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.
517 struct vm_area_struct *vma; /* Target VMA */
518 unsigned int flags; /* FAULT_FLAG_xxx flags */
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 */
522 pmd_t *pmd; /* Pointer to pmd entry matching
524 pud_t *pud; /* Pointer to pud entry matching
527 pte_t orig_pte; /* Value of PTE at the time of fault */
529 struct page *cow_page; /* Page handler may use for COW fault */
530 struct page *page; /* ->fault handlers should return a
531 * page here, unless VM_FAULT_NOPAGE
532 * is set (which is also implied by
535 /* These three entries are valid only while holding ptl lock */
536 pte_t *pte; /* Pointer to pte entry matching
537 * the 'address'. NULL if the page
538 * table hasn't been allocated.
540 spinlock_t *ptl; /* Page table lock.
541 * Protects pte page table if 'pte'
542 * is not NULL, otherwise pmd.
544 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
545 * vm_ops->map_pages() sets up a page
546 * table from atomic context.
547 * do_fault_around() pre-allocates
548 * page table to avoid allocation from
553 /* page entry size for vm->huge_fault() */
554 enum page_entry_size {
561 * These are the virtual MM functions - opening of an area, closing and
562 * unmapping it (needed to keep files on disk up-to-date etc), pointer
563 * to the functions called when a no-page or a wp-page exception occurs.
565 struct vm_operations_struct {
566 void (*open)(struct vm_area_struct * area);
567 void (*close)(struct vm_area_struct * area);
568 /* Called any time before splitting to check if it's allowed */
569 int (*may_split)(struct vm_area_struct *area, unsigned long addr);
570 int (*mremap)(struct vm_area_struct *area, unsigned long flags);
572 * Called by mprotect() to make driver-specific permission
573 * checks before mprotect() is finalised. The VMA must not
574 * be modified. Returns 0 if eprotect() can proceed.
576 int (*mprotect)(struct vm_area_struct *vma, unsigned long start,
577 unsigned long end, unsigned long newflags);
578 vm_fault_t (*fault)(struct vm_fault *vmf);
579 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
580 enum page_entry_size pe_size);
581 vm_fault_t (*map_pages)(struct vm_fault *vmf,
582 pgoff_t start_pgoff, pgoff_t end_pgoff);
583 unsigned long (*pagesize)(struct vm_area_struct * area);
585 /* notification that a previously read-only page is about to become
586 * writable, if an error is returned it will cause a SIGBUS */
587 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
589 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
590 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
592 /* called by access_process_vm when get_user_pages() fails, typically
593 * for use by special VMAs that can switch between memory and hardware
595 int (*access)(struct vm_area_struct *vma, unsigned long addr,
596 void *buf, int len, int write);
598 /* Called by the /proc/PID/maps code to ask the vma whether it
599 * has a special name. Returning non-NULL will also cause this
600 * vma to be dumped unconditionally. */
601 const char *(*name)(struct vm_area_struct *vma);
605 * set_policy() op must add a reference to any non-NULL @new mempolicy
606 * to hold the policy upon return. Caller should pass NULL @new to
607 * remove a policy and fall back to surrounding context--i.e. do not
608 * install a MPOL_DEFAULT policy, nor the task or system default
611 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
614 * get_policy() op must add reference [mpol_get()] to any policy at
615 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
616 * in mm/mempolicy.c will do this automatically.
617 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
618 * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
619 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
620 * must return NULL--i.e., do not "fallback" to task or system default
623 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
627 * Called by vm_normal_page() for special PTEs to find the
628 * page for @addr. This is useful if the default behavior
629 * (using pte_page()) would not find the correct page.
631 struct page *(*find_special_page)(struct vm_area_struct *vma,
635 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
637 static const struct vm_operations_struct dummy_vm_ops = {};
639 memset(vma, 0, sizeof(*vma));
641 vma->vm_ops = &dummy_vm_ops;
642 INIT_LIST_HEAD(&vma->anon_vma_chain);
645 static inline void vma_set_anonymous(struct vm_area_struct *vma)
650 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
655 static inline bool vma_is_temporary_stack(struct vm_area_struct *vma)
657 int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP);
662 if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) ==
663 VM_STACK_INCOMPLETE_SETUP)
669 static inline bool vma_is_foreign(struct vm_area_struct *vma)
674 if (current->mm != vma->vm_mm)
680 static inline bool vma_is_accessible(struct vm_area_struct *vma)
682 return vma->vm_flags & VM_ACCESS_FLAGS;
687 * The vma_is_shmem is not inline because it is used only by slow
688 * paths in userfault.
