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_ref.h>
28 #include <linux/memremap.h>
29 #include <linux/overflow.h>
30 #include <linux/sizes.h>
31 #include <linux/sched.h>
32 #include <linux/pgtable.h>
36 struct anon_vma_chain;
39 struct writeback_control;
42 void init_mm_internals(void);
44 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
45 extern unsigned long max_mapnr;
47 static inline void set_max_mapnr(unsigned long limit)
52 static inline void set_max_mapnr(unsigned long limit) { }
55 extern atomic_long_t _totalram_pages;
56 static inline unsigned long totalram_pages(void)
58 return (unsigned long)atomic_long_read(&_totalram_pages);
61 static inline void totalram_pages_inc(void)
63 atomic_long_inc(&_totalram_pages);
66 static inline void totalram_pages_dec(void)
68 atomic_long_dec(&_totalram_pages);
71 static inline void totalram_pages_add(long count)
73 atomic_long_add(count, &_totalram_pages);
76 extern void * high_memory;
77 extern int page_cluster;
80 extern int sysctl_legacy_va_layout;
82 #define sysctl_legacy_va_layout 0
85 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
86 extern const int mmap_rnd_bits_min;
87 extern const int mmap_rnd_bits_max;
88 extern int mmap_rnd_bits __read_mostly;
90 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
91 extern const int mmap_rnd_compat_bits_min;
92 extern const int mmap_rnd_compat_bits_max;
93 extern int mmap_rnd_compat_bits __read_mostly;
97 #include <asm/processor.h>
100 * Architectures that support memory tagging (assigning tags to memory regions,
101 * embedding these tags into addresses that point to these memory regions, and
102 * checking that the memory and the pointer tags match on memory accesses)
103 * redefine this macro to strip tags from pointers.
104 * It's defined as noop for arcitectures that don't support memory tagging.
106 #ifndef untagged_addr
107 #define untagged_addr(addr) (addr)
111 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
115 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
119 #define lm_alias(x) __va(__pa_symbol(x))
123 * To prevent common memory management code establishing
124 * a zero page mapping on a read fault.
125 * This macro should be defined within <asm/pgtable.h>.
126 * s390 does this to prevent multiplexing of hardware bits
127 * related to the physical page in case of virtualization.
129 #ifndef mm_forbids_zeropage
130 #define mm_forbids_zeropage(X) (0)
134 * On some architectures it is expensive to call memset() for small sizes.
135 * If an architecture decides to implement their own version of
136 * mm_zero_struct_page they should wrap the defines below in a #ifndef and
137 * define their own version of this macro in <asm/pgtable.h>
139 #if BITS_PER_LONG == 64
140 /* This function must be updated when the size of struct page grows above 80
141 * or reduces below 56. The idea that compiler optimizes out switch()
142 * statement, and only leaves move/store instructions. Also the compiler can
143 * combine write statments if they are both assignments and can be reordered,
144 * this can result in several of the writes here being dropped.
146 #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
147 static inline void __mm_zero_struct_page(struct page *page)
149 unsigned long *_pp = (void *)page;
151 /* Check that struct page is either 56, 64, 72, or 80 bytes */
152 BUILD_BUG_ON(sizeof(struct page) & 7);
153 BUILD_BUG_ON(sizeof(struct page) < 56);
154 BUILD_BUG_ON(sizeof(struct page) > 80);
156 switch (sizeof(struct page)) {
158 _pp[9] = 0; /* fallthrough */
160 _pp[8] = 0; /* fallthrough */
162 _pp[7] = 0; /* fallthrough */
174 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
178 * Default maximum number of active map areas, this limits the number of vmas
179 * per mm struct. Users can overwrite this number by sysctl but there is a
182 * When a program's coredump is generated as ELF format, a section is created
183 * per a vma. In ELF, the number of sections is represented in unsigned short.
184 * This means the number of sections should be smaller than 65535 at coredump.
185 * Because the kernel adds some informative sections to a image of program at
186 * generating coredump, we need some margin. The number of extra sections is
187 * 1-3 now and depends on arch. We use "5" as safe margin, here.
189 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
190 * not a hard limit any more. Although some userspace tools can be surprised by
193 #define MAPCOUNT_ELF_CORE_MARGIN (5)
194 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
196 extern int sysctl_max_map_count;
198 extern unsigned long sysctl_user_reserve_kbytes;
199 extern unsigned long sysctl_admin_reserve_kbytes;
201 extern int sysctl_overcommit_memory;
202 extern int sysctl_overcommit_ratio;
203 extern unsigned long sysctl_overcommit_kbytes;
205 int overcommit_ratio_handler(struct ctl_table *, int, void *, size_t *,
207 int overcommit_kbytes_handler(struct ctl_table *, int, void *, size_t *,
209 int overcommit_policy_handler(struct ctl_table *, int, void *, size_t *,
212 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
214 /* to align the pointer to the (next) page boundary */
215 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
217 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
218 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
220 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
223 * Linux kernel virtual memory manager primitives.
224 * The idea being to have a "virtual" mm in the same way
225 * we have a virtual fs - giving a cleaner interface to the
226 * mm details, and allowing different kinds of memory mappings
227 * (from shared memory to executable loading to arbitrary
231 struct vm_area_struct *vm_area_alloc(struct mm_struct *);
232 struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
233 void vm_area_free(struct vm_area_struct *);
236 extern struct rb_root nommu_region_tree;
237 extern struct rw_semaphore nommu_region_sem;
239 extern unsigned int kobjsize(const void *objp);
243 * vm_flags in vm_area_struct, see mm_types.h.
244 * When changing, update also include/trace/events/mmflags.h
246 #define VM_NONE 0x00000000
248 #define VM_READ 0x00000001 /* currently active flags */
249 #define VM_WRITE 0x00000002
250 #define VM_EXEC 0x00000004
251 #define VM_SHARED 0x00000008
253 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
254 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
255 #define VM_MAYWRITE 0x00000020
256 #define VM_MAYEXEC 0x00000040
257 #define VM_MAYSHARE 0x00000080
259 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
260 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
261 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
262 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
263 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
265 #define VM_LOCKED 0x00002000
266 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
268 /* Used by sys_madvise() */
269 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
270 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
272 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
273 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
274 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
275 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
276 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
277 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
278 #define VM_SYNC 0x00800000 /* Synchronous page faults */
279 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
280 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
281 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
283 #ifdef CONFIG_MEM_SOFT_DIRTY
284 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
286 # define VM_SOFTDIRTY 0
289 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
290 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
291 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
292 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
294 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
295 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
296 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
297 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
298 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
299 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
300 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
301 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
302 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
303 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
304 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
305 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
307 #ifdef CONFIG_ARCH_HAS_PKEYS
308 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
309 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
310 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
311 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
312 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
314 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
316 # define VM_PKEY_BIT4 0
318 #endif /* CONFIG_ARCH_HAS_PKEYS */
320 #if defined(CONFIG_X86)
321 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
322 #elif defined(CONFIG_PARISC)
323 # define VM_GROWSUP VM_ARCH_1
324 #elif defined(CONFIG_IA64)
325 # define VM_GROWSUP VM_ARCH_1
326 #elif defined(CONFIG_SPARC64)
327 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
328 # define VM_ARCH_CLEAR VM_SPARC_ADI
329 #elif defined(CONFIG_ARM64)
330 # define VM_ARM64_BTI VM_ARCH_1 /* BTI guarded page, a.k.a. GP bit */
331 # define VM_ARCH_CLEAR VM_ARM64_BTI
332 #elif !defined(CONFIG_MMU)
333 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
337 # define VM_GROWSUP VM_NONE
340 /* Bits set in the VMA until the stack is in its final location */
341 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
343 #define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0)
345 /* Common data flag combinations */
346 #define VM_DATA_FLAGS_TSK_EXEC (VM_READ | VM_WRITE | TASK_EXEC | \
347 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
348 #define VM_DATA_FLAGS_NON_EXEC (VM_READ | VM_WRITE | VM_MAYREAD | \
349 VM_MAYWRITE | VM_MAYEXEC)
350 #define VM_DATA_FLAGS_EXEC (VM_READ | VM_WRITE | VM_EXEC | \
351 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
353 #ifndef VM_DATA_DEFAULT_FLAGS /* arch can override this */
354 #define VM_DATA_DEFAULT_FLAGS VM_DATA_FLAGS_EXEC
357 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
358 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
361 #ifdef CONFIG_STACK_GROWSUP
362 #define VM_STACK VM_GROWSUP
364 #define VM_STACK VM_GROWSDOWN
367 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
369 /* VMA basic access permission flags */
370 #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
374 * Special vmas that are non-mergable, non-mlock()able.
376 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
378 /* This mask prevents VMA from being scanned with khugepaged */
379 #define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB)
381 /* This mask defines which mm->def_flags a process can inherit its parent */
382 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
384 /* This mask is used to clear all the VMA flags used by mlock */
385 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
387 /* Arch-specific flags to clear when updating VM flags on protection change */
388 #ifndef VM_ARCH_CLEAR
389 # define VM_ARCH_CLEAR VM_NONE
391 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
394 * mapping from the currently active vm_flags protection bits (the
395 * low four bits) to a page protection mask..
397 extern pgprot_t protection_map[16];
400 * Fault flag definitions.
402 * @FAULT_FLAG_WRITE: Fault was a write fault.
403 * @FAULT_FLAG_MKWRITE: Fault was mkwrite of existing PTE.
404 * @FAULT_FLAG_ALLOW_RETRY: Allow to retry the fault if blocked.
405 * @FAULT_FLAG_RETRY_NOWAIT: Don't drop mmap_lock and wait when retrying.
406 * @FAULT_FLAG_KILLABLE: The fault task is in SIGKILL killable region.
407 * @FAULT_FLAG_TRIED: The fault has been tried once.
408 * @FAULT_FLAG_USER: The fault originated in userspace.
409 * @FAULT_FLAG_REMOTE: The fault is not for current task/mm.
410 * @FAULT_FLAG_INSTRUCTION: The fault was during an instruction fetch.
411 * @FAULT_FLAG_INTERRUPTIBLE: The fault can be interrupted by non-fatal signals.
