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_ARM64)
329 # define VM_ARM64_BTI VM_ARCH_1 /* BTI guarded page, a.k.a. GP bit */
330 # define VM_ARCH_CLEAR VM_ARM64_BTI
331 #elif !defined(CONFIG_MMU)
332 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
336 # define VM_GROWSUP VM_NONE
339 /* Bits set in the VMA until the stack is in its final location */
340 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
342 #define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0)
344 /* Common data flag combinations */
345 #define VM_DATA_FLAGS_TSK_EXEC (VM_READ | VM_WRITE | TASK_EXEC | \
346 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
347 #define VM_DATA_FLAGS_NON_EXEC (VM_READ | VM_WRITE | VM_MAYREAD | \
348 VM_MAYWRITE | VM_MAYEXEC)
349 #define VM_DATA_FLAGS_EXEC (VM_READ | VM_WRITE | VM_EXEC | \
350 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
352 #ifndef VM_DATA_DEFAULT_FLAGS /* arch can override this */
353 #define VM_DATA_DEFAULT_FLAGS VM_DATA_FLAGS_EXEC
356 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
357 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
360 #ifdef CONFIG_STACK_GROWSUP
361 #define VM_STACK VM_GROWSUP
363 #define VM_STACK VM_GROWSDOWN
366 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
368 /* VMA basic access permission flags */
369 #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
373 * Special vmas that are non-mergable, non-mlock()able.
375 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
377 /* This mask prevents VMA from being scanned with khugepaged */
378 #define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB)
380 /* This mask defines which mm->def_flags a process can inherit its parent */
381 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
383 /* This mask is used to clear all the VMA flags used by mlock */
384 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
386 /* Arch-specific flags to clear when updating VM flags on protection change */
387 #ifndef VM_ARCH_CLEAR
388 # define VM_ARCH_CLEAR VM_NONE
390 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
393 * mapping from the currently active vm_flags protection bits (the
394 * low four bits) to a page protection mask..
396 extern pgprot_t protection_map[16];
399 * Fault flag definitions.
401 * @FAULT_FLAG_WRITE: Fault was a write fault.
402 * @FAULT_FLAG_MKWRITE: Fault was mkwrite of existing PTE.
403 * @FAULT_FLAG_ALLOW_RETRY: Allow to retry the fault if blocked.
404 * @FAULT_FLAG_RETRY_NOWAIT: Don't drop mmap_sem and wait when retrying.
405 * @FAULT_FLAG_KILLABLE: The fault task is in SIGKILL killable region.
406 * @FAULT_FLAG_TRIED: The fault has been tried once.
407 * @FAULT_FLAG_USER: The fault originated in userspace.
408 * @FAULT_FLAG_REMOTE: The fault is not for current task/mm.
409 * @FAULT_FLAG_INSTRUCTION: The fault was during an instruction fetch.
410 * @FAULT_FLAG_INTERRUPTIBLE: The fault can be interrupted by non-fatal signals.
412 * About @FAULT_FLAG_ALLOW_RETRY and @FAULT_FLAG_TRIED: we can specify
413 * whether we would allow page faults to retry by specifying these two
414 * fault flags correctly. Currently there can be three legal combinations:
416 * (a) ALLOW_RETRY and !TRIED: this means the page fault allows retry, and
417 * this is the first try
419 * (b) ALLOW_RETRY and TRIED: this means the page fault allows retry, and
420 * we've already tried at least once
422 * (c) !ALLOW_RETRY and !TRIED: this means the page fault does not allow retry
424 * The unlisted combination (!ALLOW_RETRY && TRIED) is illegal and should never
425 * be used. Note that page faults can be allowed to retry for multiple times,
426 * in which case we'll have an initial fault with flags (a) then later on
427 * continuous faults with flags (b). We should always try to detect pending
428 * signals before a retry to make sure the continuous page faults can still be
429 * interrupted if necessary.
431 #define FAULT_FLAG_WRITE 0x01
432 #define FAULT_FLAG_MKWRITE 0x02
433 #define FAULT_FLAG_ALLOW_RETRY 0x04
434 #define FAULT_FLAG_RETRY_NOWAIT 0x08
435 #define FAULT_FLAG_KILLABLE 0x10
436 #define FAULT_FLAG_TRIED 0x20
437 #define FAULT_FLAG_USER 0x40
438 #define FAULT_FLAG_REMOTE 0x80
439 #define FAULT_FLAG_INSTRUCTION 0x100
440 #define FAULT_FLAG_INTERRUPTIBLE 0x200
443 * The default fault flags that should be used by most of the
444 * arch-specific page fault handlers.
446 #define FAULT_FLAG_DEFAULT (FAULT_FLAG_ALLOW_RETRY | \
447 FAULT_FLAG_KILLABLE | \
448 FAULT_FLAG_INTERRUPTIBLE)
451 * fault_flag_allow_retry_first - check ALLOW_RETRY the first time
453 * This is mostly used for places where we want to try to avoid taking
454 * the mmap_sem for too long a time when waiting for another condition
455 * to change, in which case we can try to be polite to release the
456 * mmap_sem in the first round to avoid potential starvation of other
457 * processes that would also want the mmap_sem.
459 * Return: true if the page fault allows retry and this is the first
460 * attempt of the fault handling; false otherwise.
462 static inline bool fault_flag_allow_retry_first(unsigned int flags)
464 return (flags & FAULT_FLAG_ALLOW_RETRY) &&
465 (!(flags & FAULT_FLAG_TRIED));
468 #define FAULT_FLAG_TRACE \
469 { FAULT_FLAG_WRITE, "WRITE" }, \
470 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
471 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
472 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
473 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
474 { FAULT_FLAG_TRIED, "TRIED" }, \
475 { FAULT_FLAG_USER, "USER" }, \
476 { FAULT_FLAG_REMOTE, "REMOTE" }, \
477 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }, \
478 { FAULT_FLAG_INTERRUPTIBLE, "INTERRUPTIBLE" }
481 * vm_fault is filled by the the pagefault handler and passed to the vma's
482 * ->fault function. The vma's ->fault is responsible for returning a bitmask
483 * of VM_FAULT_xxx flags that give details about how the fault was handled.
485 * MM layer fills up gfp_mask for page allocations but fault handler might
486 * alter it if its implementation requires a different allocation context.
488 * pgoff should be used in favour of virtual_address, if possible.
491 struct vm_area_struct *vma; /* Target VMA */
492 unsigned int flags; /* FAULT_FLAG_xxx flags */
493 gfp_t gfp_mask; /* gfp mask to be used for allocations */
494 pgoff_t pgoff; /* Logical page offset based on vma */
495 unsigned long address; /* Faulting virtual address */
496 pmd_t *pmd; /* Pointer to pmd entry matching
498 pud_t *pud; /* Pointer to pud entry matching
501 pte_t orig_pte; /* Value of PTE at the time of fault */
503 struct page *cow_page; /* Page handler may use for COW fault */
504 struct page *page; /* ->fault handlers should return a
505 * page here, unless VM_FAULT_NOPAGE
506 * is set (which is also implied by
509 /* These three entries are valid only while holding ptl lock */
510 pte_t *pte; /* Pointer to pte entry matching
511 * the 'address'. NULL if the page
512 * table hasn't been allocated.
514 spinlock_t *ptl; /* Page table lock.
515 * Protects pte page table if 'pte'
516 * is not NULL, otherwise pmd.
518 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
519 * vm_ops->map_pages() calls
520 * alloc_set_pte() from atomic context.
521 * do_fault_around() pre-allocates
522 * page table to avoid allocation from
527 /* page entry size for vm->huge_fault() */
528 enum page_entry_size {
535 * These are the virtual MM functions - opening of an area, closing and
536 * unmapping it (needed to keep files on disk up-to-date etc), pointer
537 * to the functions called when a no-page or a wp-page exception occurs.
539 struct vm_operations_struct {
540 void (*open)(struct vm_area_struct * area);
541 void (*close)(struct vm_area_struct * area);
542 int (*split)(struct vm_area_struct * area, unsigned long addr);
543 int (*mremap)(struct vm_area_struct * area);
544 vm_fault_t (*fault)(struct vm_fault *vmf);
545 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
546 enum page_entry_size pe_size);
547 void (*map_pages)(struct vm_fault *vmf,
548 pgoff_t start_pgoff, pgoff_t end_pgoff);
549 unsigned long (*pagesize)(struct vm_area_struct * area);
551 /* notification that a previously read-only page is about to become
552 * writable, if an error is returned it will cause a SIGBUS */
553 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
555 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
556 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
558 /* called by access_process_vm when get_user_pages() fails, typically
559 * for use by special VMAs that can switch between memory and hardware
561 int (*access)(struct vm_area_struct *vma, unsigned long addr,
562 void *buf, int len, int write);
564 /* Called by the /proc/PID/maps code to ask the vma whether it
565 * has a special name. Returning non-NULL will also cause this
566 * vma to be dumped unconditionally. */
567 const char *(*name)(struct vm_area_struct *vma);
571 * set_policy() op must add a reference to any non-NULL @new mempolicy
572 * to hold the policy upon return. Caller should pass NULL @new to
573 * remove a policy and fall back to surrounding context--i.e. do not
574 * install a MPOL_DEFAULT policy, nor the task or system default
577 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
580 * get_policy() op must add reference [mpol_get()] to any policy at
581 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
582 * in mm/mempolicy.c will do this automatically.
