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/pgalloc_tag.h>
9 #include <linux/bug.h>
10 #include <linux/list.h>
11 #include <linux/mmzone.h>
12 #include <linux/rbtree.h>
13 #include <linux/atomic.h>
14 #include <linux/debug_locks.h>
15 #include <linux/mm_types.h>
16 #include <linux/mmap_lock.h>
17 #include <linux/range.h>
18 #include <linux/pfn.h>
19 #include <linux/percpu-refcount.h>
20 #include <linux/bit_spinlock.h>
21 #include <linux/shrinker.h>
22 #include <linux/resource.h>
23 #include <linux/page_ext.h>
24 #include <linux/err.h>
25 #include <linux/page-flags.h>
26 #include <linux/page_ref.h>
27 #include <linux/overflow.h>
28 #include <linux/sizes.h>
29 #include <linux/sched.h>
30 #include <linux/pgtable.h>
31 #include <linux/kasan.h>
32 #include <linux/memremap.h>
33 #include <linux/slab.h>
34
35 struct mempolicy;
36 struct anon_vma;
37 struct anon_vma_chain;
38 struct user_struct;
39 struct pt_regs;
40 struct folio_batch;
41
42 extern int sysctl_page_lock_unfairness;
43
44 void mm_core_init(void);
45 void init_mm_internals(void);
46
47 #ifndef CONFIG_NUMA /* Don't use mapnrs, do it properly */
48 extern unsigned long max_mapnr;
49
set_max_mapnr(unsigned long limit)50 static inline void set_max_mapnr(unsigned long limit)
51 {
52 max_mapnr = limit;
53 }
54 #else
set_max_mapnr(unsigned long limit)55 static inline void set_max_mapnr(unsigned long limit) { }
56 #endif
57
58 extern atomic_long_t _totalram_pages;
totalram_pages(void)59 static inline unsigned long totalram_pages(void)
60 {
61 return (unsigned long)atomic_long_read(&_totalram_pages);
62 }
63
totalram_pages_inc(void)64 static inline void totalram_pages_inc(void)
65 {
66 atomic_long_inc(&_totalram_pages);
67 }
68
totalram_pages_dec(void)69 static inline void totalram_pages_dec(void)
70 {
71 atomic_long_dec(&_totalram_pages);
72 }
73
totalram_pages_add(long count)74 static inline void totalram_pages_add(long count)
75 {
76 atomic_long_add(count, &_totalram_pages);
77 }
78
79 extern void * high_memory;
80 extern int page_cluster;
81 extern const int page_cluster_max;
82
83 #ifdef CONFIG_SYSCTL
84 extern int sysctl_legacy_va_layout;
85 #else
86 #define sysctl_legacy_va_layout 0
87 #endif
88
89 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
90 extern const int mmap_rnd_bits_min;
91 extern int mmap_rnd_bits_max __ro_after_init;
92 extern int mmap_rnd_bits __read_mostly;
93 #endif
94 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
95 extern const int mmap_rnd_compat_bits_min;
96 extern const int mmap_rnd_compat_bits_max;
97 extern int mmap_rnd_compat_bits __read_mostly;
98 #endif
99
100 #include <asm/page.h>
101 #include <asm/processor.h>
102
103 #ifndef __pa_symbol
104 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
105 #endif
106
107 #ifndef page_to_virt
108 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
109 #endif
110
111 #ifndef lm_alias
112 #define lm_alias(x) __va(__pa_symbol(x))
113 #endif
114
115 /*
116 * To prevent common memory management code establishing
117 * a zero page mapping on a read fault.
118 * This macro should be defined within <asm/pgtable.h>.
119 * s390 does this to prevent multiplexing of hardware bits
120 * related to the physical page in case of virtualization.
121 */
122 #ifndef mm_forbids_zeropage
123 #define mm_forbids_zeropage(X) (0)
124 #endif
125
126 /*
127 * On some architectures it is expensive to call memset() for small sizes.
128 * If an architecture decides to implement their own version of
129 * mm_zero_struct_page they should wrap the defines below in a #ifndef and
130 * define their own version of this macro in <asm/pgtable.h>
131 */
132 #if BITS_PER_LONG == 64
133 /* This function must be updated when the size of struct page grows above 96
134 * or reduces below 56. The idea that compiler optimizes out switch()
135 * statement, and only leaves move/store instructions. Also the compiler can
136 * combine write statements if they are both assignments and can be reordered,
137 * this can result in several of the writes here being dropped.
138 */
139 #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
__mm_zero_struct_page(struct page * page)140 static inline void __mm_zero_struct_page(struct page *page)
141 {
142 unsigned long *_pp = (void *)page;
143
144 /* Check that struct page is either 56, 64, 72, 80, 88 or 96 bytes */
145 BUILD_BUG_ON(sizeof(struct page) & 7);
146 BUILD_BUG_ON(sizeof(struct page) < 56);
147 BUILD_BUG_ON(sizeof(struct page) > 96);
148
149 switch (sizeof(struct page)) {
150 case 96:
151 _pp[11] = 0;
152 fallthrough;
153 case 88:
154 _pp[10] = 0;
155 fallthrough;
156 case 80:
157 _pp[9] = 0;
158 fallthrough;
159 case 72:
160 _pp[8] = 0;
161 fallthrough;
162 case 64:
163 _pp[7] = 0;
164 fallthrough;
165 case 56:
166 _pp[6] = 0;
167 _pp[5] = 0;
168 _pp[4] = 0;
169 _pp[3] = 0;
170 _pp[2] = 0;
171 _pp[1] = 0;
172 _pp[0] = 0;
173 }
174 }
175 #else
176 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
177 #endif
178
179 /*
180 * Default maximum number of active map areas, this limits the number of vmas
181 * per mm struct. Users can overwrite this number by sysctl but there is a
182 * problem.
183 *
184 * When a program's coredump is generated as ELF format, a section is created
185 * per a vma. In ELF, the number of sections is represented in unsigned short.
186 * This means the number of sections should be smaller than 65535 at coredump.
187 * Because the kernel adds some informative sections to a image of program at
188 * generating coredump, we need some margin. The number of extra sections is
189 * 1-3 now and depends on arch. We use "5" as safe margin, here.
190 *
191 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
192 * not a hard limit any more. Although some userspace tools can be surprised by
193 * that.
194 */
195 #define MAPCOUNT_ELF_CORE_MARGIN (5)
196 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
197
198 extern int sysctl_max_map_count;
199
200 extern unsigned long sysctl_user_reserve_kbytes;
201 extern unsigned long sysctl_admin_reserve_kbytes;
202
203 extern int sysctl_overcommit_memory;
204 extern int sysctl_overcommit_ratio;
205 extern unsigned long sysctl_overcommit_kbytes;
206
207 int overcommit_ratio_handler(struct ctl_table *, int, void *, size_t *,
208 loff_t *);
209 int overcommit_kbytes_handler(struct ctl_table *, int, void *, size_t *,
210 loff_t *);
211 int overcommit_policy_handler(struct ctl_table *, int, void *, size_t *,
212 loff_t *);
213
214 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
215 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
216 #define folio_page_idx(folio, p) (page_to_pfn(p) - folio_pfn(folio))
217 #else
218 #define nth_page(page,n) ((page) + (n))
219 #define folio_page_idx(folio, p) ((p) - &(folio)->page)
220 #endif
221
222 /* to align the pointer to the (next) page boundary */
223 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
224
225 /* to align the pointer to the (prev) page boundary */
226 #define PAGE_ALIGN_DOWN(addr) ALIGN_DOWN(addr, PAGE_SIZE)
227
228 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
229 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
230
lru_to_folio(struct list_head * head)231 static inline struct folio *lru_to_folio(struct list_head *head)
232 {
233 return list_entry((head)->prev, struct folio, lru);
234 }
235
236 void setup_initial_init_mm(void *start_code, void *end_code,
237 void *end_data, void *brk);
238
239 /*
240 * Linux kernel virtual memory manager primitives.
241 * The idea being to have a "virtual" mm in the same way
242 * we have a virtual fs - giving a cleaner interface to the
243 * mm details, and allowing different kinds of memory mappings
244 * (from shared memory to executable loading to arbitrary
245 * mmap() functions).
246 */
247
248 struct vm_area_struct *vm_area_alloc(struct mm_struct *);
249 struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
250 void vm_area_free(struct vm_area_struct *);
251 /* Use only if VMA has no other users */
252 void __vm_area_free(struct vm_area_struct *vma);
253
254 #ifndef CONFIG_MMU
255 extern struct rb_root nommu_region_tree;
256 extern struct rw_semaphore nommu_region_sem;
257
258 extern unsigned int kobjsize(const void *objp);
259 #endif
260
261 /*
262 * vm_flags in vm_area_struct, see mm_types.h.
263 * When changing, update also include/trace/events/mmflags.h
264 */
265 #define VM_NONE 0x00000000
266
267 #define VM_READ 0x00000001 /* currently active flags */
268 #define VM_WRITE 0x00000002
269 #define VM_EXEC 0x00000004
270 #define VM_SHARED 0x00000008
271
272 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
273 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
274 #define VM_MAYWRITE 0x00000020
275 #define VM_MAYEXEC 0x00000040
276 #define VM_MAYSHARE 0x00000080
277
278 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
279 #ifdef CONFIG_MMU
280 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
281 #else /* CONFIG_MMU */
282 #define VM_MAYOVERLAY 0x00000200 /* nommu: R/O MAP_PRIVATE mapping that might overlay a file mapping */
283 #define VM_UFFD_MISSING 0
284 #endif /* CONFIG_MMU */
285 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
286 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
287
288 #define VM_LOCKED 0x00002000
289 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
290
291 /* Used by sys_madvise() */
292 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
293 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
294
295 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
296 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
297 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
298 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
299 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
300 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
301 #define VM_SYNC 0x00800000 /* Synchronous page faults */
302 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
303 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
304 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
305
306 #ifdef CONFIG_MEM_SOFT_DIRTY
307 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
308 #else
309 # define VM_SOFTDIRTY 0
310 #endif
311
312 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
313 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
314 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
315 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
316
317 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
318 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
319 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
320 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
321 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
322 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
323 #define VM_HIGH_ARCH_BIT_5 37 /* bit only usable on 64-bit architectures */
324 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
325 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
326 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
327 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
328 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
329 #define VM_HIGH_ARCH_5 BIT(VM_HIGH_ARCH_BIT_5)
330 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
331
332 #ifdef CONFIG_ARCH_HAS_PKEYS
333 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
334 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
335 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
336 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
337 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
338 #ifdef CONFIG_PPC
339 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
340 #else
341 # define VM_PKEY_BIT4 0
342 #endif
343 #endif /* CONFIG_ARCH_HAS_PKEYS */
344
345 #ifdef CONFIG_X86_USER_SHADOW_STACK
346 /*
347 * VM_SHADOW_STACK should not be set with VM_SHARED because of lack of
348 * support core mm.
349 *
350 * These VMAs will get a single end guard page. This helps userspace protect
351 * itself from attacks. A single page is enough for current shadow stack archs
352 * (x86). See the comments near alloc_shstk() in arch/x86/kernel/shstk.c
353 * for more details on the guard size.
354 */
355 # define VM_SHADOW_STACK VM_HIGH_ARCH_5
356 #else
357 # define VM_SHADOW_STACK VM_NONE
358 #endif
359
360 #if defined(CONFIG_X86)
361 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
362 #elif defined(CONFIG_PPC)
363 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
364 #elif defined(CONFIG_PARISC)
365 # define VM_GROWSUP VM_ARCH_1
366 #elif defined(CONFIG_SPARC64)
367 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
368 # define VM_ARCH_CLEAR VM_SPARC_ADI
369 #elif defined(CONFIG_ARM64)
370 # define VM_ARM64_BTI VM_ARCH_1 /* BTI guarded page, a.k.a. GP bit */
371 # define VM_ARCH_CLEAR VM_ARM64_BTI
372 #elif !defined(CONFIG_MMU)
373 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
374 #endif
375
376 #if defined(CONFIG_ARM64_MTE)
377 # define VM_MTE VM_HIGH_ARCH_0 /* Use Tagged memory for access control */
378 # define VM_MTE_ALLOWED VM_HIGH_ARCH_1 /* Tagged memory permitted */
379 #else
380 # define VM_MTE VM_NONE
381 # define VM_MTE_ALLOWED VM_NONE
382 #endif
383
384 #ifndef VM_GROWSUP
385 # define VM_GROWSUP VM_NONE
386 #endif
387
388 #ifdef CONFIG_HAVE_ARCH_USERFAULTFD_MINOR
389 # define VM_UFFD_MINOR_BIT 38
390 # define VM_UFFD_MINOR BIT(VM_UFFD_MINOR_BIT) /* UFFD minor faults */
391 #else /* !CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
392 # define VM_UFFD_MINOR VM_NONE
393 #endif /* CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
394
395 /*
396 * This flag is used to connect VFIO to arch specific KVM code. It
397 * indicates that the memory under this VMA is safe for use with any
398 * non-cachable memory type inside KVM. Some VFIO devices, on some
399 * platforms, are thought to be unsafe and can cause machine crashes
400 * if KVM does not lock down the memory type.
401 */
402 #ifdef CONFIG_64BIT
403 #define VM_ALLOW_ANY_UNCACHED_BIT 39
404 #define VM_ALLOW_ANY_UNCACHED BIT(VM_ALLOW_ANY_UNCACHED_BIT)
405 #else
406 #define VM_ALLOW_ANY_UNCACHED VM_NONE
407 #endif
408
409 #ifdef CONFIG_64BIT
410 /* VM is sealed, in vm_flags */
411 #define VM_SEALED _BITUL(63)
412 #endif
413
414 /* Bits set in the VMA until the stack is in its final location */
415 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ | VM_STACK_EARLY)
416
417 #define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0)
418
419 /* Common data flag combinations */
420 #define VM_DATA_FLAGS_TSK_EXEC (VM_READ | VM_WRITE | TASK_EXEC | \
421 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
422 #define VM_DATA_FLAGS_NON_EXEC (VM_READ | VM_WRITE | VM_MAYREAD | \
423 VM_MAYWRITE | VM_MAYEXEC)
424 #define VM_DATA_FLAGS_EXEC (VM_READ | VM_WRITE | VM_EXEC | \
425 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
426
427 #ifndef VM_DATA_DEFAULT_FLAGS /* arch can override this */
428 #define VM_DATA_DEFAULT_FLAGS VM_DATA_FLAGS_EXEC
429 #endif
430
431 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
432 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
433 #endif
434
435 #define VM_STARTGAP_FLAGS (VM_GROWSDOWN | VM_SHADOW_STACK)
436
437 #ifdef CONFIG_STACK_GROWSUP
438 #define VM_STACK VM_GROWSUP
439 #define VM_STACK_EARLY VM_GROWSDOWN
440 #else
441 #define VM_STACK VM_GROWSDOWN
442 #define VM_STACK_EARLY 0
443 #endif
444
445 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
446
447 /* VMA basic access permission flags */
448 #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
449
450
451 /*
452 * Special vmas that are non-mergable, non-mlock()able.
453 */
454 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
455
456 /* This mask prevents VMA from being scanned with khugepaged */
457 #define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB)
458
459 /* This mask defines which mm->def_flags a process can inherit its parent */
460 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
461
462 /* This mask represents all the VMA flag bits used by mlock */
463 #define VM_LOCKED_MASK (VM_LOCKED | VM_LOCKONFAULT)
464
465 /* Arch-specific flags to clear when updating VM flags on protection change */
466 #ifndef VM_ARCH_CLEAR
467 # define VM_ARCH_CLEAR VM_NONE
468 #endif
469 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
470
471 /*
472 * mapping from the currently active vm_flags protection bits (the
473 * low four bits) to a page protection mask..
474 */
475
476 /*
477 * The default fault flags that should be used by most of the
478 * arch-specific page fault handlers.
479 */
480 #define FAULT_FLAG_DEFAULT (FAULT_FLAG_ALLOW_RETRY | \
481 FAULT_FLAG_KILLABLE | \
482 FAULT_FLAG_INTERRUPTIBLE)
483
484 /**
485 * fault_flag_allow_retry_first - check ALLOW_RETRY the first time
486 * @flags: Fault flags.
487 *
488 * This is mostly used for places where we want to try to avoid taking
489 * the mmap_lock for too long a time when waiting for another condition
490 * to change, in which case we can try to be polite to release the
491 * mmap_lock in the first round to avoid potential starvation of other
492 * processes that would also want the mmap_lock.
493 *
494 * Return: true if the page fault allows retry and this is the first
495 * attempt of the fault handling; false otherwise.
496 */
fault_flag_allow_retry_first(enum fault_flag flags)497 static inline bool fault_flag_allow_retry_first(enum fault_flag flags)
498 {
499 return (flags & FAULT_FLAG_ALLOW_RETRY) &&
500 (!(flags & FAULT_FLAG_TRIED));
501 }
502
503 #define FAULT_FLAG_TRACE \
504 { FAULT_FLAG_WRITE, "WRITE" }, \
505 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
506 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
507 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
508 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
509 { FAULT_FLAG_TRIED, "TRIED" }, \
510 { FAULT_FLAG_USER, "USER" }, \
511 { FAULT_FLAG_REMOTE, "REMOTE" }, \
512 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }, \
513 { FAULT_FLAG_INTERRUPTIBLE, "INTERRUPTIBLE" }, \
514 { FAULT_FLAG_VMA_LOCK, "VMA_LOCK" }
515
516 /*
517 * vm_fault is filled by the pagefault handler and passed to the vma's
518 * ->fault function. The vma's ->fault is responsible for returning a bitmask
519 * of VM_FAULT_xxx flags that give details about how the fault was handled.
520 *
521 * MM layer fills up gfp_mask for page allocations but fault handler might
522 * alter it if its implementation requires a different allocation context.
523 *
524 * pgoff should be used in favour of virtual_address, if possible.
525 */
526 struct vm_fault {
527 const struct {
528 struct vm_area_struct *vma; /* Target VMA */
529 gfp_t gfp_mask; /* gfp mask to be used for allocations */
530 pgoff_t pgoff; /* Logical page offset based on vma */
531 unsigned long address; /* Faulting virtual address - masked */
532 unsigned long real_address; /* Faulting virtual address - unmasked */
533 };
534 enum fault_flag flags; /* FAULT_FLAG_xxx flags
535 * XXX: should really be 'const' */
536 pmd_t *pmd; /* Pointer to pmd entry matching
537 * the 'address' */
538 pud_t *pud; /* Pointer to pud entry matching
539 * the 'address'
540 */
541 union {
542 pte_t orig_pte; /* Value of PTE at the time of fault */
543 pmd_t orig_pmd; /* Value of PMD at the time of fault,
544 * used by PMD fault only.
545 */
546 };
547
548 struct page *cow_page; /* Page handler may use for COW fault */
549 struct page *page; /* ->fault handlers should return a
550 * page here, unless VM_FAULT_NOPAGE
551 * is set (which is also implied by
552 * VM_FAULT_ERROR).
553 */
554 /* These three entries are valid only while holding ptl lock */
555 pte_t *pte; /* Pointer to pte entry matching
556 * the 'address'. NULL if the page
557 * table hasn't been allocated.
558 */
559 spinlock_t *ptl; /* Page table lock.
560 * Protects pte page table if 'pte'
561 * is not NULL, otherwise pmd.
562 */
563 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
564 * vm_ops->map_pages() sets up a page
565 * table from atomic context.
566 * do_fault_around() pre-allocates
567 * page table to avoid allocation from
568 * atomic context.
569 */
570 };
571
572 /*
573 * These are the virtual MM functions - opening of an area, closing and
574 * unmapping it (needed to keep files on disk up-to-date etc), pointer
575 * to the functions called when a no-page or a wp-page exception occurs.
576 */
577 struct vm_operations_struct {
578 void (*open)(struct vm_area_struct * area);
579 /**
580 * @close: Called when the VMA is being removed from the MM.
581 * Context: User context. May sleep. Caller holds mmap_lock.
582 */
583 void (*close)(struct vm_area_struct * area);
584 /* Called any time before splitting to check if it's allowed */
585 int (*may_split)(struct vm_area_struct *area, unsigned long addr);
586 int (*mremap)(struct vm_area_struct *area);
587 /*
588 * Called by mprotect() to make driver-specific permission
589 * checks before mprotect() is finalised. The VMA must not
590 * be modified. Returns 0 if mprotect() can proceed.
591 */
592 int (*mprotect)(struct vm_area_struct *vma, unsigned long start,
593 unsigned long end, unsigned long newflags);
594 vm_fault_t (*fault)(struct vm_fault *vmf);
595 vm_fault_t (*huge_fault)(struct vm_fault *vmf, unsigned int order);
596 vm_fault_t (*map_pages)(struct vm_fault *vmf,
597 pgoff_t start_pgoff, pgoff_t end_pgoff);
598 unsigned long (*pagesize)(struct vm_area_struct * area);
599
600 /* notification that a previously read-only page is about to become
601 * writable, if an error is returned it will cause a SIGBUS */
602 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
603
604 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
605 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
606
607 /* called by access_process_vm when get_user_pages() fails, typically
608 * for use by special VMAs. See also generic_access_phys() for a generic
609 * implementation useful for any iomem mapping.