413 * About @FAULT_FLAG_ALLOW_RETRY and @FAULT_FLAG_TRIED: we can specify
414 * whether we would allow page faults to retry by specifying these two
415 * fault flags correctly. Currently there can be three legal combinations:
417 * (a) ALLOW_RETRY and !TRIED: this means the page fault allows retry, and
418 * this is the first try
420 * (b) ALLOW_RETRY and TRIED: this means the page fault allows retry, and
421 * we've already tried at least once
423 * (c) !ALLOW_RETRY and !TRIED: this means the page fault does not allow retry
425 * The unlisted combination (!ALLOW_RETRY && TRIED) is illegal and should never
426 * be used. Note that page faults can be allowed to retry for multiple times,
427 * in which case we'll have an initial fault with flags (a) then later on
428 * continuous faults with flags (b). We should always try to detect pending
429 * signals before a retry to make sure the continuous page faults can still be
430 * interrupted if necessary.
432 #define FAULT_FLAG_WRITE 0x01
433 #define FAULT_FLAG_MKWRITE 0x02
434 #define FAULT_FLAG_ALLOW_RETRY 0x04
435 #define FAULT_FLAG_RETRY_NOWAIT 0x08
436 #define FAULT_FLAG_KILLABLE 0x10
437 #define FAULT_FLAG_TRIED 0x20
438 #define FAULT_FLAG_USER 0x40
439 #define FAULT_FLAG_REMOTE 0x80
440 #define FAULT_FLAG_INSTRUCTION 0x100
441 #define FAULT_FLAG_INTERRUPTIBLE 0x200
444 * The default fault flags that should be used by most of the
445 * arch-specific page fault handlers.
447 #define FAULT_FLAG_DEFAULT (FAULT_FLAG_ALLOW_RETRY | \
448 FAULT_FLAG_KILLABLE | \
449 FAULT_FLAG_INTERRUPTIBLE)
452 * fault_flag_allow_retry_first - check ALLOW_RETRY the first time
454 * This is mostly used for places where we want to try to avoid taking
455 * the mmap_lock for too long a time when waiting for another condition
456 * to change, in which case we can try to be polite to release the
457 * mmap_lock in the first round to avoid potential starvation of other
458 * processes that would also want the mmap_lock.
460 * Return: true if the page fault allows retry and this is the first
461 * attempt of the fault handling; false otherwise.
463 static inline bool fault_flag_allow_retry_first(unsigned int flags)
465 return (flags & FAULT_FLAG_ALLOW_RETRY) &&
466 (!(flags & FAULT_FLAG_TRIED));
469 #define FAULT_FLAG_TRACE \
470 { FAULT_FLAG_WRITE, "WRITE" }, \
471 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
472 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
473 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
474 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
475 { FAULT_FLAG_TRIED, "TRIED" }, \
476 { FAULT_FLAG_USER, "USER" }, \
477 { FAULT_FLAG_REMOTE, "REMOTE" }, \
478 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }, \
479 { FAULT_FLAG_INTERRUPTIBLE, "INTERRUPTIBLE" }
482 * vm_fault is filled by the the pagefault handler and passed to the vma's
483 * ->fault function. The vma's ->fault is responsible for returning a bitmask
484 * of VM_FAULT_xxx flags that give details about how the fault was handled.
486 * MM layer fills up gfp_mask for page allocations but fault handler might
487 * alter it if its implementation requires a different allocation context.
489 * pgoff should be used in favour of virtual_address, if possible.
492 struct vm_area_struct *vma; /* Target VMA */
493 unsigned int flags; /* FAULT_FLAG_xxx flags */
494 gfp_t gfp_mask; /* gfp mask to be used for allocations */
495 pgoff_t pgoff; /* Logical page offset based on vma */
496 unsigned long address; /* Faulting virtual address */
497 pmd_t *pmd; /* Pointer to pmd entry matching
499 pud_t *pud; /* Pointer to pud entry matching
502 pte_t orig_pte; /* Value of PTE at the time of fault */
504 struct page *cow_page; /* Page handler may use for COW fault */
505 struct page *page; /* ->fault handlers should return a
506 * page here, unless VM_FAULT_NOPAGE
507 * is set (which is also implied by
510 /* These three entries are valid only while holding ptl lock */
511 pte_t *pte; /* Pointer to pte entry matching
512 * the 'address'. NULL if the page
513 * table hasn't been allocated.
515 spinlock_t *ptl; /* Page table lock.
516 * Protects pte page table if 'pte'
517 * is not NULL, otherwise pmd.
519 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
520 * vm_ops->map_pages() calls
521 * alloc_set_pte() from atomic context.
522 * do_fault_around() pre-allocates
523 * page table to avoid allocation from
528 /* page entry size for vm->huge_fault() */
529 enum page_entry_size {
536 * These are the virtual MM functions - opening of an area, closing and
537 * unmapping it (needed to keep files on disk up-to-date etc), pointer
538 * to the functions called when a no-page or a wp-page exception occurs.
540 struct vm_operations_struct {
541 void (*open)(struct vm_area_struct * area);
542 void (*close)(struct vm_area_struct * area);
543 int (*split)(struct vm_area_struct * area, unsigned long addr);
544 int (*mremap)(struct vm_area_struct * area);
545 vm_fault_t (*fault)(struct vm_fault *vmf);
546 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
547 enum page_entry_size pe_size);
548 void (*map_pages)(struct vm_fault *vmf,
549 pgoff_t start_pgoff, pgoff_t end_pgoff);
550 unsigned long (*pagesize)(struct vm_area_struct * area);
552 /* notification that a previously read-only page is about to become
553 * writable, if an error is returned it will cause a SIGBUS */
554 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
556 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
557 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
559 /* called by access_process_vm when get_user_pages() fails, typically
560 * for use by special VMAs that can switch between memory and hardware
562 int (*access)(struct vm_area_struct *vma, unsigned long addr,
563 void *buf, int len, int write);
565 /* Called by the /proc/PID/maps code to ask the vma whether it
566 * has a special name. Returning non-NULL will also cause this
567 * vma to be dumped unconditionally. */
568 const char *(*name)(struct vm_area_struct *vma);
572 * set_policy() op must add a reference to any non-NULL @new mempolicy
573 * to hold the policy upon return. Caller should pass NULL @new to
574 * remove a policy and fall back to surrounding context--i.e. do not
575 * install a MPOL_DEFAULT policy, nor the task or system default
578 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
581 * get_policy() op must add reference [mpol_get()] to any policy at
582 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
583 * in mm/mempolicy.c will do this automatically.
584 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
585 * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
586 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
587 * must return NULL--i.e., do not "fallback" to task or system default
590 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
594 * Called by vm_normal_page() for special PTEs to find the
595 * page for @addr. This is useful if the default behavior
596 * (using pte_page()) would not find the correct page.
598 struct page *(*find_special_page)(struct vm_area_struct *vma,
602 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
604 static const struct vm_operations_struct dummy_vm_ops = {};
606 memset(vma, 0, sizeof(*vma));
608 vma->vm_ops = &dummy_vm_ops;
609 INIT_LIST_HEAD(&vma->anon_vma_chain);
612 static inline void vma_set_anonymous(struct vm_area_struct *vma)
617 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
622 static inline bool vma_is_temporary_stack(struct vm_area_struct *vma)
624 int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP);
629 if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) ==
630 VM_STACK_INCOMPLETE_SETUP)
636 static inline bool vma_is_foreign(struct vm_area_struct *vma)
641 if (current->mm != vma->vm_mm)
647 static inline bool vma_is_accessible(struct vm_area_struct *vma)
649 return vma->vm_flags & VM_ACCESS_FLAGS;
654 * The vma_is_shmem is not inline because it is used only by slow
655 * paths in userfault.
657 bool vma_is_shmem(struct vm_area_struct *vma);
659 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
662 int vma_is_stack_for_current(struct vm_area_struct *vma);
664 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
665 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
671 * FIXME: take this include out, include page-flags.h in
672 * files which need it (119 of them)
674 #include <linux/page-flags.h>
675 #include <linux/huge_mm.h>
678 * Methods to modify the page usage count.
680 * What counts for a page usage:
681 * - cache mapping (page->mapping)
682 * - private data (page->private)
683 * - page mapped in a task's page tables, each mapping
684 * is counted separately
686 * Also, many kernel routines increase the page count before a critical
687 * routine so they can be sure the page doesn't go away from under them.
691 * Drop a ref, return true if the refcount fell to zero (the page has no users)
693 static inline int put_page_testzero(struct page *page)
695 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
696 return page_ref_dec_and_test(page);
700 * Try to grab a ref unless the page has a refcount of zero, return false if
702 * This can be called when MMU is off so it must not access
703 * any of the virtual mappings.
705 static inline int get_page_unless_zero(struct page *page)
707 return page_ref_add_unless(page, 1, 0);
710 extern int page_is_ram(unsigned long pfn);
718 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
721 /* Support for virtually mapped pages */
722 struct page *vmalloc_to_page(const void *addr);
723 unsigned long vmalloc_to_pfn(const void *addr);
726 * Determine if an address is within the vmalloc range
728 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
729 * is no special casing required.
732 #ifndef is_ioremap_addr
733 #define is_ioremap_addr(x) is_vmalloc_addr(x)
737 extern bool is_vmalloc_addr(const void *x);
738 extern int is_vmalloc_or_module_addr(const void *x);
740 static inline bool is_vmalloc_addr(const void *x)
744 static inline int is_vmalloc_or_module_addr(const void *x)
750 extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
751 static inline void *kvmalloc(size_t size, gfp_t flags)
753 return kvmalloc_node(size, flags, NUMA_NO_NODE);
755 static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
757 return kvmalloc_node(size, flags | __GFP_ZERO, node);
759 static inline void *kvzalloc(size_t size, gfp_t flags)
761 return kvmalloc(size, flags | __GFP_ZERO);
764 static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
768 if (unlikely(check_mul_overflow(n, size, &bytes)))
771 return kvmalloc(bytes, flags);
774 static inline void *kvcalloc(size_t n, size_t size, gfp_t flags)
776 return kvmalloc_array(n, size, flags | __GFP_ZERO);
779 extern void kvfree(const void *addr);
780 extern void kvfree_sensitive(const void *addr, size_t len);
782 static inline int head_mapcount(struct page *head)
784 return atomic_read(compound_mapcount_ptr(head)) + 1;
788 * Mapcount of compound page as a whole, does not include mapped sub-pages.
