1 /* SPDX-License-Identifier: GPL-2.0 */
5 #include <linux/errno.h>
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>
38 struct anon_vma_chain;
41 struct writeback_control;
45 extern int sysctl_page_lock_unfairness;
47 void init_mm_internals(void);
49 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
50 extern unsigned long max_mapnr;
52 static inline void set_max_mapnr(unsigned long limit)
57 static inline void set_max_mapnr(unsigned long limit) { }
60 extern atomic_long_t _totalram_pages;
61 static inline unsigned long totalram_pages(void)
63 return (unsigned long)atomic_long_read(&_totalram_pages);
66 static inline void totalram_pages_inc(void)
68 atomic_long_inc(&_totalram_pages);
71 static inline void totalram_pages_dec(void)
73 atomic_long_dec(&_totalram_pages);
76 static inline void totalram_pages_add(long count)
78 atomic_long_add(count, &_totalram_pages);
81 extern void * high_memory;
82 extern int page_cluster;
85 extern int sysctl_legacy_va_layout;
87 #define sysctl_legacy_va_layout 0
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;
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;
101 #include <asm/page.h>
102 #include <asm/processor.h>
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.
111 #ifndef untagged_addr
112 #define untagged_addr(addr) (addr)
116 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
120 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
124 #define lm_alias(x) __va(__pa_symbol(x))
128 * With CONFIG_CFI_CLANG, the compiler replaces function addresses in
129 * instrumented C code with jump table addresses. Architectures that
130 * support CFI can define this macro to return the actual function address
133 #ifndef function_nocfi
134 #define function_nocfi(x) (x)
138 * To prevent common memory management code establishing
139 * a zero page mapping on a read fault.
140 * This macro should be defined within <asm/pgtable.h>.
141 * s390 does this to prevent multiplexing of hardware bits
142 * related to the physical page in case of virtualization.
144 #ifndef mm_forbids_zeropage
145 #define mm_forbids_zeropage(X) (0)
149 * On some architectures it is expensive to call memset() for small sizes.
150 * If an architecture decides to implement their own version of
151 * mm_zero_struct_page they should wrap the defines below in a #ifndef and
152 * define their own version of this macro in <asm/pgtable.h>
154 #if BITS_PER_LONG == 64
155 /* This function must be updated when the size of struct page grows above 80
156 * or reduces below 56. The idea that compiler optimizes out switch()
157 * statement, and only leaves move/store instructions. Also the compiler can
158 * combine write statments if they are both assignments and can be reordered,
159 * this can result in several of the writes here being dropped.
161 #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
162 static inline void __mm_zero_struct_page(struct page *page)
164 unsigned long *_pp = (void *)page;
166 /* Check that struct page is either 56, 64, 72, or 80 bytes */
167 BUILD_BUG_ON(sizeof(struct page) & 7);
168 BUILD_BUG_ON(sizeof(struct page) < 56);
169 BUILD_BUG_ON(sizeof(struct page) > 80);
171 switch (sizeof(struct page)) {
192 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
196 * Default maximum number of active map areas, this limits the number of vmas
197 * per mm struct. Users can overwrite this number by sysctl but there is a
200 * When a program's coredump is generated as ELF format, a section is created
201 * per a vma. In ELF, the number of sections is represented in unsigned short.
202 * This means the number of sections should be smaller than 65535 at coredump.
203 * Because the kernel adds some informative sections to a image of program at
204 * generating coredump, we need some margin. The number of extra sections is
205 * 1-3 now and depends on arch. We use "5" as safe margin, here.
207 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
208 * not a hard limit any more. Although some userspace tools can be surprised by
211 #define MAPCOUNT_ELF_CORE_MARGIN (5)
212 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
214 extern int sysctl_max_map_count;
216 extern unsigned long sysctl_user_reserve_kbytes;
217 extern unsigned long sysctl_admin_reserve_kbytes;
219 extern int sysctl_overcommit_memory;
220 extern int sysctl_overcommit_ratio;
221 extern unsigned long sysctl_overcommit_kbytes;
223 int overcommit_ratio_handler(struct ctl_table *, int, void *, size_t *,
225 int overcommit_kbytes_handler(struct ctl_table *, int, void *, size_t *,
227 int overcommit_policy_handler(struct ctl_table *, int, void *, size_t *,
230 * Any attempt to mark this function as static leads to build failure
231 * when CONFIG_DEBUG_INFO_BTF is enabled because __add_to_page_cache_locked()
232 * is referred to by BPF code. This must be visible for error injection.
234 int __add_to_page_cache_locked(struct page *page, struct address_space *mapping,
235 pgoff_t index, gfp_t gfp, void **shadowp);
237 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
239 /* to align the pointer to the (next) page boundary */
240 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
242 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
243 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
245 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
248 * Linux kernel virtual memory manager primitives.
249 * The idea being to have a "virtual" mm in the same way
250 * we have a virtual fs - giving a cleaner interface to the
251 * mm details, and allowing different kinds of memory mappings
252 * (from shared memory to executable loading to arbitrary
256 struct vm_area_struct *vm_area_alloc(struct mm_struct *);
257 struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
258 void vm_area_free(struct vm_area_struct *);
261 extern struct rb_root nommu_region_tree;
262 extern struct rw_semaphore nommu_region_sem;
264 extern unsigned int kobjsize(const void *objp);
268 * vm_flags in vm_area_struct, see mm_types.h.
269 * When changing, update also include/trace/events/mmflags.h
271 #define VM_NONE 0x00000000
273 #define VM_READ 0x00000001 /* currently active flags */
274 #define VM_WRITE 0x00000002
275 #define VM_EXEC 0x00000004
276 #define VM_SHARED 0x00000008
278 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
279 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
280 #define VM_MAYWRITE 0x00000020
281 #define VM_MAYEXEC 0x00000040
282 #define VM_MAYSHARE 0x00000080
284 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
285 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
286 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
287 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
288 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
290 #define VM_LOCKED 0x00002000
291 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
293 /* Used by sys_madvise() */
294 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
295 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
297 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
298 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
299 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
300 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
301 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
302 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
303 #define VM_SYNC 0x00800000 /* Synchronous page faults */
304 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
305 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
306 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
308 #ifdef CONFIG_MEM_SOFT_DIRTY
309 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
311 # define VM_SOFTDIRTY 0
314 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
315 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
316 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
317 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
319 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
320 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
321 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
322 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
323 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
324 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
325 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
326 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
327 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
328 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
329 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
330 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
332 #ifdef CONFIG_ARCH_HAS_PKEYS
333 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
334 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
335 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
336 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
337 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
339 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
341 # define VM_PKEY_BIT4 0
343 #endif /* CONFIG_ARCH_HAS_PKEYS */
345 #if defined(CONFIG_X86)
346 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
347 #elif defined(CONFIG_PPC)
348 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
349 #elif defined(CONFIG_PARISC)
350 # define VM_GROWSUP VM_ARCH_1
351 #elif defined(CONFIG_IA64)
352 # define VM_GROWSUP VM_ARCH_1
353 #elif defined(CONFIG_SPARC64)
354 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
355 # define VM_ARCH_CLEAR VM_SPARC_ADI
356 #elif defined(CONFIG_ARM64)
357 # define VM_ARM64_BTI VM_ARCH_1 /* BTI guarded page, a.k.a. GP bit */
358 # define VM_ARCH_CLEAR VM_ARM64_BTI
359 #elif !defined(CONFIG_MMU)
360 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
363 #if defined(CONFIG_ARM64_MTE)
364 # define VM_MTE VM_HIGH_ARCH_0 /* Use Tagged memory for access control */
365 # define VM_MTE_ALLOWED VM_HIGH_ARCH_1 /* Tagged memory permitted */
367 # define VM_MTE VM_NONE
368 # define VM_MTE_ALLOWED VM_NONE
372 # define VM_GROWSUP VM_NONE
375 /* Bits set in the VMA until the stack is in its final location */
376 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
378 #define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0)
380 /* Common data flag combinations */
381 #define VM_DATA_FLAGS_TSK_EXEC (VM_READ | VM_WRITE | TASK_EXEC | \
382 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
383 #define VM_DATA_FLAGS_NON_EXEC (VM_READ | VM_WRITE | VM_MAYREAD | \
384 VM_MAYWRITE | VM_MAYEXEC)
385 #define VM_DATA_FLAGS_EXEC (VM_READ | VM_WRITE | VM_EXEC | \
386 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
388 #ifndef VM_DATA_DEFAULT_FLAGS /* arch can override this */
389 #define VM_DATA_DEFAULT_FLAGS VM_DATA_FLAGS_EXEC
392 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
393 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
396 #ifdef CONFIG_STACK_GROWSUP
397 #define VM_STACK VM_GROWSUP
399 #define VM_STACK VM_GROWSDOWN
402 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
404 /* VMA basic access permission flags */
405 #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
409 * Special vmas that are non-mergable, non-mlock()able.
411 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
413 /* This mask prevents VMA from being scanned with khugepaged */
414 #define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB)
416 /* This mask defines which mm->def_flags a process can inherit its parent */
417 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
419 /* This mask is used to clear all the VMA flags used by mlock */
420 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
422 /* Arch-specific flags to clear when updating VM flags on protection change */
423 #ifndef VM_ARCH_CLEAR
424 # define VM_ARCH_CLEAR VM_NONE
426 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
429 * mapping from the currently active vm_flags protection bits (the
430 * low four bits) to a page protection mask..
432 extern pgprot_t protection_map[16];
435 * Fault flag definitions.
437 * @FAULT_FLAG_WRITE: Fault was a write fault.
438 * @FAULT_FLAG_MKWRITE: Fault was mkwrite of existing PTE.
439 * @FAULT_FLAG_ALLOW_RETRY: Allow to retry the fault if blocked.
440 * @FAULT_FLAG_RETRY_NOWAIT: Don't drop mmap_lock and wait when retrying.
441 * @FAULT_FLAG_KILLABLE: The fault task is in SIGKILL killable region.
442 * @FAULT_FLAG_TRIED: The fault has been tried once.
443 * @FAULT_FLAG_USER: The fault originated in userspace.
444 * @FAULT_FLAG_REMOTE: The fault is not for current task/mm.
445 * @FAULT_FLAG_INSTRUCTION: The fault was during an instruction fetch.
446 * @FAULT_FLAG_INTERRUPTIBLE: The fault can be interrupted by non-fatal signals.
