xref: /openbmc/linux/include/linux/mm.h (revision 707e9a6c)
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 #include <linux/slab.h>
33 
34 struct mempolicy;
35 struct anon_vma;
36 struct anon_vma_chain;
37 struct user_struct;
38 struct pt_regs;
39 
40 extern int sysctl_page_lock_unfairness;
41 
42 void mm_core_init(void);
43 void init_mm_internals(void);
44 
45 #ifndef CONFIG_NUMA		/* Don't use mapnrs, do it properly */
46 extern unsigned long max_mapnr;
47 
set_max_mapnr(unsigned long limit)48 static inline void set_max_mapnr(unsigned long limit)
49 {
50 	max_mapnr = limit;
51 }
52 #else
set_max_mapnr(unsigned long limit)53 static inline void set_max_mapnr(unsigned long limit) { }
54 #endif
55 
56 extern atomic_long_t _totalram_pages;
totalram_pages(void)57 static inline unsigned long totalram_pages(void)
58 {
59 	return (unsigned long)atomic_long_read(&_totalram_pages);
60 }
61 
totalram_pages_inc(void)62 static inline void totalram_pages_inc(void)
63 {
64 	atomic_long_inc(&_totalram_pages);
65 }
66 
totalram_pages_dec(void)67 static inline void totalram_pages_dec(void)
68 {
69 	atomic_long_dec(&_totalram_pages);
70 }
71 
totalram_pages_add(long count)72 static inline void totalram_pages_add(long count)
73 {
74 	atomic_long_add(count, &_totalram_pages);
75 }
76 
77 extern void * high_memory;
78 extern int page_cluster;
79 extern const int page_cluster_max;
80 
81 #ifdef CONFIG_SYSCTL
82 extern int sysctl_legacy_va_layout;
83 #else
84 #define sysctl_legacy_va_layout 0
85 #endif
86 
87 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
88 extern const int mmap_rnd_bits_min;
89 extern const int mmap_rnd_bits_max;
90 extern int mmap_rnd_bits __read_mostly;
91 #endif
92 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
93 extern const int mmap_rnd_compat_bits_min;
94 extern const int mmap_rnd_compat_bits_max;
95 extern int mmap_rnd_compat_bits __read_mostly;
96 #endif
97 
98 #ifndef PHYSMEM_END
99 # define PHYSMEM_END	((1ULL << MAX_PHYSMEM_BITS) - 1)
100 #endif
101 
102 #include <asm/page.h>
103 #include <asm/processor.h>
104 
105 #ifndef __pa_symbol
106 #define __pa_symbol(x)  __pa(RELOC_HIDE((unsigned long)(x), 0))
107 #endif
108 
109 #ifndef page_to_virt
110 #define page_to_virt(x)	__va(PFN_PHYS(page_to_pfn(x)))
111 #endif
112 
113 #ifndef lm_alias
114 #define lm_alias(x)	__va(__pa_symbol(x))
115 #endif
116 
117 /*
118  * To prevent common memory management code establishing
119  * a zero page mapping on a read fault.
120  * This macro should be defined within <asm/pgtable.h>.
121  * s390 does this to prevent multiplexing of hardware bits
122  * related to the physical page in case of virtualization.
123  */
124 #ifndef mm_forbids_zeropage
125 #define mm_forbids_zeropage(X)	(0)
126 #endif
127 
128 /*
129  * On some architectures it is expensive to call memset() for small sizes.
130  * If an architecture decides to implement their own version of
131  * mm_zero_struct_page they should wrap the defines below in a #ifndef and
132  * define their own version of this macro in <asm/pgtable.h>
133  */
134 #if BITS_PER_LONG == 64
135 /* This function must be updated when the size of struct page grows above 96
136  * or reduces below 56. The idea that compiler optimizes out switch()
137  * statement, and only leaves move/store instructions. Also the compiler can
138  * combine write statements if they are both assignments and can be reordered,
139  * this can result in several of the writes here being dropped.
140  */
141 #define	mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
__mm_zero_struct_page(struct page * page)142 static inline void __mm_zero_struct_page(struct page *page)
143 {
144 	unsigned long *_pp = (void *)page;
145 
146 	 /* Check that struct page is either 56, 64, 72, 80, 88 or 96 bytes */
147 	BUILD_BUG_ON(sizeof(struct page) & 7);
148 	BUILD_BUG_ON(sizeof(struct page) < 56);
149 	BUILD_BUG_ON(sizeof(struct page) > 96);
150 
151 	switch (sizeof(struct page)) {
152 	case 96:
153 		_pp[11] = 0;
154 		fallthrough;
155 	case 88:
156 		_pp[10] = 0;
157 		fallthrough;
158 	case 80:
159 		_pp[9] = 0;
160 		fallthrough;
161 	case 72:
162 		_pp[8] = 0;
163 		fallthrough;
164 	case 64:
165 		_pp[7] = 0;
166 		fallthrough;
167 	case 56:
168 		_pp[6] = 0;
169 		_pp[5] = 0;
170 		_pp[4] = 0;
171 		_pp[3] = 0;
172 		_pp[2] = 0;
173 		_pp[1] = 0;
174 		_pp[0] = 0;
175 	}
176 }
177 #else
178 #define mm_zero_struct_page(pp)  ((void)memset((pp), 0, sizeof(struct page)))
179 #endif
180 
181 /*
182  * Default maximum number of active map areas, this limits the number of vmas
183  * per mm struct. Users can overwrite this number by sysctl but there is a
184  * problem.
185  *
186  * When a program's coredump is generated as ELF format, a section is created
187  * per a vma. In ELF, the number of sections is represented in unsigned short.
188  * This means the number of sections should be smaller than 65535 at coredump.
189  * Because the kernel adds some informative sections to a image of program at
190  * generating coredump, we need some margin. The number of extra sections is
191  * 1-3 now and depends on arch. We use "5" as safe margin, here.
192  *
193  * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
194  * not a hard limit any more. Although some userspace tools can be surprised by
195  * that.
196  */
197 #define MAPCOUNT_ELF_CORE_MARGIN	(5)
198 #define DEFAULT_MAX_MAP_COUNT	(USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
199 
200 extern int sysctl_max_map_count;
201 
202 extern unsigned long sysctl_user_reserve_kbytes;
203 extern unsigned long sysctl_admin_reserve_kbytes;
204 
205 extern int sysctl_overcommit_memory;
206 extern int sysctl_overcommit_ratio;
207 extern unsigned long sysctl_overcommit_kbytes;
208 
209 int overcommit_ratio_handler(struct ctl_table *, int, void *, size_t *,
210 		loff_t *);
211 int overcommit_kbytes_handler(struct ctl_table *, int, void *, size_t *,
212 		loff_t *);
213 int overcommit_policy_handler(struct ctl_table *, int, void *, size_t *,
214 		loff_t *);
215 
216 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
217 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
218 #define folio_page_idx(folio, p)	(page_to_pfn(p) - folio_pfn(folio))
219 #else
220 #define nth_page(page,n) ((page) + (n))
221 #define folio_page_idx(folio, p)	((p) - &(folio)->page)
222 #endif
223 
224 /* to align the pointer to the (next) page boundary */
225 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
226 
227 /* to align the pointer to the (prev) page boundary */
228 #define PAGE_ALIGN_DOWN(addr) ALIGN_DOWN(addr, PAGE_SIZE)
229 
230 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
231 #define PAGE_ALIGNED(addr)	IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
232 
233 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
lru_to_folio(struct list_head * head)234 static inline struct folio *lru_to_folio(struct list_head *head)
235 {
236 	return list_entry((head)->prev, struct folio, lru);
237 }
238 
239 void setup_initial_init_mm(void *start_code, void *end_code,
240 			   void *end_data, void *brk);
241 
242 /*
243  * Linux kernel virtual memory manager primitives.
244  * The idea being to have a "virtual" mm in the same way
245  * we have a virtual fs - giving a cleaner interface to the
246  * mm details, and allowing different kinds of memory mappings
247  * (from shared memory to executable loading to arbitrary
248  * mmap() functions).
249  */
250 
251 struct vm_area_struct *vm_area_alloc(struct mm_struct *);
252 struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
253 void vm_area_free(struct vm_area_struct *);
254 /* Use only if VMA has no other users */
255 void __vm_area_free(struct vm_area_struct *vma);
256 
257 #ifndef CONFIG_MMU
258 extern struct rb_root nommu_region_tree;
259 extern struct rw_semaphore nommu_region_sem;
260 
261 extern unsigned int kobjsize(const void *objp);
262 #endif
263 
264 /*
265  * vm_flags in vm_area_struct, see mm_types.h.
266  * When changing, update also include/trace/events/mmflags.h
267  */
268 #define VM_NONE		0x00000000
269 
270 #define VM_READ		0x00000001	/* currently active flags */
271 #define VM_WRITE	0x00000002
272 #define VM_EXEC		0x00000004
273 #define VM_SHARED	0x00000008
274 
275 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
276 #define VM_MAYREAD	0x00000010	/* limits for mprotect() etc */
277 #define VM_MAYWRITE	0x00000020
278 #define VM_MAYEXEC	0x00000040
279 #define VM_MAYSHARE	0x00000080
280 
281 #define VM_GROWSDOWN	0x00000100	/* general info on the segment */
282 #ifdef CONFIG_MMU
283 #define VM_UFFD_MISSING	0x00000200	/* missing pages tracking */
284 #else /* CONFIG_MMU */
285 #define VM_MAYOVERLAY	0x00000200	/* nommu: R/O MAP_PRIVATE mapping that might overlay a file mapping */
286 #define VM_UFFD_MISSING	0
287 #endif /* CONFIG_MMU */
288 #define VM_PFNMAP	0x00000400	/* Page-ranges managed without "struct page", just pure PFN */
289 #define VM_UFFD_WP	0x00001000	/* wrprotect pages tracking */
290 
291 #define VM_LOCKED	0x00002000
292 #define VM_IO           0x00004000	/* Memory mapped I/O or similar */
293 
294 					/* Used by sys_madvise() */
295 #define VM_SEQ_READ	0x00008000	/* App will access data sequentially */
296 #define VM_RAND_READ	0x00010000	/* App will not benefit from clustered reads */
297 
298 #define VM_DONTCOPY	0x00020000      /* Do not copy this vma on fork */
299 #define VM_DONTEXPAND	0x00040000	/* Cannot expand with mremap() */
300 #define VM_LOCKONFAULT	0x00080000	/* Lock the pages covered when they are faulted in */
301 #define VM_ACCOUNT	0x00100000	/* Is a VM accounted object */
302 #define VM_NORESERVE	0x00200000	/* should the VM suppress accounting */
303 #define VM_HUGETLB	0x00400000	/* Huge TLB Page VM */
304 #define VM_SYNC		0x00800000	/* Synchronous page faults */
305 #define VM_ARCH_1	0x01000000	/* Architecture-specific flag */
306 #define VM_WIPEONFORK	0x02000000	/* Wipe VMA contents in child. */
307 #define VM_DONTDUMP	0x04000000	/* Do not include in the core dump */
308 
309 #ifdef CONFIG_MEM_SOFT_DIRTY
310 # define VM_SOFTDIRTY	0x08000000	/* Not soft dirty clean area */
311 #else
312 # define VM_SOFTDIRTY	0
313 #endif
314 
315 #define VM_MIXEDMAP	0x10000000	/* Can contain "struct page" and pure PFN pages */
316 #define VM_HUGEPAGE	0x20000000	/* MADV_HUGEPAGE marked this vma */
317 #define VM_NOHUGEPAGE	0x40000000	/* MADV_NOHUGEPAGE marked this vma */
318 #define VM_MERGEABLE	0x80000000	/* KSM may merge identical pages */
319 
320 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
321 #define VM_HIGH_ARCH_BIT_0	32	/* bit only usable on 64-bit architectures */
322 #define VM_HIGH_ARCH_BIT_1	33	/* bit only usable on 64-bit architectures */
323 #define VM_HIGH_ARCH_BIT_2	34	/* bit only usable on 64-bit architectures */
324 #define VM_HIGH_ARCH_BIT_3	35	/* bit only usable on 64-bit architectures */
325 #define VM_HIGH_ARCH_BIT_4	36	/* bit only usable on 64-bit architectures */
326 #define VM_HIGH_ARCH_BIT_5	37	/* bit only usable on 64-bit architectures */
327 #define VM_HIGH_ARCH_0	BIT(VM_HIGH_ARCH_BIT_0)
328 #define VM_HIGH_ARCH_1	BIT(VM_HIGH_ARCH_BIT_1)
329 #define VM_HIGH_ARCH_2	BIT(VM_HIGH_ARCH_BIT_2)
330 #define VM_HIGH_ARCH_3	BIT(VM_HIGH_ARCH_BIT_3)
331 #define VM_HIGH_ARCH_4	BIT(VM_HIGH_ARCH_BIT_4)
332 #define VM_HIGH_ARCH_5	BIT(VM_HIGH_ARCH_BIT_5)
333 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
334 
335 #ifdef CONFIG_ARCH_HAS_PKEYS
336 # define VM_PKEY_SHIFT	VM_HIGH_ARCH_BIT_0
337 # define VM_PKEY_BIT0	VM_HIGH_ARCH_0	/* A protection key is a 4-bit value */
338 # define VM_PKEY_BIT1	VM_HIGH_ARCH_1	/* on x86 and 5-bit value on ppc64   */
339 # define VM_PKEY_BIT2	VM_HIGH_ARCH_2
340 # define VM_PKEY_BIT3	VM_HIGH_ARCH_3
341 #ifdef CONFIG_PPC
342 # define VM_PKEY_BIT4  VM_HIGH_ARCH_4
343 #else
344 # define VM_PKEY_BIT4  0
345 #endif
346 #endif /* CONFIG_ARCH_HAS_PKEYS */
347 
348 #ifdef CONFIG_X86_USER_SHADOW_STACK
349 /*
350  * VM_SHADOW_STACK should not be set with VM_SHARED because of lack of
351  * support core mm.
352  *
353  * These VMAs will get a single end guard page. This helps userspace protect
354  * itself from attacks. A single page is enough for current shadow stack archs
355  * (x86). See the comments near alloc_shstk() in arch/x86/kernel/shstk.c
356  * for more details on the guard size.
357  */
358 # define VM_SHADOW_STACK	VM_HIGH_ARCH_5
359 #else
360 # define VM_SHADOW_STACK	VM_NONE
361 #endif
362 
363 #if defined(CONFIG_X86)
364 # define VM_PAT		VM_ARCH_1	/* PAT reserves whole VMA at once (x86) */
365 #elif defined(CONFIG_PPC)
366 # define VM_SAO		VM_ARCH_1	/* Strong Access Ordering (powerpc) */
367 #elif defined(CONFIG_PARISC)
368 # define VM_GROWSUP	VM_ARCH_1
369 #elif defined(CONFIG_IA64)
370 # define VM_GROWSUP	VM_ARCH_1
371 #elif defined(CONFIG_SPARC64)
372 # define VM_SPARC_ADI	VM_ARCH_1	/* Uses ADI tag for access control */
373 # define VM_ARCH_CLEAR	VM_SPARC_ADI
374 #elif defined(CONFIG_ARM64)
375 # define VM_ARM64_BTI	VM_ARCH_1	/* BTI guarded page, a.k.a. GP bit */
376 # define VM_ARCH_CLEAR	VM_ARM64_BTI
377 #elif !defined(CONFIG_MMU)
378 # define VM_MAPPED_COPY	VM_ARCH_1	/* T if mapped copy of data (nommu mmap) */
379 #endif
380 
381 #if defined(CONFIG_ARM64_MTE)
382 # define VM_MTE		VM_HIGH_ARCH_0	/* Use Tagged memory for access control */
383 # define VM_MTE_ALLOWED	VM_HIGH_ARCH_1	/* Tagged memory permitted */
384 #else
385 # define VM_MTE		VM_NONE
386 # define VM_MTE_ALLOWED	VM_NONE
387 #endif
388 
389 #ifndef VM_GROWSUP
390 # define VM_GROWSUP	VM_NONE
391 #endif
392 
393 #ifdef CONFIG_HAVE_ARCH_USERFAULTFD_MINOR
394 # define VM_UFFD_MINOR_BIT	38
395 # define VM_UFFD_MINOR		BIT(VM_UFFD_MINOR_BIT)	/* UFFD minor faults */
396 #else /* !CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
397 # define VM_UFFD_MINOR		VM_NONE
398 #endif /* CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
399 
400 /* Bits set in the VMA until the stack is in its final location */
401 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ | VM_STACK_EARLY)
402 
403 #define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0)
404 
405 /* Common data flag combinations */
406 #define VM_DATA_FLAGS_TSK_EXEC	(VM_READ | VM_WRITE | TASK_EXEC | \
407 				 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
408 #define VM_DATA_FLAGS_NON_EXEC	(VM_READ | VM_WRITE | VM_MAYREAD | \
409 				 VM_MAYWRITE | VM_MAYEXEC)
410 #define VM_DATA_FLAGS_EXEC	(VM_READ | VM_WRITE | VM_EXEC | \
411 				 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
412 
413 #ifndef VM_DATA_DEFAULT_FLAGS		/* arch can override this */
414 #define VM_DATA_DEFAULT_FLAGS  VM_DATA_FLAGS_EXEC
415 #endif
416 
417 #ifndef VM_STACK_DEFAULT_FLAGS		/* arch can override this */
418 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
419 #endif
420 
421 #define VM_STARTGAP_FLAGS (VM_GROWSDOWN | VM_SHADOW_STACK)
422 
423 #ifdef CONFIG_STACK_GROWSUP
424 #define VM_STACK	VM_GROWSUP
425 #define VM_STACK_EARLY	VM_GROWSDOWN
426 #else
427 #define VM_STACK	VM_GROWSDOWN
428 #define VM_STACK_EARLY	0
429 #endif
430 
431 #define VM_STACK_FLAGS	(VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
432 
433 /* VMA basic access permission flags */
434 #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
435 
436 
437 /*
438  * Special vmas that are non-mergable, non-mlock()able.
439  */
440 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
441 
442 /* This mask prevents VMA from being scanned with khugepaged */
443 #define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB)
444 
445 /* This mask defines which mm->def_flags a process can inherit its parent */
446 #define VM_INIT_DEF_MASK	VM_NOHUGEPAGE
447 
448 /* This mask represents all the VMA flag bits used by mlock */
449 #define VM_LOCKED_MASK	(VM_LOCKED | VM_LOCKONFAULT)
450 
451 /* Arch-specific flags to clear when updating VM flags on protection change */
452 #ifndef VM_ARCH_CLEAR
453 # define VM_ARCH_CLEAR	VM_NONE
454 #endif
455 #define VM_FLAGS_CLEAR	(ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
456 
457 /*
458  * mapping from the currently active vm_flags protection bits (the
459  * low four bits) to a page protection mask..
460  */
461 
462 /*
463  * The default fault flags that should be used by most of the
464  * arch-specific page fault handlers.
465  */
466 #define FAULT_FLAG_DEFAULT  (FAULT_FLAG_ALLOW_RETRY | \
467 			     FAULT_FLAG_KILLABLE | \
468 			     FAULT_FLAG_INTERRUPTIBLE)
469 
470 /**
471  * fault_flag_allow_retry_first - check ALLOW_RETRY the first time
472  * @flags: Fault flags.
473  *
474  * This is mostly used for places where we want to try to avoid taking
475  * the mmap_lock for too long a time when waiting for another condition
476  * to change, in which case we can try to be polite to release the
477  * mmap_lock in the first round to avoid potential starvation of other
478  * processes that would also want the mmap_lock.
479  *
480  * Return: true if the page fault allows retry and this is the first
481  * attempt of the fault handling; false otherwise.
482  */
fault_flag_allow_retry_first(enum fault_flag flags)483 static inline bool fault_flag_allow_retry_first(enum fault_flag flags)
484 {
485 	return (flags & FAULT_FLAG_ALLOW_RETRY) &&
486 	    (!(flags & FAULT_FLAG_TRIED));
487 }
488 
489 #define FAULT_FLAG_TRACE \
490 	{ FAULT_FLAG_WRITE,		"WRITE" }, \
491 	{ FAULT_FLAG_MKWRITE,		"MKWRITE" }, \
492 	{ FAULT_FLAG_ALLOW_RETRY,	"ALLOW_RETRY" }, \
493 	{ FAULT_FLAG_RETRY_NOWAIT,	"RETRY_NOWAIT" }, \
494 	{ FAULT_FLAG_KILLABLE,		"KILLABLE" }, \
495 	{ FAULT_FLAG_TRIED,		"TRIED" }, \
496 	{ FAULT_FLAG_USER,		"USER" }, \
497 	{ FAULT_FLAG_REMOTE,		"REMOTE" }, \
498 	{ FAULT_FLAG_INSTRUCTION,	"INSTRUCTION" }, \
499 	{ FAULT_FLAG_INTERRUPTIBLE,	"INTERRUPTIBLE" }, \
500 	{ FAULT_FLAG_VMA_LOCK,		"VMA_LOCK" }
501 
502 /*
503  * vm_fault is filled by the pagefault handler and passed to the vma's
504  * ->fault function. The vma's ->fault is responsible for returning a bitmask
505  * of VM_FAULT_xxx flags that give details about how the fault was handled.
