1 /* SPDX-License-Identifier: GPL-2.0 */
5 #include <linux/errno.h>
9 #include <linux/mmdebug.h>
10 #include <linux/gfp.h>
11 #include <linux/bug.h>
12 #include <linux/list.h>
13 #include <linux/mmzone.h>
14 #include <linux/rbtree.h>
15 #include <linux/atomic.h>
16 #include <linux/debug_locks.h>
17 #include <linux/mm_types.h>
18 #include <linux/mmap_lock.h>
19 #include <linux/range.h>
20 #include <linux/pfn.h>
21 #include <linux/percpu-refcount.h>
22 #include <linux/bit_spinlock.h>
23 #include <linux/shrinker.h>
24 #include <linux/resource.h>
25 #include <linux/page_ext.h>
26 #include <linux/err.h>
27 #include <linux/page-flags.h>
28 #include <linux/page_ref.h>
29 #include <linux/memremap.h>
30 #include <linux/overflow.h>
31 #include <linux/sizes.h>
32 #include <linux/sched.h>
33 #include <linux/pgtable.h>
34 #include <linux/kasan.h>
38 struct anon_vma_chain;
41 struct writeback_control;
45 extern int sysctl_page_lock_unfairness;
47 void init_mm_internals(void);
49 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
50 extern unsigned long max_mapnr;
52 static inline void set_max_mapnr(unsigned long limit)
57 static inline void set_max_mapnr(unsigned long limit) { }
60 extern atomic_long_t _totalram_pages;
61 static inline unsigned long totalram_pages(void)
63 return (unsigned long)atomic_long_read(&_totalram_pages);
66 static inline void totalram_pages_inc(void)
68 atomic_long_inc(&_totalram_pages);
71 static inline void totalram_pages_dec(void)
73 atomic_long_dec(&_totalram_pages);
76 static inline void totalram_pages_add(long count)
78 atomic_long_add(count, &_totalram_pages);
81 extern void * high_memory;
82 extern int page_cluster;
85 extern int sysctl_legacy_va_layout;
87 #define sysctl_legacy_va_layout 0
90 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
91 extern const int mmap_rnd_bits_min;
92 extern const int mmap_rnd_bits_max;
93 extern int mmap_rnd_bits __read_mostly;
95 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
96 extern const int mmap_rnd_compat_bits_min;
97 extern const int mmap_rnd_compat_bits_max;
98 extern int mmap_rnd_compat_bits __read_mostly;
101 #include <asm/page.h>
102 #include <asm/processor.h>
105 * Architectures that support memory tagging (assigning tags to memory regions,
106 * embedding these tags into addresses that point to these memory regions, and
107 * checking that the memory and the pointer tags match on memory accesses)
108 * redefine this macro to strip tags from pointers.
109 * It's defined as noop for arcitectures that don't support memory tagging.
111 #ifndef untagged_addr
112 #define untagged_addr(addr) (addr)
116 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
120 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
124 #define lm_alias(x) __va(__pa_symbol(x))
128 * To prevent common memory management code establishing
129 * a zero page mapping on a read fault.
130 * This macro should be defined within <asm/pgtable.h>.
131 * s390 does this to prevent multiplexing of hardware bits
132 * related to the physical page in case of virtualization.
134 #ifndef mm_forbids_zeropage
135 #define mm_forbids_zeropage(X) (0)
139 * On some architectures it is expensive to call memset() for small sizes.
140 * If an architecture decides to implement their own version of
141 * mm_zero_struct_page they should wrap the defines below in a #ifndef and
142 * define their own version of this macro in <asm/pgtable.h>
144 #if BITS_PER_LONG == 64
145 /* This function must be updated when the size of struct page grows above 80
146 * or reduces below 56. The idea that compiler optimizes out switch()
147 * statement, and only leaves move/store instructions. Also the compiler can
148 * combine write statments if they are both assignments and can be reordered,
149 * this can result in several of the writes here being dropped.
151 #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
152 static inline void __mm_zero_struct_page(struct page *page)
154 unsigned long *_pp = (void *)page;
156 /* Check that struct page is either 56, 64, 72, or 80 bytes */
157 BUILD_BUG_ON(sizeof(struct page) & 7);
158 BUILD_BUG_ON(sizeof(struct page) < 56);
159 BUILD_BUG_ON(sizeof(struct page) > 80);
161 switch (sizeof(struct page)) {
182 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
186 * Default maximum number of active map areas, this limits the number of vmas
187 * per mm struct. Users can overwrite this number by sysctl but there is a
190 * When a program's coredump is generated as ELF format, a section is created
191 * per a vma. In ELF, the number of sections is represented in unsigned short.
192 * This means the number of sections should be smaller than 65535 at coredump.
193 * Because the kernel adds some informative sections to a image of program at
194 * generating coredump, we need some margin. The number of extra sections is
195 * 1-3 now and depends on arch. We use "5" as safe margin, here.
197 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
198 * not a hard limit any more. Although some userspace tools can be surprised by
201 #define MAPCOUNT_ELF_CORE_MARGIN (5)
202 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
204 extern int sysctl_max_map_count;
206 extern unsigned long sysctl_user_reserve_kbytes;
207 extern unsigned long sysctl_admin_reserve_kbytes;
209 extern int sysctl_overcommit_memory;
210 extern int sysctl_overcommit_ratio;
211 extern unsigned long sysctl_overcommit_kbytes;
213 int overcommit_ratio_handler(struct ctl_table *, int, void *, size_t *,
215 int overcommit_kbytes_handler(struct ctl_table *, int, void *, size_t *,
217 int overcommit_policy_handler(struct ctl_table *, int, void *, size_t *,
220 * Any attempt to mark this function as static leads to build failure
221 * when CONFIG_DEBUG_INFO_BTF is enabled because __add_to_page_cache_locked()
222 * is referred to by BPF code. This must be visible for error injection.
224 int __add_to_page_cache_locked(struct page *page, struct address_space *mapping,
225 pgoff_t index, gfp_t gfp, void **shadowp);
227 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
229 /* to align the pointer to the (next) page boundary */
230 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
232 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
233 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
235 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
238 * Linux kernel virtual memory manager primitives.
239 * The idea being to have a "virtual" mm in the same way
240 * we have a virtual fs - giving a cleaner interface to the
241 * mm details, and allowing different kinds of memory mappings
242 * (from shared memory to executable loading to arbitrary
246 struct vm_area_struct *vm_area_alloc(struct mm_struct *);
247 struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
248 void vm_area_free(struct vm_area_struct *);
251 extern struct rb_root nommu_region_tree;
252 extern struct rw_semaphore nommu_region_sem;
254 extern unsigned int kobjsize(const void *objp);
258 * vm_flags in vm_area_struct, see mm_types.h.
259 * When changing, update also include/trace/events/mmflags.h
261 #define VM_NONE 0x00000000
263 #define VM_READ 0x00000001 /* currently active flags */
264 #define VM_WRITE 0x00000002
265 #define VM_EXEC 0x00000004
266 #define VM_SHARED 0x00000008
268 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
269 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
270 #define VM_MAYWRITE 0x00000020
271 #define VM_MAYEXEC 0x00000040
272 #define VM_MAYSHARE 0x00000080
274 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
275 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
276 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
277 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
278 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
280 #define VM_LOCKED 0x00002000
281 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
283 /* Used by sys_madvise() */
284 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
285 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
287 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
288 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
289 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
290 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
291 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
292 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
293 #define VM_SYNC 0x00800000 /* Synchronous page faults */
294 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
295 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
296 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
298 #ifdef CONFIG_MEM_SOFT_DIRTY
299 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
301 # define VM_SOFTDIRTY 0
304 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
305 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
306 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
307 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
309 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
310 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
311 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
312 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
313 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
314 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
315 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
316 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
317 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
318 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
319 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
320 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
322 #ifdef CONFIG_ARCH_HAS_PKEYS
323 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
324 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
325 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
326 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
327 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
329 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
331 # define VM_PKEY_BIT4 0
333 #endif /* CONFIG_ARCH_HAS_PKEYS */
335 #if defined(CONFIG_X86)
336 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
337 #elif defined(CONFIG_PPC)
338 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
339 #elif defined(CONFIG_PARISC)
340 # define VM_GROWSUP VM_ARCH_1
341 #elif defined(CONFIG_IA64)
342 # define VM_GROWSUP VM_ARCH_1
343 #elif defined(CONFIG_SPARC64)
344 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
345 # define VM_ARCH_CLEAR VM_SPARC_ADI
346 #elif defined(CONFIG_ARM64)
347 # define VM_ARM64_BTI VM_ARCH_1 /* BTI guarded page, a.k.a. GP bit */
348 # define VM_ARCH_CLEAR VM_ARM64_BTI
349 #elif !defined(CONFIG_MMU)
350 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
353 #if defined(CONFIG_ARM64_MTE)
354 # define VM_MTE VM_HIGH_ARCH_0 /* Use Tagged memory for access control */
355 # define VM_MTE_ALLOWED VM_HIGH_ARCH_1 /* Tagged memory permitted */
357 # define VM_MTE VM_NONE
358 # define VM_MTE_ALLOWED VM_NONE
362 # define VM_GROWSUP VM_NONE
365 /* Bits set in the VMA until the stack is in its final location */
366 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
368 #define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0)
370 /* Common data flag combinations */
371 #define VM_DATA_FLAGS_TSK_EXEC (VM_READ | VM_WRITE | TASK_EXEC | \
372 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
373 #define VM_DATA_FLAGS_NON_EXEC (VM_READ | VM_WRITE | VM_MAYREAD | \
374 VM_MAYWRITE | VM_MAYEXEC)
375 #define VM_DATA_FLAGS_EXEC (VM_READ | VM_WRITE | VM_EXEC | \
376 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
378 #ifndef VM_DATA_DEFAULT_FLAGS /* arch can override this */
379 #define VM_DATA_DEFAULT_FLAGS VM_DATA_FLAGS_EXEC
382 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
383 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
386 #ifdef CONFIG_STACK_GROWSUP
387 #define VM_STACK VM_GROWSUP
389 #define VM_STACK VM_GROWSDOWN
392 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
394 /* VMA basic access permission flags */
395 #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
399 * Special vmas that are non-mergable, non-mlock()able.
401 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
403 /* This mask prevents VMA from being scanned with khugepaged */
404 #define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB)
406 /* This mask defines which mm->def_flags a process can inherit its parent */
407 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
409 /* This mask is used to clear all the VMA flags used by mlock */
410 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
412 /* Arch-specific flags to clear when updating VM flags on protection change */
413 #ifndef VM_ARCH_CLEAR
414 # define VM_ARCH_CLEAR VM_NONE
416 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
419 * mapping from the currently active vm_flags protection bits (the
420 * low four bits) to a page protection mask..