583 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
584 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
585 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
586 * must return NULL--i.e., do not "fallback" to task or system default
589 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
593 * Called by vm_normal_page() for special PTEs to find the
594 * page for @addr. This is useful if the default behavior
595 * (using pte_page()) would not find the correct page.
597 struct page *(*find_special_page)(struct vm_area_struct *vma,
601 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
603 static const struct vm_operations_struct dummy_vm_ops = {};
605 memset(vma, 0, sizeof(*vma));
607 vma->vm_ops = &dummy_vm_ops;
608 INIT_LIST_HEAD(&vma->anon_vma_chain);
611 static inline void vma_set_anonymous(struct vm_area_struct *vma)
616 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
621 static inline bool vma_is_temporary_stack(struct vm_area_struct *vma)
623 int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP);
628 if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) ==
629 VM_STACK_INCOMPLETE_SETUP)
635 static inline bool vma_is_foreign(struct vm_area_struct *vma)
640 if (current->mm != vma->vm_mm)
646 static inline bool vma_is_accessible(struct vm_area_struct *vma)
648 return vma->vm_flags & VM_ACCESS_FLAGS;
653 * The vma_is_shmem is not inline because it is used only by slow
654 * paths in userfault.
656 bool vma_is_shmem(struct vm_area_struct *vma);
658 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
661 int vma_is_stack_for_current(struct vm_area_struct *vma);
663 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
664 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
670 * FIXME: take this include out, include page-flags.h in
671 * files which need it (119 of them)
673 #include <linux/page-flags.h>
674 #include <linux/huge_mm.h>
677 * Methods to modify the page usage count.
679 * What counts for a page usage:
680 * - cache mapping (page->mapping)
681 * - private data (page->private)
682 * - page mapped in a task's page tables, each mapping
683 * is counted separately
685 * Also, many kernel routines increase the page count before a critical
686 * routine so they can be sure the page doesn't go away from under them.
690 * Drop a ref, return true if the refcount fell to zero (the page has no users)
692 static inline int put_page_testzero(struct page *page)
694 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
695 return page_ref_dec_and_test(page);
699 * Try to grab a ref unless the page has a refcount of zero, return false if
701 * This can be called when MMU is off so it must not access
702 * any of the virtual mappings.
704 static inline int get_page_unless_zero(struct page *page)
706 return page_ref_add_unless(page, 1, 0);
709 extern int page_is_ram(unsigned long pfn);
717 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
720 /* Support for virtually mapped pages */
721 struct page *vmalloc_to_page(const void *addr);
722 unsigned long vmalloc_to_pfn(const void *addr);
725 * Determine if an address is within the vmalloc range
727 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
728 * is no special casing required.
731 #ifndef is_ioremap_addr
732 #define is_ioremap_addr(x) is_vmalloc_addr(x)
736 extern bool is_vmalloc_addr(const void *x);
737 extern int is_vmalloc_or_module_addr(const void *x);
739 static inline bool is_vmalloc_addr(const void *x)
743 static inline int is_vmalloc_or_module_addr(const void *x)
749 extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
750 static inline void *kvmalloc(size_t size, gfp_t flags)
752 return kvmalloc_node(size, flags, NUMA_NO_NODE);
754 static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
756 return kvmalloc_node(size, flags | __GFP_ZERO, node);
758 static inline void *kvzalloc(size_t size, gfp_t flags)
760 return kvmalloc(size, flags | __GFP_ZERO);
763 static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
767 if (unlikely(check_mul_overflow(n, size, &bytes)))
770 return kvmalloc(bytes, flags);
773 static inline void *kvcalloc(size_t n, size_t size, gfp_t flags)
775 return kvmalloc_array(n, size, flags | __GFP_ZERO);
778 extern void kvfree(const void *addr);
781 * Mapcount of compound page as a whole, does not include mapped sub-pages.
783 * Must be called only for compound pages or any their tail sub-pages.
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);
805 * Mapcount of 0-order page; when compound sub-page, includes
806 * compound_mapcount().
808 * Result is undefined for pages which cannot be mapped into userspace.
809 * For example SLAB or special types of pages. See function page_has_type().
810 * They use this place in struct page differently.
812 static inline int page_mapcount(struct page *page)
814 if (unlikely(PageCompound(page)))
815 return __page_mapcount(page);
816 return atomic_read(&page->_mapcount) + 1;
819 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
820 int total_mapcount(struct page *page);
821 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
823 static inline int total_mapcount(struct page *page)
825 return page_mapcount(page);
827 static inline int page_trans_huge_mapcount(struct page *page,
830 int mapcount = page_mapcount(page);
832 *total_mapcount = mapcount;
837 static inline struct page *virt_to_head_page(const void *x)
839 struct page *page = virt_to_page(x);
841 return compound_head(page);
844 void __put_page(struct page *page);
846 void put_pages_list(struct list_head *pages);
848 void split_page(struct page *page, unsigned int order);
851 * Compound pages have a destructor function. Provide a
852 * prototype for that function and accessor functions.
853 * These are _only_ valid on the head of a compound page.
855 typedef void compound_page_dtor(struct page *);
857 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
858 enum compound_dtor_id {
861 #ifdef CONFIG_HUGETLB_PAGE
864 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
869 extern compound_page_dtor * const compound_page_dtors[NR_COMPOUND_DTORS];
871 static inline void set_compound_page_dtor(struct page *page,
872 enum compound_dtor_id compound_dtor)
874 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
875 page[1].compound_dtor = compound_dtor;
878 static inline void destroy_compound_page(struct page *page)
880 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
881 compound_page_dtors[page[1].compound_dtor](page);
884 static inline unsigned int compound_order(struct page *page)
888 return page[1].compound_order;
891 static inline bool hpage_pincount_available(struct page *page)
894 * Can the page->hpage_pinned_refcount field be used? That field is in
895 * the 3rd page of the compound page, so the smallest (2-page) compound
896 * pages cannot support it.
898 page = compound_head(page);
899 return PageCompound(page) && compound_order(page) > 1;
902 static inline int compound_pincount(struct page *page)
904 VM_BUG_ON_PAGE(!hpage_pincount_available(page), page);
905 page = compound_head(page);
906 return atomic_read(compound_pincount_ptr(page));
909 static inline void set_compound_order(struct page *page, unsigned int order)
911 page[1].compound_order = order;
914 /* Returns the number of pages in this potentially compound page. */
915 static inline unsigned long compound_nr(struct page *page)
917 return 1UL << compound_order(page);
920 /* Returns the number of bytes in this potentially compound page. */
921 static inline unsigned long page_size(struct page *page)
923 return PAGE_SIZE << compound_order(page);
926 /* Returns the number of bits needed for the number of bytes in a page */
927 static inline unsigned int page_shift(struct page *page)
929 return PAGE_SHIFT + compound_order(page);
932 void free_compound_page(struct page *page);
936 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
937 * servicing faults for write access. In the normal case, do always want
938 * pte_mkwrite. But get_user_pages can cause write faults for mappings
939 * that do not have writing enabled, when used by access_process_vm.
941 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
943 if (likely(vma->vm_flags & VM_WRITE))
944 pte = pte_mkwrite(pte);
948 vm_fault_t alloc_set_pte(struct vm_fault *vmf, struct page *page);
949 vm_fault_t finish_fault(struct vm_fault *vmf);
950 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
954 * Multiple processes may "see" the same page. E.g. for untouched
955 * mappings of /dev/null, all processes see the same page full of
956 * zeroes, and text pages of executables and shared libraries have
957 * only one copy in memory, at most, normally.
959 * For the non-reserved pages, page_count(page) denotes a reference count.
960 * page_count() == 0 means the page is free. page->lru is then used for
961 * freelist management in the buddy allocator.
962 * page_count() > 0 means the page has been allocated.
964 * Pages are allocated by the slab allocator in order to provide memory
965 * to kmalloc and kmem_cache_alloc. In this case, the management of the
966 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
967 * unless a particular usage is carefully commented. (the responsibility of
968 * freeing the kmalloc memory is the caller's, of course).
970 * A page may be used by anyone else who does a __get_free_page().
971 * In this case, page_count still tracks the references, and should only
972 * be used through the normal accessor functions. The top bits of page->flags
973 * and page->virtual store page management information, but all other fields
974 * are unused and could be used privately, carefully. The management of this
975 * page is the responsibility of the one who allocated it, and those who have
976 * subsequently been given references to it.
978 * The other pages (we may call them "pagecache pages") are completely
979 * managed by the Linux memory manager: I/O, buffers, swapping etc.
980 * The following discussion applies only to them.
982 * A pagecache page contains an opaque `private' member, which belongs to the
983 * page's address_space. Usually, this is the address of a circular list of
984 * the page's disk buffers. PG_private must be set to tell the VM to call
985 * into the filesystem to release these pages.
987 * A page may belong to an inode's memory mapping. In this case, page->mapping
988 * is the pointer to the inode, and page->index is the file offset of the page,
989 * in units of PAGE_SIZE.
991 * If pagecache pages are not associated with an inode, they are said to be
992 * anonymous pages. These may become associated with the swapcache, and in that
993 * case PG_swapcache is set, and page->private is an offset into the swapcache.
995 * In either case (swapcache or inode backed), the pagecache itself holds one
996 * reference to the page. Setting PG_private should also increment the
997 * refcount. The each user mapping also has a reference to the page.
999 * The pagecache pages are stored in a per-mapping radix tree, which is
1000 * rooted at mapping->i_pages, and indexed by offset.