506  *
507  * MM layer fills up gfp_mask for page allocations but fault handler might
508  * alter it if its implementation requires a different allocation context.
509  *
510  * pgoff should be used in favour of virtual_address, if possible.
511  */
512 struct vm_fault {
513 	const struct {
514 		struct vm_area_struct *vma;	/* Target VMA */
515 		gfp_t gfp_mask;			/* gfp mask to be used for allocations */
516 		pgoff_t pgoff;			/* Logical page offset based on vma */
517 		unsigned long address;		/* Faulting virtual address - masked */
518 		unsigned long real_address;	/* Faulting virtual address - unmasked */
519 	};
520 	enum fault_flag flags;		/* FAULT_FLAG_xxx flags
521 					 * XXX: should really be 'const' */
522 	pmd_t *pmd;			/* Pointer to pmd entry matching
523 					 * the 'address' */
524 	pud_t *pud;			/* Pointer to pud entry matching
525 					 * the 'address'
526 					 */
527 	union {
528 		pte_t orig_pte;		/* Value of PTE at the time of fault */
529 		pmd_t orig_pmd;		/* Value of PMD at the time of fault,
530 					 * used by PMD fault only.
531 					 */
532 	};
533 
534 	struct page *cow_page;		/* Page handler may use for COW fault */
535 	struct page *page;		/* ->fault handlers should return a
536 					 * page here, unless VM_FAULT_NOPAGE
537 					 * is set (which is also implied by
538 					 * VM_FAULT_ERROR).
539 					 */
540 	/* These three entries are valid only while holding ptl lock */
541 	pte_t *pte;			/* Pointer to pte entry matching
542 					 * the 'address'. NULL if the page
543 					 * table hasn't been allocated.
544 					 */
545 	spinlock_t *ptl;		/* Page table lock.
546 					 * Protects pte page table if 'pte'
547 					 * is not NULL, otherwise pmd.
548 					 */
549 	pgtable_t prealloc_pte;		/* Pre-allocated pte page table.
550 					 * vm_ops->map_pages() sets up a page
551 					 * table from atomic context.
552 					 * do_fault_around() pre-allocates
553 					 * page table to avoid allocation from
554 					 * atomic context.
555 					 */
556 };
557 
558 /*
559  * These are the virtual MM functions - opening of an area, closing and
560  * unmapping it (needed to keep files on disk up-to-date etc), pointer
561  * to the functions called when a no-page or a wp-page exception occurs.
562  */
563 struct vm_operations_struct {
564 	void (*open)(struct vm_area_struct * area);
565 	/**
566 	 * @close: Called when the VMA is being removed from the MM.
567 	 * Context: User context.  May sleep.  Caller holds mmap_lock.
568 	 */
569 	void (*close)(struct vm_area_struct * area);
570 	/* Called any time before splitting to check if it's allowed */
571 	int (*may_split)(struct vm_area_struct *area, unsigned long addr);
572 	int (*mremap)(struct vm_area_struct *area);
573 	/*
574 	 * Called by mprotect() to make driver-specific permission
575 	 * checks before mprotect() is finalised.   The VMA must not
576 	 * be modified.  Returns 0 if mprotect() can proceed.
577 	 */
578 	int (*mprotect)(struct vm_area_struct *vma, unsigned long start,
579 			unsigned long end, unsigned long newflags);
580 	vm_fault_t (*fault)(struct vm_fault *vmf);
581 	vm_fault_t (*huge_fault)(struct vm_fault *vmf, unsigned int order);
582 	vm_fault_t (*map_pages)(struct vm_fault *vmf,
583 			pgoff_t start_pgoff, pgoff_t end_pgoff);
584 	unsigned long (*pagesize)(struct vm_area_struct * area);
585 
586 	/* notification that a previously read-only page is about to become
587 	 * writable, if an error is returned it will cause a SIGBUS */
588 	vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
589 
590 	/* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
591 	vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
592 
593 	/* called by access_process_vm when get_user_pages() fails, typically
594 	 * for use by special VMAs. See also generic_access_phys() for a generic
595 	 * implementation useful for any iomem mapping.
596 	 */
597 	int (*access)(struct vm_area_struct *vma, unsigned long addr,
598 		      void *buf, int len, int write);
599 
600 	/* Called by the /proc/PID/maps code to ask the vma whether it
601 	 * has a special name.  Returning non-NULL will also cause this
602 	 * vma to be dumped unconditionally. */
603 	const char *(*name)(struct vm_area_struct *vma);
604 
605 #ifdef CONFIG_NUMA
606 	/*
607 	 * set_policy() op must add a reference to any non-NULL @new mempolicy
608 	 * to hold the policy upon return.  Caller should pass NULL @new to
609 	 * remove a policy and fall back to surrounding context--i.e. do not
610 	 * install a MPOL_DEFAULT policy, nor the task or system default
611 	 * mempolicy.
612 	 */
613 	int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
614 
615 	/*
616 	 * get_policy() op must add reference [mpol_get()] to any policy at
617 	 * (vma,addr) marked as MPOL_SHARED.  The shared policy infrastructure
618 	 * in mm/mempolicy.c will do this automatically.
619 	 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
620 	 * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
621 	 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
622 	 * must return NULL--i.e., do not "fallback" to task or system default
623 	 * policy.
624 	 */
625 	struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
626 					unsigned long addr);
627 #endif
628 	/*
629 	 * Called by vm_normal_page() for special PTEs to find the
630 	 * page for @addr.  This is useful if the default behavior
631 	 * (using pte_page()) would not find the correct page.
632 	 */
633 	struct page *(*find_special_page)(struct vm_area_struct *vma,
634 					  unsigned long addr);
635 };
636 
637 #ifdef CONFIG_NUMA_BALANCING
vma_numab_state_init(struct vm_area_struct * vma)638 static inline void vma_numab_state_init(struct vm_area_struct *vma)
639 {
640 	vma->numab_state = NULL;
641 }
vma_numab_state_free(struct vm_area_struct * vma)642 static inline void vma_numab_state_free(struct vm_area_struct *vma)
643 {
644 	kfree(vma->numab_state);
645 }
646 #else
vma_numab_state_init(struct vm_area_struct * vma)647 static inline void vma_numab_state_init(struct vm_area_struct *vma) {}
vma_numab_state_free(struct vm_area_struct * vma)648 static inline void vma_numab_state_free(struct vm_area_struct *vma) {}
649 #endif /* CONFIG_NUMA_BALANCING */
650 
651 #ifdef CONFIG_PER_VMA_LOCK
652 /*
653  * Try to read-lock a vma. The function is allowed to occasionally yield false
654  * locked result to avoid performance overhead, in which case we fall back to
655  * using mmap_lock. The function should never yield false unlocked result.
656  */
vma_start_read(struct vm_area_struct * vma)657 static inline bool vma_start_read(struct vm_area_struct *vma)
658 {
659 	/*
660 	 * Check before locking. A race might cause false locked result.
661 	 * We can use READ_ONCE() for the mm_lock_seq here, and don't need
662 	 * ACQUIRE semantics, because this is just a lockless check whose result
663 	 * we don't rely on for anything - the mm_lock_seq read against which we
664 	 * need ordering is below.
665 	 */
666 	if (READ_ONCE(vma->vm_lock_seq) == READ_ONCE(vma->vm_mm->mm_lock_seq))
667 		return false;
668 
669 	if (unlikely(down_read_trylock(&vma->vm_lock->lock) == 0))
670 		return false;
671 
672 	/*
673 	 * Overflow might produce false locked result.
674 	 * False unlocked result is impossible because we modify and check
675 	 * vma->vm_lock_seq under vma->vm_lock protection and mm->mm_lock_seq
676 	 * modification invalidates all existing locks.
677 	 *
678 	 * We must use ACQUIRE semantics for the mm_lock_seq so that if we are
679 	 * racing with vma_end_write_all(), we only start reading from the VMA
680 	 * after it has been unlocked.
681 	 * This pairs with RELEASE semantics in vma_end_write_all().
682 	 */
683 	if (unlikely(vma->vm_lock_seq == smp_load_acquire(&vma->vm_mm->mm_lock_seq))) {
684 		up_read(&vma->vm_lock->lock);
685 		return false;
686 	}
687 	return true;
688 }
689 
vma_end_read(struct vm_area_struct * vma)690 static inline void vma_end_read(struct vm_area_struct *vma)
691 {
692 	rcu_read_lock(); /* keeps vma alive till the end of up_read */
693 	up_read(&vma->vm_lock->lock);
694 	rcu_read_unlock();
695 }
696 
697 /* WARNING! Can only be used if mmap_lock is expected to be write-locked */
__is_vma_write_locked(struct vm_area_struct * vma,int * mm_lock_seq)698 static bool __is_vma_write_locked(struct vm_area_struct *vma, int *mm_lock_seq)
699 {
700 	mmap_assert_write_locked(vma->vm_mm);
701 
702 	/*
703 	 * current task is holding mmap_write_lock, both vma->vm_lock_seq and
704 	 * mm->mm_lock_seq can't be concurrently modified.
705 	 */
706 	*mm_lock_seq = vma->vm_mm->mm_lock_seq;
707 	return (vma->vm_lock_seq == *mm_lock_seq);
708 }
709 
710 /*
711  * Begin writing to a VMA.
712  * Exclude concurrent readers under the per-VMA lock until the currently
713  * write-locked mmap_lock is dropped or downgraded.
714  */
vma_start_write(struct vm_area_struct * vma)715 static inline void vma_start_write(struct vm_area_struct *vma)
716 {
717 	int mm_lock_seq;
718 
719 	if (__is_vma_write_locked(vma, &mm_lock_seq))
720 		return;
721 
722 	down_write(&vma->vm_lock->lock);
723 	/*
724 	 * We should use WRITE_ONCE() here because we can have concurrent reads
725 	 * from the early lockless pessimistic check in vma_start_read().
726 	 * We don't really care about the correctness of that early check, but
727 	 * we should use WRITE_ONCE() for cleanliness and to keep KCSAN happy.
728 	 */
729 	WRITE_ONCE(vma->vm_lock_seq, mm_lock_seq);
730 	up_write(&vma->vm_lock->lock);
731 }
732 
vma_assert_write_locked(struct vm_area_struct * vma)733 static inline void vma_assert_write_locked(struct vm_area_struct *vma)
734 {
735 	int mm_lock_seq;
736 
737 	VM_BUG_ON_VMA(!__is_vma_write_locked(vma, &mm_lock_seq), vma);
738 }
739 
vma_assert_locked(struct vm_area_struct * vma)740 static inline void vma_assert_locked(struct vm_area_struct *vma)
741 {
742 	if (!rwsem_is_locked(&vma->vm_lock->lock))
743 		vma_assert_write_locked(vma);
744 }
745 
vma_mark_detached(struct vm_area_struct * vma,bool detached)746 static inline void vma_mark_detached(struct vm_area_struct *vma, bool detached)
747 {
748 	/* When detaching vma should be write-locked */
749 	if (detached)
750 		vma_assert_write_locked(vma);
751 	vma->detached = detached;
752 }
753 
release_fault_lock(struct vm_fault * vmf)754 static inline void release_fault_lock(struct vm_fault *vmf)
755 {
756 	if (vmf->flags & FAULT_FLAG_VMA_LOCK)
757 		vma_end_read(vmf->vma);
758 	else
759 		mmap_read_unlock(vmf->vma->vm_mm);
760 }
761 
assert_fault_locked(struct vm_fault * vmf)762 static inline void assert_fault_locked(struct vm_fault *vmf)
763 {
764 	if (vmf->flags & FAULT_FLAG_VMA_LOCK)
765 		vma_assert_locked(vmf->vma);
766 	else
767 		mmap_assert_locked(vmf->vma->vm_mm);
768 }
769 
770 struct vm_area_struct *lock_vma_under_rcu(struct mm_struct *mm,
771 					  unsigned long address);
772 
773 #else /* CONFIG_PER_VMA_LOCK */
774 
vma_start_read(struct vm_area_struct * vma)775 static inline bool vma_start_read(struct vm_area_struct *vma)
776 		{ return false; }
vma_end_read(struct vm_area_struct * vma)777 static inline void vma_end_read(struct vm_area_struct *vma) {}
vma_start_write(struct vm_area_struct * vma)778 static inline void vma_start_write(struct vm_area_struct *vma) {}
vma_assert_write_locked(struct vm_area_struct * vma)779 static inline void vma_assert_write_locked(struct vm_area_struct *vma)
780 		{ mmap_assert_write_locked(vma->vm_mm); }
vma_mark_detached(struct vm_area_struct * vma,bool detached)781 static inline void vma_mark_detached(struct vm_area_struct *vma,
782 				     bool detached) {}
783 
lock_vma_under_rcu(struct mm_struct * mm,unsigned long address)784 static inline struct vm_area_struct *lock_vma_under_rcu(struct mm_struct *mm,
785 		unsigned long address)
786 {
787 	return NULL;
788 }
789 
release_fault_lock(struct vm_fault * vmf)790 static inline void release_fault_lock(struct vm_fault *vmf)
791 {
792 	mmap_read_unlock(vmf->vma->vm_mm);
793 }
794 
assert_fault_locked(struct vm_fault * vmf)795 static inline void assert_fault_locked(struct vm_fault *vmf)
796 {
797 	mmap_assert_locked(vmf->vma->vm_mm);
798 }
799 
800 #endif /* CONFIG_PER_VMA_LOCK */
801 
802 extern const struct vm_operations_struct vma_dummy_vm_ops;
803 
804 /*
805  * WARNING: vma_init does not initialize vma->vm_lock.
806  * Use vm_area_alloc()/vm_area_free() if vma needs locking.
807  */
vma_init(struct vm_area_struct * vma,struct mm_struct * mm)808 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
809 {
810 	memset(vma, 0, sizeof(*vma));
811 	vma->vm_mm = mm;
812 	vma->vm_ops = &vma_dummy_vm_ops;
813 	INIT_LIST_HEAD(&vma->anon_vma_chain);
814 	vma_mark_detached(vma, false);
815 	vma_numab_state_init(vma);
816 }
817 
818 /* Use when VMA is not part of the VMA tree and needs no locking */
vm_flags_init(struct vm_area_struct * vma,vm_flags_t flags)819 static inline void vm_flags_init(struct vm_area_struct *vma,
820 				 vm_flags_t flags)
821 {
822 	ACCESS_PRIVATE(vma, __vm_flags) = flags;
823 }
824 
825 /*
826  * Use when VMA is part of the VMA tree and modifications need coordination
827  * Note: vm_flags_reset and vm_flags_reset_once do not lock the vma and
828  * it should be locked explicitly beforehand.
829  */
vm_flags_reset(struct vm_area_struct * vma,vm_flags_t flags)830 static inline void vm_flags_reset(struct vm_area_struct *vma,
831 				  vm_flags_t flags)
832 {
833 	vma_assert_write_locked(vma);
834 	vm_flags_init(vma, flags);
835 }
836 
vm_flags_reset_once(struct vm_area_struct * vma,vm_flags_t flags)837 static inline void vm_flags_reset_once(struct vm_area_struct *vma,
838 				       vm_flags_t flags)
839 {
840 	vma_assert_write_locked(vma);
841 	WRITE_ONCE(ACCESS_PRIVATE(vma, __vm_flags), flags);
842 }
843 
vm_flags_set(struct vm_area_struct * vma,vm_flags_t flags)844 static inline void vm_flags_set(struct vm_area_struct *vma,
845 				vm_flags_t flags)
846 {
847 	vma_start_write(vma);
848 	ACCESS_PRIVATE(vma, __vm_flags) |= flags;
849 }
850 
vm_flags_clear(struct vm_area_struct * vma,vm_flags_t flags)851 static inline void vm_flags_clear(struct vm_area_struct *vma,
852 				  vm_flags_t flags)
853 {
854 	vma_start_write(vma);
855 	ACCESS_PRIVATE(vma, __vm_flags) &= ~flags;
856 }
857 
858 /*
859  * Use only if VMA is not part of the VMA tree or has no other users and
860  * therefore needs no locking.
861  */
__vm_flags_mod(struct vm_area_struct * vma,vm_flags_t set,vm_flags_t clear)862 static inline void __vm_flags_mod(struct vm_area_struct *vma,
863 				  vm_flags_t set, vm_flags_t clear)
864 {
865 	vm_flags_init(vma, (vma->vm_flags | set) & ~clear);
866 }
867 
868 /*
869  * Use only when the order of set/clear operations is unimportant, otherwise
870  * use vm_flags_{set|clear} explicitly.
871  */
vm_flags_mod(struct vm_area_struct * vma,vm_flags_t set,vm_flags_t clear)872 static inline void vm_flags_mod(struct vm_area_struct *vma,
873 				vm_flags_t set, vm_flags_t clear)
874 {
875 	vma_start_write(vma);
876 	__vm_flags_mod(vma, set, clear);
877 }
878 
vma_set_anonymous(struct vm_area_struct * vma)879 static inline void vma_set_anonymous(struct vm_area_struct *vma)
880 {
881 	vma->vm_ops = NULL;
882 }
883 
vma_is_anonymous(struct vm_area_struct * vma)884 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
885 {
886 	return !vma->vm_ops;
887 }
888 
889 /*
890  * Indicate if the VMA is a heap for the given task; for
891  * /proc/PID/maps that is the heap of the main task.
892  */
vma_is_initial_heap(const struct vm_area_struct * vma)893 static inline bool vma_is_initial_heap(const struct vm_area_struct *vma)
894 {
895        return vma->vm_start <= vma->vm_mm->brk &&
896 		vma->vm_end >= vma->vm_mm->start_brk;
897 }
898 
899 /*
900  * Indicate if the VMA is a stack for the given task; for
901  * /proc/PID/maps that is the stack of the main task.
902  */
vma_is_initial_stack(const struct vm_area_struct * vma)903 static inline bool vma_is_initial_stack(const struct vm_area_struct *vma)
904 {
905 	/*
906 	 * We make no effort to guess what a given thread considers to be
907 	 * its "stack".  It's not even well-defined for programs written
908 	 * languages like Go.
909 	 */
910        return vma->vm_start <= vma->vm_mm->start_stack &&
911 	       vma->vm_end >= vma->vm_mm->start_stack;
912 }
913 
vma_is_temporary_stack(struct vm_area_struct * vma)914 static inline bool vma_is_temporary_stack(struct vm_area_struct *vma)
915 {
916 	int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP);
917 
918 	if (!maybe_stack)
919 		return false;
920 
921 	if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) ==
922 						VM_STACK_INCOMPLETE_SETUP)
923 		return true;
924 
925 	return false;
926 }
927 
vma_is_foreign(struct vm_area_struct * vma)928 static inline bool vma_is_foreign(struct vm_area_struct *vma)
929 {
930 	if (!current->mm)
931 		return true;
932 
933 	if (current->mm != vma->vm_mm)
934 		return true;
935 
936 	return false;
937 }
938 
vma_is_accessible(struct vm_area_struct * vma)939 static inline bool vma_is_accessible(struct vm_area_struct *vma)
940 {
941 	return vma->vm_flags & VM_ACCESS_FLAGS;
942 }
943 
944 static inline
vma_find(struct vma_iterator * vmi,unsigned long max)945 struct vm_area_struct *vma_find(struct vma_iterator *vmi, unsigned long max)
946 {
947 	return mas_find(&vmi->mas, max - 1);
948 }
949 
vma_next(struct vma_iterator * vmi)950 static inline struct vm_area_struct *vma_next(struct vma_iterator *vmi)
951 {
952 	/*
953 	 * Uses mas_find() to get the first VMA when the iterator starts.
954 	 * Calling mas_next() could skip the first entry.