448 * About @FAULT_FLAG_ALLOW_RETRY and @FAULT_FLAG_TRIED: we can specify
449 * whether we would allow page faults to retry by specifying these two
450 * fault flags correctly. Currently there can be three legal combinations:
452 * (a) ALLOW_RETRY and !TRIED: this means the page fault allows retry, and
453 * this is the first try
455 * (b) ALLOW_RETRY and TRIED: this means the page fault allows retry, and
456 * we've already tried at least once
458 * (c) !ALLOW_RETRY and !TRIED: this means the page fault does not allow retry
460 * The unlisted combination (!ALLOW_RETRY && TRIED) is illegal and should never
461 * be used. Note that page faults can be allowed to retry for multiple times,
462 * in which case we'll have an initial fault with flags (a) then later on
463 * continuous faults with flags (b). We should always try to detect pending
464 * signals before a retry to make sure the continuous page faults can still be
465 * interrupted if necessary.
467 #define FAULT_FLAG_WRITE 0x01
468 #define FAULT_FLAG_MKWRITE 0x02
469 #define FAULT_FLAG_ALLOW_RETRY 0x04
470 #define FAULT_FLAG_RETRY_NOWAIT 0x08
471 #define FAULT_FLAG_KILLABLE 0x10
472 #define FAULT_FLAG_TRIED 0x20
473 #define FAULT_FLAG_USER 0x40
474 #define FAULT_FLAG_REMOTE 0x80
475 #define FAULT_FLAG_INSTRUCTION 0x100
476 #define FAULT_FLAG_INTERRUPTIBLE 0x200
479 * The default fault flags that should be used by most of the
480 * arch-specific page fault handlers.
482 #define FAULT_FLAG_DEFAULT (FAULT_FLAG_ALLOW_RETRY | \
483 FAULT_FLAG_KILLABLE | \
484 FAULT_FLAG_INTERRUPTIBLE)
487 * fault_flag_allow_retry_first - check ALLOW_RETRY the first time
489 * This is mostly used for places where we want to try to avoid taking
490 * the mmap_lock for too long a time when waiting for another condition
491 * to change, in which case we can try to be polite to release the
492 * mmap_lock in the first round to avoid potential starvation of other
493 * processes that would also want the mmap_lock.
495 * Return: true if the page fault allows retry and this is the first
496 * attempt of the fault handling; false otherwise.
498 static inline bool fault_flag_allow_retry_first(unsigned int flags)
500 return (flags & FAULT_FLAG_ALLOW_RETRY) &&
501 (!(flags & FAULT_FLAG_TRIED));
504 #define FAULT_FLAG_TRACE \
505 { FAULT_FLAG_WRITE, "WRITE" }, \
506 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
507 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
508 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
509 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
510 { FAULT_FLAG_TRIED, "TRIED" }, \
511 { FAULT_FLAG_USER, "USER" }, \
512 { FAULT_FLAG_REMOTE, "REMOTE" }, \
513 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }, \
514 { FAULT_FLAG_INTERRUPTIBLE, "INTERRUPTIBLE" }
517 * vm_fault is filled by the pagefault handler and passed to the vma's
518 * ->fault function. The vma's ->fault is responsible for returning a bitmask
519 * of VM_FAULT_xxx flags that give details about how the fault was handled.
521 * MM layer fills up gfp_mask for page allocations but fault handler might
522 * alter it if its implementation requires a different allocation context.
524 * pgoff should be used in favour of virtual_address, if possible.
528 struct vm_area_struct *vma; /* Target VMA */
529 gfp_t gfp_mask; /* gfp mask to be used for allocations */
530 pgoff_t pgoff; /* Logical page offset based on vma */
531 unsigned long address; /* Faulting virtual address */
533 unsigned int flags; /* FAULT_FLAG_xxx flags
534 * XXX: should really be 'const' */
535 pmd_t *pmd; /* Pointer to pmd entry matching
537 pud_t *pud; /* Pointer to pud entry matching
540 pte_t orig_pte; /* Value of PTE at the time of fault */
542 struct page *cow_page; /* Page handler may use for COW fault */
543 struct page *page; /* ->fault handlers should return a
544 * page here, unless VM_FAULT_NOPAGE
545 * is set (which is also implied by
548 /* These three entries are valid only while holding ptl lock */
549 pte_t *pte; /* Pointer to pte entry matching
550 * the 'address'. NULL if the page
551 * table hasn't been allocated.
553 spinlock_t *ptl; /* Page table lock.
554 * Protects pte page table if 'pte'
555 * is not NULL, otherwise pmd.
557 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
558 * vm_ops->map_pages() sets up a page
559 * table from atomic context.
560 * do_fault_around() pre-allocates
561 * page table to avoid allocation from
566 /* page entry size for vm->huge_fault() */
567 enum page_entry_size {
574 * These are the virtual MM functions - opening of an area, closing and
575 * unmapping it (needed to keep files on disk up-to-date etc), pointer
576 * to the functions called when a no-page or a wp-page exception occurs.
578 struct vm_operations_struct {
579 void (*open)(struct vm_area_struct * area);
580 void (*close)(struct vm_area_struct * area);
581 /* Called any time before splitting to check if it's allowed */
582 int (*may_split)(struct vm_area_struct *area, unsigned long addr);
583 int (*mremap)(struct vm_area_struct *area, unsigned long flags);
585 * Called by mprotect() to make driver-specific permission
586 * checks before mprotect() is finalised. The VMA must not
587 * be modified. Returns 0 if eprotect() can proceed.
589 int (*mprotect)(struct vm_area_struct *vma, unsigned long start,
590 unsigned long end, unsigned long newflags);
591 vm_fault_t (*fault)(struct vm_fault *vmf);
592 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
593 enum page_entry_size pe_size);
594 vm_fault_t (*map_pages)(struct vm_fault *vmf,
595 pgoff_t start_pgoff, pgoff_t end_pgoff);
596 unsigned long (*pagesize)(struct vm_area_struct * area);
598 /* notification that a previously read-only page is about to become
599 * writable, if an error is returned it will cause a SIGBUS */
600 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
602 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
603 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
605 /* called by access_process_vm when get_user_pages() fails, typically
606 * for use by special VMAs. See also generic_access_phys() for a generic
607 * implementation useful for any iomem mapping.
609 int (*access)(struct vm_area_struct *vma, unsigned long addr,
610 void *buf, int len, int write);
612 /* Called by the /proc/PID/maps code to ask the vma whether it
613 * has a special name. Returning non-NULL will also cause this
614 * vma to be dumped unconditionally. */
615 const char *(*name)(struct vm_area_struct *vma);
619 * set_policy() op must add a reference to any non-NULL @new mempolicy
620 * to hold the policy upon return. Caller should pass NULL @new to
621 * remove a policy and fall back to surrounding context--i.e. do not
622 * install a MPOL_DEFAULT policy, nor the task or system default
625 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
628 * get_policy() op must add reference [mpol_get()] to any policy at
629 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
630 * in mm/mempolicy.c will do this automatically.
631 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
632 * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
633 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
634 * must return NULL--i.e., do not "fallback" to task or system default
637 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
641 * Called by vm_normal_page() for special PTEs to find the
642 * page for @addr. This is useful if the default behavior
643 * (using pte_page()) would not find the correct page.
645 struct page *(*find_special_page)(struct vm_area_struct *vma,
649 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
651 static const struct vm_operations_struct dummy_vm_ops = {};
653 memset(vma, 0, sizeof(*vma));
655 vma->vm_ops = &dummy_vm_ops;
656 INIT_LIST_HEAD(&vma->anon_vma_chain);
659 static inline void vma_set_anonymous(struct vm_area_struct *vma)
664 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
669 static inline bool vma_is_temporary_stack(struct vm_area_struct *vma)
671 int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP);
676 if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) ==
677 VM_STACK_INCOMPLETE_SETUP)
683 static inline bool vma_is_foreign(struct vm_area_struct *vma)
688 if (current->mm != vma->vm_mm)
694 static inline bool vma_is_accessible(struct vm_area_struct *vma)
696 return vma->vm_flags & VM_ACCESS_FLAGS;
701 * The vma_is_shmem is not inline because it is used only by slow
702 * paths in userfault.
704 bool vma_is_shmem(struct vm_area_struct *vma);
706 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
709 int vma_is_stack_for_current(struct vm_area_struct *vma);
711 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
712 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
717 #include <linux/huge_mm.h>
720 * Methods to modify the page usage count.
722 * What counts for a page usage:
723 * - cache mapping (page->mapping)
724 * - private data (page->private)
725 * - page mapped in a task's page tables, each mapping
726 * is counted separately
728 * Also, many kernel routines increase the page count before a critical
729 * routine so they can be sure the page doesn't go away from under them.
733 * Drop a ref, return true if the refcount fell to zero (the page has no users)
735 static inline int put_page_testzero(struct page *page)
737 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
738 return page_ref_dec_and_test(page);
742 * Try to grab a ref unless the page has a refcount of zero, return false if
744 * This can be called when MMU is off so it must not access
745 * any of the virtual mappings.
747 static inline int get_page_unless_zero(struct page *page)
749 return page_ref_add_unless(page, 1, 0);
752 extern int page_is_ram(unsigned long pfn);
760 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
763 /* Support for virtually mapped pages */
764 struct page *vmalloc_to_page(const void *addr);
765 unsigned long vmalloc_to_pfn(const void *addr);
768 * Determine if an address is within the vmalloc range
770 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
771 * is no special casing required.
774 #ifndef is_ioremap_addr
775 #define is_ioremap_addr(x) is_vmalloc_addr(x)
779 extern bool is_vmalloc_addr(const void *x);
780 extern int is_vmalloc_or_module_addr(const void *x);
782 static inline bool is_vmalloc_addr(const void *x)
786 static inline int is_vmalloc_or_module_addr(const void *x)
792 extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
793 static inline void *kvmalloc(size_t size, gfp_t flags)
795 return kvmalloc_node(size, flags, NUMA_NO_NODE);
797 static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
799 return kvmalloc_node(size, flags | __GFP_ZERO, node);
801 static inline void *kvzalloc(size_t size, gfp_t flags)
803 return kvmalloc(size, flags | __GFP_ZERO);
806 static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
810 if (unlikely(check_mul_overflow(n, size, &bytes)))
813 return kvmalloc(bytes, flags);
816 static inline void *kvcalloc(size_t n, size_t size, gfp_t flags)
818 return kvmalloc_array(n, size, flags | __GFP_ZERO);
821 extern void kvfree(const void *addr);
822 extern void kvfree_sensitive(const void *addr, size_t len);
824 static inline int head_compound_mapcount(struct page *head)
826 return atomic_read(compound_mapcount_ptr(head)) + 1;
830 * Mapcount of compound page as a whole, does not include mapped sub-pages.