690 bool vma_is_shmem(struct vm_area_struct *vma);
692 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
695 int vma_is_stack_for_current(struct vm_area_struct *vma);
697 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
698 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
703 #include <linux/huge_mm.h>
706 * Methods to modify the page usage count.
708 * What counts for a page usage:
709 * - cache mapping (page->mapping)
710 * - private data (page->private)
711 * - page mapped in a task's page tables, each mapping
712 * is counted separately
714 * Also, many kernel routines increase the page count before a critical
715 * routine so they can be sure the page doesn't go away from under them.
719 * Drop a ref, return true if the refcount fell to zero (the page has no users)
721 static inline int put_page_testzero(struct page *page)
723 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
724 return page_ref_dec_and_test(page);
728 * Try to grab a ref unless the page has a refcount of zero, return false if
730 * This can be called when MMU is off so it must not access
731 * any of the virtual mappings.
733 static inline int get_page_unless_zero(struct page *page)
735 return page_ref_add_unless(page, 1, 0);
738 extern int page_is_ram(unsigned long pfn);
746 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
749 /* Support for virtually mapped pages */
750 struct page *vmalloc_to_page(const void *addr);
751 unsigned long vmalloc_to_pfn(const void *addr);
754 * Determine if an address is within the vmalloc range
756 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
757 * is no special casing required.
760 #ifndef is_ioremap_addr
761 #define is_ioremap_addr(x) is_vmalloc_addr(x)
765 extern bool is_vmalloc_addr(const void *x);
766 extern int is_vmalloc_or_module_addr(const void *x);
768 static inline bool is_vmalloc_addr(const void *x)
772 static inline int is_vmalloc_or_module_addr(const void *x)
778 extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
779 static inline void *kvmalloc(size_t size, gfp_t flags)
781 return kvmalloc_node(size, flags, NUMA_NO_NODE);
783 static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
785 return kvmalloc_node(size, flags | __GFP_ZERO, node);
787 static inline void *kvzalloc(size_t size, gfp_t flags)
789 return kvmalloc(size, flags | __GFP_ZERO);
792 static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
796 if (unlikely(check_mul_overflow(n, size, &bytes)))
799 return kvmalloc(bytes, flags);
802 static inline void *kvcalloc(size_t n, size_t size, gfp_t flags)
804 return kvmalloc_array(n, size, flags | __GFP_ZERO);
807 extern void kvfree(const void *addr);
808 extern void kvfree_sensitive(const void *addr, size_t len);
810 static inline int head_compound_mapcount(struct page *head)
812 return atomic_read(compound_mapcount_ptr(head)) + 1;
816 * Mapcount of compound page as a whole, does not include mapped sub-pages.
818 * Must be called only for compound pages or any their tail sub-pages.
820 static inline int compound_mapcount(struct page *page)
822 VM_BUG_ON_PAGE(!PageCompound(page), page);
823 page = compound_head(page);
824 return head_compound_mapcount(page);
828 * The atomic page->_mapcount, starts from -1: so that transitions
829 * both from it and to it can be tracked, using atomic_inc_and_test
830 * and atomic_add_negative(-1).
832 static inline void page_mapcount_reset(struct page *page)
834 atomic_set(&(page)->_mapcount, -1);
837 int __page_mapcount(struct page *page);
840 * Mapcount of 0-order page; when compound sub-page, includes
841 * compound_mapcount().
843 * Result is undefined for pages which cannot be mapped into userspace.
844 * For example SLAB or special types of pages. See function page_has_type().
845 * They use this place in struct page differently.
