Line data Source code
1 : /* SPDX-License-Identifier: GPL-2.0 */
2 : #ifndef _LINUX_MM_H
3 : #define _LINUX_MM_H
4 :
5 : #include <linux/errno.h>
6 :
7 : #ifdef __KERNEL__
8 :
9 : #include <linux/mmdebug.h>
10 : #include <linux/gfp.h>
11 : #include <linux/bug.h>
12 : #include <linux/list.h>
13 : #include <linux/mmzone.h>
14 : #include <linux/rbtree.h>
15 : #include <linux/atomic.h>
16 : #include <linux/debug_locks.h>
17 : #include <linux/mm_types.h>
18 : #include <linux/mmap_lock.h>
19 : #include <linux/range.h>
20 : #include <linux/pfn.h>
21 : #include <linux/percpu-refcount.h>
22 : #include <linux/bit_spinlock.h>
23 : #include <linux/shrinker.h>
24 : #include <linux/resource.h>
25 : #include <linux/page_ext.h>
26 : #include <linux/err.h>
27 : #include <linux/page-flags.h>
28 : #include <linux/page_ref.h>
29 : #include <linux/memremap.h>
30 : #include <linux/overflow.h>
31 : #include <linux/sizes.h>
32 : #include <linux/sched.h>
33 : #include <linux/pgtable.h>
34 : #include <linux/kasan.h>
35 :
36 : struct mempolicy;
37 : struct anon_vma;
38 : struct anon_vma_chain;
39 : struct file_ra_state;
40 : struct user_struct;
41 : struct writeback_control;
42 : struct bdi_writeback;
43 : struct pt_regs;
44 :
45 : extern int sysctl_page_lock_unfairness;
46 :
47 : void init_mm_internals(void);
48 :
49 : #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
50 : extern unsigned long max_mapnr;
51 :
52 : static inline void set_max_mapnr(unsigned long limit)
53 : {
54 : max_mapnr = limit;
55 : }
56 : #else
57 : static inline void set_max_mapnr(unsigned long limit) { }
58 : #endif
59 :
60 : extern atomic_long_t _totalram_pages;
61 59518 : static inline unsigned long totalram_pages(void)
62 : {
63 59518 : return (unsigned long)atomic_long_read(&_totalram_pages);
64 : }
65 :
66 0 : static inline void totalram_pages_inc(void)
67 : {
68 0 : atomic_long_inc(&_totalram_pages);
69 0 : }
70 :
71 : static inline void totalram_pages_dec(void)
72 : {
73 : atomic_long_dec(&_totalram_pages);
74 : }
75 :
76 965 : static inline void totalram_pages_add(long count)
77 : {
78 965 : atomic_long_add(count, &_totalram_pages);
79 965 : }
80 :
81 : extern void * high_memory;
82 : extern int page_cluster;
83 :
84 : #ifdef CONFIG_SYSCTL
85 : extern int sysctl_legacy_va_layout;
86 : #else
87 : #define sysctl_legacy_va_layout 0
88 : #endif
89 :
90 : #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
91 : extern const int mmap_rnd_bits_min;
92 : extern const int mmap_rnd_bits_max;
93 : extern int mmap_rnd_bits __read_mostly;
94 : #endif
95 : #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
96 : extern const int mmap_rnd_compat_bits_min;
97 : extern const int mmap_rnd_compat_bits_max;
98 : extern int mmap_rnd_compat_bits __read_mostly;
99 : #endif
100 :
101 : #include <asm/page.h>
102 : #include <asm/processor.h>
103 :
104 : /*
105 : * Architectures that support memory tagging (assigning tags to memory regions,
106 : * embedding these tags into addresses that point to these memory regions, and
107 : * checking that the memory and the pointer tags match on memory accesses)
108 : * redefine this macro to strip tags from pointers.
109 : * It's defined as noop for arcitectures that don't support memory tagging.
110 : */
111 : #ifndef untagged_addr
112 : #define untagged_addr(addr) (addr)
113 : #endif
114 :
115 : #ifndef __pa_symbol
116 : #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
117 : #endif
118 :
119 : #ifndef page_to_virt
120 : #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
121 : #endif
122 :
123 : #ifndef lm_alias
124 : #define lm_alias(x) __va(__pa_symbol(x))
125 : #endif
126 :
127 : /*
128 : * To prevent common memory management code establishing
129 : * a zero page mapping on a read fault.
130 : * This macro should be defined within <asm/pgtable.h>.
131 : * s390 does this to prevent multiplexing of hardware bits
132 : * related to the physical page in case of virtualization.
133 : */
134 : #ifndef mm_forbids_zeropage
135 : #define mm_forbids_zeropage(X) (0)
136 : #endif
137 :
138 : /*
139 : * On some architectures it is expensive to call memset() for small sizes.
140 : * If an architecture decides to implement their own version of
141 : * mm_zero_struct_page they should wrap the defines below in a #ifndef and
142 : * define their own version of this macro in <asm/pgtable.h>
143 : */
144 : #if BITS_PER_LONG == 64
145 : /* This function must be updated when the size of struct page grows above 80
146 : * or reduces below 56. The idea that compiler optimizes out switch()
147 : * statement, and only leaves move/store instructions. Also the compiler can
148 : * combine write statments if they are both assignments and can be reordered,
149 : * this can result in several of the writes here being dropped.
150 : */
151 : #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
152 262144 : static inline void __mm_zero_struct_page(struct page *page)
153 : {
154 262144 : unsigned long *_pp = (void *)page;
155 :
156 : /* Check that struct page is either 56, 64, 72, or 80 bytes */
157 262144 : BUILD_BUG_ON(sizeof(struct page) & 7);
158 262144 : BUILD_BUG_ON(sizeof(struct page) < 56);
159 262144 : BUILD_BUG_ON(sizeof(struct page) > 80);
160 :
161 262144 : switch (sizeof(struct page)) {
162 : case 80:
163 : _pp[9] = 0;
164 : fallthrough;
165 : case 72:
166 : _pp[8] = 0;
167 262144 : fallthrough;
168 : case 64:
169 262144 : _pp[7] = 0;
170 262144 : fallthrough;
171 : case 56:
172 262144 : _pp[6] = 0;
173 262144 : _pp[5] = 0;
174 262144 : _pp[4] = 0;
175 262144 : _pp[3] = 0;
176 262144 : _pp[2] = 0;
177 262144 : _pp[1] = 0;
178 262144 : _pp[0] = 0;
179 : }
180 : }
181 : #else
182 : #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
183 : #endif
184 :
185 : /*
186 : * Default maximum number of active map areas, this limits the number of vmas
187 : * per mm struct. Users can overwrite this number by sysctl but there is a
188 : * problem.
189 : *
190 : * When a program's coredump is generated as ELF format, a section is created
191 : * per a vma. In ELF, the number of sections is represented in unsigned short.
192 : * This means the number of sections should be smaller than 65535 at coredump.
193 : * Because the kernel adds some informative sections to a image of program at
194 : * generating coredump, we need some margin. The number of extra sections is
195 : * 1-3 now and depends on arch. We use "5" as safe margin, here.
196 : *
197 : * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
198 : * not a hard limit any more. Although some userspace tools can be surprised by
199 : * that.
200 : */
201 : #define MAPCOUNT_ELF_CORE_MARGIN (5)
202 : #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
203 :
204 : extern int sysctl_max_map_count;
205 :
206 : extern unsigned long sysctl_user_reserve_kbytes;
207 : extern unsigned long sysctl_admin_reserve_kbytes;
208 :
209 : extern int sysctl_overcommit_memory;
210 : extern int sysctl_overcommit_ratio;
211 : extern unsigned long sysctl_overcommit_kbytes;
212 :
213 : int overcommit_ratio_handler(struct ctl_table *, int, void *, size_t *,
214 : loff_t *);
215 : int overcommit_kbytes_handler(struct ctl_table *, int, void *, size_t *,
216 : loff_t *);
217 : int overcommit_policy_handler(struct ctl_table *, int, void *, size_t *,
218 : loff_t *);
219 : /*
220 : * Any attempt to mark this function as static leads to build failure
221 : * when CONFIG_DEBUG_INFO_BTF is enabled because __add_to_page_cache_locked()
222 : * is referred to by BPF code. This must be visible for error injection.
223 : */
224 : int __add_to_page_cache_locked(struct page *page, struct address_space *mapping,
225 : pgoff_t index, gfp_t gfp, void **shadowp);
226 :
227 : #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
228 :
229 : /* to align the pointer to the (next) page boundary */
230 : #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
231 :
232 : /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
233 : #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
234 :
235 : #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
236 :
237 : /*
238 : * Linux kernel virtual memory manager primitives.
239 : * The idea being to have a "virtual" mm in the same way
240 : * we have a virtual fs - giving a cleaner interface to the
241 : * mm details, and allowing different kinds of memory mappings
242 : * (from shared memory to executable loading to arbitrary
243 : * mmap() functions).
244 : */
245 :
246 : struct vm_area_struct *vm_area_alloc(struct mm_struct *);
247 : struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
248 : void vm_area_free(struct vm_area_struct *);
249 :
250 : #ifndef CONFIG_MMU
251 : extern struct rb_root nommu_region_tree;
252 : extern struct rw_semaphore nommu_region_sem;
253 :
254 : extern unsigned int kobjsize(const void *objp);
255 : #endif
256 :
257 : /*
258 : * vm_flags in vm_area_struct, see mm_types.h.
259 : * When changing, update also include/trace/events/mmflags.h
260 : */
261 : #define VM_NONE 0x00000000
262 :
263 : #define VM_READ 0x00000001 /* currently active flags */
264 : #define VM_WRITE 0x00000002
265 : #define VM_EXEC 0x00000004
266 : #define VM_SHARED 0x00000008
267 :
268 : /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
269 : #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
270 : #define VM_MAYWRITE 0x00000020
271 : #define VM_MAYEXEC 0x00000040
272 : #define VM_MAYSHARE 0x00000080
273 :
274 : #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
275 : #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
276 : #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
277 : #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
278 : #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
279 :
280 : #define VM_LOCKED 0x00002000
281 : #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
282 :
283 : /* Used by sys_madvise() */
284 : #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
285 : #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
286 :
287 : #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
288 : #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
289 : #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
290 : #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
291 : #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
292 : #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
293 : #define VM_SYNC 0x00800000 /* Synchronous page faults */
294 : #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
295 : #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
296 : #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
297 :
298 : #ifdef CONFIG_MEM_SOFT_DIRTY
299 : # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
300 : #else
301 : # define VM_SOFTDIRTY 0
302 : #endif
303 :
304 : #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
305 : #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
306 : #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
307 : #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
308 :
309 : #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
310 : #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
311 : #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
312 : #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
313 : #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
314 : #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
315 : #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
316 : #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
317 : #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
318 : #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
319 : #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
320 : #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
321 :
322 : #ifdef CONFIG_ARCH_HAS_PKEYS
323 : # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
324 : # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
325 : # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
326 : # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
327 : # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
328 : #ifdef CONFIG_PPC
329 : # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
330 : #else
331 : # define VM_PKEY_BIT4 0
332 : #endif
333 : #endif /* CONFIG_ARCH_HAS_PKEYS */
334 :
335 : #if defined(CONFIG_X86)
336 : # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
337 : #elif defined(CONFIG_PPC)
338 : # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
339 : #elif defined(CONFIG_PARISC)
340 : # define VM_GROWSUP VM_ARCH_1
341 : #elif defined(CONFIG_IA64)
342 : # define VM_GROWSUP VM_ARCH_1
343 : #elif defined(CONFIG_SPARC64)
344 : # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
345 : # define VM_ARCH_CLEAR VM_SPARC_ADI
346 : #elif defined(CONFIG_ARM64)
347 : # define VM_ARM64_BTI VM_ARCH_1 /* BTI guarded page, a.k.a. GP bit */
348 : # define VM_ARCH_CLEAR VM_ARM64_BTI
349 : #elif !defined(CONFIG_MMU)
350 : # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
351 : #endif
352 :
353 : #if defined(CONFIG_ARM64_MTE)
354 : # define VM_MTE VM_HIGH_ARCH_0 /* Use Tagged memory for access control */
355 : # define VM_MTE_ALLOWED VM_HIGH_ARCH_1 /* Tagged memory permitted */
356 : #else
357 : # define VM_MTE VM_NONE
358 : # define VM_MTE_ALLOWED VM_NONE
359 : #endif
360 :
361 : #ifndef VM_GROWSUP
362 : # define VM_GROWSUP VM_NONE
363 : #endif
364 :
365 : /* Bits set in the VMA until the stack is in its final location */
366 : #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
367 :
368 : #define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0)
369 :
370 : /* Common data flag combinations */
371 : #define VM_DATA_FLAGS_TSK_EXEC (VM_READ | VM_WRITE | TASK_EXEC | \
372 : VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
373 : #define VM_DATA_FLAGS_NON_EXEC (VM_READ | VM_WRITE | VM_MAYREAD | \
374 : VM_MAYWRITE | VM_MAYEXEC)
375 : #define VM_DATA_FLAGS_EXEC (VM_READ | VM_WRITE | VM_EXEC | \
376 : VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
377 :
378 : #ifndef VM_DATA_DEFAULT_FLAGS /* arch can override this */
379 : #define VM_DATA_DEFAULT_FLAGS VM_DATA_FLAGS_EXEC
380 : #endif
381 :
382 : #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
383 : #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
384 : #endif
385 :
386 : #ifdef CONFIG_STACK_GROWSUP
387 : #define VM_STACK VM_GROWSUP
388 : #else
389 : #define VM_STACK VM_GROWSDOWN
390 : #endif
391 :
392 : #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
393 :
394 : /* VMA basic access permission flags */
395 : #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
396 :
397 :
398 : /*
399 : * Special vmas that are non-mergable, non-mlock()able.
400 : */
401 : #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
402 :
403 : /* This mask prevents VMA from being scanned with khugepaged */
404 : #define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB)
405 :
406 : /* This mask defines which mm->def_flags a process can inherit its parent */
407 : #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
408 :
409 : /* This mask is used to clear all the VMA flags used by mlock */
410 : #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
411 :
412 : /* Arch-specific flags to clear when updating VM flags on protection change */
413 : #ifndef VM_ARCH_CLEAR
414 : # define VM_ARCH_CLEAR VM_NONE
415 : #endif
416 : #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
417 :
418 : /*
419 : * mapping from the currently active vm_flags protection bits (the
420 : * low four bits) to a page protection mask..
421 : */
422 : extern pgprot_t protection_map[16];
423 :
424 : /**
425 : * Fault flag definitions.
426 : *
427 : * @FAULT_FLAG_WRITE: Fault was a write fault.
428 : * @FAULT_FLAG_MKWRITE: Fault was mkwrite of existing PTE.
