1 /* SPDX-License-Identifier: GPL-2.0 */
5 #include <linux/errno.h>
9 #include <linux/mmdebug.h>
10 #include <linux/gfp.h>
11 #include <linux/bug.h>
12 #include <linux/list.h>
13 #include <linux/mmzone.h>
14 #include <linux/rbtree.h>
15 #include <linux/atomic.h>
16 #include <linux/debug_locks.h>
17 #include <linux/mm_types.h>
18 #include <linux/mmap_lock.h>
19 #include <linux/range.h>
20 #include <linux/pfn.h>
21 #include <linux/percpu-refcount.h>
22 #include <linux/bit_spinlock.h>
23 #include <linux/shrinker.h>
24 #include <linux/resource.h>
25 #include <linux/page_ext.h>
26 #include <linux/err.h>
27 #include <linux/page-flags.h>
28 #include <linux/page_ref.h>
29 #include <linux/memremap.h>
30 #include <linux/overflow.h>
31 #include <linux/sizes.h>
32 #include <linux/sched.h>
33 #include <linux/pgtable.h>
37 struct anon_vma_chain;
40 struct writeback_control;
44 void init_mm_internals(void);
46 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
47 extern unsigned long max_mapnr;
49 static inline void set_max_mapnr(unsigned long limit)
54 static inline void set_max_mapnr(unsigned long limit) { }
57 extern atomic_long_t _totalram_pages;
58 static inline unsigned long totalram_pages(void)
60 return (unsigned long)atomic_long_read(&_totalram_pages);
63 static inline void totalram_pages_inc(void)
65 atomic_long_inc(&_totalram_pages);
68 static inline void totalram_pages_dec(void)
70 atomic_long_dec(&_totalram_pages);
73 static inline void totalram_pages_add(long count)
75 atomic_long_add(count, &_totalram_pages);
78 extern void * high_memory;
79 extern int page_cluster;
82 extern int sysctl_legacy_va_layout;
84 #define sysctl_legacy_va_layout 0
87 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
88 extern const int mmap_rnd_bits_min;
89 extern const int mmap_rnd_bits_max;
90 extern int mmap_rnd_bits __read_mostly;
92 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
93 extern const int mmap_rnd_compat_bits_min;
94 extern const int mmap_rnd_compat_bits_max;
95 extern int mmap_rnd_compat_bits __read_mostly;
99 #include <asm/processor.h>
102 * Architectures that support memory tagging (assigning tags to memory regions,
103 * embedding these tags into addresses that point to these memory regions, and
104 * checking that the memory and the pointer tags match on memory accesses)
105 * redefine this macro to strip tags from pointers.
106 * It's defined as noop for arcitectures that don't support memory tagging.
108 #ifndef untagged_addr
109 #define untagged_addr(addr) (addr)
113 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
117 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
121 #define lm_alias(x) __va(__pa_symbol(x))
125 * To prevent common memory management code establishing
126 * a zero page mapping on a read fault.
127 * This macro should be defined within <asm/pgtable.h>.
128 * s390 does this to prevent multiplexing of hardware bits
129 * related to the physical page in case of virtualization.
131 #ifndef mm_forbids_zeropage
132 #define mm_forbids_zeropage(X) (0)
136 * On some architectures it is expensive to call memset() for small sizes.
137 * If an architecture decides to implement their own version of
138 * mm_zero_struct_page they should wrap the defines below in a #ifndef and
139 * define their own version of this macro in <asm/pgtable.h>
141 #if BITS_PER_LONG == 64
142 /* This function must be updated when the size of struct page grows above 80
143 * or reduces below 56. The idea that compiler optimizes out switch()
144 * statement, and only leaves move/store instructions. Also the compiler can
145 * combine write statments if they are both assignments and can be reordered,
146 * this can result in several of the writes here being dropped.
148 #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
149 static inline void __mm_zero_struct_page(struct page *page)
151 unsigned long *_pp = (void *)page;
153 /* Check that struct page is either 56, 64, 72, or 80 bytes */
154 BUILD_BUG_ON(sizeof(struct page) & 7);
155 BUILD_BUG_ON(sizeof(struct page) < 56);
156 BUILD_BUG_ON(sizeof(struct page) > 80);
158 switch (sizeof(struct page)) {
160 _pp[9] = 0; /* fallthrough */
162 _pp[8] = 0; /* fallthrough */
164 _pp[7] = 0; /* fallthrough */
176 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
180 * Default maximum number of active map areas, this limits the number of vmas
181 * per mm struct. Users can overwrite this number by sysctl but there is a
184 * When a program's coredump is generated as ELF format, a section is created
185 * per a vma. In ELF, the number of sections is represented in unsigned short.
186 * This means the number of sections should be smaller than 65535 at coredump.
187 * Because the kernel adds some informative sections to a image of program at
188 * generating coredump, we need some margin. The number of extra sections is
189 * 1-3 now and depends on arch. We use "5" as safe margin, here.
191 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
192 * not a hard limit any more. Although some userspace tools can be surprised by
195 #define MAPCOUNT_ELF_CORE_MARGIN (5)
196 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
198 extern int sysctl_max_map_count;
200 extern unsigned long sysctl_user_reserve_kbytes;
201 extern unsigned long sysctl_admin_reserve_kbytes;
203 extern int sysctl_overcommit_memory;
204 extern int sysctl_overcommit_ratio;
205 extern unsigned long sysctl_overcommit_kbytes;
207 int overcommit_ratio_handler(struct ctl_table *, int, void *, size_t *,
209 int overcommit_kbytes_handler(struct ctl_table *, int, void *, size_t *,
211 int overcommit_policy_handler(struct ctl_table *, int, void *, size_t *,
214 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
216 /* to align the pointer to the (next) page boundary */
217 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
219 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
220 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
222 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
225 * Linux kernel virtual memory manager primitives.
226 * The idea being to have a "virtual" mm in the same way
227 * we have a virtual fs - giving a cleaner interface to the
228 * mm details, and allowing different kinds of memory mappings
229 * (from shared memory to executable loading to arbitrary
233 struct vm_area_struct *vm_area_alloc(struct mm_struct *);
234 struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
235 void vm_area_free(struct vm_area_struct *);
238 extern struct rb_root nommu_region_tree;
239 extern struct rw_semaphore nommu_region_sem;
241 extern unsigned int kobjsize(const void *objp);
245 * vm_flags in vm_area_struct, see mm_types.h.
246 * When changing, update also include/trace/events/mmflags.h
248 #define VM_NONE 0x00000000
250 #define VM_READ 0x00000001 /* currently active flags */
251 #define VM_WRITE 0x00000002
252 #define VM_EXEC 0x00000004
253 #define VM_SHARED 0x00000008
255 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
256 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
257 #define VM_MAYWRITE 0x00000020
258 #define VM_MAYEXEC 0x00000040
259 #define VM_MAYSHARE 0x00000080
261 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
262 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
263 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
264 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
265 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
267 #define VM_LOCKED 0x00002000
268 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
270 /* Used by sys_madvise() */
271 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
272 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
274 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
275 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
276 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
277 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
278 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
279 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
280 #define VM_SYNC 0x00800000 /* Synchronous page faults */
281 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
282 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
283 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
285 #ifdef CONFIG_MEM_SOFT_DIRTY
286 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
288 # define VM_SOFTDIRTY 0
291 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
292 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
293 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
294 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
296 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
297 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
298 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
299 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
300 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
301 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
302 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
303 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
304 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
305 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
306 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
307 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
309 #ifdef CONFIG_ARCH_HAS_PKEYS
310 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
311 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
312 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
313 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
314 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
316 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
318 # define VM_PKEY_BIT4 0
320 #endif /* CONFIG_ARCH_HAS_PKEYS */
322 #if defined(CONFIG_X86)
323 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
324 #elif defined(CONFIG_PARISC)
325 # define VM_GROWSUP VM_ARCH_1
326 #elif defined(CONFIG_IA64)
327 # define VM_GROWSUP VM_ARCH_1
328 #elif defined(CONFIG_SPARC64)
329 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
330 # define VM_ARCH_CLEAR VM_SPARC_ADI
331 #elif defined(CONFIG_ARM64)
332 # define VM_ARM64_BTI VM_ARCH_1 /* BTI guarded page, a.k.a. GP bit */
333 # define VM_ARCH_CLEAR VM_ARM64_BTI
334 #elif !defined(CONFIG_MMU)
335 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
339 # define VM_GROWSUP VM_NONE
342 /* Bits set in the VMA until the stack is in its final location */
343 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
345 #define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0)
347 /* Common data flag combinations */
348 #define VM_DATA_FLAGS_TSK_EXEC (VM_READ | VM_WRITE | TASK_EXEC | \
349 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
350 #define VM_DATA_FLAGS_NON_EXEC (VM_READ | VM_WRITE | VM_MAYREAD | \
351 VM_MAYWRITE | VM_MAYEXEC)
352 #define VM_DATA_FLAGS_EXEC (VM_READ | VM_WRITE | VM_EXEC | \
353 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
355 #ifndef VM_DATA_DEFAULT_FLAGS /* arch can override this */
356 #define VM_DATA_DEFAULT_FLAGS VM_DATA_FLAGS_EXEC
359 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
360 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
363 #ifdef CONFIG_STACK_GROWSUP
364 #define VM_STACK VM_GROWSUP
366 #define VM_STACK VM_GROWSDOWN
369 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
371 /* VMA basic access permission flags */
372 #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
376 * Special vmas that are non-mergable, non-mlock()able.
378 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
380 /* This mask prevents VMA from being scanned with khugepaged */
381 #define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB)
383 /* This mask defines which mm->def_flags a process can inherit its parent */
384 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
386 /* This mask is used to clear all the VMA flags used by mlock */
387 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
389 /* Arch-specific flags to clear when updating VM flags on protection change */
390 #ifndef VM_ARCH_CLEAR
391 # define VM_ARCH_CLEAR VM_NONE
393 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
396 * mapping from the currently active vm_flags protection bits (the
397 * low four bits) to a page protection mask..