955 	 */
956 	return mas_find(&vmi->mas, ULONG_MAX);
957 }
958 
959 static inline
vma_iter_next_range(struct vma_iterator * vmi)960 struct vm_area_struct *vma_iter_next_range(struct vma_iterator *vmi)
961 {
962 	return mas_next_range(&vmi->mas, ULONG_MAX);
963 }
964 
965 
vma_prev(struct vma_iterator * vmi)966 static inline struct vm_area_struct *vma_prev(struct vma_iterator *vmi)
967 {
968 	return mas_prev(&vmi->mas, 0);
969 }
970 
971 static inline
vma_iter_prev_range(struct vma_iterator * vmi)972 struct vm_area_struct *vma_iter_prev_range(struct vma_iterator *vmi)
973 {
974 	return mas_prev_range(&vmi->mas, 0);
975 }
976 
vma_iter_addr(struct vma_iterator * vmi)977 static inline unsigned long vma_iter_addr(struct vma_iterator *vmi)
978 {
979 	return vmi->mas.index;
980 }
981 
vma_iter_end(struct vma_iterator * vmi)982 static inline unsigned long vma_iter_end(struct vma_iterator *vmi)
983 {
984 	return vmi->mas.last + 1;
985 }
vma_iter_bulk_alloc(struct vma_iterator * vmi,unsigned long count)986 static inline int vma_iter_bulk_alloc(struct vma_iterator *vmi,
987 				      unsigned long count)
988 {
989 	return mas_expected_entries(&vmi->mas, count);
990 }
991 
992 /* Free any unused preallocations */
vma_iter_free(struct vma_iterator * vmi)993 static inline void vma_iter_free(struct vma_iterator *vmi)
994 {
995 	mas_destroy(&vmi->mas);
996 }
997 
vma_iter_bulk_store(struct vma_iterator * vmi,struct vm_area_struct * vma)998 static inline int vma_iter_bulk_store(struct vma_iterator *vmi,
999 				      struct vm_area_struct *vma)
1000 {
1001 	vmi->mas.index = vma->vm_start;
1002 	vmi->mas.last = vma->vm_end - 1;
1003 	mas_store(&vmi->mas, vma);
1004 	if (unlikely(mas_is_err(&vmi->mas)))
1005 		return -ENOMEM;
1006 
1007 	return 0;
1008 }
1009 
vma_iter_invalidate(struct vma_iterator * vmi)1010 static inline void vma_iter_invalidate(struct vma_iterator *vmi)
1011 {
1012 	mas_pause(&vmi->mas);
1013 }
1014 
vma_iter_set(struct vma_iterator * vmi,unsigned long addr)1015 static inline void vma_iter_set(struct vma_iterator *vmi, unsigned long addr)
1016 {
1017 	mas_set(&vmi->mas, addr);
1018 }
1019 
1020 #define for_each_vma(__vmi, __vma)					\
1021 	while (((__vma) = vma_next(&(__vmi))) != NULL)
1022 
1023 /* The MM code likes to work with exclusive end addresses */
1024 #define for_each_vma_range(__vmi, __vma, __end)				\
1025 	while (((__vma) = vma_find(&(__vmi), (__end))) != NULL)
1026 
1027 #ifdef CONFIG_SHMEM
1028 /*
1029  * The vma_is_shmem is not inline because it is used only by slow
1030  * paths in userfault.
1031  */
1032 bool vma_is_shmem(struct vm_area_struct *vma);
1033 bool vma_is_anon_shmem(struct vm_area_struct *vma);
1034 #else
vma_is_shmem(struct vm_area_struct * vma)1035 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
vma_is_anon_shmem(struct vm_area_struct * vma)1036 static inline bool vma_is_anon_shmem(struct vm_area_struct *vma) { return false; }
1037 #endif
1038 
1039 int vma_is_stack_for_current(struct vm_area_struct *vma);
1040 
1041 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
1042 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
1043 
1044 struct mmu_gather;
1045 struct inode;
1046 
1047 /*
1048  * compound_order() can be called without holding a reference, which means
1049  * that niceties like page_folio() don't work.  These callers should be
1050  * prepared to handle wild return values.  For example, PG_head may be
1051  * set before the order is initialised, or this may be a tail page.
1052  * See compaction.c for some good examples.
1053  */
compound_order(struct page * page)1054 static inline unsigned int compound_order(struct page *page)
1055 {
1056 	struct folio *folio = (struct folio *)page;
1057 
1058 	if (!test_bit(PG_head, &folio->flags))
1059 		return 0;
1060 	return folio->_flags_1 & 0xff;
1061 }
1062 
1063 /**
1064  * folio_order - The allocation order of a folio.
1065  * @folio: The folio.
1066  *
1067  * A folio is composed of 2^order pages.  See get_order() for the definition
1068  * of order.
1069  *
1070  * Return: The order of the folio.
1071  */
folio_order(struct folio * folio)1072 static inline unsigned int folio_order(struct folio *folio)
1073 {
1074 	if (!folio_test_large(folio))
1075 		return 0;
1076 	return folio->_flags_1 & 0xff;
1077 }
1078 
1079 #include <linux/huge_mm.h>
1080 
1081 /*
1082  * Methods to modify the page usage count.
1083  *
1084  * What counts for a page usage:
1085  * - cache mapping   (page->mapping)
1086  * - private data    (page->private)
1087  * - page mapped in a task's page tables, each mapping
1088  *   is counted separately
1089  *
1090  * Also, many kernel routines increase the page count before a critical
1091  * routine so they can be sure the page doesn't go away from under them.
1092  */
1093 
1094 /*
1095  * Drop a ref, return true if the refcount fell to zero (the page has no users)
1096  */
put_page_testzero(struct page * page)1097 static inline int put_page_testzero(struct page *page)
1098 {
1099 	VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
1100 	return page_ref_dec_and_test(page);
1101 }
1102 
folio_put_testzero(struct folio * folio)1103 static inline int folio_put_testzero(struct folio *folio)
1104 {
1105 	return put_page_testzero(&folio->page);
1106 }
1107 
1108 /*
1109  * Try to grab a ref unless the page has a refcount of zero, return false if
1110  * that is the case.
1111  * This can be called when MMU is off so it must not access
1112  * any of the virtual mappings.
1113  */
get_page_unless_zero(struct page * page)1114 static inline bool get_page_unless_zero(struct page *page)
1115 {
1116 	return page_ref_add_unless(page, 1, 0);
1117 }
1118 
folio_get_nontail_page(struct page * page)1119 static inline struct folio *folio_get_nontail_page(struct page *page)
1120 {
1121 	if (unlikely(!get_page_unless_zero(page)))
1122 		return NULL;
1123 	return (struct folio *)page;
1124 }
1125 
1126 extern int page_is_ram(unsigned long pfn);
1127 
1128 enum {
1129 	REGION_INTERSECTS,
1130 	REGION_DISJOINT,
1131 	REGION_MIXED,
1132 };
1133 
1134 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
1135 		      unsigned long desc);
1136 
1137 /* Support for virtually mapped pages */
1138 struct page *vmalloc_to_page(const void *addr);
1139 unsigned long vmalloc_to_pfn(const void *addr);
1140 
1141 /*
1142  * Determine if an address is within the vmalloc range
1143  *
1144  * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
1145  * is no special casing required.
1146  */
1147 #ifdef CONFIG_MMU
1148 extern bool is_vmalloc_addr(const void *x);
1149 extern int is_vmalloc_or_module_addr(const void *x);
1150 #else
is_vmalloc_addr(const void * x)1151 static inline bool is_vmalloc_addr(const void *x)
1152 {
1153 	return false;
1154 }
is_vmalloc_or_module_addr(const void * x)1155 static inline int is_vmalloc_or_module_addr(const void *x)
1156 {
1157 	return 0;
1158 }
1159 #endif
1160 
1161 /*
1162  * How many times the entire folio is mapped as a single unit (eg by a
1163  * PMD or PUD entry).  This is probably not what you want, except for
1164  * debugging purposes - it does not include PTE-mapped sub-pages; look
1165  * at folio_mapcount() or page_mapcount() or total_mapcount() instead.
1166  */
folio_entire_mapcount(struct folio * folio)1167 static inline int folio_entire_mapcount(struct folio *folio)
1168 {
1169 	VM_BUG_ON_FOLIO(!folio_test_large(folio), folio);
1170 	return atomic_read(&folio->_entire_mapcount) + 1;
1171 }
1172 
1173 /*
1174  * The atomic page->_mapcount, starts from -1: so that transitions
1175  * both from it and to it can be tracked, using atomic_inc_and_test
1176  * and atomic_add_negative(-1).
1177  */
page_mapcount_reset(struct page * page)1178 static inline void page_mapcount_reset(struct page *page)
1179 {
1180 	atomic_set(&(page)->_mapcount, -1);
1181 }
1182 
1183 /**
1184  * page_mapcount() - Number of times this precise page is mapped.
1185  * @page: The page.
1186  *
1187  * The number of times this page is mapped.  If this page is part of
1188  * a large folio, it includes the number of times this page is mapped
1189  * as part of that folio.
1190  *
1191  * Will report 0 for pages which cannot be mapped into userspace, eg
1192  * slab, page tables and similar.
1193  */
page_mapcount(struct page * page)1194 static inline int page_mapcount(struct page *page)
1195 {
1196 	int mapcount = atomic_read(&page->_mapcount) + 1;
1197 
1198 	/* Handle page_has_type() pages */
1199 	if (mapcount < 0)
1200 		mapcount = 0;
1201 	if (unlikely(PageCompound(page)))
1202 		mapcount += folio_entire_mapcount(page_folio(page));
1203 
1204 	return mapcount;
1205 }
1206 
1207 int folio_total_mapcount(struct folio *folio);
1208 
1209 /**
1210  * folio_mapcount() - Calculate the number of mappings of this folio.
1211  * @folio: The folio.
1212  *
1213  * A large folio tracks both how many times the entire folio is mapped,
1214  * and how many times each individual page in the folio is mapped.
1215  * This function calculates the total number of times the folio is
1216  * mapped.
1217  *
1218  * Return: The number of times this folio is mapped.
1219  */
folio_mapcount(struct folio * folio)1220 static inline int folio_mapcount(struct folio *folio)
1221 {
1222 	if (likely(!folio_test_large(folio)))
1223 		return atomic_read(&folio->_mapcount) + 1;
1224 	return folio_total_mapcount(folio);
1225 }
1226 
total_mapcount(struct page * page)1227 static inline int total_mapcount(struct page *page)
1228 {
1229 	if (likely(!PageCompound(page)))
1230 		return atomic_read(&page->_mapcount) + 1;
1231 	return folio_total_mapcount(page_folio(page));
1232 }
1233 
folio_large_is_mapped(struct folio * folio)1234 static inline bool folio_large_is_mapped(struct folio *folio)
1235 {
1236 	/*
1237 	 * Reading _entire_mapcount below could be omitted if hugetlb
1238 	 * participated in incrementing nr_pages_mapped when compound mapped.
1239 	 */
1240 	return atomic_read(&folio->_nr_pages_mapped) > 0 ||
1241 		atomic_read(&folio->_entire_mapcount) >= 0;
1242 }
1243 
1244 /**
1245  * folio_mapped - Is this folio mapped into userspace?
1246  * @folio: The folio.
1247  *
1248  * Return: True if any page in this folio is referenced by user page tables.
1249  */
folio_mapped(struct folio * folio)1250 static inline bool folio_mapped(struct folio *folio)
1251 {
1252 	if (likely(!folio_test_large(folio)))
1253 		return atomic_read(&folio->_mapcount) >= 0;
1254 	return folio_large_is_mapped(folio);
1255 }
1256 
1257 /*
1258  * Return true if this page is mapped into pagetables.
1259  * For compound page it returns true if any sub-page of compound page is mapped,
1260  * even if this particular sub-page is not itself mapped by any PTE or PMD.
1261  */
page_mapped(struct page * page)1262 static inline bool page_mapped(struct page *page)
1263 {
1264 	if (likely(!PageCompound(page)))
1265 		return atomic_read(&page->_mapcount) >= 0;
1266 	return folio_large_is_mapped(page_folio(page));
1267 }
1268 
virt_to_head_page(const void * x)1269 static inline struct page *virt_to_head_page(const void *x)
1270 {
1271 	struct page *page = virt_to_page(x);
1272 
1273 	return compound_head(page);
1274 }
1275 
virt_to_folio(const void * x)1276 static inline struct folio *virt_to_folio(const void *x)
1277 {
1278 	struct page *page = virt_to_page(x);
1279 
1280 	return page_folio(page);
1281 }
1282 
1283 void __folio_put(struct folio *folio);
1284 
1285 void put_pages_list(struct list_head *pages);
1286 
1287 void split_page(struct page *page, unsigned int order);
1288 void folio_copy(struct folio *dst, struct folio *src);
1289 
1290 unsigned long nr_free_buffer_pages(void);
1291 
1292 void destroy_large_folio(struct folio *folio);
1293 
1294 /* Returns the number of bytes in this potentially compound page. */
page_size(struct page * page)1295 static inline unsigned long page_size(struct page *page)
1296 {
1297 	return PAGE_SIZE << compound_order(page);
1298 }
1299 
1300 /* Returns the number of bits needed for the number of bytes in a page */
page_shift(struct page * page)1301 static inline unsigned int page_shift(struct page *page)
1302 {
1303 	return PAGE_SHIFT + compound_order(page);
1304 }
1305 
1306 /**
1307  * thp_order - Order of a transparent huge page.
1308  * @page: Head page of a transparent huge page.
1309  */
thp_order(struct page * page)1310 static inline unsigned int thp_order(struct page *page)
1311 {
1312 	VM_BUG_ON_PGFLAGS(PageTail(page), page);
1313 	return compound_order(page);
1314 }
1315 
1316 /**
1317  * thp_size - Size of a transparent huge page.
1318  * @page: Head page of a transparent huge page.
1319  *
1320  * Return: Number of bytes in this page.
1321  */
thp_size(struct page * page)1322 static inline unsigned long thp_size(struct page *page)
1323 {
1324 	return PAGE_SIZE << thp_order(page);
1325 }
1326 
1327 #ifdef CONFIG_MMU
1328 /*
1329  * Do pte_mkwrite, but only if the vma says VM_WRITE.  We do this when
1330  * servicing faults for write access.  In the normal case, do always want
1331  * pte_mkwrite.  But get_user_pages can cause write faults for mappings
1332  * that do not have writing enabled, when used by access_process_vm.
1333  */
maybe_mkwrite(pte_t pte,struct vm_area_struct * vma)1334 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
1335 {
1336 	if (likely(vma->vm_flags & VM_WRITE))
1337 		pte = pte_mkwrite(pte, vma);
1338 	return pte;
1339 }
1340 
1341 vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page);
1342 void set_pte_range(struct vm_fault *vmf, struct folio *folio,
1343 		struct page *page, unsigned int nr, unsigned long addr);
1344 
1345 vm_fault_t finish_fault(struct vm_fault *vmf);
1346 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
1347 #endif
1348 
1349 /*
1350  * Multiple processes may "see" the same page. E.g. for untouched
1351  * mappings of /dev/null, all processes see the same page full of
1352  * zeroes, and text pages of executables and shared libraries have
1353  * only one copy in memory, at most, normally.
1354  *
1355  * For the non-reserved pages, page_count(page) denotes a reference count.
1356  *   page_count() == 0 means the page is free. page->lru is then used for
1357  *   freelist management in the buddy allocator.
1358  *   page_count() > 0  means the page has been allocated.
1359  *
1360  * Pages are allocated by the slab allocator in order to provide memory
1361  * to kmalloc and kmem_cache_alloc. In this case, the management of the
1362  * page, and the fields in 'struct page' are the responsibility of mm/slab.c
1363  * unless a particular usage is carefully commented. (the responsibility of
1364  * freeing the kmalloc memory is the caller's, of course).
1365  *
1366  * A page may be used by anyone else who does a __get_free_page().
1367  * In this case, page_count still tracks the references, and should only
1368  * be used through the normal accessor functions. The top bits of page->flags
1369  * and page->virtual store page management information, but all other fields
1370  * are unused and could be used privately, carefully. The management of this
1371  * page is the responsibility of the one who allocated it, and those who have
1372  * subsequently been given references to it.
1373  *
1374  * The other pages (we may call them "pagecache pages") are completely
1375  * managed by the Linux memory manager: I/O, buffers, swapping etc.
1376  * The following discussion applies only to them.
1377  *
1378  * A pagecache page contains an opaque `private' member, which belongs to the
1379  * page's address_space. Usually, this is the address of a circular list of
1380  * the page's disk buffers. PG_private must be set to tell the VM to call
1381  * into the filesystem to release these pages.
1382  *
1383  * A page may belong to an inode's memory mapping. In this case, page->mapping
1384  * is the pointer to the inode, and page->index is the file offset of the page,
1385  * in units of PAGE_SIZE.
1386  *
1387  * If pagecache pages are not associated with an inode, they are said to be
1388  * anonymous pages. These may become associated with the swapcache, and in that
1389  * case PG_swapcache is set, and page->private is an offset into the swapcache.
1390  *
1391  * In either case (swapcache or inode backed), the pagecache itself holds one
1392  * reference to the page. Setting PG_private should also increment the
1393  * refcount. The each user mapping also has a reference to the page.
1394  *
1395  * The pagecache pages are stored in a per-mapping radix tree, which is
1396  * rooted at mapping->i_pages, and indexed by offset.
1397  * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
1398  * lists, we instead now tag pages as dirty/writeback in the radix tree.
1399  *
1400  * All pagecache pages may be subject to I/O:
1401  * - inode pages may need to be read from disk,
1402  * - inode pages which have been modified and are MAP_SHARED may need
1403  *   to be written back to the inode on disk,
1404  * - anonymous pages (including MAP_PRIVATE file mappings) which have been
1405  *   modified may need to be swapped out to swap space and (later) to be read
1406  *   back into memory.
1407  */
1408 
1409 #if defined(CONFIG_ZONE_DEVICE) && defined(CONFIG_FS_DAX)
1410 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
1411 
1412 bool __put_devmap_managed_page_refs(struct page *page, int refs);
put_devmap_managed_page_refs(struct page * page,int refs)1413 static inline bool put_devmap_managed_page_refs(struct page *page, int refs)
1414 {
1415 	if (!static_branch_unlikely(&devmap_managed_key))
1416 		return false;
1417 	if (!is_zone_device_page(page))
1418 		return false;
1419 	return __put_devmap_managed_page_refs(page, refs);
1420 }
1421 #else /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
put_devmap_managed_page_refs(struct page * page,int refs)1422 static inline bool put_devmap_managed_page_refs(struct page *page, int refs)
1423 {
1424 	return false;
1425 }
1426 #endif /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1427 
put_devmap_managed_page(struct page * page)1428 static inline bool put_devmap_managed_page(struct page *page)
1429 {
1430 	return put_devmap_managed_page_refs(page, 1);
1431 }
1432 
1433 /* 127: arbitrary random number, small enough to assemble well */
1434 #define folio_ref_zero_or_close_to_overflow(folio) \
1435 	((unsigned int) folio_ref_count(folio) + 127u <= 127u)
1436 
1437 /**
1438  * folio_get - Increment the reference count on a folio.
1439  * @folio: The folio.
1440  *
1441  * Context: May be called in any context, as long as you know that
1442  * you have a refcount on the folio.  If you do not already have one,
1443  * folio_try_get() may be the right interface for you to use.
1444  */
folio_get(struct folio * folio)1445 static inline void folio_get(struct folio *folio)
1446 {
1447 	VM_BUG_ON_FOLIO(folio_ref_zero_or_close_to_overflow(folio), folio);
1448 	folio_ref_inc(folio);
1449 }
1450 
get_page(struct page * page)1451 static inline void get_page(struct page *page)
1452 {
1453 	folio_get(page_folio(page));
1454 }
1455 
try_get_page(struct page * page)1456 static inline __must_check bool try_get_page(struct page *page)
1457 {
1458 	page = compound_head(page);
1459 	if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1460 		return false;
1461 	page_ref_inc(page);
1462 	return true;
1463 }
1464 
1465 /**
1466  * folio_put - Decrement the reference count on a folio.
1467  * @folio: The folio.
1468  *
1469  * If the folio's reference count reaches zero, the memory will be
1470  * released back to the page allocator and may be used by another
1471  * allocation immediately.  Do not access the memory or the struct folio
1472  * after calling folio_put() unless you can be sure that it wasn't the
1473  * last reference.
1474  *
1475  * Context: May be called in process or interrupt context, but not in NMI
1476  * context.  May be called while holding a spinlock.
1477  */
folio_put(struct folio * folio)1478 static inline void folio_put(struct folio *folio)
1479 {
1480 	if (folio_put_testzero(folio))
1481 		__folio_put(folio);
1482 }
1483 
1484 /**
1485  * folio_put_refs - Reduce the reference count on a folio.
1486  * @folio: The folio.
1487  * @refs: The amount to subtract from the folio's reference count.
1488  *
1489  * If the folio's reference count reaches zero, the memory will be
1490  * released back to the page allocator and may be used by another
1491  * allocation immediately.  Do not access the memory or the struct folio
1492  * after calling folio_put_refs() unless you can be sure that these weren't
1493  * the last references.
1494  *
1495  * Context: May be called in process or interrupt context, but not in NMI
1496  * context.  May be called while holding a spinlock.
1497  */
folio_put_refs(struct folio * folio,int refs)1498 static inline void folio_put_refs(struct folio *folio, int refs)
1499 {
1500 	if (folio_ref_sub_and_test(folio, refs))
1501 		__folio_put(folio);
1502 }
1503 
1504 /*
1505  * union release_pages_arg - an array of pages or folios
1506  *
1507  * release_pages() releases a simple array of multiple pages, and
1508  * accepts various different forms of said page array: either
1509  * a regular old boring array of pages, an array of folios, or
1510  * an array of encoded page pointers.