847 static inline int page_mapcount(struct page *page)
849 if (unlikely(PageCompound(page)))
850 return __page_mapcount(page);
851 return atomic_read(&page->_mapcount) + 1;
854 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
855 int total_mapcount(struct page *page);
856 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
858 static inline int total_mapcount(struct page *page)
860 return page_mapcount(page);
862 static inline int page_trans_huge_mapcount(struct page *page,
865 int mapcount = page_mapcount(page);
867 *total_mapcount = mapcount;
872 static inline struct page *virt_to_head_page(const void *x)
874 struct page *page = virt_to_page(x);
876 return compound_head(page);
879 void __put_page(struct page *page);
881 void put_pages_list(struct list_head *pages);
883 void split_page(struct page *page, unsigned int order);
886 * Compound pages have a destructor function. Provide a
887 * prototype for that function and accessor functions.
888 * These are _only_ valid on the head of a compound page.
890 typedef void compound_page_dtor(struct page *);
892 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
893 enum compound_dtor_id {
896 #ifdef CONFIG_HUGETLB_PAGE
899 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
904 extern compound_page_dtor * const compound_page_dtors[NR_COMPOUND_DTORS];
906 static inline void set_compound_page_dtor(struct page *page,
907 enum compound_dtor_id compound_dtor)
909 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
910 page[1].compound_dtor = compound_dtor;
913 static inline void destroy_compound_page(struct page *page)
915 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
916 compound_page_dtors[page[1].compound_dtor](page);
919 static inline unsigned int compound_order(struct page *page)
923 return page[1].compound_order;
926 static inline bool hpage_pincount_available(struct page *page)
929 * Can the page->hpage_pinned_refcount field be used? That field is in
930 * the 3rd page of the compound page, so the smallest (2-page) compound
931 * pages cannot support it.
933 page = compound_head(page);
934 return PageCompound(page) && compound_order(page) > 1;
937 static inline int head_compound_pincount(struct page *head)
939 return atomic_read(compound_pincount_ptr(head));
942 static inline int compound_pincount(struct page *page)
944 VM_BUG_ON_PAGE(!hpage_pincount_available(page), page);
945 page = compound_head(page);
946 return head_compound_pincount(page);
949 static inline void set_compound_order(struct page *page, unsigned int order)
951 page[1].compound_order = order;
952 page[1].compound_nr = 1U << order;
955 /* Returns the number of pages in this potentially compound page. */
956 static inline unsigned long compound_nr(struct page *page)
960 return page[1].compound_nr;
963 /* Returns the number of bytes in this potentially compound page. */
964 static inline unsigned long page_size(struct page *page)
966 return PAGE_SIZE << compound_order(page);
969 /* Returns the number of bits needed for the number of bytes in a page */
970 static inline unsigned int page_shift(struct page *page)
972 return PAGE_SHIFT + compound_order(page);
975 void free_compound_page(struct page *page);
979 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
980 * servicing faults for write access. In the normal case, do always want
981 * pte_mkwrite. But get_user_pages can cause write faults for mappings
982 * that do not have writing enabled, when used by access_process_vm.
984 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
986 if (likely(vma->vm_flags & VM_WRITE))
987 pte = pte_mkwrite(pte);
991 vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page);
992 void do_set_pte(struct vm_fault *vmf, struct page *page);
994 vm_fault_t finish_fault(struct vm_fault *vmf);
995 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
999 * Multiple processes may "see" the same page. E.g. for untouched
1000 * mappings of /dev/null, all processes see the same page full of
1001 * zeroes, and text pages of executables and shared libraries have
1002 * only one copy in memory, at most, normally.
1004 * For the non-reserved pages, page_count(page) denotes a reference count.
1005 * page_count() == 0 means the page is free. page->lru is then used for
1006 * freelist management in the buddy allocator.
1007 * page_count() > 0 means the page has been allocated.
1009 * Pages are allocated by the slab allocator in order to provide memory
1010 * to kmalloc and kmem_cache_alloc. In this case, the management of the
1011 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
1012 * unless a particular usage is carefully commented. (the responsibility of
1013 * freeing the kmalloc memory is the caller's, of course).
1015 * A page may be used by anyone else who does a __get_free_page().