790 * Must be called only for compound pages or any their tail sub-pages.
792 static inline int compound_mapcount(struct page *page)
794 VM_BUG_ON_PAGE(!PageCompound(page), page);
795 page = compound_head(page);
796 return head_mapcount(page);
800 * The atomic page->_mapcount, starts from -1: so that transitions
801 * both from it and to it can be tracked, using atomic_inc_and_test
802 * and atomic_add_negative(-1).
804 static inline void page_mapcount_reset(struct page *page)
806 atomic_set(&(page)->_mapcount, -1);
809 int __page_mapcount(struct page *page);
812 * Mapcount of 0-order page; when compound sub-page, includes
813 * compound_mapcount().
815 * Result is undefined for pages which cannot be mapped into userspace.
816 * For example SLAB or special types of pages. See function page_has_type().
817 * They use this place in struct page differently.
819 static inline int page_mapcount(struct page *page)
821 if (unlikely(PageCompound(page)))
822 return __page_mapcount(page);
823 return atomic_read(&page->_mapcount) + 1;
826 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
827 int total_mapcount(struct page *page);
828 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
830 static inline int total_mapcount(struct page *page)
832 return page_mapcount(page);
834 static inline int page_trans_huge_mapcount(struct page *page,
837 int mapcount = page_mapcount(page);
839 *total_mapcount = mapcount;
844 static inline struct page *virt_to_head_page(const void *x)
846 struct page *page = virt_to_page(x);
848 return compound_head(page);
851 void __put_page(struct page *page);
853 void put_pages_list(struct list_head *pages);
855 void split_page(struct page *page, unsigned int order);
858 * Compound pages have a destructor function. Provide a
859 * prototype for that function and accessor functions.
860 * These are _only_ valid on the head of a compound page.
862 typedef void compound_page_dtor(struct page *);
864 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
865 enum compound_dtor_id {
868 #ifdef CONFIG_HUGETLB_PAGE
871 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
876 extern compound_page_dtor * const compound_page_dtors[NR_COMPOUND_DTORS];
878 static inline void set_compound_page_dtor(struct page *page,
879 enum compound_dtor_id compound_dtor)
881 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
882 page[1].compound_dtor = compound_dtor;
885 static inline void destroy_compound_page(struct page *page)
887 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
888 compound_page_dtors[page[1].compound_dtor](page);
891 static inline unsigned int compound_order(struct page *page)
895 return page[1].compound_order;
898 static inline bool hpage_pincount_available(struct page *page)
901 * Can the page->hpage_pinned_refcount field be used? That field is in
902 * the 3rd page of the compound page, so the smallest (2-page) compound
903 * pages cannot support it.
905 page = compound_head(page);
906 return PageCompound(page) && compound_order(page) > 1;
909 static inline int head_pincount(struct page *head)
911 return atomic_read(compound_pincount_ptr(head));
914 static inline int compound_pincount(struct page *page)
916 VM_BUG_ON_PAGE(!hpage_pincount_available(page), page);
917 page = compound_head(page);
918 return head_pincount(page);
921 static inline void set_compound_order(struct page *page, unsigned int order)
923 page[1].compound_order = order;
926 /* Returns the number of pages in this potentially compound page. */
927 static inline unsigned long compound_nr(struct page *page)
929 return 1UL << compound_order(page);
932 /* Returns the number of bytes in this potentially compound page. */
933 static inline unsigned long page_size(struct page *page)
935 return PAGE_SIZE << compound_order(page);
938 /* Returns the number of bits needed for the number of bytes in a page */
939 static inline unsigned int page_shift(struct page *page)
941 return PAGE_SHIFT + compound_order(page);
944 void free_compound_page(struct page *page);
948 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
949 * servicing faults for write access. In the normal case, do always want
950 * pte_mkwrite. But get_user_pages can cause write faults for mappings
951 * that do not have writing enabled, when used by access_process_vm.
953 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
955 if (likely(vma->vm_flags & VM_WRITE))
956 pte = pte_mkwrite(pte);
960 vm_fault_t alloc_set_pte(struct vm_fault *vmf, struct page *page);
961 vm_fault_t finish_fault(struct vm_fault *vmf);
962 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
966 * Multiple processes may "see" the same page. E.g. for untouched
967 * mappings of /dev/null, all processes see the same page full of
968 * zeroes, and text pages of executables and shared libraries have
969 * only one copy in memory, at most, normally.
971 * For the non-reserved pages, page_count(page) denotes a reference count.
972 * page_count() == 0 means the page is free. page->lru is then used for
973 * freelist management in the buddy allocator.
974 * page_count() > 0 means the page has been allocated.
976 * Pages are allocated by the slab allocator in order to provide memory
977 * to kmalloc and kmem_cache_alloc. In this case, the management of the
978 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
979 * unless a particular usage is carefully commented. (the responsibility of
980 * freeing the kmalloc memory is the caller's, of course).
982 * A page may be used by anyone else who does a __get_free_page().
983 * In this case, page_count still tracks the references, and should only
984 * be used through the normal accessor functions. The top bits of page->flags
985 * and page->virtual store page management information, but all other fields
986 * are unused and could be used privately, carefully. The management of this
987 * page is the responsibility of the one who allocated it, and those who have
988 * subsequently been given references to it.
990 * The other pages (we may call them "pagecache pages") are completely
991 * managed by the Linux memory manager: I/O, buffers, swapping etc.
992 * The following discussion applies only to them.
994 * A pagecache page contains an opaque `private' member, which belongs to the
995 * page's address_space. Usually, this is the address of a circular list of
996 * the page's disk buffers. PG_private must be set to tell the VM to call
997 * into the filesystem to release these pages.
999 * A page may belong to an inode's memory mapping. In this case, page->mapping
1000 * is the pointer to the inode, and page->index is the file offset of the page,
1001 * in units of PAGE_SIZE.
1003 * If pagecache pages are not associated with an inode, they are said to be
1004 * anonymous pages. These may become associated with the swapcache, and in that
1005 * case PG_swapcache is set, and page->private is an offset into the swapcache.
1007 * In either case (swapcache or inode backed), the pagecache itself holds one
1008 * reference to the page. Setting PG_private should also increment the
1009 * refcount. The each user mapping also has a reference to the page.
1011 * The pagecache pages are stored in a per-mapping radix tree, which is
1012 * rooted at mapping->i_pages, and indexed by offset.
1013 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
1014 * lists, we instead now tag pages as dirty/writeback in the radix tree.
1016 * All pagecache pages may be subject to I/O:
1017 * - inode pages may need to be read from disk,
1018 * - inode pages which have been modified and are MAP_SHARED may need
1019 * to be written back to the inode on disk,
1020 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
1021 * modified may need to be swapped out to swap space and (later) to be read
1026 * The zone field is never updated after free_area_init_core()
1027 * sets it, so none of the operations on it need to be atomic.
1030 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
1031 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
1032 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
1033 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
1034 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
1035 #define KASAN_TAG_PGOFF (LAST_CPUPID_PGOFF - KASAN_TAG_WIDTH)
1038 * Define the bit shifts to access each section. For non-existent
1039 * sections we define the shift as 0; that plus a 0 mask ensures
1040 * the compiler will optimise away reference to them.
1042 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
1043 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
1044 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
1045 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
1046 #define KASAN_TAG_PGSHIFT (KASAN_TAG_PGOFF * (KASAN_TAG_WIDTH != 0))
1048 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
1049 #ifdef NODE_NOT_IN_PAGE_FLAGS
1050 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
1051 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
1052 SECTIONS_PGOFF : ZONES_PGOFF)
1054 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
1055 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
1056 NODES_PGOFF : ZONES_PGOFF)
1059 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
1061 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
1062 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
1063 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
1064 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
1065 #define KASAN_TAG_MASK ((1UL << KASAN_TAG_WIDTH) - 1)
1066 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
1068 static inline enum zone_type page_zonenum(const struct page *page)
1070 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
1073 #ifdef CONFIG_ZONE_DEVICE
1074 static inline bool is_zone_device_page(const struct page *page)
1076 return page_zonenum(page) == ZONE_DEVICE;
1078 extern void memmap_init_zone_device(struct zone *, unsigned long,
1079 unsigned long, struct dev_pagemap *);
1081 static inline bool is_zone_device_page(const struct page *page)
1087 #ifdef CONFIG_DEV_PAGEMAP_OPS
1088 void free_devmap_managed_page(struct page *page);
1089 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
1091 static inline bool page_is_devmap_managed(struct page *page)
1093 if (!static_branch_unlikely(&devmap_managed_key))
1095 if (!is_zone_device_page(page))
1097 switch (page->pgmap->type) {
1098 case MEMORY_DEVICE_PRIVATE:
1099 case MEMORY_DEVICE_FS_DAX:
1107 void put_devmap_managed_page(struct page *page);
1109 #else /* CONFIG_DEV_PAGEMAP_OPS */
1110 static inline bool page_is_devmap_managed(struct page *page)
1115 static inline void put_devmap_managed_page(struct page *page)
1118 #endif /* CONFIG_DEV_PAGEMAP_OPS */
1120 static inline bool is_device_private_page(const struct page *page)
1122 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
1123 IS_ENABLED(CONFIG_DEVICE_PRIVATE) &&
1124 is_zone_device_page(page) &&
1125 page->pgmap->type == MEMORY_DEVICE_PRIVATE;
1128 static inline bool is_pci_p2pdma_page(const struct page *page)
1130 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
1131 IS_ENABLED(CONFIG_PCI_P2PDMA) &&
1132 is_zone_device_page(page) &&
1133 page->pgmap->type == MEMORY_DEVICE_PCI_P2PDMA;
1136 /* 127: arbitrary random number, small enough to assemble well */
1137 #define page_ref_zero_or_close_to_overflow(page) \
1138 ((unsigned int) page_ref_count(page) + 127u <= 127u)
1140 static inline void get_page(struct page *page)
1142 page = compound_head(page);
1144 * Getting a normal page or the head of a compound page
1145 * requires to already have an elevated page->_refcount.