832 * Must be called only for compound pages or any their tail sub-pages.
834 static inline int compound_mapcount(struct page *page)
836 VM_BUG_ON_PAGE(!PageCompound(page), page);
837 page = compound_head(page);
838 return head_compound_mapcount(page);
842 * The atomic page->_mapcount, starts from -1: so that transitions
843 * both from it and to it can be tracked, using atomic_inc_and_test
844 * and atomic_add_negative(-1).
846 static inline void page_mapcount_reset(struct page *page)
848 atomic_set(&(page)->_mapcount, -1);
851 int __page_mapcount(struct page *page);
854 * Mapcount of 0-order page; when compound sub-page, includes
855 * compound_mapcount().
857 * Result is undefined for pages which cannot be mapped into userspace.
858 * For example SLAB or special types of pages. See function page_has_type().
859 * They use this place in struct page differently.
861 static inline int page_mapcount(struct page *page)
863 if (unlikely(PageCompound(page)))
864 return __page_mapcount(page);
865 return atomic_read(&page->_mapcount) + 1;
868 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
869 int total_mapcount(struct page *page);
870 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
872 static inline int total_mapcount(struct page *page)
874 return page_mapcount(page);
876 static inline int page_trans_huge_mapcount(struct page *page,
879 int mapcount = page_mapcount(page);
881 *total_mapcount = mapcount;
886 static inline struct page *virt_to_head_page(const void *x)
888 struct page *page = virt_to_page(x);
890 return compound_head(page);
893 void __put_page(struct page *page);
895 void put_pages_list(struct list_head *pages);
897 void split_page(struct page *page, unsigned int order);
900 * Compound pages have a destructor function. Provide a
901 * prototype for that function and accessor functions.
902 * These are _only_ valid on the head of a compound page.
904 typedef void compound_page_dtor(struct page *);
906 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
907 enum compound_dtor_id {
910 #ifdef CONFIG_HUGETLB_PAGE
913 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
918 extern compound_page_dtor * const compound_page_dtors[NR_COMPOUND_DTORS];
920 static inline void set_compound_page_dtor(struct page *page,
921 enum compound_dtor_id compound_dtor)
923 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
924 page[1].compound_dtor = compound_dtor;
927 static inline void destroy_compound_page(struct page *page)
929 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
930 compound_page_dtors[page[1].compound_dtor](page);
933 static inline unsigned int compound_order(struct page *page)
937 return page[1].compound_order;
940 static inline bool hpage_pincount_available(struct page *page)
943 * Can the page->hpage_pinned_refcount field be used? That field is in
944 * the 3rd page of the compound page, so the smallest (2-page) compound
945 * pages cannot support it.
947 page = compound_head(page);
948 return PageCompound(page) && compound_order(page) > 1;
951 static inline int head_compound_pincount(struct page *head)
953 return atomic_read(compound_pincount_ptr(head));
956 static inline int compound_pincount(struct page *page)
958 VM_BUG_ON_PAGE(!hpage_pincount_available(page), page);
959 page = compound_head(page);
960 return head_compound_pincount(page);
963 static inline void set_compound_order(struct page *page, unsigned int order)
965 page[1].compound_order = order;
966 page[1].compound_nr = 1U << order;
969 /* Returns the number of pages in this potentially compound page. */
970 static inline unsigned long compound_nr(struct page *page)
974 return page[1].compound_nr;
977 /* Returns the number of bytes in this potentially compound page. */
978 static inline unsigned long page_size(struct page *page)
980 return PAGE_SIZE << compound_order(page);
983 /* Returns the number of bits needed for the number of bytes in a page */
984 static inline unsigned int page_shift(struct page *page)
986 return PAGE_SHIFT + compound_order(page);
989 void free_compound_page(struct page *page);
993 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
994 * servicing faults for write access. In the normal case, do always want
995 * pte_mkwrite. But get_user_pages can cause write faults for mappings
996 * that do not have writing enabled, when used by access_process_vm.
998 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
1000 if (likely(vma->vm_flags & VM_WRITE))
1001 pte = pte_mkwrite(pte);
1005 vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page);
1006 void do_set_pte(struct vm_fault *vmf, struct page *page, unsigned long addr);
1008 vm_fault_t finish_fault(struct vm_fault *vmf);
1009 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
1013 * Multiple processes may "see" the same page. E.g. for untouched
1014 * mappings of /dev/null, all processes see the same page full of
1015 * zeroes, and text pages of executables and shared libraries have
1016 * only one copy in memory, at most, normally.
1018 * For the non-reserved pages, page_count(page) denotes a reference count.
1019 * page_count() == 0 means the page is free. page->lru is then used for
1020 * freelist management in the buddy allocator.
1021 * page_count() > 0 means the page has been allocated.
1023 * Pages are allocated by the slab allocator in order to provide memory
1024 * to kmalloc and kmem_cache_alloc. In this case, the management of the
1025 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
1026 * unless a particular usage is carefully commented. (the responsibility of
1027 * freeing the kmalloc memory is the caller's, of course).
1029 * A page may be used by anyone else who does a __get_free_page().
1030 * In this case, page_count still tracks the references, and should only
1031 * be used through the normal accessor functions. The top bits of page->flags
1032 * and page->virtual store page management information, but all other fields
1033 * are unused and could be used privately, carefully. The management of this
1034 * page is the responsibility of the one who allocated it, and those who have
1035 * subsequently been given references to it.
1037 * The other pages (we may call them "pagecache pages") are completely
1038 * managed by the Linux memory manager: I/O, buffers, swapping etc.
1039 * The following discussion applies only to them.
1041 * A pagecache page contains an opaque `private' member, which belongs to the
1042 * page's address_space. Usually, this is the address of a circular list of
1043 * the page's disk buffers. PG_private must be set to tell the VM to call
1044 * into the filesystem to release these pages.
1046 * A page may belong to an inode's memory mapping. In this case, page->mapping
1047 * is the pointer to the inode, and page->index is the file offset of the page,
1048 * in units of PAGE_SIZE.
1050 * If pagecache pages are not associated with an inode, they are said to be
1051 * anonymous pages. These may become associated with the swapcache, and in that
1052 * case PG_swapcache is set, and page->private is an offset into the swapcache.
1054 * In either case (swapcache or inode backed), the pagecache itself holds one
1055 * reference to the page. Setting PG_private should also increment the
1056 * refcount. The each user mapping also has a reference to the page.
1058 * The pagecache pages are stored in a per-mapping radix tree, which is
1059 * rooted at mapping->i_pages, and indexed by offset.
1060 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
1061 * lists, we instead now tag pages as dirty/writeback in the radix tree.
1063 * All pagecache pages may be subject to I/O:
1064 * - inode pages may need to be read from disk,
1065 * - inode pages which have been modified and are MAP_SHARED may need
1066 * to be written back to the inode on disk,
1067 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
1068 * modified may need to be swapped out to swap space and (later) to be read
1073 * The zone field is never updated after free_area_init_core()
1074 * sets it, so none of the operations on it need to be atomic.
1077 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
1078 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
1079 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
1080 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
1081 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
1082 #define KASAN_TAG_PGOFF (LAST_CPUPID_PGOFF - KASAN_TAG_WIDTH)
1085 * Define the bit shifts to access each section. For non-existent
1086 * sections we define the shift as 0; that plus a 0 mask ensures
1087 * the compiler will optimise away reference to them.
1089 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
1090 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
1091 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
1092 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
1093 #define KASAN_TAG_PGSHIFT (KASAN_TAG_PGOFF * (KASAN_TAG_WIDTH != 0))
1095 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
1096 #ifdef NODE_NOT_IN_PAGE_FLAGS
1097 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
1098 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
1099 SECTIONS_PGOFF : ZONES_PGOFF)
1101 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
1102 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
1103 NODES_PGOFF : ZONES_PGOFF)
1106 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
1108 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
1109 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
1110 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
1111 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
1112 #define KASAN_TAG_MASK ((1UL << KASAN_TAG_WIDTH) - 1)
1113 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
1115 static inline enum zone_type page_zonenum(const struct page *page)
1117 ASSERT_EXCLUSIVE_BITS(page->flags, ZONES_MASK << ZONES_PGSHIFT);
1118 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
1121 #ifdef CONFIG_ZONE_DEVICE
1122 static inline bool is_zone_device_page(const struct page *page)
1124 return page_zonenum(page) == ZONE_DEVICE;
1126 extern void memmap_init_zone_device(struct zone *, unsigned long,
1127 unsigned long, struct dev_pagemap *);
1129 static inline bool is_zone_device_page(const struct page *page)
1135 #ifdef CONFIG_DEV_PAGEMAP_OPS
1136 void free_devmap_managed_page(struct page *page);
1137 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
1139 static inline bool page_is_devmap_managed(struct page *page)
1141 if (!static_branch_unlikely(&devmap_managed_key))
1143 if (!is_zone_device_page(page))
1145 switch (page->pgmap->type) {
1146 case MEMORY_DEVICE_PRIVATE:
1147 case MEMORY_DEVICE_FS_DAX:
1155 void put_devmap_managed_page(struct page *page);
1157 #else /* CONFIG_DEV_PAGEMAP_OPS */
1158 static inline bool page_is_devmap_managed(struct page *page)
1163 static inline void put_devmap_managed_page(struct page *page)
1166 #endif /* CONFIG_DEV_PAGEMAP_OPS */
1168 static inline bool is_device_private_page(const struct page *page)
1170 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
1171 IS_ENABLED(CONFIG_DEVICE_PRIVATE) &&
1172 is_zone_device_page(page) &&
1173 page->pgmap->type == MEMORY_DEVICE_PRIVATE;
1176 static inline bool is_pci_p2pdma_page(const struct page *page)
1178 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
1179 IS_ENABLED(CONFIG_PCI_P2PDMA) &&
1180 is_zone_device_page(page) &&
1181 page->pgmap->type == MEMORY_DEVICE_PCI_P2PDMA;
1184 /* 127: arbitrary random number, small enough to assemble well */
1185 #define page_ref_zero_or_close_to_overflow(page) \
1186 ((unsigned int) page_ref_count(page) + 127u <= 127u)
1188 static inline void get_page(struct page *page)
1190 page = compound_head(page);
1192 * Getting a normal page or the head of a compound page
1193 * requires to already have an elevated page->_refcount.