610 */
611 int (*access)(struct vm_area_struct *vma, unsigned long addr,
612 void *buf, int len, int write);
613
614 /* Called by the /proc/PID/maps code to ask the vma whether it
615 * has a special name. Returning non-NULL will also cause this
616 * vma to be dumped unconditionally. */
617 const char *(*name)(struct vm_area_struct *vma);
618
619 #ifdef CONFIG_NUMA
620 /*
621 * set_policy() op must add a reference to any non-NULL @new mempolicy
622 * to hold the policy upon return. Caller should pass NULL @new to
623 * remove a policy and fall back to surrounding context--i.e. do not
624 * install a MPOL_DEFAULT policy, nor the task or system default
625 * mempolicy.
626 */
627 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
628
629 /*
630 * get_policy() op must add reference [mpol_get()] to any policy at
631 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
632 * in mm/mempolicy.c will do this automatically.
633 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
634 * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
635 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
636 * must return NULL--i.e., do not "fallback" to task or system default
637 * policy.
638 */
639 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
640 unsigned long addr, pgoff_t *ilx);
641 #endif
642 /*
643 * Called by vm_normal_page() for special PTEs to find the
644 * page for @addr. This is useful if the default behavior
645 * (using pte_page()) would not find the correct page.
646 */
647 struct page *(*find_special_page)(struct vm_area_struct *vma,
648 unsigned long addr);
649 };
650
651 #ifdef CONFIG_NUMA_BALANCING
vma_numab_state_init(struct vm_area_struct * vma)652 static inline void vma_numab_state_init(struct vm_area_struct *vma)
653 {
654 vma->numab_state = NULL;
655 }
vma_numab_state_free(struct vm_area_struct * vma)656 static inline void vma_numab_state_free(struct vm_area_struct *vma)
657 {
658 kfree(vma->numab_state);
659 }
660 #else
vma_numab_state_init(struct vm_area_struct * vma)661 static inline void vma_numab_state_init(struct vm_area_struct *vma) {}
vma_numab_state_free(struct vm_area_struct * vma)662 static inline void vma_numab_state_free(struct vm_area_struct *vma) {}
663 #endif /* CONFIG_NUMA_BALANCING */
664
665 #ifdef CONFIG_PER_VMA_LOCK
666 /*
667 * Try to read-lock a vma. The function is allowed to occasionally yield false
668 * locked result to avoid performance overhead, in which case we fall back to
669 * using mmap_lock. The function should never yield false unlocked result.
670 */
vma_start_read(struct vm_area_struct * vma)671 static inline bool vma_start_read(struct vm_area_struct *vma)
672 {
673 /*
674 * Check before locking. A race might cause false locked result.
675 * We can use READ_ONCE() for the mm_lock_seq here, and don't need
676 * ACQUIRE semantics, because this is just a lockless check whose result
677 * we don't rely on for anything - the mm_lock_seq read against which we
678 * need ordering is below.
679 */
680 if (READ_ONCE(vma->vm_lock_seq) == READ_ONCE(vma->vm_mm->mm_lock_seq))
681 return false;
682
683 if (unlikely(down_read_trylock(&vma->vm_lock->lock) == 0))
684 return false;
685
686 /*
687 * Overflow might produce false locked result.
688 * False unlocked result is impossible because we modify and check
689 * vma->vm_lock_seq under vma->vm_lock protection and mm->mm_lock_seq
690 * modification invalidates all existing locks.
691 *
692 * We must use ACQUIRE semantics for the mm_lock_seq so that if we are
693 * racing with vma_end_write_all(), we only start reading from the VMA
694 * after it has been unlocked.
695 * This pairs with RELEASE semantics in vma_end_write_all().
696 */
697 if (unlikely(vma->vm_lock_seq == smp_load_acquire(&vma->vm_mm->mm_lock_seq))) {
698 up_read(&vma->vm_lock->lock);
699 return false;
700 }
701 return true;
702 }
703
vma_end_read(struct vm_area_struct * vma)704 static inline void vma_end_read(struct vm_area_struct *vma)
705 {
706 rcu_read_lock(); /* keeps vma alive till the end of up_read */
707 up_read(&vma->vm_lock->lock);
708 rcu_read_unlock();
709 }
710
711 /* 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)712 static bool __is_vma_write_locked(struct vm_area_struct *vma, int *mm_lock_seq)
713 {
714 mmap_assert_write_locked(vma->vm_mm);
715
716 /*
717 * current task is holding mmap_write_lock, both vma->vm_lock_seq and
718 * mm->mm_lock_seq can't be concurrently modified.
719 */
720 *mm_lock_seq = vma->vm_mm->mm_lock_seq;
721 return (vma->vm_lock_seq == *mm_lock_seq);
722 }
723
724 /*
725 * Begin writing to a VMA.
726 * Exclude concurrent readers under the per-VMA lock until the currently
727 * write-locked mmap_lock is dropped or downgraded.
728 */
vma_start_write(struct vm_area_struct * vma)729 static inline void vma_start_write(struct vm_area_struct *vma)
730 {
731 int mm_lock_seq;
732
733 if (__is_vma_write_locked(vma, &mm_lock_seq))
734 return;
735
736 down_write(&vma->vm_lock->lock);
737 /*
738 * We should use WRITE_ONCE() here because we can have concurrent reads
739 * from the early lockless pessimistic check in vma_start_read().
740 * We don't really care about the correctness of that early check, but
741 * we should use WRITE_ONCE() for cleanliness and to keep KCSAN happy.
742 */
743 WRITE_ONCE(vma->vm_lock_seq, mm_lock_seq);
744 up_write(&vma->vm_lock->lock);
745 }
746
vma_assert_write_locked(struct vm_area_struct * vma)747 static inline void vma_assert_write_locked(struct vm_area_struct *vma)
748 {
749 int mm_lock_seq;
750
751 VM_BUG_ON_VMA(!__is_vma_write_locked(vma, &mm_lock_seq), vma);
752 }
753
vma_assert_locked(struct vm_area_struct * vma)754 static inline void vma_assert_locked(struct vm_area_struct *vma)
755 {
756 if (!rwsem_is_locked(&vma->vm_lock->lock))
757 vma_assert_write_locked(vma);
758 }
759
vma_mark_detached(struct vm_area_struct * vma,bool detached)760 static inline void vma_mark_detached(struct vm_area_struct *vma, bool detached)
761 {
762 /* When detaching vma should be write-locked */
763 if (detached)
764 vma_assert_write_locked(vma);
765 vma->detached = detached;
766 }
767
release_fault_lock(struct vm_fault * vmf)768 static inline void release_fault_lock(struct vm_fault *vmf)
769 {
770 if (vmf->flags & FAULT_FLAG_VMA_LOCK)
771 vma_end_read(vmf->vma);
772 else
773 mmap_read_unlock(vmf->vma->vm_mm);
774 }
775
assert_fault_locked(struct vm_fault * vmf)776 static inline void assert_fault_locked(struct vm_fault *vmf)
777 {
778 if (vmf->flags & FAULT_FLAG_VMA_LOCK)
779 vma_assert_locked(vmf->vma);
780 else
781 mmap_assert_locked(vmf->vma->vm_mm);
782 }
783
784 struct vm_area_struct *lock_vma_under_rcu(struct mm_struct *mm,
785 unsigned long address);
786
787 #else /* CONFIG_PER_VMA_LOCK */
788
vma_start_read(struct vm_area_struct * vma)789 static inline bool vma_start_read(struct vm_area_struct *vma)
790 { return false; }
vma_end_read(struct vm_area_struct * vma)791 static inline void vma_end_read(struct vm_area_struct *vma) {}
vma_start_write(struct vm_area_struct * vma)792 static inline void vma_start_write(struct vm_area_struct *vma) {}
vma_assert_write_locked(struct vm_area_struct * vma)793 static inline void vma_assert_write_locked(struct vm_area_struct *vma)
794 { mmap_assert_write_locked(vma->vm_mm); }
vma_mark_detached(struct vm_area_struct * vma,bool detached)795 static inline void vma_mark_detached(struct vm_area_struct *vma,
796 bool detached) {}
797
lock_vma_under_rcu(struct mm_struct * mm,unsigned long address)798 static inline struct vm_area_struct *lock_vma_under_rcu(struct mm_struct *mm,
799 unsigned long address)
800 {
801 return NULL;
802 }
803
vma_assert_locked(struct vm_area_struct * vma)804 static inline void vma_assert_locked(struct vm_area_struct *vma)
805 {
806 mmap_assert_locked(vma->vm_mm);
807 }
808
release_fault_lock(struct vm_fault * vmf)809 static inline void release_fault_lock(struct vm_fault *vmf)
810 {
811 mmap_read_unlock(vmf->vma->vm_mm);
812 }
813
assert_fault_locked(struct vm_fault * vmf)814 static inline void assert_fault_locked(struct vm_fault *vmf)
815 {
816 mmap_assert_locked(vmf->vma->vm_mm);
817 }
818
819 #endif /* CONFIG_PER_VMA_LOCK */
820
821 extern const struct vm_operations_struct vma_dummy_vm_ops;
822
823 /*
824 * WARNING: vma_init does not initialize vma->vm_lock.
825 * Use vm_area_alloc()/vm_area_free() if vma needs locking.
826 */
vma_init(struct vm_area_struct * vma,struct mm_struct * mm)827 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
828 {
829 memset(vma, 0, sizeof(*vma));
830 vma->vm_mm = mm;
831 vma->vm_ops = &vma_dummy_vm_ops;
832 INIT_LIST_HEAD(&vma->anon_vma_chain);
833 vma_mark_detached(vma, false);
834 vma_numab_state_init(vma);
835 }
836
837 /* 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)838 static inline void vm_flags_init(struct vm_area_struct *vma,
839 vm_flags_t flags)
840 {
841 ACCESS_PRIVATE(vma, __vm_flags) = flags;
842 }
843
844 /*
845 * Use when VMA is part of the VMA tree and modifications need coordination
846 * Note: vm_flags_reset and vm_flags_reset_once do not lock the vma and
847 * it should be locked explicitly beforehand.
848 */
vm_flags_reset(struct vm_area_struct * vma,vm_flags_t flags)849 static inline void vm_flags_reset(struct vm_area_struct *vma,
850 vm_flags_t flags)
851 {
852 vma_assert_write_locked(vma);
853 vm_flags_init(vma, flags);
854 }
855
vm_flags_reset_once(struct vm_area_struct * vma,vm_flags_t flags)856 static inline void vm_flags_reset_once(struct vm_area_struct *vma,
857 vm_flags_t flags)
858 {
859 vma_assert_write_locked(vma);
860 WRITE_ONCE(ACCESS_PRIVATE(vma, __vm_flags), flags);
861 }
862
vm_flags_set(struct vm_area_struct * vma,vm_flags_t flags)863 static inline void vm_flags_set(struct vm_area_struct *vma,
864 vm_flags_t flags)
865 {
866 vma_start_write(vma);
867 ACCESS_PRIVATE(vma, __vm_flags) |= flags;
868 }
869
vm_flags_clear(struct vm_area_struct * vma,vm_flags_t flags)870 static inline void vm_flags_clear(struct vm_area_struct *vma,
871 vm_flags_t flags)
872 {
873 vma_start_write(vma);
874 ACCESS_PRIVATE(vma, __vm_flags) &= ~flags;
875 }
876
877 /*
878 * Use only if VMA is not part of the VMA tree or has no other users and
879 * therefore needs no locking.
880 */
__vm_flags_mod(struct vm_area_struct * vma,vm_flags_t set,vm_flags_t clear)881 static inline void __vm_flags_mod(struct vm_area_struct *vma,
882 vm_flags_t set, vm_flags_t clear)
883 {
884 vm_flags_init(vma, (vma->vm_flags | set) & ~clear);
885 }
886
887 /*
888 * Use only when the order of set/clear operations is unimportant, otherwise
889 * use vm_flags_{set|clear} explicitly.
890 */
vm_flags_mod(struct vm_area_struct * vma,vm_flags_t set,vm_flags_t clear)891 static inline void vm_flags_mod(struct vm_area_struct *vma,
892 vm_flags_t set, vm_flags_t clear)
893 {
894 vma_start_write(vma);
895 __vm_flags_mod(vma, set, clear);
896 }
897
vma_set_anonymous(struct vm_area_struct * vma)898 static inline void vma_set_anonymous(struct vm_area_struct *vma)
899 {
900 vma->vm_ops = NULL;
901 }
902
vma_is_anonymous(struct vm_area_struct * vma)903 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
904 {
905 return !vma->vm_ops;
906 }
907
908 /*
909 * Indicate if the VMA is a heap for the given task; for
910 * /proc/PID/maps that is the heap of the main task.
911 */
vma_is_initial_heap(const struct vm_area_struct * vma)912 static inline bool vma_is_initial_heap(const struct vm_area_struct *vma)
913 {
914 return vma->vm_start < vma->vm_mm->brk &&
915 vma->vm_end > vma->vm_mm->start_brk;
916 }
917
918 /*
919 * Indicate if the VMA is a stack for the given task; for
920 * /proc/PID/maps that is the stack of the main task.
921 */
vma_is_initial_stack(const struct vm_area_struct * vma)922 static inline bool vma_is_initial_stack(const struct vm_area_struct *vma)
923 {
924 /*
925 * We make no effort to guess what a given thread considers to be
926 * its "stack". It's not even well-defined for programs written
927 * languages like Go.
928 */
929 return vma->vm_start <= vma->vm_mm->start_stack &&
930 vma->vm_end >= vma->vm_mm->start_stack;
931 }
932
vma_is_temporary_stack(struct vm_area_struct * vma)933 static inline bool vma_is_temporary_stack(struct vm_area_struct *vma)
934 {
935 int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP);
936
937 if (!maybe_stack)
938 return false;
939
940 if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) ==
941 VM_STACK_INCOMPLETE_SETUP)
942 return true;
943
944 return false;
945 }
946
vma_is_foreign(struct vm_area_struct * vma)947 static inline bool vma_is_foreign(struct vm_area_struct *vma)
948 {
949 if (!current->mm)
950 return true;
951
952 if (current->mm != vma->vm_mm)
953 return true;
954
955 return false;
956 }
957
vma_is_accessible(struct vm_area_struct * vma)958 static inline bool vma_is_accessible(struct vm_area_struct *vma)
959 {
960 return vma->vm_flags & VM_ACCESS_FLAGS;
961 }
962
is_shared_maywrite(vm_flags_t vm_flags)963 static inline bool is_shared_maywrite(vm_flags_t vm_flags)
964 {
965 return (vm_flags & (VM_SHARED | VM_MAYWRITE)) ==
966 (VM_SHARED | VM_MAYWRITE);
967 }
968
vma_is_shared_maywrite(struct vm_area_struct * vma)969 static inline bool vma_is_shared_maywrite(struct vm_area_struct *vma)
970 {
971 return is_shared_maywrite(vma->vm_flags);
972 }
973
974 static inline
vma_find(struct vma_iterator * vmi,unsigned long max)975 struct vm_area_struct *vma_find(struct vma_iterator *vmi, unsigned long max)
976 {
977 return mas_find(&vmi->mas, max - 1);
978 }
979
vma_next(struct vma_iterator * vmi)980 static inline struct vm_area_struct *vma_next(struct vma_iterator *vmi)
981 {
982 /*
983 * Uses mas_find() to get the first VMA when the iterator starts.
984 * Calling mas_next() could skip the first entry.
985 */
986 return mas_find(&vmi->mas, ULONG_MAX);
987 }
988
989 static inline
vma_iter_next_range(struct vma_iterator * vmi)990 struct vm_area_struct *vma_iter_next_range(struct vma_iterator *vmi)
991 {
992 return mas_next_range(&vmi->mas, ULONG_MAX);
993 }
994
995
vma_prev(struct vma_iterator * vmi)996 static inline struct vm_area_struct *vma_prev(struct vma_iterator *vmi)
997 {
998 return mas_prev(&vmi->mas, 0);
999 }
1000
1001 static inline
vma_iter_prev_range(struct vma_iterator * vmi)1002 struct vm_area_struct *vma_iter_prev_range(struct vma_iterator *vmi)
1003 {
1004 return mas_prev_range(&vmi->mas, 0);
1005 }
1006
vma_iter_addr(struct vma_iterator * vmi)1007 static inline unsigned long vma_iter_addr(struct vma_iterator *vmi)
1008 {
1009 return vmi->mas.index;
1010 }
1011
vma_iter_end(struct vma_iterator * vmi)1012 static inline unsigned long vma_iter_end(struct vma_iterator *vmi)
1013 {
1014 return vmi->mas.last + 1;
1015 }
vma_iter_bulk_alloc(struct vma_iterator * vmi,unsigned long count)1016 static inline int vma_iter_bulk_alloc(struct vma_iterator *vmi,
1017 unsigned long count)
1018 {
1019 return mas_expected_entries(&vmi->mas, count);
1020 }
1021
vma_iter_clear_gfp(struct vma_iterator * vmi,unsigned long start,unsigned long end,gfp_t gfp)1022 static inline int vma_iter_clear_gfp(struct vma_iterator *vmi,
1023 unsigned long start, unsigned long end, gfp_t gfp)
1024 {
1025 __mas_set_range(&vmi->mas, start, end - 1);
1026 mas_store_gfp(&vmi->mas, NULL, gfp);
1027 if (unlikely(mas_is_err(&vmi->mas)))
1028 return -ENOMEM;
1029
1030 return 0;
1031 }
1032
1033 /* Free any unused preallocations */
vma_iter_free(struct vma_iterator * vmi)1034 static inline void vma_iter_free(struct vma_iterator *vmi)
1035 {
1036 mas_destroy(&vmi->mas);
1037 }
1038
vma_iter_bulk_store(struct vma_iterator * vmi,struct vm_area_struct * vma)1039 static inline int vma_iter_bulk_store(struct vma_iterator *vmi,
1040 struct vm_area_struct *vma)
1041 {
1042 vmi->mas.index = vma->vm_start;
1043 vmi->mas.last = vma->vm_end - 1;
1044 mas_store(&vmi->mas, vma);
1045 if (unlikely(mas_is_err(&vmi->mas)))
1046 return -ENOMEM;
1047
1048 return 0;
1049 }
1050
vma_iter_invalidate(struct vma_iterator * vmi)1051 static inline void vma_iter_invalidate(struct vma_iterator *vmi)
1052 {
1053 mas_pause(&vmi->mas);
1054 }
1055
vma_iter_set(struct vma_iterator * vmi,unsigned long addr)1056 static inline void vma_iter_set(struct vma_iterator *vmi, unsigned long addr)
1057 {
1058 mas_set(&vmi->mas, addr);
1059 }
1060
1061 #define for_each_vma(__vmi, __vma) \
1062 while (((__vma) = vma_next(&(__vmi))) != NULL)
1063
1064 /* The MM code likes to work with exclusive end addresses */
1065 #define for_each_vma_range(__vmi, __vma, __end) \
1066 while (((__vma) = vma_find(&(__vmi), (__end))) != NULL)
1067
1068 #ifdef CONFIG_SHMEM
1069 /*
1070 * The vma_is_shmem is not inline because it is used only by slow
1071 * paths in userfault.
1072 */
1073 bool vma_is_shmem(struct vm_area_struct *vma);
1074 bool vma_is_anon_shmem(struct vm_area_struct *vma);
1075 #else
vma_is_shmem(struct vm_area_struct * vma)1076 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
vma_is_anon_shmem(struct vm_area_struct * vma)1077 static inline bool vma_is_anon_shmem(struct vm_area_struct *vma) { return false; }
1078 #endif
1079
1080 int vma_is_stack_for_current(struct vm_area_struct *vma);
1081
1082 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
1083 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
1084
1085 struct mmu_gather;
1086 struct inode;
1087
1088 /*
1089 * compound_order() can be called without holding a reference, which means
1090 * that niceties like page_folio() don't work. These callers should be
1091 * prepared to handle wild return values. For example, PG_head may be
1092 * set before the order is initialised, or this may be a tail page.
1093 * See compaction.c for some good examples.
1094 */
compound_order(struct page * page)1095 static inline unsigned int compound_order(struct page *page)
1096 {
1097 struct folio *folio = (struct folio *)page;
1098
1099 if (!test_bit(PG_head, &folio->flags))
1100 return 0;
1101 return folio->_flags_1 & 0xff;
1102 }
1103
1104 /**
1105 * folio_order - The allocation order of a folio.
1106 * @folio: The folio.
1107 *
1108 * A folio is composed of 2^order pages. See get_order() for the definition
1109 * of order.
1110 *
1111 * Return: The order of the folio.
