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/range.h>
19 #include <linux/pfn.h>
20 #include <linux/percpu-refcount.h>
21 #include <linux/bit_spinlock.h>
22 #include <linux/shrinker.h>
23 #include <linux/resource.h>
24 #include <linux/page_ext.h>
25 #include <linux/err.h>
26 #include <linux/page_ref.h>
27 #include <linux/memremap.h>
28 #include <linux/overflow.h>
29 #include <linux/sizes.h>
30 #include <linux/sched.h>
34 struct anon_vma_chain;
37 struct writeback_control;
40 void init_mm_internals(void);
42 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
43 extern unsigned long max_mapnr;
45 static inline void set_max_mapnr(unsigned long limit)
50 static inline void set_max_mapnr(unsigned long limit) { }
53 extern atomic_long_t _totalram_pages;
54 static inline unsigned long totalram_pages(void)
56 return (unsigned long)atomic_long_read(&_totalram_pages);
59 static inline void totalram_pages_inc(void)
61 atomic_long_inc(&_totalram_pages);
64 static inline void totalram_pages_dec(void)
66 atomic_long_dec(&_totalram_pages);
69 static inline void totalram_pages_add(long count)
71 atomic_long_add(count, &_totalram_pages);
74 extern void * high_memory;
75 extern int page_cluster;
78 extern int sysctl_legacy_va_layout;
80 #define sysctl_legacy_va_layout 0
83 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
84 extern const int mmap_rnd_bits_min;
85 extern const int mmap_rnd_bits_max;
86 extern int mmap_rnd_bits __read_mostly;
88 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
89 extern const int mmap_rnd_compat_bits_min;
90 extern const int mmap_rnd_compat_bits_max;
91 extern int mmap_rnd_compat_bits __read_mostly;
95 #include <asm/pgtable.h>
96 #include <asm/processor.h>
99 * Architectures that support memory tagging (assigning tags to memory regions,
100 * embedding these tags into addresses that point to these memory regions, and
101 * checking that the memory and the pointer tags match on memory accesses)
102 * redefine this macro to strip tags from pointers.
103 * It's defined as noop for arcitectures that don't support memory tagging.
105 #ifndef untagged_addr
106 #define untagged_addr(addr) (addr)
110 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
114 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
118 #define lm_alias(x) __va(__pa_symbol(x))
122 * To prevent common memory management code establishing
123 * a zero page mapping on a read fault.
124 * This macro should be defined within <asm/pgtable.h>.
125 * s390 does this to prevent multiplexing of hardware bits
126 * related to the physical page in case of virtualization.
128 #ifndef mm_forbids_zeropage
129 #define mm_forbids_zeropage(X) (0)
133 * On some architectures it is expensive to call memset() for small sizes.
134 * If an architecture decides to implement their own version of
135 * mm_zero_struct_page they should wrap the defines below in a #ifndef and
136 * define their own version of this macro in <asm/pgtable.h>
138 #if BITS_PER_LONG == 64
139 /* This function must be updated when the size of struct page grows above 80
140 * or reduces below 56. The idea that compiler optimizes out switch()
141 * statement, and only leaves move/store instructions. Also the compiler can
142 * combine write statments if they are both assignments and can be reordered,
143 * this can result in several of the writes here being dropped.
145 #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
146 static inline void __mm_zero_struct_page(struct page *page)
148 unsigned long *_pp = (void *)page;
150 /* Check that struct page is either 56, 64, 72, or 80 bytes */
151 BUILD_BUG_ON(sizeof(struct page) & 7);
152 BUILD_BUG_ON(sizeof(struct page) < 56);
153 BUILD_BUG_ON(sizeof(struct page) > 80);
155 switch (sizeof(struct page)) {
157 _pp[9] = 0; /* fallthrough */
159 _pp[8] = 0; /* fallthrough */
161 _pp[7] = 0; /* fallthrough */
173 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
177 * Default maximum number of active map areas, this limits the number of vmas
178 * per mm struct. Users can overwrite this number by sysctl but there is a
181 * When a program's coredump is generated as ELF format, a section is created
182 * per a vma. In ELF, the number of sections is represented in unsigned short.
183 * This means the number of sections should be smaller than 65535 at coredump.
184 * Because the kernel adds some informative sections to a image of program at
185 * generating coredump, we need some margin. The number of extra sections is
186 * 1-3 now and depends on arch. We use "5" as safe margin, here.
188 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
189 * not a hard limit any more. Although some userspace tools can be surprised by
192 #define MAPCOUNT_ELF_CORE_MARGIN (5)
193 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
195 extern int sysctl_max_map_count;
197 extern unsigned long sysctl_user_reserve_kbytes;
198 extern unsigned long sysctl_admin_reserve_kbytes;
200 extern int sysctl_overcommit_memory;
201 extern int sysctl_overcommit_ratio;
202 extern unsigned long sysctl_overcommit_kbytes;
204 int overcommit_ratio_handler(struct ctl_table *, int, void *, size_t *,
206 int overcommit_kbytes_handler(struct ctl_table *, int, void *, size_t *,
209 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
211 /* to align the pointer to the (next) page boundary */
212 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
214 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
215 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
217 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
220 * Linux kernel virtual memory manager primitives.
221 * The idea being to have a "virtual" mm in the same way
222 * we have a virtual fs - giving a cleaner interface to the
223 * mm details, and allowing different kinds of memory mappings
224 * (from shared memory to executable loading to arbitrary
228 struct vm_area_struct *vm_area_alloc(struct mm_struct *);
229 struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
230 void vm_area_free(struct vm_area_struct *);
233 extern struct rb_root nommu_region_tree;
234 extern struct rw_semaphore nommu_region_sem;
236 extern unsigned int kobjsize(const void *objp);
240 * vm_flags in vm_area_struct, see mm_types.h.
241 * When changing, update also include/trace/events/mmflags.h
243 #define VM_NONE 0x00000000
245 #define VM_READ 0x00000001 /* currently active flags */
246 #define VM_WRITE 0x00000002
247 #define VM_EXEC 0x00000004
248 #define VM_SHARED 0x00000008
250 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
251 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
252 #define VM_MAYWRITE 0x00000020
253 #define VM_MAYEXEC 0x00000040
254 #define VM_MAYSHARE 0x00000080
256 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
257 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
258 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
259 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
260 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
262 #define VM_LOCKED 0x00002000
263 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
265 /* Used by sys_madvise() */
266 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
267 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
269 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
270 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
271 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
272 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
273 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
274 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
275 #define VM_SYNC 0x00800000 /* Synchronous page faults */
276 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
277 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
278 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
280 #ifdef CONFIG_MEM_SOFT_DIRTY
281 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
283 # define VM_SOFTDIRTY 0
286 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
287 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
288 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
289 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
291 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
292 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
293 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
294 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
295 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
296 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
297 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
298 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
299 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
300 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
301 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
302 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
304 #ifdef CONFIG_ARCH_HAS_PKEYS
305 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
306 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
307 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
308 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
309 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
311 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
313 # define VM_PKEY_BIT4 0
315 #endif /* CONFIG_ARCH_HAS_PKEYS */
317 #if defined(CONFIG_X86)
318 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
319 #elif defined(CONFIG_PPC)
320 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
321 #elif defined(CONFIG_PARISC)
322 # define VM_GROWSUP VM_ARCH_1
323 #elif defined(CONFIG_IA64)
324 # define VM_GROWSUP VM_ARCH_1
325 #elif defined(CONFIG_SPARC64)
326 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
327 # define VM_ARCH_CLEAR VM_SPARC_ADI
328 #elif !defined(CONFIG_MMU)
329 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
332 #if defined(CONFIG_X86_INTEL_MPX)
333 /* MPX specific bounds table or bounds directory */
334 # define VM_MPX VM_HIGH_ARCH_4
336 # define VM_MPX VM_NONE
340 # define VM_GROWSUP VM_NONE
343 /* Bits set in the VMA until the stack is in its final location */
344 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
346 #define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0)
348 /* Common data flag combinations */
349 #define VM_DATA_FLAGS_TSK_EXEC (VM_READ | VM_WRITE | TASK_EXEC | \
350 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
351 #define VM_DATA_FLAGS_NON_EXEC (VM_READ | VM_WRITE | VM_MAYREAD | \
352 VM_MAYWRITE | VM_MAYEXEC)
353 #define VM_DATA_FLAGS_EXEC (VM_READ | VM_WRITE | VM_EXEC | \
354 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
356 #ifndef VM_DATA_DEFAULT_FLAGS /* arch can override this */
357 #define VM_DATA_DEFAULT_FLAGS VM_DATA_FLAGS_EXEC
360 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
361 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
364 #ifdef CONFIG_STACK_GROWSUP
365 #define VM_STACK VM_GROWSUP
367 #define VM_STACK VM_GROWSDOWN
370 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
372 /* VMA basic access permission flags */
373 #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
377 * Special vmas that are non-mergable, non-mlock()able.
379 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
381 /* This mask prevents VMA from being scanned with khugepaged */
382 #define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB)
384 /* This mask defines which mm->def_flags a process can inherit its parent */
385 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
387 /* This mask is used to clear all the VMA flags used by mlock */
388 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
390 /* Arch-specific flags to clear when updating VM flags on protection change */
391 #ifndef VM_ARCH_CLEAR
392 # define VM_ARCH_CLEAR VM_NONE
394 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
397 * mapping from the currently active vm_flags protection bits (the
398 * low four bits) to a page protection mask..
400 extern pgprot_t protection_map[16];
403 * Fault flag definitions.
405 * @FAULT_FLAG_WRITE: Fault was a write fault.
406 * @FAULT_FLAG_MKWRITE: Fault was mkwrite of existing PTE.
407 * @FAULT_FLAG_ALLOW_RETRY: Allow to retry the fault if blocked.
408 * @FAULT_FLAG_RETRY_NOWAIT: Don't drop mmap_sem and wait when retrying.
409 * @FAULT_FLAG_KILLABLE: The fault task is in SIGKILL killable region.
410 * @FAULT_FLAG_TRIED: The fault has been tried once.
411 * @FAULT_FLAG_USER: The fault originated in userspace.
412 * @FAULT_FLAG_REMOTE: The fault is not for current task/mm.