399 extern pgprot_t protection_map[16];
402 * Fault flag definitions.
404 * @FAULT_FLAG_WRITE: Fault was a write fault.
405 * @FAULT_FLAG_MKWRITE: Fault was mkwrite of existing PTE.
406 * @FAULT_FLAG_ALLOW_RETRY: Allow to retry the fault if blocked.
407 * @FAULT_FLAG_RETRY_NOWAIT: Don't drop mmap_lock and wait when retrying.
408 * @FAULT_FLAG_KILLABLE: The fault task is in SIGKILL killable region.
409 * @FAULT_FLAG_TRIED: The fault has been tried once.
410 * @FAULT_FLAG_USER: The fault originated in userspace.
411 * @FAULT_FLAG_REMOTE: The fault is not for current task/mm.
412 * @FAULT_FLAG_INSTRUCTION: The fault was during an instruction fetch.
413 * @FAULT_FLAG_INTERRUPTIBLE: The fault can be interrupted by non-fatal signals.
415 * About @FAULT_FLAG_ALLOW_RETRY and @FAULT_FLAG_TRIED: we can specify
416 * whether we would allow page faults to retry by specifying these two
417 * fault flags correctly. Currently there can be three legal combinations:
419 * (a) ALLOW_RETRY and !TRIED: this means the page fault allows retry, and
420 * this is the first try
422 * (b) ALLOW_RETRY and TRIED: this means the page fault allows retry, and
423 * we've already tried at least once
425 * (c) !ALLOW_RETRY and !TRIED: this means the page fault does not allow retry
427 * The unlisted combination (!ALLOW_RETRY && TRIED) is illegal and should never
428 * be used. Note that page faults can be allowed to retry for multiple times,
429 * in which case we'll have an initial fault with flags (a) then later on
430 * continuous faults with flags (b). We should always try to detect pending
431 * signals before a retry to make sure the continuous page faults can still be
432 * interrupted if necessary.
434 #define FAULT_FLAG_WRITE 0x01
435 #define FAULT_FLAG_MKWRITE 0x02
436 #define FAULT_FLAG_ALLOW_RETRY 0x04
437 #define FAULT_FLAG_RETRY_NOWAIT 0x08
438 #define FAULT_FLAG_KILLABLE 0x10
439 #define FAULT_FLAG_TRIED 0x20
440 #define FAULT_FLAG_USER 0x40
441 #define FAULT_FLAG_REMOTE 0x80
442 #define FAULT_FLAG_INSTRUCTION 0x100
443 #define FAULT_FLAG_INTERRUPTIBLE 0x200
446 * The default fault flags that should be used by most of the
447 * arch-specific page fault handlers.
449 #define FAULT_FLAG_DEFAULT (FAULT_FLAG_ALLOW_RETRY | \
450 FAULT_FLAG_KILLABLE | \
451 FAULT_FLAG_INTERRUPTIBLE)
454 * fault_flag_allow_retry_first - check ALLOW_RETRY the first time
456 * This is mostly used for places where we want to try to avoid taking
457 * the mmap_lock for too long a time when waiting for another condition
458 * to change, in which case we can try to be polite to release the
459 * mmap_lock in the first round to avoid potential starvation of other
460 * processes that would also want the mmap_lock.
462 * Return: true if the page fault allows retry and this is the first
463 * attempt of the fault handling; false otherwise.
465 static inline bool fault_flag_allow_retry_first(unsigned int flags)
467 return (flags & FAULT_FLAG_ALLOW_RETRY) &&
468 (!(flags & FAULT_FLAG_TRIED));
471 #define FAULT_FLAG_TRACE \
472 { FAULT_FLAG_WRITE, "WRITE" }, \
473 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
474 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
475 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
476 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
477 { FAULT_FLAG_TRIED, "TRIED" }, \
478 { FAULT_FLAG_USER, "USER" }, \
479 { FAULT_FLAG_REMOTE, "REMOTE" }, \
480 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }, \
481 { FAULT_FLAG_INTERRUPTIBLE, "INTERRUPTIBLE" }
484 * vm_fault is filled by the pagefault handler and passed to the vma's
485 * ->fault function. The vma's ->fault is responsible for returning a bitmask
486 * of VM_FAULT_xxx flags that give details about how the fault was handled.
488 * MM layer fills up gfp_mask for page allocations but fault handler might
489 * alter it if its implementation requires a different allocation context.
491 * pgoff should be used in favour of virtual_address, if possible.
494 struct vm_area_struct *vma; /* Target VMA */
495 unsigned int flags; /* FAULT_FLAG_xxx flags */
496 gfp_t gfp_mask; /* gfp mask to be used for allocations */
497 pgoff_t pgoff; /* Logical page offset based on vma */
498 unsigned long address; /* Faulting virtual address */
499 pmd_t *pmd; /* Pointer to pmd entry matching
501 pud_t *pud; /* Pointer to pud entry matching
504 pte_t orig_pte; /* Value of PTE at the time of fault */
506 struct page *cow_page; /* Page handler may use for COW fault */
507 struct page *page; /* ->fault handlers should return a
508 * page here, unless VM_FAULT_NOPAGE
509 * is set (which is also implied by
512 /* These three entries are valid only while holding ptl lock */
513 pte_t *pte; /* Pointer to pte entry matching
514 * the 'address'. NULL if the page
515 * table hasn't been allocated.
517 spinlock_t *ptl; /* Page table lock.
518 * Protects pte page table if 'pte'
519 * is not NULL, otherwise pmd.
521 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
522 * vm_ops->map_pages() calls
523 * alloc_set_pte() from atomic context.
524 * do_fault_around() pre-allocates
525 * page table to avoid allocation from
530 /* page entry size for vm->huge_fault() */
531 enum page_entry_size {
538 * These are the virtual MM functions - opening of an area, closing and
539 * unmapping it (needed to keep files on disk up-to-date etc), pointer
540 * to the functions called when a no-page or a wp-page exception occurs.
542 struct vm_operations_struct {
543 void (*open)(struct vm_area_struct * area);
544 void (*close)(struct vm_area_struct * area);
545 int (*split)(struct vm_area_struct * area, unsigned long addr);
546 int (*mremap)(struct vm_area_struct * area);
547 vm_fault_t (*fault)(struct vm_fault *vmf);
548 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
549 enum page_entry_size pe_size);
550 void (*map_pages)(struct vm_fault *vmf,
551 pgoff_t start_pgoff, pgoff_t end_pgoff);
552 unsigned long (*pagesize)(struct vm_area_struct * area);
554 /* notification that a previously read-only page is about to become
555 * writable, if an error is returned it will cause a SIGBUS */
556 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
558 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
559 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
561 /* called by access_process_vm when get_user_pages() fails, typically
562 * for use by special VMAs that can switch between memory and hardware
564 int (*access)(struct vm_area_struct *vma, unsigned long addr,
565 void *buf, int len, int write);
567 /* Called by the /proc/PID/maps code to ask the vma whether it
568 * has a special name. Returning non-NULL will also cause this
569 * vma to be dumped unconditionally. */
570 const char *(*name)(struct vm_area_struct *vma);
574 * set_policy() op must add a reference to any non-NULL @new mempolicy
575 * to hold the policy upon return. Caller should pass NULL @new to
576 * remove a policy and fall back to surrounding context--i.e. do not
577 * install a MPOL_DEFAULT policy, nor the task or system default
580 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
583 * get_policy() op must add reference [mpol_get()] to any policy at
584 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
585 * in mm/mempolicy.c will do this automatically.
586 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
587 * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
588 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
589 * must return NULL--i.e., do not "fallback" to task or system default
592 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
596 * Called by vm_normal_page() for special PTEs to find the
597 * page for @addr. This is useful if the default behavior
598 * (using pte_page()) would not find the correct page.
600 struct page *(*find_special_page)(struct vm_area_struct *vma,
604 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
606 static const struct vm_operations_struct dummy_vm_ops = {};
608 memset(vma, 0, sizeof(*vma));
610 vma->vm_ops = &dummy_vm_ops;
611 INIT_LIST_HEAD(&vma->anon_vma_chain);
614 static inline void vma_set_anonymous(struct vm_area_struct *vma)
619 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
624 static inline bool vma_is_temporary_stack(struct vm_area_struct *vma)
626 int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP);
631 if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) ==
632 VM_STACK_INCOMPLETE_SETUP)
638 static inline bool vma_is_foreign(struct vm_area_struct *vma)
643 if (current->mm != vma->vm_mm)
649 static inline bool vma_is_accessible(struct vm_area_struct *vma)
651 return vma->vm_flags & VM_ACCESS_FLAGS;
656 * The vma_is_shmem is not inline because it is used only by slow
657 * paths in userfault.
659 bool vma_is_shmem(struct vm_area_struct *vma);
661 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
664 int vma_is_stack_for_current(struct vm_area_struct *vma);
666 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
667 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
672 #include <linux/huge_mm.h>
675 * Methods to modify the page usage count.
677 * What counts for a page usage:
678 * - cache mapping (page->mapping)
679 * - private data (page->private)
680 * - page mapped in a task's page tables, each mapping
681 * is counted separately
683 * Also, many kernel routines increase the page count before a critical
684 * routine so they can be sure the page doesn't go away from under them.
688 * Drop a ref, return true if the refcount fell to zero (the page has no users)
690 static inline int put_page_testzero(struct page *page)
692 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
693 return page_ref_dec_and_test(page);
697 * Try to grab a ref unless the page has a refcount of zero, return false if
699 * This can be called when MMU is off so it must not access
700 * any of the virtual mappings.