1511  *
1512  * The transparent union syntax for this kind of "any of these
1513  * argument types" is all kinds of ugly, so look away.
1514  */
1515 typedef union {
1516 	struct page **pages;
1517 	struct folio **folios;
1518 	struct encoded_page **encoded_pages;
1519 } release_pages_arg __attribute__ ((__transparent_union__));
1520 
1521 void release_pages(release_pages_arg, int nr);
1522 
1523 /**
1524  * folios_put - Decrement the reference count on an array of folios.
1525  * @folios: The folios.
1526  * @nr: How many folios there are.
1527  *
1528  * Like folio_put(), but for an array of folios.  This is more efficient
1529  * than writing the loop yourself as it will optimise the locks which
1530  * need to be taken if the folios are freed.
1531  *
1532  * Context: May be called in process or interrupt context, but not in NMI
1533  * context.  May be called while holding a spinlock.
1534  */
folios_put(struct folio ** folios,unsigned int nr)1535 static inline void folios_put(struct folio **folios, unsigned int nr)
1536 {
1537 	release_pages(folios, nr);
1538 }
1539 
put_page(struct page * page)1540 static inline void put_page(struct page *page)
1541 {
1542 	struct folio *folio = page_folio(page);
1543 
1544 	/*
1545 	 * For some devmap managed pages we need to catch refcount transition
1546 	 * from 2 to 1:
1547 	 */
1548 	if (put_devmap_managed_page(&folio->page))
1549 		return;
1550 	folio_put(folio);
1551 }
1552 
1553 /*
1554  * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
1555  * the page's refcount so that two separate items are tracked: the original page
1556  * reference count, and also a new count of how many pin_user_pages() calls were
1557  * made against the page. ("gup-pinned" is another term for the latter).
1558  *
1559  * With this scheme, pin_user_pages() becomes special: such pages are marked as
1560  * distinct from normal pages. As such, the unpin_user_page() call (and its
1561  * variants) must be used in order to release gup-pinned pages.
1562  *
1563  * Choice of value:
1564  *
1565  * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
1566  * counts with respect to pin_user_pages() and unpin_user_page() becomes
1567  * simpler, due to the fact that adding an even power of two to the page
1568  * refcount has the effect of using only the upper N bits, for the code that
1569  * counts up using the bias value. This means that the lower bits are left for
1570  * the exclusive use of the original code that increments and decrements by one
1571  * (or at least, by much smaller values than the bias value).
1572  *
1573  * Of course, once the lower bits overflow into the upper bits (and this is
1574  * OK, because subtraction recovers the original values), then visual inspection
1575  * no longer suffices to directly view the separate counts. However, for normal
1576  * applications that don't have huge page reference counts, this won't be an
1577  * issue.
1578  *
1579  * Locking: the lockless algorithm described in folio_try_get_rcu()
1580  * provides safe operation for get_user_pages(), page_mkclean() and
1581  * other calls that race to set up page table entries.
1582  */
1583 #define GUP_PIN_COUNTING_BIAS (1U << 10)
1584 
1585 void unpin_user_page(struct page *page);
1586 void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1587 				 bool make_dirty);
1588 void unpin_user_page_range_dirty_lock(struct page *page, unsigned long npages,
1589 				      bool make_dirty);
1590 void unpin_user_pages(struct page **pages, unsigned long npages);
1591 
is_cow_mapping(vm_flags_t flags)1592 static inline bool is_cow_mapping(vm_flags_t flags)
1593 {
1594 	return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
1595 }
1596 
1597 #ifndef CONFIG_MMU
is_nommu_shared_mapping(vm_flags_t flags)1598 static inline bool is_nommu_shared_mapping(vm_flags_t flags)
1599 {
1600 	/*
1601 	 * NOMMU shared mappings are ordinary MAP_SHARED mappings and selected
1602 	 * R/O MAP_PRIVATE file mappings that are an effective R/O overlay of
1603 	 * a file mapping. R/O MAP_PRIVATE mappings might still modify
1604 	 * underlying memory if ptrace is active, so this is only possible if
1605 	 * ptrace does not apply. Note that there is no mprotect() to upgrade
1606 	 * write permissions later.
1607 	 */
1608 	return flags & (VM_MAYSHARE | VM_MAYOVERLAY);
1609 }
1610 #endif
1611 
1612 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1613 #define SECTION_IN_PAGE_FLAGS
1614 #endif
1615 
1616 /*
1617  * The identification function is mainly used by the buddy allocator for
1618  * determining if two pages could be buddies. We are not really identifying
1619  * the zone since we could be using the section number id if we do not have
1620  * node id available in page flags.
1621  * We only guarantee that it will return the same value for two combinable
1622  * pages in a zone.
1623  */
page_zone_id(struct page * page)1624 static inline int page_zone_id(struct page *page)
1625 {
1626 	return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1627 }
1628 
1629 #ifdef NODE_NOT_IN_PAGE_FLAGS
1630 extern int page_to_nid(const struct page *page);
1631 #else
page_to_nid(const struct page * page)1632 static inline int page_to_nid(const struct page *page)
1633 {
1634 	struct page *p = (struct page *)page;
1635 
1636 	return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1637 }
1638 #endif
1639 
folio_nid(const struct folio * folio)1640 static inline int folio_nid(const struct folio *folio)
1641 {
1642 	return page_to_nid(&folio->page);
1643 }
1644 
1645 #ifdef CONFIG_NUMA_BALANCING
1646 /* page access time bits needs to hold at least 4 seconds */
1647 #define PAGE_ACCESS_TIME_MIN_BITS	12
1648 #if LAST_CPUPID_SHIFT < PAGE_ACCESS_TIME_MIN_BITS
1649 #define PAGE_ACCESS_TIME_BUCKETS				\
1650 	(PAGE_ACCESS_TIME_MIN_BITS - LAST_CPUPID_SHIFT)
1651 #else
1652 #define PAGE_ACCESS_TIME_BUCKETS	0
1653 #endif
1654 
1655 #define PAGE_ACCESS_TIME_MASK				\
1656 	(LAST_CPUPID_MASK << PAGE_ACCESS_TIME_BUCKETS)
1657 
cpu_pid_to_cpupid(int cpu,int pid)1658 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1659 {
1660 	return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1661 }
1662 
cpupid_to_pid(int cpupid)1663 static inline int cpupid_to_pid(int cpupid)
1664 {
1665 	return cpupid & LAST__PID_MASK;
1666 }
1667 
cpupid_to_cpu(int cpupid)1668 static inline int cpupid_to_cpu(int cpupid)
1669 {
1670 	return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1671 }
1672 
cpupid_to_nid(int cpupid)1673 static inline int cpupid_to_nid(int cpupid)
1674 {
1675 	return cpu_to_node(cpupid_to_cpu(cpupid));
1676 }
1677 
cpupid_pid_unset(int cpupid)1678 static inline bool cpupid_pid_unset(int cpupid)
1679 {
1680 	return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1681 }
1682 
cpupid_cpu_unset(int cpupid)1683 static inline bool cpupid_cpu_unset(int cpupid)
1684 {
1685 	return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1686 }
1687 
__cpupid_match_pid(pid_t task_pid,int cpupid)1688 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1689 {
1690 	return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1691 }
1692 
1693 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1694 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
page_cpupid_xchg_last(struct page * page,int cpupid)1695 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1696 {
1697 	return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1698 }
1699 
page_cpupid_last(struct page * page)1700 static inline int page_cpupid_last(struct page *page)
1701 {
1702 	return page->_last_cpupid;
1703 }
page_cpupid_reset_last(struct page * page)1704 static inline void page_cpupid_reset_last(struct page *page)
1705 {
1706 	page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1707 }
1708 #else
page_cpupid_last(struct page * page)1709 static inline int page_cpupid_last(struct page *page)
1710 {
1711 	return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1712 }
1713 
1714 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1715 
page_cpupid_reset_last(struct page * page)1716 static inline void page_cpupid_reset_last(struct page *page)
1717 {
1718 	page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1719 }
1720 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1721 
xchg_page_access_time(struct page * page,int time)1722 static inline int xchg_page_access_time(struct page *page, int time)
1723 {
1724 	int last_time;
1725 
1726 	last_time = page_cpupid_xchg_last(page, time >> PAGE_ACCESS_TIME_BUCKETS);
1727 	return last_time << PAGE_ACCESS_TIME_BUCKETS;
1728 }
1729 
vma_set_access_pid_bit(struct vm_area_struct * vma)1730 static inline void vma_set_access_pid_bit(struct vm_area_struct *vma)
1731 {
1732 	unsigned int pid_bit;
1733 
1734 	pid_bit = hash_32(current->pid, ilog2(BITS_PER_LONG));
1735 	if (vma->numab_state && !test_bit(pid_bit, &vma->numab_state->pids_active[1])) {
1736 		__set_bit(pid_bit, &vma->numab_state->pids_active[1]);
1737 	}
1738 }
1739 #else /* !CONFIG_NUMA_BALANCING */
page_cpupid_xchg_last(struct page * page,int cpupid)1740 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1741 {
1742 	return page_to_nid(page); /* XXX */
1743 }
1744 
xchg_page_access_time(struct page * page,int time)1745 static inline int xchg_page_access_time(struct page *page, int time)
1746 {
1747 	return 0;
1748 }
1749 
page_cpupid_last(struct page * page)1750 static inline int page_cpupid_last(struct page *page)
1751 {
1752 	return page_to_nid(page); /* XXX */
1753 }
1754 
cpupid_to_nid(int cpupid)1755 static inline int cpupid_to_nid(int cpupid)
1756 {
1757 	return -1;
1758 }
1759 
cpupid_to_pid(int cpupid)1760 static inline int cpupid_to_pid(int cpupid)
1761 {
1762 	return -1;
1763 }
1764 
cpupid_to_cpu(int cpupid)1765 static inline int cpupid_to_cpu(int cpupid)
1766 {
1767 	return -1;
1768 }
1769 
cpu_pid_to_cpupid(int nid,int pid)1770 static inline int cpu_pid_to_cpupid(int nid, int pid)
1771 {
1772 	return -1;
1773 }
1774 
cpupid_pid_unset(int cpupid)1775 static inline bool cpupid_pid_unset(int cpupid)
1776 {
1777 	return true;
1778 }
1779 
page_cpupid_reset_last(struct page * page)1780 static inline void page_cpupid_reset_last(struct page *page)
1781 {
1782 }
1783 
cpupid_match_pid(struct task_struct * task,int cpupid)1784 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1785 {
1786 	return false;
1787 }
1788 
vma_set_access_pid_bit(struct vm_area_struct * vma)1789 static inline void vma_set_access_pid_bit(struct vm_area_struct *vma)
1790 {
1791 }
1792 #endif /* CONFIG_NUMA_BALANCING */
1793 
1794 #if defined(CONFIG_KASAN_SW_TAGS) || defined(CONFIG_KASAN_HW_TAGS)
1795 
1796 /*
1797  * KASAN per-page tags are stored xor'ed with 0xff. This allows to avoid
1798  * setting tags for all pages to native kernel tag value 0xff, as the default
1799  * value 0x00 maps to 0xff.
1800  */
1801 
page_kasan_tag(const struct page * page)1802 static inline u8 page_kasan_tag(const struct page *page)
1803 {
1804 	u8 tag = 0xff;
1805 
1806 	if (kasan_enabled()) {
1807 		tag = (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1808 		tag ^= 0xff;
1809 	}
1810 
1811 	return tag;
1812 }
1813 
page_kasan_tag_set(struct page * page,u8 tag)1814 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1815 {
1816 	unsigned long old_flags, flags;
1817 
1818 	if (!kasan_enabled())
1819 		return;
1820 
1821 	tag ^= 0xff;
1822 	old_flags = READ_ONCE(page->flags);
1823 	do {
1824 		flags = old_flags;
1825 		flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1826 		flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1827 	} while (unlikely(!try_cmpxchg(&page->flags, &old_flags, flags)));
1828 }
1829 
page_kasan_tag_reset(struct page * page)1830 static inline void page_kasan_tag_reset(struct page *page)
1831 {
1832 	if (kasan_enabled())
1833 		page_kasan_tag_set(page, 0xff);
1834 }
1835 
1836 #else /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1837 
page_kasan_tag(const struct page * page)1838 static inline u8 page_kasan_tag(const struct page *page)
1839 {
1840 	return 0xff;
1841 }
1842 
page_kasan_tag_set(struct page * page,u8 tag)1843 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
page_kasan_tag_reset(struct page * page)1844 static inline void page_kasan_tag_reset(struct page *page) { }
1845 
1846 #endif /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1847 
page_zone(const struct page * page)1848 static inline struct zone *page_zone(const struct page *page)
1849 {
1850 	return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1851 }
1852 
page_pgdat(const struct page * page)1853 static inline pg_data_t *page_pgdat(const struct page *page)
1854 {
1855 	return NODE_DATA(page_to_nid(page));
1856 }
1857 
folio_zone(const struct folio * folio)1858 static inline struct zone *folio_zone(const struct folio *folio)
1859 {
1860 	return page_zone(&folio->page);
1861 }
1862 
folio_pgdat(const struct folio * folio)1863 static inline pg_data_t *folio_pgdat(const struct folio *folio)
1864 {
1865 	return page_pgdat(&folio->page);
1866 }
1867 
1868 #ifdef SECTION_IN_PAGE_FLAGS
set_page_section(struct page * page,unsigned long section)1869 static inline void set_page_section(struct page *page, unsigned long section)
1870 {
1871 	page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1872 	page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1873 }
1874 
page_to_section(const struct page * page)1875 static inline unsigned long page_to_section(const struct page *page)
1876 {
1877 	return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1878 }
1879 #endif
1880 
1881 /**
1882  * folio_pfn - Return the Page Frame Number of a folio.
1883  * @folio: The folio.
1884  *
1885  * A folio may contain multiple pages.  The pages have consecutive
1886  * Page Frame Numbers.
1887  *
1888  * Return: The Page Frame Number of the first page in the folio.
1889  */
folio_pfn(struct folio * folio)1890 static inline unsigned long folio_pfn(struct folio *folio)
1891 {
1892 	return page_to_pfn(&folio->page);
1893 }
1894 
pfn_folio(unsigned long pfn)1895 static inline struct folio *pfn_folio(unsigned long pfn)
1896 {
1897 	return page_folio(pfn_to_page(pfn));
1898 }
1899 
1900 /**
1901  * folio_maybe_dma_pinned - Report if a folio may be pinned for DMA.
1902  * @folio: The folio.
1903  *
1904  * This function checks if a folio has been pinned via a call to
1905  * a function in the pin_user_pages() family.
1906  *
1907  * For small folios, the return value is partially fuzzy: false is not fuzzy,
1908  * because it means "definitely not pinned for DMA", but true means "probably
1909  * pinned for DMA, but possibly a false positive due to having at least
1910  * GUP_PIN_COUNTING_BIAS worth of normal folio references".
1911  *
1912  * False positives are OK, because: a) it's unlikely for a folio to
1913  * get that many refcounts, and b) all the callers of this routine are
1914  * expected to be able to deal gracefully with a false positive.
1915  *
1916  * For large folios, the result will be exactly correct. That's because
1917  * we have more tracking data available: the _pincount field is used
1918  * instead of the GUP_PIN_COUNTING_BIAS scheme.
1919  *
1920  * For more information, please see Documentation/core-api/pin_user_pages.rst.
1921  *
1922  * Return: True, if it is likely that the page has been "dma-pinned".
1923  * False, if the page is definitely not dma-pinned.
1924  */
folio_maybe_dma_pinned(struct folio * folio)1925 static inline bool folio_maybe_dma_pinned(struct folio *folio)
1926 {
1927 	if (folio_test_large(folio))
1928 		return atomic_read(&folio->_pincount) > 0;
1929 
1930 	/*
1931 	 * folio_ref_count() is signed. If that refcount overflows, then
1932 	 * folio_ref_count() returns a negative value, and callers will avoid
1933 	 * further incrementing the refcount.
1934 	 *
1935 	 * Here, for that overflow case, use the sign bit to count a little
1936 	 * bit higher via unsigned math, and thus still get an accurate result.
1937 	 */
1938 	return ((unsigned int)folio_ref_count(folio)) >=
1939 		GUP_PIN_COUNTING_BIAS;
1940 }
1941 
page_maybe_dma_pinned(struct page * page)1942 static inline bool page_maybe_dma_pinned(struct page *page)
1943 {
1944 	return folio_maybe_dma_pinned(page_folio(page));
1945 }
1946 
1947 /*
1948  * This should most likely only be called during fork() to see whether we
1949  * should break the cow immediately for an anon page on the src mm.
1950  *
1951  * The caller has to hold the PT lock and the vma->vm_mm->->write_protect_seq.
1952  */
page_needs_cow_for_dma(struct vm_area_struct * vma,struct page * page)1953 static inline bool page_needs_cow_for_dma(struct vm_area_struct *vma,
1954 					  struct page *page)
1955 {
1956 	VM_BUG_ON(!(raw_read_seqcount(&vma->vm_mm->write_protect_seq) & 1));
1957 
1958 	if (!test_bit(MMF_HAS_PINNED, &vma->vm_mm->flags))
1959 		return false;
1960 
1961 	return page_maybe_dma_pinned(page);
1962 }
1963 
1964 /**
1965  * is_zero_page - Query if a page is a zero page
1966  * @page: The page to query
1967  *
1968  * This returns true if @page is one of the permanent zero pages.
1969  */
is_zero_page(const struct page * page)1970 static inline bool is_zero_page(const struct page *page)
1971 {
1972 	return is_zero_pfn(page_to_pfn(page));
1973 }
1974 
1975 /**
1976  * is_zero_folio - Query if a folio is a zero page
1977  * @folio: The folio to query
1978  *
1979  * This returns true if @folio is one of the permanent zero pages.
1980  */
is_zero_folio(const struct folio * folio)1981 static inline bool is_zero_folio(const struct folio *folio)
1982 {
1983 	return is_zero_page(&folio->page);
1984 }
1985 
1986 /* MIGRATE_CMA and ZONE_MOVABLE do not allow pin folios */
1987 #ifdef CONFIG_MIGRATION
folio_is_longterm_pinnable(struct folio * folio)1988 static inline bool folio_is_longterm_pinnable(struct folio *folio)
1989 {
1990 #ifdef CONFIG_CMA
1991 	int mt = folio_migratetype(folio);
1992 
1993 	if (mt == MIGRATE_CMA || mt == MIGRATE_ISOLATE)
1994 		return false;
1995 #endif
1996 	/* The zero page can be "pinned" but gets special handling. */
1997 	if (is_zero_folio(folio))
1998 		return true;
1999 
2000 	/* Coherent device memory must always allow eviction. */
2001 	if (folio_is_device_coherent(folio))
2002 		return false;
2003 
2004 	/* Otherwise, non-movable zone folios can be pinned. */
2005 	return !folio_is_zone_movable(folio);
2006 
2007 }
2008 #else
folio_is_longterm_pinnable(struct folio * folio)2009 static inline bool folio_is_longterm_pinnable(struct folio *folio)
2010 {
2011 	return true;
2012 }
2013 #endif
2014 
set_page_zone(struct page * page,enum zone_type zone)2015 static inline void set_page_zone(struct page *page, enum zone_type zone)
2016 {
2017 	page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
2018 	page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
2019 }
2020 
set_page_node(struct page * page,unsigned long node)2021 static inline void set_page_node(struct page *page, unsigned long node)
2022 {
2023 	page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
2024 	page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
2025 }
2026 
set_page_links(struct page * page,enum zone_type zone,unsigned long node,unsigned long pfn)2027 static inline void set_page_links(struct page *page, enum zone_type zone,
2028 	unsigned long node, unsigned long pfn)
2029 {
2030 	set_page_zone(page, zone);
2031 	set_page_node(page, node);
2032 #ifdef SECTION_IN_PAGE_FLAGS
2033 	set_page_section(page, pfn_to_section_nr(pfn));
2034 #endif
2035 }
2036 
2037 /**
2038  * folio_nr_pages - The number of pages in the folio.
2039  * @folio: The folio.
2040  *
2041  * Return: A positive power of two.
2042  */
folio_nr_pages(struct folio * folio)2043 static inline long folio_nr_pages(struct folio *folio)
2044 {
2045 	if (!folio_test_large(folio))
2046 		return 1;
2047 #ifdef CONFIG_64BIT
2048 	return folio->_folio_nr_pages;
2049 #else
2050 	return 1L << (folio->_flags_1 & 0xff);
2051 #endif
2052 }
2053 
2054 /*
2055  * compound_nr() returns the number of pages in this potentially compound
2056  * page.  compound_nr() can be called on a tail page, and is defined to
2057  * return 1 in that case.