413 * @FAULT_FLAG_INSTRUCTION: The fault was during an instruction fetch.
414 * @FAULT_FLAG_INTERRUPTIBLE: The fault can be interrupted by non-fatal signals.
416 * About @FAULT_FLAG_ALLOW_RETRY and @FAULT_FLAG_TRIED: we can specify
417 * whether we would allow page faults to retry by specifying these two
418 * fault flags correctly. Currently there can be three legal combinations:
420 * (a) ALLOW_RETRY and !TRIED: this means the page fault allows retry, and
421 * this is the first try
423 * (b) ALLOW_RETRY and TRIED: this means the page fault allows retry, and
424 * we've already tried at least once
426 * (c) !ALLOW_RETRY and !TRIED: this means the page fault does not allow retry
428 * The unlisted combination (!ALLOW_RETRY && TRIED) is illegal and should never
429 * be used. Note that page faults can be allowed to retry for multiple times,
430 * in which case we'll have an initial fault with flags (a) then later on
431 * continuous faults with flags (b). We should always try to detect pending
432 * signals before a retry to make sure the continuous page faults can still be
433 * interrupted if necessary.
435 #define FAULT_FLAG_WRITE 0x01
436 #define FAULT_FLAG_MKWRITE 0x02
437 #define FAULT_FLAG_ALLOW_RETRY 0x04
438 #define FAULT_FLAG_RETRY_NOWAIT 0x08
439 #define FAULT_FLAG_KILLABLE 0x10
440 #define FAULT_FLAG_TRIED 0x20
441 #define FAULT_FLAG_USER 0x40
442 #define FAULT_FLAG_REMOTE 0x80
443 #define FAULT_FLAG_INSTRUCTION 0x100
444 #define FAULT_FLAG_INTERRUPTIBLE 0x200
447 * The default fault flags that should be used by most of the
448 * arch-specific page fault handlers.
450 #define FAULT_FLAG_DEFAULT (FAULT_FLAG_ALLOW_RETRY | \
451 FAULT_FLAG_KILLABLE | \
452 FAULT_FLAG_INTERRUPTIBLE)
455 * fault_flag_allow_retry_first - check ALLOW_RETRY the first time
457 * This is mostly used for places where we want to try to avoid taking
458 * the mmap_sem for too long a time when waiting for another condition
459 * to change, in which case we can try to be polite to release the
460 * mmap_sem in the first round to avoid potential starvation of other
461 * processes that would also want the mmap_sem.
463 * Return: true if the page fault allows retry and this is the first
464 * attempt of the fault handling; false otherwise.
466 static inline bool fault_flag_allow_retry_first(unsigned int flags)
468 return (flags & FAULT_FLAG_ALLOW_RETRY) &&
469 (!(flags & FAULT_FLAG_TRIED));
472 #define FAULT_FLAG_TRACE \
473 { FAULT_FLAG_WRITE, "WRITE" }, \
474 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
475 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
476 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
477 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
478 { FAULT_FLAG_TRIED, "TRIED" }, \
479 { FAULT_FLAG_USER, "USER" }, \
480 { FAULT_FLAG_REMOTE, "REMOTE" }, \
481 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }, \
482 { FAULT_FLAG_INTERRUPTIBLE, "INTERRUPTIBLE" }
485 * vm_fault is filled by the the pagefault handler and passed to the vma's
486 * ->fault function. The vma's ->fault is responsible for returning a bitmask
487 * of VM_FAULT_xxx flags that give details about how the fault was handled.
489 * MM layer fills up gfp_mask for page allocations but fault handler might
490 * alter it if its implementation requires a different allocation context.
492 * pgoff should be used in favour of virtual_address, if possible.
495 struct vm_area_struct *vma; /* Target VMA */
496 unsigned int flags; /* FAULT_FLAG_xxx flags */
497 gfp_t gfp_mask; /* gfp mask to be used for allocations */
498 pgoff_t pgoff; /* Logical page offset based on vma */
499 unsigned long address; /* Faulting virtual address */
500 pmd_t *pmd; /* Pointer to pmd entry matching
502 pud_t *pud; /* Pointer to pud entry matching
505 pte_t orig_pte; /* Value of PTE at the time of fault */
507 struct page *cow_page; /* Page handler may use for COW fault */
508 struct mem_cgroup *memcg; /* Cgroup cow_page belongs to */
509 struct page *page; /* ->fault handlers should return a
510 * page here, unless VM_FAULT_NOPAGE
511 * is set (which is also implied by
514 /* These three entries are valid only while holding ptl lock */
515 pte_t *pte; /* Pointer to pte entry matching
516 * the 'address'. NULL if the page
517 * table hasn't been allocated.
519 spinlock_t *ptl; /* Page table lock.
520 * Protects pte page table if 'pte'
521 * is not NULL, otherwise pmd.
523 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
524 * vm_ops->map_pages() calls
525 * alloc_set_pte() from atomic context.
526 * do_fault_around() pre-allocates
527 * page table to avoid allocation from
532 /* page entry size for vm->huge_fault() */
533 enum page_entry_size {
540 * These are the virtual MM functions - opening of an area, closing and
541 * unmapping it (needed to keep files on disk up-to-date etc), pointer
542 * to the functions called when a no-page or a wp-page exception occurs.
544 struct vm_operations_struct {
545 void (*open)(struct vm_area_struct * area);
546 void (*close)(struct vm_area_struct * area);
547 int (*split)(struct vm_area_struct * area, unsigned long addr);
548 int (*mremap)(struct vm_area_struct * area);
549 vm_fault_t (*fault)(struct vm_fault *vmf);
550 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
551 enum page_entry_size pe_size);
552 void (*map_pages)(struct vm_fault *vmf,
553 pgoff_t start_pgoff, pgoff_t end_pgoff);
554 unsigned long (*pagesize)(struct vm_area_struct * area);
556 /* notification that a previously read-only page is about to become
557 * writable, if an error is returned it will cause a SIGBUS */
558 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
560 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
561 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
563 /* called by access_process_vm when get_user_pages() fails, typically
564 * for use by special VMAs that can switch between memory and hardware
566 int (*access)(struct vm_area_struct *vma, unsigned long addr,
567 void *buf, int len, int write);
569 /* Called by the /proc/PID/maps code to ask the vma whether it
570 * has a special name. Returning non-NULL will also cause this
571 * vma to be dumped unconditionally. */
572 const char *(*name)(struct vm_area_struct *vma);
576 * set_policy() op must add a reference to any non-NULL @new mempolicy
577 * to hold the policy upon return. Caller should pass NULL @new to
578 * remove a policy and fall back to surrounding context--i.e. do not
579 * install a MPOL_DEFAULT policy, nor the task or system default
582 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
585 * get_policy() op must add reference [mpol_get()] to any policy at
586 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
587 * in mm/mempolicy.c will do this automatically.
588 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
589 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
590 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
591 * must return NULL--i.e., do not "fallback" to task or system default
594 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
598 * Called by vm_normal_page() for special PTEs to find the
599 * page for @addr. This is useful if the default behavior
600 * (using pte_page()) would not find the correct page.
602 struct page *(*find_special_page)(struct vm_area_struct *vma,
606 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
608 static const struct vm_operations_struct dummy_vm_ops = {};
610 memset(vma, 0, sizeof(*vma));
612 vma->vm_ops = &dummy_vm_ops;
613 INIT_LIST_HEAD(&vma->anon_vma_chain);
616 static inline void vma_set_anonymous(struct vm_area_struct *vma)
621 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
626 static inline bool vma_is_temporary_stack(struct vm_area_struct *vma)
628 int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP);
633 if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) ==
634 VM_STACK_INCOMPLETE_SETUP)
640 static inline bool vma_is_foreign(struct vm_area_struct *vma)
645 if (current->mm != vma->vm_mm)
651 static inline bool vma_is_accessible(struct vm_area_struct *vma)
653 return vma->vm_flags & VM_ACCESS_FLAGS;
658 * The vma_is_shmem is not inline because it is used only by slow
659 * paths in userfault.
661 bool vma_is_shmem(struct vm_area_struct *vma);
663 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
666 int vma_is_stack_for_current(struct vm_area_struct *vma);
668 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
669 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
675 * FIXME: take this include out, include page-flags.h in
676 * files which need it (119 of them)
678 #include <linux/page-flags.h>
679 #include <linux/huge_mm.h>
682 * Methods to modify the page usage count.
684 * What counts for a page usage:
685 * - cache mapping (page->mapping)
686 * - private data (page->private)
687 * - page mapped in a task's page tables, each mapping
688 * is counted separately
690 * Also, many kernel routines increase the page count before a critical
691 * routine so they can be sure the page doesn't go away from under them.
695 * Drop a ref, return true if the refcount fell to zero (the page has no users)
697 static inline int put_page_testzero(struct page *page)
699 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
700 return page_ref_dec_and_test(page);
704 * Try to grab a ref unless the page has a refcount of zero, return false if
706 * This can be called when MMU is off so it must not access
707 * any of the virtual mappings.
709 static inline int get_page_unless_zero(struct page *page)
711 return page_ref_add_unless(page, 1, 0);
714 extern int page_is_ram(unsigned long pfn);
722 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
725 /* Support for virtually mapped pages */
726 struct page *vmalloc_to_page(const void *addr);
727 unsigned long vmalloc_to_pfn(const void *addr);
730 * Determine if an address is within the vmalloc range
732 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
733 * is no special casing required.
736 #ifndef is_ioremap_addr
737 #define is_ioremap_addr(x) is_vmalloc_addr(x)
741 extern bool is_vmalloc_addr(const void *x);
742 extern int is_vmalloc_or_module_addr(const void *x);
744 static inline bool is_vmalloc_addr(const void *x)
748 static inline int is_vmalloc_or_module_addr(const void *x)
754 extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
755 static inline void *kvmalloc(size_t size, gfp_t flags)
757 return kvmalloc_node(size, flags, NUMA_NO_NODE);
759 static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
761 return kvmalloc_node(size, flags | __GFP_ZERO, node);
763 static inline void *kvzalloc(size_t size, gfp_t flags)
765 return kvmalloc(size, flags | __GFP_ZERO);
768 static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
772 if (unlikely(check_mul_overflow(n, size, &bytes)))
775 return kvmalloc(bytes, flags);
778 static inline void *kvcalloc(size_t n, size_t size, gfp_t flags)
780 return kvmalloc_array(n, size, flags | __GFP_ZERO);
783 extern void kvfree(const void *addr);
785 static inline int compound_mapcount(struct page *page)
787 VM_BUG_ON_PAGE(!PageCompound(page), page);
788 page = compound_head(page);
789 return atomic_read(compound_mapcount_ptr(page)) + 1;
793 * The atomic page->_mapcount, starts from -1: so that transitions
794 * both from it and to it can be tracked, using atomic_inc_and_test
795 * and atomic_add_negative(-1).