702 static inline int get_page_unless_zero(struct page *page)
704 return page_ref_add_unless(page, 1, 0);
707 extern int page_is_ram(unsigned long pfn);
715 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
718 /* Support for virtually mapped pages */
719 struct page *vmalloc_to_page(const void *addr);
720 unsigned long vmalloc_to_pfn(const void *addr);
723 * Determine if an address is within the vmalloc range
725 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
726 * is no special casing required.
729 #ifndef is_ioremap_addr
730 #define is_ioremap_addr(x) is_vmalloc_addr(x)
734 extern bool is_vmalloc_addr(const void *x);
735 extern int is_vmalloc_or_module_addr(const void *x);
737 static inline bool is_vmalloc_addr(const void *x)
741 static inline int is_vmalloc_or_module_addr(const void *x)
747 extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
748 static inline void *kvmalloc(size_t size, gfp_t flags)
750 return kvmalloc_node(size, flags, NUMA_NO_NODE);
752 static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
754 return kvmalloc_node(size, flags | __GFP_ZERO, node);
756 static inline void *kvzalloc(size_t size, gfp_t flags)
758 return kvmalloc(size, flags | __GFP_ZERO);
761 static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
765 if (unlikely(check_mul_overflow(n, size, &bytes)))
768 return kvmalloc(bytes, flags);
771 static inline void *kvcalloc(size_t n, size_t size, gfp_t flags)
773 return kvmalloc_array(n, size, flags | __GFP_ZERO);
776 extern void kvfree(const void *addr);
777 extern void kvfree_sensitive(const void *addr, size_t len);
779 static inline int head_mapcount(struct page *head)
781 return atomic_read(compound_mapcount_ptr(head)) + 1;
785 * Mapcount of compound page as a whole, does not include mapped sub-pages.
787 * Must be called only for compound pages or any their tail sub-pages.
789 static inline int compound_mapcount(struct page *page)
791 VM_BUG_ON_PAGE(!PageCompound(page), page);
792 page = compound_head(page);
793 return head_mapcount(page);
797 * The atomic page->_mapcount, starts from -1: so that transitions
798 * both from it and to it can be tracked, using atomic_inc_and_test
799 * and atomic_add_negative(-1).
801 static inline void page_mapcount_reset(struct page *page)
803 atomic_set(&(page)->_mapcount, -1);
806 int __page_mapcount(struct page *page);
809 * Mapcount of 0-order page; when compound sub-page, includes
810 * compound_mapcount().
812 * Result is undefined for pages which cannot be mapped into userspace.
813 * For example SLAB or special types of pages. See function page_has_type().
814 * They use this place in struct page differently.
816 static inline int page_mapcount(struct page *page)
818 if (unlikely(PageCompound(page)))
819 return __page_mapcount(page);
820 return atomic_read(&page->_mapcount) + 1;
823 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
824 int total_mapcount(struct page *page);
825 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
827 static inline int total_mapcount(struct page *page)
829 return page_mapcount(page);
831 static inline int page_trans_huge_mapcount(struct page *page,
834 int mapcount = page_mapcount(page);
836 *total_mapcount = mapcount;
841 static inline struct page *virt_to_head_page(const void *x)
843 struct page *page = virt_to_page(x);
845 return compound_head(page);
848 void __put_page(struct page *page);
850 void put_pages_list(struct list_head *pages);
852 void split_page(struct page *page, unsigned int order);
855 * Compound pages have a destructor function. Provide a
856 * prototype for that function and accessor functions.
857 * These are _only_ valid on the head of a compound page.
859 typedef void compound_page_dtor(struct page *);
861 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
862 enum compound_dtor_id {
865 #ifdef CONFIG_HUGETLB_PAGE
868 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
873 extern compound_page_dtor * const compound_page_dtors[NR_COMPOUND_DTORS];
875 static inline void set_compound_page_dtor(struct page *page,
876 enum compound_dtor_id compound_dtor)
878 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
879 page[1].compound_dtor = compound_dtor;
882 static inline void destroy_compound_page(struct page *page)
884 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
885 compound_page_dtors[page[1].compound_dtor](page);
888 static inline unsigned int compound_order(struct page *page)
892 return page[1].compound_order;
895 static inline bool hpage_pincount_available(struct page *page)
898 * Can the page->hpage_pinned_refcount field be used? That field is in
899 * the 3rd page of the compound page, so the smallest (2-page) compound
900 * pages cannot support it.
902 page = compound_head(page);
903 return PageCompound(page) && compound_order(page) > 1;
906 static inline int head_pincount(struct page *head)
908 return atomic_read(compound_pincount_ptr(head));
911 static inline int compound_pincount(struct page *page)
913 VM_BUG_ON_PAGE(!hpage_pincount_available(page), page);
914 page = compound_head(page);
915 return head_pincount(page);
918 static inline void set_compound_order(struct page *page, unsigned int order)
920 page[1].compound_order = order;
921 page[1].compound_nr = 1U << order;
924 /* Returns the number of pages in this potentially compound page. */
925 static inline unsigned long compound_nr(struct page *page)
929 return page[1].compound_nr;
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 ASSERT_EXCLUSIVE_BITS(page->flags, ZONES_MASK << ZONES_PGSHIFT);
1071 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
1074 #ifdef CONFIG_ZONE_DEVICE
1075 static inline bool is_zone_device_page(const struct page *page)
1077 return page_zonenum(page) == ZONE_DEVICE;
1079 extern void memmap_init_zone_device(struct zone *, unsigned long,
1080 unsigned long, struct dev_pagemap *);
1082 static inline bool is_zone_device_page(const struct page *page)
1088 #ifdef CONFIG_DEV_PAGEMAP_OPS
1089 void free_devmap_managed_page(struct page *page);
1090 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
1092 static inline bool page_is_devmap_managed(struct page *page)
1094 if (!static_branch_unlikely(&devmap_managed_key))
1096 if (!is_zone_device_page(page))
1098 switch (page->pgmap->type) {
1099 case MEMORY_DEVICE_PRIVATE:
1100 case MEMORY_DEVICE_FS_DAX:
1108 void put_devmap_managed_page(struct page *page);
1110 #else /* CONFIG_DEV_PAGEMAP_OPS */
1111 static inline bool page_is_devmap_managed(struct page *page)
1116 static inline void put_devmap_managed_page(struct page *page)
1119 #endif /* CONFIG_DEV_PAGEMAP_OPS */
1121 static inline bool is_device_private_page(const struct page *page)
1123 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
1124 IS_ENABLED(CONFIG_DEVICE_PRIVATE) &&
1125 is_zone_device_page(page) &&
1126 page->pgmap->type == MEMORY_DEVICE_PRIVATE;
1129 static inline bool is_pci_p2pdma_page(const struct page *page)
1131 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
1132 IS_ENABLED(CONFIG_PCI_P2PDMA) &&
1133 is_zone_device_page(page) &&
1134 page->pgmap->type == MEMORY_DEVICE_PCI_P2PDMA;
1137 /* 127: arbitrary random number, small enough to assemble well */
1138 #define page_ref_zero_or_close_to_overflow(page) \
1139 ((unsigned int) page_ref_count(page) + 127u <= 127u)
1141 static inline void get_page(struct page *page)
1143 page = compound_head(page);
1145 * Getting a normal page or the head of a compound page
1146 * requires to already have an elevated page->_refcount.
1148 VM_BUG_ON_PAGE(page_ref_zero_or_close_to_overflow(page), page);
1152 bool __must_check try_grab_page(struct page *page, unsigned int flags);
1154 static inline __must_check bool try_get_page(struct page *page)
1156 page = compound_head(page);
1157 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1163 static inline void put_page(struct page *page)
1165 page = compound_head(page);
1168 * For devmap managed pages we need to catch refcount transition from
1169 * 2 to 1, when refcount reach one it means the page is free and we
1170 * need to inform the device driver through callback. See
1171 * include/linux/memremap.h and HMM for details.
1173 if (page_is_devmap_managed(page)) {
1174 put_devmap_managed_page(page);
1178 if (put_page_testzero(page))
1183 * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
1184 * the page's refcount so that two separate items are tracked: the original page
1185 * reference count, and also a new count of how many pin_user_pages() calls were
1186 * made against the page. ("gup-pinned" is another term for the latter).
1188 * With this scheme, pin_user_pages() becomes special: such pages are marked as
1189 * distinct from normal pages. As such, the unpin_user_page() call (and its
1190 * variants) must be used in order to release gup-pinned pages.
1194 * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
1195 * counts with respect to pin_user_pages() and unpin_user_page() becomes
1196 * simpler, due to the fact that adding an even power of two to the page
1197 * refcount has the effect of using only the upper N bits, for the code that
1198 * counts up using the bias value. This means that the lower bits are left for
1199 * the exclusive use of the original code that increments and decrements by one
1200 * (or at least, by much smaller values than the bias value).
1202 * Of course, once the lower bits overflow into the upper bits (and this is
1203 * OK, because subtraction recovers the original values), then visual inspection
1204 * no longer suffices to directly view the separate counts. However, for normal
1205 * applications that don't have huge page reference counts, this won't be an
1208 * Locking: the lockless algorithm described in page_cache_get_speculative()
1209 * and page_cache_gup_pin_speculative() provides safe operation for
1210 * get_user_pages and page_mkclean and other calls that race to set up page
1213 #define GUP_PIN_COUNTING_BIAS (1U << 10)
1215 void unpin_user_page(struct page *page);
1216 void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1218 void unpin_user_pages(struct page **pages, unsigned long npages);
1221 * page_maybe_dma_pinned() - report if a page is pinned for DMA.
1223 * This function checks if a page has been pinned via a call to
1224 * pin_user_pages*().