1195 VM_BUG_ON_PAGE(page_ref_zero_or_close_to_overflow(page), page);
1199 bool __must_check try_grab_page(struct page *page, unsigned int flags);
1200 __maybe_unused struct page *try_grab_compound_head(struct page *page, int refs,
1201 unsigned int flags);
1204 static inline __must_check bool try_get_page(struct page *page)
1206 page = compound_head(page);
1207 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1213 static inline void put_page(struct page *page)
1215 page = compound_head(page);
1218 * For devmap managed pages we need to catch refcount transition from
1219 * 2 to 1, when refcount reach one it means the page is free and we
1220 * need to inform the device driver through callback. See
1221 * include/linux/memremap.h and HMM for details.
1223 if (page_is_devmap_managed(page)) {
1224 put_devmap_managed_page(page);
1228 if (put_page_testzero(page))
1233 * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
1234 * the page's refcount so that two separate items are tracked: the original page
1235 * reference count, and also a new count of how many pin_user_pages() calls were
1236 * made against the page. ("gup-pinned" is another term for the latter).
1238 * With this scheme, pin_user_pages() becomes special: such pages are marked as
1239 * distinct from normal pages. As such, the unpin_user_page() call (and its
1240 * variants) must be used in order to release gup-pinned pages.
1244 * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
1245 * counts with respect to pin_user_pages() and unpin_user_page() becomes
1246 * simpler, due to the fact that adding an even power of two to the page
1247 * refcount has the effect of using only the upper N bits, for the code that
1248 * counts up using the bias value. This means that the lower bits are left for
1249 * the exclusive use of the original code that increments and decrements by one
1250 * (or at least, by much smaller values than the bias value).
1252 * Of course, once the lower bits overflow into the upper bits (and this is
1253 * OK, because subtraction recovers the original values), then visual inspection
1254 * no longer suffices to directly view the separate counts. However, for normal
1255 * applications that don't have huge page reference counts, this won't be an
1258 * Locking: the lockless algorithm described in page_cache_get_speculative()
1259 * and page_cache_gup_pin_speculative() provides safe operation for
1260 * get_user_pages and page_mkclean and other calls that race to set up page
1263 #define GUP_PIN_COUNTING_BIAS (1U << 10)
1265 void unpin_user_page(struct page *page);
1266 void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1268 void unpin_user_pages(struct page **pages, unsigned long npages);
1271 * page_maybe_dma_pinned() - report if a page is pinned for DMA.
1273 * This function checks if a page has been pinned via a call to
1274 * pin_user_pages*().
1276 * For non-huge pages, the return value is partially fuzzy: false is not fuzzy,
1277 * because it means "definitely not pinned for DMA", but true means "probably
1278 * pinned for DMA, but possibly a false positive due to having at least
1279 * GUP_PIN_COUNTING_BIAS worth of normal page references".
1281 * False positives are OK, because: a) it's unlikely for a page to get that many
1282 * refcounts, and b) all the callers of this routine are expected to be able to
1283 * deal gracefully with a false positive.
1285 * For huge pages, the result will be exactly correct. That's because we have
1286 * more tracking data available: the 3rd struct page in the compound page is
1287 * used to track the pincount (instead using of the GUP_PIN_COUNTING_BIAS
1290 * For more information, please see Documentation/core-api/pin_user_pages.rst.
1292 * @page: pointer to page to be queried.
1293 * @Return: True, if it is likely that the page has been "dma-pinned".
1294 * False, if the page is definitely not dma-pinned.
1296 static inline bool page_maybe_dma_pinned(struct page *page)
1298 if (hpage_pincount_available(page))
1299 return compound_pincount(page) > 0;
1302 * page_ref_count() is signed. If that refcount overflows, then
1303 * page_ref_count() returns a negative value, and callers will avoid
1304 * further incrementing the refcount.
1306 * Here, for that overflow case, use the signed bit to count a little
1307 * bit higher via unsigned math, and thus still get an accurate result.
1309 return ((unsigned int)page_ref_count(compound_head(page))) >=
1310 GUP_PIN_COUNTING_BIAS;
1313 static inline bool is_cow_mapping(vm_flags_t flags)
1315 return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
1319 * This should most likely only be called during fork() to see whether we
1320 * should break the cow immediately for a page on the src mm.
1322 static inline bool page_needs_cow_for_dma(struct vm_area_struct *vma,
1325 if (!is_cow_mapping(vma->vm_flags))
1328 if (!atomic_read(&vma->vm_mm->has_pinned))
1331 return page_maybe_dma_pinned(page);
1334 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1335 #define SECTION_IN_PAGE_FLAGS
1339 * The identification function is mainly used by the buddy allocator for
1340 * determining if two pages could be buddies. We are not really identifying
1341 * the zone since we could be using the section number id if we do not have
1342 * node id available in page flags.
1343 * We only guarantee that it will return the same value for two combinable
1346 static inline int page_zone_id(struct page *page)
1348 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1351 #ifdef NODE_NOT_IN_PAGE_FLAGS
1352 extern int page_to_nid(const struct page *page);
1354 static inline int page_to_nid(const struct page *page)
1356 struct page *p = (struct page *)page;
1358 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1362 #ifdef CONFIG_NUMA_BALANCING
1363 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1365 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1368 static inline int cpupid_to_pid(int cpupid)
1370 return cpupid & LAST__PID_MASK;
1373 static inline int cpupid_to_cpu(int cpupid)
1375 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1378 static inline int cpupid_to_nid(int cpupid)
1380 return cpu_to_node(cpupid_to_cpu(cpupid));
1383 static inline bool cpupid_pid_unset(int cpupid)
1385 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1388 static inline bool cpupid_cpu_unset(int cpupid)
1390 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1393 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1395 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1398 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1399 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1400 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1402 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1405 static inline int page_cpupid_last(struct page *page)
1407 return page->_last_cpupid;
1409 static inline void page_cpupid_reset_last(struct page *page)
1411 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1414 static inline int page_cpupid_last(struct page *page)
1416 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1419 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1421 static inline void page_cpupid_reset_last(struct page *page)
1423 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1425 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1426 #else /* !CONFIG_NUMA_BALANCING */
1427 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1429 return page_to_nid(page); /* XXX */
1432 static inline int page_cpupid_last(struct page *page)
1434 return page_to_nid(page); /* XXX */
1437 static inline int cpupid_to_nid(int cpupid)
1442 static inline int cpupid_to_pid(int cpupid)
1447 static inline int cpupid_to_cpu(int cpupid)
1452 static inline int cpu_pid_to_cpupid(int nid, int pid)
1457 static inline bool cpupid_pid_unset(int cpupid)
1462 static inline void page_cpupid_reset_last(struct page *page)
1466 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1470 #endif /* CONFIG_NUMA_BALANCING */
1472 #if defined(CONFIG_KASAN_SW_TAGS) || defined(CONFIG_KASAN_HW_TAGS)
1475 * KASAN per-page tags are stored xor'ed with 0xff. This allows to avoid
1476 * setting tags for all pages to native kernel tag value 0xff, as the default
1477 * value 0x00 maps to 0xff.
1480 static inline u8 page_kasan_tag(const struct page *page)
1484 if (kasan_enabled()) {
1485 tag = (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1492 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1494 if (kasan_enabled()) {
1496 page->flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1497 page->flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1501 static inline void page_kasan_tag_reset(struct page *page)
1503 if (kasan_enabled())
1504 page_kasan_tag_set(page, 0xff);
1507 #else /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1509 static inline u8 page_kasan_tag(const struct page *page)
1514 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1515 static inline void page_kasan_tag_reset(struct page *page) { }
1517 #endif /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1519 static inline struct zone *page_zone(const struct page *page)
1521 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1524 static inline pg_data_t *page_pgdat(const struct page *page)
1526 return NODE_DATA(page_to_nid(page));
1529 #ifdef SECTION_IN_PAGE_FLAGS
1530 static inline void set_page_section(struct page *page, unsigned long section)
1532 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1533 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1536 static inline unsigned long page_to_section(const struct page *page)
1538 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1542 static inline void set_page_zone(struct page *page, enum zone_type zone)
1544 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1545 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1548 static inline void set_page_node(struct page *page, unsigned long node)
1550 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1551 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1554 static inline void set_page_links(struct page *page, enum zone_type zone,
1555 unsigned long node, unsigned long pfn)
1557 set_page_zone(page, zone);
1558 set_page_node(page, node);
1559 #ifdef SECTION_IN_PAGE_FLAGS
1560 set_page_section(page, pfn_to_section_nr(pfn));
1565 * Some inline functions in vmstat.h depend on page_zone()
1567 #include <linux/vmstat.h>
1569 static __always_inline void *lowmem_page_address(const struct page *page)
1571 return page_to_virt(page);
1574 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1575 #define HASHED_PAGE_VIRTUAL
1578 #if defined(WANT_PAGE_VIRTUAL)
1579 static inline void *page_address(const struct page *page)
1581 return page->virtual;
1583 static inline void set_page_address(struct page *page, void *address)
1585 page->virtual = address;
1587 #define page_address_init() do { } while(0)
1590 #if defined(HASHED_PAGE_VIRTUAL)
1591 void *page_address(const struct page *page);
1592 void set_page_address(struct page *page, void *virtual);
1593 void page_address_init(void);
1596 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1597 #define page_address(page) lowmem_page_address(page)
1598 #define set_page_address(page, address) do { } while(0)
1599 #define page_address_init() do { } while(0)
1602 extern void *page_rmapping(struct page *page);
1603 extern struct anon_vma *page_anon_vma(struct page *page);
1604 extern struct address_space *page_mapping(struct page *page);
1606 extern struct address_space *__page_file_mapping(struct page *);
1609 struct address_space *page_file_mapping(struct page *page)
1611 if (unlikely(PageSwapCache(page)))
1612 return __page_file_mapping(page);
1614 return page->mapping;
1617 extern pgoff_t __page_file_index(struct page *page);
1620 * Return the pagecache index of the passed page. Regular pagecache pages
1621 * use ->index whereas swapcache pages use swp_offset(->private)
1623 static inline pgoff_t page_index(struct page *page)
1625 if (unlikely(PageSwapCache(page)))
1626 return __page_file_index(page);
1630 bool page_mapped(struct page *page);
1631 struct address_space *page_mapping(struct page *page);
1632 struct address_space *page_mapping_file(struct page *page);
1635 * Return true only if the page has been allocated with
1636 * ALLOC_NO_WATERMARKS and the low watermark was not
1637 * met implying that the system is under some pressure.