429 : * @FAULT_FLAG_ALLOW_RETRY: Allow to retry the fault if blocked.
430 : * @FAULT_FLAG_RETRY_NOWAIT: Don't drop mmap_lock and wait when retrying.
431 : * @FAULT_FLAG_KILLABLE: The fault task is in SIGKILL killable region.
432 : * @FAULT_FLAG_TRIED: The fault has been tried once.
433 : * @FAULT_FLAG_USER: The fault originated in userspace.
434 : * @FAULT_FLAG_REMOTE: The fault is not for current task/mm.
435 : * @FAULT_FLAG_INSTRUCTION: The fault was during an instruction fetch.
436 : * @FAULT_FLAG_INTERRUPTIBLE: The fault can be interrupted by non-fatal signals.
437 : *
438 : * About @FAULT_FLAG_ALLOW_RETRY and @FAULT_FLAG_TRIED: we can specify
439 : * whether we would allow page faults to retry by specifying these two
440 : * fault flags correctly. Currently there can be three legal combinations:
441 : *
442 : * (a) ALLOW_RETRY and !TRIED: this means the page fault allows retry, and
443 : * this is the first try
444 : *
445 : * (b) ALLOW_RETRY and TRIED: this means the page fault allows retry, and
446 : * we've already tried at least once
447 : *
448 : * (c) !ALLOW_RETRY and !TRIED: this means the page fault does not allow retry
449 : *
450 : * The unlisted combination (!ALLOW_RETRY && TRIED) is illegal and should never
451 : * be used. Note that page faults can be allowed to retry for multiple times,
452 : * in which case we'll have an initial fault with flags (a) then later on
453 : * continuous faults with flags (b). We should always try to detect pending
454 : * signals before a retry to make sure the continuous page faults can still be
455 : * interrupted if necessary.
456 : */
457 : #define FAULT_FLAG_WRITE 0x01
458 : #define FAULT_FLAG_MKWRITE 0x02
459 : #define FAULT_FLAG_ALLOW_RETRY 0x04
460 : #define FAULT_FLAG_RETRY_NOWAIT 0x08
461 : #define FAULT_FLAG_KILLABLE 0x10
462 : #define FAULT_FLAG_TRIED 0x20
463 : #define FAULT_FLAG_USER 0x40
464 : #define FAULT_FLAG_REMOTE 0x80
465 : #define FAULT_FLAG_INSTRUCTION 0x100
466 : #define FAULT_FLAG_INTERRUPTIBLE 0x200
467 :
468 : /*
469 : * The default fault flags that should be used by most of the
470 : * arch-specific page fault handlers.
471 : */
472 : #define FAULT_FLAG_DEFAULT (FAULT_FLAG_ALLOW_RETRY | \
473 : FAULT_FLAG_KILLABLE | \
474 : FAULT_FLAG_INTERRUPTIBLE)
475 :
476 : /**
477 : * fault_flag_allow_retry_first - check ALLOW_RETRY the first time
478 : *
479 : * This is mostly used for places where we want to try to avoid taking
480 : * the mmap_lock for too long a time when waiting for another condition
481 : * to change, in which case we can try to be polite to release the
482 : * mmap_lock in the first round to avoid potential starvation of other
483 : * processes that would also want the mmap_lock.
484 : *
485 : * Return: true if the page fault allows retry and this is the first
486 : * attempt of the fault handling; false otherwise.
487 : */
488 307 : static inline bool fault_flag_allow_retry_first(unsigned int flags)
489 : {
490 307 : return (flags & FAULT_FLAG_ALLOW_RETRY) &&
491 : (!(flags & FAULT_FLAG_TRIED));
492 : }
493 :
494 : #define FAULT_FLAG_TRACE \
495 : { FAULT_FLAG_WRITE, "WRITE" }, \
496 : { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
497 : { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
498 : { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
499 : { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
500 : { FAULT_FLAG_TRIED, "TRIED" }, \
501 : { FAULT_FLAG_USER, "USER" }, \
502 : { FAULT_FLAG_REMOTE, "REMOTE" }, \
503 : { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }, \
504 : { FAULT_FLAG_INTERRUPTIBLE, "INTERRUPTIBLE" }
505 :
506 : /*
507 : * vm_fault is filled by the pagefault handler and passed to the vma's
508 : * ->fault function. The vma's ->fault is responsible for returning a bitmask
509 : * of VM_FAULT_xxx flags that give details about how the fault was handled.
510 : *
511 : * MM layer fills up gfp_mask for page allocations but fault handler might
512 : * alter it if its implementation requires a different allocation context.
513 : *
514 : * pgoff should be used in favour of virtual_address, if possible.
515 : */
516 : struct vm_fault {
517 : const struct {
518 : struct vm_area_struct *vma; /* Target VMA */
519 : gfp_t gfp_mask; /* gfp mask to be used for allocations */
520 : pgoff_t pgoff; /* Logical page offset based on vma */
521 : unsigned long address; /* Faulting virtual address */
522 : };
523 : unsigned int flags; /* FAULT_FLAG_xxx flags
524 : * XXX: should really be 'const' */
525 : pmd_t *pmd; /* Pointer to pmd entry matching
526 : * the 'address' */
527 : pud_t *pud; /* Pointer to pud entry matching
528 : * the 'address'
529 : */
530 : pte_t orig_pte; /* Value of PTE at the time of fault */
531 :
532 : struct page *cow_page; /* Page handler may use for COW fault */
533 : struct page *page; /* ->fault handlers should return a
534 : * page here, unless VM_FAULT_NOPAGE
535 : * is set (which is also implied by
536 : * VM_FAULT_ERROR).
537 : */
538 : /* These three entries are valid only while holding ptl lock */
539 : pte_t *pte; /* Pointer to pte entry matching
540 : * the 'address'. NULL if the page
541 : * table hasn't been allocated.
542 : */
543 : spinlock_t *ptl; /* Page table lock.
544 : * Protects pte page table if 'pte'
545 : * is not NULL, otherwise pmd.
546 : */
547 : pgtable_t prealloc_pte; /* Pre-allocated pte page table.
548 : * vm_ops->map_pages() sets up a page
549 : * table from atomic context.
550 : * do_fault_around() pre-allocates
551 : * page table to avoid allocation from
552 : * atomic context.
553 : */
554 : };
555 :
556 : /* page entry size for vm->huge_fault() */
557 : enum page_entry_size {
558 : PE_SIZE_PTE = 0,
559 : PE_SIZE_PMD,
560 : PE_SIZE_PUD,
561 : };
562 :
563 : /*
564 : * These are the virtual MM functions - opening of an area, closing and
565 : * unmapping it (needed to keep files on disk up-to-date etc), pointer
566 : * to the functions called when a no-page or a wp-page exception occurs.
567 : */
568 : struct vm_operations_struct {
569 : void (*open)(struct vm_area_struct * area);
570 : void (*close)(struct vm_area_struct * area);
571 : /* Called any time before splitting to check if it's allowed */
572 : int (*may_split)(struct vm_area_struct *area, unsigned long addr);
573 : int (*mremap)(struct vm_area_struct *area, unsigned long flags);
574 : /*
575 : * Called by mprotect() to make driver-specific permission
576 : * checks before mprotect() is finalised. The VMA must not
577 : * be modified. Returns 0 if eprotect() can proceed.
578 : */
579 : int (*mprotect)(struct vm_area_struct *vma, unsigned long start,
580 : unsigned long end, unsigned long newflags);
581 : vm_fault_t (*fault)(struct vm_fault *vmf);
582 : vm_fault_t (*huge_fault)(struct vm_fault *vmf,
583 : enum page_entry_size pe_size);
584 : vm_fault_t (*map_pages)(struct vm_fault *vmf,
585 : pgoff_t start_pgoff, pgoff_t end_pgoff);
586 : unsigned long (*pagesize)(struct vm_area_struct * area);
587 :
588 : /* notification that a previously read-only page is about to become
589 : * writable, if an error is returned it will cause a SIGBUS */
590 : vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
591 :
592 : /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
593 : vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
594 :
595 : /* called by access_process_vm when get_user_pages() fails, typically
596 : * for use by special VMAs. See also generic_access_phys() for a generic
597 : * implementation useful for any iomem mapping.
598 : */
599 : int (*access)(struct vm_area_struct *vma, unsigned long addr,
600 : void *buf, int len, int write);
601 :
602 : /* Called by the /proc/PID/maps code to ask the vma whether it
603 : * has a special name. Returning non-NULL will also cause this
604 : * vma to be dumped unconditionally. */
605 : const char *(*name)(struct vm_area_struct *vma);
606 :
607 : #ifdef CONFIG_NUMA
608 : /*
609 : * set_policy() op must add a reference to any non-NULL @new mempolicy
610 : * to hold the policy upon return. Caller should pass NULL @new to
611 : * remove a policy and fall back to surrounding context--i.e. do not
612 : * install a MPOL_DEFAULT policy, nor the task or system default
613 : * mempolicy.
614 : */
615 : int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
616 :
617 : /*
618 : * get_policy() op must add reference [mpol_get()] to any policy at
619 : * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
620 : * in mm/mempolicy.c will do this automatically.
621 : * get_policy() must NOT add a ref if the policy at (vma,addr) is not
622 : * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
623 : * If no [shared/vma] mempolicy exists at the addr, get_policy() op
624 : * must return NULL--i.e., do not "fallback" to task or system default
625 : * policy.
626 : */
627 : struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
628 : unsigned long addr);
629 : #endif
630 : /*
631 : * Called by vm_normal_page() for special PTEs to find the
632 : * page for @addr. This is useful if the default behavior
633 : * (using pte_page()) would not find the correct page.
634 : */
635 : struct page *(*find_special_page)(struct vm_area_struct *vma,
636 : unsigned long addr);
637 : };
638 :
639 25131 : static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
640 : {
641 25131 : static const struct vm_operations_struct dummy_vm_ops = {};
642 :
643 25131 : memset(vma, 0, sizeof(*vma));
644 25131 : vma->vm_mm = mm;
645 25131 : vma->vm_ops = &dummy_vm_ops;
646 25131 : INIT_LIST_HEAD(&vma->anon_vma_chain);
647 23146 : }
648 :
649 7926 : static inline void vma_set_anonymous(struct vm_area_struct *vma)
650 : {
651 5062 : vma->vm_ops = NULL;
652 2864 : }
653 :
654 137955 : static inline bool vma_is_anonymous(struct vm_area_struct *vma)
655 : {
656 137955 : return !vma->vm_ops;
657 : }
658 :
659 15671 : static inline bool vma_is_temporary_stack(struct vm_area_struct *vma)
660 : {
661 15671 : int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP);
662 :
663 15671 : if (!maybe_stack)
664 : return false;
665 :
666 5995 : if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) ==
667 : VM_STACK_INCOMPLETE_SETUP)
668 0 : return true;
669 :
670 : return false;
671 : }
672 :
673 349795 : static inline bool vma_is_foreign(struct vm_area_struct *vma)
674 : {
675 349795 : if (!current->mm)
676 : return true;
677 :
678 349795 : if (current->mm != vma->vm_mm)
679 : return true;
680 :
681 : return false;
682 : }
683 :
684 154635 : static inline bool vma_is_accessible(struct vm_area_struct *vma)
685 : {
686 154635 : return vma->vm_flags & VM_ACCESS_FLAGS;
687 : }
688 :
689 : #ifdef CONFIG_SHMEM
690 : /*
691 : * The vma_is_shmem is not inline because it is used only by slow
692 : * paths in userfault.
693 : */
694 : bool vma_is_shmem(struct vm_area_struct *vma);
695 : #else
696 : static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
697 : #endif
698 :
699 : int vma_is_stack_for_current(struct vm_area_struct *vma);
700 :
701 : /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
702 : #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
703 :
704 : struct mmu_gather;
705 : struct inode;
706 :
707 : #include <linux/huge_mm.h>
708 :
709 : /*
710 : * Methods to modify the page usage count.
711 : *
712 : * What counts for a page usage:
713 : * - cache mapping (page->mapping)
714 : * - private data (page->private)
715 : * - page mapped in a task's page tables, each mapping
716 : * is counted separately
717 : *
718 : * Also, many kernel routines increase the page count before a critical
719 : * routine so they can be sure the page doesn't go away from under them.
720 : */
721 :
722 : /*
723 : * Drop a ref, return true if the refcount fell to zero (the page has no users)
724 : */
725 1175622 : static inline int put_page_testzero(struct page *page)
726 : {
727 1175622 : VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
728 1175507 : return page_ref_dec_and_test(page);
729 : }
730 :
731 : /*
732 : * Try to grab a ref unless the page has a refcount of zero, return false if
733 : * that is the case.
734 : * This can be called when MMU is off so it must not access
735 : * any of the virtual mappings.
736 : */
737 0 : static inline int get_page_unless_zero(struct page *page)
738 : {
739 0 : return page_ref_add_unless(page, 1, 0);
740 : }
741 :
742 : extern int page_is_ram(unsigned long pfn);
743 :
744 : enum {
745 : REGION_INTERSECTS,
746 : REGION_DISJOINT,
747 : REGION_MIXED,
748 : };
749 :
750 : int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
751 : unsigned long desc);
752 :
753 : /* Support for virtually mapped pages */
754 : struct page *vmalloc_to_page(const void *addr);
755 : unsigned long vmalloc_to_pfn(const void *addr);
756 :
757 : /*
758 : * Determine if an address is within the vmalloc range
759 : *
760 : * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
761 : * is no special casing required.
762 : */
763 :
764 : #ifndef is_ioremap_addr
765 : #define is_ioremap_addr(x) is_vmalloc_addr(x)
766 : #endif
767 :
768 : #ifdef CONFIG_MMU
769 : extern bool is_vmalloc_addr(const void *x);
770 : extern int is_vmalloc_or_module_addr(const void *x);
771 : #else
772 : static inline bool is_vmalloc_addr(const void *x)
773 : {
774 : return false;
775 : }
776 : static inline int is_vmalloc_or_module_addr(const void *x)
777 : {
778 : return 0;
779 : }
780 : #endif
781 :
782 : extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
783 2560 : static inline void *kvmalloc(size_t size, gfp_t flags)
784 : {
785 2527 : return kvmalloc_node(size, flags, NUMA_NO_NODE);
786 : }
787 0 : static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
788 : {
789 0 : return kvmalloc_node(size, flags | __GFP_ZERO, node);
790 : }
791 143 : static inline void *kvzalloc(size_t size, gfp_t flags)
792 : {
793 143 : return kvmalloc(size, flags | __GFP_ZERO);
794 : }
795 :
796 11 : static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
797 : {
798 11 : size_t bytes;
799 :
800 11 : if (unlikely(check_mul_overflow(n, size, &bytes)))
801 : return NULL;
802 :
803 11 : return kvmalloc(bytes, flags);
804 : }
805 :
806 0 : static inline void *kvcalloc(size_t n, size_t size, gfp_t flags)
807 : {
808 0 : return kvmalloc_array(n, size, flags | __GFP_ZERO);
809 : }
810 :
811 : extern void kvfree(const void *addr);
812 : extern void kvfree_sensitive(const void *addr, size_t len);
813 :
814 13486 : static inline int head_compound_mapcount(struct page *head)
815 : {
816 26972 : return atomic_read(compound_mapcount_ptr(head)) + 1;
817 : }
818 :
819 : /*
820 : * Mapcount of compound page as a whole, does not include mapped sub-pages.