1226 * For non-huge pages, the return value is partially fuzzy: false is not fuzzy,
1227 * because it means "definitely not pinned for DMA", but true means "probably
1228 * pinned for DMA, but possibly a false positive due to having at least
1229 * GUP_PIN_COUNTING_BIAS worth of normal page references".
1231 * False positives are OK, because: a) it's unlikely for a page to get that many
1232 * refcounts, and b) all the callers of this routine are expected to be able to
1233 * deal gracefully with a false positive.
1235 * For huge pages, the result will be exactly correct. That's because we have
1236 * more tracking data available: the 3rd struct page in the compound page is
1237 * used to track the pincount (instead using of the GUP_PIN_COUNTING_BIAS
1240 * For more information, please see Documentation/core-api/pin_user_pages.rst.
1242 * @page: pointer to page to be queried.
1243 * @Return: True, if it is likely that the page has been "dma-pinned".
1244 * False, if the page is definitely not dma-pinned.
1246 static inline bool page_maybe_dma_pinned(struct page *page)
1248 if (hpage_pincount_available(page))
1249 return compound_pincount(page) > 0;
1252 * page_ref_count() is signed. If that refcount overflows, then
1253 * page_ref_count() returns a negative value, and callers will avoid
1254 * further incrementing the refcount.
1256 * Here, for that overflow case, use the signed bit to count a little
1257 * bit higher via unsigned math, and thus still get an accurate result.
1259 return ((unsigned int)page_ref_count(compound_head(page))) >=
1260 GUP_PIN_COUNTING_BIAS;
1263 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1264 #define SECTION_IN_PAGE_FLAGS
1268 * The identification function is mainly used by the buddy allocator for
1269 * determining if two pages could be buddies. We are not really identifying
1270 * the zone since we could be using the section number id if we do not have
1271 * node id available in page flags.
1272 * We only guarantee that it will return the same value for two combinable
1275 static inline int page_zone_id(struct page *page)
1277 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1280 #ifdef NODE_NOT_IN_PAGE_FLAGS
1281 extern int page_to_nid(const struct page *page);
1283 static inline int page_to_nid(const struct page *page)
1285 struct page *p = (struct page *)page;
1287 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1291 #ifdef CONFIG_NUMA_BALANCING
1292 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1294 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1297 static inline int cpupid_to_pid(int cpupid)
1299 return cpupid & LAST__PID_MASK;
1302 static inline int cpupid_to_cpu(int cpupid)
1304 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1307 static inline int cpupid_to_nid(int cpupid)
1309 return cpu_to_node(cpupid_to_cpu(cpupid));
1312 static inline bool cpupid_pid_unset(int cpupid)
1314 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1317 static inline bool cpupid_cpu_unset(int cpupid)
1319 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1322 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1324 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1327 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1328 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1329 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1331 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1334 static inline int page_cpupid_last(struct page *page)
1336 return page->_last_cpupid;
1338 static inline void page_cpupid_reset_last(struct page *page)
1340 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1343 static inline int page_cpupid_last(struct page *page)
1345 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1348 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1350 static inline void page_cpupid_reset_last(struct page *page)
1352 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1354 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1355 #else /* !CONFIG_NUMA_BALANCING */
1356 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1358 return page_to_nid(page); /* XXX */
1361 static inline int page_cpupid_last(struct page *page)
1363 return page_to_nid(page); /* XXX */
1366 static inline int cpupid_to_nid(int cpupid)
1371 static inline int cpupid_to_pid(int cpupid)
1376 static inline int cpupid_to_cpu(int cpupid)
1381 static inline int cpu_pid_to_cpupid(int nid, int pid)
1386 static inline bool cpupid_pid_unset(int cpupid)
1391 static inline void page_cpupid_reset_last(struct page *page)
1395 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1399 #endif /* CONFIG_NUMA_BALANCING */
1401 #ifdef CONFIG_KASAN_SW_TAGS
1402 static inline u8 page_kasan_tag(const struct page *page)
1404 return (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1407 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1409 page->flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1410 page->flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1413 static inline void page_kasan_tag_reset(struct page *page)
1415 page_kasan_tag_set(page, 0xff);
1418 static inline u8 page_kasan_tag(const struct page *page)
1423 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1424 static inline void page_kasan_tag_reset(struct page *page) { }
1427 static inline struct zone *page_zone(const struct page *page)
1429 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1432 static inline pg_data_t *page_pgdat(const struct page *page)
1434 return NODE_DATA(page_to_nid(page));
1437 #ifdef SECTION_IN_PAGE_FLAGS
1438 static inline void set_page_section(struct page *page, unsigned long section)
1440 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1441 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1444 static inline unsigned long page_to_section(const struct page *page)
1446 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1450 static inline void set_page_zone(struct page *page, enum zone_type zone)
1452 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1453 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1456 static inline void set_page_node(struct page *page, unsigned long node)
1458 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1459 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1462 static inline void set_page_links(struct page *page, enum zone_type zone,
1463 unsigned long node, unsigned long pfn)
1465 set_page_zone(page, zone);
1466 set_page_node(page, node);
1467 #ifdef SECTION_IN_PAGE_FLAGS
1468 set_page_section(page, pfn_to_section_nr(pfn));
1473 static inline struct mem_cgroup *page_memcg(struct page *page)
1475 return page->mem_cgroup;
1477 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1479 WARN_ON_ONCE(!rcu_read_lock_held());
1480 return READ_ONCE(page->mem_cgroup);
1483 static inline struct mem_cgroup *page_memcg(struct page *page)
1487 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1489 WARN_ON_ONCE(!rcu_read_lock_held());
1495 * Some inline functions in vmstat.h depend on page_zone()
1497 #include <linux/vmstat.h>
1499 static __always_inline void *lowmem_page_address(const struct page *page)
1501 return page_to_virt(page);
1504 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1505 #define HASHED_PAGE_VIRTUAL
1508 #if defined(WANT_PAGE_VIRTUAL)
1509 static inline void *page_address(const struct page *page)
1511 return page->virtual;
1513 static inline void set_page_address(struct page *page, void *address)
1515 page->virtual = address;
1517 #define page_address_init() do { } while(0)
1520 #if defined(HASHED_PAGE_VIRTUAL)
1521 void *page_address(const struct page *page);
1522 void set_page_address(struct page *page, void *virtual);
1523 void page_address_init(void);
1526 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1527 #define page_address(page) lowmem_page_address(page)
1528 #define set_page_address(page, address) do { } while(0)
1529 #define page_address_init() do { } while(0)
1532 extern void *page_rmapping(struct page *page);
1533 extern struct anon_vma *page_anon_vma(struct page *page);
1534 extern struct address_space *page_mapping(struct page *page);
1536 extern struct address_space *__page_file_mapping(struct page *);
1539 struct address_space *page_file_mapping(struct page *page)
1541 if (unlikely(PageSwapCache(page)))
1542 return __page_file_mapping(page);
1544 return page->mapping;
1547 extern pgoff_t __page_file_index(struct page *page);
1550 * Return the pagecache index of the passed page. Regular pagecache pages
1551 * use ->index whereas swapcache pages use swp_offset(->private)
1553 static inline pgoff_t page_index(struct page *page)
1555 if (unlikely(PageSwapCache(page)))
1556 return __page_file_index(page);
1560 bool page_mapped(struct page *page);
1561 struct address_space *page_mapping(struct page *page);
1562 struct address_space *page_mapping_file(struct page *page);
1565 * Return true only if the page has been allocated with
1566 * ALLOC_NO_WATERMARKS and the low watermark was not
1567 * met implying that the system is under some pressure.
1569 static inline bool page_is_pfmemalloc(struct page *page)
1572 * Page index cannot be this large so this must be
1573 * a pfmemalloc page.
1575 return page->index == -1UL;
1579 * Only to be called by the page allocator on a freshly allocated
1582 static inline void set_page_pfmemalloc(struct page *page)
1587 static inline void clear_page_pfmemalloc(struct page *page)
1593 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1595 extern void pagefault_out_of_memory(void);
1597 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1598 #define offset_in_thp(page, p) ((unsigned long)(p) & (thp_size(page) - 1))
1601 * Flags passed to show_mem() and show_free_areas() to suppress output in
1604 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1606 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1609 extern bool can_do_mlock(void);
1611 static inline bool can_do_mlock(void) { return false; }
1613 extern int user_shm_lock(size_t, struct user_struct *);
1614 extern void user_shm_unlock(size_t, struct user_struct *);
1617 * Parameter block passed down to zap_pte_range in exceptional cases.