1016 * In this case, page_count still tracks the references, and should only
1017 * be used through the normal accessor functions. The top bits of page->flags
1018 * and page->virtual store page management information, but all other fields
1019 * are unused and could be used privately, carefully. The management of this
1020 * page is the responsibility of the one who allocated it, and those who have
1021 * subsequently been given references to it.
1023 * The other pages (we may call them "pagecache pages") are completely
1024 * managed by the Linux memory manager: I/O, buffers, swapping etc.
1025 * The following discussion applies only to them.
1027 * A pagecache page contains an opaque `private' member, which belongs to the
1028 * page's address_space. Usually, this is the address of a circular list of
1029 * the page's disk buffers. PG_private must be set to tell the VM to call
1030 * into the filesystem to release these pages.
1032 * A page may belong to an inode's memory mapping. In this case, page->mapping
1033 * is the pointer to the inode, and page->index is the file offset of the page,
1034 * in units of PAGE_SIZE.
1036 * If pagecache pages are not associated with an inode, they are said to be
1037 * anonymous pages. These may become associated with the swapcache, and in that
1038 * case PG_swapcache is set, and page->private is an offset into the swapcache.
1040 * In either case (swapcache or inode backed), the pagecache itself holds one
1041 * reference to the page. Setting PG_private should also increment the
1042 * refcount. The each user mapping also has a reference to the page.
1044 * The pagecache pages are stored in a per-mapping radix tree, which is
1045 * rooted at mapping->i_pages, and indexed by offset.
1046 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
1047 * lists, we instead now tag pages as dirty/writeback in the radix tree.
1049 * All pagecache pages may be subject to I/O:
1050 * - inode pages may need to be read from disk,
1051 * - inode pages which have been modified and are MAP_SHARED may need
1052 * to be written back to the inode on disk,
1053 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
1054 * modified may need to be swapped out to swap space and (later) to be read
1059 * The zone field is never updated after free_area_init_core()
1060 * sets it, so none of the operations on it need to be atomic.
1063 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
1064 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
1065 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
1066 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
1067 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
1068 #define KASAN_TAG_PGOFF (LAST_CPUPID_PGOFF - KASAN_TAG_WIDTH)
1071 * Define the bit shifts to access each section. For non-existent
1072 * sections we define the shift as 0; that plus a 0 mask ensures
1073 * the compiler will optimise away reference to them.
1075 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
1076 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
1077 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
1078 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
1079 #define KASAN_TAG_PGSHIFT (KASAN_TAG_PGOFF * (KASAN_TAG_WIDTH != 0))
1081 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
1082 #ifdef NODE_NOT_IN_PAGE_FLAGS
1083 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
1084 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
1085 SECTIONS_PGOFF : ZONES_PGOFF)
1087 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
1088 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
1089 NODES_PGOFF : ZONES_PGOFF)
1092 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
1094 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
1095 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
1096 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
1097 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
1098 #define KASAN_TAG_MASK ((1UL << KASAN_TAG_WIDTH) - 1)
1099 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
1101 static inline enum zone_type page_zonenum(const struct page *page)
1103 ASSERT_EXCLUSIVE_BITS(page->flags, ZONES_MASK << ZONES_PGSHIFT);
1104 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
1107 #ifdef CONFIG_ZONE_DEVICE
1108 static inline bool is_zone_device_page(const struct page *page)
1110 return page_zonenum(page) == ZONE_DEVICE;
1112 extern void memmap_init_zone_device(struct zone *, unsigned long,
1113 unsigned long, struct dev_pagemap *);
1115 static inline bool is_zone_device_page(const struct page *page)
1121 #ifdef CONFIG_DEV_PAGEMAP_OPS
1122 void free_devmap_managed_page(struct page *page);
1123 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
1125 static inline bool page_is_devmap_managed(struct page *page)
1127 if (!static_branch_unlikely(&devmap_managed_key))
1129 if (!is_zone_device_page(page))
1131 switch (page->pgmap->type) {
1132 case MEMORY_DEVICE_PRIVATE:
1133 case MEMORY_DEVICE_FS_DAX:
1141 void put_devmap_managed_page(struct page *page);
1143 #else /* CONFIG_DEV_PAGEMAP_OPS */
1144 static inline bool page_is_devmap_managed(struct page *page)
1149 static inline void put_devmap_managed_page(struct page *page)
1152 #endif /* CONFIG_DEV_PAGEMAP_OPS */
1154 static inline bool is_device_private_page(const struct page *page)
1156 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
1157 IS_ENABLED(CONFIG_DEVICE_PRIVATE) &&
1158 is_zone_device_page(page) &&
1159 page->pgmap->type == MEMORY_DEVICE_PRIVATE;
1162 static inline bool is_pci_p2pdma_page(const struct page *page)
1164 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
1165 IS_ENABLED(CONFIG_PCI_P2PDMA) &&
1166 is_zone_device_page(page) &&
1167 page->pgmap->type == MEMORY_DEVICE_PCI_P2PDMA;
1170 /* 127: arbitrary random number, small enough to assemble well */
1171 #define page_ref_zero_or_close_to_overflow(page) \
1172 ((unsigned int) page_ref_count(page) + 127u <= 127u)
1174 static inline void get_page(struct page *page)
1176 page = compound_head(page);
1178 * Getting a normal page or the head of a compound page
1179 * requires to already have an elevated page->_refcount.