1639 static inline bool page_is_pfmemalloc(const struct page *page)
1642 * Page index cannot be this large so this must be
1643 * a pfmemalloc page.
1645 return page->index == -1UL;
1649 * Only to be called by the page allocator on a freshly allocated
1652 static inline void set_page_pfmemalloc(struct page *page)
1657 static inline void clear_page_pfmemalloc(struct page *page)
1663 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1665 extern void pagefault_out_of_memory(void);
1667 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1668 #define offset_in_thp(page, p) ((unsigned long)(p) & (thp_size(page) - 1))
1671 * Flags passed to show_mem() and show_free_areas() to suppress output in
1674 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1676 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1679 extern bool can_do_mlock(void);
1681 static inline bool can_do_mlock(void) { return false; }
1683 extern int user_shm_lock(size_t, struct user_struct *);
1684 extern void user_shm_unlock(size_t, struct user_struct *);
1687 * Parameter block passed down to zap_pte_range in exceptional cases.
1689 struct zap_details {
1690 struct address_space *check_mapping; /* Check page->mapping if set */
1691 pgoff_t first_index; /* Lowest page->index to unmap */
1692 pgoff_t last_index; /* Highest page->index to unmap */
1695 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1697 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1700 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1701 unsigned long size);
1702 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1703 unsigned long size);
1704 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1705 unsigned long start, unsigned long end);
1707 struct mmu_notifier_range;
1709 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1710 unsigned long end, unsigned long floor, unsigned long ceiling);
1712 copy_page_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma);
1713 int follow_invalidate_pte(struct mm_struct *mm, unsigned long address,
1714 struct mmu_notifier_range *range, pte_t **ptepp,
1715 pmd_t **pmdpp, spinlock_t **ptlp);
1716 int follow_pte(struct mm_struct *mm, unsigned long address,
1717 pte_t **ptepp, spinlock_t **ptlp);
1718 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1719 unsigned long *pfn);
1720 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1721 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1722 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1723 void *buf, int len, int write);
1725 extern void truncate_pagecache(struct inode *inode, loff_t new);
1726 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1727 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1728 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1729 int truncate_inode_page(struct address_space *mapping, struct page *page);
1730 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1731 int invalidate_inode_page(struct page *page);
1734 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1735 unsigned long address, unsigned int flags,
1736 struct pt_regs *regs);
1737 extern int fixup_user_fault(struct mm_struct *mm,
1738 unsigned long address, unsigned int fault_flags,
1740 void unmap_mapping_pages(struct address_space *mapping,
1741 pgoff_t start, pgoff_t nr, bool even_cows);
1742 void unmap_mapping_range(struct address_space *mapping,
1743 loff_t const holebegin, loff_t const holelen, int even_cows);
1745 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1746 unsigned long address, unsigned int flags,
1747 struct pt_regs *regs)
1749 /* should never happen if there's no MMU */
1751 return VM_FAULT_SIGBUS;
1753 static inline int fixup_user_fault(struct mm_struct *mm, unsigned long address,
1754 unsigned int fault_flags, bool *unlocked)
1756 /* should never happen if there's no MMU */
1760 static inline void unmap_mapping_pages(struct address_space *mapping,
1761 pgoff_t start, pgoff_t nr, bool even_cows) { }
1762 static inline void unmap_mapping_range(struct address_space *mapping,
1763 loff_t const holebegin, loff_t const holelen, int even_cows) { }
1766 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1767 loff_t const holebegin, loff_t const holelen)
1769 unmap_mapping_range(mapping, holebegin, holelen, 0);
1772 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1773 void *buf, int len, unsigned int gup_flags);
1774 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1775 void *buf, int len, unsigned int gup_flags);
1776 extern int __access_remote_vm(struct mm_struct *mm, unsigned long addr,
1777 void *buf, int len, unsigned int gup_flags);
1779 long get_user_pages_remote(struct mm_struct *mm,
1780 unsigned long start, unsigned long nr_pages,
1781 unsigned int gup_flags, struct page **pages,
1782 struct vm_area_struct **vmas, int *locked);
1783 long pin_user_pages_remote(struct mm_struct *mm,
1784 unsigned long start, unsigned long nr_pages,
1785 unsigned int gup_flags, struct page **pages,
1786 struct vm_area_struct **vmas, int *locked);
1787 long get_user_pages(unsigned long start, unsigned long nr_pages,
1788 unsigned int gup_flags, struct page **pages,
1789 struct vm_area_struct **vmas);
1790 long pin_user_pages(unsigned long start, unsigned long nr_pages,
1791 unsigned int gup_flags, struct page **pages,
1792 struct vm_area_struct **vmas);
1793 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1794 unsigned int gup_flags, struct page **pages, int *locked);
1795 long pin_user_pages_locked(unsigned long start, unsigned long nr_pages,
1796 unsigned int gup_flags, struct page **pages, int *locked);
1797 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1798 struct page **pages, unsigned int gup_flags);
1799 long pin_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1800 struct page **pages, unsigned int gup_flags);
1802 int get_user_pages_fast(unsigned long start, int nr_pages,
1803 unsigned int gup_flags, struct page **pages);
1804 int pin_user_pages_fast(unsigned long start, int nr_pages,
1805 unsigned int gup_flags, struct page **pages);
1807 int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
1808 int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
1809 struct task_struct *task, bool bypass_rlim);
1812 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1813 struct page **pages);
1814 int get_kernel_page(unsigned long start, int write, struct page **pages);
1815 struct page *get_dump_page(unsigned long addr);
1817 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1818 extern void do_invalidatepage(struct page *page, unsigned int offset,
1819 unsigned int length);
1821 void __set_page_dirty(struct page *, struct address_space *, int warn);
1822 int __set_page_dirty_nobuffers(struct page *page);
1823 int __set_page_dirty_no_writeback(struct page *page);
1824 int redirty_page_for_writepage(struct writeback_control *wbc,
1826 void account_page_dirtied(struct page *page, struct address_space *mapping);
1827 void account_page_cleaned(struct page *page, struct address_space *mapping,
1828 struct bdi_writeback *wb);
1829 int set_page_dirty(struct page *page);
1830 int set_page_dirty_lock(struct page *page);
1831 void __cancel_dirty_page(struct page *page);
1832 static inline void cancel_dirty_page(struct page *page)
1834 /* Avoid atomic ops, locking, etc. when not actually needed. */
1835 if (PageDirty(page))
1836 __cancel_dirty_page(page);
1838 int clear_page_dirty_for_io(struct page *page);
1840 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1842 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1843 unsigned long old_addr, struct vm_area_struct *new_vma,
1844 unsigned long new_addr, unsigned long len,
1845 bool need_rmap_locks);
1848 * Flags used by change_protection(). For now we make it a bitmap so
1849 * that we can pass in multiple flags just like parameters. However
1850 * for now all the callers are only use one of the flags at the same
1853 /* Whether we should allow dirty bit accounting */
1854 #define MM_CP_DIRTY_ACCT (1UL << 0)
1855 /* Whether this protection change is for NUMA hints */
1856 #define MM_CP_PROT_NUMA (1UL << 1)
1857 /* Whether this change is for write protecting */
1858 #define MM_CP_UFFD_WP (1UL << 2) /* do wp */
1859 #define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */
1860 #define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \
1861 MM_CP_UFFD_WP_RESOLVE)
1863 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1864 unsigned long end, pgprot_t newprot,
1865 unsigned long cp_flags);
1866 extern int mprotect_fixup(struct vm_area_struct *vma,
1867 struct vm_area_struct **pprev, unsigned long start,
1868 unsigned long end, unsigned long newflags);
1871 * doesn't attempt to fault and will return short.
1873 int get_user_pages_fast_only(unsigned long start, int nr_pages,
1874 unsigned int gup_flags, struct page **pages);
1875 int pin_user_pages_fast_only(unsigned long start, int nr_pages,
1876 unsigned int gup_flags, struct page **pages);
1878 static inline bool get_user_page_fast_only(unsigned long addr,
1879 unsigned int gup_flags, struct page **pagep)
1881 return get_user_pages_fast_only(addr, 1, gup_flags, pagep) == 1;
1884 * per-process(per-mm_struct) statistics.
1886 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1888 long val = atomic_long_read(&mm->rss_stat.count[member]);
1890 #ifdef SPLIT_RSS_COUNTING
1892 * counter is updated in asynchronous manner and may go to minus.
1893 * But it's never be expected number for users.