821 : *
822 : * Must be called only for compound pages or any their tail sub-pages.
823 : */
824 13486 : static inline int compound_mapcount(struct page *page)
825 : {
826 26972 : VM_BUG_ON_PAGE(!PageCompound(page), page);
827 13486 : page = compound_head(page);
828 13486 : return head_compound_mapcount(page);
829 : }
830 :
831 : /*
832 : * The atomic page->_mapcount, starts from -1: so that transitions
833 : * both from it and to it can be tracked, using atomic_inc_and_test
834 : * and atomic_add_negative(-1).
835 : */
836 262144 : static inline void page_mapcount_reset(struct page *page)
837 : {
838 262144 : atomic_set(&(page)->_mapcount, -1);
839 : }
840 :
841 : int __page_mapcount(struct page *page);
842 :
843 : /*
844 : * Mapcount of 0-order page; when compound sub-page, includes
845 : * compound_mapcount().
846 : *
847 : * Result is undefined for pages which cannot be mapped into userspace.
848 : * For example SLAB or special types of pages. See function page_has_type().
849 : * They use this place in struct page differently.
850 : */
851 896086 : static inline int page_mapcount(struct page *page)
852 : {
853 1792088 : if (unlikely(PageCompound(page)))
854 17 : return __page_mapcount(page);
855 895985 : return atomic_read(&page->_mapcount) + 1;
856 : }
857 :
858 : #ifdef CONFIG_TRANSPARENT_HUGEPAGE
859 : int total_mapcount(struct page *page);
860 : int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
861 : #else
862 : static inline int total_mapcount(struct page *page)
863 : {
864 : return page_mapcount(page);
865 : }
866 : static inline int page_trans_huge_mapcount(struct page *page,
867 : int *total_mapcount)
868 : {
869 : int mapcount = page_mapcount(page);
870 : if (total_mapcount)
871 : *total_mapcount = mapcount;
872 : return mapcount;
873 : }
874 : #endif
875 :
876 5789790 : static inline struct page *virt_to_head_page(const void *x)
877 : {
878 5789790 : struct page *page = virt_to_page(x);
879 :
880 5789790 : return compound_head(page);
881 : }
882 :
883 : void __put_page(struct page *page);
884 :
885 : void put_pages_list(struct list_head *pages);
886 :
887 : void split_page(struct page *page, unsigned int order);
888 :
889 : /*
890 : * Compound pages have a destructor function. Provide a
891 : * prototype for that function and accessor functions.
892 : * These are _only_ valid on the head of a compound page.
893 : */
894 : typedef void compound_page_dtor(struct page *);
895 :
896 : /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
897 : enum compound_dtor_id {
898 : NULL_COMPOUND_DTOR,
899 : COMPOUND_PAGE_DTOR,
900 : #ifdef CONFIG_HUGETLB_PAGE
901 : HUGETLB_PAGE_DTOR,
902 : #endif
903 : #ifdef CONFIG_TRANSPARENT_HUGEPAGE
904 : TRANSHUGE_PAGE_DTOR,
905 : #endif
906 : NR_COMPOUND_DTORS,
907 : };
908 : extern compound_page_dtor * const compound_page_dtors[NR_COMPOUND_DTORS];
909 :
910 21802 : static inline void set_compound_page_dtor(struct page *page,
911 : enum compound_dtor_id compound_dtor)
912 : {
913 21802 : VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
914 21802 : page[1].compound_dtor = compound_dtor;
915 21802 : }
916 :
917 63 : static inline void destroy_compound_page(struct page *page)
918 : {
919 63 : VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
920 63 : compound_page_dtors[page[1].compound_dtor](page);
921 63 : }
922 :
923 94170 : static inline unsigned int compound_order(struct page *page)
924 : {
925 94170 : if (!PageHead(page))
926 : return 0;
927 70562 : return page[1].compound_order;
928 : }
929 :
930 21802 : static inline bool hpage_pincount_available(struct page *page)
931 : {
932 : /*
933 : * Can the page->hpage_pinned_refcount field be used? That field is in
934 : * the 3rd page of the compound page, so the smallest (2-page) compound
935 : * pages cannot support it.
936 : */
937 21802 : page = compound_head(page);
938 43604 : return PageCompound(page) && compound_order(page) > 1;
939 : }
940 :
941 0 : static inline int head_compound_pincount(struct page *head)
942 : {
943 0 : return atomic_read(compound_pincount_ptr(head));
944 : }
945 :
946 0 : static inline int compound_pincount(struct page *page)
947 : {
948 0 : VM_BUG_ON_PAGE(!hpage_pincount_available(page), page);
949 0 : page = compound_head(page);
950 0 : return head_compound_pincount(page);
951 : }
952 :
953 21783 : static inline void set_compound_order(struct page *page, unsigned int order)
954 : {
955 21783 : page[1].compound_order = order;
956 21783 : page[1].compound_nr = 1U << order;
957 : }
958 :
959 : /* Returns the number of pages in this potentially compound page. */
960 137391 : static inline unsigned long compound_nr(struct page *page)
961 : {
962 137391 : if (!PageHead(page))
963 : return 1;
964 118861 : return page[1].compound_nr;
965 : }
966 :
967 : /* Returns the number of bytes in this potentially compound page. */
968 29545 : static inline unsigned long page_size(struct page *page)
969 : {
970 28914 : return PAGE_SIZE << compound_order(page);
971 : }
972 :
973 : /* Returns the number of bits needed for the number of bytes in a page */
974 : static inline unsigned int page_shift(struct page *page)
975 : {
976 : return PAGE_SHIFT + compound_order(page);
977 : }
978 :
979 : void free_compound_page(struct page *page);
980 :
981 : #ifdef CONFIG_MMU
982 : /*
983 : * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
984 : * servicing faults for write access. In the normal case, do always want
985 : * pte_mkwrite. But get_user_pages can cause write faults for mappings
986 : * that do not have writing enabled, when used by access_process_vm.
987 : */
988 58077 : static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
989 : {
990 58077 : if (likely(vma->vm_flags & VM_WRITE))
991 58077 : pte = pte_mkwrite(pte);
992 48454 : return pte;
993 : }
994 :
995 : vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page);
996 : void do_set_pte(struct vm_fault *vmf, struct page *page, unsigned long addr);
997 :
998 : vm_fault_t finish_fault(struct vm_fault *vmf);
999 : vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
1000 : #endif
1001 :
1002 : /*
1003 : * Multiple processes may "see" the same page. E.g. for untouched
1004 : * mappings of /dev/null, all processes see the same page full of
1005 : * zeroes, and text pages of executables and shared libraries have
1006 : * only one copy in memory, at most, normally.
1007 : *
1008 : * For the non-reserved pages, page_count(page) denotes a reference count.
1009 : * page_count() == 0 means the page is free. page->lru is then used for
1010 : * freelist management in the buddy allocator.
1011 : * page_count() > 0 means the page has been allocated.
1012 : *
1013 : * Pages are allocated by the slab allocator in order to provide memory
1014 : * to kmalloc and kmem_cache_alloc. In this case, the management of the
1015 : * page, and the fields in 'struct page' are the responsibility of mm/slab.c
1016 : * unless a particular usage is carefully commented. (the responsibility of
1017 : * freeing the kmalloc memory is the caller's, of course).
1018 : *
1019 : * A page may be used by anyone else who does a __get_free_page().
1020 : * In this case, page_count still tracks the references, and should only
1021 : * be used through the normal accessor functions. The top bits of page->flags
1022 : * and page->virtual store page management information, but all other fields
1023 : * are unused and could be used privately, carefully. The management of this
1024 : * page is the responsibility of the one who allocated it, and those who have
1025 : * subsequently been given references to it.
1026 : *
1027 : * The other pages (we may call them "pagecache pages") are completely
1028 : * managed by the Linux memory manager: I/O, buffers, swapping etc.
1029 : * The following discussion applies only to them.
1030 : *
1031 : * A pagecache page contains an opaque `private' member, which belongs to the
1032 : * page's address_space. Usually, this is the address of a circular list of
1033 : * the page's disk buffers. PG_private must be set to tell the VM to call
1034 : * into the filesystem to release these pages.
1035 : *
1036 : * A page may belong to an inode's memory mapping. In this case, page->mapping
1037 : * is the pointer to the inode, and page->index is the file offset of the page,
1038 : * in units of PAGE_SIZE.
1039 : *
1040 : * If pagecache pages are not associated with an inode, they are said to be
1041 : * anonymous pages. These may become associated with the swapcache, and in that
1042 : * case PG_swapcache is set, and page->private is an offset into the swapcache.
1043 : *
1044 : * In either case (swapcache or inode backed), the pagecache itself holds one
1045 : * reference to the page. Setting PG_private should also increment the
1046 : * refcount. The each user mapping also has a reference to the page.
1047 : *
1048 : * The pagecache pages are stored in a per-mapping radix tree, which is
1049 : * rooted at mapping->i_pages, and indexed by offset.
1050 : * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
1051 : * lists, we instead now tag pages as dirty/writeback in the radix tree.
1052 : *
1053 : * All pagecache pages may be subject to I/O:
1054 : * - inode pages may need to be read from disk,
1055 : * - inode pages which have been modified and are MAP_SHARED may need
1056 : * to be written back to the inode on disk,
1057 : * - anonymous pages (including MAP_PRIVATE file mappings) which have been
1058 : * modified may need to be swapped out to swap space and (later) to be read
1059 : * back into memory.
1060 : */
1061 :
1062 : /*
1063 : * The zone field is never updated after free_area_init_core()
1064 : * sets it, so none of the operations on it need to be atomic.
1065 : */
1066 :
1067 : /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
1068 : #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
1069 : #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
1070 : #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
1071 : #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
1072 : #define KASAN_TAG_PGOFF (LAST_CPUPID_PGOFF - KASAN_TAG_WIDTH)
1073 :
1074 : /*
1075 : * Define the bit shifts to access each section. For non-existent
1076 : * sections we define the shift as 0; that plus a 0 mask ensures
1077 : * the compiler will optimise away reference to them.
1078 : */
1079 : #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
1080 : #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
1081 : #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
1082 : #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
1083 : #define KASAN_TAG_PGSHIFT (KASAN_TAG_PGOFF * (KASAN_TAG_WIDTH != 0))
1084 :
1085 : /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
1086 : #ifdef NODE_NOT_IN_PAGE_FLAGS
1087 : #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
1088 : #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
1089 : SECTIONS_PGOFF : ZONES_PGOFF)
1090 : #else
1091 : #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
1092 : #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
1093 : NODES_PGOFF : ZONES_PGOFF)
1094 : #endif
1095 :
1096 : #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
1097 :
1098 : #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
1099 : #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
1100 : #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
1101 : #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
1102 : #define KASAN_TAG_MASK ((1UL << KASAN_TAG_WIDTH) - 1)
1103 : #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
1104 :
1105 839997 : static inline enum zone_type page_zonenum(const struct page *page)
1106 : {
1107 399213 : ASSERT_EXCLUSIVE_BITS(page->flags, ZONES_MASK << ZONES_PGSHIFT);
1108 839997 : return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
1109 : }
1110 :
1111 : #ifdef CONFIG_ZONE_DEVICE
1112 : static inline bool is_zone_device_page(const struct page *page)
1113 : {
1114 : return page_zonenum(page) == ZONE_DEVICE;
1115 : }
1116 : extern void memmap_init_zone_device(struct zone *, unsigned long,
1117 : unsigned long, struct dev_pagemap *);
1118 : #else
1119 1006627 : static inline bool is_zone_device_page(const struct page *page)
1120 : {
1121 1006627 : return false;
1122 : }
1123 : #endif
1124 :
1125 : #ifdef CONFIG_DEV_PAGEMAP_OPS
1126 : void free_devmap_managed_page(struct page *page);
1127 : DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
1128 :
1129 : static inline bool page_is_devmap_managed(struct page *page)
1130 : {
1131 : if (!static_branch_unlikely(&devmap_managed_key))
1132 : return false;
1133 : if (!is_zone_device_page(page))
1134 : return false;
1135 : switch (page->pgmap->type) {
1136 : case MEMORY_DEVICE_PRIVATE:
1137 : case MEMORY_DEVICE_FS_DAX:
1138 : return true;
1139 : default:
1140 : break;
1141 : }
1142 : return false;
1143 : }
1144 :
1145 : void put_devmap_managed_page(struct page *page);
1146 :
1147 : #else /* CONFIG_DEV_PAGEMAP_OPS */
1148 155175 : static inline bool page_is_devmap_managed(struct page *page)
1149 : {
1150 155175 : return false;
1151 : }
1152 :
1153 : static inline void put_devmap_managed_page(struct page *page)
1154 : {
1155 : }
1156 : #endif /* CONFIG_DEV_PAGEMAP_OPS */
1157 :
1158 0 : static inline bool is_device_private_page(const struct page *page)
1159 : {
1160 0 : return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
1161 : IS_ENABLED(CONFIG_DEVICE_PRIVATE) &&
1162 : is_zone_device_page(page) &&
1163 : page->pgmap->type == MEMORY_DEVICE_PRIVATE;
1164 : }
1165 :
1166 : static inline bool is_pci_p2pdma_page(const struct page *page)
1167 : {
1168 : return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
1169 : IS_ENABLED(CONFIG_PCI_P2PDMA) &&
1170 : is_zone_device_page(page) &&
1171 : page->pgmap->type == MEMORY_DEVICE_PCI_P2PDMA;
1172 : }
1173 :
1174 : /* 127: arbitrary random number, small enough to assemble well */
1175 : #define page_ref_zero_or_close_to_overflow(page) \
1176 : ((unsigned int) page_ref_count(page) + 127u <= 127u)
1177 :
1178 233371 : static inline void get_page(struct page *page)
1179 : {
1180 233371 : page = compound_head(page);
1181 : /*
1182 : * Getting a normal page or the head of a compound page
1183 : * requires to already have an elevated page->_refcount.