1181 VM_BUG_ON_PAGE(page_ref_zero_or_close_to_overflow(page), page);
1185 bool __must_check try_grab_page(struct page *page, unsigned int flags);
1187 static inline __must_check bool try_get_page(struct page *page)
1189 page = compound_head(page);
1190 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1196 static inline void put_page(struct page *page)
1198 page = compound_head(page);
1201 * For devmap managed pages we need to catch refcount transition from
1202 * 2 to 1, when refcount reach one it means the page is free and we
1203 * need to inform the device driver through callback. See
1204 * include/linux/memremap.h and HMM for details.
1206 if (page_is_devmap_managed(page)) {
1207 put_devmap_managed_page(page);
1211 if (put_page_testzero(page))
1216 * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
1217 * the page's refcount so that two separate items are tracked: the original page
1218 * reference count, and also a new count of how many pin_user_pages() calls were
1219 * made against the page. ("gup-pinned" is another term for the latter).
1221 * With this scheme, pin_user_pages() becomes special: such pages are marked as
1222 * distinct from normal pages. As such, the unpin_user_page() call (and its
1223 * variants) must be used in order to release gup-pinned pages.
1227 * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
1228 * counts with respect to pin_user_pages() and unpin_user_page() becomes
1229 * simpler, due to the fact that adding an even power of two to the page
1230 * refcount has the effect of using only the upper N bits, for the code that
1231 * counts up using the bias value. This means that the lower bits are left for
1232 * the exclusive use of the original code that increments and decrements by one
1233 * (or at least, by much smaller values than the bias value).
1235 * Of course, once the lower bits overflow into the upper bits (and this is
1236 * OK, because subtraction recovers the original values), then visual inspection
1237 * no longer suffices to directly view the separate counts. However, for normal
1238 * applications that don't have huge page reference counts, this won't be an
1241 * Locking: the lockless algorithm described in page_cache_get_speculative()
1242 * and page_cache_gup_pin_speculative() provides safe operation for
1243 * get_user_pages and page_mkclean and other calls that race to set up page
1246 #define GUP_PIN_COUNTING_BIAS (1U << 10)
1248 void unpin_user_page(struct page *page);
1249 void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1251 void unpin_user_pages(struct page **pages, unsigned long npages);
1254 * page_maybe_dma_pinned() - report if a page is pinned for DMA.
1256 * This function checks if a page has been pinned via a call to
1257 * pin_user_pages*().
1259 * For non-huge pages, the return value is partially fuzzy: false is not fuzzy,
1260 * because it means "definitely not pinned for DMA", but true means "probably
1261 * pinned for DMA, but possibly a false positive due to having at least
1262 * GUP_PIN_COUNTING_BIAS worth of normal page references".
1264 * False positives are OK, because: a) it's unlikely for a page to get that many
1265 * refcounts, and b) all the callers of this routine are expected to be able to
1266 * deal gracefully with a false positive.