1001 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
1002 * lists, we instead now tag pages as dirty/writeback in the radix tree.
1004 * All pagecache pages may be subject to I/O:
1005 * - inode pages may need to be read from disk,
1006 * - inode pages which have been modified and are MAP_SHARED may need
1007 * to be written back to the inode on disk,
1008 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
1009 * modified may need to be swapped out to swap space and (later) to be read
1014 * The zone field is never updated after free_area_init_core()
1015 * sets it, so none of the operations on it need to be atomic.
1018 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
1019 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
1020 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
1021 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
1022 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
1023 #define KASAN_TAG_PGOFF (LAST_CPUPID_PGOFF - KASAN_TAG_WIDTH)
1026 * Define the bit shifts to access each section. For non-existent
1027 * sections we define the shift as 0; that plus a 0 mask ensures
1028 * the compiler will optimise away reference to them.
1030 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
1031 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
1032 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
1033 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
1034 #define KASAN_TAG_PGSHIFT (KASAN_TAG_PGOFF * (KASAN_TAG_WIDTH != 0))
1036 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
1037 #ifdef NODE_NOT_IN_PAGE_FLAGS
1038 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
1039 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
1040 SECTIONS_PGOFF : ZONES_PGOFF)
1042 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
1043 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
1044 NODES_PGOFF : ZONES_PGOFF)
1047 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
1049 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
1050 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
1051 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
1052 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
1053 #define KASAN_TAG_MASK ((1UL << KASAN_TAG_WIDTH) - 1)
1054 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
1056 static inline enum zone_type page_zonenum(const struct page *page)
1058 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
1061 #ifdef CONFIG_ZONE_DEVICE
1062 static inline bool is_zone_device_page(const struct page *page)
1064 return page_zonenum(page) == ZONE_DEVICE;
1066 extern void memmap_init_zone_device(struct zone *, unsigned long,
1067 unsigned long, struct dev_pagemap *);
1069 static inline bool is_zone_device_page(const struct page *page)
1075 #ifdef CONFIG_DEV_PAGEMAP_OPS
1076 void free_devmap_managed_page(struct page *page);
1077 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
1079 static inline bool page_is_devmap_managed(struct page *page)
1081 if (!static_branch_unlikely(&devmap_managed_key))
1083 if (!is_zone_device_page(page))
1085 switch (page->pgmap->type) {
1086 case MEMORY_DEVICE_PRIVATE:
1087 case MEMORY_DEVICE_FS_DAX:
1095 void put_devmap_managed_page(struct page *page);
1097 #else /* CONFIG_DEV_PAGEMAP_OPS */
1098 static inline bool page_is_devmap_managed(struct page *page)
1103 static inline void put_devmap_managed_page(struct page *page)
1106 #endif /* CONFIG_DEV_PAGEMAP_OPS */
1108 static inline bool is_device_private_page(const struct page *page)
1110 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
1111 IS_ENABLED(CONFIG_DEVICE_PRIVATE) &&
1112 is_zone_device_page(page) &&
1113 page->pgmap->type == MEMORY_DEVICE_PRIVATE;
1116 static inline bool is_pci_p2pdma_page(const struct page *page)
1118 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
1119 IS_ENABLED(CONFIG_PCI_P2PDMA) &&
1120 is_zone_device_page(page) &&
1121 page->pgmap->type == MEMORY_DEVICE_PCI_P2PDMA;
1124 /* 127: arbitrary random number, small enough to assemble well */
1125 #define page_ref_zero_or_close_to_overflow(page) \
1126 ((unsigned int) page_ref_count(page) + 127u <= 127u)
1128 static inline void get_page(struct page *page)
1130 page = compound_head(page);
1132 * Getting a normal page or the head of a compound page
1133 * requires to already have an elevated page->_refcount.
1135 VM_BUG_ON_PAGE(page_ref_zero_or_close_to_overflow(page), page);
1139 bool __must_check try_grab_page(struct page *page, unsigned int flags);
1141 static inline __must_check bool try_get_page(struct page *page)
1143 page = compound_head(page);
1144 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1150 static inline void put_page(struct page *page)
1152 page = compound_head(page);
1155 * For devmap managed pages we need to catch refcount transition from
1156 * 2 to 1, when refcount reach one it means the page is free and we
1157 * need to inform the device driver through callback. See
1158 * include/linux/memremap.h and HMM for details.
1160 if (page_is_devmap_managed(page)) {
1161 put_devmap_managed_page(page);
1165 if (put_page_testzero(page))
1170 * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
1171 * the page's refcount so that two separate items are tracked: the original page
1172 * reference count, and also a new count of how many pin_user_pages() calls were
1173 * made against the page. ("gup-pinned" is another term for the latter).
1175 * With this scheme, pin_user_pages() becomes special: such pages are marked as
1176 * distinct from normal pages. As such, the unpin_user_page() call (and its
1177 * variants) must be used in order to release gup-pinned pages.
1181 * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
1182 * counts with respect to pin_user_pages() and unpin_user_page() becomes
1183 * simpler, due to the fact that adding an even power of two to the page
1184 * refcount has the effect of using only the upper N bits, for the code that
1185 * counts up using the bias value. This means that the lower bits are left for
1186 * the exclusive use of the original code that increments and decrements by one
1187 * (or at least, by much smaller values than the bias value).
1189 * Of course, once the lower bits overflow into the upper bits (and this is
1190 * OK, because subtraction recovers the original values), then visual inspection
1191 * no longer suffices to directly view the separate counts. However, for normal
1192 * applications that don't have huge page reference counts, this won't be an
1195 * Locking: the lockless algorithm described in page_cache_get_speculative()
1196 * and page_cache_gup_pin_speculative() provides safe operation for
1197 * get_user_pages and page_mkclean and other calls that race to set up page
1200 #define GUP_PIN_COUNTING_BIAS (1U << 10)
1202 void unpin_user_page(struct page *page);
1203 void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1205 void unpin_user_pages(struct page **pages, unsigned long npages);
1208 * page_maybe_dma_pinned() - report if a page is pinned for DMA.
1210 * This function checks if a page has been pinned via a call to
1211 * pin_user_pages*().
1213 * For non-huge pages, the return value is partially fuzzy: false is not fuzzy,
1214 * because it means "definitely not pinned for DMA", but true means "probably
1215 * pinned for DMA, but possibly a false positive due to having at least
1216 * GUP_PIN_COUNTING_BIAS worth of normal page references".
1218 * False positives are OK, because: a) it's unlikely for a page to get that many
1219 * refcounts, and b) all the callers of this routine are expected to be able to
1220 * deal gracefully with a false positive.
1222 * For huge pages, the result will be exactly correct. That's because we have
1223 * more tracking data available: the 3rd struct page in the compound page is
1224 * used to track the pincount (instead using of the GUP_PIN_COUNTING_BIAS
1227 * For more information, please see Documentation/core-api/pin_user_pages.rst.
1229 * @page: pointer to page to be queried.
1230 * @Return: True, if it is likely that the page has been "dma-pinned".
1231 * False, if the page is definitely not dma-pinned.
1233 static inline bool page_maybe_dma_pinned(struct page *page)
1235 if (hpage_pincount_available(page))
1236 return compound_pincount(page) > 0;
1239 * page_ref_count() is signed. If that refcount overflows, then
1240 * page_ref_count() returns a negative value, and callers will avoid
1241 * further incrementing the refcount.
1243 * Here, for that overflow case, use the signed bit to count a little
1244 * bit higher via unsigned math, and thus still get an accurate result.
1246 return ((unsigned int)page_ref_count(compound_head(page))) >=
1247 GUP_PIN_COUNTING_BIAS;
1250 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1251 #define SECTION_IN_PAGE_FLAGS
1255 * The identification function is mainly used by the buddy allocator for
1256 * determining if two pages could be buddies. We are not really identifying
1257 * the zone since we could be using the section number id if we do not have
1258 * node id available in page flags.