1898 return (unsigned long)val;
1901 void mm_trace_rss_stat(struct mm_struct *mm, int member, long count);
1903 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1905 long count = atomic_long_add_return(value, &mm->rss_stat.count[member]);
1907 mm_trace_rss_stat(mm, member, count);
1910 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1912 long count = atomic_long_inc_return(&mm->rss_stat.count[member]);
1914 mm_trace_rss_stat(mm, member, count);
1917 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1919 long count = atomic_long_dec_return(&mm->rss_stat.count[member]);
1921 mm_trace_rss_stat(mm, member, count);
1924 /* Optimized variant when page is already known not to be PageAnon */
1925 static inline int mm_counter_file(struct page *page)
1927 if (PageSwapBacked(page))
1928 return MM_SHMEMPAGES;
1929 return MM_FILEPAGES;
1932 static inline int mm_counter(struct page *page)
1935 return MM_ANONPAGES;
1936 return mm_counter_file(page);
1939 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1941 return get_mm_counter(mm, MM_FILEPAGES) +
1942 get_mm_counter(mm, MM_ANONPAGES) +
1943 get_mm_counter(mm, MM_SHMEMPAGES);
1946 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1948 return max(mm->hiwater_rss, get_mm_rss(mm));
1951 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1953 return max(mm->hiwater_vm, mm->total_vm);
1956 static inline void update_hiwater_rss(struct mm_struct *mm)
1958 unsigned long _rss = get_mm_rss(mm);
1960 if ((mm)->hiwater_rss < _rss)
1961 (mm)->hiwater_rss = _rss;
1964 static inline void update_hiwater_vm(struct mm_struct *mm)
1966 if (mm->hiwater_vm < mm->total_vm)
1967 mm->hiwater_vm = mm->total_vm;
1970 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1972 mm->hiwater_rss = get_mm_rss(mm);
1975 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1976 struct mm_struct *mm)
1978 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1980 if (*maxrss < hiwater_rss)
1981 *maxrss = hiwater_rss;
1984 #if defined(SPLIT_RSS_COUNTING)
1985 void sync_mm_rss(struct mm_struct *mm);
1987 static inline void sync_mm_rss(struct mm_struct *mm)
1992 #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
1993 static inline int pte_special(pte_t pte)
1998 static inline pte_t pte_mkspecial(pte_t pte)
2004 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
2005 static inline int pte_devmap(pte_t pte)
2011 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
2013 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
2015 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
2019 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
2023 #ifdef __PAGETABLE_P4D_FOLDED
2024 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2025 unsigned long address)
2030 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
2033 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
2034 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2035 unsigned long address)
2039 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
2040 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
2043 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
2045 static inline void mm_inc_nr_puds(struct mm_struct *mm)
2047 if (mm_pud_folded(mm))
2049 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2052 static inline void mm_dec_nr_puds(struct mm_struct *mm)
2054 if (mm_pud_folded(mm))
2056 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2060 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
2061 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
2062 unsigned long address)
2067 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
2068 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
2071 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
2073 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
2075 if (mm_pmd_folded(mm))
2077 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2080 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
2082 if (mm_pmd_folded(mm))
2084 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2089 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
2091 atomic_long_set(&mm->pgtables_bytes, 0);
2094 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2096 return atomic_long_read(&mm->pgtables_bytes);
2099 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
2101 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2104 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
2106 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2110 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
2111 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2116 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
2117 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
2120 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
2121 int __pte_alloc_kernel(pmd_t *pmd);
2123 #if defined(CONFIG_MMU)
2125 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2126 unsigned long address)
2128 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
2129 NULL : p4d_offset(pgd, address);
2132 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2133 unsigned long address)
2135 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
2136 NULL : pud_offset(p4d, address);
2139 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
2141 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
2142 NULL: pmd_offset(pud, address);
2144 #endif /* CONFIG_MMU */
2146 #if USE_SPLIT_PTE_PTLOCKS
2147 #if ALLOC_SPLIT_PTLOCKS
2148 void __init ptlock_cache_init(void);
2149 extern bool ptlock_alloc(struct page *page);
2150 extern void ptlock_free(struct page *page);
2152 static inline spinlock_t *ptlock_ptr(struct page *page)
2156 #else /* ALLOC_SPLIT_PTLOCKS */
2157 static inline void ptlock_cache_init(void)
2161 static inline bool ptlock_alloc(struct page *page)
2166 static inline void ptlock_free(struct page *page)
2170 static inline spinlock_t *ptlock_ptr(struct page *page)
2174 #endif /* ALLOC_SPLIT_PTLOCKS */
2176 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2178 return ptlock_ptr(pmd_page(*pmd));
2181 static inline bool ptlock_init(struct page *page)
2184 * prep_new_page() initialize page->private (and therefore page->ptl)
2185 * with 0. Make sure nobody took it in use in between.
2187 * It can happen if arch try to use slab for page table allocation:
2188 * slab code uses page->slab_cache, which share storage with page->ptl.
2190 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
2191 if (!ptlock_alloc(page))
2193 spin_lock_init(ptlock_ptr(page));
2197 #else /* !USE_SPLIT_PTE_PTLOCKS */
2199 * We use mm->page_table_lock to guard all pagetable pages of the mm.
2201 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2203 return &mm->page_table_lock;
2205 static inline void ptlock_cache_init(void) {}
2206 static inline bool ptlock_init(struct page *page) { return true; }
2207 static inline void ptlock_free(struct page *page) {}
2208 #endif /* USE_SPLIT_PTE_PTLOCKS */
2210 static inline void pgtable_init(void)
2212 ptlock_cache_init();
2213 pgtable_cache_init();
2216 static inline bool pgtable_pte_page_ctor(struct page *page)
2218 if (!ptlock_init(page))
2220 __SetPageTable(page);
2221 inc_lruvec_page_state(page, NR_PAGETABLE);
2225 static inline void pgtable_pte_page_dtor(struct page *page)
2228 __ClearPageTable(page);
2229 dec_lruvec_page_state(page, NR_PAGETABLE);
2232 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
2234 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
2235 pte_t *__pte = pte_offset_map(pmd, address); \
2241 #define pte_unmap_unlock(pte, ptl) do { \
2246 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2248 #define pte_alloc_map(mm, pmd, address) \
2249 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2251 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2252 (pte_alloc(mm, pmd) ? \
2253 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2255 #define pte_alloc_kernel(pmd, address) \
2256 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2257 NULL: pte_offset_kernel(pmd, address))
2259 #if USE_SPLIT_PMD_PTLOCKS
2261 static struct page *pmd_to_page(pmd_t *pmd)
2263 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2264 return virt_to_page((void *)((unsigned long) pmd & mask));
2267 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2269 return ptlock_ptr(pmd_to_page(pmd));
2272 static inline bool pmd_ptlock_init(struct page *page)
2274 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2275 page->pmd_huge_pte = NULL;
2277 return ptlock_init(page);
2280 static inline void pmd_ptlock_free(struct page *page)
2282 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2283 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2288 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2292 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2294 return &mm->page_table_lock;
2297 static inline bool pmd_ptlock_init(struct page *page) { return true; }
2298 static inline void pmd_ptlock_free(struct page *page) {}
2300 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2304 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2306 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2311 static inline bool pgtable_pmd_page_ctor(struct page *page)
2313 if (!pmd_ptlock_init(page))
2315 __SetPageTable(page);
2316 inc_lruvec_page_state(page, NR_PAGETABLE);
2320 static inline void pgtable_pmd_page_dtor(struct page *page)
2322 pmd_ptlock_free(page);
2323 __ClearPageTable(page);
2324 dec_lruvec_page_state(page, NR_PAGETABLE);
2328 * No scalability reason to split PUD locks yet, but follow the same pattern
2329 * as the PMD locks to make it easier if we decide to. The VM should not be
2330 * considered ready to switch to split PUD locks yet; there may be places
2331 * which need to be converted from page_table_lock.
2333 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2335 return &mm->page_table_lock;
2338 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2340 spinlock_t *ptl = pud_lockptr(mm, pud);
2346 extern void __init pagecache_init(void);
2347 extern void __init free_area_init_memoryless_node(int nid);
2348 extern void free_initmem(void);
2351 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2352 * into the buddy system. The freed pages will be poisoned with pattern
2353 * "poison" if it's within range [0, UCHAR_MAX].
2354 * Return pages freed into the buddy system.
2356 extern unsigned long free_reserved_area(void *start, void *end,
2357 int poison, const char *s);
2359 extern void adjust_managed_page_count(struct page *page, long count);
2360 extern void mem_init_print_info(const char *str);
2362 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2364 /* Free the reserved page into the buddy system, so it gets managed. */
2365 static inline void free_reserved_page(struct page *page)
2367 ClearPageReserved(page);
2368 init_page_count(page);
2370 adjust_managed_page_count(page, 1);
2372 #define free_highmem_page(page) free_reserved_page(page)
2374 static inline void mark_page_reserved(struct page *page)
2376 SetPageReserved(page);
2377 adjust_managed_page_count(page, -1);
2381 * Default method to free all the __init memory into the buddy system.
2382 * The freed pages will be poisoned with pattern "poison" if it's within
2383 * range [0, UCHAR_MAX].
2384 * Return pages freed into the buddy system.
2386 static inline unsigned long free_initmem_default(int poison)
2388 extern char __init_begin[], __init_end[];
2390 return free_reserved_area(&__init_begin, &__init_end,
2391 poison, "unused kernel");
2394 static inline unsigned long get_num_physpages(void)
2397 unsigned long phys_pages = 0;
2399 for_each_online_node(nid)
2400 phys_pages += node_present_pages(nid);
2406 * Using memblock node mappings, an architecture may initialise its
2407 * zones, allocate the backing mem_map and account for memory holes in an
2408 * architecture independent manner.