1147 VM_BUG_ON_PAGE(page_ref_zero_or_close_to_overflow(page), page);
1151 bool __must_check try_grab_page(struct page *page, unsigned int flags);
1153 static inline __must_check bool try_get_page(struct page *page)
1155 page = compound_head(page);
1156 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1162 static inline void put_page(struct page *page)
1164 page = compound_head(page);
1167 * For devmap managed pages we need to catch refcount transition from
1168 * 2 to 1, when refcount reach one it means the page is free and we
1169 * need to inform the device driver through callback. See
1170 * include/linux/memremap.h and HMM for details.
1172 if (page_is_devmap_managed(page)) {
1173 put_devmap_managed_page(page);
1177 if (put_page_testzero(page))
1182 * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
1183 * the page's refcount so that two separate items are tracked: the original page
1184 * reference count, and also a new count of how many pin_user_pages() calls were
1185 * made against the page. ("gup-pinned" is another term for the latter).
1187 * With this scheme, pin_user_pages() becomes special: such pages are marked as
1188 * distinct from normal pages. As such, the unpin_user_page() call (and its
1189 * variants) must be used in order to release gup-pinned pages.
1193 * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
1194 * counts with respect to pin_user_pages() and unpin_user_page() becomes
1195 * simpler, due to the fact that adding an even power of two to the page
1196 * refcount has the effect of using only the upper N bits, for the code that
1197 * counts up using the bias value. This means that the lower bits are left for
1198 * the exclusive use of the original code that increments and decrements by one
1199 * (or at least, by much smaller values than the bias value).
1201 * Of course, once the lower bits overflow into the upper bits (and this is
1202 * OK, because subtraction recovers the original values), then visual inspection
1203 * no longer suffices to directly view the separate counts. However, for normal
1204 * applications that don't have huge page reference counts, this won't be an
1207 * Locking: the lockless algorithm described in page_cache_get_speculative()
1208 * and page_cache_gup_pin_speculative() provides safe operation for
1209 * get_user_pages and page_mkclean and other calls that race to set up page
1212 #define GUP_PIN_COUNTING_BIAS (1U << 10)
1214 void unpin_user_page(struct page *page);
1215 void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1217 void unpin_user_pages(struct page **pages, unsigned long npages);
1220 * page_maybe_dma_pinned() - report if a page is pinned for DMA.
1222 * This function checks if a page has been pinned via a call to
1223 * pin_user_pages*().
1225 * For non-huge pages, the return value is partially fuzzy: false is not fuzzy,
1226 * because it means "definitely not pinned for DMA", but true means "probably
1227 * pinned for DMA, but possibly a false positive due to having at least
1228 * GUP_PIN_COUNTING_BIAS worth of normal page references".
1230 * False positives are OK, because: a) it's unlikely for a page to get that many
1231 * refcounts, and b) all the callers of this routine are expected to be able to
1232 * deal gracefully with a false positive.
1234 * For huge pages, the result will be exactly correct. That's because we have
1235 * more tracking data available: the 3rd struct page in the compound page is
1236 * used to track the pincount (instead using of the GUP_PIN_COUNTING_BIAS
1239 * For more information, please see Documentation/core-api/pin_user_pages.rst.
1241 * @page: pointer to page to be queried.
1242 * @Return: True, if it is likely that the page has been "dma-pinned".
1243 * False, if the page is definitely not dma-pinned.
1245 static inline bool page_maybe_dma_pinned(struct page *page)
1247 if (hpage_pincount_available(page))
1248 return compound_pincount(page) > 0;
1251 * page_ref_count() is signed. If that refcount overflows, then
1252 * page_ref_count() returns a negative value, and callers will avoid
1253 * further incrementing the refcount.
1255 * Here, for that overflow case, use the signed bit to count a little
1256 * bit higher via unsigned math, and thus still get an accurate result.
1258 return ((unsigned int)page_ref_count(compound_head(page))) >=
1259 GUP_PIN_COUNTING_BIAS;
1262 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1263 #define SECTION_IN_PAGE_FLAGS
1267 * The identification function is mainly used by the buddy allocator for
1268 * determining if two pages could be buddies. We are not really identifying
1269 * the zone since we could be using the section number id if we do not have
1270 * node id available in page flags.
1271 * We only guarantee that it will return the same value for two combinable
1274 static inline int page_zone_id(struct page *page)
1276 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1279 #ifdef NODE_NOT_IN_PAGE_FLAGS
1280 extern int page_to_nid(const struct page *page);
1282 static inline int page_to_nid(const struct page *page)
1284 struct page *p = (struct page *)page;
1286 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1290 #ifdef CONFIG_NUMA_BALANCING
1291 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1293 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1296 static inline int cpupid_to_pid(int cpupid)
1298 return cpupid & LAST__PID_MASK;
1301 static inline int cpupid_to_cpu(int cpupid)
1303 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1306 static inline int cpupid_to_nid(int cpupid)
1308 return cpu_to_node(cpupid_to_cpu(cpupid));
1311 static inline bool cpupid_pid_unset(int cpupid)
1313 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1316 static inline bool cpupid_cpu_unset(int cpupid)
1318 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1321 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1323 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1326 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1327 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1328 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1330 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1333 static inline int page_cpupid_last(struct page *page)
1335 return page->_last_cpupid;
1337 static inline void page_cpupid_reset_last(struct page *page)
1339 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1342 static inline int page_cpupid_last(struct page *page)
1344 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1347 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1349 static inline void page_cpupid_reset_last(struct page *page)
1351 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1353 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1354 #else /* !CONFIG_NUMA_BALANCING */
1355 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1357 return page_to_nid(page); /* XXX */
1360 static inline int page_cpupid_last(struct page *page)
1362 return page_to_nid(page); /* XXX */
1365 static inline int cpupid_to_nid(int cpupid)
1370 static inline int cpupid_to_pid(int cpupid)
1375 static inline int cpupid_to_cpu(int cpupid)
1380 static inline int cpu_pid_to_cpupid(int nid, int pid)
1385 static inline bool cpupid_pid_unset(int cpupid)
1390 static inline void page_cpupid_reset_last(struct page *page)
1394 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1398 #endif /* CONFIG_NUMA_BALANCING */
1400 #ifdef CONFIG_KASAN_SW_TAGS
1401 static inline u8 page_kasan_tag(const struct page *page)
1403 return (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1406 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1408 page->flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1409 page->flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1412 static inline void page_kasan_tag_reset(struct page *page)
1414 page_kasan_tag_set(page, 0xff);
1417 static inline u8 page_kasan_tag(const struct page *page)
1422 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1423 static inline void page_kasan_tag_reset(struct page *page) { }
1426 static inline struct zone *page_zone(const struct page *page)
1428 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1431 static inline pg_data_t *page_pgdat(const struct page *page)
1433 return NODE_DATA(page_to_nid(page));
1436 #ifdef SECTION_IN_PAGE_FLAGS
1437 static inline void set_page_section(struct page *page, unsigned long section)
1439 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1440 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1443 static inline unsigned long page_to_section(const struct page *page)
1445 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1449 static inline void set_page_zone(struct page *page, enum zone_type zone)
1451 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1452 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1455 static inline void set_page_node(struct page *page, unsigned long node)
1457 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1458 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1461 static inline void set_page_links(struct page *page, enum zone_type zone,
1462 unsigned long node, unsigned long pfn)
1464 set_page_zone(page, zone);
1465 set_page_node(page, node);
1466 #ifdef SECTION_IN_PAGE_FLAGS
1467 set_page_section(page, pfn_to_section_nr(pfn));
1472 static inline struct mem_cgroup *page_memcg(struct page *page)
1474 return page->mem_cgroup;
1476 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1478 WARN_ON_ONCE(!rcu_read_lock_held());
1479 return READ_ONCE(page->mem_cgroup);
1482 static inline struct mem_cgroup *page_memcg(struct page *page)
1486 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1488 WARN_ON_ONCE(!rcu_read_lock_held());
1494 * Some inline functions in vmstat.h depend on page_zone()
1496 #include <linux/vmstat.h>
1498 static __always_inline void *lowmem_page_address(const struct page *page)
1500 return page_to_virt(page);
1503 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1504 #define HASHED_PAGE_VIRTUAL
1507 #if defined(WANT_PAGE_VIRTUAL)
1508 static inline void *page_address(const struct page *page)
1510 return page->virtual;
1512 static inline void set_page_address(struct page *page, void *address)
1514 page->virtual = address;
1516 #define page_address_init() do { } while(0)
1519 #if defined(HASHED_PAGE_VIRTUAL)
1520 void *page_address(const struct page *page);
1521 void set_page_address(struct page *page, void *virtual);
1522 void page_address_init(void);
1525 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1526 #define page_address(page) lowmem_page_address(page)
1527 #define set_page_address(page, address) do { } while(0)
1528 #define page_address_init() do { } while(0)
1531 extern void *page_rmapping(struct page *page);
1532 extern struct anon_vma *page_anon_vma(struct page *page);
1533 extern struct address_space *page_mapping(struct page *page);
1535 extern struct address_space *__page_file_mapping(struct page *);
1538 struct address_space *page_file_mapping(struct page *page)
1540 if (unlikely(PageSwapCache(page)))
1541 return __page_file_mapping(page);
1543 return page->mapping;
1546 extern pgoff_t __page_file_index(struct page *page);
1549 * Return the pagecache index of the passed page. Regular pagecache pages
1550 * use ->index whereas swapcache pages use swp_offset(->private)
1552 static inline pgoff_t page_index(struct page *page)
1554 if (unlikely(PageSwapCache(page)))
1555 return __page_file_index(page);
1559 bool page_mapped(struct page *page);
1560 struct address_space *page_mapping(struct page *page);
1561 struct address_space *page_mapping_file(struct page *page);
1564 * Return true only if the page has been allocated with
1565 * ALLOC_NO_WATERMARKS and the low watermark was not
1566 * met implying that the system is under some pressure.
1568 static inline bool page_is_pfmemalloc(struct page *page)
1571 * Page index cannot be this large so this must be
1572 * a pfmemalloc page.