1619 struct zap_details {
1620 struct address_space *check_mapping; /* Check page->mapping if set */
1621 pgoff_t first_index; /* Lowest page->index to unmap */
1622 pgoff_t last_index; /* Highest page->index to unmap */
1625 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1627 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1630 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1631 unsigned long size);
1632 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1633 unsigned long size);
1634 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1635 unsigned long start, unsigned long end);
1637 struct mmu_notifier_range;
1639 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1640 unsigned long end, unsigned long floor, unsigned long ceiling);
1641 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1642 struct vm_area_struct *vma);
1643 int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
1644 struct mmu_notifier_range *range,
1645 pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp);
1646 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1647 unsigned long *pfn);
1648 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1649 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1650 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1651 void *buf, int len, int write);
1653 extern void truncate_pagecache(struct inode *inode, loff_t new);
1654 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1655 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1656 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1657 int truncate_inode_page(struct address_space *mapping, struct page *page);
1658 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1659 int invalidate_inode_page(struct page *page);
1662 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1663 unsigned long address, unsigned int flags,
1664 struct pt_regs *regs);
1665 extern int fixup_user_fault(struct mm_struct *mm,
1666 unsigned long address, unsigned int fault_flags,
1668 void unmap_mapping_pages(struct address_space *mapping,
1669 pgoff_t start, pgoff_t nr, bool even_cows);
1670 void unmap_mapping_range(struct address_space *mapping,
1671 loff_t const holebegin, loff_t const holelen, int even_cows);
1673 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1674 unsigned long address, unsigned int flags,
1675 struct pt_regs *regs)
1677 /* should never happen if there's no MMU */
1679 return VM_FAULT_SIGBUS;
1681 static inline int fixup_user_fault(struct mm_struct *mm, unsigned long address,
1682 unsigned int fault_flags, bool *unlocked)
1684 /* should never happen if there's no MMU */
1688 static inline void unmap_mapping_pages(struct address_space *mapping,
1689 pgoff_t start, pgoff_t nr, bool even_cows) { }
1690 static inline void unmap_mapping_range(struct address_space *mapping,
1691 loff_t const holebegin, loff_t const holelen, int even_cows) { }
1694 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1695 loff_t const holebegin, loff_t const holelen)
1697 unmap_mapping_range(mapping, holebegin, holelen, 0);
1700 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1701 void *buf, int len, unsigned int gup_flags);
1702 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1703 void *buf, int len, unsigned int gup_flags);
1704 extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1705 unsigned long addr, void *buf, int len, unsigned int gup_flags);
1707 long get_user_pages_remote(struct mm_struct *mm,
1708 unsigned long start, unsigned long nr_pages,
1709 unsigned int gup_flags, struct page **pages,
1710 struct vm_area_struct **vmas, int *locked);
1711 long pin_user_pages_remote(struct mm_struct *mm,
1712 unsigned long start, unsigned long nr_pages,
1713 unsigned int gup_flags, struct page **pages,
1714 struct vm_area_struct **vmas, int *locked);
1715 long get_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 pin_user_pages(unsigned long start, unsigned long nr_pages,
1719 unsigned int gup_flags, struct page **pages,
1720 struct vm_area_struct **vmas);
1721 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1722 unsigned int gup_flags, struct page **pages, int *locked);
1723 long pin_user_pages_locked(unsigned long start, unsigned long nr_pages,
1724 unsigned int gup_flags, struct page **pages, int *locked);
1725 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1726 struct page **pages, unsigned int gup_flags);
1727 long pin_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1728 struct page **pages, unsigned int gup_flags);
1730 int get_user_pages_fast(unsigned long start, int nr_pages,
1731 unsigned int gup_flags, struct page **pages);
1732 int pin_user_pages_fast(unsigned long start, int nr_pages,
1733 unsigned int gup_flags, struct page **pages);
1735 int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
1736 int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
1737 struct task_struct *task, bool bypass_rlim);
1739 /* Container for pinned pfns / pages */
1740 struct frame_vector {
1741 unsigned int nr_allocated; /* Number of frames we have space for */
1742 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1743 bool got_ref; /* Did we pin pages by getting page ref? */
1744 bool is_pfns; /* Does array contain pages or pfns? */
1745 void *ptrs[]; /* Array of pinned pfns / pages. Use
1746 * pfns_vector_pages() or pfns_vector_pfns()
1750 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1751 void frame_vector_destroy(struct frame_vector *vec);
1752 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1753 unsigned int gup_flags, struct frame_vector *vec);
1754 void put_vaddr_frames(struct frame_vector *vec);
1755 int frame_vector_to_pages(struct frame_vector *vec);
1756 void frame_vector_to_pfns(struct frame_vector *vec);
1758 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1760 return vec->nr_frames;
1763 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1766 int err = frame_vector_to_pages(vec);
1769 return ERR_PTR(err);
1771 return (struct page **)(vec->ptrs);
1774 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1777 frame_vector_to_pfns(vec);
1778 return (unsigned long *)(vec->ptrs);
1782 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1783 struct page **pages);
1784 int get_kernel_page(unsigned long start, int write, struct page **pages);
1785 struct page *get_dump_page(unsigned long addr);
1787 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1788 extern void do_invalidatepage(struct page *page, unsigned int offset,
1789 unsigned int length);
1791 void __set_page_dirty(struct page *, struct address_space *, int warn);
1792 int __set_page_dirty_nobuffers(struct page *page);
1793 int __set_page_dirty_no_writeback(struct page *page);
1794 int redirty_page_for_writepage(struct writeback_control *wbc,
1796 void account_page_dirtied(struct page *page, struct address_space *mapping);
1797 void account_page_cleaned(struct page *page, struct address_space *mapping,
1798 struct bdi_writeback *wb);
1799 int set_page_dirty(struct page *page);
1800 int set_page_dirty_lock(struct page *page);
1801 void __cancel_dirty_page(struct page *page);
1802 static inline void cancel_dirty_page(struct page *page)
1804 /* Avoid atomic ops, locking, etc. when not actually needed. */
1805 if (PageDirty(page))
1806 __cancel_dirty_page(page);
1808 int clear_page_dirty_for_io(struct page *page);
1810 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1812 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1813 unsigned long old_addr, struct vm_area_struct *new_vma,
1814 unsigned long new_addr, unsigned long len,
1815 bool need_rmap_locks);
1818 * Flags used by change_protection(). For now we make it a bitmap so
1819 * that we can pass in multiple flags just like parameters. However
1820 * for now all the callers are only use one of the flags at the same
1823 /* Whether we should allow dirty bit accounting */
1824 #define MM_CP_DIRTY_ACCT (1UL << 0)
1825 /* Whether this protection change is for NUMA hints */
1826 #define MM_CP_PROT_NUMA (1UL << 1)
1827 /* Whether this change is for write protecting */
1828 #define MM_CP_UFFD_WP (1UL << 2) /* do wp */
1829 #define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */
1830 #define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \
1831 MM_CP_UFFD_WP_RESOLVE)
1833 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1834 unsigned long end, pgprot_t newprot,
1835 unsigned long cp_flags);
1836 extern int mprotect_fixup(struct vm_area_struct *vma,
1837 struct vm_area_struct **pprev, unsigned long start,
1838 unsigned long end, unsigned long newflags);
1841 * doesn't attempt to fault and will return short.
1843 int get_user_pages_fast_only(unsigned long start, int nr_pages,
1844 unsigned int gup_flags, struct page **pages);
1845 int pin_user_pages_fast_only(unsigned long start, int nr_pages,
1846 unsigned int gup_flags, struct page **pages);
1848 static inline bool get_user_page_fast_only(unsigned long addr,
1849 unsigned int gup_flags, struct page **pagep)
1851 return get_user_pages_fast_only(addr, 1, gup_flags, pagep) == 1;
1854 * per-process(per-mm_struct) statistics.
1856 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1858 long val = atomic_long_read(&mm->rss_stat.count[member]);
1860 #ifdef SPLIT_RSS_COUNTING
1862 * counter is updated in asynchronous manner and may go to minus.
1863 * But it's never be expected number for users.
1868 return (unsigned long)val;
1871 void mm_trace_rss_stat(struct mm_struct *mm, int member, long count);
1873 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1875 long count = atomic_long_add_return(value, &mm->rss_stat.count[member]);
1877 mm_trace_rss_stat(mm, member, count);
1880 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1882 long count = atomic_long_inc_return(&mm->rss_stat.count[member]);
1884 mm_trace_rss_stat(mm, member, count);
1887 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1889 long count = atomic_long_dec_return(&mm->rss_stat.count[member]);
1891 mm_trace_rss_stat(mm, member, count);
1894 /* Optimized variant when page is already known not to be PageAnon */
1895 static inline int mm_counter_file(struct page *page)
1897 if (PageSwapBacked(page))
1898 return MM_SHMEMPAGES;
1899 return MM_FILEPAGES;
1902 static inline int mm_counter(struct page *page)
1905 return MM_ANONPAGES;
1906 return mm_counter_file(page);
1909 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1911 return get_mm_counter(mm, MM_FILEPAGES) +
1912 get_mm_counter(mm, MM_ANONPAGES) +
1913 get_mm_counter(mm, MM_SHMEMPAGES);
1916 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1918 return max(mm->hiwater_rss, get_mm_rss(mm));
1921 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1923 return max(mm->hiwater_vm, mm->total_vm);
1926 static inline void update_hiwater_rss(struct mm_struct *mm)
1928 unsigned long _rss = get_mm_rss(mm);
1930 if ((mm)->hiwater_rss < _rss)
1931 (mm)->hiwater_rss = _rss;
1934 static inline void update_hiwater_vm(struct mm_struct *mm)
1936 if (mm->hiwater_vm < mm->total_vm)
1937 mm->hiwater_vm = mm->total_vm;
1940 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1942 mm->hiwater_rss = get_mm_rss(mm);
1945 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1946 struct mm_struct *mm)
1948 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1950 if (*maxrss < hiwater_rss)
1951 *maxrss = hiwater_rss;
1954 #if defined(SPLIT_RSS_COUNTING)
1955 void sync_mm_rss(struct mm_struct *mm);
1957 static inline void sync_mm_rss(struct mm_struct *mm)
1962 #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
1963 static inline int pte_special(pte_t pte)
1968 static inline pte_t pte_mkspecial(pte_t pte)
1974 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
1975 static inline int pte_devmap(pte_t pte)
1981 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1983 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1985 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1989 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1993 #ifdef __PAGETABLE_P4D_FOLDED
1994 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1995 unsigned long address)
2000 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
2003 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
2004 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2005 unsigned long address)
2009 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
2010 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
2013 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
2015 static inline void mm_inc_nr_puds(struct mm_struct *mm)
2017 if (mm_pud_folded(mm))
2019 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2022 static inline void mm_dec_nr_puds(struct mm_struct *mm)
2024 if (mm_pud_folded(mm))
2026 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2030 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
2031 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
2032 unsigned long address)
2037 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
2038 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
2041 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
2043 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
2045 if (mm_pmd_folded(mm))
2047 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2050 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
2052 if (mm_pmd_folded(mm))
2054 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2059 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
2061 atomic_long_set(&mm->pgtables_bytes, 0);
2064 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2066 return atomic_long_read(&mm->pgtables_bytes);
2069 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
2071 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2074 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
2076 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2080 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
2081 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2086 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
2087 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
2090 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
2091 int __pte_alloc_kernel(pmd_t *pmd);
2093 #if defined(CONFIG_MMU)
2095 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2096 unsigned long address)
2098 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
2099 NULL : p4d_offset(pgd, address);
2102 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2103 unsigned long address)
2105 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
2106 NULL : pud_offset(p4d, address);
2109 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
2111 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
2112 NULL: pmd_offset(pud, address);
2114 #endif /* CONFIG_MMU */
2116 #if USE_SPLIT_PTE_PTLOCKS
2117 #if ALLOC_SPLIT_PTLOCKS
2118 void __init ptlock_cache_init(void);
2119 extern bool ptlock_alloc(struct page *page);
2120 extern void ptlock_free(struct page *page);
2122 static inline spinlock_t *ptlock_ptr(struct page *page)
2126 #else /* ALLOC_SPLIT_PTLOCKS */
2127 static inline void ptlock_cache_init(void)
2131 static inline bool ptlock_alloc(struct page *page)
2136 static inline void ptlock_free(struct page *page)
2140 static inline spinlock_t *ptlock_ptr(struct page *page)
2144 #endif /* ALLOC_SPLIT_PTLOCKS */
2146 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2148 return ptlock_ptr(pmd_page(*pmd));
2151 static inline bool ptlock_init(struct page *page)
2154 * prep_new_page() initialize page->private (and therefore page->ptl)
2155 * with 0. Make sure nobody took it in use in between.