1184 : */
1185 233371 : VM_BUG_ON_PAGE(page_ref_zero_or_close_to_overflow(page), page);
1186 233367 : page_ref_inc(page);
1187 233401 : }
1188 :
1189 : bool __must_check try_grab_page(struct page *page, unsigned int flags);
1190 : __maybe_unused struct page *try_grab_compound_head(struct page *page, int refs,
1191 : unsigned int flags);
1192 :
1193 :
1194 7772 : static inline __must_check bool try_get_page(struct page *page)
1195 : {
1196 7772 : page = compound_head(page);
1197 7772 : if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1198 : return false;
1199 7772 : page_ref_inc(page);
1200 7772 : return true;
1201 : }
1202 :
1203 155175 : static inline void put_page(struct page *page)
1204 : {
1205 155175 : page = compound_head(page);
1206 :
1207 : /*
1208 : * For devmap managed pages we need to catch refcount transition from
1209 : * 2 to 1, when refcount reach one it means the page is free and we
1210 : * need to inform the device driver through callback. See
1211 : * include/linux/memremap.h and HMM for details.
1212 : */
1213 155175 : if (page_is_devmap_managed(page)) {
1214 : put_devmap_managed_page(page);
1215 : return;
1216 : }
1217 :
1218 155175 : if (put_page_testzero(page))
1219 33630 : __put_page(page);
1220 : }
1221 :
1222 : /*
1223 : * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
1224 : * the page's refcount so that two separate items are tracked: the original page
1225 : * reference count, and also a new count of how many pin_user_pages() calls were
1226 : * made against the page. ("gup-pinned" is another term for the latter).
1227 : *
1228 : * With this scheme, pin_user_pages() becomes special: such pages are marked as
1229 : * distinct from normal pages. As such, the unpin_user_page() call (and its
1230 : * variants) must be used in order to release gup-pinned pages.
1231 : *
1232 : * Choice of value:
1233 : *
1234 : * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
1235 : * counts with respect to pin_user_pages() and unpin_user_page() becomes
1236 : * simpler, due to the fact that adding an even power of two to the page
1237 : * refcount has the effect of using only the upper N bits, for the code that
1238 : * counts up using the bias value. This means that the lower bits are left for
1239 : * the exclusive use of the original code that increments and decrements by one
1240 : * (or at least, by much smaller values than the bias value).
1241 : *
1242 : * Of course, once the lower bits overflow into the upper bits (and this is
1243 : * OK, because subtraction recovers the original values), then visual inspection
1244 : * no longer suffices to directly view the separate counts. However, for normal
1245 : * applications that don't have huge page reference counts, this won't be an
1246 : * issue.
1247 : *
1248 : * Locking: the lockless algorithm described in page_cache_get_speculative()
1249 : * and page_cache_gup_pin_speculative() provides safe operation for
1250 : * get_user_pages and page_mkclean and other calls that race to set up page
1251 : * table entries.
1252 : */
1253 : #define GUP_PIN_COUNTING_BIAS (1U << 10)
1254 :
1255 : void unpin_user_page(struct page *page);
1256 : void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1257 : bool make_dirty);
1258 : void unpin_user_pages(struct page **pages, unsigned long npages);
1259 :
1260 : /**
1261 : * page_maybe_dma_pinned() - report if a page is pinned for DMA.
1262 : *
1263 : * This function checks if a page has been pinned via a call to
1264 : * pin_user_pages*().
1265 : *
1266 : * For non-huge pages, the return value is partially fuzzy: false is not fuzzy,
1267 : * because it means "definitely not pinned for DMA", but true means "probably
1268 : * pinned for DMA, but possibly a false positive due to having at least
1269 : * GUP_PIN_COUNTING_BIAS worth of normal page references".
1270 : *
1271 : * False positives are OK, because: a) it's unlikely for a page to get that many
1272 : * refcounts, and b) all the callers of this routine are expected to be able to
1273 : * deal gracefully with a false positive.
1274 : *
1275 : * For huge pages, the result will be exactly correct. That's because we have
1276 : * more tracking data available: the 3rd struct page in the compound page is
1277 : * used to track the pincount (instead using of the GUP_PIN_COUNTING_BIAS
1278 : * scheme).
1279 : *
1280 : * For more information, please see Documentation/core-api/pin_user_pages.rst.
1281 : *
1282 : * @page: pointer to page to be queried.
1283 : * @Return: True, if it is likely that the page has been "dma-pinned".
1284 : * False, if the page is definitely not dma-pinned.
1285 : */
1286 0 : static inline bool page_maybe_dma_pinned(struct page *page)
1287 : {
1288 0 : if (hpage_pincount_available(page))
1289 0 : return compound_pincount(page) > 0;
1290 :
1291 : /*
1292 : * page_ref_count() is signed. If that refcount overflows, then
1293 : * page_ref_count() returns a negative value, and callers will avoid
1294 : * further incrementing the refcount.
1295 : *
1296 : * Here, for that overflow case, use the signed bit to count a little
1297 : * bit higher via unsigned math, and thus still get an accurate result.
1298 : */
1299 0 : return ((unsigned int)page_ref_count(compound_head(page))) >=
1300 : GUP_PIN_COUNTING_BIAS;
1301 : }
1302 :
1303 192638 : static inline bool is_cow_mapping(vm_flags_t flags)
1304 : {
1305 110822 : return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
1306 : }
1307 :
1308 : /*
1309 : * This should most likely only be called during fork() to see whether we
1310 : * should break the cow immediately for a page on the src mm.
1311 : */
1312 81816 : static inline bool page_needs_cow_for_dma(struct vm_area_struct *vma,
1313 : struct page *page)
1314 : {
1315 81816 : if (!is_cow_mapping(vma->vm_flags))
1316 : return false;
1317 :
1318 81816 : if (!atomic_read(&vma->vm_mm->has_pinned))
1319 : return false;
1320 :
1321 0 : return page_maybe_dma_pinned(page);
1322 : }
1323 :
1324 : #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1325 : #define SECTION_IN_PAGE_FLAGS
1326 : #endif
1327 :
1328 : /*
1329 : * The identification function is mainly used by the buddy allocator for
1330 : * determining if two pages could be buddies. We are not really identifying
1331 : * the zone since we could be using the section number id if we do not have
1332 : * node id available in page flags.
1333 : * We only guarantee that it will return the same value for two combinable
1334 : * pages in a zone.
1335 : */
1336 20030 : static inline int page_zone_id(struct page *page)
1337 : {
1338 20030 : return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1339 : }
1340 :
1341 : #ifdef NODE_NOT_IN_PAGE_FLAGS
1342 : extern int page_to_nid(const struct page *page);
1343 : #else
1344 1466877 : static inline int page_to_nid(const struct page *page)
1345 : {
1346 1466877 : struct page *p = (struct page *)page;
1347 :
1348 976843 : return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1349 : }
1350 : #endif
1351 :
1352 : #ifdef CONFIG_NUMA_BALANCING
1353 : static inline int cpu_pid_to_cpupid(int cpu, int pid)
1354 : {
1355 : return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1356 : }
1357 :
1358 : static inline int cpupid_to_pid(int cpupid)
1359 : {
1360 : return cpupid & LAST__PID_MASK;
1361 : }
1362 :
1363 : static inline int cpupid_to_cpu(int cpupid)
1364 : {
1365 : return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1366 : }
1367 :
1368 : static inline int cpupid_to_nid(int cpupid)
1369 : {
1370 : return cpu_to_node(cpupid_to_cpu(cpupid));
1371 : }
1372 :
1373 : static inline bool cpupid_pid_unset(int cpupid)
1374 : {
1375 : return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1376 : }
1377 :
1378 : static inline bool cpupid_cpu_unset(int cpupid)
1379 : {
1380 : return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1381 : }
1382 :
1383 : static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1384 : {
1385 : return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1386 : }
1387 :
1388 : #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1389 : #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1390 : static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1391 : {
1392 : return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1393 : }
1394 :
1395 : static inline int page_cpupid_last(struct page *page)
1396 : {
1397 : return page->_last_cpupid;
1398 : }
1399 : static inline void page_cpupid_reset_last(struct page *page)
1400 : {
1401 : page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1402 : }
1403 : #else
1404 : static inline int page_cpupid_last(struct page *page)
1405 : {
1406 : return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1407 : }
1408 :
1409 : extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1410 :
1411 : static inline void page_cpupid_reset_last(struct page *page)
1412 : {
1413 : page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1414 : }
1415 : #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1416 : #else /* !CONFIG_NUMA_BALANCING */
1417 0 : static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1418 : {
1419 0 : return page_to_nid(page); /* XXX */
1420 : }
1421 :
1422 0 : static inline int page_cpupid_last(struct page *page)
1423 : {
1424 0 : return page_to_nid(page); /* XXX */
1425 : }
1426 :
1427 : static inline int cpupid_to_nid(int cpupid)
1428 : {
1429 : return -1;
1430 : }
1431 :
1432 : static inline int cpupid_to_pid(int cpupid)
1433 : {
1434 : return -1;
1435 : }
1436 :
1437 : static inline int cpupid_to_cpu(int cpupid)
1438 : {
1439 : return -1;
1440 : }
1441 :
1442 : static inline int cpu_pid_to_cpupid(int nid, int pid)
1443 : {
1444 : return -1;
1445 : }
1446 :
1447 : static inline bool cpupid_pid_unset(int cpupid)
1448 : {
1449 : return true;
1450 : }
1451 :
1452 262144 : static inline void page_cpupid_reset_last(struct page *page)
1453 : {
1454 262144 : }
1455 :
1456 : static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1457 : {
1458 : return false;
1459 : }
1460 : #endif /* CONFIG_NUMA_BALANCING */
1461 :
1462 : #if defined(CONFIG_KASAN_SW_TAGS) || defined(CONFIG_KASAN_HW_TAGS)
1463 :
1464 : static inline u8 page_kasan_tag(const struct page *page)
1465 : {
1466 : if (kasan_enabled())
1467 : return (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1468 : return 0xff;
1469 : }
1470 :
1471 : static inline void page_kasan_tag_set(struct page *page, u8 tag)
1472 : {
1473 : if (kasan_enabled()) {
1474 : page->flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1475 : page->flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1476 : }
1477 : }
1478 :
1479 : static inline void page_kasan_tag_reset(struct page *page)
1480 : {
1481 : if (kasan_enabled())
1482 : page_kasan_tag_set(page, 0xff);
1483 : }
1484 :
1485 : #else /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1486 :
1487 84685 : static inline u8 page_kasan_tag(const struct page *page)
1488 : {
1489 84685 : return 0xff;
1490 : }
1491 :
1492 84685 : static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1493 188335 : static inline void page_kasan_tag_reset(struct page *page) { }
1494 :
1495 : #endif /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1496 :
1497 451369 : static inline struct zone *page_zone(const struct page *page)
1498 : {
1499 451369 : return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1500 : }
1501 :
1502 719902 : static inline pg_data_t *page_pgdat(const struct page *page)
1503 : {
1504 719902 : return NODE_DATA(page_to_nid(page));
1505 : }
1506 :
1507 : #ifdef SECTION_IN_PAGE_FLAGS
1508 262144 : static inline void set_page_section(struct page *page, unsigned long section)
1509 : {
1510 262144 : page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1511 262144 : page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1512 : }
1513 :
1514 4469699 : static inline unsigned long page_to_section(const struct page *page)
1515 : {
1516 3946463 : return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1517 : }
1518 : #endif
1519 :
1520 262144 : static inline void set_page_zone(struct page *page, enum zone_type zone)
1521 : {
1522 262144 : page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1523 262144 : page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1524 : }
1525 :
1526 262144 : static inline void set_page_node(struct page *page, unsigned long node)
1527 : {
1528 262144 : page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1529 262144 : page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1530 : }
1531 :
1532 262144 : static inline void set_page_links(struct page *page, enum zone_type zone,
1533 : unsigned long node, unsigned long pfn)
1534 : {
1535 262144 : set_page_zone(page, zone);
1536 262144 : set_page_node(page, node);
1537 : #ifdef SECTION_IN_PAGE_FLAGS
1538 262144 : set_page_section(page, pfn_to_section_nr(pfn));
1539 : #endif
1540 262144 : }
1541 :
1542 : /*
1543 : * Some inline functions in vmstat.h depend on page_zone()
1544 : */
1545 : #include <linux/vmstat.h>
1546 :
1547 3009043 : static __always_inline void *lowmem_page_address(const struct page *page)
1548 : {
1549 2690153 : return page_to_virt(page);
1550 : }
1551 :
1552 : #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1553 : #define HASHED_PAGE_VIRTUAL
1554 : #endif
1555 :
1556 : #if defined(WANT_PAGE_VIRTUAL)
1557 : static inline void *page_address(const struct page *page)
1558 : {
1559 : return page->virtual;
1560 : }
1561 : static inline void set_page_address(struct page *page, void *address)
1562 : {
1563 : page->virtual = address;
1564 : }
1565 : #define page_address_init() do { } while(0)
1566 : #endif
1567 :
1568 : #if defined(HASHED_PAGE_VIRTUAL)
1569 : void *page_address(const struct page *page);
1570 : void set_page_address(struct page *page, void *virtual);
1571 : void page_address_init(void);
1572 : #endif
1573 :
1574 : #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1575 : #define page_address(page) lowmem_page_address(page)
1576 : #define set_page_address(page, address) do { } while(0)
1577 : #define page_address_init() do { } while(0)
1578 : #endif
1579 :
1580 : extern void *page_rmapping(struct page *page);
1581 : extern struct anon_vma *page_anon_vma(struct page *page);
1582 : extern struct address_space *page_mapping(struct page *page);
1583 :
1584 : extern struct address_space *__page_file_mapping(struct page *);
1585 :
1586 : static inline
1587 : struct address_space *page_file_mapping(struct page *page)
1588 : {
1589 : if (unlikely(PageSwapCache(page)))
1590 : return __page_file_mapping(page);
1591 :
1592 : return page->mapping;
1593 : }
1594 :
1595 : extern pgoff_t __page_file_index(struct page *page);
1596 :
1597 : /*
1598 : * Return the pagecache index of the passed page. Regular pagecache pages
1599 : * use ->index whereas swapcache pages use swp_offset(->private)
1600 : */
1601 5289 : static inline pgoff_t page_index(struct page *page)
1602 : {
1603 5289 : if (unlikely(PageSwapCache(page)))
1604 : return __page_file_index(page);
1605 5289 : return page->index;
1606 : }
1607 :
1608 : bool page_mapped(struct page *page);
1609 : struct address_space *page_mapping(struct page *page);
1610 : struct address_space *page_mapping_file(struct page *page);
1611 :
1612 : /*
1613 : * Return true only if the page has been allocated with
1614 : * ALLOC_NO_WATERMARKS and the low watermark was not
1615 : * met implying that the system is under some pressure.