422 extern pgprot_t protection_map[16];
425 * Fault flag definitions.
427 * @FAULT_FLAG_WRITE: Fault was a write fault.
428 * @FAULT_FLAG_MKWRITE: Fault was mkwrite of existing PTE.
429 * @FAULT_FLAG_ALLOW_RETRY: Allow to retry the fault if blocked.
430 * @FAULT_FLAG_RETRY_NOWAIT: Don't drop mmap_lock and wait when retrying.
431 * @FAULT_FLAG_KILLABLE: The fault task is in SIGKILL killable region.
432 * @FAULT_FLAG_TRIED: The fault has been tried once.
433 * @FAULT_FLAG_USER: The fault originated in userspace.
434 * @FAULT_FLAG_REMOTE: The fault is not for current task/mm.
435 * @FAULT_FLAG_INSTRUCTION: The fault was during an instruction fetch.
436 * @FAULT_FLAG_INTERRUPTIBLE: The fault can be interrupted by non-fatal signals.
438 * About @FAULT_FLAG_ALLOW_RETRY and @FAULT_FLAG_TRIED: we can specify
439 * whether we would allow page faults to retry by specifying these two
440 * fault flags correctly. Currently there can be three legal combinations:
442 * (a) ALLOW_RETRY and !TRIED: this means the page fault allows retry, and
443 * this is the first try
445 * (b) ALLOW_RETRY and TRIED: this means the page fault allows retry, and
446 * we've already tried at least once
448 * (c) !ALLOW_RETRY and !TRIED: this means the page fault does not allow retry
450 * The unlisted combination (!ALLOW_RETRY && TRIED) is illegal and should never
451 * be used. Note that page faults can be allowed to retry for multiple times,
452 * in which case we'll have an initial fault with flags (a) then later on
453 * continuous faults with flags (b). We should always try to detect pending
454 * signals before a retry to make sure the continuous page faults can still be
455 * interrupted if necessary.
457 #define FAULT_FLAG_WRITE 0x01
458 #define FAULT_FLAG_MKWRITE 0x02
459 #define FAULT_FLAG_ALLOW_RETRY 0x04
460 #define FAULT_FLAG_RETRY_NOWAIT 0x08
461 #define FAULT_FLAG_KILLABLE 0x10
462 #define FAULT_FLAG_TRIED 0x20
463 #define FAULT_FLAG_USER 0x40
464 #define FAULT_FLAG_REMOTE 0x80
465 #define FAULT_FLAG_INSTRUCTION 0x100
466 #define FAULT_FLAG_INTERRUPTIBLE 0x200
469 * The default fault flags that should be used by most of the
470 * arch-specific page fault handlers.
472 #define FAULT_FLAG_DEFAULT (FAULT_FLAG_ALLOW_RETRY | \
473 FAULT_FLAG_KILLABLE | \
474 FAULT_FLAG_INTERRUPTIBLE)
477 * fault_flag_allow_retry_first - check ALLOW_RETRY the first time
479 * This is mostly used for places where we want to try to avoid taking
480 * the mmap_lock for too long a time when waiting for another condition
481 * to change, in which case we can try to be polite to release the
482 * mmap_lock in the first round to avoid potential starvation of other
483 * processes that would also want the mmap_lock.
485 * Return: true if the page fault allows retry and this is the first
486 * attempt of the fault handling; false otherwise.
488 static inline bool fault_flag_allow_retry_first(unsigned int flags)
490 return (flags & FAULT_FLAG_ALLOW_RETRY) &&
491 (!(flags & FAULT_FLAG_TRIED));
494 #define FAULT_FLAG_TRACE \
495 { FAULT_FLAG_WRITE, "WRITE" }, \
496 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
497 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
498 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
499 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
500 { FAULT_FLAG_TRIED, "TRIED" }, \
501 { FAULT_FLAG_USER, "USER" }, \
502 { FAULT_FLAG_REMOTE, "REMOTE" }, \
503 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }, \
504 { FAULT_FLAG_INTERRUPTIBLE, "INTERRUPTIBLE" }
507 * vm_fault is filled by the pagefault handler and passed to the vma's
508 * ->fault function. The vma's ->fault is responsible for returning a bitmask
509 * of VM_FAULT_xxx flags that give details about how the fault was handled.
511 * MM layer fills up gfp_mask for page allocations but fault handler might
512 * alter it if its implementation requires a different allocation context.
514 * pgoff should be used in favour of virtual_address, if possible.
518 struct vm_area_struct *vma; /* Target VMA */
519 gfp_t gfp_mask; /* gfp mask to be used for allocations */
520 pgoff_t pgoff; /* Logical page offset based on vma */
521 unsigned long address; /* Faulting virtual address */
523 unsigned int flags; /* FAULT_FLAG_xxx flags
524 * XXX: should really be 'const' */
525 pmd_t *pmd; /* Pointer to pmd entry matching
527 pud_t *pud; /* Pointer to pud entry matching
530 pte_t orig_pte; /* Value of PTE at the time of fault */
532 struct page *cow_page; /* Page handler may use for COW fault */
533 struct page *page; /* ->fault handlers should return a
534 * page here, unless VM_FAULT_NOPAGE
535 * is set (which is also implied by
538 /* These three entries are valid only while holding ptl lock */
539 pte_t *pte; /* Pointer to pte entry matching
540 * the 'address'. NULL if the page
541 * table hasn't been allocated.
543 spinlock_t *ptl; /* Page table lock.
544 * Protects pte page table if 'pte'
545 * is not NULL, otherwise pmd.
547 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
548 * vm_ops->map_pages() sets up a page
549 * table from atomic context.
550 * do_fault_around() pre-allocates
551 * page table to avoid allocation from
556 /* page entry size for vm->huge_fault() */
557 enum page_entry_size {
564 * These are the virtual MM functions - opening of an area, closing and
565 * unmapping it (needed to keep files on disk up-to-date etc), pointer
566 * to the functions called when a no-page or a wp-page exception occurs.
568 struct vm_operations_struct {
569 void (*open)(struct vm_area_struct * area);
570 void (*close)(struct vm_area_struct * area);
571 /* Called any time before splitting to check if it's allowed */
572 int (*may_split)(struct vm_area_struct *area, unsigned long addr);
573 int (*mremap)(struct vm_area_struct *area, unsigned long flags);
575 * Called by mprotect() to make driver-specific permission
576 * checks before mprotect() is finalised. The VMA must not
577 * be modified. Returns 0 if eprotect() can proceed.
579 int (*mprotect)(struct vm_area_struct *vma, unsigned long start,
580 unsigned long end, unsigned long newflags);
581 vm_fault_t (*fault)(struct vm_fault *vmf);
582 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
583 enum page_entry_size pe_size);
584 vm_fault_t (*map_pages)(struct vm_fault *vmf,
585 pgoff_t start_pgoff, pgoff_t end_pgoff);
586 unsigned long (*pagesize)(struct vm_area_struct * area);
588 /* notification that a previously read-only page is about to become
589 * writable, if an error is returned it will cause a SIGBUS */
590 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
592 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
593 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
595 /* called by access_process_vm when get_user_pages() fails, typically
596 * for use by special VMAs. See also generic_access_phys() for a generic
597 * implementation useful for any iomem mapping.
599 int (*access)(struct vm_area_struct *vma, unsigned long addr,
600 void *buf, int len, int write);
602 /* Called by the /proc/PID/maps code to ask the vma whether it
603 * has a special name. Returning non-NULL will also cause this
604 * vma to be dumped unconditionally. */
605 const char *(*name)(struct vm_area_struct *vma);
609 * set_policy() op must add a reference to any non-NULL @new mempolicy
610 * to hold the policy upon return. Caller should pass NULL @new to
611 * remove a policy and fall back to surrounding context--i.e. do not
612 * install a MPOL_DEFAULT policy, nor the task or system default
615 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
618 * get_policy() op must add reference [mpol_get()] to any policy at
619 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
620 * in mm/mempolicy.c will do this automatically.
621 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
622 * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
623 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
624 * must return NULL--i.e., do not "fallback" to task or system default
627 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
631 * Called by vm_normal_page() for special PTEs to find the
632 * page for @addr. This is useful if the default behavior
633 * (using pte_page()) would not find the correct page.
635 struct page *(*find_special_page)(struct vm_area_struct *vma,
639 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
641 static const struct vm_operations_struct dummy_vm_ops = {};
643 memset(vma, 0, sizeof(*vma));
645 vma->vm_ops = &dummy_vm_ops;
646 INIT_LIST_HEAD(&vma->anon_vma_chain);
649 static inline void vma_set_anonymous(struct vm_area_struct *vma)
654 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
659 static inline bool vma_is_temporary_stack(struct vm_area_struct *vma)
661 int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP);
666 if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) ==
667 VM_STACK_INCOMPLETE_SETUP)
673 static inline bool vma_is_foreign(struct vm_area_struct *vma)
678 if (current->mm != vma->vm_mm)
684 static inline bool vma_is_accessible(struct vm_area_struct *vma)
686 return vma->vm_flags & VM_ACCESS_FLAGS;
691 * The vma_is_shmem is not inline because it is used only by slow
692 * paths in userfault.
694 bool vma_is_shmem(struct vm_area_struct *vma);
696 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
699 int vma_is_stack_for_current(struct vm_area_struct *vma);
701 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
702 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
707 #include <linux/huge_mm.h>
710 * Methods to modify the page usage count.
712 * What counts for a page usage:
713 * - cache mapping (page->mapping)
714 * - private data (page->private)
715 * - page mapped in a task's page tables, each mapping
716 * is counted separately
718 * Also, many kernel routines increase the page count before a critical
719 * routine so they can be sure the page doesn't go away from under them.
723 * Drop a ref, return true if the refcount fell to zero (the page has no users)
725 static inline int put_page_testzero(struct page *page)
727 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
728 return page_ref_dec_and_test(page);
732 * Try to grab a ref unless the page has a refcount of zero, return false if
734 * This can be called when MMU is off so it must not access
735 * any of the virtual mappings.
737 static inline int get_page_unless_zero(struct page *page)
739 return page_ref_add_unless(page, 1, 0);
742 extern int page_is_ram(unsigned long pfn);
750 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
753 /* Support for virtually mapped pages */
754 struct page *vmalloc_to_page(const void *addr);
755 unsigned long vmalloc_to_pfn(const void *addr);
758 * Determine if an address is within the vmalloc range
760 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
761 * is no special casing required.