2157 * It can happen if arch try to use slab for page table allocation:
2158 * slab code uses page->slab_cache, which share storage with page->ptl.
2160 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
2161 if (!ptlock_alloc(page))
2163 spin_lock_init(ptlock_ptr(page));
2167 #else /* !USE_SPLIT_PTE_PTLOCKS */
2169 * We use mm->page_table_lock to guard all pagetable pages of the mm.
2171 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2173 return &mm->page_table_lock;
2175 static inline void ptlock_cache_init(void) {}
2176 static inline bool ptlock_init(struct page *page) { return true; }
2177 static inline void ptlock_free(struct page *page) {}
2178 #endif /* USE_SPLIT_PTE_PTLOCKS */
2180 static inline void pgtable_init(void)
2182 ptlock_cache_init();
2183 pgtable_cache_init();
2186 static inline bool pgtable_pte_page_ctor(struct page *page)
2188 if (!ptlock_init(page))
2190 __SetPageTable(page);
2191 inc_zone_page_state(page, NR_PAGETABLE);
2195 static inline void pgtable_pte_page_dtor(struct page *page)
2198 __ClearPageTable(page);
2199 dec_zone_page_state(page, NR_PAGETABLE);
2202 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
2204 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
2205 pte_t *__pte = pte_offset_map(pmd, address); \
2211 #define pte_unmap_unlock(pte, ptl) do { \
2216 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2218 #define pte_alloc_map(mm, pmd, address) \
2219 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2221 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2222 (pte_alloc(mm, pmd) ? \
2223 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2225 #define pte_alloc_kernel(pmd, address) \
2226 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2227 NULL: pte_offset_kernel(pmd, address))
2229 #if USE_SPLIT_PMD_PTLOCKS
2231 static struct page *pmd_to_page(pmd_t *pmd)
2233 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2234 return virt_to_page((void *)((unsigned long) pmd & mask));
2237 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2239 return ptlock_ptr(pmd_to_page(pmd));
2242 static inline bool pgtable_pmd_page_ctor(struct page *page)
2244 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2245 page->pmd_huge_pte = NULL;
2247 return ptlock_init(page);
2250 static inline void pgtable_pmd_page_dtor(struct page *page)
2252 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2253 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2258 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2262 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2264 return &mm->page_table_lock;
2267 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
2268 static inline void pgtable_pmd_page_dtor(struct page *page) {}
2270 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2274 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2276 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2282 * No scalability reason to split PUD locks yet, but follow the same pattern
2283 * as the PMD locks to make it easier if we decide to. The VM should not be
2284 * considered ready to switch to split PUD locks yet; there may be places
2285 * which need to be converted from page_table_lock.
2287 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2289 return &mm->page_table_lock;
2292 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2294 spinlock_t *ptl = pud_lockptr(mm, pud);
2300 extern void __init pagecache_init(void);
2301 extern void __init free_area_init_memoryless_node(int nid);
2302 extern void free_initmem(void);
2305 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2306 * into the buddy system. The freed pages will be poisoned with pattern
2307 * "poison" if it's within range [0, UCHAR_MAX].
2308 * Return pages freed into the buddy system.
2310 extern unsigned long free_reserved_area(void *start, void *end,
2311 int poison, const char *s);
2313 #ifdef CONFIG_HIGHMEM
2315 * Free a highmem page into the buddy system, adjusting totalhigh_pages
2316 * and totalram_pages.
2318 extern void free_highmem_page(struct page *page);
2321 extern void adjust_managed_page_count(struct page *page, long count);
2322 extern void mem_init_print_info(const char *str);
2324 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2326 /* Free the reserved page into the buddy system, so it gets managed. */
2327 static inline void __free_reserved_page(struct page *page)
2329 ClearPageReserved(page);
2330 init_page_count(page);
2334 static inline void free_reserved_page(struct page *page)
2336 __free_reserved_page(page);
2337 adjust_managed_page_count(page, 1);
2340 static inline void mark_page_reserved(struct page *page)
2342 SetPageReserved(page);
2343 adjust_managed_page_count(page, -1);
2347 * Default method to free all the __init memory into the buddy system.
2348 * The freed pages will be poisoned with pattern "poison" if it's within
2349 * range [0, UCHAR_MAX].
2350 * Return pages freed into the buddy system.
2352 static inline unsigned long free_initmem_default(int poison)
2354 extern char __init_begin[], __init_end[];
2356 return free_reserved_area(&__init_begin, &__init_end,
2357 poison, "unused kernel");
2360 static inline unsigned long get_num_physpages(void)
2363 unsigned long phys_pages = 0;
2365 for_each_online_node(nid)
2366 phys_pages += node_present_pages(nid);
2372 * Using memblock node mappings, an architecture may initialise its
2373 * zones, allocate the backing mem_map and account for memory holes in an
2374 * architecture independent manner.
2376 * An architecture is expected to register range of page frames backed by
2377 * physical memory with memblock_add[_node]() before calling
2378 * free_area_init() passing in the PFN each zone ends at. At a basic
2379 * usage, an architecture is expected to do something like
2381 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2383 * for_each_valid_physical_page_range()
2384 * memblock_add_node(base, size, nid)
2385 * free_area_init(max_zone_pfns);
2387 void free_area_init(unsigned long *max_zone_pfn);
2388 unsigned long node_map_pfn_alignment(void);
2389 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2390 unsigned long end_pfn);
2391 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2392 unsigned long end_pfn);
2393 extern void get_pfn_range_for_nid(unsigned int nid,
2394 unsigned long *start_pfn, unsigned long *end_pfn);
2395 extern unsigned long find_min_pfn_with_active_regions(void);
2397 #ifndef CONFIG_NEED_MULTIPLE_NODES
2398 static inline int early_pfn_to_nid(unsigned long pfn)
2403 /* please see mm/page_alloc.c */
2404 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2405 /* there is a per-arch backend function. */
2406 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
2407 struct mminit_pfnnid_cache *state);
2410 extern void set_dma_reserve(unsigned long new_dma_reserve);
2411 extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long,
2412 enum memmap_context, struct vmem_altmap *);
2413 extern void setup_per_zone_wmarks(void);
2414 extern int __meminit init_per_zone_wmark_min(void);
2415 extern void mem_init(void);
2416 extern void __init mmap_init(void);
2417 extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2418 extern long si_mem_available(void);
2419 extern void si_meminfo(struct sysinfo * val);
2420 extern void si_meminfo_node(struct sysinfo *val, int nid);
2421 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2422 extern unsigned long arch_reserved_kernel_pages(void);
2425 extern __printf(3, 4)
2426 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2428 extern void setup_per_cpu_pageset(void);
2431 extern int min_free_kbytes;
2432 extern int watermark_boost_factor;
2433 extern int watermark_scale_factor;
2434 extern bool arch_has_descending_max_zone_pfns(void);
2437 extern atomic_long_t mmap_pages_allocated;
2438 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2440 /* interval_tree.c */
2441 void vma_interval_tree_insert(struct vm_area_struct *node,
2442 struct rb_root_cached *root);
2443 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2444 struct vm_area_struct *prev,
2445 struct rb_root_cached *root);
2446 void vma_interval_tree_remove(struct vm_area_struct *node,
2447 struct rb_root_cached *root);
2448 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2449 unsigned long start, unsigned long last);
2450 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2451 unsigned long start, unsigned long last);
2453 #define vma_interval_tree_foreach(vma, root, start, last) \
2454 for (vma = vma_interval_tree_iter_first(root, start, last); \
2455 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2457 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2458 struct rb_root_cached *root);
2459 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2460 struct rb_root_cached *root);
2461 struct anon_vma_chain *
2462 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2463 unsigned long start, unsigned long last);
2464 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2465 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2466 #ifdef CONFIG_DEBUG_VM_RB
2467 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2470 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2471 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2472 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2475 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2476 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2477 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2478 struct vm_area_struct *expand);
2479 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2480 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2482 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2484 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2485 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2486 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2487 struct mempolicy *, struct vm_userfaultfd_ctx);
2488 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2489 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2490 unsigned long addr, int new_below);
2491 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2492 unsigned long addr, int new_below);
2493 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2494 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2495 struct rb_node **, struct rb_node *);
2496 extern void unlink_file_vma(struct vm_area_struct *);
2497 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2498 unsigned long addr, unsigned long len, pgoff_t pgoff,
2499 bool *need_rmap_locks);
2500 extern void exit_mmap(struct mm_struct *);
2502 static inline int check_data_rlimit(unsigned long rlim,
2504 unsigned long start,
2505 unsigned long end_data,
2506 unsigned long start_data)
2508 if (rlim < RLIM_INFINITY) {
2509 if (((new - start) + (end_data - start_data)) > rlim)
2516 extern int mm_take_all_locks(struct mm_struct *mm);