1574 return page->index == -1UL;
1578 * Only to be called by the page allocator on a freshly allocated
1581 static inline void set_page_pfmemalloc(struct page *page)
1586 static inline void clear_page_pfmemalloc(struct page *page)
1592 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1594 extern void pagefault_out_of_memory(void);
1596 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1599 * Flags passed to show_mem() and show_free_areas() to suppress output in
1602 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1604 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1607 extern bool can_do_mlock(void);
1609 static inline bool can_do_mlock(void) { return false; }
1611 extern int user_shm_lock(size_t, struct user_struct *);
1612 extern void user_shm_unlock(size_t, struct user_struct *);
1615 * Parameter block passed down to zap_pte_range in exceptional cases.
1617 struct zap_details {
1618 struct address_space *check_mapping; /* Check page->mapping if set */
1619 pgoff_t first_index; /* Lowest page->index to unmap */
1620 pgoff_t last_index; /* Highest page->index to unmap */
1623 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1625 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1628 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1629 unsigned long size);
1630 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1631 unsigned long size);
1632 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1633 unsigned long start, unsigned long end);
1635 struct mmu_notifier_range;
1637 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1638 unsigned long end, unsigned long floor, unsigned long ceiling);
1639 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1640 struct vm_area_struct *vma);
1641 int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
1642 struct mmu_notifier_range *range,
1643 pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp);
1644 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1645 unsigned long *pfn);
1646 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1647 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1648 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1649 void *buf, int len, int write);
1651 extern void truncate_pagecache(struct inode *inode, loff_t new);
1652 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1653 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1654 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1655 int truncate_inode_page(struct address_space *mapping, struct page *page);
1656 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1657 int invalidate_inode_page(struct page *page);
1660 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1661 unsigned long address, unsigned int flags);
1662 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1663 unsigned long address, unsigned int fault_flags,
1665 void unmap_mapping_pages(struct address_space *mapping,
1666 pgoff_t start, pgoff_t nr, bool even_cows);
1667 void unmap_mapping_range(struct address_space *mapping,
1668 loff_t const holebegin, loff_t const holelen, int even_cows);
1670 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1671 unsigned long address, unsigned int flags)
1673 /* should never happen if there's no MMU */
1675 return VM_FAULT_SIGBUS;
1677 static inline int fixup_user_fault(struct task_struct *tsk,
1678 struct mm_struct *mm, unsigned long address,
1679 unsigned int fault_flags, bool *unlocked)
1681 /* should never happen if there's no MMU */
1685 static inline void unmap_mapping_pages(struct address_space *mapping,
1686 pgoff_t start, pgoff_t nr, bool even_cows) { }
1687 static inline void unmap_mapping_range(struct address_space *mapping,
1688 loff_t const holebegin, loff_t const holelen, int even_cows) { }
1691 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1692 loff_t const holebegin, loff_t const holelen)
1694 unmap_mapping_range(mapping, holebegin, holelen, 0);
1697 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1698 void *buf, int len, unsigned int gup_flags);
1699 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1700 void *buf, int len, unsigned int gup_flags);
1701 extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1702 unsigned long addr, void *buf, int len, unsigned int gup_flags);
1704 long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1705 unsigned long start, unsigned long nr_pages,
1706 unsigned int gup_flags, struct page **pages,
1707 struct vm_area_struct **vmas, int *locked);
1708 long pin_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1709 unsigned long start, unsigned long nr_pages,
1710 unsigned int gup_flags, struct page **pages,
1711 struct vm_area_struct **vmas, int *locked);
1712 long get_user_pages(unsigned long start, unsigned long nr_pages,
1713 unsigned int gup_flags, struct page **pages,
1714 struct vm_area_struct **vmas);
1715 long pin_user_pages(unsigned long start, unsigned long nr_pages,
1716 unsigned int gup_flags, struct page **pages,
1717 struct vm_area_struct **vmas);
1718 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1719 unsigned int gup_flags, struct page **pages, int *locked);
1720 long pin_user_pages_locked(unsigned long start, unsigned long nr_pages,
1721 unsigned int gup_flags, struct page **pages, int *locked);
1722 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1723 struct page **pages, unsigned int gup_flags);
1724 long pin_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1725 struct page **pages, unsigned int gup_flags);
1727 int get_user_pages_fast(unsigned long start, int nr_pages,
1728 unsigned int gup_flags, struct page **pages);
1729 int pin_user_pages_fast(unsigned long start, int nr_pages,
1730 unsigned int gup_flags, struct page **pages);
1732 int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
1733 int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
1734 struct task_struct *task, bool bypass_rlim);
1736 /* Container for pinned pfns / pages */
1737 struct frame_vector {
1738 unsigned int nr_allocated; /* Number of frames we have space for */
1739 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1740 bool got_ref; /* Did we pin pages by getting page ref? */
1741 bool is_pfns; /* Does array contain pages or pfns? */
1742 void *ptrs[]; /* Array of pinned pfns / pages. Use
1743 * pfns_vector_pages() or pfns_vector_pfns()
1747 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1748 void frame_vector_destroy(struct frame_vector *vec);
1749 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1750 unsigned int gup_flags, struct frame_vector *vec);
1751 void put_vaddr_frames(struct frame_vector *vec);
1752 int frame_vector_to_pages(struct frame_vector *vec);
1753 void frame_vector_to_pfns(struct frame_vector *vec);
1755 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1757 return vec->nr_frames;
1760 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1763 int err = frame_vector_to_pages(vec);
1766 return ERR_PTR(err);
1768 return (struct page **)(vec->ptrs);
1771 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1774 frame_vector_to_pfns(vec);
1775 return (unsigned long *)(vec->ptrs);
1779 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1780 struct page **pages);
1781 int get_kernel_page(unsigned long start, int write, struct page **pages);
1782 struct page *get_dump_page(unsigned long addr);
1784 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1785 extern void do_invalidatepage(struct page *page, unsigned int offset,
1786 unsigned int length);
1788 void __set_page_dirty(struct page *, struct address_space *, int warn);
1789 int __set_page_dirty_nobuffers(struct page *page);
1790 int __set_page_dirty_no_writeback(struct page *page);
1791 int redirty_page_for_writepage(struct writeback_control *wbc,
1793 void account_page_dirtied(struct page *page, struct address_space *mapping);
1794 void account_page_cleaned(struct page *page, struct address_space *mapping,
1795 struct bdi_writeback *wb);
1796 int set_page_dirty(struct page *page);
1797 int set_page_dirty_lock(struct page *page);
1798 void __cancel_dirty_page(struct page *page);
1799 static inline void cancel_dirty_page(struct page *page)
1801 /* Avoid atomic ops, locking, etc. when not actually needed. */
1802 if (PageDirty(page))
1803 __cancel_dirty_page(page);
1805 int clear_page_dirty_for_io(struct page *page);
1807 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1809 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1810 unsigned long old_addr, struct vm_area_struct *new_vma,
1811 unsigned long new_addr, unsigned long len,
1812 bool need_rmap_locks);
1815 * Flags used by change_protection(). For now we make it a bitmap so
1816 * that we can pass in multiple flags just like parameters. However
1817 * for now all the callers are only use one of the flags at the same
1820 /* Whether we should allow dirty bit accounting */
1821 #define MM_CP_DIRTY_ACCT (1UL << 0)
1822 /* Whether this protection change is for NUMA hints */
1823 #define MM_CP_PROT_NUMA (1UL << 1)
1824 /* Whether this change is for write protecting */
1825 #define MM_CP_UFFD_WP (1UL << 2) /* do wp */
1826 #define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */
1827 #define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \
1828 MM_CP_UFFD_WP_RESOLVE)
1830 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1831 unsigned long end, pgprot_t newprot,
1832 unsigned long cp_flags);
1833 extern int mprotect_fixup(struct vm_area_struct *vma,
1834 struct vm_area_struct **pprev, unsigned long start,
1835 unsigned long end, unsigned long newflags);
1838 * doesn't attempt to fault and will return short.
1840 int get_user_pages_fast_only(unsigned long start, int nr_pages,
1841 unsigned int gup_flags, struct page **pages);
1842 int pin_user_pages_fast_only(unsigned long start, int nr_pages,
1843 unsigned int gup_flags, struct page **pages);
1845 static inline bool get_user_page_fast_only(unsigned long addr,
1846 unsigned int gup_flags, struct page **pagep)
1848 return get_user_pages_fast_only(addr, 1, gup_flags, pagep) == 1;
1851 * per-process(per-mm_struct) statistics.
1853 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1855 long val = atomic_long_read(&mm->rss_stat.count[member]);
1857 #ifdef SPLIT_RSS_COUNTING
1859 * counter is updated in asynchronous manner and may go to minus.
1860 * But it's never be expected number for users.