764 #ifndef is_ioremap_addr
765 #define is_ioremap_addr(x) is_vmalloc_addr(x)
769 extern bool is_vmalloc_addr(const void *x);
770 extern int is_vmalloc_or_module_addr(const void *x);
772 static inline bool is_vmalloc_addr(const void *x)
776 static inline int is_vmalloc_or_module_addr(const void *x)
782 extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
783 static inline void *kvmalloc(size_t size, gfp_t flags)
785 return kvmalloc_node(size, flags, NUMA_NO_NODE);
787 static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
789 return kvmalloc_node(size, flags | __GFP_ZERO, node);
791 static inline void *kvzalloc(size_t size, gfp_t flags)
793 return kvmalloc(size, flags | __GFP_ZERO);
796 static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
800 if (unlikely(check_mul_overflow(n, size, &bytes)))
803 return kvmalloc(bytes, flags);
806 static inline void *kvcalloc(size_t n, size_t size, gfp_t flags)
808 return kvmalloc_array(n, size, flags | __GFP_ZERO);
811 extern void kvfree(const void *addr);
812 extern void kvfree_sensitive(const void *addr, size_t len);
814 static inline int head_compound_mapcount(struct page *head)
816 return atomic_read(compound_mapcount_ptr(head)) + 1;
820 * Mapcount of compound page as a whole, does not include mapped sub-pages.
822 * Must be called only for compound pages or any their tail sub-pages.
824 static inline int compound_mapcount(struct page *page)
826 VM_BUG_ON_PAGE(!PageCompound(page), page);
827 page = compound_head(page);
828 return head_compound_mapcount(page);
832 * The atomic page->_mapcount, starts from -1: so that transitions
833 * both from it and to it can be tracked, using atomic_inc_and_test
834 * and atomic_add_negative(-1).
836 static inline void page_mapcount_reset(struct page *page)
838 atomic_set(&(page)->_mapcount, -1);
841 int __page_mapcount(struct page *page);
844 * Mapcount of 0-order page; when compound sub-page, includes
845 * compound_mapcount().
847 * Result is undefined for pages which cannot be mapped into userspace.
848 * For example SLAB or special types of pages. See function page_has_type().
849 * They use this place in struct page differently.
851 static inline int page_mapcount(struct page *page)
853 if (unlikely(PageCompound(page)))
854 return __page_mapcount(page);
855 return atomic_read(&page->_mapcount) + 1;
858 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
859 int total_mapcount(struct page *page);
860 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
862 static inline int total_mapcount(struct page *page)
864 return page_mapcount(page);
866 static inline int page_trans_huge_mapcount(struct page *page,
869 int mapcount = page_mapcount(page);
871 *total_mapcount = mapcount;
876 static inline struct page *virt_to_head_page(const void *x)
878 struct page *page = virt_to_page(x);
880 return compound_head(page);
883 void __put_page(struct page *page);
885 void put_pages_list(struct list_head *pages);
887 void split_page(struct page *page, unsigned int order);
890 * Compound pages have a destructor function. Provide a
891 * prototype for that function and accessor functions.
892 * These are _only_ valid on the head of a compound page.
894 typedef void compound_page_dtor(struct page *);
896 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
897 enum compound_dtor_id {
900 #ifdef CONFIG_HUGETLB_PAGE
903 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
908 extern compound_page_dtor * const compound_page_dtors[NR_COMPOUND_DTORS];
910 static inline void set_compound_page_dtor(struct page *page,
911 enum compound_dtor_id compound_dtor)
913 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
914 page[1].compound_dtor = compound_dtor;
917 static inline void destroy_compound_page(struct page *page)
919 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
920 compound_page_dtors[page[1].compound_dtor](page);
923 static inline unsigned int compound_order(struct page *page)
927 return page[1].compound_order;
930 static inline bool hpage_pincount_available(struct page *page)
933 * Can the page->hpage_pinned_refcount field be used? That field is in
934 * the 3rd page of the compound page, so the smallest (2-page) compound
935 * pages cannot support it.
937 page = compound_head(page);
938 return PageCompound(page) && compound_order(page) > 1;
941 static inline int head_compound_pincount(struct page *head)
943 return atomic_read(compound_pincount_ptr(head));
946 static inline int compound_pincount(struct page *page)
948 VM_BUG_ON_PAGE(!hpage_pincount_available(page), page);
949 page = compound_head(page);
950 return head_compound_pincount(page);
953 static inline void set_compound_order(struct page *page, unsigned int order)
955 page[1].compound_order = order;
956 page[1].compound_nr = 1U << order;
959 /* Returns the number of pages in this potentially compound page. */
960 static inline unsigned long compound_nr(struct page *page)
964 return page[1].compound_nr;
967 /* Returns the number of bytes in this potentially compound page. */
968 static inline unsigned long page_size(struct page *page)
970 return PAGE_SIZE << compound_order(page);
973 /* Returns the number of bits needed for the number of bytes in a page */
974 static inline unsigned int page_shift(struct page *page)
976 return PAGE_SHIFT + compound_order(page);
979 void free_compound_page(struct page *page);
983 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
984 * servicing faults for write access. In the normal case, do always want
985 * pte_mkwrite. But get_user_pages can cause write faults for mappings
986 * that do not have writing enabled, when used by access_process_vm.
988 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
990 if (likely(vma->vm_flags & VM_WRITE))
991 pte = pte_mkwrite(pte);
995 vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page);
996 void do_set_pte(struct vm_fault *vmf, struct page *page, unsigned long addr);
998 vm_fault_t finish_fault(struct vm_fault *vmf);
999 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
1003 * Multiple processes may "see" the same page. E.g. for untouched
1004 * mappings of /dev/null, all processes see the same page full of
1005 * zeroes, and text pages of executables and shared libraries have
1006 * only one copy in memory, at most, normally.
1008 * For the non-reserved pages, page_count(page) denotes a reference count.
1009 * page_count() == 0 means the page is free. page->lru is then used for
1010 * freelist management in the buddy allocator.
1011 * page_count() > 0 means the page has been allocated.
1013 * Pages are allocated by the slab allocator in order to provide memory
1014 * to kmalloc and kmem_cache_alloc. In this case, the management of the
1015 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
1016 * unless a particular usage is carefully commented. (the responsibility of
1017 * freeing the kmalloc memory is the caller's, of course).
1019 * A page may be used by anyone else who does a __get_free_page().
1020 * In this case, page_count still tracks the references, and should only
1021 * be used through the normal accessor functions. The top bits of page->flags
1022 * and page->virtual store page management information, but all other fields
1023 * are unused and could be used privately, carefully. The management of this
1024 * page is the responsibility of the one who allocated it, and those who have
1025 * subsequently been given references to it.
1027 * The other pages (we may call them "pagecache pages") are completely
1028 * managed by the Linux memory manager: I/O, buffers, swapping etc.
1029 * The following discussion applies only to them.
1031 * A pagecache page contains an opaque `private' member, which belongs to the
1032 * page's address_space. Usually, this is the address of a circular list of
1033 * the page's disk buffers. PG_private must be set to tell the VM to call
1034 * into the filesystem to release these pages.
1036 * A page may belong to an inode's memory mapping. In this case, page->mapping
1037 * is the pointer to the inode, and page->index is the file offset of the page,
1038 * in units of PAGE_SIZE.
1040 * If pagecache pages are not associated with an inode, they are said to be
1041 * anonymous pages. These may become associated with the swapcache, and in that
1042 * case PG_swapcache is set, and page->private is an offset into the swapcache.
1044 * In either case (swapcache or inode backed), the pagecache itself holds one
1045 * reference to the page. Setting PG_private should also increment the
1046 * refcount. The each user mapping also has a reference to the page.
1048 * The pagecache pages are stored in a per-mapping radix tree, which is
1049 * rooted at mapping->i_pages, and indexed by offset.
1050 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
1051 * lists, we instead now tag pages as dirty/writeback in the radix tree.
1053 * All pagecache pages may be subject to I/O:
1054 * - inode pages may need to be read from disk,
1055 * - inode pages which have been modified and are MAP_SHARED may need
1056 * to be written back to the inode on disk,
1057 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
1058 * modified may need to be swapped out to swap space and (later) to be read
1063 * The zone field is never updated after free_area_init_core()
1064 * sets it, so none of the operations on it need to be atomic.
1067 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
1068 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
1069 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
1070 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
1071 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
1072 #define KASAN_TAG_PGOFF (LAST_CPUPID_PGOFF - KASAN_TAG_WIDTH)
1075 * Define the bit shifts to access each section. For non-existent
1076 * sections we define the shift as 0; that plus a 0 mask ensures
1077 * the compiler will optimise away reference to them.
1079 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
1080 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
1081 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
1082 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
1083 #define KASAN_TAG_PGSHIFT (KASAN_TAG_PGOFF * (KASAN_TAG_WIDTH != 0))
1085 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
1086 #ifdef NODE_NOT_IN_PAGE_FLAGS
1087 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
1088 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
1089 SECTIONS_PGOFF : ZONES_PGOFF)
1091 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
1092 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
1093 NODES_PGOFF : ZONES_PGOFF)
1096 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
1098 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
1099 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
1100 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
1101 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
1102 #define KASAN_TAG_MASK ((1UL << KASAN_TAG_WIDTH) - 1)
1103 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
1105 static inline enum zone_type page_zonenum(const struct page *page)
1107 ASSERT_EXCLUSIVE_BITS(page->flags, ZONES_MASK << ZONES_PGSHIFT);
1108 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
1111 #ifdef CONFIG_ZONE_DEVICE
1112 static inline bool is_zone_device_page(const struct page *page)
1114 return page_zonenum(page) == ZONE_DEVICE;
1116 extern void memmap_init_zone_device(struct zone *, unsigned long,
1117 unsigned long, struct dev_pagemap *);
1119 static inline bool is_zone_device_page(const struct page *page)
1125 #ifdef CONFIG_DEV_PAGEMAP_OPS
1126 void free_devmap_managed_page(struct page *page);
1127 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
1129 static inline bool page_is_devmap_managed(struct page *page)
1131 if (!static_branch_unlikely(&devmap_managed_key))
1133 if (!is_zone_device_page(page))
1135 switch (page->pgmap->type) {
1136 case MEMORY_DEVICE_PRIVATE:
1137 case MEMORY_DEVICE_FS_DAX:
1145 void put_devmap_managed_page(struct page *page);
1147 #else /* CONFIG_DEV_PAGEMAP_OPS */
1148 static inline bool page_is_devmap_managed(struct page *page)
1153 static inline void put_devmap_managed_page(struct page *page)
1156 #endif /* CONFIG_DEV_PAGEMAP_OPS */
1158 static inline bool is_device_private_page(const struct page *page)
1160 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
1161 IS_ENABLED(CONFIG_DEVICE_PRIVATE) &&
1162 is_zone_device_page(page) &&
1163 page->pgmap->type == MEMORY_DEVICE_PRIVATE;
1166 static inline bool is_pci_p2pdma_page(const struct page *page)
1168 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
1169 IS_ENABLED(CONFIG_PCI_P2PDMA) &&
1170 is_zone_device_page(page) &&
1171 page->pgmap->type == MEMORY_DEVICE_PCI_P2PDMA;
1174 /* 127: arbitrary random number, small enough to assemble well */
1175 #define page_ref_zero_or_close_to_overflow(page) \
1176 ((unsigned int) page_ref_count(page) + 127u <= 127u)
1178 static inline void get_page(struct page *page)
1180 page = compound_head(page);
1182 * Getting a normal page or the head of a compound page
1183 * requires to already have an elevated page->_refcount.