797 static inline void page_mapcount_reset(struct page *page)
799 atomic_set(&(page)->_mapcount, -1);
802 int __page_mapcount(struct page *page);
804 static inline int page_mapcount(struct page *page)
806 VM_BUG_ON_PAGE(PageSlab(page), page);
808 if (unlikely(PageCompound(page)))
809 return __page_mapcount(page);
810 return atomic_read(&page->_mapcount) + 1;
813 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
814 int total_mapcount(struct page *page);
815 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
817 static inline int total_mapcount(struct page *page)
819 return page_mapcount(page);
821 static inline int page_trans_huge_mapcount(struct page *page,
824 int mapcount = page_mapcount(page);
826 *total_mapcount = mapcount;
831 static inline struct page *virt_to_head_page(const void *x)
833 struct page *page = virt_to_page(x);
835 return compound_head(page);
838 void __put_page(struct page *page);
840 void put_pages_list(struct list_head *pages);
842 void split_page(struct page *page, unsigned int order);
845 * Compound pages have a destructor function. Provide a
846 * prototype for that function and accessor functions.
847 * These are _only_ valid on the head of a compound page.
849 typedef void compound_page_dtor(struct page *);
851 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
852 enum compound_dtor_id {
855 #ifdef CONFIG_HUGETLB_PAGE
858 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
863 extern compound_page_dtor * const compound_page_dtors[];
865 static inline void set_compound_page_dtor(struct page *page,
866 enum compound_dtor_id compound_dtor)
868 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
869 page[1].compound_dtor = compound_dtor;
872 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
874 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
875 return compound_page_dtors[page[1].compound_dtor];
878 static inline unsigned int compound_order(struct page *page)
882 return page[1].compound_order;
885 static inline bool hpage_pincount_available(struct page *page)
888 * Can the page->hpage_pinned_refcount field be used? That field is in
889 * the 3rd page of the compound page, so the smallest (2-page) compound
890 * pages cannot support it.
892 page = compound_head(page);
893 return PageCompound(page) && compound_order(page) > 1;
896 static inline int compound_pincount(struct page *page)
898 VM_BUG_ON_PAGE(!hpage_pincount_available(page), page);
899 page = compound_head(page);
900 return atomic_read(compound_pincount_ptr(page));
903 static inline void set_compound_order(struct page *page, unsigned int order)
905 page[1].compound_order = order;
908 /* Returns the number of pages in this potentially compound page. */
909 static inline unsigned long compound_nr(struct page *page)
911 return 1UL << compound_order(page);
914 /* Returns the number of bytes in this potentially compound page. */
915 static inline unsigned long page_size(struct page *page)
917 return PAGE_SIZE << compound_order(page);
920 /* Returns the number of bits needed for the number of bytes in a page */
921 static inline unsigned int page_shift(struct page *page)
923 return PAGE_SHIFT + compound_order(page);
926 void free_compound_page(struct page *page);
930 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
931 * servicing faults for write access. In the normal case, do always want
932 * pte_mkwrite. But get_user_pages can cause write faults for mappings
933 * that do not have writing enabled, when used by access_process_vm.
935 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
937 if (likely(vma->vm_flags & VM_WRITE))
938 pte = pte_mkwrite(pte);
942 vm_fault_t alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
944 vm_fault_t finish_fault(struct vm_fault *vmf);
945 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
949 * Multiple processes may "see" the same page. E.g. for untouched
950 * mappings of /dev/null, all processes see the same page full of
951 * zeroes, and text pages of executables and shared libraries have
952 * only one copy in memory, at most, normally.
954 * For the non-reserved pages, page_count(page) denotes a reference count.
955 * page_count() == 0 means the page is free. page->lru is then used for
956 * freelist management in the buddy allocator.
957 * page_count() > 0 means the page has been allocated.
959 * Pages are allocated by the slab allocator in order to provide memory
960 * to kmalloc and kmem_cache_alloc. In this case, the management of the
961 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
962 * unless a particular usage is carefully commented. (the responsibility of
963 * freeing the kmalloc memory is the caller's, of course).
965 * A page may be used by anyone else who does a __get_free_page().
966 * In this case, page_count still tracks the references, and should only
967 * be used through the normal accessor functions. The top bits of page->flags
968 * and page->virtual store page management information, but all other fields
969 * are unused and could be used privately, carefully. The management of this
970 * page is the responsibility of the one who allocated it, and those who have
971 * subsequently been given references to it.
973 * The other pages (we may call them "pagecache pages") are completely
974 * managed by the Linux memory manager: I/O, buffers, swapping etc.
975 * The following discussion applies only to them.
977 * A pagecache page contains an opaque `private' member, which belongs to the
978 * page's address_space. Usually, this is the address of a circular list of
979 * the page's disk buffers. PG_private must be set to tell the VM to call
980 * into the filesystem to release these pages.
982 * A page may belong to an inode's memory mapping. In this case, page->mapping
983 * is the pointer to the inode, and page->index is the file offset of the page,
984 * in units of PAGE_SIZE.
986 * If pagecache pages are not associated with an inode, they are said to be
987 * anonymous pages. These may become associated with the swapcache, and in that
988 * case PG_swapcache is set, and page->private is an offset into the swapcache.
990 * In either case (swapcache or inode backed), the pagecache itself holds one
991 * reference to the page. Setting PG_private should also increment the
992 * refcount. The each user mapping also has a reference to the page.
994 * The pagecache pages are stored in a per-mapping radix tree, which is
995 * rooted at mapping->i_pages, and indexed by offset.
996 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
997 * lists, we instead now tag pages as dirty/writeback in the radix tree.
999 * All pagecache pages may be subject to I/O:
1000 * - inode pages may need to be read from disk,
1001 * - inode pages which have been modified and are MAP_SHARED may need
1002 * to be written back to the inode on disk,
1003 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
1004 * modified may need to be swapped out to swap space and (later) to be read
1009 * The zone field is never updated after free_area_init_core()
1010 * sets it, so none of the operations on it need to be atomic.
1013 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
1014 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
1015 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
1016 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
1017 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
1018 #define KASAN_TAG_PGOFF (LAST_CPUPID_PGOFF - KASAN_TAG_WIDTH)
1021 * Define the bit shifts to access each section. For non-existent
1022 * sections we define the shift as 0; that plus a 0 mask ensures
1023 * the compiler will optimise away reference to them.
1025 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
1026 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
1027 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
1028 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
1029 #define KASAN_TAG_PGSHIFT (KASAN_TAG_PGOFF * (KASAN_TAG_WIDTH != 0))
1031 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
1032 #ifdef NODE_NOT_IN_PAGE_FLAGS
1033 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
1034 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
1035 SECTIONS_PGOFF : ZONES_PGOFF)
1037 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
1038 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
1039 NODES_PGOFF : ZONES_PGOFF)
1042 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
1044 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
1045 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
1046 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
1047 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
1048 #define KASAN_TAG_MASK ((1UL << KASAN_TAG_WIDTH) - 1)
1049 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
1051 static inline enum zone_type page_zonenum(const struct page *page)
1053 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
1056 #ifdef CONFIG_ZONE_DEVICE
1057 static inline bool is_zone_device_page(const struct page *page)
1059 return page_zonenum(page) == ZONE_DEVICE;
1061 extern void memmap_init_zone_device(struct zone *, unsigned long,
1062 unsigned long, struct dev_pagemap *);
1064 static inline bool is_zone_device_page(const struct page *page)
1070 #ifdef CONFIG_DEV_PAGEMAP_OPS
1071 void free_devmap_managed_page(struct page *page);
1072 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
1074 static inline bool page_is_devmap_managed(struct page *page)
1076 if (!static_branch_unlikely(&devmap_managed_key))
1078 if (!is_zone_device_page(page))
1080 switch (page->pgmap->type) {
1081 case MEMORY_DEVICE_PRIVATE:
1082 case MEMORY_DEVICE_FS_DAX:
1090 void put_devmap_managed_page(struct page *page);
1092 #else /* CONFIG_DEV_PAGEMAP_OPS */
1093 static inline bool page_is_devmap_managed(struct page *page)
1098 static inline void put_devmap_managed_page(struct page *page)
1101 #endif /* CONFIG_DEV_PAGEMAP_OPS */
1103 static inline bool is_device_private_page(const struct page *page)
1105 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
1106 IS_ENABLED(CONFIG_DEVICE_PRIVATE) &&
1107 is_zone_device_page(page) &&
1108 page->pgmap->type == MEMORY_DEVICE_PRIVATE;
1111 static inline bool is_pci_p2pdma_page(const struct page *page)
1113 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
1114 IS_ENABLED(CONFIG_PCI_P2PDMA) &&
1115 is_zone_device_page(page) &&
1116 page->pgmap->type == MEMORY_DEVICE_PCI_P2PDMA;
1119 /* 127: arbitrary random number, small enough to assemble well */
1120 #define page_ref_zero_or_close_to_overflow(page) \
1121 ((unsigned int) page_ref_count(page) + 127u <= 127u)
1123 static inline void get_page(struct page *page)
1125 page = compound_head(page);
1127 * Getting a normal page or the head of a compound page
1128 * requires to already have an elevated page->_refcount.
1130 VM_BUG_ON_PAGE(page_ref_zero_or_close_to_overflow(page), page);
1134 bool __must_check try_grab_page(struct page *page, unsigned int flags);
1136 static inline __must_check bool try_get_page(struct page *page)
1138 page = compound_head(page);
1139 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1145 static inline void put_page(struct page *page)
1147 page = compound_head(page);
1150 * For devmap managed pages we need to catch refcount transition from
1151 * 2 to 1, when refcount reach one it means the page is free and we
1152 * need to inform the device driver through callback. See
1153 * include/linux/memremap.h and HMM for details.
