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