2517 extern void mm_drop_all_locks(struct mm_struct *mm);
2519 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2520 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2521 extern struct file *get_task_exe_file(struct task_struct *task);
2523 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2524 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2526 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2527 const struct vm_special_mapping *sm);
2528 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2529 unsigned long addr, unsigned long len,
2530 unsigned long flags,
2531 const struct vm_special_mapping *spec);
2532 /* This is an obsolete alternative to _install_special_mapping. */
2533 extern int install_special_mapping(struct mm_struct *mm,
2534 unsigned long addr, unsigned long len,
2535 unsigned long flags, struct page **pages);
2537 unsigned long randomize_stack_top(unsigned long stack_top);
2539 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2541 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2542 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2543 struct list_head *uf);
2544 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2545 unsigned long len, unsigned long prot, unsigned long flags,
2546 unsigned long pgoff, unsigned long *populate, struct list_head *uf);
2547 extern int __do_munmap(struct mm_struct *, unsigned long, size_t,
2548 struct list_head *uf, bool downgrade);
2549 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2550 struct list_head *uf);
2551 extern int do_madvise(unsigned long start, size_t len_in, int behavior);
2554 extern int __mm_populate(unsigned long addr, unsigned long len,
2556 static inline void mm_populate(unsigned long addr, unsigned long len)
2559 (void) __mm_populate(addr, len, 1);
2562 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2565 /* These take the mm semaphore themselves */
2566 extern int __must_check vm_brk(unsigned long, unsigned long);
2567 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2568 extern int vm_munmap(unsigned long, size_t);
2569 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2570 unsigned long, unsigned long,
2571 unsigned long, unsigned long);
2573 struct vm_unmapped_area_info {
2574 #define VM_UNMAPPED_AREA_TOPDOWN 1
2575 unsigned long flags;
2576 unsigned long length;
2577 unsigned long low_limit;
2578 unsigned long high_limit;
2579 unsigned long align_mask;
2580 unsigned long align_offset;
2583 extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info);
2586 extern void truncate_inode_pages(struct address_space *, loff_t);
2587 extern void truncate_inode_pages_range(struct address_space *,
2588 loff_t lstart, loff_t lend);
2589 extern void truncate_inode_pages_final(struct address_space *);
2591 /* generic vm_area_ops exported for stackable file systems */
2592 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2593 extern void filemap_map_pages(struct vm_fault *vmf,
2594 pgoff_t start_pgoff, pgoff_t end_pgoff);
2595 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2597 /* mm/page-writeback.c */
2598 int __must_check write_one_page(struct page *page);
2599 void task_dirty_inc(struct task_struct *tsk);
2601 extern unsigned long stack_guard_gap;
2602 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2603 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2605 /* CONFIG_STACK_GROWSUP still needs to grow downwards at some places */
2606 extern int expand_downwards(struct vm_area_struct *vma,
2607 unsigned long address);
2609 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2611 #define expand_upwards(vma, address) (0)
2614 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2615 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2616 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2617 struct vm_area_struct **pprev);
2619 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2620 NULL if none. Assume start_addr < end_addr. */
2621 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2623 struct vm_area_struct * vma = find_vma(mm,start_addr);
2625 if (vma && end_addr <= vma->vm_start)
2630 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2632 unsigned long vm_start = vma->vm_start;
2634 if (vma->vm_flags & VM_GROWSDOWN) {
2635 vm_start -= stack_guard_gap;
2636 if (vm_start > vma->vm_start)
2642 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2644 unsigned long vm_end = vma->vm_end;
2646 if (vma->vm_flags & VM_GROWSUP) {
2647 vm_end += stack_guard_gap;
2648 if (vm_end < vma->vm_end)
2649 vm_end = -PAGE_SIZE;
2654 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2656 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2659 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2660 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2661 unsigned long vm_start, unsigned long vm_end)
2663 struct vm_area_struct *vma = find_vma(mm, vm_start);
2665 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2671 static inline bool range_in_vma(struct vm_area_struct *vma,
2672 unsigned long start, unsigned long end)
2674 return (vma && vma->vm_start <= start && end <= vma->vm_end);
2678 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2679 void vma_set_page_prot(struct vm_area_struct *vma);
2681 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2685 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2687 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2691 #ifdef CONFIG_NUMA_BALANCING
2692 unsigned long change_prot_numa(struct vm_area_struct *vma,
2693 unsigned long start, unsigned long end);
2696 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2697 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2698 unsigned long pfn, unsigned long size, pgprot_t);
2699 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2700 int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
2701 struct page **pages, unsigned long *num);
2702 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
2704 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
2706 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2708 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2709 unsigned long pfn, pgprot_t pgprot);
2710 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2712 vm_fault_t vmf_insert_mixed_prot(struct vm_area_struct *vma, unsigned long addr,
2713 pfn_t pfn, pgprot_t pgprot);
2714 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
2715 unsigned long addr, pfn_t pfn);
2716 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2718 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2719 unsigned long addr, struct page *page)
2721 int err = vm_insert_page(vma, addr, page);
2724 return VM_FAULT_OOM;
2725 if (err < 0 && err != -EBUSY)
2726 return VM_FAULT_SIGBUS;
2728 return VM_FAULT_NOPAGE;
2731 static inline vm_fault_t vmf_error(int err)
2734 return VM_FAULT_OOM;
2735 return VM_FAULT_SIGBUS;
2738 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
2739 unsigned int foll_flags);
2741 #define FOLL_WRITE 0x01 /* check pte is writable */
2742 #define FOLL_TOUCH 0x02 /* mark page accessed */
2743 #define FOLL_GET 0x04 /* do get_page on page */
2744 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2745 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2746 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2747 * and return without waiting upon it */
2748 #define FOLL_POPULATE 0x40 /* fault in page */
2749 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2750 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2751 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2752 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2753 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2754 #define FOLL_MLOCK 0x1000 /* lock present pages */
2755 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2756 #define FOLL_COW 0x4000 /* internal GUP flag */
2757 #define FOLL_ANON 0x8000 /* don't do file mappings */
2758 #define FOLL_LONGTERM 0x10000 /* mapping lifetime is indefinite: see below */
2759 #define FOLL_SPLIT_PMD 0x20000 /* split huge pmd before returning */
2760 #define FOLL_PIN 0x40000 /* pages must be released via unpin_user_page */
2761 #define FOLL_FAST_ONLY 0x80000 /* gup_fast: prevent fall-back to slow gup */
2764 * FOLL_PIN and FOLL_LONGTERM may be used in various combinations with each
2765 * other. Here is what they mean, and how to use them:
2767 * FOLL_LONGTERM indicates that the page will be held for an indefinite time
2768 * period _often_ under userspace control. This is in contrast to
2769 * iov_iter_get_pages(), whose usages are transient.
2771 * FIXME: For pages which are part of a filesystem, mappings are subject to the
2772 * lifetime enforced by the filesystem and we need guarantees that longterm
2773 * users like RDMA and V4L2 only establish mappings which coordinate usage with
2774 * the filesystem. Ideas for this coordination include revoking the longterm
2775 * pin, delaying writeback, bounce buffer page writeback, etc. As FS DAX was
2776 * added after the problem with filesystems was found FS DAX VMAs are
2777 * specifically failed. Filesystem pages are still subject to bugs and use of
2778 * FOLL_LONGTERM should be avoided on those pages.
2780 * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call.
2781 * Currently only get_user_pages() and get_user_pages_fast() support this flag
2782 * and calls to get_user_pages_[un]locked are specifically not allowed. This
2783 * is due to an incompatibility with the FS DAX check and
2784 * FAULT_FLAG_ALLOW_RETRY.
2786 * In the CMA case: long term pins in a CMA region would unnecessarily fragment
2787 * that region. And so, CMA attempts to migrate the page before pinning, when
2788 * FOLL_LONGTERM is specified.
2790 * FOLL_PIN indicates that a special kind of tracking (not just page->_refcount,
2791 * but an additional pin counting system) will be invoked. This is intended for
2792 * anything that gets a page reference and then touches page data (for example,
2793 * Direct IO). This lets the filesystem know that some non-file-system entity is
2794 * potentially changing the pages' data. In contrast to FOLL_GET (whose pages
2795 * are released via put_page()), FOLL_PIN pages must be released, ultimately, by
2796 * a call to unpin_user_page().
2798 * FOLL_PIN is similar to FOLL_GET: both of these pin pages. They use different
2799 * and separate refcounting mechanisms, however, and that means that each has
2800 * its own acquire and release mechanisms:
2802 * FOLL_GET: get_user_pages*() to acquire, and put_page() to release.
2804 * FOLL_PIN: pin_user_pages*() to acquire, and unpin_user_pages to release.
2806 * FOLL_PIN and FOLL_GET are mutually exclusive for a given function call.
2807 * (The underlying pages may experience both FOLL_GET-based and FOLL_PIN-based
2808 * calls applied to them, and that's perfectly OK. This is a constraint on the
2809 * callers, not on the pages.)