1865 return (unsigned long)val;
1868 void mm_trace_rss_stat(struct mm_struct *mm, int member, long count);
1870 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1872 long count = atomic_long_add_return(value, &mm->rss_stat.count[member]);
1874 mm_trace_rss_stat(mm, member, count);
1877 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1879 long count = atomic_long_inc_return(&mm->rss_stat.count[member]);
1881 mm_trace_rss_stat(mm, member, count);
1884 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1886 long count = atomic_long_dec_return(&mm->rss_stat.count[member]);
1888 mm_trace_rss_stat(mm, member, count);
1891 /* Optimized variant when page is already known not to be PageAnon */
1892 static inline int mm_counter_file(struct page *page)
1894 if (PageSwapBacked(page))
1895 return MM_SHMEMPAGES;
1896 return MM_FILEPAGES;
1899 static inline int mm_counter(struct page *page)
1902 return MM_ANONPAGES;
1903 return mm_counter_file(page);
1906 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1908 return get_mm_counter(mm, MM_FILEPAGES) +
1909 get_mm_counter(mm, MM_ANONPAGES) +
1910 get_mm_counter(mm, MM_SHMEMPAGES);
1913 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1915 return max(mm->hiwater_rss, get_mm_rss(mm));
1918 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1920 return max(mm->hiwater_vm, mm->total_vm);
1923 static inline void update_hiwater_rss(struct mm_struct *mm)
1925 unsigned long _rss = get_mm_rss(mm);
1927 if ((mm)->hiwater_rss < _rss)
1928 (mm)->hiwater_rss = _rss;
1931 static inline void update_hiwater_vm(struct mm_struct *mm)
1933 if (mm->hiwater_vm < mm->total_vm)
1934 mm->hiwater_vm = mm->total_vm;
1937 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1939 mm->hiwater_rss = get_mm_rss(mm);
1942 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1943 struct mm_struct *mm)
1945 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1947 if (*maxrss < hiwater_rss)
1948 *maxrss = hiwater_rss;
1951 #if defined(SPLIT_RSS_COUNTING)
1952 void sync_mm_rss(struct mm_struct *mm);
1954 static inline void sync_mm_rss(struct mm_struct *mm)
1959 #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
1960 static inline int pte_special(pte_t pte)
1965 static inline pte_t pte_mkspecial(pte_t pte)
1971 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
1972 static inline int pte_devmap(pte_t pte)
1978 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1980 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1982 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1986 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1990 #ifdef __PAGETABLE_P4D_FOLDED
1991 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1992 unsigned long address)
1997 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
2000 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
2001 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2002 unsigned long address)
2006 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
2007 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
2010 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
2012 static inline void mm_inc_nr_puds(struct mm_struct *mm)
2014 if (mm_pud_folded(mm))
2016 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2019 static inline void mm_dec_nr_puds(struct mm_struct *mm)
2021 if (mm_pud_folded(mm))
2023 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2027 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
2028 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
2029 unsigned long address)
2034 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
2035 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
2038 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
2040 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
2042 if (mm_pmd_folded(mm))
2044 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2047 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
2049 if (mm_pmd_folded(mm))
2051 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2056 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
2058 atomic_long_set(&mm->pgtables_bytes, 0);
2061 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2063 return atomic_long_read(&mm->pgtables_bytes);
2066 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
2068 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2071 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
2073 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2077 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
2078 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2083 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
2084 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
2087 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
2088 int __pte_alloc_kernel(pmd_t *pmd);
2090 #if defined(CONFIG_MMU)
2092 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2093 unsigned long address)
2095 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
2096 NULL : p4d_offset(pgd, address);
2099 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2100 unsigned long address)
2102 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
2103 NULL : pud_offset(p4d, address);
2106 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
2108 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
2109 NULL: pmd_offset(pud, address);
2111 #endif /* CONFIG_MMU */
2113 #if USE_SPLIT_PTE_PTLOCKS
2114 #if ALLOC_SPLIT_PTLOCKS
2115 void __init ptlock_cache_init(void);
2116 extern bool ptlock_alloc(struct page *page);
2117 extern void ptlock_free(struct page *page);
2119 static inline spinlock_t *ptlock_ptr(struct page *page)
2123 #else /* ALLOC_SPLIT_PTLOCKS */
2124 static inline void ptlock_cache_init(void)
2128 static inline bool ptlock_alloc(struct page *page)
2133 static inline void ptlock_free(struct page *page)
2137 static inline spinlock_t *ptlock_ptr(struct page *page)
2141 #endif /* ALLOC_SPLIT_PTLOCKS */
2143 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2145 return ptlock_ptr(pmd_page(*pmd));
2148 static inline bool ptlock_init(struct page *page)
2151 * prep_new_page() initialize page->private (and therefore page->ptl)
2152 * with 0. Make sure nobody took it in use in between.
2154 * It can happen if arch try to use slab for page table allocation:
2155 * slab code uses page->slab_cache, which share storage with page->ptl.
2157 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
2158 if (!ptlock_alloc(page))
2160 spin_lock_init(ptlock_ptr(page));
2164 #else /* !USE_SPLIT_PTE_PTLOCKS */
2166 * We use mm->page_table_lock to guard all pagetable pages of the mm.
2168 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2170 return &mm->page_table_lock;
2172 static inline void ptlock_cache_init(void) {}
2173 static inline bool ptlock_init(struct page *page) { return true; }
2174 static inline void ptlock_free(struct page *page) {}
2175 #endif /* USE_SPLIT_PTE_PTLOCKS */
2177 static inline void pgtable_init(void)
2179 ptlock_cache_init();
2180 pgtable_cache_init();
2183 static inline bool pgtable_pte_page_ctor(struct page *page)
2185 if (!ptlock_init(page))
2187 __SetPageTable(page);
2188 inc_zone_page_state(page, NR_PAGETABLE);
2192 static inline void pgtable_pte_page_dtor(struct page *page)
2195 __ClearPageTable(page);
2196 dec_zone_page_state(page, NR_PAGETABLE);
2199 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
2201 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
2202 pte_t *__pte = pte_offset_map(pmd, address); \
2208 #define pte_unmap_unlock(pte, ptl) do { \
2213 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2215 #define pte_alloc_map(mm, pmd, address) \
2216 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2218 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2219 (pte_alloc(mm, pmd) ? \
2220 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2222 #define pte_alloc_kernel(pmd, address) \
2223 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2224 NULL: pte_offset_kernel(pmd, address))
2226 #if USE_SPLIT_PMD_PTLOCKS
2228 static struct page *pmd_to_page(pmd_t *pmd)
2230 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2231 return virt_to_page((void *)((unsigned long) pmd & mask));
2234 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2236 return ptlock_ptr(pmd_to_page(pmd));
2239 static inline bool pgtable_pmd_page_ctor(struct page *page)
2241 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2242 page->pmd_huge_pte = NULL;
2244 return ptlock_init(page);
2247 static inline void pgtable_pmd_page_dtor(struct page *page)
2249 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2250 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2255 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2259 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2261 return &mm->page_table_lock;
2264 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
2265 static inline void pgtable_pmd_page_dtor(struct page *page) {}
2267 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2271 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2273 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2279 * No scalability reason to split PUD locks yet, but follow the same pattern
2280 * as the PMD locks to make it easier if we decide to. The VM should not be
2281 * considered ready to switch to split PUD locks yet; there may be places
2282 * which need to be converted from page_table_lock.
2284 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2286 return &mm->page_table_lock;
2289 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2291 spinlock_t *ptl = pud_lockptr(mm, pud);
2297 extern void __init pagecache_init(void);
2298 extern void __init free_area_init_memoryless_node(int nid);
2299 extern void free_initmem(void);
2302 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2303 * into the buddy system. The freed pages will be poisoned with pattern
2304 * "poison" if it's within range [0, UCHAR_MAX].
2305 * Return pages freed into the buddy system.
2307 extern unsigned long free_reserved_area(void *start, void *end,
2308 int poison, const char *s);
2310 #ifdef CONFIG_HIGHMEM
2312 * Free a highmem page into the buddy system, adjusting totalhigh_pages
2313 * and totalram_pages.
2315 extern void free_highmem_page(struct page *page);
2318 extern void adjust_managed_page_count(struct page *page, long count);
2319 extern void mem_init_print_info(const char *str);
2321 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2323 /* Free the reserved page into the buddy system, so it gets managed. */
2324 static inline void __free_reserved_page(struct page *page)
2326 ClearPageReserved(page);
2327 init_page_count(page);
2331 static inline void free_reserved_page(struct page *page)
2333 __free_reserved_page(page);
2334 adjust_managed_page_count(page, 1);
2337 static inline void mark_page_reserved(struct page *page)
2339 SetPageReserved(page);
2340 adjust_managed_page_count(page, -1);
2344 * Default method to free all the __init memory into the buddy system.
2345 * The freed pages will be poisoned with pattern "poison" if it's within
2346 * range [0, UCHAR_MAX].
2347 * Return pages freed into the buddy system.
2349 static inline unsigned long free_initmem_default(int poison)
2351 extern char __init_begin[], __init_end[];
2353 return free_reserved_area(&__init_begin, &__init_end,
2354 poison, "unused kernel");
2357 static inline unsigned long get_num_physpages(void)
2360 unsigned long phys_pages = 0;
2362 for_each_online_node(nid)
2363 phys_pages += node_present_pages(nid);
2369 * Using memblock node mappings, an architecture may initialise its
2370 * zones, allocate the backing mem_map and account for memory holes in an
2371 * architecture independent manner.