1155 if (page_is_devmap_managed(page)) {
1156 put_devmap_managed_page(page);
1160 if (put_page_testzero(page))
1165 * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
1166 * the page's refcount so that two separate items are tracked: the original page
1167 * reference count, and also a new count of how many pin_user_pages() calls were
1168 * made against the page. ("gup-pinned" is another term for the latter).
1170 * With this scheme, pin_user_pages() becomes special: such pages are marked as
1171 * distinct from normal pages. As such, the unpin_user_page() call (and its
1172 * variants) must be used in order to release gup-pinned pages.
1176 * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
1177 * counts with respect to pin_user_pages() and unpin_user_page() becomes
1178 * simpler, due to the fact that adding an even power of two to the page
1179 * refcount has the effect of using only the upper N bits, for the code that
1180 * counts up using the bias value. This means that the lower bits are left for
1181 * the exclusive use of the original code that increments and decrements by one
1182 * (or at least, by much smaller values than the bias value).
1184 * Of course, once the lower bits overflow into the upper bits (and this is
1185 * OK, because subtraction recovers the original values), then visual inspection
1186 * no longer suffices to directly view the separate counts. However, for normal
1187 * applications that don't have huge page reference counts, this won't be an
1190 * Locking: the lockless algorithm described in page_cache_get_speculative()
1191 * and page_cache_gup_pin_speculative() provides safe operation for
1192 * get_user_pages and page_mkclean and other calls that race to set up page
1195 #define GUP_PIN_COUNTING_BIAS (1U << 10)
1197 void unpin_user_page(struct page *page);
1198 void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1200 void unpin_user_pages(struct page **pages, unsigned long npages);
1203 * page_maybe_dma_pinned() - report if a page is pinned for DMA.
1205 * This function checks if a page has been pinned via a call to
1206 * pin_user_pages*().
1208 * For non-huge pages, the return value is partially fuzzy: false is not fuzzy,
1209 * because it means "definitely not pinned for DMA", but true means "probably
1210 * pinned for DMA, but possibly a false positive due to having at least
1211 * GUP_PIN_COUNTING_BIAS worth of normal page references".
1213 * False positives are OK, because: a) it's unlikely for a page to get that many
1214 * refcounts, and b) all the callers of this routine are expected to be able to
1215 * deal gracefully with a false positive.
1217 * For huge pages, the result will be exactly correct. That's because we have
1218 * more tracking data available: the 3rd struct page in the compound page is
1219 * used to track the pincount (instead using of the GUP_PIN_COUNTING_BIAS
1222 * For more information, please see Documentation/vm/pin_user_pages.rst.
1224 * @page: pointer to page to be queried.
1225 * @Return: True, if it is likely that the page has been "dma-pinned".
1226 * False, if the page is definitely not dma-pinned.
1228 static inline bool page_maybe_dma_pinned(struct page *page)
1230 if (hpage_pincount_available(page))
1231 return compound_pincount(page) > 0;
1234 * page_ref_count() is signed. If that refcount overflows, then
1235 * page_ref_count() returns a negative value, and callers will avoid
1236 * further incrementing the refcount.
1238 * Here, for that overflow case, use the signed bit to count a little
1239 * bit higher via unsigned math, and thus still get an accurate result.
1241 return ((unsigned int)page_ref_count(compound_head(page))) >=
1242 GUP_PIN_COUNTING_BIAS;
1245 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1246 #define SECTION_IN_PAGE_FLAGS
1250 * The identification function is mainly used by the buddy allocator for
1251 * determining if two pages could be buddies. We are not really identifying
1252 * the zone since we could be using the section number id if we do not have
1253 * node id available in page flags.
1254 * We only guarantee that it will return the same value for two combinable
1257 static inline int page_zone_id(struct page *page)
1259 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1262 #ifdef NODE_NOT_IN_PAGE_FLAGS
1263 extern int page_to_nid(const struct page *page);
1265 static inline int page_to_nid(const struct page *page)
1267 struct page *p = (struct page *)page;
1269 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1273 #ifdef CONFIG_NUMA_BALANCING
1274 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1276 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1279 static inline int cpupid_to_pid(int cpupid)
1281 return cpupid & LAST__PID_MASK;
1284 static inline int cpupid_to_cpu(int cpupid)
1286 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1289 static inline int cpupid_to_nid(int cpupid)
1291 return cpu_to_node(cpupid_to_cpu(cpupid));
1294 static inline bool cpupid_pid_unset(int cpupid)
1296 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1299 static inline bool cpupid_cpu_unset(int cpupid)
1301 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1304 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1306 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1309 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1310 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1311 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1313 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1316 static inline int page_cpupid_last(struct page *page)
1318 return page->_last_cpupid;
1320 static inline void page_cpupid_reset_last(struct page *page)
1322 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1325 static inline int page_cpupid_last(struct page *page)
1327 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1330 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1332 static inline void page_cpupid_reset_last(struct page *page)
1334 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1336 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1337 #else /* !CONFIG_NUMA_BALANCING */
1338 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1340 return page_to_nid(page); /* XXX */
1343 static inline int page_cpupid_last(struct page *page)
1345 return page_to_nid(page); /* XXX */
1348 static inline int cpupid_to_nid(int cpupid)
1353 static inline int cpupid_to_pid(int cpupid)
1358 static inline int cpupid_to_cpu(int cpupid)
1363 static inline int cpu_pid_to_cpupid(int nid, int pid)
1368 static inline bool cpupid_pid_unset(int cpupid)
1373 static inline void page_cpupid_reset_last(struct page *page)
1377 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1381 #endif /* CONFIG_NUMA_BALANCING */
1383 #ifdef CONFIG_KASAN_SW_TAGS
1384 static inline u8 page_kasan_tag(const struct page *page)
1386 return (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1389 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1391 page->flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1392 page->flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1395 static inline void page_kasan_tag_reset(struct page *page)
1397 page_kasan_tag_set(page, 0xff);
1400 static inline u8 page_kasan_tag(const struct page *page)
1405 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1406 static inline void page_kasan_tag_reset(struct page *page) { }
1409 static inline struct zone *page_zone(const struct page *page)
1411 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1414 static inline pg_data_t *page_pgdat(const struct page *page)
1416 return NODE_DATA(page_to_nid(page));
1419 #ifdef SECTION_IN_PAGE_FLAGS
1420 static inline void set_page_section(struct page *page, unsigned long section)
1422 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1423 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1426 static inline unsigned long page_to_section(const struct page *page)
1428 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1432 static inline void set_page_zone(struct page *page, enum zone_type zone)
1434 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1435 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1438 static inline void set_page_node(struct page *page, unsigned long node)
1440 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1441 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1444 static inline void set_page_links(struct page *page, enum zone_type zone,
1445 unsigned long node, unsigned long pfn)
1447 set_page_zone(page, zone);
1448 set_page_node(page, node);
1449 #ifdef SECTION_IN_PAGE_FLAGS
1450 set_page_section(page, pfn_to_section_nr(pfn));
1455 static inline struct mem_cgroup *page_memcg(struct page *page)
1457 return page->mem_cgroup;
1459 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1461 WARN_ON_ONCE(!rcu_read_lock_held());
1462 return READ_ONCE(page->mem_cgroup);
1465 static inline struct mem_cgroup *page_memcg(struct page *page)
1469 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1471 WARN_ON_ONCE(!rcu_read_lock_held());
1477 * Some inline functions in vmstat.h depend on page_zone()
1479 #include <linux/vmstat.h>
1481 static __always_inline void *lowmem_page_address(const struct page *page)
1483 return page_to_virt(page);
1486 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1487 #define HASHED_PAGE_VIRTUAL
1490 #if defined(WANT_PAGE_VIRTUAL)
1491 static inline void *page_address(const struct page *page)
1493 return page->virtual;
1495 static inline void set_page_address(struct page *page, void *address)
1497 page->virtual = address;
1499 #define page_address_init() do { } while(0)
1502 #if defined(HASHED_PAGE_VIRTUAL)
1503 void *page_address(const struct page *page);
1504 void set_page_address(struct page *page, void *virtual);
1505 void page_address_init(void);
1508 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1509 #define page_address(page) lowmem_page_address(page)
1510 #define set_page_address(page, address) do { } while(0)
1511 #define page_address_init() do { } while(0)
1514 extern void *page_rmapping(struct page *page);
1515 extern struct anon_vma *page_anon_vma(struct page *page);
1516 extern struct address_space *page_mapping(struct page *page);
1518 extern struct address_space *__page_file_mapping(struct page *);
1521 struct address_space *page_file_mapping(struct page *page)
1523 if (unlikely(PageSwapCache(page)))
1524 return __page_file_mapping(page);
1526 return page->mapping;
1529 extern pgoff_t __page_file_index(struct page *page);
1532 * Return the pagecache index of the passed page. Regular pagecache pages
1533 * use ->index whereas swapcache pages use swp_offset(->private)
1535 static inline pgoff_t page_index(struct page *page)
1537 if (unlikely(PageSwapCache(page)))
1538 return __page_file_index(page);
1542 bool page_mapped(struct page *page);
1543 struct address_space *page_mapping(struct page *page);
1544 struct address_space *page_mapping_file(struct page *page);
1547 * Return true only if the page has been allocated with
1548 * ALLOC_NO_WATERMARKS and the low watermark was not
1549 * met implying that the system is under some pressure.
1551 static inline bool page_is_pfmemalloc(struct page *page)
1554 * Page index cannot be this large so this must be
1555 * a pfmemalloc page.
1557 return page->index == -1UL;
1561 * Only to be called by the page allocator on a freshly allocated
1564 static inline void set_page_pfmemalloc(struct page *page)
1569 static inline void clear_page_pfmemalloc(struct page *page)
1575 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1577 extern void pagefault_out_of_memory(void);
1579 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1582 * Flags passed to show_mem() and show_free_areas() to suppress output in
1585 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1587 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1590 extern bool can_do_mlock(void);
1592 static inline bool can_do_mlock(void) { return false; }
1594 extern int user_shm_lock(size_t, struct user_struct *);
1595 extern void user_shm_unlock(size_t, struct user_struct *);
1598 * Parameter block passed down to zap_pte_range in exceptional cases.