1259 * We only guarantee that it will return the same value for two combinable
1262 static inline int page_zone_id(struct page *page)
1264 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1267 #ifdef NODE_NOT_IN_PAGE_FLAGS
1268 extern int page_to_nid(const struct page *page);
1270 static inline int page_to_nid(const struct page *page)
1272 struct page *p = (struct page *)page;
1274 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1278 #ifdef CONFIG_NUMA_BALANCING
1279 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1281 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1284 static inline int cpupid_to_pid(int cpupid)
1286 return cpupid & LAST__PID_MASK;
1289 static inline int cpupid_to_cpu(int cpupid)
1291 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1294 static inline int cpupid_to_nid(int cpupid)
1296 return cpu_to_node(cpupid_to_cpu(cpupid));
1299 static inline bool cpupid_pid_unset(int cpupid)
1301 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1304 static inline bool cpupid_cpu_unset(int cpupid)
1306 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1309 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1311 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1314 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1315 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1316 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1318 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1321 static inline int page_cpupid_last(struct page *page)
1323 return page->_last_cpupid;
1325 static inline void page_cpupid_reset_last(struct page *page)
1327 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1330 static inline int page_cpupid_last(struct page *page)
1332 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1335 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1337 static inline void page_cpupid_reset_last(struct page *page)
1339 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1341 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1342 #else /* !CONFIG_NUMA_BALANCING */
1343 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1345 return page_to_nid(page); /* XXX */
1348 static inline int page_cpupid_last(struct page *page)
1350 return page_to_nid(page); /* XXX */
1353 static inline int cpupid_to_nid(int cpupid)
1358 static inline int cpupid_to_pid(int cpupid)
1363 static inline int cpupid_to_cpu(int cpupid)
1368 static inline int cpu_pid_to_cpupid(int nid, int pid)
1373 static inline bool cpupid_pid_unset(int cpupid)
1378 static inline void page_cpupid_reset_last(struct page *page)
1382 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1386 #endif /* CONFIG_NUMA_BALANCING */
1388 #ifdef CONFIG_KASAN_SW_TAGS
1389 static inline u8 page_kasan_tag(const struct page *page)
1391 return (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1394 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1396 page->flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1397 page->flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1400 static inline void page_kasan_tag_reset(struct page *page)
1402 page_kasan_tag_set(page, 0xff);
1405 static inline u8 page_kasan_tag(const struct page *page)
1410 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1411 static inline void page_kasan_tag_reset(struct page *page) { }
1414 static inline struct zone *page_zone(const struct page *page)
1416 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1419 static inline pg_data_t *page_pgdat(const struct page *page)
1421 return NODE_DATA(page_to_nid(page));
1424 #ifdef SECTION_IN_PAGE_FLAGS
1425 static inline void set_page_section(struct page *page, unsigned long section)
1427 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1428 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1431 static inline unsigned long page_to_section(const struct page *page)
1433 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1437 static inline void set_page_zone(struct page *page, enum zone_type zone)
1439 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1440 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1443 static inline void set_page_node(struct page *page, unsigned long node)
1445 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1446 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1449 static inline void set_page_links(struct page *page, enum zone_type zone,
1450 unsigned long node, unsigned long pfn)
1452 set_page_zone(page, zone);
1453 set_page_node(page, node);
1454 #ifdef SECTION_IN_PAGE_FLAGS
1455 set_page_section(page, pfn_to_section_nr(pfn));
1460 static inline struct mem_cgroup *page_memcg(struct page *page)
1462 return page->mem_cgroup;
1464 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1466 WARN_ON_ONCE(!rcu_read_lock_held());
1467 return READ_ONCE(page->mem_cgroup);
1470 static inline struct mem_cgroup *page_memcg(struct page *page)
1474 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1476 WARN_ON_ONCE(!rcu_read_lock_held());
1482 * Some inline functions in vmstat.h depend on page_zone()
1484 #include <linux/vmstat.h>
1486 static __always_inline void *lowmem_page_address(const struct page *page)
1488 return page_to_virt(page);
1491 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1492 #define HASHED_PAGE_VIRTUAL
1495 #if defined(WANT_PAGE_VIRTUAL)
1496 static inline void *page_address(const struct page *page)
1498 return page->virtual;
1500 static inline void set_page_address(struct page *page, void *address)
1502 page->virtual = address;
1504 #define page_address_init() do { } while(0)
1507 #if defined(HASHED_PAGE_VIRTUAL)
1508 void *page_address(const struct page *page);
1509 void set_page_address(struct page *page, void *virtual);
1510 void page_address_init(void);
1513 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1514 #define page_address(page) lowmem_page_address(page)
1515 #define set_page_address(page, address) do { } while(0)
1516 #define page_address_init() do { } while(0)
1519 extern void *page_rmapping(struct page *page);
1520 extern struct anon_vma *page_anon_vma(struct page *page);
1521 extern struct address_space *page_mapping(struct page *page);
1523 extern struct address_space *__page_file_mapping(struct page *);
1526 struct address_space *page_file_mapping(struct page *page)
1528 if (unlikely(PageSwapCache(page)))
1529 return __page_file_mapping(page);
1531 return page->mapping;
1534 extern pgoff_t __page_file_index(struct page *page);
1537 * Return the pagecache index of the passed page. Regular pagecache pages
1538 * use ->index whereas swapcache pages use swp_offset(->private)
1540 static inline pgoff_t page_index(struct page *page)
1542 if (unlikely(PageSwapCache(page)))
1543 return __page_file_index(page);
1547 bool page_mapped(struct page *page);
1548 struct address_space *page_mapping(struct page *page);
1549 struct address_space *page_mapping_file(struct page *page);
1552 * Return true only if the page has been allocated with
1553 * ALLOC_NO_WATERMARKS and the low watermark was not
1554 * met implying that the system is under some pressure.
1556 static inline bool page_is_pfmemalloc(struct page *page)
1559 * Page index cannot be this large so this must be
1560 * a pfmemalloc page.
1562 return page->index == -1UL;
1566 * Only to be called by the page allocator on a freshly allocated
1569 static inline void set_page_pfmemalloc(struct page *page)
1574 static inline void clear_page_pfmemalloc(struct page *page)
1580 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1582 extern void pagefault_out_of_memory(void);
1584 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1587 * Flags passed to show_mem() and show_free_areas() to suppress output in
1590 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1592 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1595 extern bool can_do_mlock(void);
1597 static inline bool can_do_mlock(void) { return false; }
1599 extern int user_shm_lock(size_t, struct user_struct *);
1600 extern void user_shm_unlock(size_t, struct user_struct *);
1603 * Parameter block passed down to zap_pte_range in exceptional cases.
1605 struct zap_details {
1606 struct address_space *check_mapping; /* Check page->mapping if set */
1607 pgoff_t first_index; /* Lowest page->index to unmap */
1608 pgoff_t last_index; /* Highest page->index to unmap */
1611 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1613 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1616 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1617 unsigned long size);
1618 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1619 unsigned long size);
1620 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1621 unsigned long start, unsigned long end);
1623 struct mmu_notifier_range;
1625 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1626 unsigned long end, unsigned long floor, unsigned long ceiling);
1627 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1628 struct vm_area_struct *vma);
1629 int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
1630 struct mmu_notifier_range *range,
1631 pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp);
1632 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1633 unsigned long *pfn);
1634 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1635 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1636 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1637 void *buf, int len, int write);
1639 extern void truncate_pagecache(struct inode *inode, loff_t new);
1640 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1641 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1642 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1643 int truncate_inode_page(struct address_space *mapping, struct page *page);
1644 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1645 int invalidate_inode_page(struct page *page);
1648 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1649 unsigned long address, unsigned int flags);
1650 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1651 unsigned long address, unsigned int fault_flags,
1653 void unmap_mapping_pages(struct address_space *mapping,
1654 pgoff_t start, pgoff_t nr, bool even_cows);
1655 void unmap_mapping_range(struct address_space *mapping,
1656 loff_t const holebegin, loff_t const holelen, int even_cows);
1658 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1659 unsigned long address, unsigned int flags)
1661 /* should never happen if there's no MMU */
1663 return VM_FAULT_SIGBUS;
1665 static inline int fixup_user_fault(struct task_struct *tsk,
1666 struct mm_struct *mm, unsigned long address,
1667 unsigned int fault_flags, bool *unlocked)
1669 /* should never happen if there's no MMU */
1673 static inline void unmap_mapping_pages(struct address_space *mapping,
1674 pgoff_t start, pgoff_t nr, bool even_cows) { }
1675 static inline void unmap_mapping_range(struct address_space *mapping,
1676 loff_t const holebegin, loff_t const holelen, int even_cows) { }
1679 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1680 loff_t const holebegin, loff_t const holelen)
1682 unmap_mapping_range(mapping, holebegin, holelen, 0);
1685 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1686 void *buf, int len, unsigned int gup_flags);
1687 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1688 void *buf, int len, unsigned int gup_flags);
1689 extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1690 unsigned long addr, void *buf, int len, unsigned int gup_flags);
1692 long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1693 unsigned long start, unsigned long nr_pages,
1694 unsigned int gup_flags, struct page **pages,
1695 struct vm_area_struct **vmas, int *locked);
1696 long pin_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1697 unsigned long start, unsigned long nr_pages,
1698 unsigned int gup_flags, struct page **pages,
1699 struct vm_area_struct **vmas, int *locked);
1700 long get_user_pages(unsigned long start, unsigned long nr_pages,
1701 unsigned int gup_flags, struct page **pages,
1702 struct vm_area_struct **vmas);
1703 long pin_user_pages(unsigned long start, unsigned long nr_pages,
1704 unsigned int gup_flags, struct page **pages,
1705 struct vm_area_struct **vmas);
1706 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1707 unsigned int gup_flags, struct page **pages, int *locked);
1708 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1709 struct page **pages, unsigned int gup_flags);
1710 long pin_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1711 struct page **pages, unsigned int gup_flags);
1713 int get_user_pages_fast(unsigned long start, int nr_pages,
1714 unsigned int gup_flags, struct page **pages);
1715 int pin_user_pages_fast(unsigned long start, int nr_pages,
1716 unsigned int gup_flags, struct page **pages);
1718 int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
1719 int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
1720 struct task_struct *task, bool bypass_rlim);
1722 /* Container for pinned pfns / pages */
1723 struct frame_vector {
1724 unsigned int nr_allocated; /* Number of frames we have space for */
1725 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1726 bool got_ref; /* Did we pin pages by getting page ref? */
1727 bool is_pfns; /* Does array contain pages or pfns? */
1728 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1729 * pfns_vector_pages() or pfns_vector_pfns()
1733 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1734 void frame_vector_destroy(struct frame_vector *vec);
1735 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1736 unsigned int gup_flags, struct frame_vector *vec);
1737 void put_vaddr_frames(struct frame_vector *vec);
1738 int frame_vector_to_pages(struct frame_vector *vec);
1739 void frame_vector_to_pfns(struct frame_vector *vec);
1741 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1743 return vec->nr_frames;
1746 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1749 int err = frame_vector_to_pages(vec);
1752 return ERR_PTR(err);
1754 return (struct page **)(vec->ptrs);
1757 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1760 frame_vector_to_pfns(vec);
1761 return (unsigned long *)(vec->ptrs);
1765 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1766 struct page **pages);
1767 int get_kernel_page(unsigned long start, int write, struct page **pages);
1768 struct page *get_dump_page(unsigned long addr);
1770 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1771 extern void do_invalidatepage(struct page *page, unsigned int offset,
1772 unsigned int length);
1774 void __set_page_dirty(struct page *, struct address_space *, int warn);
1775 int __set_page_dirty_nobuffers(struct page *page);
1776 int __set_page_dirty_no_writeback(struct page *page);
1777 int redirty_page_for_writepage(struct writeback_control *wbc,
1779 void account_page_dirtied(struct page *page, struct address_space *mapping);
1780 void account_page_cleaned(struct page *page, struct address_space *mapping,
1781 struct bdi_writeback *wb);
1782 int set_page_dirty(struct page *page);
1783 int set_page_dirty_lock(struct page *page);
1784 void __cancel_dirty_page(struct page *page);
1785 static inline void cancel_dirty_page(struct page *page)
1787 /* Avoid atomic ops, locking, etc. when not actually needed. */
1788 if (PageDirty(page))
1789 __cancel_dirty_page(page);
1791 int clear_page_dirty_for_io(struct page *page);
1793 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1795 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1796 unsigned long old_addr, struct vm_area_struct *new_vma,
1797 unsigned long new_addr, unsigned long len,
1798 bool need_rmap_locks);
1801 * Flags used by change_protection(). For now we make it a bitmap so
1802 * that we can pass in multiple flags just like parameters. However
1803 * for now all the callers are only use one of the flags at the same
1806 /* Whether we should allow dirty bit accounting */
1807 #define MM_CP_DIRTY_ACCT (1UL << 0)
1808 /* Whether this protection change is for NUMA hints */
1809 #define MM_CP_PROT_NUMA (1UL << 1)
1810 /* Whether this change is for write protecting */
1811 #define MM_CP_UFFD_WP (1UL << 2) /* do wp */
1812 #define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */
1813 #define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \
1814 MM_CP_UFFD_WP_RESOLVE)
1816 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1817 unsigned long end, pgprot_t newprot,
1818 unsigned long cp_flags);
1819 extern int mprotect_fixup(struct vm_area_struct *vma,
1820 struct vm_area_struct **pprev, unsigned long start,
1821 unsigned long end, unsigned long newflags);
1824 * doesn't attempt to fault and will return short.