1616 : */
1617 28835 : static inline bool page_is_pfmemalloc(const struct page *page)
1618 : {
1619 : /*
1620 : * Page index cannot be this large so this must be
1621 : * a pfmemalloc page.
1622 : */
1623 28835 : return page->index == -1UL;
1624 : }
1625 :
1626 : /*
1627 : * Only to be called by the page allocator on a freshly allocated
1628 : * page.
1629 : */
1630 0 : static inline void set_page_pfmemalloc(struct page *page)
1631 : {
1632 0 : page->index = -1UL;
1633 0 : }
1634 :
1635 193025 : static inline void clear_page_pfmemalloc(struct page *page)
1636 : {
1637 193025 : page->index = 0;
1638 193025 : }
1639 :
1640 : /*
1641 : * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1642 : */
1643 : extern void pagefault_out_of_memory(void);
1644 :
1645 : #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1646 : #define offset_in_thp(page, p) ((unsigned long)(p) & (thp_size(page) - 1))
1647 :
1648 : /*
1649 : * Flags passed to show_mem() and show_free_areas() to suppress output in
1650 : * various contexts.
1651 : */
1652 : #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1653 :
1654 : extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1655 :
1656 : #ifdef CONFIG_MMU
1657 : extern bool can_do_mlock(void);
1658 : #else
1659 : static inline bool can_do_mlock(void) { return false; }
1660 : #endif
1661 : extern int user_shm_lock(size_t, struct user_struct *);
1662 : extern void user_shm_unlock(size_t, struct user_struct *);
1663 :
1664 : /*
1665 : * Parameter block passed down to zap_pte_range in exceptional cases.
1666 : */
1667 : struct zap_details {
1668 : struct address_space *check_mapping; /* Check page->mapping if set */
1669 : pgoff_t first_index; /* Lowest page->index to unmap */
1670 : pgoff_t last_index; /* Highest page->index to unmap */
1671 : };
1672 :
1673 : struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1674 : pte_t pte);
1675 : struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1676 : pmd_t pmd);
1677 :
1678 : void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1679 : unsigned long size);
1680 : void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1681 : unsigned long size);
1682 : void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1683 : unsigned long start, unsigned long end);
1684 :
1685 : struct mmu_notifier_range;
1686 :
1687 : void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1688 : unsigned long end, unsigned long floor, unsigned long ceiling);
1689 : int
1690 : copy_page_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma);
1691 : int follow_invalidate_pte(struct mm_struct *mm, unsigned long address,
1692 : struct mmu_notifier_range *range, pte_t **ptepp,
1693 : pmd_t **pmdpp, spinlock_t **ptlp);
1694 : int follow_pte(struct mm_struct *mm, unsigned long address,
1695 : pte_t **ptepp, spinlock_t **ptlp);
1696 : int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1697 : unsigned long *pfn);
1698 : int follow_phys(struct vm_area_struct *vma, unsigned long address,
1699 : unsigned int flags, unsigned long *prot, resource_size_t *phys);
1700 : int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1701 : void *buf, int len, int write);
1702 :
1703 : extern void truncate_pagecache(struct inode *inode, loff_t new);
1704 : extern void truncate_setsize(struct inode *inode, loff_t newsize);
1705 : void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1706 : void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1707 : int truncate_inode_page(struct address_space *mapping, struct page *page);
1708 : int generic_error_remove_page(struct address_space *mapping, struct page *page);
1709 : int invalidate_inode_page(struct page *page);
1710 :
1711 : #ifdef CONFIG_MMU
1712 : extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1713 : unsigned long address, unsigned int flags,
1714 : struct pt_regs *regs);
1715 : extern int fixup_user_fault(struct mm_struct *mm,
1716 : unsigned long address, unsigned int fault_flags,
1717 : bool *unlocked);
1718 : void unmap_mapping_pages(struct address_space *mapping,
1719 : pgoff_t start, pgoff_t nr, bool even_cows);
1720 : void unmap_mapping_range(struct address_space *mapping,
1721 : loff_t const holebegin, loff_t const holelen, int even_cows);
1722 : #else
1723 : static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1724 : unsigned long address, unsigned int flags,
1725 : struct pt_regs *regs)
1726 : {
1727 : /* should never happen if there's no MMU */
1728 : BUG();
1729 : return VM_FAULT_SIGBUS;
1730 : }
1731 : static inline int fixup_user_fault(struct mm_struct *mm, unsigned long address,
1732 : unsigned int fault_flags, bool *unlocked)
1733 : {
1734 : /* should never happen if there's no MMU */
1735 : BUG();
1736 : return -EFAULT;
1737 : }
1738 : static inline void unmap_mapping_pages(struct address_space *mapping,
1739 : pgoff_t start, pgoff_t nr, bool even_cows) { }
1740 : static inline void unmap_mapping_range(struct address_space *mapping,
1741 : loff_t const holebegin, loff_t const holelen, int even_cows) { }
1742 : #endif
1743 :
1744 : static inline void unmap_shared_mapping_range(struct address_space *mapping,
1745 : loff_t const holebegin, loff_t const holelen)
1746 : {
1747 : unmap_mapping_range(mapping, holebegin, holelen, 0);
1748 : }
1749 :
1750 : extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1751 : void *buf, int len, unsigned int gup_flags);
1752 : extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1753 : void *buf, int len, unsigned int gup_flags);
1754 : extern int __access_remote_vm(struct mm_struct *mm, unsigned long addr,
1755 : void *buf, int len, unsigned int gup_flags);
1756 :
1757 : long get_user_pages_remote(struct mm_struct *mm,
1758 : unsigned long start, unsigned long nr_pages,
1759 : unsigned int gup_flags, struct page **pages,
1760 : struct vm_area_struct **vmas, int *locked);
1761 : long pin_user_pages_remote(struct mm_struct *mm,
1762 : unsigned long start, unsigned long nr_pages,
1763 : unsigned int gup_flags, struct page **pages,
1764 : struct vm_area_struct **vmas, int *locked);
1765 : long get_user_pages(unsigned long start, unsigned long nr_pages,
1766 : unsigned int gup_flags, struct page **pages,
1767 : struct vm_area_struct **vmas);
1768 : long pin_user_pages(unsigned long start, unsigned long nr_pages,
1769 : unsigned int gup_flags, struct page **pages,
1770 : struct vm_area_struct **vmas);
1771 : long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1772 : unsigned int gup_flags, struct page **pages, int *locked);
1773 : long pin_user_pages_locked(unsigned long start, unsigned long nr_pages,
1774 : unsigned int gup_flags, struct page **pages, int *locked);
1775 : long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1776 : struct page **pages, unsigned int gup_flags);
1777 : long pin_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1778 : struct page **pages, unsigned int gup_flags);
1779 :
1780 : int get_user_pages_fast(unsigned long start, int nr_pages,
1781 : unsigned int gup_flags, struct page **pages);
1782 : int pin_user_pages_fast(unsigned long start, int nr_pages,
1783 : unsigned int gup_flags, struct page **pages);
1784 :
1785 : int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
1786 : int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
1787 : struct task_struct *task, bool bypass_rlim);
1788 :
1789 : struct kvec;
1790 : int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1791 : struct page **pages);
1792 : int get_kernel_page(unsigned long start, int write, struct page **pages);
1793 : struct page *get_dump_page(unsigned long addr);
1794 :
1795 : extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1796 : extern void do_invalidatepage(struct page *page, unsigned int offset,
1797 : unsigned int length);
1798 :
1799 : void __set_page_dirty(struct page *, struct address_space *, int warn);
1800 : int __set_page_dirty_nobuffers(struct page *page);
1801 : int __set_page_dirty_no_writeback(struct page *page);
1802 : int redirty_page_for_writepage(struct writeback_control *wbc,
1803 : struct page *page);
1804 : void account_page_dirtied(struct page *page, struct address_space *mapping);
1805 : void account_page_cleaned(struct page *page, struct address_space *mapping,
1806 : struct bdi_writeback *wb);
1807 : int set_page_dirty(struct page *page);
1808 : int set_page_dirty_lock(struct page *page);
1809 : void __cancel_dirty_page(struct page *page);
1810 2477 : static inline void cancel_dirty_page(struct page *page)
1811 : {
1812 : /* Avoid atomic ops, locking, etc. when not actually needed. */
1813 4954 : if (PageDirty(page))
1814 1164 : __cancel_dirty_page(page);
1815 2477 : }
1816 : int clear_page_dirty_for_io(struct page *page);
1817 :
1818 : int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1819 :
1820 : extern unsigned long move_page_tables(struct vm_area_struct *vma,
1821 : unsigned long old_addr, struct vm_area_struct *new_vma,
1822 : unsigned long new_addr, unsigned long len,
1823 : bool need_rmap_locks);
1824 :
1825 : /*
1826 : * Flags used by change_protection(). For now we make it a bitmap so
1827 : * that we can pass in multiple flags just like parameters. However
1828 : * for now all the callers are only use one of the flags at the same
1829 : * time.
1830 : */
1831 : /* Whether we should allow dirty bit accounting */
1832 : #define MM_CP_DIRTY_ACCT (1UL << 0)
1833 : /* Whether this protection change is for NUMA hints */
1834 : #define MM_CP_PROT_NUMA (1UL << 1)
1835 : /* Whether this change is for write protecting */
1836 : #define MM_CP_UFFD_WP (1UL << 2) /* do wp */
1837 : #define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */
1838 : #define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \
1839 : MM_CP_UFFD_WP_RESOLVE)
1840 :
1841 : extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1842 : unsigned long end, pgprot_t newprot,
1843 : unsigned long cp_flags);
1844 : extern int mprotect_fixup(struct vm_area_struct *vma,
1845 : struct vm_area_struct **pprev, unsigned long start,
1846 : unsigned long end, unsigned long newflags);
1847 :
1848 : /*
1849 : * doesn't attempt to fault and will return short.
1850 : */
1851 : int get_user_pages_fast_only(unsigned long start, int nr_pages,
1852 : unsigned int gup_flags, struct page **pages);
1853 : int pin_user_pages_fast_only(unsigned long start, int nr_pages,
1854 : unsigned int gup_flags, struct page **pages);
1855 :
1856 0 : static inline bool get_user_page_fast_only(unsigned long addr,
1857 : unsigned int gup_flags, struct page **pagep)
1858 : {
1859 0 : return get_user_pages_fast_only(addr, 1, gup_flags, pagep) == 1;
1860 : }
1861 : /*
1862 : * per-process(per-mm_struct) statistics.
1863 : */
1864 88458 : static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1865 : {
1866 117941 : long val = atomic_long_read(&mm->rss_stat.count[member]);
1867 :
1868 : #ifdef SPLIT_RSS_COUNTING
1869 : /*
1870 : * counter is updated in asynchronous manner and may go to minus.
1871 : * But it's never be expected number for users.