1600 struct zap_details {
1601 struct address_space *check_mapping; /* Check page->mapping if set */
1602 pgoff_t first_index; /* Lowest page->index to unmap */
1603 pgoff_t last_index; /* Highest page->index to unmap */
1606 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1608 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1611 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1612 unsigned long size);
1613 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1614 unsigned long size);
1615 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1616 unsigned long start, unsigned long end);
1618 struct mmu_notifier_range;
1620 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1621 unsigned long end, unsigned long floor, unsigned long ceiling);
1622 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1623 struct vm_area_struct *vma);
1624 int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
1625 struct mmu_notifier_range *range,
1626 pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp);
1627 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1628 unsigned long *pfn);
1629 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1630 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1631 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1632 void *buf, int len, int write);
1634 extern void truncate_pagecache(struct inode *inode, loff_t new);
1635 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1636 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1637 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1638 int truncate_inode_page(struct address_space *mapping, struct page *page);
1639 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1640 int invalidate_inode_page(struct page *page);
1643 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1644 unsigned long address, unsigned int flags);
1645 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1646 unsigned long address, unsigned int fault_flags,
1648 void unmap_mapping_pages(struct address_space *mapping,
1649 pgoff_t start, pgoff_t nr, bool even_cows);
1650 void unmap_mapping_range(struct address_space *mapping,
1651 loff_t const holebegin, loff_t const holelen, int even_cows);
1653 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1654 unsigned long address, unsigned int flags)
1656 /* should never happen if there's no MMU */
1658 return VM_FAULT_SIGBUS;
1660 static inline int fixup_user_fault(struct task_struct *tsk,
1661 struct mm_struct *mm, unsigned long address,
1662 unsigned int fault_flags, bool *unlocked)
1664 /* should never happen if there's no MMU */
1668 static inline void unmap_mapping_pages(struct address_space *mapping,
1669 pgoff_t start, pgoff_t nr, bool even_cows) { }
1670 static inline void unmap_mapping_range(struct address_space *mapping,
1671 loff_t const holebegin, loff_t const holelen, int even_cows) { }
1674 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1675 loff_t const holebegin, loff_t const holelen)
1677 unmap_mapping_range(mapping, holebegin, holelen, 0);
1680 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1681 void *buf, int len, unsigned int gup_flags);
1682 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1683 void *buf, int len, unsigned int gup_flags);
1684 extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1685 unsigned long addr, void *buf, int len, unsigned int gup_flags);
1687 long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1688 unsigned long start, unsigned long nr_pages,
1689 unsigned int gup_flags, struct page **pages,
1690 struct vm_area_struct **vmas, int *locked);
1691 long pin_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1692 unsigned long start, unsigned long nr_pages,
1693 unsigned int gup_flags, struct page **pages,
1694 struct vm_area_struct **vmas, int *locked);
1695 long get_user_pages(unsigned long start, unsigned long nr_pages,
1696 unsigned int gup_flags, struct page **pages,
1697 struct vm_area_struct **vmas);
1698 long pin_user_pages(unsigned long start, unsigned long nr_pages,
1699 unsigned int gup_flags, struct page **pages,
1700 struct vm_area_struct **vmas);
1701 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1702 unsigned int gup_flags, struct page **pages, int *locked);
1703 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1704 struct page **pages, unsigned int gup_flags);
1706 int get_user_pages_fast(unsigned long start, int nr_pages,
1707 unsigned int gup_flags, struct page **pages);
1708 int pin_user_pages_fast(unsigned long start, int nr_pages,
1709 unsigned int gup_flags, struct page **pages);
1711 int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
1712 int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
1713 struct task_struct *task, bool bypass_rlim);
1715 /* Container for pinned pfns / pages */
1716 struct frame_vector {
1717 unsigned int nr_allocated; /* Number of frames we have space for */
1718 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1719 bool got_ref; /* Did we pin pages by getting page ref? */
1720 bool is_pfns; /* Does array contain pages or pfns? */
1721 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1722 * pfns_vector_pages() or pfns_vector_pfns()
1726 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1727 void frame_vector_destroy(struct frame_vector *vec);
1728 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1729 unsigned int gup_flags, struct frame_vector *vec);
1730 void put_vaddr_frames(struct frame_vector *vec);
1731 int frame_vector_to_pages(struct frame_vector *vec);
1732 void frame_vector_to_pfns(struct frame_vector *vec);
1734 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1736 return vec->nr_frames;
1739 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1742 int err = frame_vector_to_pages(vec);
1745 return ERR_PTR(err);
1747 return (struct page **)(vec->ptrs);
1750 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1753 frame_vector_to_pfns(vec);
1754 return (unsigned long *)(vec->ptrs);
1758 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1759 struct page **pages);
1760 int get_kernel_page(unsigned long start, int write, struct page **pages);
1761 struct page *get_dump_page(unsigned long addr);
1763 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1764 extern void do_invalidatepage(struct page *page, unsigned int offset,
1765 unsigned int length);
1767 void __set_page_dirty(struct page *, struct address_space *, int warn);
1768 int __set_page_dirty_nobuffers(struct page *page);
1769 int __set_page_dirty_no_writeback(struct page *page);
1770 int redirty_page_for_writepage(struct writeback_control *wbc,
1772 void account_page_dirtied(struct page *page, struct address_space *mapping);
1773 void account_page_cleaned(struct page *page, struct address_space *mapping,
1774 struct bdi_writeback *wb);
1775 int set_page_dirty(struct page *page);
1776 int set_page_dirty_lock(struct page *page);
1777 void __cancel_dirty_page(struct page *page);
1778 static inline void cancel_dirty_page(struct page *page)
1780 /* Avoid atomic ops, locking, etc. when not actually needed. */
1781 if (PageDirty(page))
1782 __cancel_dirty_page(page);
1784 int clear_page_dirty_for_io(struct page *page);
1786 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1788 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1789 unsigned long old_addr, struct vm_area_struct *new_vma,
1790 unsigned long new_addr, unsigned long len,
1791 bool need_rmap_locks);
1794 * Flags used by change_protection(). For now we make it a bitmap so
1795 * that we can pass in multiple flags just like parameters. However
1796 * for now all the callers are only use one of the flags at the same
1799 /* Whether we should allow dirty bit accounting */
1800 #define MM_CP_DIRTY_ACCT (1UL << 0)
1801 /* Whether this protection change is for NUMA hints */
1802 #define MM_CP_PROT_NUMA (1UL << 1)
1803 /* Whether this change is for write protecting */
1804 #define MM_CP_UFFD_WP (1UL << 2) /* do wp */
1805 #define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */
1806 #define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \
1807 MM_CP_UFFD_WP_RESOLVE)
1809 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1810 unsigned long end, pgprot_t newprot,
1811 unsigned long cp_flags);
1812 extern int mprotect_fixup(struct vm_area_struct *vma,
1813 struct vm_area_struct **pprev, unsigned long start,
1814 unsigned long end, unsigned long newflags);
1817 * doesn't attempt to fault and will return short.
1819 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1820 struct page **pages);
1822 * per-process(per-mm_struct) statistics.
1824 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1826 long val = atomic_long_read(&mm->rss_stat.count[member]);
1828 #ifdef SPLIT_RSS_COUNTING
1830 * counter is updated in asynchronous manner and may go to minus.
1831 * But it's never be expected number for users.
1836 return (unsigned long)val;
1839 void mm_trace_rss_stat(struct mm_struct *mm, int member, long count);
1841 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1843 long count = atomic_long_add_return(value, &mm->rss_stat.count[member]);
1845 mm_trace_rss_stat(mm, member, count);
1848 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1850 long count = atomic_long_inc_return(&mm->rss_stat.count[member]);
1852 mm_trace_rss_stat(mm, member, count);
1855 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1857 long count = atomic_long_dec_return(&mm->rss_stat.count[member]);
1859 mm_trace_rss_stat(mm, member, count);
1862 /* Optimized variant when page is already known not to be PageAnon */
1863 static inline int mm_counter_file(struct page *page)
1865 if (PageSwapBacked(page))
1866 return MM_SHMEMPAGES;
1867 return MM_FILEPAGES;
1870 static inline int mm_counter(struct page *page)
1873 return MM_ANONPAGES;
1874 return mm_counter_file(page);
1877 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1879 return get_mm_counter(mm, MM_FILEPAGES) +
1880 get_mm_counter(mm, MM_ANONPAGES) +
1881 get_mm_counter(mm, MM_SHMEMPAGES);
1884 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1886 return max(mm->hiwater_rss, get_mm_rss(mm));
1889 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1891 return max(mm->hiwater_vm, mm->total_vm);
1894 static inline void update_hiwater_rss(struct mm_struct *mm)
1896 unsigned long _rss = get_mm_rss(mm);
1898 if ((mm)->hiwater_rss < _rss)
1899 (mm)->hiwater_rss = _rss;
1902 static inline void update_hiwater_vm(struct mm_struct *mm)
1904 if (mm->hiwater_vm < mm->total_vm)
1905 mm->hiwater_vm = mm->total_vm;
1908 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1910 mm->hiwater_rss = get_mm_rss(mm);
1913 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1914 struct mm_struct *mm)
1916 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1918 if (*maxrss < hiwater_rss)
1919 *maxrss = hiwater_rss;
1922 #if defined(SPLIT_RSS_COUNTING)
1923 void sync_mm_rss(struct mm_struct *mm);
1925 static inline void sync_mm_rss(struct mm_struct *mm)
1930 #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
1931 static inline int pte_special(pte_t pte)
1936 static inline pte_t pte_mkspecial(pte_t pte)
1942 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
1943 static inline int pte_devmap(pte_t pte)
1949 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1951 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1953 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1957 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1961 #ifdef __PAGETABLE_P4D_FOLDED
1962 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1963 unsigned long address)
1968 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1971 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
1972 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1973 unsigned long address)
1977 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
1978 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
1981 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
1983 static inline void mm_inc_nr_puds(struct mm_struct *mm)
1985 if (mm_pud_folded(mm))
1987 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1990 static inline void mm_dec_nr_puds(struct mm_struct *mm)
1992 if (mm_pud_folded(mm))
1994 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1998 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1999 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
2000 unsigned long address)
2005 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
2006 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
2009 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
2011 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
2013 if (mm_pmd_folded(mm))
2015 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2018 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
2020 if (mm_pmd_folded(mm))
2022 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2027 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
2029 atomic_long_set(&mm->pgtables_bytes, 0);
2032 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2034 return atomic_long_read(&mm->pgtables_bytes);
2037 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
2039 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2042 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
2044 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2048 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
2049 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2054 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
2055 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
2058 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
2059 int __pte_alloc_kernel(pmd_t *pmd);
2061 #if defined(CONFIG_MMU)
2064 * The following ifdef needed to get the 5level-fixup.h header to work.