1826 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1827 struct page **pages);
1828 int pin_user_pages_fast_only(unsigned long start, int nr_pages,
1829 unsigned int gup_flags, struct page **pages);
1831 * per-process(per-mm_struct) statistics.
1833 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1835 long val = atomic_long_read(&mm->rss_stat.count[member]);
1837 #ifdef SPLIT_RSS_COUNTING
1839 * counter is updated in asynchronous manner and may go to minus.
1840 * But it's never be expected number for users.
1845 return (unsigned long)val;
1848 void mm_trace_rss_stat(struct mm_struct *mm, int member, long count);
1850 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1852 long count = atomic_long_add_return(value, &mm->rss_stat.count[member]);
1854 mm_trace_rss_stat(mm, member, count);
1857 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1859 long count = atomic_long_inc_return(&mm->rss_stat.count[member]);
1861 mm_trace_rss_stat(mm, member, count);
1864 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1866 long count = atomic_long_dec_return(&mm->rss_stat.count[member]);
1868 mm_trace_rss_stat(mm, member, count);
1871 /* Optimized variant when page is already known not to be PageAnon */
1872 static inline int mm_counter_file(struct page *page)
1874 if (PageSwapBacked(page))
1875 return MM_SHMEMPAGES;
1876 return MM_FILEPAGES;
1879 static inline int mm_counter(struct page *page)
1882 return MM_ANONPAGES;
1883 return mm_counter_file(page);
1886 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1888 return get_mm_counter(mm, MM_FILEPAGES) +
1889 get_mm_counter(mm, MM_ANONPAGES) +
1890 get_mm_counter(mm, MM_SHMEMPAGES);
1893 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1895 return max(mm->hiwater_rss, get_mm_rss(mm));
1898 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1900 return max(mm->hiwater_vm, mm->total_vm);
1903 static inline void update_hiwater_rss(struct mm_struct *mm)
1905 unsigned long _rss = get_mm_rss(mm);
1907 if ((mm)->hiwater_rss < _rss)
1908 (mm)->hiwater_rss = _rss;
1911 static inline void update_hiwater_vm(struct mm_struct *mm)
1913 if (mm->hiwater_vm < mm->total_vm)
1914 mm->hiwater_vm = mm->total_vm;
1917 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1919 mm->hiwater_rss = get_mm_rss(mm);
1922 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1923 struct mm_struct *mm)
1925 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1927 if (*maxrss < hiwater_rss)
1928 *maxrss = hiwater_rss;
1931 #if defined(SPLIT_RSS_COUNTING)
1932 void sync_mm_rss(struct mm_struct *mm);
1934 static inline void sync_mm_rss(struct mm_struct *mm)
1939 #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
1940 static inline int pte_special(pte_t pte)
1945 static inline pte_t pte_mkspecial(pte_t pte)
1951 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
1952 static inline int pte_devmap(pte_t pte)
1958 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1960 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1962 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1966 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1970 #ifdef __PAGETABLE_P4D_FOLDED
1971 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1972 unsigned long address)
1977 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1980 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
1981 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1982 unsigned long address)
1986 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
1987 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
1990 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
1992 static inline void mm_inc_nr_puds(struct mm_struct *mm)
1994 if (mm_pud_folded(mm))
1996 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1999 static inline void mm_dec_nr_puds(struct mm_struct *mm)
2001 if (mm_pud_folded(mm))
2003 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2007 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
2008 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
2009 unsigned long address)
2014 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
2015 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
2018 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
2020 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
2022 if (mm_pmd_folded(mm))
2024 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2027 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
2029 if (mm_pmd_folded(mm))
2031 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2036 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
2038 atomic_long_set(&mm->pgtables_bytes, 0);
2041 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2043 return atomic_long_read(&mm->pgtables_bytes);
2046 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
2048 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2051 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
2053 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2057 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
2058 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2063 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
2064 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
2067 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
2068 int __pte_alloc_kernel(pmd_t *pmd);
2070 #if defined(CONFIG_MMU)
2073 * The following ifdef needed to get the 5level-fixup.h header to work.
2074 * Remove it when 5level-fixup.h has been removed.
2076 #ifndef __ARCH_HAS_5LEVEL_HACK
2077 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2078 unsigned long address)
2080 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
2081 NULL : p4d_offset(pgd, address);
2084 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2085 unsigned long address)
2087 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
2088 NULL : pud_offset(p4d, address);
2091 static inline p4d_t *p4d_alloc_track(struct mm_struct *mm, pgd_t *pgd,
2092 unsigned long address,
2093 pgtbl_mod_mask *mod_mask)
2096 if (unlikely(pgd_none(*pgd))) {
2097 if (__p4d_alloc(mm, pgd, address))
2099 *mod_mask |= PGTBL_PGD_MODIFIED;
2102 return p4d_offset(pgd, address);
2105 #endif /* !__ARCH_HAS_5LEVEL_HACK */
2107 static inline pud_t *pud_alloc_track(struct mm_struct *mm, p4d_t *p4d,
2108 unsigned long address,
2109 pgtbl_mod_mask *mod_mask)
2111 if (unlikely(p4d_none(*p4d))) {
2112 if (__pud_alloc(mm, p4d, address))
2114 *mod_mask |= PGTBL_P4D_MODIFIED;
2117 return pud_offset(p4d, address);
2120 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
2122 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
2123 NULL: pmd_offset(pud, address);
2126 static inline pmd_t *pmd_alloc_track(struct mm_struct *mm, pud_t *pud,
2127 unsigned long address,
2128 pgtbl_mod_mask *mod_mask)
2130 if (unlikely(pud_none(*pud))) {
2131 if (__pmd_alloc(mm, pud, address))
2133 *mod_mask |= PGTBL_PUD_MODIFIED;
2136 return pmd_offset(pud, address);
2138 #endif /* CONFIG_MMU */
2140 #if USE_SPLIT_PTE_PTLOCKS
2141 #if ALLOC_SPLIT_PTLOCKS
2142 void __init ptlock_cache_init(void);
2143 extern bool ptlock_alloc(struct page *page);
2144 extern void ptlock_free(struct page *page);
2146 static inline spinlock_t *ptlock_ptr(struct page *page)
2150 #else /* ALLOC_SPLIT_PTLOCKS */
2151 static inline void ptlock_cache_init(void)
2155 static inline bool ptlock_alloc(struct page *page)
2160 static inline void ptlock_free(struct page *page)
2164 static inline spinlock_t *ptlock_ptr(struct page *page)
2168 #endif /* ALLOC_SPLIT_PTLOCKS */
2170 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2172 return ptlock_ptr(pmd_page(*pmd));
2175 static inline bool ptlock_init(struct page *page)
2178 * prep_new_page() initialize page->private (and therefore page->ptl)
2179 * with 0. Make sure nobody took it in use in between.