2373 * An architecture is expected to register range of page frames backed by
2374 * physical memory with memblock_add[_node]() before calling
2375 * free_area_init() passing in the PFN each zone ends at. At a basic
2376 * usage, an architecture is expected to do something like
2378 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2380 * for_each_valid_physical_page_range()
2381 * memblock_add_node(base, size, nid)
2382 * free_area_init(max_zone_pfns);
2384 void free_area_init(unsigned long *max_zone_pfn);
2385 unsigned long node_map_pfn_alignment(void);
2386 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2387 unsigned long end_pfn);
2388 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2389 unsigned long end_pfn);
2390 extern void get_pfn_range_for_nid(unsigned int nid,
2391 unsigned long *start_pfn, unsigned long *end_pfn);
2392 extern unsigned long find_min_pfn_with_active_regions(void);
2394 #ifndef CONFIG_NEED_MULTIPLE_NODES
2395 static inline int early_pfn_to_nid(unsigned long pfn)
2400 /* please see mm/page_alloc.c */
2401 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2402 /* there is a per-arch backend function. */
2403 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
2404 struct mminit_pfnnid_cache *state);
2407 extern void set_dma_reserve(unsigned long new_dma_reserve);
2408 extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long,
2409 enum memmap_context, struct vmem_altmap *);
2410 extern void setup_per_zone_wmarks(void);
2411 extern int __meminit init_per_zone_wmark_min(void);
2412 extern void mem_init(void);
2413 extern void __init mmap_init(void);
2414 extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2415 extern long si_mem_available(void);
2416 extern void si_meminfo(struct sysinfo * val);
2417 extern void si_meminfo_node(struct sysinfo *val, int nid);
2418 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2419 extern unsigned long arch_reserved_kernel_pages(void);
2422 extern __printf(3, 4)
2423 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2425 extern void setup_per_cpu_pageset(void);
2428 extern int min_free_kbytes;
2429 extern int watermark_boost_factor;
2430 extern int watermark_scale_factor;
2431 extern bool arch_has_descending_max_zone_pfns(void);
2434 extern atomic_long_t mmap_pages_allocated;
2435 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2437 /* interval_tree.c */
2438 void vma_interval_tree_insert(struct vm_area_struct *node,
2439 struct rb_root_cached *root);
2440 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2441 struct vm_area_struct *prev,
2442 struct rb_root_cached *root);
2443 void vma_interval_tree_remove(struct vm_area_struct *node,
2444 struct rb_root_cached *root);
2445 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2446 unsigned long start, unsigned long last);
2447 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2448 unsigned long start, unsigned long last);
2450 #define vma_interval_tree_foreach(vma, root, start, last) \
2451 for (vma = vma_interval_tree_iter_first(root, start, last); \
2452 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2454 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2455 struct rb_root_cached *root);
2456 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2457 struct rb_root_cached *root);
2458 struct anon_vma_chain *
2459 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2460 unsigned long start, unsigned long last);
2461 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2462 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2463 #ifdef CONFIG_DEBUG_VM_RB
2464 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2467 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2468 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2469 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2472 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2473 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2474 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2475 struct vm_area_struct *expand);
2476 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2477 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2479 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2481 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2482 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2483 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2484 struct mempolicy *, struct vm_userfaultfd_ctx);
2485 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2486 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2487 unsigned long addr, int new_below);
2488 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2489 unsigned long addr, int new_below);
2490 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2491 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2492 struct rb_node **, struct rb_node *);
2493 extern void unlink_file_vma(struct vm_area_struct *);
2494 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2495 unsigned long addr, unsigned long len, pgoff_t pgoff,
2496 bool *need_rmap_locks);
2497 extern void exit_mmap(struct mm_struct *);
2499 static inline int check_data_rlimit(unsigned long rlim,
2501 unsigned long start,
2502 unsigned long end_data,
2503 unsigned long start_data)
2505 if (rlim < RLIM_INFINITY) {
2506 if (((new - start) + (end_data - start_data)) > rlim)
2513 extern int mm_take_all_locks(struct mm_struct *mm);
2514 extern void mm_drop_all_locks(struct mm_struct *mm);
2516 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2517 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2518 extern struct file *get_task_exe_file(struct task_struct *task);
2520 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2521 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2523 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2524 const struct vm_special_mapping *sm);
2525 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2526 unsigned long addr, unsigned long len,
2527 unsigned long flags,
2528 const struct vm_special_mapping *spec);
2529 /* This is an obsolete alternative to _install_special_mapping. */
2530 extern int install_special_mapping(struct mm_struct *mm,
2531 unsigned long addr, unsigned long len,
2532 unsigned long flags, struct page **pages);
2534 unsigned long randomize_stack_top(unsigned long stack_top);
2536 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2538 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2539 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2540 struct list_head *uf);
2541 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2542 unsigned long len, unsigned long prot, unsigned long flags,
2543 unsigned long pgoff, unsigned long *populate, struct list_head *uf);
2544 extern int __do_munmap(struct mm_struct *, unsigned long, size_t,
2545 struct list_head *uf, bool downgrade);
2546 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2547 struct list_head *uf);
2548 extern int do_madvise(unsigned long start, size_t len_in, int behavior);
2551 extern int __mm_populate(unsigned long addr, unsigned long len,
2553 static inline void mm_populate(unsigned long addr, unsigned long len)
2556 (void) __mm_populate(addr, len, 1);
2559 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2562 /* These take the mm semaphore themselves */
2563 extern int __must_check vm_brk(unsigned long, unsigned long);
2564 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2565 extern int vm_munmap(unsigned long, size_t);
2566 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2567 unsigned long, unsigned long,
2568 unsigned long, unsigned long);
2570 struct vm_unmapped_area_info {
2571 #define VM_UNMAPPED_AREA_TOPDOWN 1
2572 unsigned long flags;
2573 unsigned long length;
2574 unsigned long low_limit;
2575 unsigned long high_limit;
2576 unsigned long align_mask;
2577 unsigned long align_offset;
2580 extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info);
2583 extern void truncate_inode_pages(struct address_space *, loff_t);
2584 extern void truncate_inode_pages_range(struct address_space *,
2585 loff_t lstart, loff_t lend);
2586 extern void truncate_inode_pages_final(struct address_space *);
2588 /* generic vm_area_ops exported for stackable file systems */
2589 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2590 extern void filemap_map_pages(struct vm_fault *vmf,
2591 pgoff_t start_pgoff, pgoff_t end_pgoff);
2592 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2594 /* mm/page-writeback.c */
2595 int __must_check write_one_page(struct page *page);
2596 void task_dirty_inc(struct task_struct *tsk);
2598 extern unsigned long stack_guard_gap;
2599 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2600 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2602 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2603 extern int expand_downwards(struct vm_area_struct *vma,
2604 unsigned long address);
2606 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2608 #define expand_upwards(vma, address) (0)
2611 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2612 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2613 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2614 struct vm_area_struct **pprev);
2616 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2617 NULL if none. Assume start_addr < end_addr. */
2618 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2620 struct vm_area_struct * vma = find_vma(mm,start_addr);
2622 if (vma && end_addr <= vma->vm_start)
2627 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2629 unsigned long vm_start = vma->vm_start;
2631 if (vma->vm_flags & VM_GROWSDOWN) {
2632 vm_start -= stack_guard_gap;
2633 if (vm_start > vma->vm_start)
2639 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2641 unsigned long vm_end = vma->vm_end;
2643 if (vma->vm_flags & VM_GROWSUP) {
2644 vm_end += stack_guard_gap;
2645 if (vm_end < vma->vm_end)
2646 vm_end = -PAGE_SIZE;
2651 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2653 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2656 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2657 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2658 unsigned long vm_start, unsigned long vm_end)
2660 struct vm_area_struct *vma = find_vma(mm, vm_start);
2662 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2668 static inline bool range_in_vma(struct vm_area_struct *vma,
2669 unsigned long start, unsigned long end)
2671 return (vma && vma->vm_start <= start && end <= vma->vm_end);
2675 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2676 void vma_set_page_prot(struct vm_area_struct *vma);
2678 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2682 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2684 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2688 #ifdef CONFIG_NUMA_BALANCING
2689 unsigned long change_prot_numa(struct vm_area_struct *vma,
2690 unsigned long start, unsigned long end);
2693 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2694 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2695 unsigned long pfn, unsigned long size, pgprot_t);
2696 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2697 int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
2698 struct page **pages, unsigned long *num);
2699 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
2701 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
2703 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2705 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2706 unsigned long pfn, pgprot_t pgprot);
2707 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2709 vm_fault_t vmf_insert_mixed_prot(struct vm_area_struct *vma, unsigned long addr,
2710 pfn_t pfn, pgprot_t pgprot);
2711 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
2712 unsigned long addr, pfn_t pfn);
2713 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2715 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2716 unsigned long addr, struct page *page)
2718 int err = vm_insert_page(vma, addr, page);
2721 return VM_FAULT_OOM;
2722 if (err < 0 && err != -EBUSY)
2723 return VM_FAULT_SIGBUS;
2725 return VM_FAULT_NOPAGE;
2728 static inline vm_fault_t vmf_error(int err)
2731 return VM_FAULT_OOM;
2732 return VM_FAULT_SIGBUS;
2735 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
2736 unsigned int foll_flags);
2738 #define FOLL_WRITE 0x01 /* check pte is writable */
2739 #define FOLL_TOUCH 0x02 /* mark page accessed */
2740 #define FOLL_GET 0x04 /* do get_page on page */
2741 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2742 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2743 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2744 * and return without waiting upon it */
2745 #define FOLL_POPULATE 0x40 /* fault in page */
2746 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2747 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2748 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2749 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2750 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2751 #define FOLL_MLOCK 0x1000 /* lock present pages */
2752 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2753 #define FOLL_COW 0x4000 /* internal GUP flag */
2754 #define FOLL_ANON 0x8000 /* don't do file mappings */
2755 #define FOLL_LONGTERM 0x10000 /* mapping lifetime is indefinite: see below */
2756 #define FOLL_SPLIT_PMD 0x20000 /* split huge pmd before returning */
2757 #define FOLL_PIN 0x40000 /* pages must be released via unpin_user_page */
2758 #define FOLL_FAST_ONLY 0x80000 /* gup_fast: prevent fall-back to slow gup */
2761 * FOLL_PIN and FOLL_LONGTERM may be used in various combinations with each
2762 * other. Here is what they mean, and how to use them:
2764 * FOLL_LONGTERM indicates that the page will be held for an indefinite time
2765 * period _often_ under userspace control. This is in contrast to
2766 * iov_iter_get_pages(), whose usages are transient.
2768 * FIXME: For pages which are part of a filesystem, mappings are subject to the
2769 * lifetime enforced by the filesystem and we need guarantees that longterm
2770 * users like RDMA and V4L2 only establish mappings which coordinate usage with
2771 * the filesystem. Ideas for this coordination include revoking the longterm
2772 * pin, delaying writeback, bounce buffer page writeback, etc. As FS DAX was
2773 * added after the problem with filesystems was found FS DAX VMAs are
2774 * specifically failed. Filesystem pages are still subject to bugs and use of
2775 * FOLL_LONGTERM should be avoided on those pages.
2777 * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call.
2778 * Currently only get_user_pages() and get_user_pages_fast() support this flag
2779 * and calls to get_user_pages_[un]locked are specifically not allowed. This
2780 * is due to an incompatibility with the FS DAX check and
2781 * FAULT_FLAG_ALLOW_RETRY.
2783 * In the CMA case: long term pins in a CMA region would unnecessarily fragment
2784 * that region. And so, CMA attempts to migrate the page before pinning, when
2785 * FOLL_LONGTERM is specified.
2787 * FOLL_PIN indicates that a special kind of tracking (not just page->_refcount,
2788 * but an additional pin counting system) will be invoked. This is intended for
2789 * anything that gets a page reference and then touches page data (for example,
2790 * Direct IO). This lets the filesystem know that some non-file-system entity is
2791 * potentially changing the pages' data. In contrast to FOLL_GET (whose pages
2792 * are released via put_page()), FOLL_PIN pages must be released, ultimately, by
2793 * a call to unpin_user_page().
2795 * FOLL_PIN is similar to FOLL_GET: both of these pin pages. They use different
2796 * and separate refcounting mechanisms, however, and that means that each has
2797 * its own acquire and release mechanisms:
2799 * FOLL_GET: get_user_pages*() to acquire, and put_page() to release.