1185 VM_BUG_ON_PAGE(page_ref_zero_or_close_to_overflow(page), page);
1189 bool __must_check try_grab_page(struct page *page, unsigned int flags);
1190 __maybe_unused struct page *try_grab_compound_head(struct page *page, int refs,
1191 unsigned int flags);
1194 static inline __must_check bool try_get_page(struct page *page)
1196 page = compound_head(page);
1197 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1203 static inline void put_page(struct page *page)
1205 page = compound_head(page);
1208 * For devmap managed pages we need to catch refcount transition from
1209 * 2 to 1, when refcount reach one it means the page is free and we
1210 * need to inform the device driver through callback. See
1211 * include/linux/memremap.h and HMM for details.
1213 if (page_is_devmap_managed(page)) {
1214 put_devmap_managed_page(page);
1218 if (put_page_testzero(page))
1223 * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
1224 * the page's refcount so that two separate items are tracked: the original page
1225 * reference count, and also a new count of how many pin_user_pages() calls were
1226 * made against the page. ("gup-pinned" is another term for the latter).
1228 * With this scheme, pin_user_pages() becomes special: such pages are marked as
1229 * distinct from normal pages. As such, the unpin_user_page() call (and its
1230 * variants) must be used in order to release gup-pinned pages.
1234 * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
1235 * counts with respect to pin_user_pages() and unpin_user_page() becomes
1236 * simpler, due to the fact that adding an even power of two to the page
1237 * refcount has the effect of using only the upper N bits, for the code that
1238 * counts up using the bias value. This means that the lower bits are left for
1239 * the exclusive use of the original code that increments and decrements by one
1240 * (or at least, by much smaller values than the bias value).
1242 * Of course, once the lower bits overflow into the upper bits (and this is
1243 * OK, because subtraction recovers the original values), then visual inspection
1244 * no longer suffices to directly view the separate counts. However, for normal
1245 * applications that don't have huge page reference counts, this won't be an
1248 * Locking: the lockless algorithm described in page_cache_get_speculative()
1249 * and page_cache_gup_pin_speculative() provides safe operation for
1250 * get_user_pages and page_mkclean and other calls that race to set up page
1253 #define GUP_PIN_COUNTING_BIAS (1U << 10)
1255 void unpin_user_page(struct page *page);
1256 void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1258 void unpin_user_pages(struct page **pages, unsigned long npages);
1261 * page_maybe_dma_pinned() - report if a page is pinned for DMA.
1263 * This function checks if a page has been pinned via a call to
1264 * pin_user_pages*().
1266 * For non-huge pages, the return value is partially fuzzy: false is not fuzzy,
1267 * because it means "definitely not pinned for DMA", but true means "probably
1268 * pinned for DMA, but possibly a false positive due to having at least
1269 * GUP_PIN_COUNTING_BIAS worth of normal page references".
1271 * False positives are OK, because: a) it's unlikely for a page to get that many
1272 * refcounts, and b) all the callers of this routine are expected to be able to
1273 * deal gracefully with a false positive.
1275 * For huge pages, the result will be exactly correct. That's because we have
1276 * more tracking data available: the 3rd struct page in the compound page is
1277 * used to track the pincount (instead using of the GUP_PIN_COUNTING_BIAS
1280 * For more information, please see Documentation/core-api/pin_user_pages.rst.
1282 * @page: pointer to page to be queried.
1283 * @Return: True, if it is likely that the page has been "dma-pinned".
1284 * False, if the page is definitely not dma-pinned.
1286 static inline bool page_maybe_dma_pinned(struct page *page)
1288 if (hpage_pincount_available(page))
1289 return compound_pincount(page) > 0;
1292 * page_ref_count() is signed. If that refcount overflows, then
1293 * page_ref_count() returns a negative value, and callers will avoid
1294 * further incrementing the refcount.
1296 * Here, for that overflow case, use the signed bit to count a little
1297 * bit higher via unsigned math, and thus still get an accurate result.
1299 return ((unsigned int)page_ref_count(compound_head(page))) >=
1300 GUP_PIN_COUNTING_BIAS;
1303 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1304 #define SECTION_IN_PAGE_FLAGS
1308 * The identification function is mainly used by the buddy allocator for
1309 * determining if two pages could be buddies. We are not really identifying
1310 * the zone since we could be using the section number id if we do not have
1311 * node id available in page flags.
1312 * We only guarantee that it will return the same value for two combinable
1315 static inline int page_zone_id(struct page *page)
1317 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1320 #ifdef NODE_NOT_IN_PAGE_FLAGS
1321 extern int page_to_nid(const struct page *page);
1323 static inline int page_to_nid(const struct page *page)
1325 struct page *p = (struct page *)page;
1327 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1331 #ifdef CONFIG_NUMA_BALANCING
1332 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1334 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1337 static inline int cpupid_to_pid(int cpupid)
1339 return cpupid & LAST__PID_MASK;
1342 static inline int cpupid_to_cpu(int cpupid)
1344 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1347 static inline int cpupid_to_nid(int cpupid)
1349 return cpu_to_node(cpupid_to_cpu(cpupid));
1352 static inline bool cpupid_pid_unset(int cpupid)
1354 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1357 static inline bool cpupid_cpu_unset(int cpupid)
1359 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1362 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1364 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1367 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1368 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1369 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1371 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1374 static inline int page_cpupid_last(struct page *page)
1376 return page->_last_cpupid;
1378 static inline void page_cpupid_reset_last(struct page *page)
1380 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1383 static inline int page_cpupid_last(struct page *page)
1385 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1388 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1390 static inline void page_cpupid_reset_last(struct page *page)
1392 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1394 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1395 #else /* !CONFIG_NUMA_BALANCING */
1396 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1398 return page_to_nid(page); /* XXX */
1401 static inline int page_cpupid_last(struct page *page)
1403 return page_to_nid(page); /* XXX */
1406 static inline int cpupid_to_nid(int cpupid)
1411 static inline int cpupid_to_pid(int cpupid)
1416 static inline int cpupid_to_cpu(int cpupid)
1421 static inline int cpu_pid_to_cpupid(int nid, int pid)
1426 static inline bool cpupid_pid_unset(int cpupid)
1431 static inline void page_cpupid_reset_last(struct page *page)
1435 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1439 #endif /* CONFIG_NUMA_BALANCING */
1441 #if defined(CONFIG_KASAN_SW_TAGS) || defined(CONFIG_KASAN_HW_TAGS)
1443 static inline u8 page_kasan_tag(const struct page *page)
1445 if (kasan_enabled())
1446 return (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1450 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1452 if (kasan_enabled()) {
1453 page->flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1454 page->flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1458 static inline void page_kasan_tag_reset(struct page *page)
1460 if (kasan_enabled())
1461 page_kasan_tag_set(page, 0xff);
1464 #else /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1466 static inline u8 page_kasan_tag(const struct page *page)
1471 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1472 static inline void page_kasan_tag_reset(struct page *page) { }
1474 #endif /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1476 static inline struct zone *page_zone(const struct page *page)
1478 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1481 static inline pg_data_t *page_pgdat(const struct page *page)
1483 return NODE_DATA(page_to_nid(page));
1486 #ifdef SECTION_IN_PAGE_FLAGS
1487 static inline void set_page_section(struct page *page, unsigned long section)
1489 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1490 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1493 static inline unsigned long page_to_section(const struct page *page)
1495 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1499 static inline void set_page_zone(struct page *page, enum zone_type zone)
1501 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1502 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1505 static inline void set_page_node(struct page *page, unsigned long node)
1507 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1508 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1511 static inline void set_page_links(struct page *page, enum zone_type zone,
1512 unsigned long node, unsigned long pfn)
1514 set_page_zone(page, zone);
1515 set_page_node(page, node);
1516 #ifdef SECTION_IN_PAGE_FLAGS
1517 set_page_section(page, pfn_to_section_nr(pfn));
1522 * Some inline functions in vmstat.h depend on page_zone()
1524 #include <linux/vmstat.h>
1526 static __always_inline void *lowmem_page_address(const struct page *page)
1528 return page_to_virt(page);
1531 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1532 #define HASHED_PAGE_VIRTUAL
1535 #if defined(WANT_PAGE_VIRTUAL)
1536 static inline void *page_address(const struct page *page)
1538 return page->virtual;
1540 static inline void set_page_address(struct page *page, void *address)
1542 page->virtual = address;
1544 #define page_address_init() do { } while(0)
1547 #if defined(HASHED_PAGE_VIRTUAL)
1548 void *page_address(const struct page *page);
1549 void set_page_address(struct page *page, void *virtual);
1550 void page_address_init(void);
1553 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1554 #define page_address(page) lowmem_page_address(page)
1555 #define set_page_address(page, address) do { } while(0)
1556 #define page_address_init() do { } while(0)
1559 extern void *page_rmapping(struct page *page);
1560 extern struct anon_vma *page_anon_vma(struct page *page);
1561 extern struct address_space *page_mapping(struct page *page);
1563 extern struct address_space *__page_file_mapping(struct page *);
1566 struct address_space *page_file_mapping(struct page *page)
1568 if (unlikely(PageSwapCache(page)))
1569 return __page_file_mapping(page);
1571 return page->mapping;
1574 extern pgoff_t __page_file_index(struct page *page);
1577 * Return the pagecache index of the passed page. Regular pagecache pages
1578 * use ->index whereas swapcache pages use swp_offset(->private)
1580 static inline pgoff_t page_index(struct page *page)
1582 if (unlikely(PageSwapCache(page)))
1583 return __page_file_index(page);
1587 bool page_mapped(struct page *page);
1588 struct address_space *page_mapping(struct page *page);
1589 struct address_space *page_mapping_file(struct page *page);
1592 * Return true only if the page has been allocated with
1593 * ALLOC_NO_WATERMARKS and the low watermark was not
1594 * met implying that the system is under some pressure.