2181 * It can happen if arch try to use slab for page table allocation:
2182 * slab code uses page->slab_cache, which share storage with page->ptl.
2184 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
2185 if (!ptlock_alloc(page))
2187 spin_lock_init(ptlock_ptr(page));
2191 #else /* !USE_SPLIT_PTE_PTLOCKS */
2193 * We use mm->page_table_lock to guard all pagetable pages of the mm.
2195 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2197 return &mm->page_table_lock;
2199 static inline void ptlock_cache_init(void) {}
2200 static inline bool ptlock_init(struct page *page) { return true; }
2201 static inline void ptlock_free(struct page *page) {}
2202 #endif /* USE_SPLIT_PTE_PTLOCKS */
2204 static inline void pgtable_init(void)
2206 ptlock_cache_init();
2207 pgtable_cache_init();
2210 static inline bool pgtable_pte_page_ctor(struct page *page)
2212 if (!ptlock_init(page))
2214 __SetPageTable(page);
2215 inc_zone_page_state(page, NR_PAGETABLE);
2219 static inline void pgtable_pte_page_dtor(struct page *page)
2222 __ClearPageTable(page);
2223 dec_zone_page_state(page, NR_PAGETABLE);
2226 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
2228 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
2229 pte_t *__pte = pte_offset_map(pmd, address); \
2235 #define pte_unmap_unlock(pte, ptl) do { \
2240 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2242 #define pte_alloc_map(mm, pmd, address) \
2243 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2245 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2246 (pte_alloc(mm, pmd) ? \
2247 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2249 #define pte_alloc_kernel(pmd, address) \
2250 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2251 NULL: pte_offset_kernel(pmd, address))
2253 #define pte_alloc_kernel_track(pmd, address, mask) \
2254 ((unlikely(pmd_none(*(pmd))) && \
2255 (__pte_alloc_kernel(pmd) || ({*(mask)|=PGTBL_PMD_MODIFIED;0;})))?\
2256 NULL: pte_offset_kernel(pmd, address))
2258 #if USE_SPLIT_PMD_PTLOCKS
2260 static struct page *pmd_to_page(pmd_t *pmd)
2262 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2263 return virt_to_page((void *)((unsigned long) pmd & mask));
2266 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2268 return ptlock_ptr(pmd_to_page(pmd));
2271 static inline bool pgtable_pmd_page_ctor(struct page *page)
2273 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2274 page->pmd_huge_pte = NULL;
2276 return ptlock_init(page);
2279 static inline void pgtable_pmd_page_dtor(struct page *page)
2281 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2282 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2287 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2291 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2293 return &mm->page_table_lock;
2296 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
2297 static inline void pgtable_pmd_page_dtor(struct page *page) {}
2299 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2303 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2305 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2311 * No scalability reason to split PUD locks yet, but follow the same pattern
2312 * as the PMD locks to make it easier if we decide to. The VM should not be
2313 * considered ready to switch to split PUD locks yet; there may be places
2314 * which need to be converted from page_table_lock.
2316 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2318 return &mm->page_table_lock;
2321 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2323 spinlock_t *ptl = pud_lockptr(mm, pud);
2329 extern void __init pagecache_init(void);
2330 extern void __init free_area_init_memoryless_node(int nid);
2331 extern void free_initmem(void);
2334 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2335 * into the buddy system. The freed pages will be poisoned with pattern
2336 * "poison" if it's within range [0, UCHAR_MAX].
2337 * Return pages freed into the buddy system.
2339 extern unsigned long free_reserved_area(void *start, void *end,
2340 int poison, const char *s);
2342 #ifdef CONFIG_HIGHMEM
2344 * Free a highmem page into the buddy system, adjusting totalhigh_pages
2345 * and totalram_pages.
2347 extern void free_highmem_page(struct page *page);
2350 extern void adjust_managed_page_count(struct page *page, long count);
2351 extern void mem_init_print_info(const char *str);
2353 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2355 /* Free the reserved page into the buddy system, so it gets managed. */
2356 static inline void __free_reserved_page(struct page *page)
2358 ClearPageReserved(page);
2359 init_page_count(page);
2363 static inline void free_reserved_page(struct page *page)
2365 __free_reserved_page(page);
2366 adjust_managed_page_count(page, 1);
2369 static inline void mark_page_reserved(struct page *page)
2371 SetPageReserved(page);
2372 adjust_managed_page_count(page, -1);
2376 * Default method to free all the __init memory into the buddy system.
2377 * The freed pages will be poisoned with pattern "poison" if it's within
2378 * range [0, UCHAR_MAX].
2379 * Return pages freed into the buddy system.
2381 static inline unsigned long free_initmem_default(int poison)
2383 extern char __init_begin[], __init_end[];
2385 return free_reserved_area(&__init_begin, &__init_end,
2386 poison, "unused kernel");
2389 static inline unsigned long get_num_physpages(void)
2392 unsigned long phys_pages = 0;
2394 for_each_online_node(nid)
2395 phys_pages += node_present_pages(nid);
2401 * Using memblock node mappings, an architecture may initialise its
2402 * zones, allocate the backing mem_map and account for memory holes in an
2403 * architecture independent manner.
2405 * An architecture is expected to register range of page frames backed by
2406 * physical memory with memblock_add[_node]() before calling
2407 * free_area_init() passing in the PFN each zone ends at. At a basic
2408 * usage, an architecture is expected to do something like
2410 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2412 * for_each_valid_physical_page_range()
2413 * memblock_add_node(base, size, nid)
2414 * free_area_init(max_zone_pfns);
2416 * sparse_memory_present_with_active_regions() calls memory_present() for
2417 * each range when SPARSEMEM is enabled.
2419 void free_area_init(unsigned long *max_zone_pfn);
2420 unsigned long node_map_pfn_alignment(void);
2421 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2422 unsigned long end_pfn);
2423 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2424 unsigned long end_pfn);
2425 extern void get_pfn_range_for_nid(unsigned int nid,
2426 unsigned long *start_pfn, unsigned long *end_pfn);
2427 extern unsigned long find_min_pfn_with_active_regions(void);
2428 extern void sparse_memory_present_with_active_regions(int nid);
2430 #ifndef CONFIG_NEED_MULTIPLE_NODES
2431 static inline int early_pfn_to_nid(unsigned long pfn)
2436 /* please see mm/page_alloc.c */
2437 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2438 /* there is a per-arch backend function. */
2439 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
2440 struct mminit_pfnnid_cache *state);
2443 extern void set_dma_reserve(unsigned long new_dma_reserve);
2444 extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long,
2445 enum memmap_context, struct vmem_altmap *);
2446 extern void setup_per_zone_wmarks(void);
2447 extern int __meminit init_per_zone_wmark_min(void);
2448 extern void mem_init(void);
2449 extern void __init mmap_init(void);
2450 extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2451 extern long si_mem_available(void);
2452 extern void si_meminfo(struct sysinfo * val);
2453 extern void si_meminfo_node(struct sysinfo *val, int nid);
2454 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2455 extern unsigned long arch_reserved_kernel_pages(void);
2458 extern __printf(3, 4)
2459 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2461 extern void setup_per_cpu_pageset(void);
2464 extern int min_free_kbytes;
2465 extern int watermark_boost_factor;
2466 extern int watermark_scale_factor;
2467 extern bool arch_has_descending_max_zone_pfns(void);
2470 extern atomic_long_t mmap_pages_allocated;
2471 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2473 /* interval_tree.c */
2474 void vma_interval_tree_insert(struct vm_area_struct *node,
2475 struct rb_root_cached *root);
2476 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2477 struct vm_area_struct *prev,
2478 struct rb_root_cached *root);
2479 void vma_interval_tree_remove(struct vm_area_struct *node,
2480 struct rb_root_cached *root);
2481 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2482 unsigned long start, unsigned long last);
2483 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2484 unsigned long start, unsigned long last);
2486 #define vma_interval_tree_foreach(vma, root, start, last) \
2487 for (vma = vma_interval_tree_iter_first(root, start, last); \
2488 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2490 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2491 struct rb_root_cached *root);
2492 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2493 struct rb_root_cached *root);
2494 struct anon_vma_chain *
2495 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2496 unsigned long start, unsigned long last);
2497 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2498 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2499 #ifdef CONFIG_DEBUG_VM_RB
2500 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2503 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2504 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2505 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2508 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2509 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2510 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2511 struct vm_area_struct *expand);
2512 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2513 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2515 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2517 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2518 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2519 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2520 struct mempolicy *, struct vm_userfaultfd_ctx);
2521 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2522 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2523 unsigned long addr, int new_below);
2524 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2525 unsigned long addr, int new_below);
2526 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2527 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2528 struct rb_node **, struct rb_node *);
2529 extern void unlink_file_vma(struct vm_area_struct *);
2530 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2531 unsigned long addr, unsigned long len, pgoff_t pgoff,
2532 bool *need_rmap_locks);
2533 extern void exit_mmap(struct mm_struct *);
2535 static inline int check_data_rlimit(unsigned long rlim,
2537 unsigned long start,
2538 unsigned long end_data,
2539 unsigned long start_data)
2541 if (rlim < RLIM_INFINITY) {
2542 if (((new - start) + (end_data - start_data)) > rlim)
2549 extern int mm_take_all_locks(struct mm_struct *mm);
2550 extern void mm_drop_all_locks(struct mm_struct *mm);