1112 */
folio_order(struct folio * folio)1113 static inline unsigned int folio_order(struct folio *folio)
1114 {
1115 if (!folio_test_large(folio))
1116 return 0;
1117 return folio->_flags_1 & 0xff;
1118 }
1119
1120 #include <linux/huge_mm.h>
1121
1122 /*
1123 * Methods to modify the page usage count.
1124 *
1125 * What counts for a page usage:
1126 * - cache mapping (page->mapping)
1127 * - private data (page->private)
1128 * - page mapped in a task's page tables, each mapping
1129 * is counted separately
1130 *
1131 * Also, many kernel routines increase the page count before a critical
1132 * routine so they can be sure the page doesn't go away from under them.
1133 */
1134
1135 /*
1136 * Drop a ref, return true if the refcount fell to zero (the page has no users)
1137 */
put_page_testzero(struct page * page)1138 static inline int put_page_testzero(struct page *page)
1139 {
1140 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
1141 return page_ref_dec_and_test(page);
1142 }
1143
folio_put_testzero(struct folio * folio)1144 static inline int folio_put_testzero(struct folio *folio)
1145 {
1146 return put_page_testzero(&folio->page);
1147 }
1148
1149 /*
1150 * Try to grab a ref unless the page has a refcount of zero, return false if
1151 * that is the case.
1152 * This can be called when MMU is off so it must not access
1153 * any of the virtual mappings.
1154 */
get_page_unless_zero(struct page * page)1155 static inline bool get_page_unless_zero(struct page *page)
1156 {
1157 return page_ref_add_unless(page, 1, 0);
1158 }
1159
folio_get_nontail_page(struct page * page)1160 static inline struct folio *folio_get_nontail_page(struct page *page)
1161 {
1162 if (unlikely(!get_page_unless_zero(page)))
1163 return NULL;
1164 return (struct folio *)page;
1165 }
1166
1167 extern int page_is_ram(unsigned long pfn);
1168
1169 enum {
1170 REGION_INTERSECTS,
1171 REGION_DISJOINT,
1172 REGION_MIXED,
1173 };
1174
1175 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
1176 unsigned long desc);
1177
1178 /* Support for virtually mapped pages */
1179 struct page *vmalloc_to_page(const void *addr);
1180 unsigned long vmalloc_to_pfn(const void *addr);
1181
1182 /*
1183 * Determine if an address is within the vmalloc range
1184 *
1185 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
1186 * is no special casing required.
1187 */
1188 #ifdef CONFIG_MMU
1189 extern bool is_vmalloc_addr(const void *x);
1190 extern int is_vmalloc_or_module_addr(const void *x);
1191 #else
is_vmalloc_addr(const void * x)1192 static inline bool is_vmalloc_addr(const void *x)
1193 {
1194 return false;
1195 }
is_vmalloc_or_module_addr(const void * x)1196 static inline int is_vmalloc_or_module_addr(const void *x)
1197 {
1198 return 0;
1199 }
1200 #endif
1201
1202 /*
1203 * How many times the entire folio is mapped as a single unit (eg by a
1204 * PMD or PUD entry). This is probably not what you want, except for
1205 * debugging purposes - it does not include PTE-mapped sub-pages; look
1206 * at folio_mapcount() or page_mapcount() instead.
1207 */
folio_entire_mapcount(const struct folio * folio)1208 static inline int folio_entire_mapcount(const struct folio *folio)
1209 {
1210 VM_BUG_ON_FOLIO(!folio_test_large(folio), folio);
1211 return atomic_read(&folio->_entire_mapcount) + 1;
1212 }
1213
1214 /*
1215 * The atomic page->_mapcount, starts from -1: so that transitions
1216 * both from it and to it can be tracked, using atomic_inc_and_test
1217 * and atomic_add_negative(-1).
1218 */
page_mapcount_reset(struct page * page)1219 static inline void page_mapcount_reset(struct page *page)
1220 {
1221 atomic_set(&(page)->_mapcount, -1);
1222 }
1223
1224 /**
1225 * page_mapcount() - Number of times this precise page is mapped.
1226 * @page: The page.
1227 *
1228 * The number of times this page is mapped. If this page is part of
1229 * a large folio, it includes the number of times this page is mapped
1230 * as part of that folio.
1231 *
1232 * Will report 0 for pages which cannot be mapped into userspace, eg
1233 * slab, page tables and similar.
1234 */
page_mapcount(struct page * page)1235 static inline int page_mapcount(struct page *page)
1236 {
1237 int mapcount = atomic_read(&page->_mapcount) + 1;
1238
1239 /* Handle page_has_type() pages */
1240 if (mapcount < PAGE_MAPCOUNT_RESERVE + 1)
1241 mapcount = 0;
1242 if (unlikely(PageCompound(page)))
1243 mapcount += folio_entire_mapcount(page_folio(page));
1244
1245 return mapcount;
1246 }
1247
folio_large_mapcount(const struct folio * folio)1248 static inline int folio_large_mapcount(const struct folio *folio)
1249 {
1250 VM_WARN_ON_FOLIO(!folio_test_large(folio), folio);
1251 return atomic_read(&folio->_large_mapcount) + 1;
1252 }
1253
1254 /**
1255 * folio_mapcount() - Number of mappings of this folio.
1256 * @folio: The folio.
1257 *
1258 * The folio mapcount corresponds to the number of present user page table
1259 * entries that reference any part of a folio. Each such present user page
1260 * table entry must be paired with exactly on folio reference.
1261 *
1262 * For ordindary folios, each user page table entry (PTE/PMD/PUD/...) counts
1263 * exactly once.
1264 *
1265 * For hugetlb folios, each abstracted "hugetlb" user page table entry that
1266 * references the entire folio counts exactly once, even when such special
1267 * page table entries are comprised of multiple ordinary page table entries.
1268 *
1269 * Will report 0 for pages which cannot be mapped into userspace, such as
1270 * slab, page tables and similar.
1271 *
1272 * Return: The number of times this folio is mapped.
1273 */
folio_mapcount(const struct folio * folio)1274 static inline int folio_mapcount(const struct folio *folio)
1275 {
1276 int mapcount;
1277
1278 if (likely(!folio_test_large(folio))) {
1279 mapcount = atomic_read(&folio->_mapcount) + 1;
1280 /* Handle page_has_type() pages */
1281 if (mapcount < PAGE_MAPCOUNT_RESERVE + 1)
1282 mapcount = 0;
1283 return mapcount;
1284 }
1285 return folio_large_mapcount(folio);
1286 }
1287
1288 /**
1289 * folio_mapped - Is this folio mapped into userspace?
1290 * @folio: The folio.
1291 *
1292 * Return: True if any page in this folio is referenced by user page tables.
1293 */
folio_mapped(const struct folio * folio)1294 static inline bool folio_mapped(const struct folio *folio)
1295 {
1296 return folio_mapcount(folio) >= 1;
1297 }
1298
1299 /*
1300 * Return true if this page is mapped into pagetables.
1301 * For compound page it returns true if any sub-page of compound page is mapped,
1302 * even if this particular sub-page is not itself mapped by any PTE or PMD.
1303 */
page_mapped(const struct page * page)1304 static inline bool page_mapped(const struct page *page)
1305 {
1306 return folio_mapped(page_folio(page));
1307 }
1308
virt_to_head_page(const void * x)1309 static inline struct page *virt_to_head_page(const void *x)
1310 {
1311 struct page *page = virt_to_page(x);
1312
1313 return compound_head(page);
1314 }
1315
virt_to_folio(const void * x)1316 static inline struct folio *virt_to_folio(const void *x)
1317 {
1318 struct page *page = virt_to_page(x);
1319
1320 return page_folio(page);
1321 }
1322
1323 void __folio_put(struct folio *folio);
1324
1325 void put_pages_list(struct list_head *pages);
1326
1327 void split_page(struct page *page, unsigned int order);
1328 void folio_copy(struct folio *dst, struct folio *src);
1329
1330 unsigned long nr_free_buffer_pages(void);
1331
1332 /* Returns the number of bytes in this potentially compound page. */
page_size(struct page * page)1333 static inline unsigned long page_size(struct page *page)
1334 {
1335 return PAGE_SIZE << compound_order(page);
1336 }
1337
1338 /* Returns the number of bits needed for the number of bytes in a page */
page_shift(struct page * page)1339 static inline unsigned int page_shift(struct page *page)
1340 {
1341 return PAGE_SHIFT + compound_order(page);
1342 }
1343
1344 /**
1345 * thp_order - Order of a transparent huge page.
1346 * @page: Head page of a transparent huge page.
1347 */
thp_order(struct page * page)1348 static inline unsigned int thp_order(struct page *page)
1349 {
1350 VM_BUG_ON_PGFLAGS(PageTail(page), page);
1351 return compound_order(page);
1352 }
1353
1354 /**
1355 * thp_size - Size of a transparent huge page.
1356 * @page: Head page of a transparent huge page.
1357 *
1358 * Return: Number of bytes in this page.
1359 */
thp_size(struct page * page)1360 static inline unsigned long thp_size(struct page *page)
1361 {
1362 return PAGE_SIZE << thp_order(page);
1363 }
1364
1365 #ifdef CONFIG_MMU
1366 /*
1367 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
1368 * servicing faults for write access. In the normal case, do always want
1369 * pte_mkwrite. But get_user_pages can cause write faults for mappings
1370 * that do not have writing enabled, when used by access_process_vm.
1371 */
maybe_mkwrite(pte_t pte,struct vm_area_struct * vma)1372 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
1373 {
1374 if (likely(vma->vm_flags & VM_WRITE))
1375 pte = pte_mkwrite(pte, vma);
1376 return pte;
1377 }
1378
1379 vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page);
1380 void set_pte_range(struct vm_fault *vmf, struct folio *folio,
1381 struct page *page, unsigned int nr, unsigned long addr);
1382
1383 vm_fault_t finish_fault(struct vm_fault *vmf);
1384 #endif
1385
1386 /*
1387 * Multiple processes may "see" the same page. E.g. for untouched
1388 * mappings of /dev/null, all processes see the same page full of
1389 * zeroes, and text pages of executables and shared libraries have
1390 * only one copy in memory, at most, normally.
1391 *
1392 * For the non-reserved pages, page_count(page) denotes a reference count.
1393 * page_count() == 0 means the page is free. page->lru is then used for
1394 * freelist management in the buddy allocator.
1395 * page_count() > 0 means the page has been allocated.
1396 *
1397 * Pages are allocated by the slab allocator in order to provide memory
1398 * to kmalloc and kmem_cache_alloc. In this case, the management of the
1399 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
1400 * unless a particular usage is carefully commented. (the responsibility of
1401 * freeing the kmalloc memory is the caller's, of course).
1402 *
1403 * A page may be used by anyone else who does a __get_free_page().
1404 * In this case, page_count still tracks the references, and should only
1405 * be used through the normal accessor functions. The top bits of page->flags
1406 * and page->virtual store page management information, but all other fields
1407 * are unused and could be used privately, carefully. The management of this
1408 * page is the responsibility of the one who allocated it, and those who have
1409 * subsequently been given references to it.
1410 *
1411 * The other pages (we may call them "pagecache pages") are completely
1412 * managed by the Linux memory manager: I/O, buffers, swapping etc.
1413 * The following discussion applies only to them.
1414 *
1415 * A pagecache page contains an opaque `private' member, which belongs to the
1416 * page's address_space. Usually, this is the address of a circular list of
1417 * the page's disk buffers. PG_private must be set to tell the VM to call
1418 * into the filesystem to release these pages.
1419 *
1420 * A page may belong to an inode's memory mapping. In this case, page->mapping
1421 * is the pointer to the inode, and page->index is the file offset of the page,
1422 * in units of PAGE_SIZE.
1423 *
1424 * If pagecache pages are not associated with an inode, they are said to be
1425 * anonymous pages. These may become associated with the swapcache, and in that
1426 * case PG_swapcache is set, and page->private is an offset into the swapcache.
1427 *
1428 * In either case (swapcache or inode backed), the pagecache itself holds one
1429 * reference to the page. Setting PG_private should also increment the
1430 * refcount. The each user mapping also has a reference to the page.
1431 *
1432 * The pagecache pages are stored in a per-mapping radix tree, which is
1433 * rooted at mapping->i_pages, and indexed by offset.
1434 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
1435 * lists, we instead now tag pages as dirty/writeback in the radix tree.
1436 *
1437 * All pagecache pages may be subject to I/O:
1438 * - inode pages may need to be read from disk,
1439 * - inode pages which have been modified and are MAP_SHARED may need
1440 * to be written back to the inode on disk,
1441 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
1442 * modified may need to be swapped out to swap space and (later) to be read
1443 * back into memory.
1444 */
1445
1446 #if defined(CONFIG_ZONE_DEVICE) && defined(CONFIG_FS_DAX)
1447 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
1448
1449 bool __put_devmap_managed_folio_refs(struct folio *folio, int refs);
put_devmap_managed_folio_refs(struct folio * folio,int refs)1450 static inline bool put_devmap_managed_folio_refs(struct folio *folio, int refs)
1451 {
1452 if (!static_branch_unlikely(&devmap_managed_key))
1453 return false;
1454 if (!folio_is_zone_device(folio))
1455 return false;
1456 return __put_devmap_managed_folio_refs(folio, refs);
1457 }
1458 #else /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
put_devmap_managed_folio_refs(struct folio * folio,int refs)1459 static inline bool put_devmap_managed_folio_refs(struct folio *folio, int refs)
1460 {
1461 return false;
1462 }
1463 #endif /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1464
1465 /* 127: arbitrary random number, small enough to assemble well */
1466 #define folio_ref_zero_or_close_to_overflow(folio) \
1467 ((unsigned int) folio_ref_count(folio) + 127u <= 127u)
1468
1469 /**
1470 * folio_get - Increment the reference count on a folio.
1471 * @folio: The folio.
1472 *
1473 * Context: May be called in any context, as long as you know that
1474 * you have a refcount on the folio. If you do not already have one,
1475 * folio_try_get() may be the right interface for you to use.
1476 */
folio_get(struct folio * folio)1477 static inline void folio_get(struct folio *folio)
1478 {
1479 VM_BUG_ON_FOLIO(folio_ref_zero_or_close_to_overflow(folio), folio);
1480 folio_ref_inc(folio);
1481 }
1482
get_page(struct page * page)1483 static inline void get_page(struct page *page)
1484 {
1485 folio_get(page_folio(page));
1486 }
1487
try_get_page(struct page * page)1488 static inline __must_check bool try_get_page(struct page *page)
1489 {
1490 page = compound_head(page);
1491 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1492 return false;
1493 page_ref_inc(page);
1494 return true;
1495 }
1496
1497 /**
1498 * folio_put - Decrement the reference count on a folio.
1499 * @folio: The folio.
1500 *
1501 * If the folio's reference count reaches zero, the memory will be
1502 * released back to the page allocator and may be used by another
1503 * allocation immediately. Do not access the memory or the struct folio
1504 * after calling folio_put() unless you can be sure that it wasn't the
1505 * last reference.
1506 *
1507 * Context: May be called in process or interrupt context, but not in NMI
1508 * context. May be called while holding a spinlock.
1509 */
folio_put(struct folio * folio)1510 static inline void folio_put(struct folio *folio)
1511 {
1512 if (folio_put_testzero(folio))
1513 __folio_put(folio);
1514 }
1515
1516 /**
1517 * folio_put_refs - Reduce the reference count on a folio.
1518 * @folio: The folio.
1519 * @refs: The amount to subtract from the folio's reference count.
1520 *
1521 * If the folio's reference count reaches zero, the memory will be
1522 * released back to the page allocator and may be used by another
1523 * allocation immediately. Do not access the memory or the struct folio
1524 * after calling folio_put_refs() unless you can be sure that these weren't
1525 * the last references.
1526 *
1527 * Context: May be called in process or interrupt context, but not in NMI
1528 * context. May be called while holding a spinlock.
1529 */
folio_put_refs(struct folio * folio,int refs)1530 static inline void folio_put_refs(struct folio *folio, int refs)
1531 {
1532 if (folio_ref_sub_and_test(folio, refs))
1533 __folio_put(folio);
1534 }
1535
1536 void folios_put_refs(struct folio_batch *folios, unsigned int *refs);
1537
1538 /*
1539 * union release_pages_arg - an array of pages or folios
1540 *
1541 * release_pages() releases a simple array of multiple pages, and
1542 * accepts various different forms of said page array: either
1543 * a regular old boring array of pages, an array of folios, or
1544 * an array of encoded page pointers.
1545 *
1546 * The transparent union syntax for this kind of "any of these
1547 * argument types" is all kinds of ugly, so look away.
1548 */
1549 typedef union {
1550 struct page **pages;
1551 struct folio **folios;
1552 struct encoded_page **encoded_pages;
1553 } release_pages_arg __attribute__ ((__transparent_union__));
1554
1555 void release_pages(release_pages_arg, int nr);
1556
1557 /**
1558 * folios_put - Decrement the reference count on an array of folios.
1559 * @folios: The folios.
1560 *
1561 * Like folio_put(), but for a batch of folios. This is more efficient
1562 * than writing the loop yourself as it will optimise the locks which need
1563 * to be taken if the folios are freed. The folios batch is returned
1564 * empty and ready to be reused for another batch; there is no need to
1565 * reinitialise it.
1566 *
1567 * Context: May be called in process or interrupt context, but not in NMI
1568 * context. May be called while holding a spinlock.
1569 */
folios_put(struct folio_batch * folios)1570 static inline void folios_put(struct folio_batch *folios)
1571 {
1572 folios_put_refs(folios, NULL);
1573 }
1574
put_page(struct page * page)1575 static inline void put_page(struct page *page)
1576 {
1577 struct folio *folio = page_folio(page);
1578
1579 /*
1580 * For some devmap managed pages we need to catch refcount transition
1581 * from 2 to 1:
1582 */
1583 if (put_devmap_managed_folio_refs(folio, 1))
1584 return;
1585 folio_put(folio);
1586 }
1587
1588 /*
1589 * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
1590 * the page's refcount so that two separate items are tracked: the original page
1591 * reference count, and also a new count of how many pin_user_pages() calls were
1592 * made against the page. ("gup-pinned" is another term for the latter).
1593 *
1594 * With this scheme, pin_user_pages() becomes special: such pages are marked as
1595 * distinct from normal pages. As such, the unpin_user_page() call (and its
1596 * variants) must be used in order to release gup-pinned pages.
1597 *
1598 * Choice of value:
1599 *
1600 * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
1601 * counts with respect to pin_user_pages() and unpin_user_page() becomes
1602 * simpler, due to the fact that adding an even power of two to the page
1603 * refcount has the effect of using only the upper N bits, for the code that
1604 * counts up using the bias value. This means that the lower bits are left for
1605 * the exclusive use of the original code that increments and decrements by one
1606 * (or at least, by much smaller values than the bias value).
1607 *
1608 * Of course, once the lower bits overflow into the upper bits (and this is
1609 * OK, because subtraction recovers the original values), then visual inspection
1610 * no longer suffices to directly view the separate counts. However, for normal
1611 * applications that don't have huge page reference counts, this won't be an
1612 * issue.
1613 *
1614 * Locking: the lockless algorithm described in folio_try_get_rcu()
1615 * provides safe operation for get_user_pages(), page_mkclean() and
1616 * other calls that race to set up page table entries.
1617 */
1618 #define GUP_PIN_COUNTING_BIAS (1U << 10)
1619
1620 void unpin_user_page(struct page *page);
1621 void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1622 bool make_dirty);
1623 void unpin_user_page_range_dirty_lock(struct page *page, unsigned long npages,
1624 bool make_dirty);
1625 void unpin_user_pages(struct page **pages, unsigned long npages);
1626
is_cow_mapping(vm_flags_t flags)1627 static inline bool is_cow_mapping(vm_flags_t flags)
1628 {
1629 return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
1630 }
1631
1632 #ifndef CONFIG_MMU
is_nommu_shared_mapping(vm_flags_t flags)1633 static inline bool is_nommu_shared_mapping(vm_flags_t flags)
1634 {
1635 /*
1636 * NOMMU shared mappings are ordinary MAP_SHARED mappings and selected
1637 * R/O MAP_PRIVATE file mappings that are an effective R/O overlay of
1638 * a file mapping. R/O MAP_PRIVATE mappings might still modify
1639 * underlying memory if ptrace is active, so this is only possible if
1640 * ptrace does not apply. Note that there is no mprotect() to upgrade
1641 * write permissions later.
1642 */
1643 return flags & (VM_MAYSHARE | VM_MAYOVERLAY);
1644 }
1645 #endif
1646
1647 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1648 #define SECTION_IN_PAGE_FLAGS
1649 #endif
1650
1651 /*
1652 * The identification function is mainly used by the buddy allocator for
1653 * determining if two pages could be buddies. We are not really identifying
1654 * the zone since we could be using the section number id if we do not have
1655 * node id available in page flags.
1656 * We only guarantee that it will return the same value for two combinable
1657 * pages in a zone.