1596 static inline bool page_is_pfmemalloc(const struct page *page)
1599 * Page index cannot be this large so this must be
1600 * a pfmemalloc page.
1602 return page->index == -1UL;
1606 * Only to be called by the page allocator on a freshly allocated
1609 static inline void set_page_pfmemalloc(struct page *page)
1614 static inline void clear_page_pfmemalloc(struct page *page)
1620 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1622 extern void pagefault_out_of_memory(void);
1624 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1625 #define offset_in_thp(page, p) ((unsigned long)(p) & (thp_size(page) - 1))
1628 * Flags passed to show_mem() and show_free_areas() to suppress output in
1631 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1633 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1636 extern bool can_do_mlock(void);
1638 static inline bool can_do_mlock(void) { return false; }
1640 extern int user_shm_lock(size_t, struct user_struct *);
1641 extern void user_shm_unlock(size_t, struct user_struct *);
1644 * Parameter block passed down to zap_pte_range in exceptional cases.
1646 struct zap_details {
1647 struct address_space *check_mapping; /* Check page->mapping if set */
1648 pgoff_t first_index; /* Lowest page->index to unmap */
1649 pgoff_t last_index; /* Highest page->index to unmap */
1652 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1654 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1657 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1658 unsigned long size);
1659 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1660 unsigned long size);
1661 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1662 unsigned long start, unsigned long end);
1664 struct mmu_notifier_range;
1666 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1667 unsigned long end, unsigned long floor, unsigned long ceiling);
1669 copy_page_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma);
1670 int follow_invalidate_pte(struct mm_struct *mm, unsigned long address,
1671 struct mmu_notifier_range *range, pte_t **ptepp,
1672 pmd_t **pmdpp, spinlock_t **ptlp);
1673 int follow_pte(struct mm_struct *mm, unsigned long address,
1674 pte_t **ptepp, spinlock_t **ptlp);
1675 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1676 unsigned long *pfn);
1677 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1678 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1679 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1680 void *buf, int len, int write);
1682 extern void truncate_pagecache(struct inode *inode, loff_t new);
1683 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1684 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1685 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1686 int truncate_inode_page(struct address_space *mapping, struct page *page);
1687 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1688 int invalidate_inode_page(struct page *page);
1691 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1692 unsigned long address, unsigned int flags,
1693 struct pt_regs *regs);
1694 extern int fixup_user_fault(struct mm_struct *mm,
1695 unsigned long address, unsigned int fault_flags,
1697 void unmap_mapping_pages(struct address_space *mapping,
1698 pgoff_t start, pgoff_t nr, bool even_cows);
1699 void unmap_mapping_range(struct address_space *mapping,
1700 loff_t const holebegin, loff_t const holelen, int even_cows);
1702 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1703 unsigned long address, unsigned int flags,
1704 struct pt_regs *regs)
1706 /* should never happen if there's no MMU */
1708 return VM_FAULT_SIGBUS;
1710 static inline int fixup_user_fault(struct mm_struct *mm, unsigned long address,
1711 unsigned int fault_flags, bool *unlocked)
1713 /* should never happen if there's no MMU */
1717 static inline void unmap_mapping_pages(struct address_space *mapping,
1718 pgoff_t start, pgoff_t nr, bool even_cows) { }
1719 static inline void unmap_mapping_range(struct address_space *mapping,
1720 loff_t const holebegin, loff_t const holelen, int even_cows) { }
1723 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1724 loff_t const holebegin, loff_t const holelen)
1726 unmap_mapping_range(mapping, holebegin, holelen, 0);
1729 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1730 void *buf, int len, unsigned int gup_flags);
1731 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1732 void *buf, int len, unsigned int gup_flags);
1733 extern int __access_remote_vm(struct mm_struct *mm, unsigned long addr,
1734 void *buf, int len, unsigned int gup_flags);
1736 long get_user_pages_remote(struct mm_struct *mm,
1737 unsigned long start, unsigned long nr_pages,
1738 unsigned int gup_flags, struct page **pages,
1739 struct vm_area_struct **vmas, int *locked);
1740 long pin_user_pages_remote(struct mm_struct *mm,
1741 unsigned long start, unsigned long nr_pages,
1742 unsigned int gup_flags, struct page **pages,
1743 struct vm_area_struct **vmas, int *locked);
1744 long get_user_pages(unsigned long start, unsigned long nr_pages,
1745 unsigned int gup_flags, struct page **pages,
1746 struct vm_area_struct **vmas);
1747 long pin_user_pages(unsigned long start, unsigned long nr_pages,
1748 unsigned int gup_flags, struct page **pages,
1749 struct vm_area_struct **vmas);
1750 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1751 unsigned int gup_flags, struct page **pages, int *locked);
1752 long pin_user_pages_locked(unsigned long start, unsigned long nr_pages,
1753 unsigned int gup_flags, struct page **pages, int *locked);
1754 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1755 struct page **pages, unsigned int gup_flags);
1756 long pin_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1757 struct page **pages, unsigned int gup_flags);
1759 int get_user_pages_fast(unsigned long start, int nr_pages,
1760 unsigned int gup_flags, struct page **pages);
1761 int pin_user_pages_fast(unsigned long start, int nr_pages,
1762 unsigned int gup_flags, struct page **pages);
1764 int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
1765 int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
1766 struct task_struct *task, bool bypass_rlim);
1769 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1770 struct page **pages);
1771 int get_kernel_page(unsigned long start, int write, struct page **pages);
1772 struct page *get_dump_page(unsigned long addr);
1774 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1775 extern void do_invalidatepage(struct page *page, unsigned int offset,
1776 unsigned int length);
1778 void __set_page_dirty(struct page *, struct address_space *, int warn);
1779 int __set_page_dirty_nobuffers(struct page *page);
1780 int __set_page_dirty_no_writeback(struct page *page);
1781 int redirty_page_for_writepage(struct writeback_control *wbc,
1783 void account_page_dirtied(struct page *page, struct address_space *mapping);
1784 void account_page_cleaned(struct page *page, struct address_space *mapping,
1785 struct bdi_writeback *wb);
1786 int set_page_dirty(struct page *page);
1787 int set_page_dirty_lock(struct page *page);
1788 void __cancel_dirty_page(struct page *page);
1789 static inline void cancel_dirty_page(struct page *page)
1791 /* Avoid atomic ops, locking, etc. when not actually needed. */
1792 if (PageDirty(page))
1793 __cancel_dirty_page(page);
1795 int clear_page_dirty_for_io(struct page *page);
1797 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1799 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1800 unsigned long old_addr, struct vm_area_struct *new_vma,
1801 unsigned long new_addr, unsigned long len,
1802 bool need_rmap_locks);
1805 * Flags used by change_protection(). For now we make it a bitmap so
1806 * that we can pass in multiple flags just like parameters. However
1807 * for now all the callers are only use one of the flags at the same
1810 /* Whether we should allow dirty bit accounting */
1811 #define MM_CP_DIRTY_ACCT (1UL << 0)
1812 /* Whether this protection change is for NUMA hints */
1813 #define MM_CP_PROT_NUMA (1UL << 1)
1814 /* Whether this change is for write protecting */
1815 #define MM_CP_UFFD_WP (1UL << 2) /* do wp */
1816 #define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */
1817 #define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \
1818 MM_CP_UFFD_WP_RESOLVE)
1820 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1821 unsigned long end, pgprot_t newprot,
1822 unsigned long cp_flags);
1823 extern int mprotect_fixup(struct vm_area_struct *vma,
1824 struct vm_area_struct **pprev, unsigned long start,
1825 unsigned long end, unsigned long newflags);
1828 * doesn't attempt to fault and will return short.
1830 int get_user_pages_fast_only(unsigned long start, int nr_pages,
1831 unsigned int gup_flags, struct page **pages);
1832 int pin_user_pages_fast_only(unsigned long start, int nr_pages,
1833 unsigned int gup_flags, struct page **pages);
1835 static inline bool get_user_page_fast_only(unsigned long addr,
1836 unsigned int gup_flags, struct page **pagep)
1838 return get_user_pages_fast_only(addr, 1, gup_flags, pagep) == 1;
1841 * per-process(per-mm_struct) statistics.
1843 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1845 long val = atomic_long_read(&mm->rss_stat.count[member]);
1847 #ifdef SPLIT_RSS_COUNTING
1849 * counter is updated in asynchronous manner and may go to minus.
1850 * But it's never be expected number for users.