2058  */
compound_nr(struct page * page)2059 static inline unsigned long compound_nr(struct page *page)
2060 {
2061 	struct folio *folio = (struct folio *)page;
2062 
2063 	if (!test_bit(PG_head, &folio->flags))
2064 		return 1;
2065 #ifdef CONFIG_64BIT
2066 	return folio->_folio_nr_pages;
2067 #else
2068 	return 1L << (folio->_flags_1 & 0xff);
2069 #endif
2070 }
2071 
2072 /**
2073  * thp_nr_pages - The number of regular pages in this huge page.
2074  * @page: The head page of a huge page.
2075  */
thp_nr_pages(struct page * page)2076 static inline int thp_nr_pages(struct page *page)
2077 {
2078 	return folio_nr_pages((struct folio *)page);
2079 }
2080 
2081 /**
2082  * folio_next - Move to the next physical folio.
2083  * @folio: The folio we're currently operating on.
2084  *
2085  * If you have physically contiguous memory which may span more than
2086  * one folio (eg a &struct bio_vec), use this function to move from one
2087  * folio to the next.  Do not use it if the memory is only virtually
2088  * contiguous as the folios are almost certainly not adjacent to each
2089  * other.  This is the folio equivalent to writing ``page++``.
2090  *
2091  * Context: We assume that the folios are refcounted and/or locked at a
2092  * higher level and do not adjust the reference counts.
2093  * Return: The next struct folio.
2094  */
folio_next(struct folio * folio)2095 static inline struct folio *folio_next(struct folio *folio)
2096 {
2097 	return (struct folio *)folio_page(folio, folio_nr_pages(folio));
2098 }
2099 
2100 /**
2101  * folio_shift - The size of the memory described by this folio.
2102  * @folio: The folio.
2103  *
2104  * A folio represents a number of bytes which is a power-of-two in size.
2105  * This function tells you which power-of-two the folio is.  See also
2106  * folio_size() and folio_order().
2107  *
2108  * Context: The caller should have a reference on the folio to prevent
2109  * it from being split.  It is not necessary for the folio to be locked.
2110  * Return: The base-2 logarithm of the size of this folio.
2111  */
folio_shift(struct folio * folio)2112 static inline unsigned int folio_shift(struct folio *folio)
2113 {
2114 	return PAGE_SHIFT + folio_order(folio);
2115 }
2116 
2117 /**
2118  * folio_size - The number of bytes in a folio.
2119  * @folio: The folio.
2120  *
2121  * Context: The caller should have a reference on the folio to prevent
2122  * it from being split.  It is not necessary for the folio to be locked.
2123  * Return: The number of bytes in this folio.
2124  */
folio_size(struct folio * folio)2125 static inline size_t folio_size(struct folio *folio)
2126 {
2127 	return PAGE_SIZE << folio_order(folio);
2128 }
2129 
2130 /**
2131  * folio_estimated_sharers - Estimate the number of sharers of a folio.
2132  * @folio: The folio.
2133  *
2134  * folio_estimated_sharers() aims to serve as a function to efficiently
2135  * estimate the number of processes sharing a folio. This is done by
2136  * looking at the precise mapcount of the first subpage in the folio, and
2137  * assuming the other subpages are the same. This may not be true for large
2138  * folios. If you want exact mapcounts for exact calculations, look at
2139  * page_mapcount() or folio_total_mapcount().
2140  *
2141  * Return: The estimated number of processes sharing a folio.
2142  */
folio_estimated_sharers(struct folio * folio)2143 static inline int folio_estimated_sharers(struct folio *folio)
2144 {
2145 	return page_mapcount(folio_page(folio, 0));
2146 }
2147 
2148 #ifndef HAVE_ARCH_MAKE_PAGE_ACCESSIBLE
arch_make_page_accessible(struct page * page)2149 static inline int arch_make_page_accessible(struct page *page)
2150 {
2151 	return 0;
2152 }
2153 #endif
2154 
2155 #ifndef HAVE_ARCH_MAKE_FOLIO_ACCESSIBLE
arch_make_folio_accessible(struct folio * folio)2156 static inline int arch_make_folio_accessible(struct folio *folio)
2157 {
2158 	int ret;
2159 	long i, nr = folio_nr_pages(folio);
2160 
2161 	for (i = 0; i < nr; i++) {
2162 		ret = arch_make_page_accessible(folio_page(folio, i));
2163 		if (ret)
2164 			break;
2165 	}
2166 
2167 	return ret;
2168 }
2169 #endif
2170 
2171 /*
2172  * Some inline functions in vmstat.h depend on page_zone()
2173  */
2174 #include <linux/vmstat.h>
2175 
lowmem_page_address(const struct page * page)2176 static __always_inline void *lowmem_page_address(const struct page *page)
2177 {
2178 	return page_to_virt(page);
2179 }
2180 
2181 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
2182 #define HASHED_PAGE_VIRTUAL
2183 #endif
2184 
2185 #if defined(WANT_PAGE_VIRTUAL)
page_address(const struct page * page)2186 static inline void *page_address(const struct page *page)
2187 {
2188 	return page->virtual;
2189 }
set_page_address(struct page * page,void * address)2190 static inline void set_page_address(struct page *page, void *address)
2191 {
2192 	page->virtual = address;
2193 }
2194 #define page_address_init()  do { } while(0)
2195 #endif
2196 
2197 #if defined(HASHED_PAGE_VIRTUAL)
2198 void *page_address(const struct page *page);
2199 void set_page_address(struct page *page, void *virtual);
2200 void page_address_init(void);
2201 #endif
2202 
2203 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
2204 #define page_address(page) lowmem_page_address(page)
2205 #define set_page_address(page, address)  do { } while(0)
2206 #define page_address_init()  do { } while(0)
2207 #endif
2208 
folio_address(const struct folio * folio)2209 static inline void *folio_address(const struct folio *folio)
2210 {
2211 	return page_address(&folio->page);
2212 }
2213 
2214 extern pgoff_t __page_file_index(struct page *page);
2215 
2216 /*
2217  * Return the pagecache index of the passed page.  Regular pagecache pages
2218  * use ->index whereas swapcache pages use swp_offset(->private)
2219  */
page_index(struct page * page)2220 static inline pgoff_t page_index(struct page *page)
2221 {
2222 	if (unlikely(PageSwapCache(page)))
2223 		return __page_file_index(page);
2224 	return page->index;
2225 }
2226 
2227 /*
2228  * Return true only if the page has been allocated with
2229  * ALLOC_NO_WATERMARKS and the low watermark was not
2230  * met implying that the system is under some pressure.
2231  */
page_is_pfmemalloc(const struct page * page)2232 static inline bool page_is_pfmemalloc(const struct page *page)
2233 {
2234 	/*
2235 	 * lru.next has bit 1 set if the page is allocated from the
2236 	 * pfmemalloc reserves.  Callers may simply overwrite it if
2237 	 * they do not need to preserve that information.
2238 	 */
2239 	return (uintptr_t)page->lru.next & BIT(1);
2240 }
2241 
2242 /*
2243  * Return true only if the folio has been allocated with
2244  * ALLOC_NO_WATERMARKS and the low watermark was not
2245  * met implying that the system is under some pressure.
2246  */
folio_is_pfmemalloc(const struct folio * folio)2247 static inline bool folio_is_pfmemalloc(const struct folio *folio)
2248 {
2249 	/*
2250 	 * lru.next has bit 1 set if the page is allocated from the
2251 	 * pfmemalloc reserves.  Callers may simply overwrite it if
2252 	 * they do not need to preserve that information.
2253 	 */
2254 	return (uintptr_t)folio->lru.next & BIT(1);
2255 }
2256 
2257 /*
2258  * Only to be called by the page allocator on a freshly allocated
2259  * page.
2260  */
set_page_pfmemalloc(struct page * page)2261 static inline void set_page_pfmemalloc(struct page *page)
2262 {
2263 	page->lru.next = (void *)BIT(1);
2264 }
2265 
clear_page_pfmemalloc(struct page * page)2266 static inline void clear_page_pfmemalloc(struct page *page)
2267 {
2268 	page->lru.next = NULL;
2269 }
2270 
2271 /*
2272  * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
2273  */
2274 extern void pagefault_out_of_memory(void);
2275 
2276 #define offset_in_page(p)	((unsigned long)(p) & ~PAGE_MASK)
2277 #define offset_in_thp(page, p)	((unsigned long)(p) & (thp_size(page) - 1))
2278 #define offset_in_folio(folio, p) ((unsigned long)(p) & (folio_size(folio) - 1))
2279 
2280 /*
2281  * Parameter block passed down to zap_pte_range in exceptional cases.
2282  */
2283 struct zap_details {
2284 	struct folio *single_folio;	/* Locked folio to be unmapped */
2285 	bool even_cows;			/* Zap COWed private pages too? */
2286 	zap_flags_t zap_flags;		/* Extra flags for zapping */
2287 };
2288 
2289 /*
2290  * Whether to drop the pte markers, for example, the uffd-wp information for
2291  * file-backed memory.  This should only be specified when we will completely
2292  * drop the page in the mm, either by truncation or unmapping of the vma.  By
2293  * default, the flag is not set.
2294  */
2295 #define  ZAP_FLAG_DROP_MARKER        ((__force zap_flags_t) BIT(0))
2296 /* Set in unmap_vmas() to indicate a final unmap call.  Only used by hugetlb */
2297 #define  ZAP_FLAG_UNMAP              ((__force zap_flags_t) BIT(1))
2298 
2299 #ifdef CONFIG_SCHED_MM_CID
2300 void sched_mm_cid_before_execve(struct task_struct *t);
2301 void sched_mm_cid_after_execve(struct task_struct *t);
2302 void sched_mm_cid_fork(struct task_struct *t);
2303 void sched_mm_cid_exit_signals(struct task_struct *t);
task_mm_cid(struct task_struct * t)2304 static inline int task_mm_cid(struct task_struct *t)
2305 {
2306 	return t->mm_cid;
2307 }
2308 #else
sched_mm_cid_before_execve(struct task_struct * t)2309 static inline void sched_mm_cid_before_execve(struct task_struct *t) { }
sched_mm_cid_after_execve(struct task_struct * t)2310 static inline void sched_mm_cid_after_execve(struct task_struct *t) { }
sched_mm_cid_fork(struct task_struct * t)2311 static inline void sched_mm_cid_fork(struct task_struct *t) { }
sched_mm_cid_exit_signals(struct task_struct * t)2312 static inline void sched_mm_cid_exit_signals(struct task_struct *t) { }
task_mm_cid(struct task_struct * t)2313 static inline int task_mm_cid(struct task_struct *t)
2314 {
2315 	/*
2316 	 * Use the processor id as a fall-back when the mm cid feature is
2317 	 * disabled. This provides functional per-cpu data structure accesses
2318 	 * in user-space, althrough it won't provide the memory usage benefits.
2319 	 */
2320 	return raw_smp_processor_id();
2321 }
2322 #endif
2323 
2324 #ifdef CONFIG_MMU
2325 extern bool can_do_mlock(void);
2326 #else
can_do_mlock(void)2327 static inline bool can_do_mlock(void) { return false; }
2328 #endif
2329 extern int user_shm_lock(size_t, struct ucounts *);
2330 extern void user_shm_unlock(size_t, struct ucounts *);
2331 
2332 struct folio *vm_normal_folio(struct vm_area_struct *vma, unsigned long addr,
2333 			     pte_t pte);
2334 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
2335 			     pte_t pte);
2336 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
2337 				pmd_t pmd);
2338 
2339 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
2340 		  unsigned long size);
2341 void zap_page_range_single(struct vm_area_struct *vma, unsigned long address,
2342 			   unsigned long size, struct zap_details *details);
zap_vma_pages(struct vm_area_struct * vma)2343 static inline void zap_vma_pages(struct vm_area_struct *vma)
2344 {
2345 	zap_page_range_single(vma, vma->vm_start,
2346 			      vma->vm_end - vma->vm_start, NULL);
2347 }
2348 void unmap_vmas(struct mmu_gather *tlb, struct ma_state *mas,
2349 		struct vm_area_struct *start_vma, unsigned long start,
2350 		unsigned long end, unsigned long tree_end, bool mm_wr_locked);
2351 
2352 struct mmu_notifier_range;
2353 
2354 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
2355 		unsigned long end, unsigned long floor, unsigned long ceiling);
2356 int
2357 copy_page_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma);
2358 int follow_pte(struct mm_struct *mm, unsigned long address,
2359 	       pte_t **ptepp, spinlock_t **ptlp);
2360 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
2361 	unsigned long *pfn);
2362 int follow_phys(struct vm_area_struct *vma, unsigned long address,
2363 		unsigned int flags, unsigned long *prot, resource_size_t *phys);
2364 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
2365 			void *buf, int len, int write);
2366 
2367 extern void truncate_pagecache(struct inode *inode, loff_t new);
2368 extern void truncate_setsize(struct inode *inode, loff_t newsize);
2369 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
2370 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
2371 int generic_error_remove_page(struct address_space *mapping, struct page *page);
2372 
2373 struct vm_area_struct *lock_mm_and_find_vma(struct mm_struct *mm,
2374 		unsigned long address, struct pt_regs *regs);
2375 
2376 #ifdef CONFIG_MMU
2377 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
2378 				  unsigned long address, unsigned int flags,
2379 				  struct pt_regs *regs);
2380 extern int fixup_user_fault(struct mm_struct *mm,
2381 			    unsigned long address, unsigned int fault_flags,
2382 			    bool *unlocked);
2383 void unmap_mapping_pages(struct address_space *mapping,
2384 		pgoff_t start, pgoff_t nr, bool even_cows);
2385 void unmap_mapping_range(struct address_space *mapping,
2386 		loff_t const holebegin, loff_t const holelen, int even_cows);
2387 #else
handle_mm_fault(struct vm_area_struct * vma,unsigned long address,unsigned int flags,struct pt_regs * regs)2388 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
2389 					 unsigned long address, unsigned int flags,
2390 					 struct pt_regs *regs)
2391 {
2392 	/* should never happen if there's no MMU */
2393 	BUG();
2394 	return VM_FAULT_SIGBUS;
2395 }
fixup_user_fault(struct mm_struct * mm,unsigned long address,unsigned int fault_flags,bool * unlocked)2396 static inline int fixup_user_fault(struct mm_struct *mm, unsigned long address,
2397 		unsigned int fault_flags, bool *unlocked)
2398 {
2399 	/* should never happen if there's no MMU */
2400 	BUG();
2401 	return -EFAULT;
2402 }
unmap_mapping_pages(struct address_space * mapping,pgoff_t start,pgoff_t nr,bool even_cows)2403 static inline void unmap_mapping_pages(struct address_space *mapping,
2404 		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)2405 static inline void unmap_mapping_range(struct address_space *mapping,
2406 		loff_t const holebegin, loff_t const holelen, int even_cows) { }
2407 #endif
2408 
unmap_shared_mapping_range(struct address_space * mapping,loff_t const holebegin,loff_t const holelen)2409 static inline void unmap_shared_mapping_range(struct address_space *mapping,
2410 		loff_t const holebegin, loff_t const holelen)
2411 {
2412 	unmap_mapping_range(mapping, holebegin, holelen, 0);
2413 }
2414 
2415 static inline struct vm_area_struct *vma_lookup(struct mm_struct *mm,
2416 						unsigned long addr);
2417 
2418 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
2419 		void *buf, int len, unsigned int gup_flags);
2420 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
2421 		void *buf, int len, unsigned int gup_flags);
2422 extern int __access_remote_vm(struct mm_struct *mm, unsigned long addr,
2423 			      void *buf, int len, unsigned int gup_flags);
2424 
2425 long get_user_pages_remote(struct mm_struct *mm,
2426 			   unsigned long start, unsigned long nr_pages,
2427 			   unsigned int gup_flags, struct page **pages,
2428 			   int *locked);
2429 long pin_user_pages_remote(struct mm_struct *mm,
2430 			   unsigned long start, unsigned long nr_pages,
2431 			   unsigned int gup_flags, struct page **pages,
2432 			   int *locked);
2433 
get_user_page_vma_remote(struct mm_struct * mm,unsigned long addr,int gup_flags,struct vm_area_struct ** vmap)2434 static inline struct page *get_user_page_vma_remote(struct mm_struct *mm,
2435 						    unsigned long addr,
2436 						    int gup_flags,
2437 						    struct vm_area_struct **vmap)
2438 {
2439 	struct page *page;
2440 	struct vm_area_struct *vma;
2441 	int got = get_user_pages_remote(mm, addr, 1, gup_flags, &page, NULL);
2442 
2443 	if (got < 0)
2444 		return ERR_PTR(got);
2445 	if (got == 0)
2446 		return NULL;
2447 
2448 	vma = vma_lookup(mm, addr);
2449 	if (WARN_ON_ONCE(!vma)) {
2450 		put_page(page);
2451 		return ERR_PTR(-EINVAL);
2452 	}
2453 
2454 	*vmap = vma;
2455 	return page;
2456 }
2457 
2458 long get_user_pages(unsigned long start, unsigned long nr_pages,
2459 		    unsigned int gup_flags, struct page **pages);
2460 long pin_user_pages(unsigned long start, unsigned long nr_pages,
2461 		    unsigned int gup_flags, struct page **pages);
2462 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
2463 		    struct page **pages, unsigned int gup_flags);
2464 long pin_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
2465 		    struct page **pages, unsigned int gup_flags);
2466 
2467 int get_user_pages_fast(unsigned long start, int nr_pages,
2468 			unsigned int gup_flags, struct page **pages);
2469 int pin_user_pages_fast(unsigned long start, int nr_pages,
2470 			unsigned int gup_flags, struct page **pages);
2471 void folio_add_pin(struct folio *folio);
2472 
2473 int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
2474 int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
2475 			struct task_struct *task, bool bypass_rlim);
2476 
2477 struct kvec;
2478 struct page *get_dump_page(unsigned long addr);
2479 
2480 bool folio_mark_dirty(struct folio *folio);
2481 bool set_page_dirty(struct page *page);
2482 int set_page_dirty_lock(struct page *page);
2483 
2484 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
2485 
2486 extern unsigned long move_page_tables(struct vm_area_struct *vma,
2487 		unsigned long old_addr, struct vm_area_struct *new_vma,
2488 		unsigned long new_addr, unsigned long len,
2489 		bool need_rmap_locks);
2490 
2491 /*
2492  * Flags used by change_protection().  For now we make it a bitmap so
2493  * that we can pass in multiple flags just like parameters.  However
2494  * for now all the callers are only use one of the flags at the same
2495  * time.
2496  */
2497 /*
2498  * Whether we should manually check if we can map individual PTEs writable,
2499  * because something (e.g., COW, uffd-wp) blocks that from happening for all
2500  * PTEs automatically in a writable mapping.
2501  */
2502 #define  MM_CP_TRY_CHANGE_WRITABLE	   (1UL << 0)
2503 /* Whether this protection change is for NUMA hints */
2504 #define  MM_CP_PROT_NUMA                   (1UL << 1)
2505 /* Whether this change is for write protecting */
2506 #define  MM_CP_UFFD_WP                     (1UL << 2) /* do wp */
2507 #define  MM_CP_UFFD_WP_RESOLVE             (1UL << 3) /* Resolve wp */
2508 #define  MM_CP_UFFD_WP_ALL                 (MM_CP_UFFD_WP | \
2509 					    MM_CP_UFFD_WP_RESOLVE)
2510 
2511 bool vma_needs_dirty_tracking(struct vm_area_struct *vma);
2512 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
vma_wants_manual_pte_write_upgrade(struct vm_area_struct * vma)2513 static inline bool vma_wants_manual_pte_write_upgrade(struct vm_area_struct *vma)
2514 {
2515 	/*
2516 	 * We want to check manually if we can change individual PTEs writable
2517 	 * if we can't do that automatically for all PTEs in a mapping. For
2518 	 * private mappings, that's always the case when we have write
2519 	 * permissions as we properly have to handle COW.
2520 	 */
2521 	if (vma->vm_flags & VM_SHARED)
2522 		return vma_wants_writenotify(vma, vma->vm_page_prot);
2523 	return !!(vma->vm_flags & VM_WRITE);
2524 
2525 }
2526 bool can_change_pte_writable(struct vm_area_struct *vma, unsigned long addr,
2527 			     pte_t pte);
2528 extern long change_protection(struct mmu_gather *tlb,
2529 			      struct vm_area_struct *vma, unsigned long start,
2530 			      unsigned long end, unsigned long cp_flags);
2531 extern int mprotect_fixup(struct vma_iterator *vmi, struct mmu_gather *tlb,
2532 	  struct vm_area_struct *vma, struct vm_area_struct **pprev,
2533 	  unsigned long start, unsigned long end, unsigned long newflags);
2534 
2535 /*
2536  * doesn't attempt to fault and will return short.
2537  */
2538 int get_user_pages_fast_only(unsigned long start, int nr_pages,
2539 			     unsigned int gup_flags, struct page **pages);
2540 
get_user_page_fast_only(unsigned long addr,unsigned int gup_flags,struct page ** pagep)2541 static inline bool get_user_page_fast_only(unsigned long addr,
2542 			unsigned int gup_flags, struct page **pagep)
2543 {
2544 	return get_user_pages_fast_only(addr, 1, gup_flags, pagep) == 1;
2545 }
2546 /*
2547  * per-process(per-mm_struct) statistics.