1872 : */
1873 88653 : if (val < 0)
1874 : val = 0;
1875 : #endif
1876 88488 : return (unsigned long)val;
1877 : }
1878 :
1879 : void mm_trace_rss_stat(struct mm_struct *mm, int member, long count);
1880 :
1881 125284 : static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1882 : {
1883 125284 : long count = atomic_long_add_return(value, &mm->rss_stat.count[member]);
1884 :
1885 125287 : mm_trace_rss_stat(mm, member, count);
1886 125284 : }
1887 :
1888 0 : static inline void inc_mm_counter(struct mm_struct *mm, int member)
1889 : {
1890 0 : long count = atomic_long_inc_return(&mm->rss_stat.count[member]);
1891 :
1892 0 : mm_trace_rss_stat(mm, member, count);
1893 0 : }
1894 :
1895 0 : static inline void dec_mm_counter(struct mm_struct *mm, int member)
1896 : {
1897 0 : long count = atomic_long_dec_return(&mm->rss_stat.count[member]);
1898 :
1899 0 : mm_trace_rss_stat(mm, member, count);
1900 0 : }
1901 :
1902 : /* Optimized variant when page is already known not to be PageAnon */
1903 1548032 : static inline int mm_counter_file(struct page *page)
1904 : {
1905 3095963 : if (PageSwapBacked(page))
1906 742 : return MM_SHMEMPAGES;
1907 : return MM_FILEPAGES;
1908 : }
1909 :
1910 960159 : static inline int mm_counter(struct page *page)
1911 : {
1912 960159 : if (PageAnon(page))
1913 : return MM_ANONPAGES;
1914 750559 : return mm_counter_file(page);
1915 : }
1916 :
1917 29453 : static inline unsigned long get_mm_rss(struct mm_struct *mm)
1918 : {
1919 29453 : return get_mm_counter(mm, MM_FILEPAGES) +
1920 29491 : get_mm_counter(mm, MM_ANONPAGES) +
1921 29483 : get_mm_counter(mm, MM_SHMEMPAGES);
1922 : }
1923 :
1924 1021 : static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1925 : {
1926 2042 : return max(mm->hiwater_rss, get_mm_rss(mm));
1927 : }
1928 :
1929 0 : static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1930 : {
1931 0 : return max(mm->hiwater_vm, mm->total_vm);
1932 : }
1933 :
1934 16782 : static inline void update_hiwater_rss(struct mm_struct *mm)
1935 : {
1936 16782 : unsigned long _rss = get_mm_rss(mm);
1937 :
1938 16781 : if ((mm)->hiwater_rss < _rss)
1939 7322 : (mm)->hiwater_rss = _rss;
1940 16781 : }
1941 :
1942 16778 : static inline void update_hiwater_vm(struct mm_struct *mm)
1943 : {
1944 16778 : if (mm->hiwater_vm < mm->total_vm)
1945 6621 : mm->hiwater_vm = mm->total_vm;
1946 : }
1947 :
1948 : static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1949 : {
1950 : mm->hiwater_rss = get_mm_rss(mm);
1951 : }
1952 :
1953 1021 : static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1954 : struct mm_struct *mm)
1955 : {
1956 1021 : unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1957 :
1958 1021 : if (*maxrss < hiwater_rss)
1959 1019 : *maxrss = hiwater_rss;
1960 1021 : }
1961 :
1962 : #if defined(SPLIT_RSS_COUNTING)
1963 : void sync_mm_rss(struct mm_struct *mm);
1964 : #else
1965 : static inline void sync_mm_rss(struct mm_struct *mm)
1966 : {
1967 : }
1968 : #endif
1969 :
1970 : #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
1971 : static inline int pte_special(pte_t pte)
1972 : {
1973 : return 0;
1974 : }
1975 :
1976 : static inline pte_t pte_mkspecial(pte_t pte)
1977 : {
1978 : return pte;
1979 : }
1980 : #endif
1981 :
1982 : #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
1983 : static inline int pte_devmap(pte_t pte)
1984 : {
1985 : return 0;
1986 : }
1987 : #endif
1988 :
1989 : int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1990 :
1991 : extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1992 : spinlock_t **ptl);
1993 250 : static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1994 : spinlock_t **ptl)
1995 : {
1996 250 : pte_t *ptep;
1997 250 : __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1998 250 : return ptep;
1999 : }
2000 :
2001 : #ifdef __PAGETABLE_P4D_FOLDED
2002 : static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2003 : unsigned long address)
2004 : {
2005 : return 0;
2006 : }
2007 : #else
2008 : int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
2009 : #endif
2010 :
2011 : #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
2012 : static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2013 : unsigned long address)
2014 : {
2015 : return 0;
2016 : }
2017 : static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
2018 : static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
2019 :
2020 : #else
2021 : int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
2022 :
2023 7132 : static inline void mm_inc_nr_puds(struct mm_struct *mm)
2024 : {
2025 7132 : if (mm_pud_folded(mm))
2026 : return;
2027 7132 : atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2028 : }
2029 :
2030 7011 : static inline void mm_dec_nr_puds(struct mm_struct *mm)
2031 : {
2032 7011 : if (mm_pud_folded(mm))
2033 : return;
2034 7011 : atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2035 : }
2036 : #endif
2037 :
2038 : #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
2039 : static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
2040 : unsigned long address)
2041 : {
2042 : return 0;
2043 : }
2044 :
2045 : static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
2046 : static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
2047 :
2048 : #else
2049 : int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
2050 :
2051 8967 : static inline void mm_inc_nr_pmds(struct mm_struct *mm)
2052 : {
2053 8967 : if (mm_pmd_folded(mm))
2054 : return;
2055 8967 : atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2056 : }
2057 :
2058 8893 : static inline void mm_dec_nr_pmds(struct mm_struct *mm)
2059 : {
2060 8893 : if (mm_pmd_folded(mm))
2061 : return;
2062 8893 : atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2063 : }
2064 : #endif
2065 :
2066 : #ifdef CONFIG_MMU
2067 2019 : static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
2068 : {
2069 2019 : atomic_long_set(&mm->pgtables_bytes, 0);
2070 : }
2071 :
2072 2047 : static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2073 : {
2074 2047 : return atomic_long_read(&mm->pgtables_bytes);
2075 : }
2076 :
2077 14054 : static inline void mm_inc_nr_ptes(struct mm_struct *mm)
2078 : {
2079 14054 : atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2080 14054 : }
2081 :
2082 16942 : static inline void mm_dec_nr_ptes(struct mm_struct *mm)
2083 : {
2084 16942 : atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2085 16942 : }
2086 : #else
2087 :
2088 : static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
2089 : static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2090 : {
2091 : return 0;
2092 : }
2093 :
2094 : static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
2095 : static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
2096 : #endif
2097 :
2098 : int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
2099 : int __pte_alloc_kernel(pmd_t *pmd);
2100 :
2101 : #if defined(CONFIG_MMU)
2102 :
2103 201226 : static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2104 : unsigned long address)
2105 : {
2106 201226 : return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
2107 201226 : NULL : p4d_offset(pgd, address);
2108 : }
2109 :
2110 201223 : static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2111 : unsigned long address)
2112 : {
2113 201223 : return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
2114 208355 : NULL : pud_offset(p4d, address);
2115 : }
2116 :
2117 201159 : static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
2118 : {
2119 201159 : return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
2120 402318 : NULL: pmd_offset(pud, address);
2121 : }
2122 : #endif /* CONFIG_MMU */
2123 :
2124 : #if USE_SPLIT_PTE_PTLOCKS
2125 : #if ALLOC_SPLIT_PTLOCKS
2126 : void __init ptlock_cache_init(void);
2127 : extern bool ptlock_alloc(struct page *page);
2128 : extern void ptlock_free(struct page *page);
2129 :
2130 392355 : static inline spinlock_t *ptlock_ptr(struct page *page)
2131 : {
2132 392355 : return page->ptl;
2133 : }
2134 : #else /* ALLOC_SPLIT_PTLOCKS */
2135 : static inline void ptlock_cache_init(void)
2136 : {
2137 : }
2138 :
2139 : static inline bool ptlock_alloc(struct page *page)
2140 : {
2141 : return true;
2142 : }
2143 :
2144 : static inline void ptlock_free(struct page *page)
2145 : {
2146 : }
2147 :
2148 : static inline spinlock_t *ptlock_ptr(struct page *page)
2149 : {
2150 : return &page->ptl;
2151 : }
2152 : #endif /* ALLOC_SPLIT_PTLOCKS */
2153 :
2154 324106 : static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2155 : {
2156 324106 : return ptlock_ptr(pmd_page(*pmd));
2157 : }
2158 :
2159 54047 : static inline bool ptlock_init(struct page *page)
2160 : {
2161 : /*
2162 : * prep_new_page() initialize page->private (and therefore page->ptl)
2163 : * with 0. Make sure nobody took it in use in between.
2164 : *
2165 : * It can happen if arch try to use slab for page table allocation:
2166 : * slab code uses page->slab_cache, which share storage with page->ptl.
2167 : */
2168 54047 : VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
2169 54047 : if (!ptlock_alloc(page))
2170 : return false;
2171 54047 : spin_lock_init(ptlock_ptr(page));
2172 54047 : return true;
2173 : }
2174 :
2175 : #else /* !USE_SPLIT_PTE_PTLOCKS */
2176 : /*
2177 : * We use mm->page_table_lock to guard all pagetable pages of the mm.
2178 : */
2179 : static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2180 : {
2181 : return &mm->page_table_lock;
2182 : }
2183 : static inline void ptlock_cache_init(void) {}
2184 : static inline bool ptlock_init(struct page *page) { return true; }
2185 : static inline void ptlock_free(struct page *page) {}
2186 : #endif /* USE_SPLIT_PTE_PTLOCKS */
2187 :
2188 1 : static inline void pgtable_init(void)
2189 : {
2190 1 : ptlock_cache_init();
2191 1 : pgtable_cache_init();
2192 1 : }
2193 :
2194 45080 : static inline bool pgtable_pte_page_ctor(struct page *page)
2195 : {
2196 45080 : if (!ptlock_init(page))
2197 : return false;
2198 45080 : __SetPageTable(page);
2199 45080 : inc_lruvec_page_state(page, NR_PAGETABLE);
2200 45080 : return true;
2201 : }
2202 :
2203 41881 : static inline void pgtable_pte_page_dtor(struct page *page)
2204 : {
2205 41881 : ptlock_free(page);
2206 41881 : __ClearPageTable(page);
2207 41881 : dec_lruvec_page_state(page, NR_PAGETABLE);
2208 41881 : }
2209 :
2210 : #define pte_offset_map_lock(mm, pmd, address, ptlp) \
2211 : ({ \
2212 : spinlock_t *__ptl = pte_lockptr(mm, pmd); \
2213 : pte_t *__pte = pte_offset_map(pmd, address); \
2214 : *(ptlp) = __ptl; \
2215 : spin_lock(__ptl); \
2216 : __pte; \
2217 : })
2218 :
2219 : #define pte_unmap_unlock(pte, ptl) do { \
2220 : spin_unlock(ptl); \
2221 : pte_unmap(pte); \
2222 : } while (0)
2223 :
2224 : #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2225 :
2226 : #define pte_alloc_map(mm, pmd, address) \
2227 : (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2228 :
2229 : #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2230 : (pte_alloc(mm, pmd) ? \
2231 : NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2232 :
2233 : #define pte_alloc_kernel(pmd, address) \
2234 : ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2235 : NULL: pte_offset_kernel(pmd, address))
2236 :
2237 : #if USE_SPLIT_PMD_PTLOCKS
2238 :
2239 14274 : static struct page *pmd_to_page(pmd_t *pmd)
2240 : {
2241 14274 : unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2242 14274 : return virt_to_page((void *)((unsigned long) pmd & mask));
2243 : }
2244 :
2245 14202 : static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2246 : {
2247 14077 : return ptlock_ptr(pmd_to_page(pmd));
2248 : }
2249 :
2250 8967 : static inline bool pmd_ptlock_init(struct page *page)
2251 : {
2252 : #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2253 8967 : page->pmd_huge_pte = NULL;
2254 : #endif
2255 8967 : return ptlock_init(page);
2256 : }
2257 :
2258 22771 : static inline void pmd_ptlock_free(struct page *page)
2259 : {
2260 : #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2261 22771 : VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2262 : #endif
2263 22771 : ptlock_free(page);
2264 22771 : }
2265 :
2266 : #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2267 :
2268 : #else
2269 :
2270 : static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2271 : {
2272 : return &mm->page_table_lock;
2273 : }
2274 :
2275 : static inline bool pmd_ptlock_init(struct page *page) { return true; }
2276 : static inline void pmd_ptlock_free(struct page *page) {}
2277 :
2278 : #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2279 :
2280 : #endif
2281 :
2282 14163 : static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2283 : {
2284 14163 : spinlock_t *ptl = pmd_lockptr(mm, pmd);
2285 14163 : spin_lock(ptl);
2286 14163 : return ptl;
2287 : }
2288 :
2289 8967 : static inline bool pgtable_pmd_page_ctor(struct page *page)
2290 : {
2291 8967 : if (!pmd_ptlock_init(page))
2292 : return false;
2293 8967 : __SetPageTable(page);
2294 8967 : inc_lruvec_page_state(page, NR_PAGETABLE);
2295 8967 : return true;
2296 : }
2297 :
2298 22771 : static inline void pgtable_pmd_page_dtor(struct page *page)
2299 : {
2300 22771 : pmd_ptlock_free(page);
2301 22771 : __ClearPageTable(page);
2302 22771 : dec_lruvec_page_state(page, NR_PAGETABLE);
2303 22771 : }
2304 :
2305 : /*
2306 : * No scalability reason to split PUD locks yet, but follow the same pattern
2307 : * as the PMD locks to make it easier if we decide to. The VM should not be
2308 : * considered ready to switch to split PUD locks yet; there may be places
2309 : * which need to be converted from page_table_lock.
2310 : */
2311 10971 : static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2312 : {
2313 8967 : return &mm->page_table_lock;
2314 : }
2315 :
2316 10971 : static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2317 : {
2318 10971 : spinlock_t *ptl = pud_lockptr(mm, pud);
2319 :
2320 10971 : spin_lock(ptl);
2321 10971 : return ptl;
2322 : }
2323 :
2324 : extern void __init pagecache_init(void);
2325 : extern void __init free_area_init_memoryless_node(int nid);
2326 : extern void free_initmem(void);
2327 :
2328 : /*
2329 : * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2330 : * into the buddy system. The freed pages will be poisoned with pattern
2331 : * "poison" if it's within range [0, UCHAR_MAX].
2332 : * Return pages freed into the buddy system.
2333 : */
2334 : extern unsigned long free_reserved_area(void *start, void *end,
2335 : int poison, const char *s);
2336 :
2337 : extern void adjust_managed_page_count(struct page *page, long count);
2338 : extern void mem_init_print_info(const char *str);
2339 :
2340 : extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2341 :
2342 : /* Free the reserved page into the buddy system, so it gets managed. */
2343 964 : static inline void free_reserved_page(struct page *page)
2344 : {
2345 964 : ClearPageReserved(page);
2346 964 : init_page_count(page);
2347 964 : __free_page(page);
2348 964 : adjust_managed_page_count(page, 1);
2349 964 : }
2350 : #define free_highmem_page(page) free_reserved_page(page)
2351 :
2352 : static inline void mark_page_reserved(struct page *page)
2353 : {
2354 : SetPageReserved(page);
2355 : adjust_managed_page_count(page, -1);
2356 : }
2357 :
2358 : /*
2359 : * Default method to free all the __init memory into the buddy system.
2360 : * The freed pages will be poisoned with pattern "poison" if it's within
2361 : * range [0, UCHAR_MAX].
2362 : * Return pages freed into the buddy system.
2363 : */
2364 0 : static inline unsigned long free_initmem_default(int poison)
2365 : {
2366 0 : extern char __init_begin[], __init_end[];
2367 :
2368 0 : return free_reserved_area(&__init_begin, &__init_end,
2369 : poison, "unused kernel");
2370 : }
2371 :
2372 1 : static inline unsigned long get_num_physpages(void)
2373 : {
2374 1 : int nid;
2375 1 : unsigned long phys_pages = 0;
2376 :
2377 2 : for_each_online_node(nid)
2378 1 : phys_pages += node_present_pages(nid);
2379 :
2380 1 : return phys_pages;
2381 : }
2382 :
2383 : /*
2384 : * Using memblock node mappings, an architecture may initialise its
2385 : * zones, allocate the backing mem_map and account for memory holes in an
2386 : * architecture independent manner.