1855 return (unsigned long)val;
1858 void mm_trace_rss_stat(struct mm_struct *mm, int member, long count);
1860 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1862 long count = atomic_long_add_return(value, &mm->rss_stat.count[member]);
1864 mm_trace_rss_stat(mm, member, count);
1867 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1869 long count = atomic_long_inc_return(&mm->rss_stat.count[member]);
1871 mm_trace_rss_stat(mm, member, count);
1874 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1876 long count = atomic_long_dec_return(&mm->rss_stat.count[member]);
1878 mm_trace_rss_stat(mm, member, count);
1881 /* Optimized variant when page is already known not to be PageAnon */
1882 static inline int mm_counter_file(struct page *page)
1884 if (PageSwapBacked(page))
1885 return MM_SHMEMPAGES;
1886 return MM_FILEPAGES;
1889 static inline int mm_counter(struct page *page)
1892 return MM_ANONPAGES;
1893 return mm_counter_file(page);
1896 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1898 return get_mm_counter(mm, MM_FILEPAGES) +
1899 get_mm_counter(mm, MM_ANONPAGES) +
1900 get_mm_counter(mm, MM_SHMEMPAGES);
1903 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1905 return max(mm->hiwater_rss, get_mm_rss(mm));
1908 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1910 return max(mm->hiwater_vm, mm->total_vm);
1913 static inline void update_hiwater_rss(struct mm_struct *mm)
1915 unsigned long _rss = get_mm_rss(mm);
1917 if ((mm)->hiwater_rss < _rss)
1918 (mm)->hiwater_rss = _rss;
1921 static inline void update_hiwater_vm(struct mm_struct *mm)
1923 if (mm->hiwater_vm < mm->total_vm)
1924 mm->hiwater_vm = mm->total_vm;
1927 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1929 mm->hiwater_rss = get_mm_rss(mm);
1932 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1933 struct mm_struct *mm)
1935 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1937 if (*maxrss < hiwater_rss)
1938 *maxrss = hiwater_rss;
1941 #if defined(SPLIT_RSS_COUNTING)
1942 void sync_mm_rss(struct mm_struct *mm);
1944 static inline void sync_mm_rss(struct mm_struct *mm)
1949 #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
1950 static inline int pte_special(pte_t pte)
1955 static inline pte_t pte_mkspecial(pte_t pte)
1961 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
1962 static inline int pte_devmap(pte_t pte)
1968 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1970 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1972 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1976 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1980 #ifdef __PAGETABLE_P4D_FOLDED
1981 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1982 unsigned long address)
1987 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1990 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
1991 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1992 unsigned long address)
1996 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
1997 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
2000 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
2002 static inline void mm_inc_nr_puds(struct mm_struct *mm)
2004 if (mm_pud_folded(mm))
2006 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2009 static inline void mm_dec_nr_puds(struct mm_struct *mm)
2011 if (mm_pud_folded(mm))
2013 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2017 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
2018 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
2019 unsigned long address)
2024 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
2025 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
2028 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
2030 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
2032 if (mm_pmd_folded(mm))
2034 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2037 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
2039 if (mm_pmd_folded(mm))
2041 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2046 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
2048 atomic_long_set(&mm->pgtables_bytes, 0);
2051 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2053 return atomic_long_read(&mm->pgtables_bytes);
2056 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
2058 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2061 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
2063 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2067 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
2068 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2073 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
2074 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
2077 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
2078 int __pte_alloc_kernel(pmd_t *pmd);
2080 #if defined(CONFIG_MMU)
2082 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2083 unsigned long address)
2085 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
2086 NULL : p4d_offset(pgd, address);
2089 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2090 unsigned long address)
2092 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
2093 NULL : pud_offset(p4d, address);
2096 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
2098 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
2099 NULL: pmd_offset(pud, address);
2101 #endif /* CONFIG_MMU */
2103 #if USE_SPLIT_PTE_PTLOCKS
2104 #if ALLOC_SPLIT_PTLOCKS
2105 void __init ptlock_cache_init(void);
2106 extern bool ptlock_alloc(struct page *page);
2107 extern void ptlock_free(struct page *page);
2109 static inline spinlock_t *ptlock_ptr(struct page *page)
2113 #else /* ALLOC_SPLIT_PTLOCKS */
2114 static inline void ptlock_cache_init(void)
2118 static inline bool ptlock_alloc(struct page *page)
2123 static inline void ptlock_free(struct page *page)
2127 static inline spinlock_t *ptlock_ptr(struct page *page)
2131 #endif /* ALLOC_SPLIT_PTLOCKS */
2133 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2135 return ptlock_ptr(pmd_page(*pmd));
2138 static inline bool ptlock_init(struct page *page)
2141 * prep_new_page() initialize page->private (and therefore page->ptl)
2142 * with 0. Make sure nobody took it in use in between.
2144 * It can happen if arch try to use slab for page table allocation:
2145 * slab code uses page->slab_cache, which share storage with page->ptl.
2147 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
2148 if (!ptlock_alloc(page))
2150 spin_lock_init(ptlock_ptr(page));
2154 #else /* !USE_SPLIT_PTE_PTLOCKS */
2156 * We use mm->page_table_lock to guard all pagetable pages of the mm.
2158 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2160 return &mm->page_table_lock;
2162 static inline void ptlock_cache_init(void) {}
2163 static inline bool ptlock_init(struct page *page) { return true; }
2164 static inline void ptlock_free(struct page *page) {}
2165 #endif /* USE_SPLIT_PTE_PTLOCKS */
2167 static inline void pgtable_init(void)
2169 ptlock_cache_init();
2170 pgtable_cache_init();
2173 static inline bool pgtable_pte_page_ctor(struct page *page)
2175 if (!ptlock_init(page))
2177 __SetPageTable(page);
2178 inc_lruvec_page_state(page, NR_PAGETABLE);
2182 static inline void pgtable_pte_page_dtor(struct page *page)
2185 __ClearPageTable(page);
2186 dec_lruvec_page_state(page, NR_PAGETABLE);
2189 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
2191 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
2192 pte_t *__pte = pte_offset_map(pmd, address); \
2198 #define pte_unmap_unlock(pte, ptl) do { \
2203 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2205 #define pte_alloc_map(mm, pmd, address) \
2206 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2208 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2209 (pte_alloc(mm, pmd) ? \
2210 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2212 #define pte_alloc_kernel(pmd, address) \
2213 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2214 NULL: pte_offset_kernel(pmd, address))
2216 #if USE_SPLIT_PMD_PTLOCKS
2218 static struct page *pmd_to_page(pmd_t *pmd)
2220 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2221 return virt_to_page((void *)((unsigned long) pmd & mask));
2224 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2226 return ptlock_ptr(pmd_to_page(pmd));
2229 static inline bool pmd_ptlock_init(struct page *page)
2231 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2232 page->pmd_huge_pte = NULL;
2234 return ptlock_init(page);
2237 static inline void pmd_ptlock_free(struct page *page)
2239 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2240 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2245 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2249 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2251 return &mm->page_table_lock;
2254 static inline bool pmd_ptlock_init(struct page *page) { return true; }
2255 static inline void pmd_ptlock_free(struct page *page) {}
2257 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2261 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2263 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2268 static inline bool pgtable_pmd_page_ctor(struct page *page)
2270 if (!pmd_ptlock_init(page))
2272 __SetPageTable(page);
2273 inc_lruvec_page_state(page, NR_PAGETABLE);
2277 static inline void pgtable_pmd_page_dtor(struct page *page)
2279 pmd_ptlock_free(page);
2280 __ClearPageTable(page);
2281 dec_lruvec_page_state(page, NR_PAGETABLE);
2285 * No scalability reason to split PUD locks yet, but follow the same pattern
2286 * as the PMD locks to make it easier if we decide to. The VM should not be
2287 * considered ready to switch to split PUD locks yet; there may be places
2288 * which need to be converted from page_table_lock.
2290 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2292 return &mm->page_table_lock;
2295 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2297 spinlock_t *ptl = pud_lockptr(mm, pud);
2303 extern void __init pagecache_init(void);
2304 extern void __init free_area_init_memoryless_node(int nid);
2305 extern void free_initmem(void);
2308 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2309 * into the buddy system. The freed pages will be poisoned with pattern
2310 * "poison" if it's within range [0, UCHAR_MAX].
2311 * Return pages freed into the buddy system.
2313 extern unsigned long free_reserved_area(void *start, void *end,
2314 int poison, const char *s);
2316 extern void adjust_managed_page_count(struct page *page, long count);
2317 extern void mem_init_print_info(const char *str);
2319 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2321 /* Free the reserved page into the buddy system, so it gets managed. */
2322 static inline void free_reserved_page(struct page *page)
2324 ClearPageReserved(page);
2325 init_page_count(page);
2327 adjust_managed_page_count(page, 1);
2329 #define free_highmem_page(page) free_reserved_page(page)
2331 static inline void mark_page_reserved(struct page *page)
2333 SetPageReserved(page);
2334 adjust_managed_page_count(page, -1);
2338 * Default method to free all the __init memory into the buddy system.
2339 * The freed pages will be poisoned with pattern "poison" if it's within
2340 * range [0, UCHAR_MAX].
2341 * Return pages freed into the buddy system.
2343 static inline unsigned long free_initmem_default(int poison)
2345 extern char __init_begin[], __init_end[];
2347 return free_reserved_area(&__init_begin, &__init_end,
2348 poison, "unused kernel");
2351 static inline unsigned long get_num_physpages(void)
2354 unsigned long phys_pages = 0;
2356 for_each_online_node(nid)
2357 phys_pages += node_present_pages(nid);
2363 * Using memblock node mappings, an architecture may initialise its
2364 * zones, allocate the backing mem_map and account for memory holes in an
2365 * architecture independent manner.