1658 */
page_zone_id(struct page * page)1659 static inline int page_zone_id(struct page *page)
1660 {
1661 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1662 }
1663
1664 #ifdef NODE_NOT_IN_PAGE_FLAGS
1665 int page_to_nid(const struct page *page);
1666 #else
page_to_nid(const struct page * page)1667 static inline int page_to_nid(const struct page *page)
1668 {
1669 return (PF_POISONED_CHECK(page)->flags >> NODES_PGSHIFT) & NODES_MASK;
1670 }
1671 #endif
1672
folio_nid(const struct folio * folio)1673 static inline int folio_nid(const struct folio *folio)
1674 {
1675 return page_to_nid(&folio->page);
1676 }
1677
1678 #ifdef CONFIG_NUMA_BALANCING
1679 /* page access time bits needs to hold at least 4 seconds */
1680 #define PAGE_ACCESS_TIME_MIN_BITS 12
1681 #if LAST_CPUPID_SHIFT < PAGE_ACCESS_TIME_MIN_BITS
1682 #define PAGE_ACCESS_TIME_BUCKETS \
1683 (PAGE_ACCESS_TIME_MIN_BITS - LAST_CPUPID_SHIFT)
1684 #else
1685 #define PAGE_ACCESS_TIME_BUCKETS 0
1686 #endif
1687
1688 #define PAGE_ACCESS_TIME_MASK \
1689 (LAST_CPUPID_MASK << PAGE_ACCESS_TIME_BUCKETS)
1690
cpu_pid_to_cpupid(int cpu,int pid)1691 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1692 {
1693 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1694 }
1695
cpupid_to_pid(int cpupid)1696 static inline int cpupid_to_pid(int cpupid)
1697 {
1698 return cpupid & LAST__PID_MASK;
1699 }
1700
cpupid_to_cpu(int cpupid)1701 static inline int cpupid_to_cpu(int cpupid)
1702 {
1703 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1704 }
1705
cpupid_to_nid(int cpupid)1706 static inline int cpupid_to_nid(int cpupid)
1707 {
1708 return cpu_to_node(cpupid_to_cpu(cpupid));
1709 }
1710
cpupid_pid_unset(int cpupid)1711 static inline bool cpupid_pid_unset(int cpupid)
1712 {
1713 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1714 }
1715
cpupid_cpu_unset(int cpupid)1716 static inline bool cpupid_cpu_unset(int cpupid)
1717 {
1718 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1719 }
1720
__cpupid_match_pid(pid_t task_pid,int cpupid)1721 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1722 {
1723 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1724 }
1725
1726 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1727 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
folio_xchg_last_cpupid(struct folio * folio,int cpupid)1728 static inline int folio_xchg_last_cpupid(struct folio *folio, int cpupid)
1729 {
1730 return xchg(&folio->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1731 }
1732
folio_last_cpupid(struct folio * folio)1733 static inline int folio_last_cpupid(struct folio *folio)
1734 {
1735 return folio->_last_cpupid;
1736 }
page_cpupid_reset_last(struct page * page)1737 static inline void page_cpupid_reset_last(struct page *page)
1738 {
1739 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1740 }
1741 #else
folio_last_cpupid(struct folio * folio)1742 static inline int folio_last_cpupid(struct folio *folio)
1743 {
1744 return (folio->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1745 }
1746
1747 int folio_xchg_last_cpupid(struct folio *folio, int cpupid);
1748
page_cpupid_reset_last(struct page * page)1749 static inline void page_cpupid_reset_last(struct page *page)
1750 {
1751 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1752 }
1753 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1754
folio_xchg_access_time(struct folio * folio,int time)1755 static inline int folio_xchg_access_time(struct folio *folio, int time)
1756 {
1757 int last_time;
1758
1759 last_time = folio_xchg_last_cpupid(folio,
1760 time >> PAGE_ACCESS_TIME_BUCKETS);
1761 return last_time << PAGE_ACCESS_TIME_BUCKETS;
1762 }
1763
vma_set_access_pid_bit(struct vm_area_struct * vma)1764 static inline void vma_set_access_pid_bit(struct vm_area_struct *vma)
1765 {
1766 unsigned int pid_bit;
1767
1768 pid_bit = hash_32(current->pid, ilog2(BITS_PER_LONG));
1769 if (vma->numab_state && !test_bit(pid_bit, &vma->numab_state->pids_active[1])) {
1770 __set_bit(pid_bit, &vma->numab_state->pids_active[1]);
1771 }
1772 }
1773 #else /* !CONFIG_NUMA_BALANCING */
folio_xchg_last_cpupid(struct folio * folio,int cpupid)1774 static inline int folio_xchg_last_cpupid(struct folio *folio, int cpupid)
1775 {
1776 return folio_nid(folio); /* XXX */
1777 }
1778
folio_xchg_access_time(struct folio * folio,int time)1779 static inline int folio_xchg_access_time(struct folio *folio, int time)
1780 {
1781 return 0;
1782 }
1783
folio_last_cpupid(struct folio * folio)1784 static inline int folio_last_cpupid(struct folio *folio)
1785 {
1786 return folio_nid(folio); /* XXX */
1787 }
1788
cpupid_to_nid(int cpupid)1789 static inline int cpupid_to_nid(int cpupid)
1790 {
1791 return -1;
1792 }
1793
cpupid_to_pid(int cpupid)1794 static inline int cpupid_to_pid(int cpupid)
1795 {
1796 return -1;
1797 }
1798
cpupid_to_cpu(int cpupid)1799 static inline int cpupid_to_cpu(int cpupid)
1800 {
1801 return -1;
1802 }
1803
cpu_pid_to_cpupid(int nid,int pid)1804 static inline int cpu_pid_to_cpupid(int nid, int pid)
1805 {
1806 return -1;
1807 }
1808
cpupid_pid_unset(int cpupid)1809 static inline bool cpupid_pid_unset(int cpupid)
1810 {
1811 return true;
1812 }
1813
page_cpupid_reset_last(struct page * page)1814 static inline void page_cpupid_reset_last(struct page *page)
1815 {
1816 }
1817
cpupid_match_pid(struct task_struct * task,int cpupid)1818 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1819 {
1820 return false;
1821 }
1822
vma_set_access_pid_bit(struct vm_area_struct * vma)1823 static inline void vma_set_access_pid_bit(struct vm_area_struct *vma)
1824 {
1825 }
1826 #endif /* CONFIG_NUMA_BALANCING */
1827
1828 #if defined(CONFIG_KASAN_SW_TAGS) || defined(CONFIG_KASAN_HW_TAGS)
1829
1830 /*
1831 * KASAN per-page tags are stored xor'ed with 0xff. This allows to avoid
1832 * setting tags for all pages to native kernel tag value 0xff, as the default
1833 * value 0x00 maps to 0xff.
1834 */
1835
page_kasan_tag(const struct page * page)1836 static inline u8 page_kasan_tag(const struct page *page)
1837 {
1838 u8 tag = KASAN_TAG_KERNEL;
1839
1840 if (kasan_enabled()) {
1841 tag = (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1842 tag ^= 0xff;
1843 }
1844
1845 return tag;
1846 }
1847
page_kasan_tag_set(struct page * page,u8 tag)1848 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1849 {
1850 unsigned long old_flags, flags;
1851
1852 if (!kasan_enabled())
1853 return;
1854
1855 tag ^= 0xff;
1856 old_flags = READ_ONCE(page->flags);
1857 do {
1858 flags = old_flags;
1859 flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1860 flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1861 } while (unlikely(!try_cmpxchg(&page->flags, &old_flags, flags)));
1862 }
1863
page_kasan_tag_reset(struct page * page)1864 static inline void page_kasan_tag_reset(struct page *page)
1865 {
1866 if (kasan_enabled())
1867 page_kasan_tag_set(page, KASAN_TAG_KERNEL);
1868 }
1869
1870 #else /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1871
page_kasan_tag(const struct page * page)1872 static inline u8 page_kasan_tag(const struct page *page)
1873 {
1874 return 0xff;
1875 }
1876
page_kasan_tag_set(struct page * page,u8 tag)1877 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
page_kasan_tag_reset(struct page * page)1878 static inline void page_kasan_tag_reset(struct page *page) { }
1879
1880 #endif /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1881
page_zone(const struct page * page)1882 static inline struct zone *page_zone(const struct page *page)
1883 {
1884 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1885 }
1886
page_pgdat(const struct page * page)1887 static inline pg_data_t *page_pgdat(const struct page *page)
1888 {
1889 return NODE_DATA(page_to_nid(page));
1890 }
1891
folio_zone(const struct folio * folio)1892 static inline struct zone *folio_zone(const struct folio *folio)
1893 {
1894 return page_zone(&folio->page);
1895 }
1896
folio_pgdat(const struct folio * folio)1897 static inline pg_data_t *folio_pgdat(const struct folio *folio)
1898 {
1899 return page_pgdat(&folio->page);
1900 }
1901
1902 #ifdef SECTION_IN_PAGE_FLAGS
set_page_section(struct page * page,unsigned long section)1903 static inline void set_page_section(struct page *page, unsigned long section)
1904 {
1905 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1906 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1907 }
1908
page_to_section(const struct page * page)1909 static inline unsigned long page_to_section(const struct page *page)
1910 {
1911 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1912 }
1913 #endif
1914
1915 /**
1916 * folio_pfn - Return the Page Frame Number of a folio.
1917 * @folio: The folio.
1918 *
1919 * A folio may contain multiple pages. The pages have consecutive
1920 * Page Frame Numbers.
1921 *
1922 * Return: The Page Frame Number of the first page in the folio.
1923 */
folio_pfn(struct folio * folio)1924 static inline unsigned long folio_pfn(struct folio *folio)
1925 {
1926 return page_to_pfn(&folio->page);
1927 }
1928
pfn_folio(unsigned long pfn)1929 static inline struct folio *pfn_folio(unsigned long pfn)
1930 {
1931 return page_folio(pfn_to_page(pfn));
1932 }
1933
1934 /**
1935 * folio_maybe_dma_pinned - Report if a folio may be pinned for DMA.
1936 * @folio: The folio.
1937 *
1938 * This function checks if a folio has been pinned via a call to
1939 * a function in the pin_user_pages() family.
1940 *
1941 * For small folios, the return value is partially fuzzy: false is not fuzzy,
1942 * because it means "definitely not pinned for DMA", but true means "probably
1943 * pinned for DMA, but possibly a false positive due to having at least
1944 * GUP_PIN_COUNTING_BIAS worth of normal folio references".
1945 *
1946 * False positives are OK, because: a) it's unlikely for a folio to
1947 * get that many refcounts, and b) all the callers of this routine are
1948 * expected to be able to deal gracefully with a false positive.
1949 *
1950 * For large folios, the result will be exactly correct. That's because
1951 * we have more tracking data available: the _pincount field is used
1952 * instead of the GUP_PIN_COUNTING_BIAS scheme.
1953 *
1954 * For more information, please see Documentation/core-api/pin_user_pages.rst.
1955 *
1956 * Return: True, if it is likely that the page has been "dma-pinned".
1957 * False, if the page is definitely not dma-pinned.
1958 */
folio_maybe_dma_pinned(struct folio * folio)1959 static inline bool folio_maybe_dma_pinned(struct folio *folio)
1960 {
1961 if (folio_test_large(folio))
1962 return atomic_read(&folio->_pincount) > 0;
1963
1964 /*
1965 * folio_ref_count() is signed. If that refcount overflows, then
1966 * folio_ref_count() returns a negative value, and callers will avoid
1967 * further incrementing the refcount.
1968 *
1969 * Here, for that overflow case, use the sign bit to count a little
1970 * bit higher via unsigned math, and thus still get an accurate result.
1971 */
1972 return ((unsigned int)folio_ref_count(folio)) >=
1973 GUP_PIN_COUNTING_BIAS;
1974 }
1975
page_maybe_dma_pinned(struct page * page)1976 static inline bool page_maybe_dma_pinned(struct page *page)
1977 {
1978 return folio_maybe_dma_pinned(page_folio(page));
1979 }
1980
1981 /*
1982 * This should most likely only be called during fork() to see whether we
1983 * should break the cow immediately for an anon page on the src mm.
1984 *
1985 * The caller has to hold the PT lock and the vma->vm_mm->->write_protect_seq.
1986 */
folio_needs_cow_for_dma(struct vm_area_struct * vma,struct folio * folio)1987 static inline bool folio_needs_cow_for_dma(struct vm_area_struct *vma,
1988 struct folio *folio)
1989 {
1990 VM_BUG_ON(!(raw_read_seqcount(&vma->vm_mm->write_protect_seq) & 1));
1991
1992 if (!test_bit(MMF_HAS_PINNED, &vma->vm_mm->flags))
1993 return false;
1994
1995 return folio_maybe_dma_pinned(folio);
1996 }
1997
1998 /**
1999 * is_zero_page - Query if a page is a zero page
2000 * @page: The page to query
2001 *
2002 * This returns true if @page is one of the permanent zero pages.
2003 */
is_zero_page(const struct page * page)2004 static inline bool is_zero_page(const struct page *page)
2005 {
2006 return is_zero_pfn(page_to_pfn(page));
2007 }
2008
2009 /**
2010 * is_zero_folio - Query if a folio is a zero page
2011 * @folio: The folio to query
2012 *
2013 * This returns true if @folio is one of the permanent zero pages.
2014 */
is_zero_folio(const struct folio * folio)2015 static inline bool is_zero_folio(const struct folio *folio)
2016 {
2017 return is_zero_page(&folio->page);
2018 }
2019
2020 /* MIGRATE_CMA and ZONE_MOVABLE do not allow pin folios */
2021 #ifdef CONFIG_MIGRATION
folio_is_longterm_pinnable(struct folio * folio)2022 static inline bool folio_is_longterm_pinnable(struct folio *folio)
2023 {
2024 #ifdef CONFIG_CMA
2025 int mt = folio_migratetype(folio);
2026
2027 if (mt == MIGRATE_CMA || mt == MIGRATE_ISOLATE)
2028 return false;
2029 #endif
2030 /* The zero page can be "pinned" but gets special handling. */
2031 if (is_zero_folio(folio))
2032 return true;
2033
2034 /* Coherent device memory must always allow eviction. */
2035 if (folio_is_device_coherent(folio))
2036 return false;
2037
2038 /* Otherwise, non-movable zone folios can be pinned. */
2039 return !folio_is_zone_movable(folio);
2040
2041 }
2042 #else
folio_is_longterm_pinnable(struct folio * folio)2043 static inline bool folio_is_longterm_pinnable(struct folio *folio)
2044 {
2045 return true;
2046 }
2047 #endif
2048
set_page_zone(struct page * page,enum zone_type zone)2049 static inline void set_page_zone(struct page *page, enum zone_type zone)
2050 {
2051 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
2052 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
2053 }
2054
set_page_node(struct page * page,unsigned long node)2055 static inline void set_page_node(struct page *page, unsigned long node)
2056 {
2057 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
2058 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
2059 }
2060
set_page_links(struct page * page,enum zone_type zone,unsigned long node,unsigned long pfn)2061 static inline void set_page_links(struct page *page, enum zone_type zone,
2062 unsigned long node, unsigned long pfn)
2063 {
2064 set_page_zone(page, zone);
2065 set_page_node(page, node);
2066 #ifdef SECTION_IN_PAGE_FLAGS
2067 set_page_section(page, pfn_to_section_nr(pfn));
2068 #endif
2069 }
2070
2071 /**
2072 * folio_nr_pages - The number of pages in the folio.
2073 * @folio: The folio.
2074 *
2075 * Return: A positive power of two.
2076 */
folio_nr_pages(const struct folio * folio)2077 static inline long folio_nr_pages(const struct folio *folio)
2078 {
2079 if (!folio_test_large(folio))
2080 return 1;
2081 #ifdef CONFIG_64BIT
2082 return folio->_folio_nr_pages;
2083 #else
2084 return 1L << (folio->_flags_1 & 0xff);
2085 #endif
2086 }
2087
2088 /* Only hugetlbfs can allocate folios larger than MAX_ORDER */
2089 #ifdef CONFIG_ARCH_HAS_GIGANTIC_PAGE
2090 #define MAX_FOLIO_NR_PAGES (1UL << PUD_ORDER)
2091 #else
2092 #define MAX_FOLIO_NR_PAGES MAX_ORDER_NR_PAGES
2093 #endif
2094
2095 /*
2096 * compound_nr() returns the number of pages in this potentially compound
2097 * page. compound_nr() can be called on a tail page, and is defined to
2098 * return 1 in that case.
2099 */
compound_nr(struct page * page)2100 static inline unsigned long compound_nr(struct page *page)
2101 {
2102 struct folio *folio = (struct folio *)page;
2103
2104 if (!test_bit(PG_head, &folio->flags))
2105 return 1;
2106 #ifdef CONFIG_64BIT
2107 return folio->_folio_nr_pages;
2108 #else
2109 return 1L << (folio->_flags_1 & 0xff);
2110 #endif
2111 }
2112
2113 /**
2114 * thp_nr_pages - The number of regular pages in this huge page.
2115 * @page: The head page of a huge page.
2116 */
thp_nr_pages(struct page * page)2117 static inline int thp_nr_pages(struct page *page)
2118 {
2119 return folio_nr_pages((struct folio *)page);
2120 }
2121
2122 /**
2123 * folio_next - Move to the next physical folio.
2124 * @folio: The folio we're currently operating on.
2125 *
2126 * If you have physically contiguous memory which may span more than
2127 * one folio (eg a &struct bio_vec), use this function to move from one
2128 * folio to the next. Do not use it if the memory is only virtually
2129 * contiguous as the folios are almost certainly not adjacent to each
2130 * other. This is the folio equivalent to writing ``page++``.
2131 *
2132 * Context: We assume that the folios are refcounted and/or locked at a
2133 * higher level and do not adjust the reference counts.
2134 * Return: The next struct folio.
2135 */
folio_next(struct folio * folio)2136 static inline struct folio *folio_next(struct folio *folio)
2137 {
2138 return (struct folio *)folio_page(folio, folio_nr_pages(folio));
2139 }
2140
2141 /**
2142 * folio_shift - The size of the memory described by this folio.
2143 * @folio: The folio.
2144 *
2145 * A folio represents a number of bytes which is a power-of-two in size.
2146 * This function tells you which power-of-two the folio is. See also
2147 * folio_size() and folio_order().
2148 *
2149 * Context: The caller should have a reference on the folio to prevent
2150 * it from being split. It is not necessary for the folio to be locked.
2151 * Return: The base-2 logarithm of the size of this folio.
2152 */
folio_shift(struct folio * folio)2153 static inline unsigned int folio_shift(struct folio *folio)
2154 {
2155 return PAGE_SHIFT + folio_order(folio);
2156 }
2157
2158 /**
2159 * folio_size - The number of bytes in a folio.
2160 * @folio: The folio.
2161 *
2162 * Context: The caller should have a reference on the folio to prevent
2163 * it from being split. It is not necessary for the folio to be locked.
2164 * Return: The number of bytes in this folio.
2165 */
folio_size(struct folio * folio)2166 static inline size_t folio_size(struct folio *folio)
2167 {
2168 return PAGE_SIZE << folio_order(folio);
2169 }
2170
2171 /**
2172 * folio_likely_mapped_shared - Estimate if the folio is mapped into the page
2173 * tables of more than one MM
2174 * @folio: The folio.
2175 *
2176 * This function checks if the folio is currently mapped into more than one
2177 * MM ("mapped shared"), or if the folio is only mapped into a single MM
2178 * ("mapped exclusively").
2179 *
2180 * As precise information is not easily available for all folios, this function
2181 * estimates the number of MMs ("sharers") that are currently mapping a folio
2182 * using the number of times the first page of the folio is currently mapped
2183 * into page tables.
2184 *
2185 * For small anonymous folios (except KSM folios) and anonymous hugetlb folios,
2186 * the return value will be exactly correct, because they can only be mapped
2187 * at most once into an MM, and they cannot be partially mapped.
2188 *
2189 * For other folios, the result can be fuzzy:
2190 * #. For partially-mappable large folios (THP), the return value can wrongly
2191 * indicate "mapped exclusively" (false negative) when the folio is
2192 * only partially mapped into at least one MM.
2193 * #. For pagecache folios (including hugetlb), the return value can wrongly
2194 * indicate "mapped shared" (false positive) when two VMAs in the same MM
2195 * cover the same file range.
2196 * #. For (small) KSM folios, the return value can wrongly indicate "mapped
2197 * shared" (false positive), when the folio is mapped multiple times into
2198 * the same MM.
2199 *
2200 * Further, this function only considers current page table mappings that
2201 * are tracked using the folio mapcount(s).
2202 *
2203 * This function does not consider:
2204 * #. If the folio might get mapped in the (near) future (e.g., swapcache,
2205 * pagecache, temporary unmapping for migration).
2206 * #. If the folio is mapped differently (VM_PFNMAP).
2207 * #. If hugetlb page table sharing applies. Callers might want to check
2208 * hugetlb_pmd_shared().