2387 : *
2388 : * An architecture is expected to register range of page frames backed by
2389 : * physical memory with memblock_add[_node]() before calling
2390 : * free_area_init() passing in the PFN each zone ends at. At a basic
2391 : * usage, an architecture is expected to do something like
2392 : *
2393 : * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2394 : * max_highmem_pfn};
2395 : * for_each_valid_physical_page_range()
2396 : * memblock_add_node(base, size, nid)
2397 : * free_area_init(max_zone_pfns);
2398 : */
2399 : void free_area_init(unsigned long *max_zone_pfn);
2400 : unsigned long node_map_pfn_alignment(void);
2401 : unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2402 : unsigned long end_pfn);
2403 : extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2404 : unsigned long end_pfn);
2405 : extern void get_pfn_range_for_nid(unsigned int nid,
2406 : unsigned long *start_pfn, unsigned long *end_pfn);
2407 : extern unsigned long find_min_pfn_with_active_regions(void);
2408 :
2409 : #ifndef CONFIG_NEED_MULTIPLE_NODES
2410 : static inline int early_pfn_to_nid(unsigned long pfn)
2411 : {
2412 : return 0;
2413 : }
2414 : #else
2415 : /* please see mm/page_alloc.c */
2416 : extern int __meminit early_pfn_to_nid(unsigned long pfn);
2417 : #endif
2418 :
2419 : extern void set_dma_reserve(unsigned long new_dma_reserve);
2420 : extern void memmap_init_range(unsigned long, int, unsigned long,
2421 : unsigned long, unsigned long, enum meminit_context,
2422 : struct vmem_altmap *, int migratetype);
2423 : extern void memmap_init_zone(struct zone *zone);
2424 : extern void setup_per_zone_wmarks(void);
2425 : extern int __meminit init_per_zone_wmark_min(void);
2426 : extern void mem_init(void);
2427 : extern void __init mmap_init(void);
2428 : extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2429 : extern long si_mem_available(void);
2430 : extern void si_meminfo(struct sysinfo * val);
2431 : extern void si_meminfo_node(struct sysinfo *val, int nid);
2432 : #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2433 : extern unsigned long arch_reserved_kernel_pages(void);
2434 : #endif
2435 :
2436 : extern __printf(3, 4)
2437 : void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2438 :
2439 : extern void setup_per_cpu_pageset(void);
2440 :
2441 : /* page_alloc.c */
2442 : extern int min_free_kbytes;
2443 : extern int watermark_boost_factor;
2444 : extern int watermark_scale_factor;
2445 : extern bool arch_has_descending_max_zone_pfns(void);
2446 :
2447 : /* nommu.c */
2448 : extern atomic_long_t mmap_pages_allocated;
2449 : extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2450 :
2451 : /* interval_tree.c */
2452 : void vma_interval_tree_insert(struct vm_area_struct *node,
2453 : struct rb_root_cached *root);
2454 : void vma_interval_tree_insert_after(struct vm_area_struct *node,
2455 : struct vm_area_struct *prev,
2456 : struct rb_root_cached *root);
2457 : void vma_interval_tree_remove(struct vm_area_struct *node,
2458 : struct rb_root_cached *root);
2459 : struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2460 : unsigned long start, unsigned long last);
2461 : struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2462 : unsigned long start, unsigned long last);
2463 :
2464 : #define vma_interval_tree_foreach(vma, root, start, last) \
2465 : for (vma = vma_interval_tree_iter_first(root, start, last); \
2466 : vma; vma = vma_interval_tree_iter_next(vma, start, last))
2467 :
2468 : void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2469 : struct rb_root_cached *root);
2470 : void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2471 : struct rb_root_cached *root);
2472 : struct anon_vma_chain *
2473 : anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2474 : unsigned long start, unsigned long last);
2475 : struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2476 : struct anon_vma_chain *node, unsigned long start, unsigned long last);
2477 : #ifdef CONFIG_DEBUG_VM_RB
2478 : void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2479 : #endif
2480 :
2481 : #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2482 : for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2483 : avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2484 :
2485 : /* mmap.c */
2486 : extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2487 : extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2488 : unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2489 : struct vm_area_struct *expand);
2490 27455 : static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2491 : unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2492 : {
2493 27455 : return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2494 : }
2495 : extern struct vm_area_struct *vma_merge(struct mm_struct *,
2496 : struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2497 : unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2498 : struct mempolicy *, struct vm_userfaultfd_ctx);
2499 : extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2500 : extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2501 : unsigned long addr, int new_below);
2502 : extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2503 : unsigned long addr, int new_below);
2504 : extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2505 : extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2506 : struct rb_node **, struct rb_node *);
2507 : extern void unlink_file_vma(struct vm_area_struct *);
2508 : extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2509 : unsigned long addr, unsigned long len, pgoff_t pgoff,
2510 : bool *need_rmap_locks);
2511 : extern void exit_mmap(struct mm_struct *);
2512 :
2513 1360 : static inline int check_data_rlimit(unsigned long rlim,
2514 : unsigned long new,
2515 : unsigned long start,
2516 : unsigned long end_data,
2517 : unsigned long start_data)
2518 : {
2519 1360 : if (rlim < RLIM_INFINITY) {
2520 0 : if (((new - start) + (end_data - start_data)) > rlim)
2521 0 : return -ENOSPC;
2522 : }
2523 :
2524 : return 0;
2525 : }
2526 :
2527 : extern int mm_take_all_locks(struct mm_struct *mm);
2528 : extern void mm_drop_all_locks(struct mm_struct *mm);
2529 :
2530 : extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2531 : extern struct file *get_mm_exe_file(struct mm_struct *mm);
2532 : extern struct file *get_task_exe_file(struct task_struct *task);
2533 :
2534 : extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2535 : extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2536 :
2537 : extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2538 : const struct vm_special_mapping *sm);
2539 : extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2540 : unsigned long addr, unsigned long len,
2541 : unsigned long flags,
2542 : const struct vm_special_mapping *spec);
2543 : /* This is an obsolete alternative to _install_special_mapping. */
2544 : extern int install_special_mapping(struct mm_struct *mm,
2545 : unsigned long addr, unsigned long len,
2546 : unsigned long flags, struct page **pages);
2547 :
2548 : unsigned long randomize_stack_top(unsigned long stack_top);
2549 :
2550 : extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2551 :
2552 : extern unsigned long mmap_region(struct file *file, unsigned long addr,
2553 : unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2554 : struct list_head *uf);
2555 : extern unsigned long do_mmap(struct file *file, unsigned long addr,
2556 : unsigned long len, unsigned long prot, unsigned long flags,
2557 : unsigned long pgoff, unsigned long *populate, struct list_head *uf);
2558 : extern int __do_munmap(struct mm_struct *, unsigned long, size_t,
2559 : struct list_head *uf, bool downgrade);
2560 : extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2561 : struct list_head *uf);
2562 : extern int do_madvise(struct mm_struct *mm, unsigned long start, size_t len_in, int behavior);
2563 :
2564 : #ifdef CONFIG_MMU
2565 : extern int __mm_populate(unsigned long addr, unsigned long len,
2566 : int ignore_errors);
2567 10 : static inline void mm_populate(unsigned long addr, unsigned long len)
2568 : {
2569 : /* Ignore errors */
2570 10 : (void) __mm_populate(addr, len, 1);
2571 10 : }
2572 : #else
2573 : static inline void mm_populate(unsigned long addr, unsigned long len) {}
2574 : #endif
2575 :
2576 : /* These take the mm semaphore themselves */
2577 : extern int __must_check vm_brk(unsigned long, unsigned long);
2578 : extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2579 : extern int vm_munmap(unsigned long, size_t);
2580 : extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2581 : unsigned long, unsigned long,
2582 : unsigned long, unsigned long);
2583 :
2584 : struct vm_unmapped_area_info {
2585 : #define VM_UNMAPPED_AREA_TOPDOWN 1
2586 : unsigned long flags;
2587 : unsigned long length;
2588 : unsigned long low_limit;
2589 : unsigned long high_limit;
2590 : unsigned long align_mask;
2591 : unsigned long align_offset;
2592 : };
2593 :
2594 : extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info);
2595 :
2596 : /* truncate.c */
2597 : extern void truncate_inode_pages(struct address_space *, loff_t);
2598 : extern void truncate_inode_pages_range(struct address_space *,
2599 : loff_t lstart, loff_t lend);
2600 : extern void truncate_inode_pages_final(struct address_space *);
2601 :
2602 : /* generic vm_area_ops exported for stackable file systems */
2603 : extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2604 : extern vm_fault_t filemap_map_pages(struct vm_fault *vmf,
2605 : pgoff_t start_pgoff, pgoff_t end_pgoff);
2606 : extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2607 :
2608 : /* mm/page-writeback.c */
2609 : int __must_check write_one_page(struct page *page);
2610 : void task_dirty_inc(struct task_struct *tsk);
2611 :
2612 : extern unsigned long stack_guard_gap;
2613 : /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2614 : extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2615 :
2616 : /* CONFIG_STACK_GROWSUP still needs to grow downwards at some places */
2617 : extern int expand_downwards(struct vm_area_struct *vma,
2618 : unsigned long address);
2619 : #if VM_GROWSUP
2620 : extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2621 : #else
2622 : #define expand_upwards(vma, address) (0)
2623 : #endif
2624 :
2625 : /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2626 : extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2627 : extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2628 : struct vm_area_struct **pprev);
2629 :
2630 : /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2631 : NULL if none. Assume start_addr < end_addr. */
2632 0 : static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2633 : {
2634 0 : struct vm_area_struct * vma = find_vma(mm,start_addr);
2635 :
2636 0 : if (vma && end_addr <= vma->vm_start)
2637 0 : vma = NULL;
2638 0 : return vma;
2639 : }
2640 :
2641 511373 : static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2642 : {
2643 511373 : unsigned long vm_start = vma->vm_start;
2644 :
2645 511373 : if (vma->vm_flags & VM_GROWSDOWN) {
2646 35559 : vm_start -= stack_guard_gap;
2647 35559 : if (vm_start > vma->vm_start)
2648 0 : vm_start = 0;
2649 : }
2650 511373 : return vm_start;
2651 : }
2652 :
2653 585041 : static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2654 : {
2655 585041 : unsigned long vm_end = vma->vm_end;
2656 :
2657 585041 : if (vma->vm_flags & VM_GROWSUP) {
2658 : vm_end += stack_guard_gap;
2659 : if (vm_end < vma->vm_end)
2660 : vm_end = -PAGE_SIZE;
2661 : }
2662 585041 : return vm_end;
2663 : }
2664 :
2665 724201 : static inline unsigned long vma_pages(struct vm_area_struct *vma)
2666 : {
2667 724201 : return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2668 : }
2669 :
2670 : /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2671 0 : static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2672 : unsigned long vm_start, unsigned long vm_end)
2673 : {
2674 0 : struct vm_area_struct *vma = find_vma(mm, vm_start);
2675 :
2676 0 : if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2677 0 : vma = NULL;
2678 :
2679 0 : return vma;
2680 : }
2681 :
2682 0 : static inline bool range_in_vma(struct vm_area_struct *vma,
2683 : unsigned long start, unsigned long end)
2684 : {
2685 0 : return (vma && vma->vm_start <= start && end <= vma->vm_end);
2686 : }
2687 :
2688 : #ifdef CONFIG_MMU
2689 : pgprot_t vm_get_page_prot(unsigned long vm_flags);
2690 : void vma_set_page_prot(struct vm_area_struct *vma);
2691 : #else
2692 : static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2693 : {
2694 : return __pgprot(0);
2695 : }
2696 : static inline void vma_set_page_prot(struct vm_area_struct *vma)
2697 : {
2698 : vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2699 : }
2700 : #endif
2701 :
2702 : void vma_set_file(struct vm_area_struct *vma, struct file *file);
2703 :
2704 : #ifdef CONFIG_NUMA_BALANCING
2705 : unsigned long change_prot_numa(struct vm_area_struct *vma,
2706 : unsigned long start, unsigned long end);
2707 : #endif
2708 :
2709 : struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2710 : int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2711 : unsigned long pfn, unsigned long size, pgprot_t);
2712 : int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2713 : int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
2714 : struct page **pages, unsigned long *num);
2715 : int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
2716 : unsigned long num);
2717 : int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
2718 : unsigned long num);
2719 : vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2720 : unsigned long pfn);
2721 : vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2722 : unsigned long pfn, pgprot_t pgprot);
2723 : vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2724 : pfn_t pfn);
2725 : vm_fault_t vmf_insert_mixed_prot(struct vm_area_struct *vma, unsigned long addr,
2726 : pfn_t pfn, pgprot_t pgprot);
2727 : vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
2728 : unsigned long addr, pfn_t pfn);
2729 : int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2730 :
2731 : static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2732 : unsigned long addr, struct page *page)
2733 : {
2734 : int err = vm_insert_page(vma, addr, page);
2735 :
2736 : if (err == -ENOMEM)
2737 : return VM_FAULT_OOM;
2738 : if (err < 0 && err != -EBUSY)
2739 : return VM_FAULT_SIGBUS;
2740 :
2741 : return VM_FAULT_NOPAGE;
2742 : }
2743 :
2744 : #ifndef io_remap_pfn_range
2745 0 : static inline int io_remap_pfn_range(struct vm_area_struct *vma,
2746 : unsigned long addr, unsigned long pfn,
2747 : unsigned long size, pgprot_t prot)
2748 : {
2749 0 : return remap_pfn_range(vma, addr, pfn, size, pgprot_decrypted(prot));
2750 : }
2751 : #endif
2752 :
2753 0 : static inline vm_fault_t vmf_error(int err)
2754 : {
2755 0 : if (err == -ENOMEM)
2756 0 : return VM_FAULT_OOM;
2757 : return VM_FAULT_SIGBUS;
2758 : }
2759 :
2760 : struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
2761 : unsigned int foll_flags);
2762 :
2763 : #define FOLL_WRITE 0x01 /* check pte is writable */
2764 : #define FOLL_TOUCH 0x02 /* mark page accessed */
2765 : #define FOLL_GET 0x04 /* do get_page on page */
2766 : #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2767 : #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2768 : #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2769 : * and return without waiting upon it */
2770 : #define FOLL_POPULATE 0x40 /* fault in page */
2771 : #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2772 : #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2773 : #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2774 : #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2775 : #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2776 : #define FOLL_MLOCK 0x1000 /* lock present pages */
2777 : #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2778 : #define FOLL_COW 0x4000 /* internal GUP flag */
2779 : #define FOLL_ANON 0x8000 /* don't do file mappings */
2780 : #define FOLL_LONGTERM 0x10000 /* mapping lifetime is indefinite: see below */
2781 : #define FOLL_SPLIT_PMD 0x20000 /* split huge pmd before returning */
2782 : #define FOLL_PIN 0x40000 /* pages must be released via unpin_user_page */
2783 : #define FOLL_FAST_ONLY 0x80000 /* gup_fast: prevent fall-back to slow gup */
2784 :
2785 : /*
2786 : * FOLL_PIN and FOLL_LONGTERM may be used in various combinations with each
2787 : * other. Here is what they mean, and how to use them:
2788 : *
2789 : * FOLL_LONGTERM indicates that the page will be held for an indefinite time
2790 : * period _often_ under userspace control. This is in contrast to
2791 : * iov_iter_get_pages(), whose usages are transient.
2792 : *
2793 : * FIXME: For pages which are part of a filesystem, mappings are subject to the
2794 : * lifetime enforced by the filesystem and we need guarantees that longterm
2795 : * users like RDMA and V4L2 only establish mappings which coordinate usage with
2796 : * the filesystem. Ideas for this coordination include revoking the longterm
2797 : * pin, delaying writeback, bounce buffer page writeback, etc. As FS DAX was
2798 : * added after the problem with filesystems was found FS DAX VMAs are
2799 : * specifically failed. Filesystem pages are still subject to bugs and use of
2800 : * FOLL_LONGTERM should be avoided on those pages.
2801 : *
2802 : * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call.
2803 : * Currently only get_user_pages() and get_user_pages_fast() support this flag
2804 : * and calls to get_user_pages_[un]locked are specifically not allowed. This
2805 : * is due to an incompatibility with the FS DAX check and
2806 : * FAULT_FLAG_ALLOW_RETRY.
2807 : *
2808 : * In the CMA case: long term pins in a CMA region would unnecessarily fragment
2809 : * that region. And so, CMA attempts to migrate the page before pinning, when
2810 : * FOLL_LONGTERM is specified.