2367 * An architecture is expected to register range of page frames backed by
2368 * physical memory with memblock_add[_node]() before calling
2369 * free_area_init() passing in the PFN each zone ends at. At a basic
2370 * usage, an architecture is expected to do something like
2372 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2374 * for_each_valid_physical_page_range()
2375 * memblock_add_node(base, size, nid)
2376 * free_area_init(max_zone_pfns);
2378 void free_area_init(unsigned long *max_zone_pfn);
2379 unsigned long node_map_pfn_alignment(void);
2380 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2381 unsigned long end_pfn);
2382 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2383 unsigned long end_pfn);
2384 extern void get_pfn_range_for_nid(unsigned int nid,
2385 unsigned long *start_pfn, unsigned long *end_pfn);
2386 extern unsigned long find_min_pfn_with_active_regions(void);
2388 #ifndef CONFIG_NEED_MULTIPLE_NODES
2389 static inline int early_pfn_to_nid(unsigned long pfn)
2394 /* please see mm/page_alloc.c */
2395 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2398 extern void set_dma_reserve(unsigned long new_dma_reserve);
2399 extern void memmap_init_range(unsigned long, int, unsigned long,
2400 unsigned long, unsigned long, enum meminit_context,
2401 struct vmem_altmap *, int migratetype);
2402 extern void memmap_init_zone(struct zone *zone);
2403 extern void setup_per_zone_wmarks(void);
2404 extern int __meminit init_per_zone_wmark_min(void);
2405 extern void mem_init(void);
2406 extern void __init mmap_init(void);
2407 extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2408 extern long si_mem_available(void);
2409 extern void si_meminfo(struct sysinfo * val);
2410 extern void si_meminfo_node(struct sysinfo *val, int nid);
2411 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2412 extern unsigned long arch_reserved_kernel_pages(void);
2415 extern __printf(3, 4)
2416 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2418 extern void setup_per_cpu_pageset(void);
2421 extern int min_free_kbytes;
2422 extern int watermark_boost_factor;
2423 extern int watermark_scale_factor;
2424 extern bool arch_has_descending_max_zone_pfns(void);
2427 extern atomic_long_t mmap_pages_allocated;
2428 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2430 /* interval_tree.c */
2431 void vma_interval_tree_insert(struct vm_area_struct *node,
2432 struct rb_root_cached *root);
2433 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2434 struct vm_area_struct *prev,
2435 struct rb_root_cached *root);
2436 void vma_interval_tree_remove(struct vm_area_struct *node,
2437 struct rb_root_cached *root);
2438 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2439 unsigned long start, unsigned long last);
2440 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2441 unsigned long start, unsigned long last);
2443 #define vma_interval_tree_foreach(vma, root, start, last) \
2444 for (vma = vma_interval_tree_iter_first(root, start, last); \
2445 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2447 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2448 struct rb_root_cached *root);
2449 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2450 struct rb_root_cached *root);
2451 struct anon_vma_chain *
2452 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2453 unsigned long start, unsigned long last);
2454 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2455 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2456 #ifdef CONFIG_DEBUG_VM_RB
2457 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2460 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2461 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2462 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2465 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2466 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2467 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2468 struct vm_area_struct *expand);
2469 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2470 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2472 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2474 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2475 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2476 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2477 struct mempolicy *, struct vm_userfaultfd_ctx);
2478 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2479 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2480 unsigned long addr, int new_below);
2481 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2482 unsigned long addr, int new_below);
2483 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2484 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2485 struct rb_node **, struct rb_node *);
2486 extern void unlink_file_vma(struct vm_area_struct *);
2487 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2488 unsigned long addr, unsigned long len, pgoff_t pgoff,
2489 bool *need_rmap_locks);
2490 extern void exit_mmap(struct mm_struct *);
2492 static inline int check_data_rlimit(unsigned long rlim,
2494 unsigned long start,
2495 unsigned long end_data,
2496 unsigned long start_data)
2498 if (rlim < RLIM_INFINITY) {
2499 if (((new - start) + (end_data - start_data)) > rlim)
2506 extern int mm_take_all_locks(struct mm_struct *mm);
2507 extern void mm_drop_all_locks(struct mm_struct *mm);
2509 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2510 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2511 extern struct file *get_task_exe_file(struct task_struct *task);
2513 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2514 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2516 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2517 const struct vm_special_mapping *sm);
2518 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2519 unsigned long addr, unsigned long len,
2520 unsigned long flags,
2521 const struct vm_special_mapping *spec);
2522 /* This is an obsolete alternative to _install_special_mapping. */
2523 extern int install_special_mapping(struct mm_struct *mm,
2524 unsigned long addr, unsigned long len,
2525 unsigned long flags, struct page **pages);
2527 unsigned long randomize_stack_top(unsigned long stack_top);
2529 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2531 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2532 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2533 struct list_head *uf);
2534 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2535 unsigned long len, unsigned long prot, unsigned long flags,
2536 unsigned long pgoff, unsigned long *populate, struct list_head *uf);
2537 extern int __do_munmap(struct mm_struct *, unsigned long, size_t,
2538 struct list_head *uf, bool downgrade);
2539 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2540 struct list_head *uf);
2541 extern int do_madvise(struct mm_struct *mm, unsigned long start, size_t len_in, int behavior);
2544 extern int __mm_populate(unsigned long addr, unsigned long len,
2546 static inline void mm_populate(unsigned long addr, unsigned long len)
2549 (void) __mm_populate(addr, len, 1);
2552 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2555 /* These take the mm semaphore themselves */
2556 extern int __must_check vm_brk(unsigned long, unsigned long);
2557 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2558 extern int vm_munmap(unsigned long, size_t);
2559 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2560 unsigned long, unsigned long,
2561 unsigned long, unsigned long);
2563 struct vm_unmapped_area_info {
2564 #define VM_UNMAPPED_AREA_TOPDOWN 1
2565 unsigned long flags;
2566 unsigned long length;
2567 unsigned long low_limit;
2568 unsigned long high_limit;
2569 unsigned long align_mask;
2570 unsigned long align_offset;
2573 extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info);
2576 extern void truncate_inode_pages(struct address_space *, loff_t);
2577 extern void truncate_inode_pages_range(struct address_space *,
2578 loff_t lstart, loff_t lend);
2579 extern void truncate_inode_pages_final(struct address_space *);
2581 /* generic vm_area_ops exported for stackable file systems */
2582 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2583 extern vm_fault_t filemap_map_pages(struct vm_fault *vmf,
2584 pgoff_t start_pgoff, pgoff_t end_pgoff);
2585 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2587 /* mm/page-writeback.c */
2588 int __must_check write_one_page(struct page *page);
2589 void task_dirty_inc(struct task_struct *tsk);
2591 extern unsigned long stack_guard_gap;
2592 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2593 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2595 /* CONFIG_STACK_GROWSUP still needs to grow downwards at some places */
2596 extern int expand_downwards(struct vm_area_struct *vma,
2597 unsigned long address);
2599 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2601 #define expand_upwards(vma, address) (0)
2604 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2605 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2606 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2607 struct vm_area_struct **pprev);
2609 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2610 NULL if none. Assume start_addr < end_addr. */
2611 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2613 struct vm_area_struct * vma = find_vma(mm,start_addr);
2615 if (vma && end_addr <= vma->vm_start)
2620 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2622 unsigned long vm_start = vma->vm_start;
2624 if (vma->vm_flags & VM_GROWSDOWN) {
2625 vm_start -= stack_guard_gap;
2626 if (vm_start > vma->vm_start)
2632 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2634 unsigned long vm_end = vma->vm_end;
2636 if (vma->vm_flags & VM_GROWSUP) {
2637 vm_end += stack_guard_gap;
2638 if (vm_end < vma->vm_end)
2639 vm_end = -PAGE_SIZE;
2644 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2646 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2649 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2650 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2651 unsigned long vm_start, unsigned long vm_end)
2653 struct vm_area_struct *vma = find_vma(mm, vm_start);
2655 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2661 static inline bool range_in_vma(struct vm_area_struct *vma,
2662 unsigned long start, unsigned long end)
2664 return (vma && vma->vm_start <= start && end <= vma->vm_end);
2668 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2669 void vma_set_page_prot(struct vm_area_struct *vma);
2671 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2675 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2677 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2681 void vma_set_file(struct vm_area_struct *vma, struct file *file);
2683 #ifdef CONFIG_NUMA_BALANCING
2684 unsigned long change_prot_numa(struct vm_area_struct *vma,
2685 unsigned long start, unsigned long end);
2688 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2689 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2690 unsigned long pfn, unsigned long size, pgprot_t);
2691 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2692 int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
2693 struct page **pages, unsigned long *num);
2694 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
2696 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
2698 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2700 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2701 unsigned long pfn, pgprot_t pgprot);
2702 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2704 vm_fault_t vmf_insert_mixed_prot(struct vm_area_struct *vma, unsigned long addr,
2705 pfn_t pfn, pgprot_t pgprot);
2706 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
2707 unsigned long addr, pfn_t pfn);
2708 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2710 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2711 unsigned long addr, struct page *page)
2713 int err = vm_insert_page(vma, addr, page);
2716 return VM_FAULT_OOM;
2717 if (err < 0 && err != -EBUSY)
2718 return VM_FAULT_SIGBUS;
2720 return VM_FAULT_NOPAGE;
2723 #ifndef io_remap_pfn_range
2724 static inline int io_remap_pfn_range(struct vm_area_struct *vma,
2725 unsigned long addr, unsigned long pfn,
2726 unsigned long size, pgprot_t prot)
2728 return remap_pfn_range(vma, addr, pfn, size, pgprot_decrypted(prot));
2732 static inline vm_fault_t vmf_error(int err)
2735 return VM_FAULT_OOM;
2736 return VM_FAULT_SIGBUS;
2739 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
2740 unsigned int foll_flags);
2742 #define FOLL_WRITE 0x01 /* check pte is writable */
2743 #define FOLL_TOUCH 0x02 /* mark page accessed */
2744 #define FOLL_GET 0x04 /* do get_page on page */
2745 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2746 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2747 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2748 * and return without waiting upon it */
2749 #define FOLL_POPULATE 0x40 /* fault in page */
2750 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2751 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2752 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2753 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2754 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2755 #define FOLL_MLOCK 0x1000 /* lock present pages */
2756 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2757 #define FOLL_COW 0x4000 /* internal GUP flag */
2758 #define FOLL_ANON 0x8000 /* don't do file mappings */
2759 #define FOLL_LONGTERM 0x10000 /* mapping lifetime is indefinite: see below */
2760 #define FOLL_SPLIT_PMD 0x20000 /* split huge pmd before returning */
2761 #define FOLL_PIN 0x40000 /* pages must be released via unpin_user_page */
2762 #define FOLL_FAST_ONLY 0x80000 /* gup_fast: prevent fall-back to slow gup */
2765 * FOLL_PIN and FOLL_LONGTERM may be used in various combinations with each
2766 * other. Here is what they mean, and how to use them:
2768 * FOLL_LONGTERM indicates that the page will be held for an indefinite time
2769 * period _often_ under userspace control. This is in contrast to
2770 * iov_iter_get_pages(), whose usages are transient.
2772 * FIXME: For pages which are part of a filesystem, mappings are subject to the
2773 * lifetime enforced by the filesystem and we need guarantees that longterm
2774 * users like RDMA and V4L2 only establish mappings which coordinate usage with
2775 * the filesystem. Ideas for this coordination include revoking the longterm
2776 * pin, delaying writeback, bounce buffer page writeback, etc. As FS DAX was
2777 * added after the problem with filesystems was found FS DAX VMAs are
2778 * specifically failed. Filesystem pages are still subject to bugs and use of
2779 * FOLL_LONGTERM should be avoided on those pages.
2781 * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call.
2782 * Currently only get_user_pages() and get_user_pages_fast() support this flag
2783 * and calls to get_user_pages_[un]locked are specifically not allowed. This
2784 * is due to an incompatibility with the FS DAX check and
2785 * FAULT_FLAG_ALLOW_RETRY.
2787 * In the CMA case: long term pins in a CMA region would unnecessarily fragment
2788 * that region. And so, CMA attempts to migrate the page before pinning, when
2789 * FOLL_LONGTERM is specified.
2791 * FOLL_PIN indicates that a special kind of tracking (not just page->_refcount,
2792 * but an additional pin counting system) will be invoked. This is intended for
2793 * anything that gets a page reference and then touches page data (for example,
2794 * Direct IO). This lets the filesystem know that some non-file-system entity is
2795 * potentially changing the pages' data. In contrast to FOLL_GET (whose pages
2796 * are released via put_page()), FOLL_PIN pages must be released, ultimately, by
2797 * a call to unpin_user_page().
2799 * FOLL_PIN is similar to FOLL_GET: both of these pin pages. They use different
2800 * and separate refcounting mechanisms, however, and that means that each has
2801 * its own acquire and release mechanisms:
2803 * FOLL_GET: get_user_pages*() to acquire, and put_page() to release.