1268 * For huge pages, the result will be exactly correct. That's because we have
1269 * more tracking data available: the 3rd struct page in the compound page is
1270 * used to track the pincount (instead using of the GUP_PIN_COUNTING_BIAS
1273 * For more information, please see Documentation/core-api/pin_user_pages.rst.
1275 * @page: pointer to page to be queried.
1276 * @Return: True, if it is likely that the page has been "dma-pinned".
1277 * False, if the page is definitely not dma-pinned.
1279 static inline bool page_maybe_dma_pinned(struct page *page)
1281 if (hpage_pincount_available(page))
1282 return compound_pincount(page) > 0;
1285 * page_ref_count() is signed. If that refcount overflows, then
1286 * page_ref_count() returns a negative value, and callers will avoid
1287 * further incrementing the refcount.
1289 * Here, for that overflow case, use the signed bit to count a little
1290 * bit higher via unsigned math, and thus still get an accurate result.
1292 return ((unsigned int)page_ref_count(compound_head(page))) >=
1293 GUP_PIN_COUNTING_BIAS;
1296 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1297 #define SECTION_IN_PAGE_FLAGS
1301 * The identification function is mainly used by the buddy allocator for
1302 * determining if two pages could be buddies. We are not really identifying
1303 * the zone since we could be using the section number id if we do not have
1304 * node id available in page flags.
1305 * We only guarantee that it will return the same value for two combinable
1308 static inline int page_zone_id(struct page *page)
1310 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1313 #ifdef NODE_NOT_IN_PAGE_FLAGS
1314 extern int page_to_nid(const struct page *page);
1316 static inline int page_to_nid(const struct page *page)
1318 struct page *p = (struct page *)page;
1320 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1324 #ifdef CONFIG_NUMA_BALANCING
1325 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1327 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1330 static inline int cpupid_to_pid(int cpupid)
1332 return cpupid & LAST__PID_MASK;
1335 static inline int cpupid_to_cpu(int cpupid)
1337 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1340 static inline int cpupid_to_nid(int cpupid)
1342 return cpu_to_node(cpupid_to_cpu(cpupid));
1345 static inline bool cpupid_pid_unset(int cpupid)
1347 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1350 static inline bool cpupid_cpu_unset(int cpupid)
1352 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1355 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1357 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1360 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1361 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1362 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1364 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1367 static inline int page_cpupid_last(struct page *page)
1369 return page->_last_cpupid;
1371 static inline void page_cpupid_reset_last(struct page *page)
1373 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1376 static inline int page_cpupid_last(struct page *page)
1378 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1381 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1383 static inline void page_cpupid_reset_last(struct page *page)
1385 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1387 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1388 #else /* !CONFIG_NUMA_BALANCING */
1389 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1391 return page_to_nid(page); /* XXX */
1394 static inline int page_cpupid_last(struct page *page)
1396 return page_to_nid(page); /* XXX */
1399 static inline int cpupid_to_nid(int cpupid)
1404 static inline int cpupid_to_pid(int cpupid)
1409 static inline int cpupid_to_cpu(int cpupid)
1414 static inline int cpu_pid_to_cpupid(int nid, int pid)
1419 static inline bool cpupid_pid_unset(int cpupid)
1424 static inline void page_cpupid_reset_last(struct page *page)
1428 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1432 #endif /* CONFIG_NUMA_BALANCING */
1434 #if defined(CONFIG_KASAN_SW_TAGS) || defined(CONFIG_KASAN_HW_TAGS)
1436 static inline u8 page_kasan_tag(const struct page *page)
1438 if (kasan_enabled())
1439 return (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1443 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1445 if (kasan_enabled()) {
1446 page->flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1447 page->flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1451 static inline void page_kasan_tag_reset(struct page *page)
1453 if (kasan_enabled())
1454 page_kasan_tag_set(page, 0xff);