2552 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2553 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2554 extern struct file *get_task_exe_file(struct task_struct *task);
2556 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2557 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2559 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2560 const struct vm_special_mapping *sm);
2561 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2562 unsigned long addr, unsigned long len,
2563 unsigned long flags,
2564 const struct vm_special_mapping *spec);
2565 /* This is an obsolete alternative to _install_special_mapping. */
2566 extern int install_special_mapping(struct mm_struct *mm,
2567 unsigned long addr, unsigned long len,
2568 unsigned long flags, struct page **pages);
2570 unsigned long randomize_stack_top(unsigned long stack_top);
2572 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2574 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2575 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2576 struct list_head *uf);
2577 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2578 unsigned long len, unsigned long prot, unsigned long flags,
2579 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate,
2580 struct list_head *uf);
2581 extern int __do_munmap(struct mm_struct *, unsigned long, size_t,
2582 struct list_head *uf, bool downgrade);
2583 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2584 struct list_head *uf);
2585 extern int do_madvise(unsigned long start, size_t len_in, int behavior);
2587 static inline unsigned long
2588 do_mmap_pgoff(struct file *file, unsigned long addr,
2589 unsigned long len, unsigned long prot, unsigned long flags,
2590 unsigned long pgoff, unsigned long *populate,
2591 struct list_head *uf)
2593 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate, uf);
2597 extern int __mm_populate(unsigned long addr, unsigned long len,
2599 static inline void mm_populate(unsigned long addr, unsigned long len)
2602 (void) __mm_populate(addr, len, 1);
2605 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2608 /* These take the mm semaphore themselves */
2609 extern int __must_check vm_brk(unsigned long, unsigned long);
2610 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2611 extern int vm_munmap(unsigned long, size_t);
2612 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2613 unsigned long, unsigned long,
2614 unsigned long, unsigned long);
2616 struct vm_unmapped_area_info {
2617 #define VM_UNMAPPED_AREA_TOPDOWN 1
2618 unsigned long flags;
2619 unsigned long length;
2620 unsigned long low_limit;
2621 unsigned long high_limit;
2622 unsigned long align_mask;
2623 unsigned long align_offset;
2626 extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info);
2629 extern void truncate_inode_pages(struct address_space *, loff_t);
2630 extern void truncate_inode_pages_range(struct address_space *,
2631 loff_t lstart, loff_t lend);
2632 extern void truncate_inode_pages_final(struct address_space *);
2634 /* generic vm_area_ops exported for stackable file systems */
2635 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2636 extern void filemap_map_pages(struct vm_fault *vmf,
2637 pgoff_t start_pgoff, pgoff_t end_pgoff);
2638 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2640 /* mm/page-writeback.c */
2641 int __must_check write_one_page(struct page *page);
2642 void task_dirty_inc(struct task_struct *tsk);
2644 extern unsigned long stack_guard_gap;
2645 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2646 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2648 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2649 extern int expand_downwards(struct vm_area_struct *vma,
2650 unsigned long address);
2652 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2654 #define expand_upwards(vma, address) (0)
2657 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2658 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2659 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2660 struct vm_area_struct **pprev);
2662 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2663 NULL if none. Assume start_addr < end_addr. */
2664 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2666 struct vm_area_struct * vma = find_vma(mm,start_addr);
2668 if (vma && end_addr <= vma->vm_start)
2673 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2675 unsigned long vm_start = vma->vm_start;
2677 if (vma->vm_flags & VM_GROWSDOWN) {
2678 vm_start -= stack_guard_gap;
2679 if (vm_start > vma->vm_start)
2685 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2687 unsigned long vm_end = vma->vm_end;
2689 if (vma->vm_flags & VM_GROWSUP) {
2690 vm_end += stack_guard_gap;
2691 if (vm_end < vma->vm_end)
2692 vm_end = -PAGE_SIZE;
2697 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2699 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2702 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2703 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2704 unsigned long vm_start, unsigned long vm_end)
2706 struct vm_area_struct *vma = find_vma(mm, vm_start);
2708 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2714 static inline bool range_in_vma(struct vm_area_struct *vma,
2715 unsigned long start, unsigned long end)
2717 return (vma && vma->vm_start <= start && end <= vma->vm_end);
2721 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2722 void vma_set_page_prot(struct vm_area_struct *vma);
2724 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2728 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2730 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2734 #ifdef CONFIG_NUMA_BALANCING
2735 unsigned long change_prot_numa(struct vm_area_struct *vma,
2736 unsigned long start, unsigned long end);
2739 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2740 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2741 unsigned long pfn, unsigned long size, pgprot_t);
2742 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2743 int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
2744 struct page **pages, unsigned long *num);
2745 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
2747 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
2749 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2751 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2752 unsigned long pfn, pgprot_t pgprot);
2753 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2755 vm_fault_t vmf_insert_mixed_prot(struct vm_area_struct *vma, unsigned long addr,
2756 pfn_t pfn, pgprot_t pgprot);
2757 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
2758 unsigned long addr, pfn_t pfn);
2759 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2761 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2762 unsigned long addr, struct page *page)
2764 int err = vm_insert_page(vma, addr, page);
2767 return VM_FAULT_OOM;
2768 if (err < 0 && err != -EBUSY)
2769 return VM_FAULT_SIGBUS;
2771 return VM_FAULT_NOPAGE;
2774 static inline vm_fault_t vmf_error(int err)
2777 return VM_FAULT_OOM;
2778 return VM_FAULT_SIGBUS;
2781 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
2782 unsigned int foll_flags);
2784 #define FOLL_WRITE 0x01 /* check pte is writable */
2785 #define FOLL_TOUCH 0x02 /* mark page accessed */
2786 #define FOLL_GET 0x04 /* do get_page on page */
2787 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2788 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2789 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2790 * and return without waiting upon it */
2791 #define FOLL_POPULATE 0x40 /* fault in page */
2792 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2793 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2794 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2795 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2796 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2797 #define FOLL_MLOCK 0x1000 /* lock present pages */
2798 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2799 #define FOLL_COW 0x4000 /* internal GUP flag */
2800 #define FOLL_ANON 0x8000 /* don't do file mappings */
2801 #define FOLL_LONGTERM 0x10000 /* mapping lifetime is indefinite: see below */
2802 #define FOLL_SPLIT_PMD 0x20000 /* split huge pmd before returning */
2803 #define FOLL_PIN 0x40000 /* pages must be released via unpin_user_page */
2804 #define FOLL_FAST_ONLY 0x80000 /* gup_fast: prevent fall-back to slow gup */
2807 * FOLL_PIN and FOLL_LONGTERM may be used in various combinations with each
2808 * other. Here is what they mean, and how to use them:
2810 * FOLL_LONGTERM indicates that the page will be held for an indefinite time
2811 * period _often_ under userspace control. This is in contrast to
2812 * iov_iter_get_pages(), whose usages are transient.
2814 * FIXME: For pages which are part of a filesystem, mappings are subject to the
2815 * lifetime enforced by the filesystem and we need guarantees that longterm
2816 * users like RDMA and V4L2 only establish mappings which coordinate usage with
2817 * the filesystem. Ideas for this coordination include revoking the longterm
2818 * pin, delaying writeback, bounce buffer page writeback, etc. As FS DAX was
2819 * added after the problem with filesystems was found FS DAX VMAs are
2820 * specifically failed. Filesystem pages are still subject to bugs and use of
2821 * FOLL_LONGTERM should be avoided on those pages.
2823 * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call.
2824 * Currently only get_user_pages() and get_user_pages_fast() support this flag
2825 * and calls to get_user_pages_[un]locked are specifically not allowed. This
2826 * is due to an incompatibility with the FS DAX check and
2827 * FAULT_FLAG_ALLOW_RETRY.
2829 * In the CMA case: long term pins in a CMA region would unnecessarily fragment
2830 * that region. And so, CMA attempts to migrate the page before pinning, when
2831 * FOLL_LONGTERM is specified.
2833 * FOLL_PIN indicates that a special kind of tracking (not just page->_refcount,
2834 * but an additional pin counting system) will be invoked. This is intended for
2835 * anything that gets a page reference and then touches page data (for example,
2836 * Direct IO). This lets the filesystem know that some non-file-system entity is
2837 * potentially changing the pages' data. In contrast to FOLL_GET (whose pages
2838 * are released via put_page()), FOLL_PIN pages must be released, ultimately, by
2839 * a call to unpin_user_page().
2841 * FOLL_PIN is similar to FOLL_GET: both of these pin pages. They use different
2842 * and separate refcounting mechanisms, however, and that means that each has
2843 * its own acquire and release mechanisms:
2845 * FOLL_GET: get_user_pages*() to acquire, and put_page() to release.
2847 * FOLL_PIN: pin_user_pages*() to acquire, and unpin_user_pages to release.
2849 * FOLL_PIN and FOLL_GET are mutually exclusive for a given function call.
2850 * (The underlying pages may experience both FOLL_GET-based and FOLL_PIN-based
2851 * calls applied to them, and that's perfectly OK. This is a constraint on the
2852 * callers, not on the pages.)
2854 * FOLL_PIN should be set internally by the pin_user_pages*() APIs, never
2855 * directly by the caller. That's in order to help avoid mismatches when
2856 * releasing pages: get_user_pages*() pages must be released via put_page(),
2857 * while pin_user_pages*() pages must be released via unpin_user_page().
2859 * Please see Documentation/core-api/pin_user_pages.rst for more information.