2805 * FOLL_PIN: pin_user_pages*() to acquire, and unpin_user_pages to release.
2807 * FOLL_PIN and FOLL_GET are mutually exclusive for a given function call.
2808 * (The underlying pages may experience both FOLL_GET-based and FOLL_PIN-based
2809 * calls applied to them, and that's perfectly OK. This is a constraint on the
2810 * callers, not on the pages.)
2812 * FOLL_PIN should be set internally by the pin_user_pages*() APIs, never
2813 * directly by the caller. That's in order to help avoid mismatches when
2814 * releasing pages: get_user_pages*() pages must be released via put_page(),
2815 * while pin_user_pages*() pages must be released via unpin_user_page().
2817 * Please see Documentation/core-api/pin_user_pages.rst for more information.
2820 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
2822 if (vm_fault & VM_FAULT_OOM)
2824 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2825 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2826 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2831 typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
2832 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2833 unsigned long size, pte_fn_t fn, void *data);
2834 extern int apply_to_existing_page_range(struct mm_struct *mm,
2835 unsigned long address, unsigned long size,
2836 pte_fn_t fn, void *data);
2838 extern void init_mem_debugging_and_hardening(void);
2839 #ifdef CONFIG_PAGE_POISONING
2840 extern void __kernel_poison_pages(struct page *page, int numpages);
2841 extern void __kernel_unpoison_pages(struct page *page, int numpages);
2842 extern bool _page_poisoning_enabled_early;
2843 DECLARE_STATIC_KEY_FALSE(_page_poisoning_enabled);
2844 static inline bool page_poisoning_enabled(void)
2846 return _page_poisoning_enabled_early;
2849 * For use in fast paths after init_mem_debugging() has run, or when a
2850 * false negative result is not harmful when called too early.
2852 static inline bool page_poisoning_enabled_static(void)
2854 return static_branch_unlikely(&_page_poisoning_enabled);
2856 static inline void kernel_poison_pages(struct page *page, int numpages)
2858 if (page_poisoning_enabled_static())
2859 __kernel_poison_pages(page, numpages);
2861 static inline void kernel_unpoison_pages(struct page *page, int numpages)
2863 if (page_poisoning_enabled_static())
2864 __kernel_unpoison_pages(page, numpages);
2867 static inline bool page_poisoning_enabled(void) { return false; }
2868 static inline bool page_poisoning_enabled_static(void) { return false; }
2869 static inline void __kernel_poison_pages(struct page *page, int nunmpages) { }
2870 static inline void kernel_poison_pages(struct page *page, int numpages) { }
2871 static inline void kernel_unpoison_pages(struct page *page, int numpages) { }
2874 DECLARE_STATIC_KEY_FALSE(init_on_alloc);
2875 static inline bool want_init_on_alloc(gfp_t flags)
2877 if (static_branch_unlikely(&init_on_alloc))
2879 return flags & __GFP_ZERO;
2882 DECLARE_STATIC_KEY_FALSE(init_on_free);
2883 static inline bool want_init_on_free(void)
2885 return static_branch_unlikely(&init_on_free);
2888 extern bool _debug_pagealloc_enabled_early;
2889 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
2891 static inline bool debug_pagealloc_enabled(void)
2893 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
2894 _debug_pagealloc_enabled_early;
2898 * For use in fast paths after init_debug_pagealloc() has run, or when a
2899 * false negative result is not harmful when called too early.
2901 static inline bool debug_pagealloc_enabled_static(void)
2903 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
2906 return static_branch_unlikely(&_debug_pagealloc_enabled);
2909 #ifdef CONFIG_DEBUG_PAGEALLOC
2911 * To support DEBUG_PAGEALLOC architecture must ensure that
2912 * __kernel_map_pages() never fails
2914 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2916 static inline void debug_pagealloc_map_pages(struct page *page, int numpages)
2918 if (debug_pagealloc_enabled_static())
2919 __kernel_map_pages(page, numpages, 1);
2922 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages)
2924 if (debug_pagealloc_enabled_static())
2925 __kernel_map_pages(page, numpages, 0);
2927 #else /* CONFIG_DEBUG_PAGEALLOC */
2928 static inline void debug_pagealloc_map_pages(struct page *page, int numpages) {}
2929 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages) {}
2930 #endif /* CONFIG_DEBUG_PAGEALLOC */
2932 #ifdef __HAVE_ARCH_GATE_AREA
2933 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2934 extern int in_gate_area_no_mm(unsigned long addr);
2935 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2937 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2941 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2942 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2946 #endif /* __HAVE_ARCH_GATE_AREA */
2948 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2950 #ifdef CONFIG_SYSCTL
2951 extern int sysctl_drop_caches;
2952 int drop_caches_sysctl_handler(struct ctl_table *, int, void *, size_t *,
2956 void drop_slab(void);
2957 void drop_slab_node(int nid);
2960 #define randomize_va_space 0
2962 extern int randomize_va_space;
2965 const char * arch_vma_name(struct vm_area_struct *vma);
2967 void print_vma_addr(char *prefix, unsigned long rip);
2969 static inline void print_vma_addr(char *prefix, unsigned long rip)
2974 void *sparse_buffer_alloc(unsigned long size);
2975 struct page * __populate_section_memmap(unsigned long pfn,
2976 unsigned long nr_pages, int nid, struct vmem_altmap *altmap);
2977 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2978 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
2979 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
2980 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2981 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node,
2982 struct vmem_altmap *altmap);
2983 void *vmemmap_alloc_block(unsigned long size, int node);
2985 void *vmemmap_alloc_block_buf(unsigned long size, int node,
2986 struct vmem_altmap *altmap);
2987 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2988 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2989 int node, struct vmem_altmap *altmap);
2990 int vmemmap_populate(unsigned long start, unsigned long end, int node,
2991 struct vmem_altmap *altmap);
2992 void vmemmap_populate_print_last(void);
2993 #ifdef CONFIG_MEMORY_HOTPLUG
2994 void vmemmap_free(unsigned long start, unsigned long end,
2995 struct vmem_altmap *altmap);
2997 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2998 unsigned long nr_pages);
3001 MF_COUNT_INCREASED = 1 << 0,
3002 MF_ACTION_REQUIRED = 1 << 1,
3003 MF_MUST_KILL = 1 << 2,
3004 MF_SOFT_OFFLINE = 1 << 3,
3006 extern int memory_failure(unsigned long pfn, int flags);
3007 extern void memory_failure_queue(unsigned long pfn, int flags);
3008 extern void memory_failure_queue_kick(int cpu);
3009 extern int unpoison_memory(unsigned long pfn);
3010 extern int sysctl_memory_failure_early_kill;
3011 extern int sysctl_memory_failure_recovery;
3012 extern void shake_page(struct page *p, int access);
3013 extern atomic_long_t num_poisoned_pages __read_mostly;
3014 extern int soft_offline_page(unsigned long pfn, int flags);
3018 * Error handlers for various types of pages.
3021 MF_IGNORED, /* Error: cannot be handled */
3022 MF_FAILED, /* Error: handling failed */
3023 MF_DELAYED, /* Will be handled later */
3024 MF_RECOVERED, /* Successfully recovered */
3027 enum mf_action_page_type {
3029 MF_MSG_KERNEL_HIGH_ORDER,
3031 MF_MSG_DIFFERENT_COMPOUND,
3032 MF_MSG_POISONED_HUGE,
3035 MF_MSG_NON_PMD_HUGE,
3036 MF_MSG_UNMAP_FAILED,
3037 MF_MSG_DIRTY_SWAPCACHE,
3038 MF_MSG_CLEAN_SWAPCACHE,
3039 MF_MSG_DIRTY_MLOCKED_LRU,
3040 MF_MSG_CLEAN_MLOCKED_LRU,
3041 MF_MSG_DIRTY_UNEVICTABLE_LRU,
3042 MF_MSG_CLEAN_UNEVICTABLE_LRU,
3045 MF_MSG_TRUNCATED_LRU,
3053 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
3054 extern void clear_huge_page(struct page *page,
3055 unsigned long addr_hint,
3056 unsigned int pages_per_huge_page);
3057 extern void copy_user_huge_page(struct page *dst, struct page *src,
3058 unsigned long addr_hint,
3059 struct vm_area_struct *vma,
3060 unsigned int pages_per_huge_page);
3061 extern long copy_huge_page_from_user(struct page *dst_page,
3062 const void __user *usr_src,
3063 unsigned int pages_per_huge_page,
3064 bool allow_pagefault);
3067 * vma_is_special_huge - Are transhuge page-table entries considered special?
3068 * @vma: Pointer to the struct vm_area_struct to consider
3070 * Whether transhuge page-table entries are considered "special" following
3071 * the definition in vm_normal_page().
3073 * Return: true if transhuge page-table entries should be considered special,
3076 static inline bool vma_is_special_huge(const struct vm_area_struct *vma)
3078 return vma_is_dax(vma) || (vma->vm_file &&
3079 (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP)));
3082 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
3084 #ifdef CONFIG_DEBUG_PAGEALLOC
3085 extern unsigned int _debug_guardpage_minorder;
3086 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
3088 static inline unsigned int debug_guardpage_minorder(void)
3090 return _debug_guardpage_minorder;
3093 static inline bool debug_guardpage_enabled(void)
3095 return static_branch_unlikely(&_debug_guardpage_enabled);
3098 static inline bool page_is_guard(struct page *page)
3100 if (!debug_guardpage_enabled())
3103 return PageGuard(page);
3106 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
3107 static inline bool debug_guardpage_enabled(void) { return false; }
3108 static inline bool page_is_guard(struct page *page) { return false; }
3109 #endif /* CONFIG_DEBUG_PAGEALLOC */
3111 #if MAX_NUMNODES > 1
3112 void __init setup_nr_node_ids(void);
3114 static inline void setup_nr_node_ids(void) {}
3117 extern int memcmp_pages(struct page *page1, struct page *page2);
3119 static inline int pages_identical(struct page *page1, struct page *page2)
3121 return !memcmp_pages(page1, page2);
3124 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
3125 unsigned long clean_record_shared_mapping_range(struct address_space *mapping,
3126 pgoff_t first_index, pgoff_t nr,
3127 pgoff_t bitmap_pgoff,
3128 unsigned long *bitmap,
3132 unsigned long wp_shared_mapping_range(struct address_space *mapping,
3133 pgoff_t first_index, pgoff_t nr);
3136 extern int sysctl_nr_trim_pages;
3138 void mem_dump_obj(void *object);
3140 #endif /* __KERNEL__ */
3141 #endif /* _LINUX_MM_H */