1457 #else /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1459 static inline u8 page_kasan_tag(const struct page *page)
1464 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1465 static inline void page_kasan_tag_reset(struct page *page) { }
1467 #endif /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1469 static inline struct zone *page_zone(const struct page *page)
1471 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1474 static inline pg_data_t *page_pgdat(const struct page *page)
1476 return NODE_DATA(page_to_nid(page));
1479 #ifdef SECTION_IN_PAGE_FLAGS
1480 static inline void set_page_section(struct page *page, unsigned long section)
1482 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1483 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1486 static inline unsigned long page_to_section(const struct page *page)
1488 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1492 static inline void set_page_zone(struct page *page, enum zone_type zone)
1494 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1495 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1498 static inline void set_page_node(struct page *page, unsigned long node)
1500 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1501 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1504 static inline void set_page_links(struct page *page, enum zone_type zone,
1505 unsigned long node, unsigned long pfn)
1507 set_page_zone(page, zone);
1508 set_page_node(page, node);
1509 #ifdef SECTION_IN_PAGE_FLAGS
1510 set_page_section(page, pfn_to_section_nr(pfn));
1515 * Some inline functions in vmstat.h depend on page_zone()
1517 #include <linux/vmstat.h>
1519 static __always_inline void *lowmem_page_address(const struct page *page)
1521 return page_to_virt(page);
1524 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1525 #define HASHED_PAGE_VIRTUAL
1528 #if defined(WANT_PAGE_VIRTUAL)
1529 static inline void *page_address(const struct page *page)
1531 return page->virtual;
1533 static inline void set_page_address(struct page *page, void *address)
1535 page->virtual = address;
1537 #define page_address_init() do { } while(0)
1540 #if defined(HASHED_PAGE_VIRTUAL)
1541 void *page_address(const struct page *page);
1542 void set_page_address(struct page *page, void *virtual);
1543 void page_address_init(void);
1546 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1547 #define page_address(page) lowmem_page_address(page)
1548 #define set_page_address(page, address) do { } while(0)
1549 #define page_address_init() do { } while(0)
1552 extern void *page_rmapping(struct page *page);
1553 extern struct anon_vma *page_anon_vma(struct page *page);
1554 extern struct address_space *page_mapping(struct page *page);
1556 extern struct address_space *__page_file_mapping(struct page *);
1559 struct address_space *page_file_mapping(struct page *page)
1561 if (unlikely(PageSwapCache(page)))
1562 return __page_file_mapping(page);
1564 return page->mapping;
1567 extern pgoff_t __page_file_index(struct page *page);
1570 * Return the pagecache index of the passed page. Regular pagecache pages
1571 * use ->index whereas swapcache pages use swp_offset(->private)
1573 static inline pgoff_t page_index(struct page *page)
1575 if (unlikely(PageSwapCache(page)))
1576 return __page_file_index(page);
1580 bool page_mapped(struct page *page);
1581 struct address_space *page_mapping(struct page *page);
1582 struct address_space *page_mapping_file(struct page *page);
1585 * Return true only if the page has been allocated with
1586 * ALLOC_NO_WATERMARKS and the low watermark was not
1587 * met implying that the system is under some pressure.
1589 static inline bool page_is_pfmemalloc(struct page *page)
1592 * Page index cannot be this large so this must be
1593 * a pfmemalloc page.
1595 return page->index == -1UL;
1599 * Only to be called by the page allocator on a freshly allocated
1602 static inline void set_page_pfmemalloc(struct page *page)
1607 static inline void clear_page_pfmemalloc(struct page *page)
1613 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1615 extern void pagefault_out_of_memory(void);
1617 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1618 #define offset_in_thp(page, p) ((unsigned long)(p) & (thp_size(page) - 1))
1621 * Flags passed to show_mem() and show_free_areas() to suppress output in
1624 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1626 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1629 extern bool can_do_mlock(void);
1631 static inline bool can_do_mlock(void) { return false; }
1633 extern int user_shm_lock(size_t, struct user_struct *);
1634 extern void user_shm_unlock(size_t, struct user_struct *);
1637 * Parameter block passed down to zap_pte_range in exceptional cases.