2801 * FOLL_PIN: pin_user_pages*() to acquire, and unpin_user_pages to release.
2803 * FOLL_PIN and FOLL_GET are mutually exclusive for a given function call.
2804 * (The underlying pages may experience both FOLL_GET-based and FOLL_PIN-based
2805 * calls applied to them, and that's perfectly OK. This is a constraint on the
2806 * callers, not on the pages.)
2808 * FOLL_PIN should be set internally by the pin_user_pages*() APIs, never
2809 * directly by the caller. That's in order to help avoid mismatches when
2810 * releasing pages: get_user_pages*() pages must be released via put_page(),
2811 * while pin_user_pages*() pages must be released via unpin_user_page().
2813 * Please see Documentation/core-api/pin_user_pages.rst for more information.
2816 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
2818 if (vm_fault & VM_FAULT_OOM)
2820 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2821 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2822 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2827 typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
2828 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2829 unsigned long size, pte_fn_t fn, void *data);
2830 extern int apply_to_existing_page_range(struct mm_struct *mm,
2831 unsigned long address, unsigned long size,
2832 pte_fn_t fn, void *data);
2834 #ifdef CONFIG_PAGE_POISONING
2835 extern bool page_poisoning_enabled(void);
2836 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2838 static inline bool page_poisoning_enabled(void) { return false; }
2839 static inline void kernel_poison_pages(struct page *page, int numpages,
2843 #ifdef CONFIG_INIT_ON_ALLOC_DEFAULT_ON
2844 DECLARE_STATIC_KEY_TRUE(init_on_alloc);
2846 DECLARE_STATIC_KEY_FALSE(init_on_alloc);
2848 static inline bool want_init_on_alloc(gfp_t flags)
2850 if (static_branch_unlikely(&init_on_alloc) &&
2851 !page_poisoning_enabled())
2853 return flags & __GFP_ZERO;
2856 #ifdef CONFIG_INIT_ON_FREE_DEFAULT_ON
2857 DECLARE_STATIC_KEY_TRUE(init_on_free);
2859 DECLARE_STATIC_KEY_FALSE(init_on_free);
2861 static inline bool want_init_on_free(void)
2863 return static_branch_unlikely(&init_on_free) &&
2864 !page_poisoning_enabled();
2867 #ifdef CONFIG_DEBUG_PAGEALLOC
2868 extern void init_debug_pagealloc(void);
2870 static inline void init_debug_pagealloc(void) {}
2872 extern bool _debug_pagealloc_enabled_early;
2873 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
2875 static inline bool debug_pagealloc_enabled(void)
2877 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
2878 _debug_pagealloc_enabled_early;
2882 * For use in fast paths after init_debug_pagealloc() has run, or when a
2883 * false negative result is not harmful when called too early.
2885 static inline bool debug_pagealloc_enabled_static(void)
2887 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
2890 return static_branch_unlikely(&_debug_pagealloc_enabled);
2893 #if defined(CONFIG_DEBUG_PAGEALLOC) || defined(CONFIG_ARCH_HAS_SET_DIRECT_MAP)
2894 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2897 * When called in DEBUG_PAGEALLOC context, the call should most likely be
2898 * guarded by debug_pagealloc_enabled() or debug_pagealloc_enabled_static()
2901 kernel_map_pages(struct page *page, int numpages, int enable)
2903 __kernel_map_pages(page, numpages, enable);
2905 #ifdef CONFIG_HIBERNATION
2906 extern bool kernel_page_present(struct page *page);
2907 #endif /* CONFIG_HIBERNATION */
2908 #else /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */
2910 kernel_map_pages(struct page *page, int numpages, int enable) {}
2911 #ifdef CONFIG_HIBERNATION
2912 static inline bool kernel_page_present(struct page *page) { return true; }
2913 #endif /* CONFIG_HIBERNATION */
2914 #endif /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */
2916 #ifdef __HAVE_ARCH_GATE_AREA
2917 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2918 extern int in_gate_area_no_mm(unsigned long addr);
2919 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2921 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2925 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2926 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2930 #endif /* __HAVE_ARCH_GATE_AREA */
2932 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2934 #ifdef CONFIG_SYSCTL
2935 extern int sysctl_drop_caches;
2936 int drop_caches_sysctl_handler(struct ctl_table *, int, void *, size_t *,
2940 void drop_slab(void);
2941 void drop_slab_node(int nid);
2944 #define randomize_va_space 0
2946 extern int randomize_va_space;
2949 const char * arch_vma_name(struct vm_area_struct *vma);
2951 void print_vma_addr(char *prefix, unsigned long rip);
2953 static inline void print_vma_addr(char *prefix, unsigned long rip)
2958 void *sparse_buffer_alloc(unsigned long size);
2959 struct page * __populate_section_memmap(unsigned long pfn,
2960 unsigned long nr_pages, int nid, struct vmem_altmap *altmap);
2961 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2962 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
2963 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
2964 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2965 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node,
2966 struct vmem_altmap *altmap);
2967 void *vmemmap_alloc_block(unsigned long size, int node);
2969 void *vmemmap_alloc_block_buf(unsigned long size, int node,
2970 struct vmem_altmap *altmap);
2971 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2972 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2973 int node, struct vmem_altmap *altmap);
2974 int vmemmap_populate(unsigned long start, unsigned long end, int node,
2975 struct vmem_altmap *altmap);
2976 void vmemmap_populate_print_last(void);
2977 #ifdef CONFIG_MEMORY_HOTPLUG
2978 void vmemmap_free(unsigned long start, unsigned long end,
2979 struct vmem_altmap *altmap);
2981 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2982 unsigned long nr_pages);
2985 MF_COUNT_INCREASED = 1 << 0,
2986 MF_ACTION_REQUIRED = 1 << 1,
2987 MF_MUST_KILL = 1 << 2,
2988 MF_SOFT_OFFLINE = 1 << 3,
2990 extern int memory_failure(unsigned long pfn, int flags);
2991 extern void memory_failure_queue(unsigned long pfn, int flags);
2992 extern void memory_failure_queue_kick(int cpu);
2993 extern int unpoison_memory(unsigned long pfn);
2994 extern int get_hwpoison_page(struct page *page);
2995 #define put_hwpoison_page(page) put_page(page)
2996 extern int sysctl_memory_failure_early_kill;
2997 extern int sysctl_memory_failure_recovery;
2998 extern void shake_page(struct page *p, int access);
2999 extern atomic_long_t num_poisoned_pages __read_mostly;
3000 extern int soft_offline_page(unsigned long pfn, int flags);
3004 * Error handlers for various types of pages.
3007 MF_IGNORED, /* Error: cannot be handled */
3008 MF_FAILED, /* Error: handling failed */
3009 MF_DELAYED, /* Will be handled later */
3010 MF_RECOVERED, /* Successfully recovered */
3013 enum mf_action_page_type {
3015 MF_MSG_KERNEL_HIGH_ORDER,
3017 MF_MSG_DIFFERENT_COMPOUND,
3018 MF_MSG_POISONED_HUGE,
3021 MF_MSG_NON_PMD_HUGE,
3022 MF_MSG_UNMAP_FAILED,
3023 MF_MSG_DIRTY_SWAPCACHE,
3024 MF_MSG_CLEAN_SWAPCACHE,
3025 MF_MSG_DIRTY_MLOCKED_LRU,
3026 MF_MSG_CLEAN_MLOCKED_LRU,
3027 MF_MSG_DIRTY_UNEVICTABLE_LRU,
3028 MF_MSG_CLEAN_UNEVICTABLE_LRU,
3031 MF_MSG_TRUNCATED_LRU,
3038 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
3039 extern void clear_huge_page(struct page *page,
3040 unsigned long addr_hint,
3041 unsigned int pages_per_huge_page);
3042 extern void copy_user_huge_page(struct page *dst, struct page *src,
3043 unsigned long addr_hint,
3044 struct vm_area_struct *vma,
3045 unsigned int pages_per_huge_page);
3046 extern long copy_huge_page_from_user(struct page *dst_page,
3047 const void __user *usr_src,
3048 unsigned int pages_per_huge_page,
3049 bool allow_pagefault);
3052 * vma_is_special_huge - Are transhuge page-table entries considered special?
3053 * @vma: Pointer to the struct vm_area_struct to consider
3055 * Whether transhuge page-table entries are considered "special" following
3056 * the definition in vm_normal_page().
3058 * Return: true if transhuge page-table entries should be considered special,
3061 static inline bool vma_is_special_huge(const struct vm_area_struct *vma)
3063 return vma_is_dax(vma) || (vma->vm_file &&
3064 (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP)));
3067 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
3069 #ifdef CONFIG_DEBUG_PAGEALLOC
3070 extern unsigned int _debug_guardpage_minorder;
3071 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
3073 static inline unsigned int debug_guardpage_minorder(void)
3075 return _debug_guardpage_minorder;
3078 static inline bool debug_guardpage_enabled(void)
3080 return static_branch_unlikely(&_debug_guardpage_enabled);
3083 static inline bool page_is_guard(struct page *page)
3085 if (!debug_guardpage_enabled())
3088 return PageGuard(page);
3091 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
3092 static inline bool debug_guardpage_enabled(void) { return false; }
3093 static inline bool page_is_guard(struct page *page) { return false; }
3094 #endif /* CONFIG_DEBUG_PAGEALLOC */
3096 #if MAX_NUMNODES > 1
3097 void __init setup_nr_node_ids(void);
3099 static inline void setup_nr_node_ids(void) {}
3102 extern int memcmp_pages(struct page *page1, struct page *page2);
3104 static inline int pages_identical(struct page *page1, struct page *page2)
3106 return !memcmp_pages(page1, page2);
3109 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
3110 unsigned long clean_record_shared_mapping_range(struct address_space *mapping,
3111 pgoff_t first_index, pgoff_t nr,
3112 pgoff_t bitmap_pgoff,
3113 unsigned long *bitmap,
3117 unsigned long wp_shared_mapping_range(struct address_space *mapping,
3118 pgoff_t first_index, pgoff_t nr);
3121 extern int sysctl_nr_trim_pages;
3123 #endif /* __KERNEL__ */
3124 #endif /* _LINUX_MM_H */