2862 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
2864 if (vm_fault & VM_FAULT_OOM)
2866 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2867 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2868 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2873 typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
2874 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2875 unsigned long size, pte_fn_t fn, void *data);
2876 extern int apply_to_existing_page_range(struct mm_struct *mm,
2877 unsigned long address, unsigned long size,
2878 pte_fn_t fn, void *data);
2880 #ifdef CONFIG_PAGE_POISONING
2881 extern bool page_poisoning_enabled(void);
2882 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2884 static inline bool page_poisoning_enabled(void) { return false; }
2885 static inline void kernel_poison_pages(struct page *page, int numpages,
2889 #ifdef CONFIG_INIT_ON_ALLOC_DEFAULT_ON
2890 DECLARE_STATIC_KEY_TRUE(init_on_alloc);
2892 DECLARE_STATIC_KEY_FALSE(init_on_alloc);
2894 static inline bool want_init_on_alloc(gfp_t flags)
2896 if (static_branch_unlikely(&init_on_alloc) &&
2897 !page_poisoning_enabled())
2899 return flags & __GFP_ZERO;
2902 #ifdef CONFIG_INIT_ON_FREE_DEFAULT_ON
2903 DECLARE_STATIC_KEY_TRUE(init_on_free);
2905 DECLARE_STATIC_KEY_FALSE(init_on_free);
2907 static inline bool want_init_on_free(void)
2909 return static_branch_unlikely(&init_on_free) &&
2910 !page_poisoning_enabled();
2913 #ifdef CONFIG_DEBUG_PAGEALLOC
2914 extern void init_debug_pagealloc(void);
2916 static inline void init_debug_pagealloc(void) {}
2918 extern bool _debug_pagealloc_enabled_early;
2919 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
2921 static inline bool debug_pagealloc_enabled(void)
2923 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
2924 _debug_pagealloc_enabled_early;
2928 * For use in fast paths after init_debug_pagealloc() has run, or when a
2929 * false negative result is not harmful when called too early.
2931 static inline bool debug_pagealloc_enabled_static(void)
2933 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
2936 return static_branch_unlikely(&_debug_pagealloc_enabled);
2939 #if defined(CONFIG_DEBUG_PAGEALLOC) || defined(CONFIG_ARCH_HAS_SET_DIRECT_MAP)
2940 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2943 * When called in DEBUG_PAGEALLOC context, the call should most likely be
2944 * guarded by debug_pagealloc_enabled() or debug_pagealloc_enabled_static()
2947 kernel_map_pages(struct page *page, int numpages, int enable)
2949 __kernel_map_pages(page, numpages, enable);
2951 #ifdef CONFIG_HIBERNATION
2952 extern bool kernel_page_present(struct page *page);
2953 #endif /* CONFIG_HIBERNATION */
2954 #else /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */
2956 kernel_map_pages(struct page *page, int numpages, int enable) {}
2957 #ifdef CONFIG_HIBERNATION
2958 static inline bool kernel_page_present(struct page *page) { return true; }
2959 #endif /* CONFIG_HIBERNATION */
2960 #endif /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */
2962 #ifdef __HAVE_ARCH_GATE_AREA
2963 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2964 extern int in_gate_area_no_mm(unsigned long addr);
2965 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2967 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2971 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2972 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2976 #endif /* __HAVE_ARCH_GATE_AREA */
2978 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2980 #ifdef CONFIG_SYSCTL
2981 extern int sysctl_drop_caches;
2982 int drop_caches_sysctl_handler(struct ctl_table *, int, void *, size_t *,
2986 void drop_slab(void);
2987 void drop_slab_node(int nid);
2990 #define randomize_va_space 0
2992 extern int randomize_va_space;
2995 const char * arch_vma_name(struct vm_area_struct *vma);
2997 void print_vma_addr(char *prefix, unsigned long rip);
2999 static inline void print_vma_addr(char *prefix, unsigned long rip)
3004 void *sparse_buffer_alloc(unsigned long size);
3005 struct page * __populate_section_memmap(unsigned long pfn,
3006 unsigned long nr_pages, int nid, struct vmem_altmap *altmap);
3007 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
3008 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
3009 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
3010 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
3011 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
3012 void *vmemmap_alloc_block(unsigned long size, int node);
3014 void *vmemmap_alloc_block_buf(unsigned long size, int node);
3015 void *altmap_alloc_block_buf(unsigned long size, struct vmem_altmap *altmap);
3016 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
3017 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
3019 int vmemmap_populate(unsigned long start, unsigned long end, int node,
3020 struct vmem_altmap *altmap);
3021 void vmemmap_populate_print_last(void);
3022 #ifdef CONFIG_MEMORY_HOTPLUG
3023 void vmemmap_free(unsigned long start, unsigned long end,
3024 struct vmem_altmap *altmap);
3026 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
3027 unsigned long nr_pages);
3030 MF_COUNT_INCREASED = 1 << 0,
3031 MF_ACTION_REQUIRED = 1 << 1,
3032 MF_MUST_KILL = 1 << 2,
3033 MF_SOFT_OFFLINE = 1 << 3,
3035 extern int memory_failure(unsigned long pfn, int flags);
3036 extern void memory_failure_queue(unsigned long pfn, int flags);
3037 extern void memory_failure_queue_kick(int cpu);
3038 extern int unpoison_memory(unsigned long pfn);
3039 extern int get_hwpoison_page(struct page *page);
3040 #define put_hwpoison_page(page) put_page(page)
3041 extern int sysctl_memory_failure_early_kill;
3042 extern int sysctl_memory_failure_recovery;
3043 extern void shake_page(struct page *p, int access);
3044 extern atomic_long_t num_poisoned_pages __read_mostly;
3045 extern int soft_offline_page(unsigned long pfn, int flags);
3049 * Error handlers for various types of pages.
3052 MF_IGNORED, /* Error: cannot be handled */
3053 MF_FAILED, /* Error: handling failed */
3054 MF_DELAYED, /* Will be handled later */
3055 MF_RECOVERED, /* Successfully recovered */
3058 enum mf_action_page_type {
3060 MF_MSG_KERNEL_HIGH_ORDER,
3062 MF_MSG_DIFFERENT_COMPOUND,
3063 MF_MSG_POISONED_HUGE,
3066 MF_MSG_NON_PMD_HUGE,
3067 MF_MSG_UNMAP_FAILED,
3068 MF_MSG_DIRTY_SWAPCACHE,
3069 MF_MSG_CLEAN_SWAPCACHE,
3070 MF_MSG_DIRTY_MLOCKED_LRU,
3071 MF_MSG_CLEAN_MLOCKED_LRU,
3072 MF_MSG_DIRTY_UNEVICTABLE_LRU,
3073 MF_MSG_CLEAN_UNEVICTABLE_LRU,
3076 MF_MSG_TRUNCATED_LRU,
3083 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
3084 extern void clear_huge_page(struct page *page,
3085 unsigned long addr_hint,
3086 unsigned int pages_per_huge_page);
3087 extern void copy_user_huge_page(struct page *dst, struct page *src,
3088 unsigned long addr_hint,
3089 struct vm_area_struct *vma,
3090 unsigned int pages_per_huge_page);
3091 extern long copy_huge_page_from_user(struct page *dst_page,
3092 const void __user *usr_src,
3093 unsigned int pages_per_huge_page,
3094 bool allow_pagefault);
3097 * vma_is_special_huge - Are transhuge page-table entries considered special?
3098 * @vma: Pointer to the struct vm_area_struct to consider
3100 * Whether transhuge page-table entries are considered "special" following
3101 * the definition in vm_normal_page().
3103 * Return: true if transhuge page-table entries should be considered special,
3106 static inline bool vma_is_special_huge(const struct vm_area_struct *vma)
3108 return vma_is_dax(vma) || (vma->vm_file &&
3109 (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP)));
3112 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
3114 #ifdef CONFIG_DEBUG_PAGEALLOC
3115 extern unsigned int _debug_guardpage_minorder;
3116 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
3118 static inline unsigned int debug_guardpage_minorder(void)
3120 return _debug_guardpage_minorder;
3123 static inline bool debug_guardpage_enabled(void)
3125 return static_branch_unlikely(&_debug_guardpage_enabled);
3128 static inline bool page_is_guard(struct page *page)
3130 if (!debug_guardpage_enabled())
3133 return PageGuard(page);
3136 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
3137 static inline bool debug_guardpage_enabled(void) { return false; }
3138 static inline bool page_is_guard(struct page *page) { return false; }
3139 #endif /* CONFIG_DEBUG_PAGEALLOC */
3141 #if MAX_NUMNODES > 1
3142 void __init setup_nr_node_ids(void);
3144 static inline void setup_nr_node_ids(void) {}
3147 extern int memcmp_pages(struct page *page1, struct page *page2);
3149 static inline int pages_identical(struct page *page1, struct page *page2)
3151 return !memcmp_pages(page1, page2);
3154 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
3155 unsigned long clean_record_shared_mapping_range(struct address_space *mapping,
3156 pgoff_t first_index, pgoff_t nr,
3157 pgoff_t bitmap_pgoff,
3158 unsigned long *bitmap,
3162 unsigned long wp_shared_mapping_range(struct address_space *mapping,
3163 pgoff_t first_index, pgoff_t nr);
3166 extern int sysctl_nr_trim_pages;
3168 #endif /* __KERNEL__ */
3169 #endif /* _LINUX_MM_H */