2811 * FOLL_PIN should be set internally by the pin_user_pages*() APIs, never
2812 * directly by the caller. That's in order to help avoid mismatches when
2813 * releasing pages: get_user_pages*() pages must be released via put_page(),
2814 * while pin_user_pages*() pages must be released via unpin_user_page().
2816 * Please see Documentation/core-api/pin_user_pages.rst for more information.
2819 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
2821 if (vm_fault & VM_FAULT_OOM)
2823 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2824 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2825 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2830 typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
2831 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2832 unsigned long size, pte_fn_t fn, void *data);
2833 extern int apply_to_existing_page_range(struct mm_struct *mm,
2834 unsigned long address, unsigned long size,
2835 pte_fn_t fn, void *data);
2837 #ifdef CONFIG_PAGE_POISONING
2838 extern bool page_poisoning_enabled(void);
2839 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2841 static inline bool page_poisoning_enabled(void) { return false; }
2842 static inline void kernel_poison_pages(struct page *page, int numpages,
2846 #ifdef CONFIG_INIT_ON_ALLOC_DEFAULT_ON
2847 DECLARE_STATIC_KEY_TRUE(init_on_alloc);
2849 DECLARE_STATIC_KEY_FALSE(init_on_alloc);
2851 static inline bool want_init_on_alloc(gfp_t flags)
2853 if (static_branch_unlikely(&init_on_alloc) &&
2854 !page_poisoning_enabled())
2856 return flags & __GFP_ZERO;
2859 #ifdef CONFIG_INIT_ON_FREE_DEFAULT_ON
2860 DECLARE_STATIC_KEY_TRUE(init_on_free);
2862 DECLARE_STATIC_KEY_FALSE(init_on_free);
2864 static inline bool want_init_on_free(void)
2866 return static_branch_unlikely(&init_on_free) &&
2867 !page_poisoning_enabled();
2870 #ifdef CONFIG_DEBUG_PAGEALLOC
2871 extern void init_debug_pagealloc(void);
2873 static inline void init_debug_pagealloc(void) {}
2875 extern bool _debug_pagealloc_enabled_early;
2876 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
2878 static inline bool debug_pagealloc_enabled(void)
2880 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
2881 _debug_pagealloc_enabled_early;
2885 * For use in fast paths after init_debug_pagealloc() has run, or when a
2886 * false negative result is not harmful when called too early.
2888 static inline bool debug_pagealloc_enabled_static(void)
2890 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
2893 return static_branch_unlikely(&_debug_pagealloc_enabled);
2896 #if defined(CONFIG_DEBUG_PAGEALLOC) || defined(CONFIG_ARCH_HAS_SET_DIRECT_MAP)
2897 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2900 * When called in DEBUG_PAGEALLOC context, the call should most likely be
2901 * guarded by debug_pagealloc_enabled() or debug_pagealloc_enabled_static()
2904 kernel_map_pages(struct page *page, int numpages, int enable)
2906 __kernel_map_pages(page, numpages, enable);
2908 #ifdef CONFIG_HIBERNATION
2909 extern bool kernel_page_present(struct page *page);
2910 #endif /* CONFIG_HIBERNATION */
2911 #else /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */
2913 kernel_map_pages(struct page *page, int numpages, int enable) {}
2914 #ifdef CONFIG_HIBERNATION
2915 static inline bool kernel_page_present(struct page *page) { return true; }
2916 #endif /* CONFIG_HIBERNATION */
2917 #endif /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */
2919 #ifdef __HAVE_ARCH_GATE_AREA
2920 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2921 extern int in_gate_area_no_mm(unsigned long addr);
2922 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2924 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2928 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2929 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2933 #endif /* __HAVE_ARCH_GATE_AREA */
2935 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2937 #ifdef CONFIG_SYSCTL
2938 extern int sysctl_drop_caches;
2939 int drop_caches_sysctl_handler(struct ctl_table *, int, void *, size_t *,
2943 void drop_slab(void);
2944 void drop_slab_node(int nid);
2947 #define randomize_va_space 0
2949 extern int randomize_va_space;
2952 const char * arch_vma_name(struct vm_area_struct *vma);
2954 void print_vma_addr(char *prefix, unsigned long rip);
2956 static inline void print_vma_addr(char *prefix, unsigned long rip)
2961 void *sparse_buffer_alloc(unsigned long size);
2962 struct page * __populate_section_memmap(unsigned long pfn,
2963 unsigned long nr_pages, int nid, struct vmem_altmap *altmap);
2964 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2965 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
2966 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
2967 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2968 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node,
2969 struct vmem_altmap *altmap);
2970 void *vmemmap_alloc_block(unsigned long size, int node);
2972 void *vmemmap_alloc_block_buf(unsigned long size, int node,
2973 struct vmem_altmap *altmap);
2974 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2975 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2976 int node, struct vmem_altmap *altmap);
2977 int vmemmap_populate(unsigned long start, unsigned long end, int node,
2978 struct vmem_altmap *altmap);
2979 void vmemmap_populate_print_last(void);
2980 #ifdef CONFIG_MEMORY_HOTPLUG
2981 void vmemmap_free(unsigned long start, unsigned long end,
2982 struct vmem_altmap *altmap);
2984 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2985 unsigned long nr_pages);
2988 MF_COUNT_INCREASED = 1 << 0,
2989 MF_ACTION_REQUIRED = 1 << 1,
2990 MF_MUST_KILL = 1 << 2,
2991 MF_SOFT_OFFLINE = 1 << 3,
2993 extern int memory_failure(unsigned long pfn, int flags);
2994 extern void memory_failure_queue(unsigned long pfn, int flags);
2995 extern void memory_failure_queue_kick(int cpu);
2996 extern int unpoison_memory(unsigned long pfn);
2997 extern int get_hwpoison_page(struct page *page);
2998 #define put_hwpoison_page(page) put_page(page)
2999 extern int sysctl_memory_failure_early_kill;
3000 extern int sysctl_memory_failure_recovery;
3001 extern void shake_page(struct page *p, int access);
3002 extern atomic_long_t num_poisoned_pages __read_mostly;
3003 extern int soft_offline_page(unsigned long pfn, int flags);
3007 * Error handlers for various types of pages.
3010 MF_IGNORED, /* Error: cannot be handled */
3011 MF_FAILED, /* Error: handling failed */
3012 MF_DELAYED, /* Will be handled later */
3013 MF_RECOVERED, /* Successfully recovered */
3016 enum mf_action_page_type {
3018 MF_MSG_KERNEL_HIGH_ORDER,
3020 MF_MSG_DIFFERENT_COMPOUND,
3021 MF_MSG_POISONED_HUGE,
3024 MF_MSG_NON_PMD_HUGE,
3025 MF_MSG_UNMAP_FAILED,
3026 MF_MSG_DIRTY_SWAPCACHE,
3027 MF_MSG_CLEAN_SWAPCACHE,
3028 MF_MSG_DIRTY_MLOCKED_LRU,
3029 MF_MSG_CLEAN_MLOCKED_LRU,
3030 MF_MSG_DIRTY_UNEVICTABLE_LRU,
3031 MF_MSG_CLEAN_UNEVICTABLE_LRU,
3034 MF_MSG_TRUNCATED_LRU,
3041 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
3042 extern void clear_huge_page(struct page *page,
3043 unsigned long addr_hint,
3044 unsigned int pages_per_huge_page);
3045 extern void copy_user_huge_page(struct page *dst, struct page *src,
3046 unsigned long addr_hint,
3047 struct vm_area_struct *vma,
3048 unsigned int pages_per_huge_page);
3049 extern long copy_huge_page_from_user(struct page *dst_page,
3050 const void __user *usr_src,
3051 unsigned int pages_per_huge_page,
3052 bool allow_pagefault);
3055 * vma_is_special_huge - Are transhuge page-table entries considered special?
3056 * @vma: Pointer to the struct vm_area_struct to consider
3058 * Whether transhuge page-table entries are considered "special" following
3059 * the definition in vm_normal_page().
3061 * Return: true if transhuge page-table entries should be considered special,
3064 static inline bool vma_is_special_huge(const struct vm_area_struct *vma)
3066 return vma_is_dax(vma) || (vma->vm_file &&
3067 (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP)));
3070 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
3072 #ifdef CONFIG_DEBUG_PAGEALLOC
3073 extern unsigned int _debug_guardpage_minorder;
3074 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
3076 static inline unsigned int debug_guardpage_minorder(void)
3078 return _debug_guardpage_minorder;
3081 static inline bool debug_guardpage_enabled(void)
3083 return static_branch_unlikely(&_debug_guardpage_enabled);
3086 static inline bool page_is_guard(struct page *page)
3088 if (!debug_guardpage_enabled())
3091 return PageGuard(page);
3094 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
3095 static inline bool debug_guardpage_enabled(void) { return false; }
3096 static inline bool page_is_guard(struct page *page) { return false; }
3097 #endif /* CONFIG_DEBUG_PAGEALLOC */
3099 #if MAX_NUMNODES > 1
3100 void __init setup_nr_node_ids(void);
3102 static inline void setup_nr_node_ids(void) {}
3105 extern int memcmp_pages(struct page *page1, struct page *page2);
3107 static inline int pages_identical(struct page *page1, struct page *page2)
3109 return !memcmp_pages(page1, page2);
3112 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
3113 unsigned long clean_record_shared_mapping_range(struct address_space *mapping,
3114 pgoff_t first_index, pgoff_t nr,
3115 pgoff_t bitmap_pgoff,
3116 unsigned long *bitmap,
3120 unsigned long wp_shared_mapping_range(struct address_space *mapping,
3121 pgoff_t first_index, pgoff_t nr);
3124 extern int sysctl_nr_trim_pages;
3126 #endif /* __KERNEL__ */
3127 #endif /* _LINUX_MM_H */