2209 *
2210 * Return: Whether the folio is estimated to be mapped into more than one MM.
2211 */
folio_likely_mapped_shared(struct folio * folio)2212 static inline bool folio_likely_mapped_shared(struct folio *folio)
2213 {
2214 int mapcount = folio_mapcount(folio);
2215
2216 /* Only partially-mappable folios require more care. */
2217 if (!folio_test_large(folio) || unlikely(folio_test_hugetlb(folio)))
2218 return mapcount > 1;
2219
2220 /* A single mapping implies "mapped exclusively". */
2221 if (mapcount <= 1)
2222 return false;
2223
2224 /* If any page is mapped more than once we treat it "mapped shared". */
2225 if (folio_entire_mapcount(folio) || mapcount > folio_nr_pages(folio))
2226 return true;
2227
2228 /* Let's guess based on the first subpage. */
2229 return atomic_read(&folio->_mapcount) > 0;
2230 }
2231
2232 #ifndef HAVE_ARCH_MAKE_PAGE_ACCESSIBLE
arch_make_page_accessible(struct page * page)2233 static inline int arch_make_page_accessible(struct page *page)
2234 {
2235 return 0;
2236 }
2237 #endif
2238
2239 #ifndef HAVE_ARCH_MAKE_FOLIO_ACCESSIBLE
arch_make_folio_accessible(struct folio * folio)2240 static inline int arch_make_folio_accessible(struct folio *folio)
2241 {
2242 int ret;
2243 long i, nr = folio_nr_pages(folio);
2244
2245 for (i = 0; i < nr; i++) {
2246 ret = arch_make_page_accessible(folio_page(folio, i));
2247 if (ret)
2248 break;
2249 }
2250
2251 return ret;
2252 }
2253 #endif
2254
2255 /*
2256 * Some inline functions in vmstat.h depend on page_zone()
2257 */
2258 #include <linux/vmstat.h>
2259
2260 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
2261 #define HASHED_PAGE_VIRTUAL
2262 #endif
2263
2264 #if defined(WANT_PAGE_VIRTUAL)
page_address(const struct page * page)2265 static inline void *page_address(const struct page *page)
2266 {
2267 return page->virtual;
2268 }
set_page_address(struct page * page,void * address)2269 static inline void set_page_address(struct page *page, void *address)
2270 {
2271 page->virtual = address;
2272 }
2273 #define page_address_init() do { } while(0)
2274 #endif
2275
2276 #if defined(HASHED_PAGE_VIRTUAL)
2277 void *page_address(const struct page *page);
2278 void set_page_address(struct page *page, void *virtual);
2279 void page_address_init(void);
2280 #endif
2281
lowmem_page_address(const struct page * page)2282 static __always_inline void *lowmem_page_address(const struct page *page)
2283 {
2284 return page_to_virt(page);
2285 }
2286
2287 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
2288 #define page_address(page) lowmem_page_address(page)
2289 #define set_page_address(page, address) do { } while(0)
2290 #define page_address_init() do { } while(0)
2291 #endif
2292
folio_address(const struct folio * folio)2293 static inline void *folio_address(const struct folio *folio)
2294 {
2295 return page_address(&folio->page);
2296 }
2297
2298 extern pgoff_t __page_file_index(struct page *page);
2299
2300 /*
2301 * Return the pagecache index of the passed page. Regular pagecache pages
2302 * use ->index whereas swapcache pages use swp_offset(->private)
2303 */
page_index(struct page * page)2304 static inline pgoff_t page_index(struct page *page)
2305 {
2306 if (unlikely(PageSwapCache(page)))
2307 return __page_file_index(page);
2308 return page->index;
2309 }
2310
2311 /*
2312 * Return true only if the page has been allocated with
2313 * ALLOC_NO_WATERMARKS and the low watermark was not
2314 * met implying that the system is under some pressure.
2315 */
page_is_pfmemalloc(const struct page * page)2316 static inline bool page_is_pfmemalloc(const struct page *page)
2317 {
2318 /*
2319 * lru.next has bit 1 set if the page is allocated from the
2320 * pfmemalloc reserves. Callers may simply overwrite it if
2321 * they do not need to preserve that information.
2322 */
2323 return (uintptr_t)page->lru.next & BIT(1);
2324 }
2325
2326 /*
2327 * Return true only if the folio has been allocated with
2328 * ALLOC_NO_WATERMARKS and the low watermark was not
2329 * met implying that the system is under some pressure.
2330 */
folio_is_pfmemalloc(const struct folio * folio)2331 static inline bool folio_is_pfmemalloc(const struct folio *folio)
2332 {
2333 /*
2334 * lru.next has bit 1 set if the page is allocated from the
2335 * pfmemalloc reserves. Callers may simply overwrite it if
2336 * they do not need to preserve that information.
2337 */
2338 return (uintptr_t)folio->lru.next & BIT(1);
2339 }
2340
2341 /*
2342 * Only to be called by the page allocator on a freshly allocated
2343 * page.
2344 */
set_page_pfmemalloc(struct page * page)2345 static inline void set_page_pfmemalloc(struct page *page)
2346 {
2347 page->lru.next = (void *)BIT(1);
2348 }
2349
clear_page_pfmemalloc(struct page * page)2350 static inline void clear_page_pfmemalloc(struct page *page)
2351 {
2352 page->lru.next = NULL;
2353 }
2354
2355 /*
2356 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
2357 */
2358 extern void pagefault_out_of_memory(void);
2359
2360 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
2361 #define offset_in_thp(page, p) ((unsigned long)(p) & (thp_size(page) - 1))
2362 #define offset_in_folio(folio, p) ((unsigned long)(p) & (folio_size(folio) - 1))
2363
2364 /*
2365 * Parameter block passed down to zap_pte_range in exceptional cases.
2366 */
2367 struct zap_details {
2368 struct folio *single_folio; /* Locked folio to be unmapped */
2369 bool even_cows; /* Zap COWed private pages too? */
2370 zap_flags_t zap_flags; /* Extra flags for zapping */
2371 };
2372
2373 /*
2374 * Whether to drop the pte markers, for example, the uffd-wp information for
2375 * file-backed memory. This should only be specified when we will completely
2376 * drop the page in the mm, either by truncation or unmapping of the vma. By
2377 * default, the flag is not set.
2378 */
2379 #define ZAP_FLAG_DROP_MARKER ((__force zap_flags_t) BIT(0))
2380 /* Set in unmap_vmas() to indicate a final unmap call. Only used by hugetlb */
2381 #define ZAP_FLAG_UNMAP ((__force zap_flags_t) BIT(1))
2382
2383 #ifdef CONFIG_SCHED_MM_CID
2384 void sched_mm_cid_before_execve(struct task_struct *t);
2385 void sched_mm_cid_after_execve(struct task_struct *t);
2386 void sched_mm_cid_fork(struct task_struct *t);
2387 void sched_mm_cid_exit_signals(struct task_struct *t);
task_mm_cid(struct task_struct * t)2388 static inline int task_mm_cid(struct task_struct *t)
2389 {
2390 return t->mm_cid;
2391 }
2392 #else
sched_mm_cid_before_execve(struct task_struct * t)2393 static inline void sched_mm_cid_before_execve(struct task_struct *t) { }
sched_mm_cid_after_execve(struct task_struct * t)2394 static inline void sched_mm_cid_after_execve(struct task_struct *t) { }
sched_mm_cid_fork(struct task_struct * t)2395 static inline void sched_mm_cid_fork(struct task_struct *t) { }
sched_mm_cid_exit_signals(struct task_struct * t)2396 static inline void sched_mm_cid_exit_signals(struct task_struct *t) { }
task_mm_cid(struct task_struct * t)2397 static inline int task_mm_cid(struct task_struct *t)
2398 {
2399 /*
2400 * Use the processor id as a fall-back when the mm cid feature is
2401 * disabled. This provides functional per-cpu data structure accesses
2402 * in user-space, althrough it won't provide the memory usage benefits.
2403 */
2404 return raw_smp_processor_id();
2405 }
2406 #endif
2407
2408 #ifdef CONFIG_MMU
2409 extern bool can_do_mlock(void);
2410 #else
can_do_mlock(void)2411 static inline bool can_do_mlock(void) { return false; }
2412 #endif
2413 extern int user_shm_lock(size_t, struct ucounts *);
2414 extern void user_shm_unlock(size_t, struct ucounts *);
2415
2416 struct folio *vm_normal_folio(struct vm_area_struct *vma, unsigned long addr,
2417 pte_t pte);
2418 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
2419 pte_t pte);
2420 struct folio *vm_normal_folio_pmd(struct vm_area_struct *vma,
2421 unsigned long addr, pmd_t pmd);
2422 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
2423 pmd_t pmd);
2424
2425 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
2426 unsigned long size);
2427 void zap_page_range_single(struct vm_area_struct *vma, unsigned long address,
2428 unsigned long size, struct zap_details *details);
zap_vma_pages(struct vm_area_struct * vma)2429 static inline void zap_vma_pages(struct vm_area_struct *vma)
2430 {
2431 zap_page_range_single(vma, vma->vm_start,
2432 vma->vm_end - vma->vm_start, NULL);
2433 }
2434 void unmap_vmas(struct mmu_gather *tlb, struct ma_state *mas,
2435 struct vm_area_struct *start_vma, unsigned long start,
2436 unsigned long end, unsigned long tree_end, bool mm_wr_locked);
2437
2438 struct mmu_notifier_range;
2439
2440 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
2441 unsigned long end, unsigned long floor, unsigned long ceiling);
2442 int
2443 copy_page_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma);
2444 int follow_pte(struct vm_area_struct *vma, unsigned long address,
2445 pte_t **ptepp, spinlock_t **ptlp);
2446 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
2447 void *buf, int len, int write);
2448
2449 extern void truncate_pagecache(struct inode *inode, loff_t new);
2450 extern void truncate_setsize(struct inode *inode, loff_t newsize);
2451 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
2452 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
2453 int generic_error_remove_folio(struct address_space *mapping,
2454 struct folio *folio);
2455
2456 struct vm_area_struct *lock_mm_and_find_vma(struct mm_struct *mm,
2457 unsigned long address, struct pt_regs *regs);
2458
2459 #ifdef CONFIG_MMU
2460 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
2461 unsigned long address, unsigned int flags,
2462 struct pt_regs *regs);
2463 extern int fixup_user_fault(struct mm_struct *mm,
2464 unsigned long address, unsigned int fault_flags,
2465 bool *unlocked);
2466 void unmap_mapping_pages(struct address_space *mapping,
2467 pgoff_t start, pgoff_t nr, bool even_cows);
2468 void unmap_mapping_range(struct address_space *mapping,
2469 loff_t const holebegin, loff_t const holelen, int even_cows);
2470 #else
handle_mm_fault(struct vm_area_struct * vma,unsigned long address,unsigned int flags,struct pt_regs * regs)2471 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
2472 unsigned long address, unsigned int flags,
2473 struct pt_regs *regs)
2474 {
2475 /* should never happen if there's no MMU */
2476 BUG();
2477 return VM_FAULT_SIGBUS;
2478 }
fixup_user_fault(struct mm_struct * mm,unsigned long address,unsigned int fault_flags,bool * unlocked)2479 static inline int fixup_user_fault(struct mm_struct *mm, unsigned long address,
2480 unsigned int fault_flags, bool *unlocked)
2481 {
2482 /* should never happen if there's no MMU */
2483 BUG();
2484 return -EFAULT;
2485 }
unmap_mapping_pages(struct address_space * mapping,pgoff_t start,pgoff_t nr,bool even_cows)2486 static inline void unmap_mapping_pages(struct address_space *mapping,
2487 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)2488 static inline void unmap_mapping_range(struct address_space *mapping,
2489 loff_t const holebegin, loff_t const holelen, int even_cows) { }
2490 #endif
2491
unmap_shared_mapping_range(struct address_space * mapping,loff_t const holebegin,loff_t const holelen)2492 static inline void unmap_shared_mapping_range(struct address_space *mapping,
2493 loff_t const holebegin, loff_t const holelen)
2494 {
2495 unmap_mapping_range(mapping, holebegin, holelen, 0);
2496 }
2497
2498 static inline struct vm_area_struct *vma_lookup(struct mm_struct *mm,
2499 unsigned long addr);
2500
2501 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
2502 void *buf, int len, unsigned int gup_flags);
2503 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
2504 void *buf, int len, unsigned int gup_flags);
2505
2506 long get_user_pages_remote(struct mm_struct *mm,
2507 unsigned long start, unsigned long nr_pages,
2508 unsigned int gup_flags, struct page **pages,
2509 int *locked);
2510 long pin_user_pages_remote(struct mm_struct *mm,
2511 unsigned long start, unsigned long nr_pages,
2512 unsigned int gup_flags, struct page **pages,
2513 int *locked);
2514
2515 /*
2516 * Retrieves a single page alongside its VMA. Does not support FOLL_NOWAIT.
2517 */
get_user_page_vma_remote(struct mm_struct * mm,unsigned long addr,int gup_flags,struct vm_area_struct ** vmap)2518 static inline struct page *get_user_page_vma_remote(struct mm_struct *mm,
2519 unsigned long addr,
2520 int gup_flags,
2521 struct vm_area_struct **vmap)
2522 {
2523 struct page *page;
2524 struct vm_area_struct *vma;
2525 int got;
2526
2527 if (WARN_ON_ONCE(unlikely(gup_flags & FOLL_NOWAIT)))
2528 return ERR_PTR(-EINVAL);
2529
2530 got = get_user_pages_remote(mm, addr, 1, gup_flags, &page, NULL);
2531
2532 if (got < 0)
2533 return ERR_PTR(got);
2534
2535 vma = vma_lookup(mm, addr);
2536 if (WARN_ON_ONCE(!vma)) {
2537 put_page(page);
2538 return ERR_PTR(-EINVAL);
2539 }
2540
2541 *vmap = vma;
2542 return page;
2543 }
2544
2545 long get_user_pages(unsigned long start, unsigned long nr_pages,
2546 unsigned int gup_flags, struct page **pages);
2547 long pin_user_pages(unsigned long start, unsigned long nr_pages,
2548 unsigned int gup_flags, struct page **pages);
2549 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
2550 struct page **pages, unsigned int gup_flags);
2551 long pin_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
2552 struct page **pages, unsigned int gup_flags);
2553
2554 int get_user_pages_fast(unsigned long start, int nr_pages,
2555 unsigned int gup_flags, struct page **pages);
2556 int pin_user_pages_fast(unsigned long start, int nr_pages,
2557 unsigned int gup_flags, struct page **pages);
2558 void folio_add_pin(struct folio *folio);
2559
2560 int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
2561 int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
2562 struct task_struct *task, bool bypass_rlim);
2563
2564 struct kvec;
2565 struct page *get_dump_page(unsigned long addr);
2566
2567 bool folio_mark_dirty(struct folio *folio);
2568 bool set_page_dirty(struct page *page);
2569 int set_page_dirty_lock(struct page *page);
2570
2571 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
2572
2573 extern unsigned long move_page_tables(struct vm_area_struct *vma,
2574 unsigned long old_addr, struct vm_area_struct *new_vma,
2575 unsigned long new_addr, unsigned long len,
2576 bool need_rmap_locks, bool for_stack);
2577
2578 /*
2579 * Flags used by change_protection(). For now we make it a bitmap so
2580 * that we can pass in multiple flags just like parameters. However
2581 * for now all the callers are only use one of the flags at the same
2582 * time.
2583 */
2584 /*
2585 * Whether we should manually check if we can map individual PTEs writable,
2586 * because something (e.g., COW, uffd-wp) blocks that from happening for all
2587 * PTEs automatically in a writable mapping.
2588 */
2589 #define MM_CP_TRY_CHANGE_WRITABLE (1UL << 0)
2590 /* Whether this protection change is for NUMA hints */
2591 #define MM_CP_PROT_NUMA (1UL << 1)
2592 /* Whether this change is for write protecting */
2593 #define MM_CP_UFFD_WP (1UL << 2) /* do wp */
2594 #define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */
2595 #define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \
2596 MM_CP_UFFD_WP_RESOLVE)
2597
2598 bool vma_needs_dirty_tracking(struct vm_area_struct *vma);
2599 bool vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
vma_wants_manual_pte_write_upgrade(struct vm_area_struct * vma)2600 static inline bool vma_wants_manual_pte_write_upgrade(struct vm_area_struct *vma)
2601 {
2602 /*
2603 * We want to check manually if we can change individual PTEs writable
2604 * if we can't do that automatically for all PTEs in a mapping. For
2605 * private mappings, that's always the case when we have write
2606 * permissions as we properly have to handle COW.
2607 */
2608 if (vma->vm_flags & VM_SHARED)
2609 return vma_wants_writenotify(vma, vma->vm_page_prot);
2610 return !!(vma->vm_flags & VM_WRITE);
2611
2612 }
2613 bool can_change_pte_writable(struct vm_area_struct *vma, unsigned long addr,
2614 pte_t pte);
2615 extern long change_protection(struct mmu_gather *tlb,
2616 struct vm_area_struct *vma, unsigned long start,
2617 unsigned long end, unsigned long cp_flags);
2618 extern int mprotect_fixup(struct vma_iterator *vmi, struct mmu_gather *tlb,
2619 struct vm_area_struct *vma, struct vm_area_struct **pprev,
2620 unsigned long start, unsigned long end, unsigned long newflags);
2621
2622 /*
2623 * doesn't attempt to fault and will return short.
2624 */
2625 int get_user_pages_fast_only(unsigned long start, int nr_pages,
2626 unsigned int gup_flags, struct page **pages);
2627
get_user_page_fast_only(unsigned long addr,unsigned int gup_flags,struct page ** pagep)2628 static inline bool get_user_page_fast_only(unsigned long addr,
2629 unsigned int gup_flags, struct page **pagep)
2630 {
2631 return get_user_pages_fast_only(addr, 1, gup_flags, pagep) == 1;
2632 }
2633 /*
2634 * per-process(per-mm_struct) statistics.
