	/* Error: cannot be handled */
3328 	MF_FAILED,	/* Error: handling failed */
3329 	MF_DELAYED,	/* Will be handled later */
3330 	MF_RECOVERED,	/* Successfully recovered */
3331 };
3332 
3333 enum mf_action_page_type {
3334 	MF_MSG_KERNEL,
3335 	MF_MSG_KERNEL_HIGH_ORDER,
3336 	MF_MSG_SLAB,
3337 	MF_MSG_DIFFERENT_COMPOUND,
3338 	MF_MSG_HUGE,
3339 	MF_MSG_FREE_HUGE,
3340 	MF_MSG_UNMAP_FAILED,
3341 	MF_MSG_DIRTY_SWAPCACHE,
3342 	MF_MSG_CLEAN_SWAPCACHE,
3343 	MF_MSG_DIRTY_MLOCKED_LRU,
3344 	MF_MSG_CLEAN_MLOCKED_LRU,
3345 	MF_MSG_DIRTY_UNEVICTABLE_LRU,
3346 	MF_MSG_CLEAN_UNEVICTABLE_LRU,
3347 	MF_MSG_DIRTY_LRU,
3348 	MF_MSG_CLEAN_LRU,
3349 	MF_MSG_TRUNCATED_LRU,
3350 	MF_MSG_BUDDY,
3351 	MF_MSG_DAX,
3352 	MF_MSG_UNSPLIT_THP,
3353 	MF_MSG_UNKNOWN,
3354 };
3355 
3356 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
3357 extern void clear_huge_page(struct page *page,
3358 			    unsigned long addr_hint,
3359 			    unsigned int pages_per_huge_page);
3360 extern void copy_user_huge_page(struct page *dst, struct page *src,
3361 				unsigned long addr_hint,
3362 				struct vm_area_struct *vma,
3363 				unsigned int pages_per_huge_page);
3364 extern long copy_huge_page_from_user(struct page *dst_page,
3365 				const void __user *usr_src,
3366 				unsigned int pages_per_huge_page,
3367 				bool allow_pagefault);
3368 
3369 /**
3370  * vma_is_special_huge - Are transhuge page-table entries considered special?
3371  * @vma: Pointer to the struct vm_area_struct to consider
3372  *
3373  * Whether transhuge page-table entries are considered "special" following
3374  * the definition in vm_normal_page().
3375  *
3376  * Return: true if transhuge page-table entries should be considered special,
3377  * false otherwise.
3378  */
vma_is_special_huge(const struct vm_area_struct * vma)3379 static inline bool vma_is_special_huge(const struct vm_area_struct *vma)
3380 {
3381 	return vma_is_dax(vma) || (vma->vm_file &&
3382 				   (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP)));
3383 }
3384 
3385 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
3386 
3387 #ifdef CONFIG_DEBUG_PAGEALLOC
3388 extern unsigned int _debug_guardpage_minorder;
3389 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
3390 
debug_guardpage_minorder(void)3391 static inline unsigned int debug_guardpage_minorder(void)
3392 {
3393 	return _debug_guardpage_minorder;
3394 }
3395 
debug_guardpage_enabled(void)3396 static inline bool debug_guardpage_enabled(void)
3397 {
3398 	return static_branch_unlikely(&_debug_guardpage_enabled);
3399 }
3400 
page_is_guard(struct page * page)3401 static inline bool page_is_guard(struct page *page)
3402 {
3403 	if (!debug_guardpage_enabled())
3404 		return false;
3405 
3406 	return PageGuard(page);
3407 }
3408 #else
debug_guardpage_minorder(void)3409 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
debug_guardpage_enabled(void)3410 static inline bool debug_guardpage_enabled(void) { return false; }
page_is_guard(struct page * page)3411 static inline bool page_is_guard(struct page *page) { return false; }
3412 #endif /* CONFIG_DEBUG_PAGEALLOC */
3413 
3414 #if MAX_NUMNODES > 1
3415 void __init setup_nr_node_ids(void);
3416 #else
setup_nr_node_ids(void)3417 static inline void setup_nr_node_ids(void) {}
3418 #endif
3419 
3420 extern int memcmp_pages(struct page *page1, struct page *page2);
3421 
pages_identical(struct page * page1,struct page * page2)3422 static inline int pages_identical(struct page *page1, struct page *page2)
3423 {
3424 	return !memcmp_pages(page1, page2);
3425 }
3426 
3427 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
3428 unsigned long clean_record_shared_mapping_range(struct address_space *mapping,
3429 						pgoff_t first_index, pgoff_t nr,
3430 						pgoff_t bitmap_pgoff,
3431 						unsigned long *bitmap,
3432 						pgoff_t *start,
3433 						pgoff_t *end);
3434 
3435 unsigned long wp_shared_mapping_range(struct address_space *mapping,
3436 				      pgoff_t first_index, pgoff_t nr);
3437 #endif
3438 
3439 extern int sysctl_nr_trim_pages;
3440 
3441 #ifdef CONFIG_PRINTK
3442 void mem_dump_obj(void *object);
3443 #else
mem_dump_obj(void * object)3444 static inline void mem_dump_obj(void *object) {}
3445 #endif
3446 
3447 /**
3448  * seal_check_future_write - Check for F_SEAL_FUTURE_WRITE flag and handle it
3449  * @seals: the seals to check
3450  * @vma: the vma to operate on
3451  *
3452  * Check whether F_SEAL_FUTURE_WRITE is set; if so, do proper check/handling on
3453  * the vma flags.  Return 0 if check pass, or <0 for errors.
3454  */
seal_check_future_write(int seals,struct vm_area_struct * vma)3455 static inline int seal_check_future_write(int seals, struct vm_area_struct *vma)
3456 {
3457 	if (seals & F_SEAL_FUTURE_WRITE) {
3458 		/*
3459 		 * New PROT_WRITE and MAP_SHARED mmaps are not allowed when
3460 		 * "future write" seal active.
3461 		 */
3462 		if ((vma->vm_flags & VM_SHARED) && (vma->vm_flags & VM_WRITE))
3463 			return -EPERM;
3464 
3465 		/*
3466 		 * Since an F_SEAL_FUTURE_WRITE sealed memfd can be mapped as
3467 		 * MAP_SHARED and read-only, take care to not allow mprotect to
3468 		 * revert protections on such mappings. Do this only for shared
3469 		 * mappings. For private mappings, don't need to mask
3470 		 * VM_MAYWRITE as we still want them to be COW-writable.
3471 		 */
3472 		if (vma->vm_flags & VM_SHARED)
3473 			vma->vm_flags &= ~(VM_MAYWRITE);
3474 	}
3475 
3476 	return 0;
3477 }
3478 
3479 #ifdef CONFIG_ANON_VMA_NAME
3480 int madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
3481 			  unsigned long len_in,
3482 			  struct anon_vma_name *anon_name);
3483 #else
3484 static inline int
madvise_set_anon_name(struct mm_struct * mm,unsigned long start,unsigned long len_in,struct anon_vma_name * anon_name)3485 madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
3486 		      unsigned long len_in, struct anon_vma_name *anon_name) {
3487 	return 0;
3488 }
3489 #endif
3490 
3491 #endif /* _LINUX_MM_H */
3492