2065 * Remove it when 5level-fixup.h has been removed.
2067 #ifndef __ARCH_HAS_5LEVEL_HACK
2068 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2069 unsigned long address)
2071 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
2072 NULL : p4d_offset(pgd, address);
2075 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2076 unsigned long address)
2078 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
2079 NULL : pud_offset(p4d, address);
2081 #endif /* !__ARCH_HAS_5LEVEL_HACK */
2083 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
2085 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
2086 NULL: pmd_offset(pud, address);
2088 #endif /* CONFIG_MMU */
2090 #if USE_SPLIT_PTE_PTLOCKS
2091 #if ALLOC_SPLIT_PTLOCKS
2092 void __init ptlock_cache_init(void);
2093 extern bool ptlock_alloc(struct page *page);
2094 extern void ptlock_free(struct page *page);
2096 static inline spinlock_t *ptlock_ptr(struct page *page)
2100 #else /* ALLOC_SPLIT_PTLOCKS */
2101 static inline void ptlock_cache_init(void)
2105 static inline bool ptlock_alloc(struct page *page)
2110 static inline void ptlock_free(struct page *page)
2114 static inline spinlock_t *ptlock_ptr(struct page *page)
2118 #endif /* ALLOC_SPLIT_PTLOCKS */
2120 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2122 return ptlock_ptr(pmd_page(*pmd));
2125 static inline bool ptlock_init(struct page *page)
2128 * prep_new_page() initialize page->private (and therefore page->ptl)
2129 * with 0. Make sure nobody took it in use in between.
2131 * It can happen if arch try to use slab for page table allocation:
2132 * slab code uses page->slab_cache, which share storage with page->ptl.
2134 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
2135 if (!ptlock_alloc(page))
2137 spin_lock_init(ptlock_ptr(page));
2141 #else /* !USE_SPLIT_PTE_PTLOCKS */
2143 * We use mm->page_table_lock to guard all pagetable pages of the mm.
2145 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2147 return &mm->page_table_lock;
2149 static inline void ptlock_cache_init(void) {}
2150 static inline bool ptlock_init(struct page *page) { return true; }
2151 static inline void ptlock_free(struct page *page) {}
2152 #endif /* USE_SPLIT_PTE_PTLOCKS */
2154 static inline void pgtable_init(void)
2156 ptlock_cache_init();
2157 pgtable_cache_init();
2160 static inline bool pgtable_pte_page_ctor(struct page *page)
2162 if (!ptlock_init(page))
2164 __SetPageTable(page);
2165 inc_zone_page_state(page, NR_PAGETABLE);
2169 static inline void pgtable_pte_page_dtor(struct page *page)
2172 __ClearPageTable(page);
2173 dec_zone_page_state(page, NR_PAGETABLE);
2176 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
2178 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
2179 pte_t *__pte = pte_offset_map(pmd, address); \
2185 #define pte_unmap_unlock(pte, ptl) do { \
2190 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2192 #define pte_alloc_map(mm, pmd, address) \
2193 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2195 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2196 (pte_alloc(mm, pmd) ? \
2197 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2199 #define pte_alloc_kernel(pmd, address) \
2200 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2201 NULL: pte_offset_kernel(pmd, address))
2203 #if USE_SPLIT_PMD_PTLOCKS
2205 static struct page *pmd_to_page(pmd_t *pmd)
2207 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2208 return virt_to_page((void *)((unsigned long) pmd & mask));
2211 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2213 return ptlock_ptr(pmd_to_page(pmd));
2216 static inline bool pgtable_pmd_page_ctor(struct page *page)
2218 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2219 page->pmd_huge_pte = NULL;
2221 return ptlock_init(page);
2224 static inline void pgtable_pmd_page_dtor(struct page *page)
2226 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2227 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2232 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2236 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2238 return &mm->page_table_lock;
2241 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
2242 static inline void pgtable_pmd_page_dtor(struct page *page) {}
2244 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2248 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2250 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2256 * No scalability reason to split PUD locks yet, but follow the same pattern
2257 * as the PMD locks to make it easier if we decide to. The VM should not be
2258 * considered ready to switch to split PUD locks yet; there may be places
2259 * which need to be converted from page_table_lock.
2261 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2263 return &mm->page_table_lock;
2266 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2268 spinlock_t *ptl = pud_lockptr(mm, pud);
2274 extern void __init pagecache_init(void);
2275 extern void free_area_init(unsigned long * zones_size);
2276 extern void __init free_area_init_node(int nid, unsigned long * zones_size,
2277 unsigned long zone_start_pfn, unsigned long *zholes_size);
2278 extern void free_initmem(void);
2281 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2282 * into the buddy system. The freed pages will be poisoned with pattern
2283 * "poison" if it's within range [0, UCHAR_MAX].
2284 * Return pages freed into the buddy system.
2286 extern unsigned long free_reserved_area(void *start, void *end,
2287 int poison, const char *s);
2289 #ifdef CONFIG_HIGHMEM
2291 * Free a highmem page into the buddy system, adjusting totalhigh_pages
2292 * and totalram_pages.
2294 extern void free_highmem_page(struct page *page);
2297 extern void adjust_managed_page_count(struct page *page, long count);
2298 extern void mem_init_print_info(const char *str);
2300 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2302 /* Free the reserved page into the buddy system, so it gets managed. */
2303 static inline void __free_reserved_page(struct page *page)
2305 ClearPageReserved(page);
2306 init_page_count(page);
2310 static inline void free_reserved_page(struct page *page)
2312 __free_reserved_page(page);
2313 adjust_managed_page_count(page, 1);
2316 static inline void mark_page_reserved(struct page *page)
2318 SetPageReserved(page);
2319 adjust_managed_page_count(page, -1);
2323 * Default method to free all the __init memory into the buddy system.
2324 * The freed pages will be poisoned with pattern "poison" if it's within
2325 * range [0, UCHAR_MAX].
2326 * Return pages freed into the buddy system.
2328 static inline unsigned long free_initmem_default(int poison)
2330 extern char __init_begin[], __init_end[];
2332 return free_reserved_area(&__init_begin, &__init_end,
2333 poison, "unused kernel");
2336 static inline unsigned long get_num_physpages(void)
2339 unsigned long phys_pages = 0;
2341 for_each_online_node(nid)
2342 phys_pages += node_present_pages(nid);
2347 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
2349 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
2350 * zones, allocate the backing mem_map and account for memory holes in a more
2351 * architecture independent manner. This is a substitute for creating the
2352 * zone_sizes[] and zholes_size[] arrays and passing them to
2353 * free_area_init_node()
2355 * An architecture is expected to register range of page frames backed by
2356 * physical memory with memblock_add[_node]() before calling
2357 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
2358 * usage, an architecture is expected to do something like
2360 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2362 * for_each_valid_physical_page_range()
2363 * memblock_add_node(base, size, nid)
2364 * free_area_init_nodes(max_zone_pfns);
2366 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
2367 * registered physical page range. Similarly
2368 * sparse_memory_present_with_active_regions() calls memory_present() for
2369 * each range when SPARSEMEM is enabled.