2635 */
get_mm_counter(struct mm_struct * mm,int member)2636 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
2637 {
2638 return percpu_counter_read_positive(&mm->rss_stat[member]);
2639 }
2640
2641 void mm_trace_rss_stat(struct mm_struct *mm, int member);
2642
add_mm_counter(struct mm_struct * mm,int member,long value)2643 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
2644 {
2645 percpu_counter_add(&mm->rss_stat[member], value);
2646
2647 mm_trace_rss_stat(mm, member);
2648 }
2649
inc_mm_counter(struct mm_struct * mm,int member)2650 static inline void inc_mm_counter(struct mm_struct *mm, int member)
2651 {
2652 percpu_counter_inc(&mm->rss_stat[member]);
2653
2654 mm_trace_rss_stat(mm, member);
2655 }
2656
dec_mm_counter(struct mm_struct * mm,int member)2657 static inline void dec_mm_counter(struct mm_struct *mm, int member)
2658 {
2659 percpu_counter_dec(&mm->rss_stat[member]);
2660
2661 mm_trace_rss_stat(mm, member);
2662 }
2663
2664 /* Optimized variant when folio is already known not to be anon */
mm_counter_file(struct folio * folio)2665 static inline int mm_counter_file(struct folio *folio)
2666 {
2667 if (folio_test_swapbacked(folio))
2668 return MM_SHMEMPAGES;
2669 return MM_FILEPAGES;
2670 }
2671
mm_counter(struct folio * folio)2672 static inline int mm_counter(struct folio *folio)
2673 {
2674 if (folio_test_anon(folio))
2675 return MM_ANONPAGES;
2676 return mm_counter_file(folio);
2677 }
2678
get_mm_rss(struct mm_struct * mm)2679 static inline unsigned long get_mm_rss(struct mm_struct *mm)
2680 {
2681 return get_mm_counter(mm, MM_FILEPAGES) +
2682 get_mm_counter(mm, MM_ANONPAGES) +
2683 get_mm_counter(mm, MM_SHMEMPAGES);
2684 }
2685
get_mm_hiwater_rss(struct mm_struct * mm)2686 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
2687 {
2688 return max(mm->hiwater_rss, get_mm_rss(mm));
2689 }
2690
get_mm_hiwater_vm(struct mm_struct * mm)2691 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
2692 {
2693 return max(mm->hiwater_vm, mm->total_vm);
2694 }
2695
update_hiwater_rss(struct mm_struct * mm)2696 static inline void update_hiwater_rss(struct mm_struct *mm)
2697 {
2698 unsigned long _rss = get_mm_rss(mm);
2699
2700 if ((mm)->hiwater_rss < _rss)
2701 (mm)->hiwater_rss = _rss;
2702 }
2703
update_hiwater_vm(struct mm_struct * mm)2704 static inline void update_hiwater_vm(struct mm_struct *mm)
2705 {
2706 if (mm->hiwater_vm < mm->total_vm)
2707 mm->hiwater_vm = mm->total_vm;
2708 }
2709
reset_mm_hiwater_rss(struct mm_struct * mm)2710 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
2711 {
2712 mm->hiwater_rss = get_mm_rss(mm);
2713 }
2714
setmax_mm_hiwater_rss(unsigned long * maxrss,struct mm_struct * mm)2715 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
2716 struct mm_struct *mm)
2717 {
2718 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
2719
2720 if (*maxrss < hiwater_rss)
2721 *maxrss = hiwater_rss;
2722 }
2723
2724 #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
pte_special(pte_t pte)2725 static inline int pte_special(pte_t pte)
2726 {
2727 return 0;
2728 }
2729
pte_mkspecial(pte_t pte)2730 static inline pte_t pte_mkspecial(pte_t pte)
2731 {
2732 return pte;
2733 }
2734 #endif
2735
2736 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
pte_devmap(pte_t pte)2737 static inline int pte_devmap(pte_t pte)
2738 {
2739 return 0;
2740 }
2741 #endif
2742
2743 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
2744 spinlock_t **ptl);
get_locked_pte(struct mm_struct * mm,unsigned long addr,spinlock_t ** ptl)2745 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
2746 spinlock_t **ptl)
2747 {
2748 pte_t *ptep;
2749 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
2750 return ptep;
2751 }
2752
2753 #ifdef __PAGETABLE_P4D_FOLDED
__p4d_alloc(struct mm_struct * mm,pgd_t * pgd,unsigned long address)2754 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2755 unsigned long address)
2756 {
2757 return 0;
2758 }
2759 #else
2760 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
2761 #endif
2762
2763 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
__pud_alloc(struct mm_struct * mm,p4d_t * p4d,unsigned long address)2764 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2765 unsigned long address)
2766 {
2767 return 0;
2768 }
mm_inc_nr_puds(struct mm_struct * mm)2769 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
mm_dec_nr_puds(struct mm_struct * mm)2770 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
2771
2772 #else
2773 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
2774
mm_inc_nr_puds(struct mm_struct * mm)2775 static inline void mm_inc_nr_puds(struct mm_struct *mm)
2776 {
2777 if (mm_pud_folded(mm))
2778 return;
2779 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2780 }
2781
mm_dec_nr_puds(struct mm_struct * mm)2782 static inline void mm_dec_nr_puds(struct mm_struct *mm)
2783 {
2784 if (mm_pud_folded(mm))
2785 return;
2786 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2787 }
2788 #endif
2789
2790 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
__pmd_alloc(struct mm_struct * mm,pud_t * pud,unsigned long address)2791 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
2792 unsigned long address)
2793 {
2794 return 0;
2795 }
2796
mm_inc_nr_pmds(struct mm_struct * mm)2797 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
mm_dec_nr_pmds(struct mm_struct * mm)2798 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
2799
2800 #else
2801 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
2802
mm_inc_nr_pmds(struct mm_struct * mm)2803 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
2804 {
2805 if (mm_pmd_folded(mm))
2806 return;
2807 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2808 }
2809
mm_dec_nr_pmds(struct mm_struct * mm)2810 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
2811 {
2812 if (mm_pmd_folded(mm))
2813 return;
2814 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2815 }
2816 #endif
2817
2818 #ifdef CONFIG_MMU
mm_pgtables_bytes_init(struct mm_struct * mm)2819 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
2820 {
2821 atomic_long_set(&mm->pgtables_bytes, 0);
2822 }
2823
mm_pgtables_bytes(const struct mm_struct * mm)2824 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2825 {
2826 return atomic_long_read(&mm->pgtables_bytes);
2827 }
2828
mm_inc_nr_ptes(struct mm_struct * mm)2829 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
2830 {
2831 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2832 }
2833
mm_dec_nr_ptes(struct mm_struct * mm)2834 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
2835 {
2836 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2837 }
2838 #else
2839
mm_pgtables_bytes_init(struct mm_struct * mm)2840 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
mm_pgtables_bytes(const struct mm_struct * mm)2841 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2842 {
2843 return 0;
2844 }
2845
mm_inc_nr_ptes(struct mm_struct * mm)2846 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
mm_dec_nr_ptes(struct mm_struct * mm)2847 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
2848 #endif
2849
2850 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
2851 int __pte_alloc_kernel(pmd_t *pmd);
2852
2853 #if defined(CONFIG_MMU)
2854
p4d_alloc(struct mm_struct * mm,pgd_t * pgd,unsigned long address)2855 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2856 unsigned long address)
2857 {
2858 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
2859 NULL : p4d_offset(pgd, address);
2860 }
2861
pud_alloc(struct mm_struct * mm,p4d_t * p4d,unsigned long address)2862 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2863 unsigned long address)
2864 {
2865 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
2866 NULL : pud_offset(p4d, address);
2867 }
2868
pmd_alloc(struct mm_struct * mm,pud_t * pud,unsigned long address)2869 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
2870 {
2871 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
2872 NULL: pmd_offset(pud, address);
2873 }
2874 #endif /* CONFIG_MMU */
2875
virt_to_ptdesc(const void * x)2876 static inline struct ptdesc *virt_to_ptdesc(const void *x)
2877 {
2878 return page_ptdesc(virt_to_page(x));
2879 }
2880
ptdesc_to_virt(const struct ptdesc * pt)2881 static inline void *ptdesc_to_virt(const struct ptdesc *pt)
2882 {
2883 return page_to_virt(ptdesc_page(pt));
2884 }
2885
ptdesc_address(const struct ptdesc * pt)2886 static inline void *ptdesc_address(const struct ptdesc *pt)
2887 {
2888 return folio_address(ptdesc_folio(pt));
2889 }
2890
pagetable_is_reserved(struct ptdesc * pt)2891 static inline bool pagetable_is_reserved(struct ptdesc *pt)
2892 {
2893 return folio_test_reserved(ptdesc_folio(pt));
2894 }
2895
2896 /**
2897 * pagetable_alloc - Allocate pagetables
2898 * @gfp: GFP flags
2899 * @order: desired pagetable order
2900 *
2901 * pagetable_alloc allocates memory for page tables as well as a page table
2902 * descriptor to describe that memory.
2903 *
2904 * Return: The ptdesc describing the allocated page tables.
2905 */
pagetable_alloc_noprof(gfp_t gfp,unsigned int order)2906 static inline struct ptdesc *pagetable_alloc_noprof(gfp_t gfp, unsigned int order)
2907 {
2908 struct page *page = alloc_pages_noprof(gfp | __GFP_COMP, order);
2909
2910 return page_ptdesc(page);
2911 }
2912 #define pagetable_alloc(...) alloc_hooks(pagetable_alloc_noprof(__VA_ARGS__))
2913
2914 /**
2915 * pagetable_free - Free pagetables
2916 * @pt: The page table descriptor
2917 *
2918 * pagetable_free frees the memory of all page tables described by a page
2919 * table descriptor and the memory for the descriptor itself.
2920 */
pagetable_free(struct ptdesc * pt)2921 static inline void pagetable_free(struct ptdesc *pt)
2922 {
2923 struct page *page = ptdesc_page(pt);
2924
2925 __free_pages(page, compound_order(page));
2926 }
2927
2928 #if USE_SPLIT_PTE_PTLOCKS
2929 #if ALLOC_SPLIT_PTLOCKS
2930 void __init ptlock_cache_init(void);
2931 bool ptlock_alloc(struct ptdesc *ptdesc);
2932 void ptlock_free(struct ptdesc *ptdesc);
2933
ptlock_ptr(struct ptdesc * ptdesc)2934 static inline spinlock_t *ptlock_ptr(struct ptdesc *ptdesc)
2935 {
2936 return ptdesc->ptl;
2937 }
2938 #else /* ALLOC_SPLIT_PTLOCKS */
ptlock_cache_init(void)2939 static inline void ptlock_cache_init(void)
2940 {
2941 }
2942
ptlock_alloc(struct ptdesc * ptdesc)2943 static inline bool ptlock_alloc(struct ptdesc *ptdesc)
2944 {
2945 return true;
2946 }
2947
ptlock_free(struct ptdesc * ptdesc)2948 static inline void ptlock_free(struct ptdesc *ptdesc)
2949 {
2950 }
2951
ptlock_ptr(struct ptdesc * ptdesc)2952 static inline spinlock_t *ptlock_ptr(struct ptdesc *ptdesc)
2953 {
2954 return &ptdesc->ptl;
2955 }
2956 #endif /* ALLOC_SPLIT_PTLOCKS */
2957
pte_lockptr(struct mm_struct * mm,pmd_t * pmd)2958 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2959 {
2960 return ptlock_ptr(page_ptdesc(pmd_page(*pmd)));
2961 }
2962
ptlock_init(struct ptdesc * ptdesc)2963 static inline bool ptlock_init(struct ptdesc *ptdesc)
2964 {
2965 /*
2966 * prep_new_page() initialize page->private (and therefore page->ptl)
2967 * with 0. Make sure nobody took it in use in between.
2968 *
2969 * It can happen if arch try to use slab for page table allocation:
2970 * slab code uses page->slab_cache, which share storage with page->ptl.
2971 */
2972 VM_BUG_ON_PAGE(*(unsigned long *)&ptdesc->ptl, ptdesc_page(ptdesc));
2973 if (!ptlock_alloc(ptdesc))
2974 return false;
2975 spin_lock_init(ptlock_ptr(ptdesc));
2976 return true;
2977 }
2978
2979 #else /* !USE_SPLIT_PTE_PTLOCKS */
2980 /*
2981 * We use mm->page_table_lock to guard all pagetable pages of the mm.
2982 */
pte_lockptr(struct mm_struct * mm,pmd_t * pmd)2983 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2984 {
2985 return &mm->page_table_lock;
2986 }
ptlock_cache_init(void)2987 static inline void ptlock_cache_init(void) {}
ptlock_init(struct ptdesc * ptdesc)2988 static inline bool ptlock_init(struct ptdesc *ptdesc) { return true; }
ptlock_free(struct ptdesc * ptdesc)2989 static inline void ptlock_free(struct ptdesc *ptdesc) {}
2990 #endif /* USE_SPLIT_PTE_PTLOCKS */
2991
pagetable_pte_ctor(struct ptdesc * ptdesc)2992 static inline bool pagetable_pte_ctor(struct ptdesc *ptdesc)
2993 {
2994 struct folio *folio = ptdesc_folio(ptdesc);
2995
2996 if (!ptlock_init(ptdesc))
2997 return false;
2998 __folio_set_pgtable(folio);
2999 lruvec_stat_add_folio(folio, NR_PAGETABLE);
3000 return true;
3001 }
3002
pagetable_pte_dtor(struct ptdesc * ptdesc)3003 static inline void pagetable_pte_dtor(struct ptdesc *ptdesc)
3004 {
3005 struct folio *folio = ptdesc_folio(ptdesc);
3006
3007 ptlock_free(ptdesc);
3008 __folio_clear_pgtable(folio);
3009 lruvec_stat_sub_folio(folio, NR_PAGETABLE);
3010 }
3011
3012 pte_t *__pte_offset_map(pmd_t *pmd, unsigned long addr, pmd_t *pmdvalp);
pte_offset_map(pmd_t * pmd,unsigned long addr)3013 static inline pte_t *pte_offset_map(pmd_t *pmd, unsigned long addr)
3014 {
3015 return __pte_offset_map(pmd, addr, NULL);
3016 }
3017
3018 pte_t *__pte_offset_map_lock(struct mm_struct *mm, pmd_t *pmd,
3019 unsigned long addr, spinlock_t **ptlp);
pte_offset_map_lock(struct mm_struct * mm,pmd_t * pmd,unsigned long addr,spinlock_t ** ptlp)3020 static inline pte_t *pte_offset_map_lock(struct mm_struct *mm, pmd_t *pmd,
3021 unsigned long addr, spinlock_t **ptlp)
3022 {
3023 pte_t *pte;
3024
3025 __cond_lock(*ptlp, pte = __pte_offset_map_lock(mm, pmd, addr, ptlp));
3026 return pte;
3027 }
3028
3029 pte_t *pte_offset_map_nolock(struct mm_struct *mm, pmd_t *pmd,
3030 unsigned long addr, spinlock_t **ptlp);
3031
3032 #define pte_unmap_unlock(pte, ptl) do { \
3033 spin_unlock(ptl); \
3034 pte_unmap(pte); \
3035 } while (0)
3036
3037 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
3038
3039 #define pte_alloc_map(mm, pmd, address) \
3040 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
3041
3042 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
3043 (pte_alloc(mm, pmd) ? \
3044 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
3045
3046 #define pte_alloc_kernel(pmd, address) \
3047 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
3048 NULL: pte_offset_kernel(pmd, address))
3049
3050 #if USE_SPLIT_PMD_PTLOCKS
3051
pmd_pgtable_page(pmd_t * pmd)3052 static inline struct page *pmd_pgtable_page(pmd_t *pmd)
3053 {
3054 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
3055 return virt_to_page((void *)((unsigned long) pmd & mask));
3056 }
3057
pmd_ptdesc(pmd_t * pmd)3058 static inline struct ptdesc *pmd_ptdesc(pmd_t *pmd)
3059 {
3060 return page_ptdesc(pmd_pgtable_page(pmd));
3061 }
3062
pmd_lockptr(struct mm_struct * mm,pmd_t * pmd)3063 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
3064 {
3065 return ptlock_ptr(pmd_ptdesc(pmd));
3066 }
3067
pmd_ptlock_init(struct ptdesc * ptdesc)3068 static inline bool pmd_ptlock_init(struct ptdesc *ptdesc)
3069 {
3070 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
3071 ptdesc->pmd_huge_pte = NULL;
3072 #endif
3073 return ptlock_init(ptdesc);
3074 }
3075
pmd_ptlock_free(struct ptdesc * ptdesc)3076 static inline void pmd_ptlock_free(struct ptdesc *ptdesc)
3077 {
3078 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
3079 VM_BUG_ON_PAGE(ptdesc->pmd_huge_pte, ptdesc_page(ptdesc));
3080 #endif
3081 ptlock_free(ptdesc);
3082 }
3083
3084 #define pmd_huge_pte(mm, pmd) (pmd_ptdesc(pmd)->pmd_huge_pte)
3085
3086 #else
3087
pmd_lockptr(struct mm_struct * mm,pmd_t * pmd)3088 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
3089 {
3090 return &mm->page_table_lock;
3091 }
3092
pmd_ptlock_init(struct ptdesc * ptdesc)3093 static inline bool pmd_ptlock_init(struct ptdesc *ptdesc) { return true; }
pmd_ptlock_free(struct ptdesc * ptdesc)3094 static inline void pmd_ptlock_free(struct ptdesc *ptdesc) {}
3095
3096 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
3097
3098 #endif
3099
pmd_lock(struct mm_struct * mm,pmd_t * pmd)3100 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
3101 {
3102 spinlock_t *ptl = pmd_lockptr(mm, pmd);
3103 spin_lock(ptl);
3104 return ptl;
3105 }
3106
pagetable_pmd_ctor(struct ptdesc * ptdesc)3107 static inline bool pagetable_pmd_ctor(struct ptdesc *ptdesc)
3108 {
3109 struct folio *folio = ptdesc_folio(ptdesc);
3110
3111 if (!pmd_ptlock_init(ptdesc))
3112 return false;
3113 __folio_set_pgtable(folio);
3114 lruvec_stat_add_folio(folio, NR_PAGETABLE);
3115 return true;
3116 }
3117
pagetable_pmd_dtor(struct ptdesc * ptdesc)3118 static inline void pagetable_pmd_dtor(struct ptdesc *ptdesc)
3119 {
3120 struct folio *folio = ptdesc_folio(ptdesc);
3121
3122 pmd_ptlock_free(ptdesc);
3123 __folio_clear_pgtable(folio);
3124 lruvec_stat_sub_folio(folio, NR_PAGETABLE);
3125 }
3126
3127 /*
3128 * No scalability reason to split PUD locks yet, but follow the same pattern
3129 * as the PMD locks to make it easier if we decide to. The VM should not be
3130 * considered ready to switch to split PUD locks yet; there may be places
3131 * which need to be converted from page_table_lock.
3132 */
pud_lockptr(struct mm_struct * mm,pud_t * pud)3133 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
3134 {
3135 return &mm->page_table_lock;
3136 }
3137
pud_lock(struct mm_struct * mm,pud_t * pud)3138 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
3139 {
3140 spinlock_t *ptl = pud_lockptr(mm, pud);
3141
3142 spin_lock(ptl);
3143 return ptl;
3144 }
3145
pagetable_pud_ctor(struct ptdesc * ptdesc)3146 static inline void pagetable_pud_ctor(struct ptdesc *ptdesc)
3147 {
3148 struct folio *folio = ptdesc_folio(ptdesc);
3149
3150 __folio_set_pgtable(folio);
3151 lruvec_stat_add_folio(folio, NR_PAGETABLE);
3152 }
3153
pagetable_pud_dtor(struct ptdesc * ptdesc)3154 static inline void pagetable_pud_dtor(struct ptdesc *ptdesc)
3155 {
3156 struct folio *folio = ptdesc_folio(ptdesc);
3157
3158 __folio_clear_pgtable(folio);
3159 lruvec_stat_sub_folio(folio, NR_PAGETABLE);
3160 }
3161
3162 extern void __init pagecache_init(void);
3163 extern void free_initmem(void);
3164
3165 /*
3166 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
3167 * into the buddy system. The freed pages will be poisoned with pattern
3168 * "poison" if it's within range [0, UCHAR_MAX].
3169 * Return pages freed into the buddy system.
3170 */
3171 extern unsigned long free_reserved_area(void *start, void *end,
3172 int poison, const char *s);
3173
3174 extern void adjust_managed_page_count(struct page *page, long count);
3175
3176 extern void reserve_bootmem_region(phys_addr_t start,
3177 phys_addr_t end, int nid);
3178
3179 /* Free the reserved page into the buddy system, so it gets managed. */
free_reserved_page(struct page * page)3180 static inline void free_reserved_page(struct page *page)
3181 {
3182 if (mem_alloc_profiling_enabled()) {
3183 union codetag_ref *ref = get_page_tag_ref(page);
3184
3185 if (ref) {
3186 set_codetag_empty(ref);
3187 put_page_tag_ref(ref);
3188 }
3189 }
3190 ClearPageReserved(page);
3191 init_page_count(page);
3192 __free_page(page);
3193 adjust_managed_page_count(page, 1);
3194 }
3195 #define free_highmem_page(page) free_reserved_page(page)
3196
mark_page_reserved(struct page * page)3197 static inline void mark_page_reserved(struct page *page)
3198 {
3199 SetPageReserved(page);
3200 adjust_managed_page_count(page, -1);
3201 }
3202
free_reserved_ptdesc(struct ptdesc * pt)3203 static inline void free_reserved_ptdesc(struct ptdesc *pt)
3204 {
3205 free_reserved_page(ptdesc_page(pt));
3206 }
3207
3208 /*
3209 * Default method to free all the __init memory into the buddy system.
3210 * The freed pages will be poisoned with pattern "poison" if it's within
3211 * range [0, UCHAR_MAX].
3212 * Return pages freed into the buddy system.
3213 */
free_initmem_default(int poison)3214 static inline unsigned long free_initmem_default(int poison)
3215 {
3216 extern char __init_begin[], __init_end[];
3217
3218 return free_reserved_area(&__init_begin, &__init_end,
3219 poison, "unused kernel image (initmem)");
3220 }
3221
get_num_physpages(void)3222 static inline unsigned long get_num_physpages(void)
3223 {
3224 int nid;
3225 unsigned long phys_pages = 0;
3226
3227 for_each_online_node(nid)
3228 phys_pages += node_present_pages(nid);
3229
3230 return phys_pages;
3231 }
3232
3233 /*
3234 * Using memblock node mappings, an architecture may initialise its
3235 * zones, allocate the backing mem_map and account for memory holes in an
3236 * architecture independent manner.