2410 * An architecture is expected to register range of page frames backed by
2411 * physical memory with memblock_add[_node]() before calling
2412 * free_area_init() passing in the PFN each zone ends at. At a basic
2413 * usage, an architecture is expected to do something like
2415 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2417 * for_each_valid_physical_page_range()
2418 * memblock_add_node(base, size, nid)
2419 * free_area_init(max_zone_pfns);
2421 void free_area_init(unsigned long *max_zone_pfn);
2422 unsigned long node_map_pfn_alignment(void);
2423 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2424 unsigned long end_pfn);
2425 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2426 unsigned long end_pfn);
2427 extern void get_pfn_range_for_nid(unsigned int nid,
2428 unsigned long *start_pfn, unsigned long *end_pfn);
2429 extern unsigned long find_min_pfn_with_active_regions(void);
2431 #ifndef CONFIG_NEED_MULTIPLE_NODES
2432 static inline int early_pfn_to_nid(unsigned long pfn)
2437 /* please see mm/page_alloc.c */
2438 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2441 extern void set_dma_reserve(unsigned long new_dma_reserve);
2442 extern void memmap_init_range(unsigned long, int, unsigned long,
2443 unsigned long, unsigned long, enum meminit_context,
2444 struct vmem_altmap *, int migratetype);
2445 extern void memmap_init_zone(struct zone *zone);
2446 extern void setup_per_zone_wmarks(void);
2447 extern int __meminit init_per_zone_wmark_min(void);
2448 extern void mem_init(void);
2449 extern void __init mmap_init(void);
2450 extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2451 extern long si_mem_available(void);
2452 extern void si_meminfo(struct sysinfo * val);
2453 extern void si_meminfo_node(struct sysinfo *val, int nid);
2454 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2455 extern unsigned long arch_reserved_kernel_pages(void);
2458 extern __printf(3, 4)
2459 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2461 extern void setup_per_cpu_pageset(void);
2464 extern int min_free_kbytes;
2465 extern int watermark_boost_factor;
2466 extern int watermark_scale_factor;
2467 extern bool arch_has_descending_max_zone_pfns(void);
2470 extern atomic_long_t mmap_pages_allocated;
2471 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2473 /* interval_tree.c */
2474 void vma_interval_tree_insert(struct vm_area_struct *node,
2475 struct rb_root_cached *root);
2476 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2477 struct vm_area_struct *prev,
2478 struct rb_root_cached *root);
2479 void vma_interval_tree_remove(struct vm_area_struct *node,
2480 struct rb_root_cached *root);
2481 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2482 unsigned long start, unsigned long last);
2483 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2484 unsigned long start, unsigned long last);
2486 #define vma_interval_tree_foreach(vma, root, start, last) \
2487 for (vma = vma_interval_tree_iter_first(root, start, last); \
2488 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2490 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2491 struct rb_root_cached *root);
2492 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2493 struct rb_root_cached *root);
2494 struct anon_vma_chain *
2495 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2496 unsigned long start, unsigned long last);
2497 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2498 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2499 #ifdef CONFIG_DEBUG_VM_RB
2500 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2503 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2504 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2505 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2508 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2509 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2510 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2511 struct vm_area_struct *expand);
2512 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2513 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2515 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2517 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2518 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2519 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2520 struct mempolicy *, struct vm_userfaultfd_ctx);
2521 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2522 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2523 unsigned long addr, int new_below);
2524 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2525 unsigned long addr, int new_below);
2526 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2527 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2528 struct rb_node **, struct rb_node *);
2529 extern void unlink_file_vma(struct vm_area_struct *);
2530 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2531 unsigned long addr, unsigned long len, pgoff_t pgoff,
2532 bool *need_rmap_locks);
2533 extern void exit_mmap(struct mm_struct *);
2535 static inline int check_data_rlimit(unsigned long rlim,
2537 unsigned long start,
2538 unsigned long end_data,
2539 unsigned long start_data)
2541 if (rlim < RLIM_INFINITY) {
2542 if (((new - start) + (end_data - start_data)) > rlim)
2549 extern int mm_take_all_locks(struct mm_struct *mm);
2550 extern void mm_drop_all_locks(struct mm_struct *mm);
2552 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2553 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2554 extern struct file *get_task_exe_file(struct task_struct *task);
2556 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2557 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2559 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2560 const struct vm_special_mapping *sm);
2561 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2562 unsigned long addr, unsigned long len,
2563 unsigned long flags,
2564 const struct vm_special_mapping *spec);
2565 /* This is an obsolete alternative to _install_special_mapping. */
2566 extern int install_special_mapping(struct mm_struct *mm,
2567 unsigned long addr, unsigned long len,
2568 unsigned long flags, struct page **pages);
2570 unsigned long randomize_stack_top(unsigned long stack_top);
2572 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2574 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2575 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2576 struct list_head *uf);
2577 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2578 unsigned long len, unsigned long prot, unsigned long flags,
2579 unsigned long pgoff, unsigned long *populate, struct list_head *uf);
2580 extern int __do_munmap(struct mm_struct *, unsigned long, size_t,
2581 struct list_head *uf, bool downgrade);
2582 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2583 struct list_head *uf);
2584 extern int do_madvise(struct mm_struct *mm, unsigned long start, size_t len_in, int behavior);
2587 extern int __mm_populate(unsigned long addr, unsigned long len,
2589 static inline void mm_populate(unsigned long addr, unsigned long len)
2592 (void) __mm_populate(addr, len, 1);
2595 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2598 /* These take the mm semaphore themselves */
2599 extern int __must_check vm_brk(unsigned long, unsigned long);
2600 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2601 extern int vm_munmap(unsigned long, size_t);
2602 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2603 unsigned long, unsigned long,
2604 unsigned long, unsigned long);
2606 struct vm_unmapped_area_info {
2607 #define VM_UNMAPPED_AREA_TOPDOWN 1
2608 unsigned long flags;
2609 unsigned long length;
2610 unsigned long low_limit;
2611 unsigned long high_limit;
2612 unsigned long align_mask;
2613 unsigned long align_offset;
2616 extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info);
2619 extern void truncate_inode_pages(struct address_space *, loff_t);
2620 extern void truncate_inode_pages_range(struct address_space *,
2621 loff_t lstart, loff_t lend);
2622 extern void truncate_inode_pages_final(struct address_space *);
2624 /* generic vm_area_ops exported for stackable file systems */
2625 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2626 extern vm_fault_t filemap_map_pages(struct vm_fault *vmf,
2627 pgoff_t start_pgoff, pgoff_t end_pgoff);
2628 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2630 /* mm/page-writeback.c */
2631 int __must_check write_one_page(struct page *page);
2632 void task_dirty_inc(struct task_struct *tsk);
2634 extern unsigned long stack_guard_gap;
2635 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2636 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2638 /* CONFIG_STACK_GROWSUP still needs to grow downwards at some places */
2639 extern int expand_downwards(struct vm_area_struct *vma,
2640 unsigned long address);
2642 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2644 #define expand_upwards(vma, address) (0)
2647 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2648 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2649 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2650 struct vm_area_struct **pprev);
2652 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2653 NULL if none. Assume start_addr < end_addr. */
2654 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2656 struct vm_area_struct * vma = find_vma(mm,start_addr);
2658 if (vma && end_addr <= vma->vm_start)
2663 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2665 unsigned long vm_start = vma->vm_start;
2667 if (vma->vm_flags & VM_GROWSDOWN) {
2668 vm_start -= stack_guard_gap;
2669 if (vm_start > vma->vm_start)
2675 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2677 unsigned long vm_end = vma->vm_end;
2679 if (vma->vm_flags & VM_GROWSUP) {
2680 vm_end += stack_guard_gap;
2681 if (vm_end < vma->vm_end)
2682 vm_end = -PAGE_SIZE;
2687 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2689 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2692 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2693 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2694 unsigned long vm_start, unsigned long vm_end)
2696 struct vm_area_struct *vma = find_vma(mm, vm_start);
2698 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2704 static inline bool range_in_vma(struct vm_area_struct *vma,
2705 unsigned long start, unsigned long end)
2707 return (vma && vma->vm_start <= start && end <= vma->vm_end);
2711 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2712 void vma_set_page_prot(struct vm_area_struct *vma);
2714 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2718 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2720 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2724 void vma_set_file(struct vm_area_struct *vma, struct file *file);
2726 #ifdef CONFIG_NUMA_BALANCING
2727 unsigned long change_prot_numa(struct vm_area_struct *vma,
2728 unsigned long start, unsigned long end);
2731 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2732 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2733 unsigned long pfn, unsigned long size, pgprot_t);
2734 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2735 int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
2736 struct page **pages, unsigned long *num);
2737 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
2739 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
2741 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2743 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2744 unsigned long pfn, pgprot_t pgprot);
2745 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2747 vm_fault_t vmf_insert_mixed_prot(struct vm_area_struct *vma, unsigned long addr,
2748 pfn_t pfn, pgprot_t pgprot);
2749 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
2750 unsigned long addr, pfn_t pfn);
2751 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2753 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2754 unsigned long addr, struct page *page)
2756 int err = vm_insert_page(vma, addr, page);
2759 return VM_FAULT_OOM;
2760 if (err < 0 && err != -EBUSY)
2761 return VM_FAULT_SIGBUS;
2763 return VM_FAULT_NOPAGE;
2766 #ifndef io_remap_pfn_range
2767 static inline int io_remap_pfn_range(struct vm_area_struct *vma,
2768 unsigned long addr, unsigned long pfn,
2769 unsigned long size, pgprot_t prot)
2771 return remap_pfn_range(vma, addr, pfn, size, pgprot_decrypted(prot));
2775 static inline vm_fault_t vmf_error(int err)
2778 return VM_FAULT_OOM;
2779 return VM_FAULT_SIGBUS;
2782 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
2783 unsigned int foll_flags);
2785 #define FOLL_WRITE 0x01 /* check pte is writable */
2786 #define FOLL_TOUCH 0x02 /* mark page accessed */
2787 #define FOLL_GET 0x04 /* do get_page on page */
2788 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2789 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2790 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2791 * and return without waiting upon it */
2792 #define FOLL_POPULATE 0x40 /* fault in page */
2793 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2794 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2795 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2796 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2797 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2798 #define FOLL_MLOCK 0x1000 /* lock present pages */
2799 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2800 #define FOLL_COW 0x4000 /* internal GUP flag */
2801 #define FOLL_ANON 0x8000 /* don't do file mappings */
2802 #define FOLL_LONGTERM 0x10000 /* mapping lifetime is indefinite: see below */
2803 #define FOLL_SPLIT_PMD 0x20000 /* split huge pmd before returning */
2804 #define FOLL_PIN 0x40000 /* pages must be released via unpin_user_page */
2805 #define FOLL_FAST_ONLY 0x80000 /* gup_fast: prevent fall-back to slow gup */
2808 * FOLL_PIN and FOLL_LONGTERM may be used in various combinations with each
2809 * other. Here is what they mean, and how to use them:
2811 * FOLL_LONGTERM indicates that the page will be held for an indefinite time
2812 * period _often_ under userspace control. This is in contrast to
2813 * iov_iter_get_pages(), whose usages are transient.
2815 * FIXME: For pages which are part of a filesystem, mappings are subject to the
2816 * lifetime enforced by the filesystem and we need guarantees that longterm
2817 * users like RDMA and V4L2 only establish mappings which coordinate usage with
2818 * the filesystem. Ideas for this coordination include revoking the longterm
2819 * pin, delaying writeback, bounce buffer page writeback, etc. As FS DAX was
2820 * added after the problem with filesystems was found FS DAX VMAs are
2821 * specifically failed. Filesystem pages are still subject to bugs and use of
2822 * FOLL_LONGTERM should be avoided on those pages.
2824 * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call.
2825 * Currently only get_user_pages() and get_user_pages_fast() support this flag
2826 * and calls to get_user_pages_[un]locked are specifically not allowed. This
2827 * is due to an incompatibility with the FS DAX check and
2828 * FAULT_FLAG_ALLOW_RETRY.
2830 * In the CMA case: long term pins in a CMA region would unnecessarily fragment
2831 * that region. And so, CMA attempts to migrate the page before pinning, when
2832 * FOLL_LONGTERM is specified.
2834 * FOLL_PIN indicates that a special kind of tracking (not just page->_refcount,
2835 * but an additional pin counting system) will be invoked. This is intended for
2836 * anything that gets a page reference and then touches page data (for example,
2837 * Direct IO). This lets the filesystem know that some non-file-system entity is
2838 * potentially changing the pages' data. In contrast to FOLL_GET (whose pages
2839 * are released via put_page()), FOLL_PIN pages must be released, ultimately, by
2840 * a call to unpin_user_page().
2842 * FOLL_PIN is similar to FOLL_GET: both of these pin pages. They use different
2843 * and separate refcounting mechanisms, however, and that means that each has
2844 * its own acquire and release mechanisms:
2846 * FOLL_GET: get_user_pages*() to acquire, and put_page() to release.
2848 * FOLL_PIN: pin_user_pages*() to acquire, and unpin_user_pages to release.
2850 * FOLL_PIN and FOLL_GET are mutually exclusive for a given function call.