2371 * See mm/page_alloc.c for more information on each function exposed by
2372 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
2374 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
2375 unsigned long node_map_pfn_alignment(void);
2376 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2377 unsigned long end_pfn);
2378 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2379 unsigned long end_pfn);
2380 extern void get_pfn_range_for_nid(unsigned int nid,
2381 unsigned long *start_pfn, unsigned long *end_pfn);
2382 extern unsigned long find_min_pfn_with_active_regions(void);
2383 extern void free_bootmem_with_active_regions(int nid,
2384 unsigned long max_low_pfn);
2385 extern void sparse_memory_present_with_active_regions(int nid);
2387 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
2389 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
2390 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
2391 static inline int __early_pfn_to_nid(unsigned long pfn,
2392 struct mminit_pfnnid_cache *state)
2397 /* please see mm/page_alloc.c */
2398 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2399 /* there is a per-arch backend function. */
2400 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
2401 struct mminit_pfnnid_cache *state);
2404 extern void set_dma_reserve(unsigned long new_dma_reserve);
2405 extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long,
2406 enum memmap_context, struct vmem_altmap *);
2407 extern void setup_per_zone_wmarks(void);
2408 extern int __meminit init_per_zone_wmark_min(void);
2409 extern void mem_init(void);
2410 extern void __init mmap_init(void);
2411 extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2412 extern long si_mem_available(void);
2413 extern void si_meminfo(struct sysinfo * val);
2414 extern void si_meminfo_node(struct sysinfo *val, int nid);
2415 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2416 extern unsigned long arch_reserved_kernel_pages(void);
2419 extern __printf(3, 4)
2420 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2422 extern void setup_per_cpu_pageset(void);
2425 extern int min_free_kbytes;
2426 extern int watermark_boost_factor;
2427 extern int watermark_scale_factor;
2430 extern atomic_long_t mmap_pages_allocated;
2431 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2433 /* interval_tree.c */
2434 void vma_interval_tree_insert(struct vm_area_struct *node,
2435 struct rb_root_cached *root);
2436 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2437 struct vm_area_struct *prev,
2438 struct rb_root_cached *root);
2439 void vma_interval_tree_remove(struct vm_area_struct *node,
2440 struct rb_root_cached *root);
2441 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2442 unsigned long start, unsigned long last);
2443 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2444 unsigned long start, unsigned long last);
2446 #define vma_interval_tree_foreach(vma, root, start, last) \
2447 for (vma = vma_interval_tree_iter_first(root, start, last); \
2448 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2450 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2451 struct rb_root_cached *root);
2452 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2453 struct rb_root_cached *root);
2454 struct anon_vma_chain *
2455 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2456 unsigned long start, unsigned long last);
2457 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2458 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2459 #ifdef CONFIG_DEBUG_VM_RB
2460 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2463 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2464 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2465 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2468 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2469 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2470 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2471 struct vm_area_struct *expand);
2472 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2473 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2475 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2477 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2478 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2479 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2480 struct mempolicy *, struct vm_userfaultfd_ctx);
2481 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2482 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2483 unsigned long addr, int new_below);
2484 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2485 unsigned long addr, int new_below);
2486 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2487 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2488 struct rb_node **, struct rb_node *);
2489 extern void unlink_file_vma(struct vm_area_struct *);
2490 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2491 unsigned long addr, unsigned long len, pgoff_t pgoff,
2492 bool *need_rmap_locks);
2493 extern void exit_mmap(struct mm_struct *);
2495 static inline int check_data_rlimit(unsigned long rlim,
2497 unsigned long start,
2498 unsigned long end_data,
2499 unsigned long start_data)
2501 if (rlim < RLIM_INFINITY) {
2502 if (((new - start) + (end_data - start_data)) > rlim)
2509 extern int mm_take_all_locks(struct mm_struct *mm);
2510 extern void mm_drop_all_locks(struct mm_struct *mm);
2512 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2513 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2514 extern struct file *get_task_exe_file(struct task_struct *task);
2516 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2517 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2519 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2520 const struct vm_special_mapping *sm);
2521 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2522 unsigned long addr, unsigned long len,
2523 unsigned long flags,
2524 const struct vm_special_mapping *spec);
2525 /* This is an obsolete alternative to _install_special_mapping. */
2526 extern int install_special_mapping(struct mm_struct *mm,
2527 unsigned long addr, unsigned long len,
2528 unsigned long flags, struct page **pages);
2530 unsigned long randomize_stack_top(unsigned long stack_top);
2532 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2534 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2535 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2536 struct list_head *uf);
2537 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2538 unsigned long len, unsigned long prot, unsigned long flags,
2539 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate,
2540 struct list_head *uf);
2541 extern int __do_munmap(struct mm_struct *, unsigned long, size_t,
2542 struct list_head *uf, bool downgrade);
2543 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2544 struct list_head *uf);
2545 extern int do_madvise(unsigned long start, size_t len_in, int behavior);
2547 static inline unsigned long
2548 do_mmap_pgoff(struct file *file, unsigned long addr,
2549 unsigned long len, unsigned long prot, unsigned long flags,
2550 unsigned long pgoff, unsigned long *populate,
2551 struct list_head *uf)
2553 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate, uf);
2557 extern int __mm_populate(unsigned long addr, unsigned long len,
2559 static inline void mm_populate(unsigned long addr, unsigned long len)
2562 (void) __mm_populate(addr, len, 1);
2565 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2568 /* These take the mm semaphore themselves */
2569 extern int __must_check vm_brk(unsigned long, unsigned long);
2570 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2571 extern int vm_munmap(unsigned long, size_t);
2572 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2573 unsigned long, unsigned long,
2574 unsigned long, unsigned long);
2576 struct vm_unmapped_area_info {
2577 #define VM_UNMAPPED_AREA_TOPDOWN 1
2578 unsigned long flags;
2579 unsigned long length;
2580 unsigned long low_limit;
2581 unsigned long high_limit;
2582 unsigned long align_mask;
2583 unsigned long align_offset;
2586 extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info);
2589 extern void truncate_inode_pages(struct address_space *, loff_t);
2590 extern void truncate_inode_pages_range(struct address_space *,
2591 loff_t lstart, loff_t lend);
2592 extern void truncate_inode_pages_final(struct address_space *);
2594 /* generic vm_area_ops exported for stackable file systems */
2595 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2596 extern void filemap_map_pages(struct vm_fault *vmf,
2597 pgoff_t start_pgoff, pgoff_t end_pgoff);
2598 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2600 /* mm/page-writeback.c */
2601 int __must_check write_one_page(struct page *page);
2602 void task_dirty_inc(struct task_struct *tsk);
2605 #define VM_READAHEAD_PAGES (SZ_128K / PAGE_SIZE)
2607 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2608 pgoff_t offset, unsigned long nr_to_read);
2610 void page_cache_sync_readahead(struct address_space *mapping,
2611 struct file_ra_state *ra,
2614 unsigned long size);
2616 void page_cache_async_readahead(struct address_space *mapping,
2617 struct file_ra_state *ra,
2621 unsigned long size);
2623 extern unsigned long stack_guard_gap;
2624 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2625 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2627 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2628 extern int expand_downwards(struct vm_area_struct *vma,
2629 unsigned long address);
2631 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2633 #define expand_upwards(vma, address) (0)
2636 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2637 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2638 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2639 struct vm_area_struct **pprev);
2641 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2642 NULL if none. Assume start_addr < end_addr. */
2643 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2645 struct vm_area_struct * vma = find_vma(mm,start_addr);
2647 if (vma && end_addr <= vma->vm_start)
2652 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2654 unsigned long vm_start = vma->vm_start;
2656 if (vma->vm_flags & VM_GROWSDOWN) {
2657 vm_start -= stack_guard_gap;
2658 if (vm_start > vma->vm_start)
2664 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2666 unsigned long vm_end = vma->vm_end;
2668 if (vma->vm_flags & VM_GROWSUP) {
2669 vm_end += stack_guard_gap;
2670 if (vm_end < vma->vm_end)
2671 vm_end = -PAGE_SIZE;
2676 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2678 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2681 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2682 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2683 unsigned long vm_start, unsigned long vm_end)
2685 struct vm_area_struct *vma = find_vma(mm, vm_start);
2687 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2693 static inline bool range_in_vma(struct vm_area_struct *vma,
2694 unsigned long start, unsigned long end)
2696 return (vma && vma->vm_start <= start && end <= vma->vm_end);
2700 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2701 void vma_set_page_prot(struct vm_area_struct *vma);
2703 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2707 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2709 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2713 #ifdef CONFIG_NUMA_BALANCING
2714 unsigned long change_prot_numa(struct vm_area_struct *vma,
2715 unsigned long start, unsigned long end);
2718 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2719 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2720 unsigned long pfn, unsigned long size, pgprot_t);
2721 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2722 int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
2723 struct page **pages, unsigned long *num);
2724 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
2726 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
2728 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2730 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2731 unsigned long pfn, pgprot_t pgprot);
2732 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2734 vm_fault_t vmf_insert_mixed_prot(struct vm_area_struct *vma, unsigned long addr,
2735 pfn_t pfn, pgprot_t pgprot);
2736 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
2737 unsigned long addr, pfn_t pfn);
2738 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2740 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2741 unsigned long addr, struct page *page)
2743 int err = vm_insert_page(vma, addr, page);
2746 return VM_FAULT_OOM;
2747 if (err < 0 && err != -EBUSY)
2748 return VM_FAULT_SIGBUS;
2750 return VM_FAULT_NOPAGE;
2753 static inline vm_fault_t vmf_error(int err)
2756 return VM_FAULT_OOM;
2757 return VM_FAULT_SIGBUS;
2760 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
2761 unsigned int foll_flags);
2763 #define FOLL_WRITE 0x01 /* check pte is writable */
2764 #define FOLL_TOUCH 0x02 /* mark page accessed */
2765 #define FOLL_GET 0x04 /* do get_page on page */
2766 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2767 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2768 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2769 * and return without waiting upon it */
2770 #define FOLL_POPULATE 0x40 /* fault in page */
2771 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2772 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2773 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2774 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2775 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2776 #define FOLL_MLOCK 0x1000 /* lock present pages */
2777 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2778 #define FOLL_COW 0x4000 /* internal GUP flag */
2779 #define FOLL_ANON 0x8000 /* don't do file mappings */
2780 #define FOLL_LONGTERM 0x10000 /* mapping lifetime is indefinite: see below */
2781 #define FOLL_SPLIT_PMD 0x20000 /* split huge pmd before returning */
2782 #define FOLL_PIN 0x40000 /* pages must be released via unpin_user_page */
2785 * FOLL_PIN and FOLL_LONGTERM may be used in various combinations with each
2786 * other. Here is what they mean, and how to use them:
2788 * FOLL_LONGTERM indicates that the page will be held for an indefinite time
2789 * period _often_ under userspace control. This is in contrast to
2790 * iov_iter_get_pages(), whose usages are transient.
2792 * FIXME: For pages which are part of a filesystem, mappings are subject to the
2793 * lifetime enforced by the filesystem and we need guarantees that longterm
2794 * users like RDMA and V4L2 only establish mappings which coordinate usage with
2795 * the filesystem. Ideas for this coordination include revoking the longterm
2796 * pin, delaying writeback, bounce buffer page writeback, etc. As FS DAX was
2797 * added after the problem with filesystems was found FS DAX VMAs are
2798 * specifically failed. Filesystem pages are still subject to bugs and use of
2799 * FOLL_LONGTERM should be avoided on those pages.
2801 * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call.
2802 * Currently only get_user_pages() and get_user_pages_fast() support this flag
2803 * and calls to get_user_pages_[un]locked are specifically not allowed. This
2804 * is due to an incompatibility with the FS DAX check and
2805 * FAULT_FLAG_ALLOW_RETRY.
2807 * In the CMA case: long term pins in a CMA region would unnecessarily fragment
2808 * that region. And so, CMA attempts to migrate the page before pinning, when
2809 * FOLL_LONGTERM is specified.
2811 * FOLL_PIN indicates that a special kind of tracking (not just page->_refcount,
2812 * but an additional pin counting system) will be invoked. This is intended for
2813 * anything that gets a page reference and then touches page data (for example,
2814 * Direct IO). This lets the filesystem know that some non-file-system entity is
2815 * potentially changing the pages' data. In contrast to FOLL_GET (whose pages
2816 * are released via put_page()), FOLL_PIN pages must be released, ultimately, by
2817 * a call to unpin_user_page().