3237 *
3238 * An architecture is expected to register range of page frames backed by
3239 * physical memory with memblock_add[_node]() before calling
3240 * free_area_init() passing in the PFN each zone ends at. At a basic
3241 * usage, an architecture is expected to do something like
3242 *
3243 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
3244 * max_highmem_pfn};
3245 * for_each_valid_physical_page_range()
3246 * memblock_add_node(base, size, nid, MEMBLOCK_NONE)
3247 * free_area_init(max_zone_pfns);
3248 */
3249 void free_area_init(unsigned long *max_zone_pfn);
3250 unsigned long node_map_pfn_alignment(void);
3251 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
3252 unsigned long end_pfn);
3253 extern void get_pfn_range_for_nid(unsigned int nid,
3254 unsigned long *start_pfn, unsigned long *end_pfn);
3255
3256 #ifndef CONFIG_NUMA
early_pfn_to_nid(unsigned long pfn)3257 static inline int early_pfn_to_nid(unsigned long pfn)
3258 {
3259 return 0;
3260 }
3261 #else
3262 /* please see mm/page_alloc.c */
3263 extern int __meminit early_pfn_to_nid(unsigned long pfn);
3264 #endif
3265
3266 extern void mem_init(void);
3267 extern void __init mmap_init(void);
3268
3269 extern void __show_mem(unsigned int flags, nodemask_t *nodemask, int max_zone_idx);
show_mem(void)3270 static inline void show_mem(void)
3271 {
3272 __show_mem(0, NULL, MAX_NR_ZONES - 1);
3273 }
3274 extern long si_mem_available(void);
3275 extern void si_meminfo(struct sysinfo * val);
3276 extern void si_meminfo_node(struct sysinfo *val, int nid);
3277
3278 extern __printf(3, 4)
3279 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
3280
3281 extern void setup_per_cpu_pageset(void);
3282
3283 /* nommu.c */
3284 extern atomic_long_t mmap_pages_allocated;
3285 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
3286
3287 /* interval_tree.c */
3288 void vma_interval_tree_insert(struct vm_area_struct *node,
3289 struct rb_root_cached *root);
3290 void vma_interval_tree_insert_after(struct vm_area_struct *node,
3291 struct vm_area_struct *prev,
3292 struct rb_root_cached *root);
3293 void vma_interval_tree_remove(struct vm_area_struct *node,
3294 struct rb_root_cached *root);
3295 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
3296 unsigned long start, unsigned long last);
3297 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
3298 unsigned long start, unsigned long last);
3299
3300 #define vma_interval_tree_foreach(vma, root, start, last) \
3301 for (vma = vma_interval_tree_iter_first(root, start, last); \
3302 vma; vma = vma_interval_tree_iter_next(vma, start, last))
3303
3304 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
3305 struct rb_root_cached *root);
3306 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
3307 struct rb_root_cached *root);
3308 struct anon_vma_chain *
3309 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
3310 unsigned long start, unsigned long last);
3311 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
3312 struct anon_vma_chain *node, unsigned long start, unsigned long last);
3313 #ifdef CONFIG_DEBUG_VM_RB
3314 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
3315 #endif
3316
3317 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
3318 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
3319 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
3320
3321 /* mmap.c */
3322 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
3323 extern int vma_expand(struct vma_iterator *vmi, struct vm_area_struct *vma,
3324 unsigned long start, unsigned long end, pgoff_t pgoff,
3325 struct vm_area_struct *next);
3326 extern int vma_shrink(struct vma_iterator *vmi, struct vm_area_struct *vma,
3327 unsigned long start, unsigned long end, pgoff_t pgoff);
3328 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
3329 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
3330 extern void unlink_file_vma(struct vm_area_struct *);
3331 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
3332 unsigned long addr, unsigned long len, pgoff_t pgoff,
3333 bool *need_rmap_locks);
3334 extern void exit_mmap(struct mm_struct *);
3335 struct vm_area_struct *vma_modify(struct vma_iterator *vmi,
3336 struct vm_area_struct *prev,
3337 struct vm_area_struct *vma,
3338 unsigned long start, unsigned long end,
3339 unsigned long vm_flags,
3340 struct mempolicy *policy,
3341 struct vm_userfaultfd_ctx uffd_ctx,
3342 struct anon_vma_name *anon_name);
3343
3344 /* We are about to modify the VMA's flags. */
3345 static inline struct vm_area_struct
vma_modify_flags(struct vma_iterator * vmi,struct vm_area_struct * prev,struct vm_area_struct * vma,unsigned long start,unsigned long end,unsigned long new_flags)3346 *vma_modify_flags(struct vma_iterator *vmi,
3347 struct vm_area_struct *prev,
3348 struct vm_area_struct *vma,
3349 unsigned long start, unsigned long end,
3350 unsigned long new_flags)
3351 {
3352 return vma_modify(vmi, prev, vma, start, end, new_flags,
3353 vma_policy(vma), vma->vm_userfaultfd_ctx,
3354 anon_vma_name(vma));
3355 }
3356
3357 /* We are about to modify the VMA's flags and/or anon_name. */
3358 static inline struct vm_area_struct
vma_modify_flags_name(struct vma_iterator * vmi,struct vm_area_struct * prev,struct vm_area_struct * vma,unsigned long start,unsigned long end,unsigned long new_flags,struct anon_vma_name * new_name)3359 *vma_modify_flags_name(struct vma_iterator *vmi,
3360 struct vm_area_struct *prev,
3361 struct vm_area_struct *vma,
3362 unsigned long start,
3363 unsigned long end,
3364 unsigned long new_flags,
3365 struct anon_vma_name *new_name)
3366 {
3367 return vma_modify(vmi, prev, vma, start, end, new_flags,
3368 vma_policy(vma), vma->vm_userfaultfd_ctx, new_name);
3369 }
3370
3371 /* We are about to modify the VMA's memory policy. */
3372 static inline struct vm_area_struct
vma_modify_policy(struct vma_iterator * vmi,struct vm_area_struct * prev,struct vm_area_struct * vma,unsigned long start,unsigned long end,struct mempolicy * new_pol)3373 *vma_modify_policy(struct vma_iterator *vmi,
3374 struct vm_area_struct *prev,
3375 struct vm_area_struct *vma,
3376 unsigned long start, unsigned long end,
3377 struct mempolicy *new_pol)
3378 {
3379 return vma_modify(vmi, prev, vma, start, end, vma->vm_flags,
3380 new_pol, vma->vm_userfaultfd_ctx, anon_vma_name(vma));
3381 }
3382
3383 /* We are about to modify the VMA's flags and/or uffd context. */
3384 static inline struct vm_area_struct
vma_modify_flags_uffd(struct vma_iterator * vmi,struct vm_area_struct * prev,struct vm_area_struct * vma,unsigned long start,unsigned long end,unsigned long new_flags,struct vm_userfaultfd_ctx new_ctx)3385 *vma_modify_flags_uffd(struct vma_iterator *vmi,
3386 struct vm_area_struct *prev,
3387 struct vm_area_struct *vma,
3388 unsigned long start, unsigned long end,
3389 unsigned long new_flags,
3390 struct vm_userfaultfd_ctx new_ctx)
3391 {
3392 return vma_modify(vmi, prev, vma, start, end, new_flags,
3393 vma_policy(vma), new_ctx, anon_vma_name(vma));
3394 }
3395
check_data_rlimit(unsigned long rlim,unsigned long new,unsigned long start,unsigned long end_data,unsigned long start_data)3396 static inline int check_data_rlimit(unsigned long rlim,
3397 unsigned long new,
3398 unsigned long start,
3399 unsigned long end_data,
3400 unsigned long start_data)
3401 {
3402 if (rlim < RLIM_INFINITY) {
3403 if (((new - start) + (end_data - start_data)) > rlim)
3404 return -ENOSPC;
3405 }
3406
3407 return 0;
3408 }
3409
3410 extern int mm_take_all_locks(struct mm_struct *mm);
3411 extern void mm_drop_all_locks(struct mm_struct *mm);
3412
3413 extern int set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
3414 extern int replace_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
3415 extern struct file *get_mm_exe_file(struct mm_struct *mm);
3416 extern struct file *get_task_exe_file(struct task_struct *task);
3417
3418 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
3419 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
3420
3421 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
3422 const struct vm_special_mapping *sm);
3423 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
3424 unsigned long addr, unsigned long len,
3425 unsigned long flags,
3426 const struct vm_special_mapping *spec);
3427 /* This is an obsolete alternative to _install_special_mapping. */
3428 extern int install_special_mapping(struct mm_struct *mm,
3429 unsigned long addr, unsigned long len,
3430 unsigned long flags, struct page **pages);
3431
3432 unsigned long randomize_stack_top(unsigned long stack_top);
3433 unsigned long randomize_page(unsigned long start, unsigned long range);
3434
3435 unsigned long
3436 __get_unmapped_area(struct file *file, unsigned long addr, unsigned long len,
3437 unsigned long pgoff, unsigned long flags, vm_flags_t vm_flags);
3438
3439 static inline unsigned long
get_unmapped_area(struct file * file,unsigned long addr,unsigned long len,unsigned long pgoff,unsigned long flags)3440 get_unmapped_area(struct file *file, unsigned long addr, unsigned long len,
3441 unsigned long pgoff, unsigned long flags)
3442 {
3443 return __get_unmapped_area(file, addr, len, pgoff, flags, 0);
3444 }
3445
3446 extern unsigned long mmap_region(struct file *file, unsigned long addr,
3447 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
3448 struct list_head *uf);
3449 extern unsigned long do_mmap(struct file *file, unsigned long addr,
3450 unsigned long len, unsigned long prot, unsigned long flags,
3451 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate,
3452 struct list_head *uf);
3453 extern int do_vmi_munmap(struct vma_iterator *vmi, struct mm_struct *mm,
3454 unsigned long start, size_t len, struct list_head *uf,
3455 bool unlock);
3456 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
3457 struct list_head *uf);
3458 extern int do_madvise(struct mm_struct *mm, unsigned long start, size_t len_in, int behavior);
3459
3460 #ifdef CONFIG_MMU
3461 extern int do_vma_munmap(struct vma_iterator *vmi, struct vm_area_struct *vma,
3462 unsigned long start, unsigned long end,
3463 struct list_head *uf, bool unlock);
3464 extern int __mm_populate(unsigned long addr, unsigned long len,
3465 int ignore_errors);
mm_populate(unsigned long addr,unsigned long len)3466 static inline void mm_populate(unsigned long addr, unsigned long len)
3467 {
3468 /* Ignore errors */
3469 (void) __mm_populate(addr, len, 1);
3470 }
3471 #else
mm_populate(unsigned long addr,unsigned long len)3472 static inline void mm_populate(unsigned long addr, unsigned long len) {}
3473 #endif
3474
3475 /* This takes the mm semaphore itself */
3476 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
3477 extern int vm_munmap(unsigned long, size_t);
3478 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
3479 unsigned long, unsigned long,
3480 unsigned long, unsigned long);
3481
3482 struct vm_unmapped_area_info {
3483 #define VM_UNMAPPED_AREA_TOPDOWN 1
3484 unsigned long flags;
3485 unsigned long length;
3486 unsigned long low_limit;
3487 unsigned long high_limit;
3488 unsigned long align_mask;
3489 unsigned long align_offset;
3490 unsigned long start_gap;
3491 };
3492
3493 extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info);
3494
3495 /* truncate.c */
3496 extern void truncate_inode_pages(struct address_space *, loff_t);
3497 extern void truncate_inode_pages_range(struct address_space *,
3498 loff_t lstart, loff_t lend);
3499 extern void truncate_inode_pages_final(struct address_space *);
3500
3501 /* generic vm_area_ops exported for stackable file systems */
3502 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
3503 extern vm_fault_t filemap_map_pages(struct vm_fault *vmf,
3504 pgoff_t start_pgoff, pgoff_t end_pgoff);
3505 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
3506
3507 extern unsigned long stack_guard_gap;
3508 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
3509 int expand_stack_locked(struct vm_area_struct *vma, unsigned long address);
3510 struct vm_area_struct *expand_stack(struct mm_struct * mm, unsigned long addr);
3511
3512 /* CONFIG_STACK_GROWSUP still needs to grow downwards at some places */
3513 int expand_downwards(struct vm_area_struct *vma, unsigned long address);
3514
3515 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
3516 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
3517 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
3518 struct vm_area_struct **pprev);
3519
3520 /*
3521 * Look up the first VMA which intersects the interval [start_addr, end_addr)
3522 * NULL if none. Assume start_addr < end_addr.
3523 */
3524 struct vm_area_struct *find_vma_intersection(struct mm_struct *mm,
3525 unsigned long start_addr, unsigned long end_addr);
3526
3527 /**
3528 * vma_lookup() - Find a VMA at a specific address
3529 * @mm: The process address space.
3530 * @addr: The user address.
3531 *
3532 * Return: The vm_area_struct at the given address, %NULL otherwise.
3533 */
3534 static inline
vma_lookup(struct mm_struct * mm,unsigned long addr)3535 struct vm_area_struct *vma_lookup(struct mm_struct *mm, unsigned long addr)
3536 {
3537 return mtree_load(&mm->mm_mt, addr);
3538 }
3539
stack_guard_start_gap(struct vm_area_struct * vma)3540 static inline unsigned long stack_guard_start_gap(struct vm_area_struct *vma)
3541 {
3542 if (vma->vm_flags & VM_GROWSDOWN)
3543 return stack_guard_gap;
3544
3545 /* See reasoning around the VM_SHADOW_STACK definition */
3546 if (vma->vm_flags & VM_SHADOW_STACK)
3547 return PAGE_SIZE;
3548
3549 return 0;
3550 }
3551
vm_start_gap(struct vm_area_struct * vma)3552 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
3553 {
3554 unsigned long gap = stack_guard_start_gap(vma);
3555 unsigned long vm_start = vma->vm_start;
3556
3557 vm_start -= gap;
3558 if (vm_start > vma->vm_start)
3559 vm_start = 0;
3560 return vm_start;
3561 }
3562
vm_end_gap(struct vm_area_struct * vma)3563 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
3564 {
3565 unsigned long vm_end = vma->vm_end;
3566
3567 if (vma->vm_flags & VM_GROWSUP) {
3568 vm_end += stack_guard_gap;
3569 if (vm_end < vma->vm_end)
3570 vm_end = -PAGE_SIZE;
3571 }
3572 return vm_end;
3573 }
3574
vma_pages(struct vm_area_struct * vma)3575 static inline unsigned long vma_pages(struct vm_area_struct *vma)
3576 {
3577 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
3578 }
3579
3580 /* 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)3581 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
3582 unsigned long vm_start, unsigned long vm_end)
3583 {
3584 struct vm_area_struct *vma = vma_lookup(mm, vm_start);
3585
3586 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
3587 vma = NULL;
3588
3589 return vma;
3590 }
3591
range_in_vma(struct vm_area_struct * vma,unsigned long start,unsigned long end)3592 static inline bool range_in_vma(struct vm_area_struct *vma,
3593 unsigned long start, unsigned long end)
3594 {
3595 return (vma && vma->vm_start <= start && end <= vma->vm_end);
3596 }
3597
3598 #ifdef CONFIG_MMU
3599 pgprot_t vm_get_page_prot(unsigned long vm_flags);
3600 void vma_set_page_prot(struct vm_area_struct *vma);
3601 #else
vm_get_page_prot(unsigned long vm_flags)3602 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
3603 {
3604 return __pgprot(0);
3605 }
vma_set_page_prot(struct vm_area_struct * vma)3606 static inline void vma_set_page_prot(struct vm_area_struct *vma)
3607 {
3608 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
3609 }
3610 #endif
3611
3612 void vma_set_file(struct vm_area_struct *vma, struct file *file);
3613
3614 #ifdef CONFIG_NUMA_BALANCING
3615 unsigned long change_prot_numa(struct vm_area_struct *vma,
3616 unsigned long start, unsigned long end);
3617 #endif
3618
3619 struct vm_area_struct *find_extend_vma_locked(struct mm_struct *,
3620 unsigned long addr);
3621 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
3622 unsigned long pfn, unsigned long size, pgprot_t);
3623 int remap_pfn_range_notrack(struct vm_area_struct *vma, unsigned long addr,
3624 unsigned long pfn, unsigned long size, pgprot_t prot);
3625 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
3626 int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
3627 struct page **pages, unsigned long *num);
3628 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
3629 unsigned long num);
3630 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
3631 unsigned long num);
3632 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
3633 unsigned long pfn);
3634 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
3635 unsigned long pfn, pgprot_t pgprot);
3636 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
3637 pfn_t pfn);
3638 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
3639 unsigned long addr, pfn_t pfn);
3640 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
3641
vmf_insert_page(struct vm_area_struct * vma,unsigned long addr,struct page * page)3642 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
3643 unsigned long addr, struct page *page)
3644 {
3645 int err = vm_insert_page(vma, addr, page);
3646
3647 if (err == -ENOMEM)
3648 return VM_FAULT_OOM;
3649 if (err < 0 && err != -EBUSY)
3650 return VM_FAULT_SIGBUS;
3651
3652 return VM_FAULT_NOPAGE;
3653 }
3654
3655 #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)3656 static inline int io_remap_pfn_range(struct vm_area_struct *vma,
3657 unsigned long addr, unsigned long pfn,
3658 unsigned long size, pgprot_t prot)
3659 {
3660 return remap_pfn_range(vma, addr, pfn, size, pgprot_decrypted(prot));
3661 }
3662 #endif
3663
vmf_error(int err)3664 static inline vm_fault_t vmf_error(int err)
3665 {
3666 if (err == -ENOMEM)
3667 return VM_FAULT_OOM;
3668 else if (err == -EHWPOISON)
3669 return VM_FAULT_HWPOISON;
3670 return VM_FAULT_SIGBUS;
3671 }
3672
3673 /*
3674 * Convert errno to return value for ->page_mkwrite() calls.
3675 *
3676 * This should eventually be merged with vmf_error() above, but will need a
3677 * careful audit of all vmf_error() callers.
3678 */
vmf_fs_error(int err)3679 static inline vm_fault_t vmf_fs_error(int err)
3680 {
3681 if (err == 0)
3682 return VM_FAULT_LOCKED;
3683 if (err == -EFAULT || err == -EAGAIN)
3684 return VM_FAULT_NOPAGE;
3685 if (err == -ENOMEM)
3686 return VM_FAULT_OOM;
3687 /* -ENOSPC, -EDQUOT, -EIO ... */
3688 return VM_FAULT_SIGBUS;
3689 }
3690
3691 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
3692 unsigned int foll_flags);
3693
vm_fault_to_errno(vm_fault_t vm_fault,int foll_flags)3694 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
3695 {
3696 if (vm_fault & VM_FAULT_OOM)
3697 return -ENOMEM;
3698 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
3699 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
3700 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
3701 return -EFAULT;
3702 return 0;
3703 }
3704
3705 /*
3706 * Indicates whether GUP can follow a PROT_NONE mapped page, or whether
3707 * a (NUMA hinting) fault is required.
3708 */
gup_can_follow_protnone(struct vm_area_struct * vma,unsigned int flags)3709 static inline bool gup_can_follow_protnone(struct vm_area_struct *vma,
3710 unsigned int flags)
3711 {
3712 /*
3713 * If callers don't want to honor NUMA hinting faults, no need to
3714 * determine if we would actually have to trigger a NUMA hinting fault.
3715 */
3716 if (!(flags & FOLL_HONOR_NUMA_FAULT))
3717 return true;
3718
3719 /*
3720 * NUMA hinting faults don't apply in inaccessible (PROT_NONE) VMAs.
3721 *
3722 * Requiring a fault here even for inaccessible VMAs would mean that
3723 * FOLL_FORCE cannot make any progress, because handle_mm_fault()
3724 * refuses to process NUMA hinting faults in inaccessible VMAs.
3725 */
3726 return !vma_is_accessible(vma);
3727 }
3728
3729 typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
3730 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
3731 unsigned long size, pte_fn_t fn, void *data);
3732 extern int apply_to_existing_page_range(struct mm_struct *mm,
3733 unsigned long address, unsigned long size,
3734 pte_fn_t fn, void *data);
3735
3736 #ifdef CONFIG_PAGE_POISONING
3737 extern void __kernel_poison_pages(struct page *page, int numpages);
3738 extern void __kernel_unpoison_pages(struct page *page, int numpages);
3739 extern bool _page_poisoning_enabled_early;
3740 DECLARE_STATIC_KEY_FALSE(_page_poisoning_enabled);
page_poisoning_enabled(void)3741 static inline bool page_poisoning_enabled(void)
3742 {
3743 return _page_poisoning_enabled_early;
3744 }
3745 /*
3746 * For use in fast paths after init_mem_debugging() has run, or when a
3747 * false negative result is not harmful when called too early.
3748 */
page_poisoning_enabled_static(void)3749 static inline bool page_poisoning_enabled_static(void)
3750 {
3751 return static_branch_unlikely(&_page_poisoning_enabled);
3752 }
kernel_poison_pages(struct page * page,int numpages)3753 static inline void kernel_poison_pages(struct page *page, int numpages)
3754 {
3755 if (page_poisoning_enabled_static())
3756 __kernel_poison_pages(page, numpages);
3757 }
kernel_unpoison_pages(struct page * page,int numpages)3758 static inline void kernel_unpoison_pages(struct page *page, int numpages)
3759 {
3760 if (page_poisoning_enabled_static())
3761 __kernel_unpoison_pages(page, numpages);
3762 }
3763 #else
page_poisoning_enabled(void)3764 static inline bool page_poisoning_enabled(void) { return false; }
page_poisoning_enabled_static(void)3765 static inline bool page_poisoning_enabled_static(void) { return false; }
__kernel_poison_pages(struct page * page,int nunmpages)3766 static inline void __kernel_poison_pages(struct page *page, int nunmpages) { }
kernel_poison_pages(struct page * page,int numpages)3767 static inline void kernel_poison_pages(struct page *page, int numpages) { }
kernel_unpoison_pages(struct page * page,int numpages)3768 static inline void kernel_unpoison_pages(struct page *page, int numpages) { }
3769 #endif
3770
3771 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_ALLOC_DEFAULT_ON, init_on_alloc);
want_init_on_alloc(gfp_t flags)3772 static inline bool want_init_on_alloc(gfp_t flags)
3773 {
3774 if (static_branch_maybe(CONFIG_INIT_ON_ALLOC_DEFAULT_ON,
3775 &init_on_alloc))
3776 return true;
3777 return flags & __GFP_ZERO;
3778 }
3779
3780 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_FREE_DEFAULT_ON, init_on_free);
want_init_on_free(void)3781 static inline bool want_init_on_free(void)
3782 {
3783 return static_branch_maybe(CONFIG_INIT_ON_FREE_DEFAULT_ON,
3784 &init_on_free);
3785 }
3786
3787 extern bool _debug_pagealloc_enabled_early;
3788 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
3789
debug_pagealloc_enabled(void)3790 static inline bool debug_pagealloc_enabled(void)
3791 {
3792 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
3793 _debug_pagealloc_enabled_early;
3794 }
3795
3796 /*
3797 * For use in fast paths after mem_debugging_and_hardening_init() has run,
3798 * or when a false negative result is not harmful when called too early.