2851 * (The underlying pages may experience both FOLL_GET-based and FOLL_PIN-based
2852 * calls applied to them, and that's perfectly OK. This is a constraint on the
2853 * callers, not on the pages.)
2855 * FOLL_PIN should be set internally by the pin_user_pages*() APIs, never
2856 * directly by the caller. That's in order to help avoid mismatches when
2857 * releasing pages: get_user_pages*() pages must be released via put_page(),
2858 * while pin_user_pages*() pages must be released via unpin_user_page().
2860 * Please see Documentation/core-api/pin_user_pages.rst for more information.
2863 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
2865 if (vm_fault & VM_FAULT_OOM)
2867 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2868 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2869 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2874 typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
2875 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2876 unsigned long size, pte_fn_t fn, void *data);
2877 extern int apply_to_existing_page_range(struct mm_struct *mm,
2878 unsigned long address, unsigned long size,
2879 pte_fn_t fn, void *data);
2881 extern void init_mem_debugging_and_hardening(void);
2882 #ifdef CONFIG_PAGE_POISONING
2883 extern void __kernel_poison_pages(struct page *page, int numpages);
2884 extern void __kernel_unpoison_pages(struct page *page, int numpages);
2885 extern bool _page_poisoning_enabled_early;
2886 DECLARE_STATIC_KEY_FALSE(_page_poisoning_enabled);
2887 static inline bool page_poisoning_enabled(void)
2889 return _page_poisoning_enabled_early;
2892 * For use in fast paths after init_mem_debugging() has run, or when a
2893 * false negative result is not harmful when called too early.
2895 static inline bool page_poisoning_enabled_static(void)
2897 return static_branch_unlikely(&_page_poisoning_enabled);
2899 static inline void kernel_poison_pages(struct page *page, int numpages)
2901 if (page_poisoning_enabled_static())
2902 __kernel_poison_pages(page, numpages);
2904 static inline void kernel_unpoison_pages(struct page *page, int numpages)
2906 if (page_poisoning_enabled_static())
2907 __kernel_unpoison_pages(page, numpages);
2910 static inline bool page_poisoning_enabled(void) { return false; }
2911 static inline bool page_poisoning_enabled_static(void) { return false; }
2912 static inline void __kernel_poison_pages(struct page *page, int nunmpages) { }
2913 static inline void kernel_poison_pages(struct page *page, int numpages) { }
2914 static inline void kernel_unpoison_pages(struct page *page, int numpages) { }
2917 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_ALLOC_DEFAULT_ON, init_on_alloc);
2918 static inline bool want_init_on_alloc(gfp_t flags)
2920 if (static_branch_maybe(CONFIG_INIT_ON_ALLOC_DEFAULT_ON,
2923 return flags & __GFP_ZERO;
2926 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_FREE_DEFAULT_ON, init_on_free);
2927 static inline bool want_init_on_free(void)
2929 return static_branch_maybe(CONFIG_INIT_ON_FREE_DEFAULT_ON,
2933 extern bool _debug_pagealloc_enabled_early;
2934 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
2936 static inline bool debug_pagealloc_enabled(void)
2938 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
2939 _debug_pagealloc_enabled_early;
2943 * For use in fast paths after init_debug_pagealloc() has run, or when a
2944 * false negative result is not harmful when called too early.
2946 static inline bool debug_pagealloc_enabled_static(void)
2948 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
2951 return static_branch_unlikely(&_debug_pagealloc_enabled);
2954 #ifdef CONFIG_DEBUG_PAGEALLOC
2956 * To support DEBUG_PAGEALLOC architecture must ensure that
2957 * __kernel_map_pages() never fails
2959 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2961 static inline void debug_pagealloc_map_pages(struct page *page, int numpages)
2963 if (debug_pagealloc_enabled_static())
2964 __kernel_map_pages(page, numpages, 1);
2967 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages)
2969 if (debug_pagealloc_enabled_static())
2970 __kernel_map_pages(page, numpages, 0);
2972 #else /* CONFIG_DEBUG_PAGEALLOC */
2973 static inline void debug_pagealloc_map_pages(struct page *page, int numpages) {}
2974 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages) {}
2975 #endif /* CONFIG_DEBUG_PAGEALLOC */
2977 #ifdef __HAVE_ARCH_GATE_AREA
2978 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2979 extern int in_gate_area_no_mm(unsigned long addr);
2980 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2982 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2986 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2987 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2991 #endif /* __HAVE_ARCH_GATE_AREA */
2993 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2995 #ifdef CONFIG_SYSCTL
2996 extern int sysctl_drop_caches;
2997 int drop_caches_sysctl_handler(struct ctl_table *, int, void *, size_t *,
3001 void drop_slab(void);
3002 void drop_slab_node(int nid);
3005 #define randomize_va_space 0
3007 extern int randomize_va_space;
3010 const char * arch_vma_name(struct vm_area_struct *vma);
3012 void print_vma_addr(char *prefix, unsigned long rip);
3014 static inline void print_vma_addr(char *prefix, unsigned long rip)
3019 void *sparse_buffer_alloc(unsigned long size);
3020 struct page * __populate_section_memmap(unsigned long pfn,
3021 unsigned long nr_pages, int nid, struct vmem_altmap *altmap);
3022 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
3023 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
3024 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
3025 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
3026 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node,
3027 struct vmem_altmap *altmap);
3028 void *vmemmap_alloc_block(unsigned long size, int node);
3030 void *vmemmap_alloc_block_buf(unsigned long size, int node,
3031 struct vmem_altmap *altmap);
3032 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
3033 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
3034 int node, struct vmem_altmap *altmap);
3035 int vmemmap_populate(unsigned long start, unsigned long end, int node,
3036 struct vmem_altmap *altmap);
3037 void vmemmap_populate_print_last(void);
3038 #ifdef CONFIG_MEMORY_HOTPLUG
3039 void vmemmap_free(unsigned long start, unsigned long end,
3040 struct vmem_altmap *altmap);
3042 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
3043 unsigned long nr_pages);
3046 MF_COUNT_INCREASED = 1 << 0,
3047 MF_ACTION_REQUIRED = 1 << 1,
3048 MF_MUST_KILL = 1 << 2,
3049 MF_SOFT_OFFLINE = 1 << 3,
3051 extern int memory_failure(unsigned long pfn, int flags);
3052 extern void memory_failure_queue(unsigned long pfn, int flags);
3053 extern void memory_failure_queue_kick(int cpu);
3054 extern int unpoison_memory(unsigned long pfn);
3055 extern int sysctl_memory_failure_early_kill;
3056 extern int sysctl_memory_failure_recovery;
3057 extern void shake_page(struct page *p, int access);
3058 extern atomic_long_t num_poisoned_pages __read_mostly;
3059 extern int soft_offline_page(unsigned long pfn, int flags);
3063 * Error handlers for various types of pages.
3066 MF_IGNORED, /* Error: cannot be handled */
3067 MF_FAILED, /* Error: handling failed */
3068 MF_DELAYED, /* Will be handled later */
3069 MF_RECOVERED, /* Successfully recovered */
3072 enum mf_action_page_type {
3074 MF_MSG_KERNEL_HIGH_ORDER,
3076 MF_MSG_DIFFERENT_COMPOUND,
3077 MF_MSG_POISONED_HUGE,
3080 MF_MSG_NON_PMD_HUGE,
3081 MF_MSG_UNMAP_FAILED,
3082 MF_MSG_DIRTY_SWAPCACHE,
3083 MF_MSG_CLEAN_SWAPCACHE,
3084 MF_MSG_DIRTY_MLOCKED_LRU,
3085 MF_MSG_CLEAN_MLOCKED_LRU,
3086 MF_MSG_DIRTY_UNEVICTABLE_LRU,
3087 MF_MSG_CLEAN_UNEVICTABLE_LRU,
3090 MF_MSG_TRUNCATED_LRU,
3098 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
3099 extern void clear_huge_page(struct page *page,
3100 unsigned long addr_hint,
3101 unsigned int pages_per_huge_page);
3102 extern void copy_user_huge_page(struct page *dst, struct page *src,
3103 unsigned long addr_hint,
3104 struct vm_area_struct *vma,
3105 unsigned int pages_per_huge_page);
3106 extern long copy_huge_page_from_user(struct page *dst_page,
3107 const void __user *usr_src,
3108 unsigned int pages_per_huge_page,
3109 bool allow_pagefault);
3112 * vma_is_special_huge - Are transhuge page-table entries considered special?
3113 * @vma: Pointer to the struct vm_area_struct to consider
3115 * Whether transhuge page-table entries are considered "special" following
3116 * the definition in vm_normal_page().
3118 * Return: true if transhuge page-table entries should be considered special,
3121 static inline bool vma_is_special_huge(const struct vm_area_struct *vma)
3123 return vma_is_dax(vma) || (vma->vm_file &&
3124 (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP)));
3127 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
3129 #ifdef CONFIG_DEBUG_PAGEALLOC
3130 extern unsigned int _debug_guardpage_minorder;
3131 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
3133 static inline unsigned int debug_guardpage_minorder(void)
3135 return _debug_guardpage_minorder;
3138 static inline bool debug_guardpage_enabled(void)
3140 return static_branch_unlikely(&_debug_guardpage_enabled);
3143 static inline bool page_is_guard(struct page *page)
3145 if (!debug_guardpage_enabled())
3148 return PageGuard(page);
3151 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
3152 static inline bool debug_guardpage_enabled(void) { return false; }
3153 static inline bool page_is_guard(struct page *page) { return false; }
3154 #endif /* CONFIG_DEBUG_PAGEALLOC */
3156 #if MAX_NUMNODES > 1
3157 void __init setup_nr_node_ids(void);
3159 static inline void setup_nr_node_ids(void) {}
3162 extern int memcmp_pages(struct page *page1, struct page *page2);
3164 static inline int pages_identical(struct page *page1, struct page *page2)
3166 return !memcmp_pages(page1, page2);
3169 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
3170 unsigned long clean_record_shared_mapping_range(struct address_space *mapping,
3171 pgoff_t first_index, pgoff_t nr,
3172 pgoff_t bitmap_pgoff,
3173 unsigned long *bitmap,
3177 unsigned long wp_shared_mapping_range(struct address_space *mapping,
3178 pgoff_t first_index, pgoff_t nr);
3181 extern int sysctl_nr_trim_pages;
3183 #ifdef CONFIG_PRINTK
3184 void mem_dump_obj(void *object);
3186 static inline void mem_dump_obj(void *object) {}
3189 #endif /* __KERNEL__ */
3190 #endif /* _LINUX_MM_H */