2548  */
get_mm_counter(struct mm_struct * mm,int member)2549 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
2550 {
2551 	return percpu_counter_read_positive(&mm->rss_stat[member]);
2552 }
2553 
2554 void mm_trace_rss_stat(struct mm_struct *mm, int member);
2555 
add_mm_counter(struct mm_struct * mm,int member,long value)2556 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
2557 {
2558 	percpu_counter_add(&mm->rss_stat[member], value);
2559 
2560 	mm_trace_rss_stat(mm, member);
2561 }
2562 
inc_mm_counter(struct mm_struct * mm,int member)2563 static inline void inc_mm_counter(struct mm_struct *mm, int member)
2564 {
2565 	percpu_counter_inc(&mm->rss_stat[member]);
2566 
2567 	mm_trace_rss_stat(mm, member);
2568 }
2569 
dec_mm_counter(struct mm_struct * mm,int member)2570 static inline void dec_mm_counter(struct mm_struct *mm, int member)
2571 {
2572 	percpu_counter_dec(&mm->rss_stat[member]);
2573 
2574 	mm_trace_rss_stat(mm, member);
2575 }
2576 
2577 /* Optimized variant when page is already known not to be PageAnon */
mm_counter_file(struct page * page)2578 static inline int mm_counter_file(struct page *page)
2579 {
2580 	if (PageSwapBacked(page))
2581 		return MM_SHMEMPAGES;
2582 	return MM_FILEPAGES;
2583 }
2584 
mm_counter(struct page * page)2585 static inline int mm_counter(struct page *page)
2586 {
2587 	if (PageAnon(page))
2588 		return MM_ANONPAGES;
2589 	return mm_counter_file(page);
2590 }
2591 
get_mm_rss(struct mm_struct * mm)2592 static inline unsigned long get_mm_rss(struct mm_struct *mm)
2593 {
2594 	return get_mm_counter(mm, MM_FILEPAGES) +
2595 		get_mm_counter(mm, MM_ANONPAGES) +
2596 		get_mm_counter(mm, MM_SHMEMPAGES);
2597 }
2598 
get_mm_hiwater_rss(struct mm_struct * mm)2599 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
2600 {
2601 	return max(mm->hiwater_rss, get_mm_rss(mm));
2602 }
2603 
get_mm_hiwater_vm(struct mm_struct * mm)2604 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
2605 {
2606 	return max(mm->hiwater_vm, mm->total_vm);
2607 }
2608 
update_hiwater_rss(struct mm_struct * mm)2609 static inline void update_hiwater_rss(struct mm_struct *mm)
2610 {
2611 	unsigned long _rss = get_mm_rss(mm);
2612 
2613 	if ((mm)->hiwater_rss < _rss)
2614 		(mm)->hiwater_rss = _rss;
2615 }
2616 
update_hiwater_vm(struct mm_struct * mm)2617 static inline void update_hiwater_vm(struct mm_struct *mm)
2618 {
2619 	if (mm->hiwater_vm < mm->total_vm)
2620 		mm->hiwater_vm = mm->total_vm;
2621 }
2622 
reset_mm_hiwater_rss(struct mm_struct * mm)2623 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
2624 {
2625 	mm->hiwater_rss = get_mm_rss(mm);
2626 }
2627 
setmax_mm_hiwater_rss(unsigned long * maxrss,struct mm_struct * mm)2628 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
2629 					 struct mm_struct *mm)
2630 {
2631 	unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
2632 
2633 	if (*maxrss < hiwater_rss)
2634 		*maxrss = hiwater_rss;
2635 }
2636 
2637 #if defined(SPLIT_RSS_COUNTING)
2638 void sync_mm_rss(struct mm_struct *mm);
2639 #else
sync_mm_rss(struct mm_struct * mm)2640 static inline void sync_mm_rss(struct mm_struct *mm)
2641 {
2642 }
2643 #endif
2644 
2645 #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
pte_special(pte_t pte)2646 static inline int pte_special(pte_t pte)
2647 {
2648 	return 0;
2649 }
2650 
pte_mkspecial(pte_t pte)2651 static inline pte_t pte_mkspecial(pte_t pte)
2652 {
2653 	return pte;
2654 }
2655 #endif
2656 
2657 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
pte_devmap(pte_t pte)2658 static inline int pte_devmap(pte_t pte)
2659 {
2660 	return 0;
2661 }
2662 #endif
2663 
2664 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
2665 			       spinlock_t **ptl);
get_locked_pte(struct mm_struct * mm,unsigned long addr,spinlock_t ** ptl)2666 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
2667 				    spinlock_t **ptl)
2668 {
2669 	pte_t *ptep;
2670 	__cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
2671 	return ptep;
2672 }
2673 
2674 #ifdef __PAGETABLE_P4D_FOLDED
__p4d_alloc(struct mm_struct * mm,pgd_t * pgd,unsigned long address)2675 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2676 						unsigned long address)
2677 {
2678 	return 0;
2679 }
2680 #else
2681 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
2682 #endif
2683 
2684 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
__pud_alloc(struct mm_struct * mm,p4d_t * p4d,unsigned long address)2685 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2686 						unsigned long address)
2687 {
2688 	return 0;
2689 }
mm_inc_nr_puds(struct mm_struct * mm)2690 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
mm_dec_nr_puds(struct mm_struct * mm)2691 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
2692 
2693 #else
2694 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
2695 
mm_inc_nr_puds(struct mm_struct * mm)2696 static inline void mm_inc_nr_puds(struct mm_struct *mm)
2697 {
2698 	if (mm_pud_folded(mm))
2699 		return;
2700 	atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2701 }
2702 
mm_dec_nr_puds(struct mm_struct * mm)2703 static inline void mm_dec_nr_puds(struct mm_struct *mm)
2704 {
2705 	if (mm_pud_folded(mm))
2706 		return;
2707 	atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2708 }
2709 #endif
2710 
2711 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
__pmd_alloc(struct mm_struct * mm,pud_t * pud,unsigned long address)2712 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
2713 						unsigned long address)
2714 {
2715 	return 0;
2716 }
2717 
mm_inc_nr_pmds(struct mm_struct * mm)2718 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
mm_dec_nr_pmds(struct mm_struct * mm)2719 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
2720 
2721 #else
2722 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
2723 
mm_inc_nr_pmds(struct mm_struct * mm)2724 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
2725 {
2726 	if (mm_pmd_folded(mm))
2727 		return;
2728 	atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2729 }
2730 
mm_dec_nr_pmds(struct mm_struct * mm)2731 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
2732 {
2733 	if (mm_pmd_folded(mm))
2734 		return;
2735 	atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2736 }
2737 #endif
2738 
2739 #ifdef CONFIG_MMU
mm_pgtables_bytes_init(struct mm_struct * mm)2740 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
2741 {
2742 	atomic_long_set(&mm->pgtables_bytes, 0);
2743 }
2744 
mm_pgtables_bytes(const struct mm_struct * mm)2745 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2746 {
2747 	return atomic_long_read(&mm->pgtables_bytes);
2748 }
2749 
mm_inc_nr_ptes(struct mm_struct * mm)2750 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
2751 {
2752 	atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2753 }
2754 
mm_dec_nr_ptes(struct mm_struct * mm)2755 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
2756 {
2757 	atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2758 }
2759 #else
2760 
mm_pgtables_bytes_init(struct mm_struct * mm)2761 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
mm_pgtables_bytes(const struct mm_struct * mm)2762 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2763 {
2764 	return 0;
2765 }
2766 
mm_inc_nr_ptes(struct mm_struct * mm)2767 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
mm_dec_nr_ptes(struct mm_struct * mm)2768 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
2769 #endif
2770 
2771 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
2772 int __pte_alloc_kernel(pmd_t *pmd);
2773 
2774 #if defined(CONFIG_MMU)
2775 
p4d_alloc(struct mm_struct * mm,pgd_t * pgd,unsigned long address)2776 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2777 		unsigned long address)
2778 {
2779 	return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
2780 		NULL : p4d_offset(pgd, address);
2781 }
2782 
pud_alloc(struct mm_struct * mm,p4d_t * p4d,unsigned long address)2783 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2784 		unsigned long address)
2785 {
2786 	return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
2787 		NULL : pud_offset(p4d, address);
2788 }
2789 
pmd_alloc(struct mm_struct * mm,pud_t * pud,unsigned long address)2790 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
2791 {
2792 	return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
2793 		NULL: pmd_offset(pud, address);
2794 }
2795 #endif /* CONFIG_MMU */
2796 
virt_to_ptdesc(const void * x)2797 static inline struct ptdesc *virt_to_ptdesc(const void *x)
2798 {
2799 	return page_ptdesc(virt_to_page(x));
2800 }
2801 
ptdesc_to_virt(const struct ptdesc * pt)2802 static inline void *ptdesc_to_virt(const struct ptdesc *pt)
2803 {
2804 	return page_to_virt(ptdesc_page(pt));
2805 }
2806 
ptdesc_address(const struct ptdesc * pt)2807 static inline void *ptdesc_address(const struct ptdesc *pt)
2808 {
2809 	return folio_address(ptdesc_folio(pt));
2810 }
2811 
pagetable_is_reserved(struct ptdesc * pt)2812 static inline bool pagetable_is_reserved(struct ptdesc *pt)
2813 {
2814 	return folio_test_reserved(ptdesc_folio(pt));
2815 }
2816 
2817 /**
2818  * pagetable_alloc - Allocate pagetables
2819  * @gfp:    GFP flags
2820  * @order:  desired pagetable order
2821  *
2822  * pagetable_alloc allocates memory for page tables as well as a page table
2823  * descriptor to describe that memory.
2824  *
2825  * Return: The ptdesc describing the allocated page tables.
2826  */
pagetable_alloc(gfp_t gfp,unsigned int order)2827 static inline struct ptdesc *pagetable_alloc(gfp_t gfp, unsigned int order)
2828 {
2829 	struct page *page = alloc_pages(gfp | __GFP_COMP, order);
2830 
2831 	return page_ptdesc(page);
2832 }
2833 
2834 /**
2835  * pagetable_free - Free pagetables
2836  * @pt:	The page table descriptor
2837  *
2838  * pagetable_free frees the memory of all page tables described by a page
2839  * table descriptor and the memory for the descriptor itself.
2840  */
pagetable_free(struct ptdesc * pt)2841 static inline void pagetable_free(struct ptdesc *pt)
2842 {
2843 	struct page *page = ptdesc_page(pt);
2844 
2845 	__free_pages(page, compound_order(page));
2846 }
2847 
2848 #if USE_SPLIT_PTE_PTLOCKS
2849 #if ALLOC_SPLIT_PTLOCKS
2850 void __init ptlock_cache_init(void);
2851 bool ptlock_alloc(struct ptdesc *ptdesc);
2852 void ptlock_free(struct ptdesc *ptdesc);
2853 
ptlock_ptr(struct ptdesc * ptdesc)2854 static inline spinlock_t *ptlock_ptr(struct ptdesc *ptdesc)
2855 {
2856 	return ptdesc->ptl;
2857 }
2858 #else /* ALLOC_SPLIT_PTLOCKS */
ptlock_cache_init(void)2859 static inline void ptlock_cache_init(void)
2860 {
2861 }
2862 
ptlock_alloc(struct ptdesc * ptdesc)2863 static inline bool ptlock_alloc(struct ptdesc *ptdesc)
2864 {
2865 	return true;
2866 }
2867 
ptlock_free(struct ptdesc * ptdesc)2868 static inline void ptlock_free(struct ptdesc *ptdesc)
2869 {
2870 }
2871 
ptlock_ptr(struct ptdesc * ptdesc)2872 static inline spinlock_t *ptlock_ptr(struct ptdesc *ptdesc)
2873 {
2874 	return &ptdesc->ptl;
2875 }
2876 #endif /* ALLOC_SPLIT_PTLOCKS */
2877 
pte_lockptr(struct mm_struct * mm,pmd_t * pmd)2878 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2879 {
2880 	return ptlock_ptr(page_ptdesc(pmd_page(*pmd)));
2881 }
2882 
ptlock_init(struct ptdesc * ptdesc)2883 static inline bool ptlock_init(struct ptdesc *ptdesc)
2884 {
2885 	/*
2886 	 * prep_new_page() initialize page->private (and therefore page->ptl)
2887 	 * with 0. Make sure nobody took it in use in between.
2888 	 *
2889 	 * It can happen if arch try to use slab for page table allocation:
2890 	 * slab code uses page->slab_cache, which share storage with page->ptl.
2891 	 */
2892 	VM_BUG_ON_PAGE(*(unsigned long *)&ptdesc->ptl, ptdesc_page(ptdesc));
2893 	if (!ptlock_alloc(ptdesc))
2894 		return false;
2895 	spin_lock_init(ptlock_ptr(ptdesc));
2896 	return true;
2897 }
2898 
2899 #else	/* !USE_SPLIT_PTE_PTLOCKS */
2900 /*
2901  * We use mm->page_table_lock to guard all pagetable pages of the mm.
2902  */
pte_lockptr(struct mm_struct * mm,pmd_t * pmd)2903 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2904 {
2905 	return &mm->page_table_lock;
2906 }
ptlock_cache_init(void)2907 static inline void ptlock_cache_init(void) {}
ptlock_init(struct ptdesc * ptdesc)2908 static inline bool ptlock_init(struct ptdesc *ptdesc) { return true; }
ptlock_free(struct ptdesc * ptdesc)2909 static inline void ptlock_free(struct ptdesc *ptdesc) {}
2910 #endif /* USE_SPLIT_PTE_PTLOCKS */
2911 
pagetable_pte_ctor(struct ptdesc * ptdesc)2912 static inline bool pagetable_pte_ctor(struct ptdesc *ptdesc)
2913 {
2914 	struct folio *folio = ptdesc_folio(ptdesc);
2915 
2916 	if (!ptlock_init(ptdesc))
2917 		return false;
2918 	__folio_set_pgtable(folio);
2919 	lruvec_stat_add_folio(folio, NR_PAGETABLE);
2920 	return true;
2921 }
2922 
pagetable_pte_dtor(struct ptdesc * ptdesc)2923 static inline void pagetable_pte_dtor(struct ptdesc *ptdesc)
2924 {
2925 	struct folio *folio = ptdesc_folio(ptdesc);
2926 
2927 	ptlock_free(ptdesc);
2928 	__folio_clear_pgtable(folio);
2929 	lruvec_stat_sub_folio(folio, NR_PAGETABLE);
2930 }
2931 
2932 pte_t *__pte_offset_map(pmd_t *pmd, unsigned long addr, pmd_t *pmdvalp);
pte_offset_map(pmd_t * pmd,unsigned long addr)2933 static inline pte_t *pte_offset_map(pmd_t *pmd, unsigned long addr)
2934 {
2935 	return __pte_offset_map(pmd, addr, NULL);
2936 }
2937 
2938 pte_t *__pte_offset_map_lock(struct mm_struct *mm, pmd_t *pmd,
2939 			unsigned long addr, spinlock_t **ptlp);
pte_offset_map_lock(struct mm_struct * mm,pmd_t * pmd,unsigned long addr,spinlock_t ** ptlp)2940 static inline pte_t *pte_offset_map_lock(struct mm_struct *mm, pmd_t *pmd,
2941 			unsigned long addr, spinlock_t **ptlp)
2942 {
2943 	pte_t *pte;
2944 
2945 	__cond_lock(*ptlp, pte = __pte_offset_map_lock(mm, pmd, addr, ptlp));
2946 	return pte;
2947 }
2948 
2949 pte_t *pte_offset_map_nolock(struct mm_struct *mm, pmd_t *pmd,
2950 			unsigned long addr, spinlock_t **ptlp);
2951 
2952 #define pte_unmap_unlock(pte, ptl)	do {		\
2953 	spin_unlock(ptl);				\
2954 	pte_unmap(pte);					\
2955 } while (0)
2956 
2957 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2958 
2959 #define pte_alloc_map(mm, pmd, address)			\
2960 	(pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2961 
2962 #define pte_alloc_map_lock(mm, pmd, address, ptlp)	\
2963 	(pte_alloc(mm, pmd) ?			\
2964 		 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2965 
2966 #define pte_alloc_kernel(pmd, address)			\
2967 	((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2968 		NULL: pte_offset_kernel(pmd, address))
2969 
2970 #if USE_SPLIT_PMD_PTLOCKS
2971 
pmd_pgtable_page(pmd_t * pmd)2972 static inline struct page *pmd_pgtable_page(pmd_t *pmd)
2973 {
2974 	unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2975 	return virt_to_page((void *)((unsigned long) pmd & mask));
2976 }
2977 
pmd_ptdesc(pmd_t * pmd)2978 static inline struct ptdesc *pmd_ptdesc(pmd_t *pmd)
2979 {
2980 	return page_ptdesc(pmd_pgtable_page(pmd));
2981 }
2982 
pmd_lockptr(struct mm_struct * mm,pmd_t * pmd)2983 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2984 {
2985 	return ptlock_ptr(pmd_ptdesc(pmd));
2986 }
2987 
pmd_ptlock_init(struct ptdesc * ptdesc)2988 static inline bool pmd_ptlock_init(struct ptdesc *ptdesc)
2989 {
2990 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2991 	ptdesc->pmd_huge_pte = NULL;
2992 #endif
2993 	return ptlock_init(ptdesc);
2994 }
2995 
pmd_ptlock_free(struct ptdesc * ptdesc)2996 static inline void pmd_ptlock_free(struct ptdesc *ptdesc)
2997 {
2998 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2999 	VM_BUG_ON_PAGE(ptdesc->pmd_huge_pte, ptdesc_page(ptdesc));
3000 #endif
3001 	ptlock_free(ptdesc);
3002 }
3003 
3004 #define pmd_huge_pte(mm, pmd) (pmd_ptdesc(pmd)->pmd_huge_pte)
3005 
3006 #else
3007 
pmd_lockptr(struct mm_struct * mm,pmd_t * pmd)3008 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
3009 {
3010 	return &mm->page_table_lock;
3011 }
3012 
pmd_ptlock_init(struct ptdesc * ptdesc)3013 static inline bool pmd_ptlock_init(struct ptdesc *ptdesc) { return true; }
pmd_ptlock_free(struct ptdesc * ptdesc)3014 static inline void pmd_ptlock_free(struct ptdesc *ptdesc) {}
3015 
3016 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
3017 
3018 #endif
3019 
pmd_lock(struct mm_struct * mm,pmd_t * pmd)3020 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
3021 {
3022 	spinlock_t *ptl = pmd_lockptr(mm, pmd);
3023 	spin_lock(ptl);
3024 	return ptl;
3025 }
3026 
pagetable_pmd_ctor(struct ptdesc * ptdesc)3027 static inline bool pagetable_pmd_ctor(struct ptdesc *ptdesc)
3028 {
3029 	struct folio *folio = ptdesc_folio(ptdesc);
3030 
3031 	if (!pmd_ptlock_init(ptdesc))
3032 		return false;
3033 	__folio_set_pgtable(folio);
3034 	lruvec_stat_add_folio(folio, NR_PAGETABLE);
3035 	return true;
3036 }
3037 
pagetable_pmd_dtor(struct ptdesc * ptdesc)3038 static inline void pagetable_pmd_dtor(struct ptdesc *ptdesc)
3039 {
3040 	struct folio *folio = ptdesc_folio(ptdesc);
3041 
3042 	pmd_ptlock_free(ptdesc);
3043 	__folio_clear_pgtable(folio);
3044 	lruvec_stat_sub_folio(folio, NR_PAGETABLE);
3045 }
3046 
3047 /*
3048  * No scalability reason to split PUD locks yet, but follow the same pattern
3049  * as the PMD locks to make it easier if we decide to.  The VM should not be
3050  * considered ready to switch to split PUD locks yet; there may be places
3051  * which need to be converted from page_table_lock.
3052  */
pud_lockptr(struct mm_struct * mm,pud_t * pud)3053 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
3054 {
3055 	return &mm->page_table_lock;
3056 }
3057 
pud_lock(struct mm_struct * mm,pud_t * pud)3058 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
3059 {
3060 	spinlock_t *ptl = pud_lockptr(mm, pud);
3061 
3062 	spin_lock(ptl);
3063 	return ptl;
3064 }
3065 
3066 extern void __init pagecache_init(void);
3067 extern void free_initmem(void);
3068 
3069 /*
3070  * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
3071  * into the buddy system. The freed pages will be poisoned with pattern
3072  * "poison" if it's within range [0, UCHAR_MAX].
3073  * Return pages freed into the buddy system.