3799 */
debug_pagealloc_enabled_static(void)3800 static inline bool debug_pagealloc_enabled_static(void)
3801 {
3802 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
3803 return false;
3804
3805 return static_branch_unlikely(&_debug_pagealloc_enabled);
3806 }
3807
3808 /*
3809 * To support DEBUG_PAGEALLOC architecture must ensure that
3810 * __kernel_map_pages() never fails
3811 */
3812 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
3813 #ifdef CONFIG_DEBUG_PAGEALLOC
debug_pagealloc_map_pages(struct page * page,int numpages)3814 static inline void debug_pagealloc_map_pages(struct page *page, int numpages)
3815 {
3816 if (debug_pagealloc_enabled_static())
3817 __kernel_map_pages(page, numpages, 1);
3818 }
3819
debug_pagealloc_unmap_pages(struct page * page,int numpages)3820 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages)
3821 {
3822 if (debug_pagealloc_enabled_static())
3823 __kernel_map_pages(page, numpages, 0);
3824 }
3825
3826 extern unsigned int _debug_guardpage_minorder;
3827 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
3828
debug_guardpage_minorder(void)3829 static inline unsigned int debug_guardpage_minorder(void)
3830 {
3831 return _debug_guardpage_minorder;
3832 }
3833
debug_guardpage_enabled(void)3834 static inline bool debug_guardpage_enabled(void)
3835 {
3836 return static_branch_unlikely(&_debug_guardpage_enabled);
3837 }
3838
page_is_guard(struct page * page)3839 static inline bool page_is_guard(struct page *page)
3840 {
3841 if (!debug_guardpage_enabled())
3842 return false;
3843
3844 return PageGuard(page);
3845 }
3846
3847 bool __set_page_guard(struct zone *zone, struct page *page, unsigned int order);
set_page_guard(struct zone * zone,struct page * page,unsigned int order)3848 static inline bool set_page_guard(struct zone *zone, struct page *page,
3849 unsigned int order)
3850 {
3851 if (!debug_guardpage_enabled())
3852 return false;
3853 return __set_page_guard(zone, page, order);
3854 }
3855
3856 void __clear_page_guard(struct zone *zone, struct page *page, unsigned int order);
clear_page_guard(struct zone * zone,struct page * page,unsigned int order)3857 static inline void clear_page_guard(struct zone *zone, struct page *page,
3858 unsigned int order)
3859 {
3860 if (!debug_guardpage_enabled())
3861 return;
3862 __clear_page_guard(zone, page, order);
3863 }
3864
3865 #else /* CONFIG_DEBUG_PAGEALLOC */
debug_pagealloc_map_pages(struct page * page,int numpages)3866 static inline void debug_pagealloc_map_pages(struct page *page, int numpages) {}
debug_pagealloc_unmap_pages(struct page * page,int numpages)3867 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages) {}
debug_guardpage_minorder(void)3868 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
debug_guardpage_enabled(void)3869 static inline bool debug_guardpage_enabled(void) { return false; }
page_is_guard(struct page * page)3870 static inline bool page_is_guard(struct page *page) { return false; }
set_page_guard(struct zone * zone,struct page * page,unsigned int order)3871 static inline bool set_page_guard(struct zone *zone, struct page *page,
3872 unsigned int order) { return false; }
clear_page_guard(struct zone * zone,struct page * page,unsigned int order)3873 static inline void clear_page_guard(struct zone *zone, struct page *page,
3874 unsigned int order) {}
3875 #endif /* CONFIG_DEBUG_PAGEALLOC */
3876
3877 #ifdef __HAVE_ARCH_GATE_AREA
3878 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
3879 extern int in_gate_area_no_mm(unsigned long addr);
3880 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
3881 #else
get_gate_vma(struct mm_struct * mm)3882 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
3883 {
3884 return NULL;
3885 }
in_gate_area_no_mm(unsigned long addr)3886 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
in_gate_area(struct mm_struct * mm,unsigned long addr)3887 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
3888 {
3889 return 0;
3890 }
3891 #endif /* __HAVE_ARCH_GATE_AREA */
3892
3893 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
3894
3895 #ifdef CONFIG_SYSCTL
3896 extern int sysctl_drop_caches;
3897 int drop_caches_sysctl_handler(struct ctl_table *, int, void *, size_t *,
3898 loff_t *);
3899 #endif
3900
3901 void drop_slab(void);
3902
3903 #ifndef CONFIG_MMU
3904 #define randomize_va_space 0
3905 #else
3906 extern int randomize_va_space;
3907 #endif
3908
3909 const char * arch_vma_name(struct vm_area_struct *vma);
3910 #ifdef CONFIG_MMU
3911 void print_vma_addr(char *prefix, unsigned long rip);
3912 #else
print_vma_addr(char * prefix,unsigned long rip)3913 static inline void print_vma_addr(char *prefix, unsigned long rip)
3914 {
3915 }
3916 #endif
3917
3918 void *sparse_buffer_alloc(unsigned long size);
3919 struct page * __populate_section_memmap(unsigned long pfn,
3920 unsigned long nr_pages, int nid, struct vmem_altmap *altmap,
3921 struct dev_pagemap *pgmap);
3922 void pmd_init(void *addr);
3923 void pud_init(void *addr);
3924 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
3925 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
3926 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
3927 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
3928 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node,
3929 struct vmem_altmap *altmap, struct page *reuse);
3930 void *vmemmap_alloc_block(unsigned long size, int node);
3931 struct vmem_altmap;
3932 void *vmemmap_alloc_block_buf(unsigned long size, int node,
3933 struct vmem_altmap *altmap);
3934 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
3935 void vmemmap_set_pmd(pmd_t *pmd, void *p, int node,
3936 unsigned long addr, unsigned long next);
3937 int vmemmap_check_pmd(pmd_t *pmd, int node,
3938 unsigned long addr, unsigned long next);
3939 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
3940 int node, struct vmem_altmap *altmap);
3941 int vmemmap_populate_hugepages(unsigned long start, unsigned long end,
3942 int node, struct vmem_altmap *altmap);
3943 int vmemmap_populate(unsigned long start, unsigned long end, int node,
3944 struct vmem_altmap *altmap);
3945 void vmemmap_populate_print_last(void);
3946 #ifdef CONFIG_MEMORY_HOTPLUG
3947 void vmemmap_free(unsigned long start, unsigned long end,
3948 struct vmem_altmap *altmap);
3949 #endif
3950
3951 #ifdef CONFIG_SPARSEMEM_VMEMMAP
vmem_altmap_offset(struct vmem_altmap * altmap)3952 static inline unsigned long vmem_altmap_offset(struct vmem_altmap *altmap)
3953 {
3954 /* number of pfns from base where pfn_to_page() is valid */
3955 if (altmap)
3956 return altmap->reserve + altmap->free;
3957 return 0;
3958 }
3959
vmem_altmap_free(struct vmem_altmap * altmap,unsigned long nr_pfns)3960 static inline void vmem_altmap_free(struct vmem_altmap *altmap,
3961 unsigned long nr_pfns)
3962 {
3963 altmap->alloc -= nr_pfns;
3964 }
3965 #else
vmem_altmap_offset(struct vmem_altmap * altmap)3966 static inline unsigned long vmem_altmap_offset(struct vmem_altmap *altmap)
3967 {
3968 return 0;
3969 }
3970
vmem_altmap_free(struct vmem_altmap * altmap,unsigned long nr_pfns)3971 static inline void vmem_altmap_free(struct vmem_altmap *altmap,
3972 unsigned long nr_pfns)
3973 {
3974 }
3975 #endif
3976
3977 #define VMEMMAP_RESERVE_NR 2
3978 #ifdef CONFIG_ARCH_WANT_OPTIMIZE_DAX_VMEMMAP
__vmemmap_can_optimize(struct vmem_altmap * altmap,struct dev_pagemap * pgmap)3979 static inline bool __vmemmap_can_optimize(struct vmem_altmap *altmap,
3980 struct dev_pagemap *pgmap)
3981 {
3982 unsigned long nr_pages;
3983 unsigned long nr_vmemmap_pages;
3984
3985 if (!pgmap || !is_power_of_2(sizeof(struct page)))
3986 return false;
3987
3988 nr_pages = pgmap_vmemmap_nr(pgmap);
3989 nr_vmemmap_pages = ((nr_pages * sizeof(struct page)) >> PAGE_SHIFT);
3990 /*
3991 * For vmemmap optimization with DAX we need minimum 2 vmemmap
3992 * pages. See layout diagram in Documentation/mm/vmemmap_dedup.rst
3993 */
3994 return !altmap && (nr_vmemmap_pages > VMEMMAP_RESERVE_NR);
3995 }
3996 /*
3997 * If we don't have an architecture override, use the generic rule
3998 */
3999 #ifndef vmemmap_can_optimize
4000 #define vmemmap_can_optimize __vmemmap_can_optimize
4001 #endif
4002
4003 #else
vmemmap_can_optimize(struct vmem_altmap * altmap,struct dev_pagemap * pgmap)4004 static inline bool vmemmap_can_optimize(struct vmem_altmap *altmap,
4005 struct dev_pagemap *pgmap)
4006 {
4007 return false;
4008 }
4009 #endif
4010
4011 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
4012 unsigned long nr_pages);
4013
4014 enum mf_flags {
4015 MF_COUNT_INCREASED = 1 << 0,
4016 MF_ACTION_REQUIRED = 1 << 1,
4017 MF_MUST_KILL = 1 << 2,
4018 MF_SOFT_OFFLINE = 1 << 3,
4019 MF_UNPOISON = 1 << 4,
4020 MF_SW_SIMULATED = 1 << 5,
4021 MF_NO_RETRY = 1 << 6,
4022 MF_MEM_PRE_REMOVE = 1 << 7,
4023 };
4024 int mf_dax_kill_procs(struct address_space *mapping, pgoff_t index,
4025 unsigned long count, int mf_flags);
4026 extern int memory_failure(unsigned long pfn, int flags);
4027 extern void memory_failure_queue_kick(int cpu);
4028 extern int unpoison_memory(unsigned long pfn);
4029 extern atomic_long_t num_poisoned_pages __read_mostly;
4030 extern int soft_offline_page(unsigned long pfn, int flags);
4031 #ifdef CONFIG_MEMORY_FAILURE
4032 /*
4033 * Sysfs entries for memory failure handling statistics.
4034 */
4035 extern const struct attribute_group memory_failure_attr_group;
4036 extern void memory_failure_queue(unsigned long pfn, int flags);
4037 extern int __get_huge_page_for_hwpoison(unsigned long pfn, int flags,
4038 bool *migratable_cleared);
4039 void num_poisoned_pages_inc(unsigned long pfn);
4040 void num_poisoned_pages_sub(unsigned long pfn, long i);
4041 struct task_struct *task_early_kill(struct task_struct *tsk, int force_early);
4042 #else
memory_failure_queue(unsigned long pfn,int flags)4043 static inline void memory_failure_queue(unsigned long pfn, int flags)
4044 {
4045 }
4046
__get_huge_page_for_hwpoison(unsigned long pfn,int flags,bool * migratable_cleared)4047 static inline int __get_huge_page_for_hwpoison(unsigned long pfn, int flags,
4048 bool *migratable_cleared)
4049 {
4050 return 0;
4051 }
4052
num_poisoned_pages_inc(unsigned long pfn)4053 static inline void num_poisoned_pages_inc(unsigned long pfn)
4054 {
4055 }
4056
num_poisoned_pages_sub(unsigned long pfn,long i)4057 static inline void num_poisoned_pages_sub(unsigned long pfn, long i)
4058 {
4059 }
4060 #endif
4061
4062 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_KSM)
4063 void add_to_kill_ksm(struct task_struct *tsk, struct page *p,
4064 struct vm_area_struct *vma, struct list_head *to_kill,
4065 unsigned long ksm_addr);
4066 #endif
4067
4068 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_MEMORY_HOTPLUG)
4069 extern void memblk_nr_poison_inc(unsigned long pfn);
4070 extern void memblk_nr_poison_sub(unsigned long pfn, long i);
4071 #else
memblk_nr_poison_inc(unsigned long pfn)4072 static inline void memblk_nr_poison_inc(unsigned long pfn)
4073 {
4074 }
4075
memblk_nr_poison_sub(unsigned long pfn,long i)4076 static inline void memblk_nr_poison_sub(unsigned long pfn, long i)
4077 {
4078 }
4079 #endif
4080
4081 #ifndef arch_memory_failure
arch_memory_failure(unsigned long pfn,int flags)4082 static inline int arch_memory_failure(unsigned long pfn, int flags)
4083 {
4084 return -ENXIO;
4085 }
4086 #endif
4087
4088 #ifndef arch_is_platform_page
arch_is_platform_page(u64 paddr)4089 static inline bool arch_is_platform_page(u64 paddr)
4090 {
4091 return false;
4092 }
4093 #endif
4094
4095 /*
4096 * Error handlers for various types of pages.
4097 */
4098 enum mf_result {
4099 MF_IGNORED, /* Error: cannot be handled */
4100 MF_FAILED, /* Error: handling failed */
4101 MF_DELAYED, /* Will be handled later */
4102 MF_RECOVERED, /* Successfully recovered */
4103 };
4104
4105 enum mf_action_page_type {
4106 MF_MSG_KERNEL,
4107 MF_MSG_KERNEL_HIGH_ORDER,
4108 MF_MSG_SLAB,
4109 MF_MSG_DIFFERENT_COMPOUND,
4110 MF_MSG_HUGE,
4111 MF_MSG_FREE_HUGE,
4112 MF_MSG_UNMAP_FAILED,
4113 MF_MSG_DIRTY_SWAPCACHE,
4114 MF_MSG_CLEAN_SWAPCACHE,
4115 MF_MSG_DIRTY_MLOCKED_LRU,
4116 MF_MSG_CLEAN_MLOCKED_LRU,
4117 MF_MSG_DIRTY_UNEVICTABLE_LRU,
4118 MF_MSG_CLEAN_UNEVICTABLE_LRU,
4119 MF_MSG_DIRTY_LRU,
4120 MF_MSG_CLEAN_LRU,
4121 MF_MSG_TRUNCATED_LRU,
4122 MF_MSG_BUDDY,
4123 MF_MSG_DAX,
4124 MF_MSG_UNSPLIT_THP,
4125 MF_MSG_UNKNOWN,
4126 };
4127
4128 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
4129 extern void clear_huge_page(struct page *page,
4130 unsigned long addr_hint,
4131 unsigned int pages_per_huge_page);
4132 int copy_user_large_folio(struct folio *dst, struct folio *src,
4133 unsigned long addr_hint,
4134 struct vm_area_struct *vma);
4135 long copy_folio_from_user(struct folio *dst_folio,
4136 const void __user *usr_src,
4137 bool allow_pagefault);
4138
4139 /**
4140 * vma_is_special_huge - Are transhuge page-table entries considered special?
4141 * @vma: Pointer to the struct vm_area_struct to consider
4142 *
4143 * Whether transhuge page-table entries are considered "special" following
4144 * the definition in vm_normal_page().
4145 *
4146 * Return: true if transhuge page-table entries should be considered special,
4147 * false otherwise.
4148 */
vma_is_special_huge(const struct vm_area_struct * vma)4149 static inline bool vma_is_special_huge(const struct vm_area_struct *vma)
4150 {
4151 return vma_is_dax(vma) || (vma->vm_file &&
4152 (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP)));
4153 }
4154
4155 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
4156
4157 #if MAX_NUMNODES > 1
4158 void __init setup_nr_node_ids(void);
4159 #else
setup_nr_node_ids(void)4160 static inline void setup_nr_node_ids(void) {}
4161 #endif
4162
4163 extern int memcmp_pages(struct page *page1, struct page *page2);
4164
pages_identical(struct page * page1,struct page * page2)4165 static inline int pages_identical(struct page *page1, struct page *page2)
4166 {
4167 return !memcmp_pages(page1, page2);
4168 }
4169
4170 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
4171 unsigned long clean_record_shared_mapping_range(struct address_space *mapping,
4172 pgoff_t first_index, pgoff_t nr,
4173 pgoff_t bitmap_pgoff,
4174 unsigned long *bitmap,
4175 pgoff_t *start,
4176 pgoff_t *end);
4177
4178 unsigned long wp_shared_mapping_range(struct address_space *mapping,
4179 pgoff_t first_index, pgoff_t nr);
4180 #endif
4181
4182 extern int sysctl_nr_trim_pages;
4183
4184 #ifdef CONFIG_PRINTK
4185 void mem_dump_obj(void *object);
4186 #else
mem_dump_obj(void * object)4187 static inline void mem_dump_obj(void *object) {}
4188 #endif
4189
4190 /**
4191 * seal_check_write - Check for F_SEAL_WRITE or F_SEAL_FUTURE_WRITE flags and
4192 * handle them.
4193 * @seals: the seals to check
4194 * @vma: the vma to operate on
4195 *
4196 * Check whether F_SEAL_WRITE or F_SEAL_FUTURE_WRITE are set; if so, do proper
4197 * check/handling on the vma flags. Return 0 if check pass, or <0 for errors.
4198 */
seal_check_write(int seals,struct vm_area_struct * vma)4199 static inline int seal_check_write(int seals, struct vm_area_struct *vma)
4200 {
4201 if (seals & (F_SEAL_WRITE | F_SEAL_FUTURE_WRITE)) {
4202 /*
4203 * New PROT_WRITE and MAP_SHARED mmaps are not allowed when
4204 * write seals are active.
4205 */
4206 if ((vma->vm_flags & VM_SHARED) && (vma->vm_flags & VM_WRITE))
4207 return -EPERM;
4208
4209 /*
4210 * Since an F_SEAL_[FUTURE_]WRITE sealed memfd can be mapped as
4211 * MAP_SHARED and read-only, take care to not allow mprotect to
4212 * revert protections on such mappings. Do this only for shared
4213 * mappings. For private mappings, don't need to mask
4214 * VM_MAYWRITE as we still want them to be COW-writable.
4215 */
4216 if (vma->vm_flags & VM_SHARED)
4217 vm_flags_clear(vma, VM_MAYWRITE);
4218 }
4219
4220 return 0;
4221 }
4222
4223 #ifdef CONFIG_ANON_VMA_NAME
4224 int madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
4225 unsigned long len_in,
4226 struct anon_vma_name *anon_name);
4227 #else
4228 static inline int
madvise_set_anon_name(struct mm_struct * mm,unsigned long start,unsigned long len_in,struct anon_vma_name * anon_name)4229 madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
4230 unsigned long len_in, struct anon_vma_name *anon_name) {
4231 return 0;
4232 }
4233 #endif
4234
4235 #ifdef CONFIG_UNACCEPTED_MEMORY
4236
4237 bool range_contains_unaccepted_memory(phys_addr_t start, phys_addr_t end);
4238 void accept_memory(phys_addr_t start, phys_addr_t end);
4239
4240 #else
4241
range_contains_unaccepted_memory(phys_addr_t start,phys_addr_t end)4242 static inline bool range_contains_unaccepted_memory(phys_addr_t start,
4243 phys_addr_t end)
4244 {
4245 return false;
4246 }
4247
accept_memory(phys_addr_t start,phys_addr_t end)4248 static inline void accept_memory(phys_addr_t start, phys_addr_t end)
4249 {
4250 }
4251
4252 #endif
4253
pfn_is_unaccepted_memory(unsigned long pfn)4254 static inline bool pfn_is_unaccepted_memory(unsigned long pfn)
4255 {
4256 phys_addr_t paddr = pfn << PAGE_SHIFT;
4257
4258 return range_contains_unaccepted_memory(paddr, paddr + PAGE_SIZE);
4259 }
4260
4261 void vma_pgtable_walk_begin(struct vm_area_struct *vma);
4262 void vma_pgtable_walk_end(struct vm_area_struct *vma);
4263
4264 #endif /* _LINUX_MM_H */
4265