2811 : *
2812 : * FOLL_PIN indicates that a special kind of tracking (not just page->_refcount,
2813 : * but an additional pin counting system) will be invoked. This is intended for
2814 : * anything that gets a page reference and then touches page data (for example,
2815 : * Direct IO). This lets the filesystem know that some non-file-system entity is
2816 : * potentially changing the pages' data. In contrast to FOLL_GET (whose pages
2817 : * are released via put_page()), FOLL_PIN pages must be released, ultimately, by
2818 : * a call to unpin_user_page().
2819 : *
2820 : * FOLL_PIN is similar to FOLL_GET: both of these pin pages. They use different
2821 : * and separate refcounting mechanisms, however, and that means that each has
2822 : * its own acquire and release mechanisms:
2823 : *
2824 : * FOLL_GET: get_user_pages*() to acquire, and put_page() to release.
2825 : *
2826 : * FOLL_PIN: pin_user_pages*() to acquire, and unpin_user_pages to release.
2827 : *
2828 : * FOLL_PIN and FOLL_GET are mutually exclusive for a given function call.
2829 : * (The underlying pages may experience both FOLL_GET-based and FOLL_PIN-based
2830 : * calls applied to them, and that's perfectly OK. This is a constraint on the
2831 : * callers, not on the pages.)
2832 : *
2833 : * FOLL_PIN should be set internally by the pin_user_pages*() APIs, never
2834 : * directly by the caller. That's in order to help avoid mismatches when
2835 : * releasing pages: get_user_pages*() pages must be released via put_page(),
2836 : * while pin_user_pages*() pages must be released via unpin_user_page().
2837 : *
2838 : * Please see Documentation/core-api/pin_user_pages.rst for more information.
2839 : */
2840 :
2841 0 : static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
2842 : {
2843 0 : if (vm_fault & VM_FAULT_OOM)
2844 : return -ENOMEM;
2845 0 : if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2846 0 : return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2847 0 : if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2848 0 : return -EFAULT;
2849 : return 0;
2850 : }
2851 :
2852 : typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
2853 : extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2854 : unsigned long size, pte_fn_t fn, void *data);
2855 : extern int apply_to_existing_page_range(struct mm_struct *mm,
2856 : unsigned long address, unsigned long size,
2857 : pte_fn_t fn, void *data);
2858 :
2859 : extern void init_mem_debugging_and_hardening(void);
2860 : #ifdef CONFIG_PAGE_POISONING
2861 : extern void __kernel_poison_pages(struct page *page, int numpages);
2862 : extern void __kernel_unpoison_pages(struct page *page, int numpages);
2863 : extern bool _page_poisoning_enabled_early;
2864 : DECLARE_STATIC_KEY_FALSE(_page_poisoning_enabled);
2865 : static inline bool page_poisoning_enabled(void)
2866 : {
2867 : return _page_poisoning_enabled_early;
2868 : }
2869 : /*
2870 : * For use in fast paths after init_mem_debugging() has run, or when a
2871 : * false negative result is not harmful when called too early.
2872 : */
2873 : static inline bool page_poisoning_enabled_static(void)
2874 : {
2875 : return static_branch_unlikely(&_page_poisoning_enabled);
2876 : }
2877 : static inline void kernel_poison_pages(struct page *page, int numpages)
2878 : {
2879 : if (page_poisoning_enabled_static())
2880 : __kernel_poison_pages(page, numpages);
2881 : }
2882 : static inline void kernel_unpoison_pages(struct page *page, int numpages)
2883 : {
2884 : if (page_poisoning_enabled_static())
2885 : __kernel_unpoison_pages(page, numpages);
2886 : }
2887 : #else
2888 0 : static inline bool page_poisoning_enabled(void) { return false; }
2889 0 : static inline bool page_poisoning_enabled_static(void) { return false; }
2890 : static inline void __kernel_poison_pages(struct page *page, int nunmpages) { }
2891 : static inline void kernel_poison_pages(struct page *page, int numpages) { }
2892 192995 : static inline void kernel_unpoison_pages(struct page *page, int numpages) { }
2893 : #endif
2894 :
2895 : DECLARE_STATIC_KEY_FALSE(init_on_alloc);
2896 193862 : static inline bool want_init_on_alloc(gfp_t flags)
2897 : {
2898 193862 : if (static_branch_unlikely(&init_on_alloc))
2899 : return true;
2900 193876 : return flags & __GFP_ZERO;
2901 : }
2902 :
2903 : DECLARE_STATIC_KEY_FALSE(init_on_free);
2904 341376 : static inline bool want_init_on_free(void)
2905 : {
2906 341376 : return static_branch_unlikely(&init_on_free);
2907 : }
2908 :
2909 : extern bool _debug_pagealloc_enabled_early;
2910 : DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
2911 :
2912 8779 : static inline bool debug_pagealloc_enabled(void)
2913 : {
2914 8779 : return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
2915 : _debug_pagealloc_enabled_early;
2916 : }
2917 :
2918 : /*
2919 : * For use in fast paths after init_debug_pagealloc() has run, or when a
2920 : * false negative result is not harmful when called too early.
2921 : */
2922 1335783 : static inline bool debug_pagealloc_enabled_static(void)
2923 : {
2924 1335783 : if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
2925 1335783 : return false;
2926 :
2927 : return static_branch_unlikely(&_debug_pagealloc_enabled);
2928 : }
2929 :
2930 : #ifdef CONFIG_DEBUG_PAGEALLOC
2931 : /*
2932 : * To support DEBUG_PAGEALLOC architecture must ensure that
2933 : * __kernel_map_pages() never fails
2934 : */
2935 : extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2936 :
2937 : static inline void debug_pagealloc_map_pages(struct page *page, int numpages)
2938 : {
2939 : if (debug_pagealloc_enabled_static())
2940 : __kernel_map_pages(page, numpages, 1);
2941 : }
2942 :
2943 : static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages)
2944 : {
2945 : if (debug_pagealloc_enabled_static())
2946 : __kernel_map_pages(page, numpages, 0);
2947 : }
2948 : #else /* CONFIG_DEBUG_PAGEALLOC */
2949 193031 : static inline void debug_pagealloc_map_pages(struct page *page, int numpages) {}
2950 : static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages) {}
2951 : #endif /* CONFIG_DEBUG_PAGEALLOC */
2952 :
2953 : #ifdef __HAVE_ARCH_GATE_AREA
2954 : extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2955 : extern int in_gate_area_no_mm(unsigned long addr);
2956 : extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2957 : #else
2958 2 : static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2959 : {
2960 2 : return NULL;
2961 : }
2962 : static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2963 : static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2964 : {
2965 : return 0;
2966 : }
2967 : #endif /* __HAVE_ARCH_GATE_AREA */
2968 :
2969 : extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2970 :
2971 : #ifdef CONFIG_SYSCTL
2972 : extern int sysctl_drop_caches;
2973 : int drop_caches_sysctl_handler(struct ctl_table *, int, void *, size_t *,
2974 : loff_t *);
2975 : #endif
2976 :
2977 : void drop_slab(void);
2978 : void drop_slab_node(int nid);
2979 :
2980 : #ifndef CONFIG_MMU
2981 : #define randomize_va_space 0
2982 : #else
2983 : extern int randomize_va_space;
2984 : #endif
2985 :
2986 : const char * arch_vma_name(struct vm_area_struct *vma);
2987 : #ifdef CONFIG_MMU
2988 : void print_vma_addr(char *prefix, unsigned long rip);
2989 : #else
2990 : static inline void print_vma_addr(char *prefix, unsigned long rip)
2991 : {
2992 : }
2993 : #endif
2994 :
2995 : void *sparse_buffer_alloc(unsigned long size);
2996 : struct page * __populate_section_memmap(unsigned long pfn,
2997 : unsigned long nr_pages, int nid, struct vmem_altmap *altmap);
2998 : pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2999 : p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
3000 : pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
3001 : pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
3002 : pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node,
3003 : struct vmem_altmap *altmap);
3004 : void *vmemmap_alloc_block(unsigned long size, int node);
3005 : struct vmem_altmap;
3006 : void *vmemmap_alloc_block_buf(unsigned long size, int node,
3007 : struct vmem_altmap *altmap);
3008 : void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
3009 : int vmemmap_populate_basepages(unsigned long start, unsigned long end,
3010 : int node, struct vmem_altmap *altmap);
3011 : int vmemmap_populate(unsigned long start, unsigned long end, int node,
3012 : struct vmem_altmap *altmap);
3013 : void vmemmap_populate_print_last(void);
3014 : #ifdef CONFIG_MEMORY_HOTPLUG
3015 : void vmemmap_free(unsigned long start, unsigned long end,
3016 : struct vmem_altmap *altmap);
3017 : #endif
3018 : void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
3019 : unsigned long nr_pages);
3020 :
3021 : enum mf_flags {
3022 : MF_COUNT_INCREASED = 1 << 0,
3023 : MF_ACTION_REQUIRED = 1 << 1,
3024 : MF_MUST_KILL = 1 << 2,
3025 : MF_SOFT_OFFLINE = 1 << 3,
3026 : };
3027 : extern int memory_failure(unsigned long pfn, int flags);
3028 : extern void memory_failure_queue(unsigned long pfn, int flags);
3029 : extern void memory_failure_queue_kick(int cpu);
3030 : extern int unpoison_memory(unsigned long pfn);
3031 : extern int sysctl_memory_failure_early_kill;
3032 : extern int sysctl_memory_failure_recovery;
3033 : extern void shake_page(struct page *p, int access);
3034 : extern atomic_long_t num_poisoned_pages __read_mostly;
3035 : extern int soft_offline_page(unsigned long pfn, int flags);
3036 :
3037 :
3038 : /*
3039 : * Error handlers for various types of pages.
3040 : */
3041 : enum mf_result {
3042 : MF_IGNORED, /* Error: cannot be handled */
3043 : MF_FAILED, /* Error: handling failed */
3044 : MF_DELAYED, /* Will be handled later */
3045 : MF_RECOVERED, /* Successfully recovered */
3046 : };
3047 :
3048 : enum mf_action_page_type {
3049 : MF_MSG_KERNEL,
3050 : MF_MSG_KERNEL_HIGH_ORDER,
3051 : MF_MSG_SLAB,
3052 : MF_MSG_DIFFERENT_COMPOUND,
3053 : MF_MSG_POISONED_HUGE,
3054 : MF_MSG_HUGE,
3055 : MF_MSG_FREE_HUGE,
3056 : MF_MSG_NON_PMD_HUGE,
3057 : MF_MSG_UNMAP_FAILED,
3058 : MF_MSG_DIRTY_SWAPCACHE,
3059 : MF_MSG_CLEAN_SWAPCACHE,
3060 : MF_MSG_DIRTY_MLOCKED_LRU,
3061 : MF_MSG_CLEAN_MLOCKED_LRU,
3062 : MF_MSG_DIRTY_UNEVICTABLE_LRU,
3063 : MF_MSG_CLEAN_UNEVICTABLE_LRU,
3064 : MF_MSG_DIRTY_LRU,
3065 : MF_MSG_CLEAN_LRU,
3066 : MF_MSG_TRUNCATED_LRU,
3067 : MF_MSG_BUDDY,
3068 : MF_MSG_BUDDY_2ND,
3069 : MF_MSG_DAX,
3070 : MF_MSG_UNSPLIT_THP,
3071 : MF_MSG_UNKNOWN,
3072 : };
3073 :
3074 : #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
3075 : extern void clear_huge_page(struct page *page,
3076 : unsigned long addr_hint,
3077 : unsigned int pages_per_huge_page);
3078 : extern void copy_user_huge_page(struct page *dst, struct page *src,
3079 : unsigned long addr_hint,
3080 : struct vm_area_struct *vma,
3081 : unsigned int pages_per_huge_page);
3082 : extern long copy_huge_page_from_user(struct page *dst_page,
3083 : const void __user *usr_src,
3084 : unsigned int pages_per_huge_page,
3085 : bool allow_pagefault);
3086 :
3087 : /**
3088 : * vma_is_special_huge - Are transhuge page-table entries considered special?
3089 : * @vma: Pointer to the struct vm_area_struct to consider
3090 : *
3091 : * Whether transhuge page-table entries are considered "special" following
3092 : * the definition in vm_normal_page().
3093 : *
3094 : * Return: true if transhuge page-table entries should be considered special,
3095 : * false otherwise.
3096 : */
3097 17 : static inline bool vma_is_special_huge(const struct vm_area_struct *vma)
3098 : {
3099 34 : return vma_is_dax(vma) || (vma->vm_file &&
3100 0 : (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP)));
3101 : }
3102 :
3103 : #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
3104 :
3105 : #ifdef CONFIG_DEBUG_PAGEALLOC
3106 : extern unsigned int _debug_guardpage_minorder;
3107 : DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
3108 :
3109 : static inline unsigned int debug_guardpage_minorder(void)
3110 : {
3111 : return _debug_guardpage_minorder;
3112 : }
3113 :
3114 : static inline bool debug_guardpage_enabled(void)
3115 : {
3116 : return static_branch_unlikely(&_debug_guardpage_enabled);
3117 : }
3118 :
3119 : static inline bool page_is_guard(struct page *page)
3120 : {
3121 : if (!debug_guardpage_enabled())
3122 : return false;
3123 :
3124 : return PageGuard(page);
3125 : }
3126 : #else
3127 1 : static inline unsigned int debug_guardpage_minorder(void) { return 0; }
3128 : static inline bool debug_guardpage_enabled(void) { return false; }
3129 102485 : static inline bool page_is_guard(struct page *page) { return false; }
3130 : #endif /* CONFIG_DEBUG_PAGEALLOC */
3131 :
3132 : #if MAX_NUMNODES > 1
3133 : void __init setup_nr_node_ids(void);
3134 : #else
3135 : static inline void setup_nr_node_ids(void) {}
3136 : #endif
3137 :
3138 : extern int memcmp_pages(struct page *page1, struct page *page2);
3139 :
3140 0 : static inline int pages_identical(struct page *page1, struct page *page2)
3141 : {
3142 0 : return !memcmp_pages(page1, page2);
3143 : }
3144 :
3145 : #ifdef CONFIG_MAPPING_DIRTY_HELPERS
3146 : unsigned long clean_record_shared_mapping_range(struct address_space *mapping,
3147 : pgoff_t first_index, pgoff_t nr,
3148 : pgoff_t bitmap_pgoff,
3149 : unsigned long *bitmap,
3150 : pgoff_t *start,
3151 : pgoff_t *end);
3152 :
3153 : unsigned long wp_shared_mapping_range(struct address_space *mapping,
3154 : pgoff_t first_index, pgoff_t nr);
3155 : #endif
3156 :
3157 : extern int sysctl_nr_trim_pages;
3158 :
3159 : void mem_dump_obj(void *object);
3160 :
3161 : #endif /* __KERNEL__ */
3162 : #endif /* _LINUX_MM_H */
|