3074  */
3075 extern unsigned long free_reserved_area(void *start, void *end,
3076 					int poison, const char *s);
3077 
3078 extern void adjust_managed_page_count(struct page *page, long count);
3079 
3080 extern void reserve_bootmem_region(phys_addr_t start,
3081 				   phys_addr_t end, int nid);
3082 
3083 /* Free the reserved page into the buddy system, so it gets managed. */
free_reserved_page(struct page * page)3084 static inline void free_reserved_page(struct page *page)
3085 {
3086 	ClearPageReserved(page);
3087 	init_page_count(page);
3088 	__free_page(page);
3089 	adjust_managed_page_count(page, 1);
3090 }
3091 #define free_highmem_page(page) free_reserved_page(page)
3092 
mark_page_reserved(struct page * page)3093 static inline void mark_page_reserved(struct page *page)
3094 {
3095 	SetPageReserved(page);
3096 	adjust_managed_page_count(page, -1);
3097 }
3098 
free_reserved_ptdesc(struct ptdesc * pt)3099 static inline void free_reserved_ptdesc(struct ptdesc *pt)
3100 {
3101 	free_reserved_page(ptdesc_page(pt));
3102 }
3103 
3104 /*
3105  * Default method to free all the __init memory into the buddy system.
3106  * The freed pages will be poisoned with pattern "poison" if it's within
3107  * range [0, UCHAR_MAX].
3108  * Return pages freed into the buddy system.
3109  */
free_initmem_default(int poison)3110 static inline unsigned long free_initmem_default(int poison)
3111 {
3112 	extern char __init_begin[], __init_end[];
3113 
3114 	return free_reserved_area(&__init_begin, &__init_end,
3115 				  poison, "unused kernel image (initmem)");
3116 }
3117 
get_num_physpages(void)3118 static inline unsigned long get_num_physpages(void)
3119 {
3120 	int nid;
3121 	unsigned long phys_pages = 0;
3122 
3123 	for_each_online_node(nid)
3124 		phys_pages += node_present_pages(nid);
3125 
3126 	return phys_pages;
3127 }
3128 
3129 /*
3130  * Using memblock node mappings, an architecture may initialise its
3131  * zones, allocate the backing mem_map and account for memory holes in an
3132  * architecture independent manner.
3133  *
3134  * An architecture is expected to register range of page frames backed by
3135  * physical memory with memblock_add[_node]() before calling
3136  * free_area_init() passing in the PFN each zone ends at. At a basic
3137  * usage, an architecture is expected to do something like
3138  *
3139  * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
3140  * 							 max_highmem_pfn};
3141  * for_each_valid_physical_page_range()
3142  *	memblock_add_node(base, size, nid, MEMBLOCK_NONE)
3143  * free_area_init(max_zone_pfns);
3144  */
3145 void free_area_init(unsigned long *max_zone_pfn);
3146 unsigned long node_map_pfn_alignment(void);
3147 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
3148 						unsigned long end_pfn);
3149 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
3150 						unsigned long end_pfn);
3151 extern void get_pfn_range_for_nid(unsigned int nid,
3152 			unsigned long *start_pfn, unsigned long *end_pfn);
3153 
3154 #ifndef CONFIG_NUMA
early_pfn_to_nid(unsigned long pfn)3155 static inline int early_pfn_to_nid(unsigned long pfn)
3156 {
3157 	return 0;
3158 }
3159 #else
3160 /* please see mm/page_alloc.c */
3161 extern int __meminit early_pfn_to_nid(unsigned long pfn);
3162 #endif
3163 
3164 extern void set_dma_reserve(unsigned long new_dma_reserve);
3165 extern void mem_init(void);
3166 extern void __init mmap_init(void);
3167 
3168 extern void __show_mem(unsigned int flags, nodemask_t *nodemask, int max_zone_idx);
show_mem(void)3169 static inline void show_mem(void)
3170 {
3171 	__show_mem(0, NULL, MAX_NR_ZONES - 1);
3172 }
3173 extern long si_mem_available(void);
3174 extern void si_meminfo(struct sysinfo * val);
3175 extern void si_meminfo_node(struct sysinfo *val, int nid);
3176 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
3177 extern unsigned long arch_reserved_kernel_pages(void);
3178 #endif
3179 
3180 extern __printf(3, 4)
3181 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
3182 
3183 extern void setup_per_cpu_pageset(void);
3184 
3185 /* nommu.c */
3186 extern atomic_long_t mmap_pages_allocated;
3187 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
3188 
3189 /* interval_tree.c */
3190 void vma_interval_tree_insert(struct vm_area_struct *node,
3191 			      struct rb_root_cached *root);
3192 void vma_interval_tree_insert_after(struct vm_area_struct *node,
3193 				    struct vm_area_struct *prev,
3194 				    struct rb_root_cached *root);
3195 void vma_interval_tree_remove(struct vm_area_struct *node,
3196 			      struct rb_root_cached *root);
3197 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
3198 				unsigned long start, unsigned long last);
3199 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
3200 				unsigned long start, unsigned long last);
3201 
3202 #define vma_interval_tree_foreach(vma, root, start, last)		\
3203 	for (vma = vma_interval_tree_iter_first(root, start, last);	\
3204 	     vma; vma = vma_interval_tree_iter_next(vma, start, last))
3205 
3206 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
3207 				   struct rb_root_cached *root);
3208 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
3209 				   struct rb_root_cached *root);
3210 struct anon_vma_chain *
3211 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
3212 				  unsigned long start, unsigned long last);
3213 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
3214 	struct anon_vma_chain *node, unsigned long start, unsigned long last);
3215 #ifdef CONFIG_DEBUG_VM_RB
3216 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
3217 #endif
3218 
3219 #define anon_vma_interval_tree_foreach(avc, root, start, last)		 \
3220 	for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
3221 	     avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
3222 
3223 /* mmap.c */
3224 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
3225 extern int vma_expand(struct vma_iterator *vmi, struct vm_area_struct *vma,
3226 		      unsigned long start, unsigned long end, pgoff_t pgoff,
3227 		      struct vm_area_struct *next);
3228 extern int vma_shrink(struct vma_iterator *vmi, struct vm_area_struct *vma,
3229 		       unsigned long start, unsigned long end, pgoff_t pgoff);
3230 extern struct vm_area_struct *vma_merge(struct vma_iterator *vmi,
3231 	struct mm_struct *, struct vm_area_struct *prev, unsigned long addr,
3232 	unsigned long end, unsigned long vm_flags, struct anon_vma *,
3233 	struct file *, pgoff_t, struct mempolicy *, struct vm_userfaultfd_ctx,
3234 	struct anon_vma_name *);
3235 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
3236 extern int __split_vma(struct vma_iterator *vmi, struct vm_area_struct *,
3237 		       unsigned long addr, int new_below);
3238 extern int split_vma(struct vma_iterator *vmi, struct vm_area_struct *,
3239 			 unsigned long addr, int new_below);
3240 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
3241 extern void unlink_file_vma(struct vm_area_struct *);
3242 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
3243 	unsigned long addr, unsigned long len, pgoff_t pgoff,
3244 	bool *need_rmap_locks);
3245 extern void exit_mmap(struct mm_struct *);
3246 
check_data_rlimit(unsigned long rlim,unsigned long new,unsigned long start,unsigned long end_data,unsigned long start_data)3247 static inline int check_data_rlimit(unsigned long rlim,
3248 				    unsigned long new,
3249 				    unsigned long start,
3250 				    unsigned long end_data,
3251 				    unsigned long start_data)
3252 {
3253 	if (rlim < RLIM_INFINITY) {
3254 		if (((new - start) + (end_data - start_data)) > rlim)
3255 			return -ENOSPC;
3256 	}
3257 
3258 	return 0;
3259 }
3260 
3261 extern int mm_take_all_locks(struct mm_struct *mm);
3262 extern void mm_drop_all_locks(struct mm_struct *mm);
3263 
3264 extern int set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
3265 extern int replace_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
3266 extern struct file *get_mm_exe_file(struct mm_struct *mm);
3267 extern struct file *get_task_exe_file(struct task_struct *task);
3268 
3269 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
3270 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
3271 
3272 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
3273 				   const struct vm_special_mapping *sm);
3274 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
3275 				   unsigned long addr, unsigned long len,
3276 				   unsigned long flags,
3277 				   const struct vm_special_mapping *spec);
3278 /* This is an obsolete alternative to _install_special_mapping. */
3279 extern int install_special_mapping(struct mm_struct *mm,
3280 				   unsigned long addr, unsigned long len,
3281 				   unsigned long flags, struct page **pages);
3282 
3283 unsigned long randomize_stack_top(unsigned long stack_top);
3284 unsigned long randomize_page(unsigned long start, unsigned long range);
3285 
3286 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
3287 
3288 extern unsigned long mmap_region(struct file *file, unsigned long addr,
3289 	unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
3290 	struct list_head *uf);
3291 extern unsigned long do_mmap(struct file *file, unsigned long addr,
3292 	unsigned long len, unsigned long prot, unsigned long flags,
3293 	vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate,
3294 	struct list_head *uf);
3295 extern int do_vmi_munmap(struct vma_iterator *vmi, struct mm_struct *mm,
3296 			 unsigned long start, size_t len, struct list_head *uf,
3297 			 bool unlock);
3298 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
3299 		     struct list_head *uf);
3300 extern int do_madvise(struct mm_struct *mm, unsigned long start, size_t len_in, int behavior);
3301 
3302 #ifdef CONFIG_MMU
3303 extern int do_vma_munmap(struct vma_iterator *vmi, struct vm_area_struct *vma,
3304 			 unsigned long start, unsigned long end,
3305 			 struct list_head *uf, bool unlock);
3306 extern int __mm_populate(unsigned long addr, unsigned long len,
3307 			 int ignore_errors);
mm_populate(unsigned long addr,unsigned long len)3308 static inline void mm_populate(unsigned long addr, unsigned long len)
3309 {
3310 	/* Ignore errors */
3311 	(void) __mm_populate(addr, len, 1);
3312 }
3313 #else
mm_populate(unsigned long addr,unsigned long len)3314 static inline void mm_populate(unsigned long addr, unsigned long len) {}
3315 #endif
3316 
3317 /* These take the mm semaphore themselves */
3318 extern int __must_check vm_brk(unsigned long, unsigned long);
3319 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
3320 extern int vm_munmap(unsigned long, size_t);
3321 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
3322         unsigned long, unsigned long,
3323         unsigned long, unsigned long);
3324 
3325 struct vm_unmapped_area_info {
3326 #define VM_UNMAPPED_AREA_TOPDOWN 1
3327 	unsigned long flags;
3328 	unsigned long length;
3329 	unsigned long low_limit;
3330 	unsigned long high_limit;
3331 	unsigned long align_mask;
3332 	unsigned long align_offset;
3333 };
3334 
3335 extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info);
3336 
3337 /* truncate.c */
3338 extern void truncate_inode_pages(struct address_space *, loff_t);
3339 extern void truncate_inode_pages_range(struct address_space *,
3340 				       loff_t lstart, loff_t lend);
3341 extern void truncate_inode_pages_final(struct address_space *);
3342 
3343 /* generic vm_area_ops exported for stackable file systems */
3344 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
3345 extern vm_fault_t filemap_map_pages(struct vm_fault *vmf,
3346 		pgoff_t start_pgoff, pgoff_t end_pgoff);
3347 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
3348 
3349 extern unsigned long stack_guard_gap;
3350 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
3351 int expand_stack_locked(struct vm_area_struct *vma, unsigned long address);
3352 struct vm_area_struct *expand_stack(struct mm_struct * mm, unsigned long addr);
3353 
3354 /* CONFIG_STACK_GROWSUP still needs to grow downwards at some places */
3355 int expand_downwards(struct vm_area_struct *vma, unsigned long address);
3356 
3357 /* Look up the first VMA which satisfies  addr < vm_end,  NULL if none. */
3358 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
3359 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
3360 					     struct vm_area_struct **pprev);
3361 
3362 /*
3363  * Look up the first VMA which intersects the interval [start_addr, end_addr)
3364  * NULL if none.  Assume start_addr < end_addr.
3365  */
3366 struct vm_area_struct *find_vma_intersection(struct mm_struct *mm,
3367 			unsigned long start_addr, unsigned long end_addr);
3368 
3369 /**
3370  * vma_lookup() - Find a VMA at a specific address
3371  * @mm: The process address space.
3372  * @addr: The user address.
3373  *
3374  * Return: The vm_area_struct at the given address, %NULL otherwise.
3375  */
3376 static inline
vma_lookup(struct mm_struct * mm,unsigned long addr)3377 struct vm_area_struct *vma_lookup(struct mm_struct *mm, unsigned long addr)
3378 {
3379 	return mtree_load(&mm->mm_mt, addr);
3380 }
3381 
stack_guard_start_gap(struct vm_area_struct * vma)3382 static inline unsigned long stack_guard_start_gap(struct vm_area_struct *vma)
3383 {
3384 	if (vma->vm_flags & VM_GROWSDOWN)
3385 		return stack_guard_gap;
3386 
3387 	/* See reasoning around the VM_SHADOW_STACK definition */
3388 	if (vma->vm_flags & VM_SHADOW_STACK)
3389 		return PAGE_SIZE;
3390 
3391 	return 0;
3392 }
3393 
vm_start_gap(struct vm_area_struct * vma)3394 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
3395 {
3396 	unsigned long gap = stack_guard_start_gap(vma);
3397 	unsigned long vm_start = vma->vm_start;
3398 
3399 	vm_start -= gap;
3400 	if (vm_start > vma->vm_start)
3401 		vm_start = 0;
3402 	return vm_start;
3403 }
3404 
vm_end_gap(struct vm_area_struct * vma)3405 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
3406 {
3407 	unsigned long vm_end = vma->vm_end;
3408 
3409 	if (vma->vm_flags & VM_GROWSUP) {
3410 		vm_end += stack_guard_gap;
3411 		if (vm_end < vma->vm_end)
3412 			vm_end = -PAGE_SIZE;
3413 	}
3414 	return vm_end;
3415 }
3416 
vma_pages(struct vm_area_struct * vma)3417 static inline unsigned long vma_pages(struct vm_area_struct *vma)
3418 {
3419 	return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
3420 }
3421 
3422 /* 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)3423 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
3424 				unsigned long vm_start, unsigned long vm_end)
3425 {
3426 	struct vm_area_struct *vma = vma_lookup(mm, vm_start);
3427 
3428 	if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
3429 		vma = NULL;
3430 
3431 	return vma;
3432 }
3433 
range_in_vma(struct vm_area_struct * vma,unsigned long start,unsigned long end)3434 static inline bool range_in_vma(struct vm_area_struct *vma,
3435 				unsigned long start, unsigned long end)
3436 {
3437 	return (vma && vma->vm_start <= start && end <= vma->vm_end);
3438 }
3439 
3440 #ifdef CONFIG_MMU
3441 pgprot_t vm_get_page_prot(unsigned long vm_flags);
3442 void vma_set_page_prot(struct vm_area_struct *vma);
3443 #else
vm_get_page_prot(unsigned long vm_flags)3444 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
3445 {
3446 	return __pgprot(0);
3447 }
vma_set_page_prot(struct vm_area_struct * vma)3448 static inline void vma_set_page_prot(struct vm_area_struct *vma)
3449 {
3450 	vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
3451 }
3452 #endif
3453 
3454 void vma_set_file(struct vm_area_struct *vma, struct file *file);
3455 
3456 #ifdef CONFIG_NUMA_BALANCING
3457 unsigned long change_prot_numa(struct vm_area_struct *vma,
3458 			unsigned long start, unsigned long end);
3459 #endif
3460 
3461 struct vm_area_struct *find_extend_vma_locked(struct mm_struct *,
3462 		unsigned long addr);
3463 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
3464 			unsigned long pfn, unsigned long size, pgprot_t);
3465 int remap_pfn_range_notrack(struct vm_area_struct *vma, unsigned long addr,
3466 		unsigned long pfn, unsigned long size, pgprot_t prot);
3467 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
3468 int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
3469 			struct page **pages, unsigned long *num);
3470 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
3471 				unsigned long num);
3472 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
3473 				unsigned long num);
3474 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
3475 			unsigned long pfn);
3476 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
3477 			unsigned long pfn, pgprot_t pgprot);
3478 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
3479 			pfn_t pfn);
3480 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
3481 		unsigned long addr, pfn_t pfn);
3482 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
3483 
vmf_insert_page(struct vm_area_struct * vma,unsigned long addr,struct page * page)3484 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
3485 				unsigned long addr, struct page *page)
3486 {
3487 	int err = vm_insert_page(vma, addr, page);
3488 
3489 	if (err == -ENOMEM)
3490 		return VM_FAULT_OOM;
3491 	if (err < 0 && err != -EBUSY)
3492 		return VM_FAULT_SIGBUS;
3493 
3494 	return VM_FAULT_NOPAGE;
3495 }
3496 
3497 #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)3498 static inline int io_remap_pfn_range(struct vm_area_struct *vma,
3499 				     unsigned long addr, unsigned long pfn,
3500 				     unsigned long size, pgprot_t prot)
3501 {
3502 	return remap_pfn_range(vma, addr, pfn, size, pgprot_decrypted(prot));
3503 }
3504 #endif
3505 
vmf_error(int err)3506 static inline vm_fault_t vmf_error(int err)
3507 {
3508 	if (err == -ENOMEM)
3509 		return VM_FAULT_OOM;
3510 	else if (err == -EHWPOISON)
3511 		return VM_FAULT_HWPOISON;
3512 	return VM_FAULT_SIGBUS;
3513 }
3514 
3515 /*
3516  * Convert errno to return value for ->page_mkwrite() calls.
3517  *
3518  * This should eventually be merged with vmf_error() above, but will need a
3519  * careful audit of all vmf_error() callers.
3520  */
vmf_fs_error(int err)3521 static inline vm_fault_t vmf_fs_error(int err)
3522 {
3523 	if (err == 0)
3524 		return VM_FAULT_LOCKED;
3525 	if (err == -EFAULT || err == -EAGAIN)
3526 		return VM_FAULT_NOPAGE;
3527 	if (err == -ENOMEM)
3528 		return VM_FAULT_OOM;
3529 	/* -ENOSPC, -EDQUOT, -EIO ... */
3530 	return VM_FAULT_SIGBUS;
3531 }
3532 
3533 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
3534 			 unsigned int foll_flags);
3535 
vm_fault_to_errno(vm_fault_t vm_fault,int foll_flags)3536 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
3537 {
3538 	if (vm_fault & VM_FAULT_OOM)
3539 		return -ENOMEM;
3540 	if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
3541 		return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
3542 	if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
3543 		return -EFAULT;
3544 	return 0;
3545 }
3546 
3547 /*
3548  * Indicates whether GUP can follow a PROT_NONE mapped page, or whether
3549  * a (NUMA hinting) fault is required.
3550  */
gup_can_follow_protnone(struct vm_area_struct * vma,unsigned int flags)3551 static inline bool gup_can_follow_protnone(struct vm_area_struct *vma,
3552 					   unsigned int flags)
3553 {
3554 	/*
3555 	 * If callers don't want to honor NUMA hinting faults, no need to
3556 	 * determine if we would actually have to trigger a NUMA hinting fault.
3557 	 */
3558 	if (!(flags & FOLL_HONOR_NUMA_FAULT))
3559 		return true;
3560 
3561 	/*
3562 	 * NUMA hinting faults don't apply in inaccessible (PROT_NONE) VMAs.
3563 	 *
3564 	 * Requiring a fault here even for inaccessible VMAs would mean that
3565 	 * FOLL_FORCE cannot make any progress, because handle_mm_fault()
3566 	 * refuses to process NUMA hinting faults in inaccessible VMAs.
3567 	 */
3568 	return !vma_is_accessible(vma);
3569 }
3570 
3571 typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
3572 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
3573 			       unsigned long size, pte_fn_t fn, void *data);
3574 extern int apply_to_existing_page_range(struct mm_struct *mm,
3575 				   unsigned long address, unsigned long size,
3576 				   pte_fn_t fn, void *data);
3577 
3578 #ifdef CONFIG_PAGE_POISONING
3579 extern void __kernel_poison_pages(struct page *page, int numpages);
3580 extern void __kernel_unpoison_pages(struct page *page, int numpages);
3581 extern bool _page_poisoning_enabled_early;
3582 DECLARE_STATIC_KEY_FALSE(_page_poisoning_enabled);
page_poisoning_enabled(void)3583 static inline bool page_poisoning_enabled(void)
3584 {
3585 	return _page_poisoning_enabled_early;
3586 }
3587 /*
3588  * For use in fast paths after init_mem_debugging() has run, or when a
3589  * false negative result is not harmful when called too early.