2819 * FOLL_PIN is similar to FOLL_GET: both of these pin pages. They use different
2820 * and separate refcounting mechanisms, however, and that means that each has
2821 * its own acquire and release mechanisms:
2823 * FOLL_GET: get_user_pages*() to acquire, and put_page() to release.
2825 * FOLL_PIN: pin_user_pages*() to acquire, and unpin_user_pages to release.
2827 * FOLL_PIN and FOLL_GET are mutually exclusive for a given function call.
2828 * (The underlying pages may experience both FOLL_GET-based and FOLL_PIN-based
2829 * calls applied to them, and that's perfectly OK. This is a constraint on the
2830 * callers, not on the pages.)
2832 * FOLL_PIN should be set internally by the pin_user_pages*() APIs, never
2833 * directly by the caller. That's in order to help avoid mismatches when
2834 * releasing pages: get_user_pages*() pages must be released via put_page(),
2835 * while pin_user_pages*() pages must be released via unpin_user_page().
2837 * Please see Documentation/vm/pin_user_pages.rst for more information.
2840 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
2842 if (vm_fault & VM_FAULT_OOM)
2844 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2845 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2846 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2851 typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
2852 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2853 unsigned long size, pte_fn_t fn, void *data);
2854 extern int apply_to_existing_page_range(struct mm_struct *mm,
2855 unsigned long address, unsigned long size,
2856 pte_fn_t fn, void *data);
2858 #ifdef CONFIG_PAGE_POISONING
2859 extern bool page_poisoning_enabled(void);
2860 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2862 static inline bool page_poisoning_enabled(void) { return false; }
2863 static inline void kernel_poison_pages(struct page *page, int numpages,
2867 #ifdef CONFIG_INIT_ON_ALLOC_DEFAULT_ON
2868 DECLARE_STATIC_KEY_TRUE(init_on_alloc);
2870 DECLARE_STATIC_KEY_FALSE(init_on_alloc);
2872 static inline bool want_init_on_alloc(gfp_t flags)
2874 if (static_branch_unlikely(&init_on_alloc) &&
2875 !page_poisoning_enabled())
2877 return flags & __GFP_ZERO;
2880 #ifdef CONFIG_INIT_ON_FREE_DEFAULT_ON
2881 DECLARE_STATIC_KEY_TRUE(init_on_free);
2883 DECLARE_STATIC_KEY_FALSE(init_on_free);
2885 static inline bool want_init_on_free(void)
2887 return static_branch_unlikely(&init_on_free) &&
2888 !page_poisoning_enabled();
2891 #ifdef CONFIG_DEBUG_PAGEALLOC
2892 extern void init_debug_pagealloc(void);
2894 static inline void init_debug_pagealloc(void) {}
2896 extern bool _debug_pagealloc_enabled_early;
2897 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
2899 static inline bool debug_pagealloc_enabled(void)
2901 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
2902 _debug_pagealloc_enabled_early;
2906 * For use in fast paths after init_debug_pagealloc() has run, or when a
2907 * false negative result is not harmful when called too early.
2909 static inline bool debug_pagealloc_enabled_static(void)
2911 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
2914 return static_branch_unlikely(&_debug_pagealloc_enabled);
2917 #if defined(CONFIG_DEBUG_PAGEALLOC) || defined(CONFIG_ARCH_HAS_SET_DIRECT_MAP)
2918 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2921 * When called in DEBUG_PAGEALLOC context, the call should most likely be
2922 * guarded by debug_pagealloc_enabled() or debug_pagealloc_enabled_static()
2925 kernel_map_pages(struct page *page, int numpages, int enable)
2927 __kernel_map_pages(page, numpages, enable);
2929 #ifdef CONFIG_HIBERNATION
2930 extern bool kernel_page_present(struct page *page);
2931 #endif /* CONFIG_HIBERNATION */
2932 #else /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */
2934 kernel_map_pages(struct page *page, int numpages, int enable) {}
2935 #ifdef CONFIG_HIBERNATION
2936 static inline bool kernel_page_present(struct page *page) { return true; }
2937 #endif /* CONFIG_HIBERNATION */
2938 #endif /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */
2940 #ifdef __HAVE_ARCH_GATE_AREA
2941 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2942 extern int in_gate_area_no_mm(unsigned long addr);
2943 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2945 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2949 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2950 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2954 #endif /* __HAVE_ARCH_GATE_AREA */
2956 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2958 #ifdef CONFIG_SYSCTL
2959 extern int sysctl_drop_caches;
2960 int drop_caches_sysctl_handler(struct ctl_table *, int, void *, size_t *,
2964 void drop_slab(void);
2965 void drop_slab_node(int nid);
2968 #define randomize_va_space 0
2970 extern int randomize_va_space;
2973 const char * arch_vma_name(struct vm_area_struct *vma);
2975 void print_vma_addr(char *prefix, unsigned long rip);
2977 static inline void print_vma_addr(char *prefix, unsigned long rip)
2982 void *sparse_buffer_alloc(unsigned long size);
2983 struct page * __populate_section_memmap(unsigned long pfn,
2984 unsigned long nr_pages, int nid, struct vmem_altmap *altmap);
2985 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2986 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
2987 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
2988 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2989 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2990 void *vmemmap_alloc_block(unsigned long size, int node);
2992 void *vmemmap_alloc_block_buf(unsigned long size, int node);
2993 void *altmap_alloc_block_buf(unsigned long size, struct vmem_altmap *altmap);
2994 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2995 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2997 int vmemmap_populate(unsigned long start, unsigned long end, int node,
2998 struct vmem_altmap *altmap);
2999 void vmemmap_populate_print_last(void);
3000 #ifdef CONFIG_MEMORY_HOTPLUG
3001 void vmemmap_free(unsigned long start, unsigned long end,
3002 struct vmem_altmap *altmap);
3004 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
3005 unsigned long nr_pages);
3008 MF_COUNT_INCREASED = 1 << 0,
3009 MF_ACTION_REQUIRED = 1 << 1,
3010 MF_MUST_KILL = 1 << 2,
3011 MF_SOFT_OFFLINE = 1 << 3,
3013 extern int memory_failure(unsigned long pfn, int flags);
3014 extern void memory_failure_queue(unsigned long pfn, int flags);
3015 extern int unpoison_memory(unsigned long pfn);
3016 extern int get_hwpoison_page(struct page *page);
3017 #define put_hwpoison_page(page) put_page(page)
3018 extern int sysctl_memory_failure_early_kill;
3019 extern int sysctl_memory_failure_recovery;
3020 extern void shake_page(struct page *p, int access);
3021 extern atomic_long_t num_poisoned_pages __read_mostly;
3022 extern int soft_offline_page(unsigned long pfn, int flags);
3026 * Error handlers for various types of pages.
3029 MF_IGNORED, /* Error: cannot be handled */
3030 MF_FAILED, /* Error: handling failed */
3031 MF_DELAYED, /* Will be handled later */
3032 MF_RECOVERED, /* Successfully recovered */
3035 enum mf_action_page_type {
3037 MF_MSG_KERNEL_HIGH_ORDER,
3039 MF_MSG_DIFFERENT_COMPOUND,
3040 MF_MSG_POISONED_HUGE,
3043 MF_MSG_NON_PMD_HUGE,
3044 MF_MSG_UNMAP_FAILED,
3045 MF_MSG_DIRTY_SWAPCACHE,
3046 MF_MSG_CLEAN_SWAPCACHE,
3047 MF_MSG_DIRTY_MLOCKED_LRU,
3048 MF_MSG_CLEAN_MLOCKED_LRU,
3049 MF_MSG_DIRTY_UNEVICTABLE_LRU,
3050 MF_MSG_CLEAN_UNEVICTABLE_LRU,
3053 MF_MSG_TRUNCATED_LRU,
3060 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
3061 extern void clear_huge_page(struct page *page,
3062 unsigned long addr_hint,
3063 unsigned int pages_per_huge_page);
3064 extern void copy_user_huge_page(struct page *dst, struct page *src,
3065 unsigned long addr_hint,
3066 struct vm_area_struct *vma,
3067 unsigned int pages_per_huge_page);
3068 extern long copy_huge_page_from_user(struct page *dst_page,
3069 const void __user *usr_src,
3070 unsigned int pages_per_huge_page,
3071 bool allow_pagefault);
3074 * vma_is_special_huge - Are transhuge page-table entries considered special?
3075 * @vma: Pointer to the struct vm_area_struct to consider
3077 * Whether transhuge page-table entries are considered "special" following
3078 * the definition in vm_normal_page().
3080 * Return: true if transhuge page-table entries should be considered special,
3083 static inline bool vma_is_special_huge(const struct vm_area_struct *vma)
3085 return vma_is_dax(vma) || (vma->vm_file &&
3086 (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP)));
3089 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
3091 #ifdef CONFIG_DEBUG_PAGEALLOC
3092 extern unsigned int _debug_guardpage_minorder;
3093 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
3095 static inline unsigned int debug_guardpage_minorder(void)
3097 return _debug_guardpage_minorder;
3100 static inline bool debug_guardpage_enabled(void)
3102 return static_branch_unlikely(&_debug_guardpage_enabled);
3105 static inline bool page_is_guard(struct page *page)
3107 if (!debug_guardpage_enabled())
3110 return PageGuard(page);
3113 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
3114 static inline bool debug_guardpage_enabled(void) { return false; }
3115 static inline bool page_is_guard(struct page *page) { return false; }
3116 #endif /* CONFIG_DEBUG_PAGEALLOC */
3118 #if MAX_NUMNODES > 1
3119 void __init setup_nr_node_ids(void);
3121 static inline void setup_nr_node_ids(void) {}
3124 extern int memcmp_pages(struct page *page1, struct page *page2);
3126 static inline int pages_identical(struct page *page1, struct page *page2)
3128 return !memcmp_pages(page1, page2);
3131 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
3132 unsigned long clean_record_shared_mapping_range(struct address_space *mapping,
3133 pgoff_t first_index, pgoff_t nr,
3134 pgoff_t bitmap_pgoff,
3135 unsigned long *bitmap,
3139 unsigned long wp_shared_mapping_range(struct address_space *mapping,
3140 pgoff_t first_index, pgoff_t nr);
3143 extern int sysctl_nr_trim_pages;
3145 #endif /* __KERNEL__ */
3146 #endif /* _LINUX_MM_H */