3590  */
page_poisoning_enabled_static(void)3591 static inline bool page_poisoning_enabled_static(void)
3592 {
3593 	return static_branch_unlikely(&_page_poisoning_enabled);
3594 }
kernel_poison_pages(struct page * page,int numpages)3595 static inline void kernel_poison_pages(struct page *page, int numpages)
3596 {
3597 	if (page_poisoning_enabled_static())
3598 		__kernel_poison_pages(page, numpages);
3599 }
kernel_unpoison_pages(struct page * page,int numpages)3600 static inline void kernel_unpoison_pages(struct page *page, int numpages)
3601 {
3602 	if (page_poisoning_enabled_static())
3603 		__kernel_unpoison_pages(page, numpages);
3604 }
3605 #else
page_poisoning_enabled(void)3606 static inline bool page_poisoning_enabled(void) { return false; }
page_poisoning_enabled_static(void)3607 static inline bool page_poisoning_enabled_static(void) { return false; }
__kernel_poison_pages(struct page * page,int nunmpages)3608 static inline void __kernel_poison_pages(struct page *page, int nunmpages) { }
kernel_poison_pages(struct page * page,int numpages)3609 static inline void kernel_poison_pages(struct page *page, int numpages) { }
kernel_unpoison_pages(struct page * page,int numpages)3610 static inline void kernel_unpoison_pages(struct page *page, int numpages) { }
3611 #endif
3612 
3613 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_ALLOC_DEFAULT_ON, init_on_alloc);
want_init_on_alloc(gfp_t flags)3614 static inline bool want_init_on_alloc(gfp_t flags)
3615 {
3616 	if (static_branch_maybe(CONFIG_INIT_ON_ALLOC_DEFAULT_ON,
3617 				&init_on_alloc))
3618 		return true;
3619 	return flags & __GFP_ZERO;
3620 }
3621 
3622 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_FREE_DEFAULT_ON, init_on_free);
want_init_on_free(void)3623 static inline bool want_init_on_free(void)
3624 {
3625 	return static_branch_maybe(CONFIG_INIT_ON_FREE_DEFAULT_ON,
3626 				   &init_on_free);
3627 }
3628 
3629 extern bool _debug_pagealloc_enabled_early;
3630 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
3631 
debug_pagealloc_enabled(void)3632 static inline bool debug_pagealloc_enabled(void)
3633 {
3634 	return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
3635 		_debug_pagealloc_enabled_early;
3636 }
3637 
3638 /*
3639  * For use in fast paths after mem_debugging_and_hardening_init() has run,
3640  * or when a false negative result is not harmful when called too early.
3641  */
debug_pagealloc_enabled_static(void)3642 static inline bool debug_pagealloc_enabled_static(void)
3643 {
3644 	if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
3645 		return false;
3646 
3647 	return static_branch_unlikely(&_debug_pagealloc_enabled);
3648 }
3649 
3650 /*
3651  * To support DEBUG_PAGEALLOC architecture must ensure that
3652  * __kernel_map_pages() never fails
3653  */
3654 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
3655 #ifdef CONFIG_DEBUG_PAGEALLOC
debug_pagealloc_map_pages(struct page * page,int numpages)3656 static inline void debug_pagealloc_map_pages(struct page *page, int numpages)
3657 {
3658 	if (debug_pagealloc_enabled_static())
3659 		__kernel_map_pages(page, numpages, 1);
3660 }
3661 
debug_pagealloc_unmap_pages(struct page * page,int numpages)3662 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages)
3663 {
3664 	if (debug_pagealloc_enabled_static())
3665 		__kernel_map_pages(page, numpages, 0);
3666 }
3667 
3668 extern unsigned int _debug_guardpage_minorder;
3669 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
3670 
debug_guardpage_minorder(void)3671 static inline unsigned int debug_guardpage_minorder(void)
3672 {
3673 	return _debug_guardpage_minorder;
3674 }
3675 
debug_guardpage_enabled(void)3676 static inline bool debug_guardpage_enabled(void)
3677 {
3678 	return static_branch_unlikely(&_debug_guardpage_enabled);
3679 }
3680 
page_is_guard(struct page * page)3681 static inline bool page_is_guard(struct page *page)
3682 {
3683 	if (!debug_guardpage_enabled())
3684 		return false;
3685 
3686 	return PageGuard(page);
3687 }
3688 
3689 bool __set_page_guard(struct zone *zone, struct page *page, unsigned int order,
3690 		      int migratetype);
set_page_guard(struct zone * zone,struct page * page,unsigned int order,int migratetype)3691 static inline bool set_page_guard(struct zone *zone, struct page *page,
3692 				  unsigned int order, int migratetype)
3693 {
3694 	if (!debug_guardpage_enabled())
3695 		return false;
3696 	return __set_page_guard(zone, page, order, migratetype);
3697 }
3698 
3699 void __clear_page_guard(struct zone *zone, struct page *page, unsigned int order,
3700 			int migratetype);
clear_page_guard(struct zone * zone,struct page * page,unsigned int order,int migratetype)3701 static inline void clear_page_guard(struct zone *zone, struct page *page,
3702 				    unsigned int order, int migratetype)
3703 {
3704 	if (!debug_guardpage_enabled())
3705 		return;
3706 	__clear_page_guard(zone, page, order, migratetype);
3707 }
3708 
3709 #else	/* CONFIG_DEBUG_PAGEALLOC */
debug_pagealloc_map_pages(struct page * page,int numpages)3710 static inline void debug_pagealloc_map_pages(struct page *page, int numpages) {}
debug_pagealloc_unmap_pages(struct page * page,int numpages)3711 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages) {}
debug_guardpage_minorder(void)3712 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
debug_guardpage_enabled(void)3713 static inline bool debug_guardpage_enabled(void) { return false; }
page_is_guard(struct page * page)3714 static inline bool page_is_guard(struct page *page) { return false; }
set_page_guard(struct zone * zone,struct page * page,unsigned int order,int migratetype)3715 static inline bool set_page_guard(struct zone *zone, struct page *page,
3716 			unsigned int order, int migratetype) { return false; }
clear_page_guard(struct zone * zone,struct page * page,unsigned int order,int migratetype)3717 static inline void clear_page_guard(struct zone *zone, struct page *page,
3718 				unsigned int order, int migratetype) {}
3719 #endif	/* CONFIG_DEBUG_PAGEALLOC */
3720 
3721 #ifdef __HAVE_ARCH_GATE_AREA
3722 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
3723 extern int in_gate_area_no_mm(unsigned long addr);
3724 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
3725 #else
get_gate_vma(struct mm_struct * mm)3726 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
3727 {
3728 	return NULL;
3729 }
in_gate_area_no_mm(unsigned long addr)3730 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
in_gate_area(struct mm_struct * mm,unsigned long addr)3731 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
3732 {
3733 	return 0;
3734 }
3735 #endif	/* __HAVE_ARCH_GATE_AREA */
3736 
3737 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
3738 
3739 #ifdef CONFIG_SYSCTL
3740 extern int sysctl_drop_caches;
3741 int drop_caches_sysctl_handler(struct ctl_table *, int, void *, size_t *,
3742 		loff_t *);
3743 #endif
3744 
3745 void drop_slab(void);
3746 
3747 #ifndef CONFIG_MMU
3748 #define randomize_va_space 0
3749 #else
3750 extern int randomize_va_space;
3751 #endif
3752 
3753 const char * arch_vma_name(struct vm_area_struct *vma);
3754 #ifdef CONFIG_MMU
3755 void print_vma_addr(char *prefix, unsigned long rip);
3756 #else
print_vma_addr(char * prefix,unsigned long rip)3757 static inline void print_vma_addr(char *prefix, unsigned long rip)
3758 {
3759 }
3760 #endif
3761 
3762 void *sparse_buffer_alloc(unsigned long size);
3763 struct page * __populate_section_memmap(unsigned long pfn,
3764 		unsigned long nr_pages, int nid, struct vmem_altmap *altmap,
3765 		struct dev_pagemap *pgmap);
3766 void pmd_init(void *addr);
3767 void pud_init(void *addr);
3768 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
3769 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
3770 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
3771 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
3772 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node,
3773 			    struct vmem_altmap *altmap, struct page *reuse);
3774 void *vmemmap_alloc_block(unsigned long size, int node);
3775 struct vmem_altmap;
3776 void *vmemmap_alloc_block_buf(unsigned long size, int node,
3777 			      struct vmem_altmap *altmap);
3778 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
3779 void vmemmap_set_pmd(pmd_t *pmd, void *p, int node,
3780 		     unsigned long addr, unsigned long next);
3781 int vmemmap_check_pmd(pmd_t *pmd, int node,
3782 		      unsigned long addr, unsigned long next);
3783 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
3784 			       int node, struct vmem_altmap *altmap);
3785 int vmemmap_populate_hugepages(unsigned long start, unsigned long end,
3786 			       int node, struct vmem_altmap *altmap);
3787 int vmemmap_populate(unsigned long start, unsigned long end, int node,
3788 		struct vmem_altmap *altmap);
3789 void vmemmap_populate_print_last(void);
3790 #ifdef CONFIG_MEMORY_HOTPLUG
3791 void vmemmap_free(unsigned long start, unsigned long end,
3792 		struct vmem_altmap *altmap);
3793 #endif
3794 
3795 #define VMEMMAP_RESERVE_NR	2
3796 #ifdef CONFIG_ARCH_WANT_OPTIMIZE_DAX_VMEMMAP
__vmemmap_can_optimize(struct vmem_altmap * altmap,struct dev_pagemap * pgmap)3797 static inline bool __vmemmap_can_optimize(struct vmem_altmap *altmap,
3798 					  struct dev_pagemap *pgmap)
3799 {
3800 	unsigned long nr_pages;
3801 	unsigned long nr_vmemmap_pages;
3802 
3803 	if (!pgmap || !is_power_of_2(sizeof(struct page)))
3804 		return false;
3805 
3806 	nr_pages = pgmap_vmemmap_nr(pgmap);
3807 	nr_vmemmap_pages = ((nr_pages * sizeof(struct page)) >> PAGE_SHIFT);
3808 	/*
3809 	 * For vmemmap optimization with DAX we need minimum 2 vmemmap
3810 	 * pages. See layout diagram in Documentation/mm/vmemmap_dedup.rst
3811 	 */
3812 	return !altmap && (nr_vmemmap_pages > VMEMMAP_RESERVE_NR);
3813 }
3814 /*
3815  * If we don't have an architecture override, use the generic rule
3816  */
3817 #ifndef vmemmap_can_optimize
3818 #define vmemmap_can_optimize __vmemmap_can_optimize
3819 #endif
3820 
3821 #else
vmemmap_can_optimize(struct vmem_altmap * altmap,struct dev_pagemap * pgmap)3822 static inline bool vmemmap_can_optimize(struct vmem_altmap *altmap,
3823 					   struct dev_pagemap *pgmap)
3824 {
3825 	return false;
3826 }
3827 #endif
3828 
3829 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
3830 				  unsigned long nr_pages);
3831 
3832 enum mf_flags {
3833 	MF_COUNT_INCREASED = 1 << 0,
3834 	MF_ACTION_REQUIRED = 1 << 1,
3835 	MF_MUST_KILL = 1 << 2,
3836 	MF_SOFT_OFFLINE = 1 << 3,
3837 	MF_UNPOISON = 1 << 4,
3838 	MF_SW_SIMULATED = 1 << 5,
3839 	MF_NO_RETRY = 1 << 6,
3840 };
3841 int mf_dax_kill_procs(struct address_space *mapping, pgoff_t index,
3842 		      unsigned long count, int mf_flags);
3843 extern int memory_failure(unsigned long pfn, int flags);
3844 extern void memory_failure_queue_kick(int cpu);
3845 extern int unpoison_memory(unsigned long pfn);
3846 extern void shake_page(struct page *p);
3847 extern atomic_long_t num_poisoned_pages __read_mostly;
3848 extern int soft_offline_page(unsigned long pfn, int flags);
3849 #ifdef CONFIG_MEMORY_FAILURE
3850 /*
3851  * Sysfs entries for memory failure handling statistics.
3852  */
3853 extern const struct attribute_group memory_failure_attr_group;
3854 extern void memory_failure_queue(unsigned long pfn, int flags);
3855 extern int __get_huge_page_for_hwpoison(unsigned long pfn, int flags,
3856 					bool *migratable_cleared);
3857 void num_poisoned_pages_inc(unsigned long pfn);
3858 void num_poisoned_pages_sub(unsigned long pfn, long i);
3859 struct task_struct *task_early_kill(struct task_struct *tsk, int force_early);
3860 #else
memory_failure_queue(unsigned long pfn,int flags)3861 static inline void memory_failure_queue(unsigned long pfn, int flags)
3862 {
3863 }
3864 
__get_huge_page_for_hwpoison(unsigned long pfn,int flags,bool * migratable_cleared)3865 static inline int __get_huge_page_for_hwpoison(unsigned long pfn, int flags,
3866 					bool *migratable_cleared)
3867 {
3868 	return 0;
3869 }
3870 
num_poisoned_pages_inc(unsigned long pfn)3871 static inline void num_poisoned_pages_inc(unsigned long pfn)
3872 {
3873 }
3874 
num_poisoned_pages_sub(unsigned long pfn,long i)3875 static inline void num_poisoned_pages_sub(unsigned long pfn, long i)
3876 {
3877 }
3878 #endif
3879 
3880 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_KSM)
3881 void add_to_kill_ksm(struct task_struct *tsk, struct page *p,
3882 		     struct vm_area_struct *vma, struct list_head *to_kill,
3883 		     unsigned long ksm_addr);
3884 #endif
3885 
3886 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_MEMORY_HOTPLUG)
3887 extern void memblk_nr_poison_inc(unsigned long pfn);
3888 extern void memblk_nr_poison_sub(unsigned long pfn, long i);
3889 #else
memblk_nr_poison_inc(unsigned long pfn)3890 static inline void memblk_nr_poison_inc(unsigned long pfn)
3891 {
3892 }
3893 
memblk_nr_poison_sub(unsigned long pfn,long i)3894 static inline void memblk_nr_poison_sub(unsigned long pfn, long i)
3895 {
3896 }
3897 #endif
3898 
3899 #ifndef arch_memory_failure
arch_memory_failure(unsigned long pfn,int flags)3900 static inline int arch_memory_failure(unsigned long pfn, int flags)
3901 {
3902 	return -ENXIO;
3903 }
3904 #endif
3905 
3906 #ifndef arch_is_platform_page
arch_is_platform_page(u64 paddr)3907 static inline bool arch_is_platform_page(u64 paddr)
3908 {
3909 	return false;
3910 }
3911 #endif
3912 
3913 /*
3914  * Error handlers for various types of pages.
3915  */
3916 enum mf_result {
3917 	MF_IGNORED,	/* Error: cannot be handled */
3918 	MF_FAILED,	/* Error: handling failed */
3919 	MF_DELAYED,	/* Will be handled later */
3920 	MF_RECOVERED,	/* Successfully recovered */
3921 };
3922 
3923 enum mf_action_page_type {
3924 	MF_MSG_KERNEL,
3925 	MF_MSG_KERNEL_HIGH_ORDER,
3926 	MF_MSG_SLAB,
3927 	MF_MSG_DIFFERENT_COMPOUND,
3928 	MF_MSG_HUGE,
3929 	MF_MSG_FREE_HUGE,
3930 	MF_MSG_UNMAP_FAILED,
3931 	MF_MSG_DIRTY_SWAPCACHE,
3932 	MF_MSG_CLEAN_SWAPCACHE,
3933 	MF_MSG_DIRTY_MLOCKED_LRU,
3934 	MF_MSG_CLEAN_MLOCKED_LRU,
3935 	MF_MSG_DIRTY_UNEVICTABLE_LRU,
3936 	MF_MSG_CLEAN_UNEVICTABLE_LRU,
3937 	MF_MSG_DIRTY_LRU,
3938 	MF_MSG_CLEAN_LRU,
3939 	MF_MSG_TRUNCATED_LRU,
3940 	MF_MSG_BUDDY,
3941 	MF_MSG_DAX,
3942 	MF_MSG_UNSPLIT_THP,
3943 	MF_MSG_UNKNOWN,
3944 };
3945 
3946 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
3947 extern void clear_huge_page(struct page *page,
3948 			    unsigned long addr_hint,
3949 			    unsigned int pages_per_huge_page);
3950 int copy_user_large_folio(struct folio *dst, struct folio *src,
3951 			  unsigned long addr_hint,
3952 			  struct vm_area_struct *vma);
3953 long copy_folio_from_user(struct folio *dst_folio,
3954 			   const void __user *usr_src,
3955 			   bool allow_pagefault);
3956 
3957 /**
3958  * vma_is_special_huge - Are transhuge page-table entries considered special?
3959  * @vma: Pointer to the struct vm_area_struct to consider
3960  *
3961  * Whether transhuge page-table entries are considered "special" following
3962  * the definition in vm_normal_page().
3963  *
3964  * Return: true if transhuge page-table entries should be considered special,
3965  * false otherwise.
3966  */
vma_is_special_huge(const struct vm_area_struct * vma)3967 static inline bool vma_is_special_huge(const struct vm_area_struct *vma)
3968 {
3969 	return vma_is_dax(vma) || (vma->vm_file &&
3970 				   (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP)));
3971 }
3972 
3973 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
3974 
3975 #if MAX_NUMNODES > 1
3976 void __init setup_nr_node_ids(void);
3977 #else
setup_nr_node_ids(void)3978 static inline void setup_nr_node_ids(void) {}
3979 #endif
3980 
3981 extern int memcmp_pages(struct page *page1, struct page *page2);
3982 
pages_identical(struct page * page1,struct page * page2)3983 static inline int pages_identical(struct page *page1, struct page *page2)
3984 {
3985 	return !memcmp_pages(page1, page2);
3986 }
3987 
3988 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
3989 unsigned long clean_record_shared_mapping_range(struct address_space *mapping,
3990 						pgoff_t first_index, pgoff_t nr,
3991 						pgoff_t bitmap_pgoff,
3992 						unsigned long *bitmap,
3993 						pgoff_t *start,
3994 						pgoff_t *end);
3995 
3996 unsigned long wp_shared_mapping_range(struct address_space *mapping,
3997 				      pgoff_t first_index, pgoff_t nr);
3998 #endif
3999 
4000 extern int sysctl_nr_trim_pages;
4001 
4002 #ifdef CONFIG_PRINTK
4003 void mem_dump_obj(void *object);
4004 #else
mem_dump_obj(void * object)4005 static inline void mem_dump_obj(void *object) {}
4006 #endif
4007 
4008 /**
4009  * seal_check_future_write - Check for F_SEAL_FUTURE_WRITE flag and handle it
4010  * @seals: the seals to check
4011  * @vma: the vma to operate on
4012  *
4013  * Check whether F_SEAL_FUTURE_WRITE is set; if so, do proper check/handling on
4014  * the vma flags.  Return 0 if check pass, or <0 for errors.
4015  */
seal_check_future_write(int seals,struct vm_area_struct * vma)4016 static inline int seal_check_future_write(int seals, struct vm_area_struct *vma)
4017 {
4018 	if (seals & F_SEAL_FUTURE_WRITE) {
4019 		/*
4020 		 * New PROT_WRITE and MAP_SHARED mmaps are not allowed when
4021 		 * "future write" seal active.
4022 		 */
4023 		if ((vma->vm_flags & VM_SHARED) && (vma->vm_flags & VM_WRITE))
4024 			return -EPERM;
4025 
4026 		/*
4027 		 * Since an F_SEAL_FUTURE_WRITE sealed memfd can be mapped as
4028 		 * MAP_SHARED and read-only, take care to not allow mprotect to
4029 		 * revert protections on such mappings. Do this only for shared
4030 		 * mappings. For private mappings, don't need to mask
4031 		 * VM_MAYWRITE as we still want them to be COW-writable.
4032 		 */
4033 		if (vma->vm_flags & VM_SHARED)
4034 			vm_flags_clear(vma, VM_MAYWRITE);
4035 	}
4036 
4037 	return 0;
4038 }
4039 
4040 #ifdef CONFIG_ANON_VMA_NAME
4041 int madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
4042 			  unsigned long len_in,
4043 			  struct anon_vma_name *anon_name);
4044 #else
4045 static inline int
madvise_set_anon_name(struct mm_struct * mm,unsigned long start,unsigned long len_in,struct anon_vma_name * anon_name)4046 madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
4047 		      unsigned long len_in, struct anon_vma_name *anon_name) {
4048 	return 0;
4049 }
4050 #endif
4051 
4052 #ifdef CONFIG_UNACCEPTED_MEMORY
4053 
4054 bool range_contains_unaccepted_memory(phys_addr_t start, phys_addr_t end);
4055 void accept_memory(phys_addr_t start, phys_addr_t end);
4056 
4057 #else
4058 
range_contains_unaccepted_memory(phys_addr_t start,phys_addr_t end)4059 static inline bool range_contains_unaccepted_memory(phys_addr_t start,
4060 						    phys_addr_t end)
4061 {
4062 	return false;
4063 }
4064 
accept_memory(phys_addr_t start,phys_addr_t end)4065 static inline void accept_memory(phys_addr_t start, phys_addr_t end)
4066 {
4067 }
4068 
4069 #endif
4070 
4071 #endif /* _LINUX_MM_H */
4072