1639 struct zap_details {
1640 struct address_space *check_mapping; /* Check page->mapping if set */
1641 pgoff_t first_index; /* Lowest page->index to unmap */
1642 pgoff_t last_index; /* Highest page->index to unmap */
1645 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1647 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1650 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1651 unsigned long size);
1652 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1653 unsigned long size);
1654 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1655 unsigned long start, unsigned long end);
1657 struct mmu_notifier_range;
1659 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1660 unsigned long end, unsigned long floor, unsigned long ceiling);
1662 copy_page_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma);
1663 int follow_pte(struct mm_struct *mm, unsigned long address,
1664 struct mmu_notifier_range *range, pte_t **ptepp, pmd_t **pmdpp,
1666 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1667 unsigned long *pfn);
1668 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1669 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1670 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1671 void *buf, int len, int write);
1673 extern void truncate_pagecache(struct inode *inode, loff_t new);
1674 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1675 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1676 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1677 int truncate_inode_page(struct address_space *mapping, struct page *page);
1678 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1679 int invalidate_inode_page(struct page *page);
1682 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1683 unsigned long address, unsigned int flags,
1684 struct pt_regs *regs);
1685 extern int fixup_user_fault(struct mm_struct *mm,
1686 unsigned long address, unsigned int fault_flags,
1688 void unmap_mapping_pages(struct address_space *mapping,
1689 pgoff_t start, pgoff_t nr, bool even_cows);
1690 void unmap_mapping_range(struct address_space *mapping,
1691 loff_t const holebegin, loff_t const holelen, int even_cows);
1693 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1694 unsigned long address, unsigned int flags,
1695 struct pt_regs *regs)
1697 /* should never happen if there's no MMU */
1699 return VM_FAULT_SIGBUS;
1701 static inline int fixup_user_fault(struct mm_struct *mm, unsigned long address,
1702 unsigned int fault_flags, bool *unlocked)
1704 /* should never happen if there's no MMU */
1708 static inline void unmap_mapping_pages(struct address_space *mapping,
1709 pgoff_t start, pgoff_t nr, bool even_cows) { }
1710 static inline void unmap_mapping_range(struct address_space *mapping,
1711 loff_t const holebegin, loff_t const holelen, int even_cows) { }
1714 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1715 loff_t const holebegin, loff_t const holelen)
1717 unmap_mapping_range(mapping, holebegin, holelen, 0);
1720 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1721 void *buf, int len, unsigned int gup_flags);
1722 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1723 void *buf, int len, unsigned int gup_flags);
1724 extern int __access_remote_vm(struct mm_struct *mm, unsigned long addr,
1725 void *buf, int len, unsigned int gup_flags);
1727 long get_user_pages_remote(struct mm_struct *mm,
1728 unsigned long start, unsigned long nr_pages,
1729 unsigned int gup_flags, struct page **pages,
1730 struct vm_area_struct **vmas, int *locked);
1731 long pin_user_pages_remote(struct mm_struct *mm,
1732 unsigned long start, unsigned long nr_pages,
1733 unsigned int gup_flags, struct page **pages,
1734 struct vm_area_struct **vmas, int *locked);
1735 long get_user_pages(unsigned long start, unsigned long nr_pages,
1736 unsigned int gup_flags, struct page **pages,
1737 struct vm_area_struct **vmas);
1738 long pin_user_pages(unsigned long start, unsigned long nr_pages,
1739 unsigned int gup_flags, struct page **pages,
1740 struct vm_area_struct **vmas);
1741 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1742 unsigned int gup_flags, struct page **pages, int *locked);
1743 long pin_user_pages_locked(unsigned long start, unsigned long nr_pages,
1744 unsigned int gup_flags, struct page **pages, int *locked);
1745 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1746 struct page **pages, unsigned int gup_flags);
1747 long pin_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1748 struct page **pages, unsigned int gup_flags);
1750 int get_user_pages_fast(unsigned long start, int nr_pages,
1751 unsigned int gup_flags, struct page **pages);
1752 int pin_user_pages_fast(unsigned long start, int nr_pages,
1753 unsigned int gup_flags, struct page **pages);
1755 int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
1756 int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
1757 struct task_struct *task, bool bypass_rlim);
1759 /* Container for pinned pfns / pages */
1760 struct frame_vector {
1761 unsigned int nr_allocated; /* Number of frames we have space for */
1762 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1763 bool got_ref; /* Did we pin pages by getting page ref? */
1764 bool is_pfns; /* Does array contain pages or pfns? */
1765 void *ptrs[]; /* Array of pinned pfns / pages. Use
1766 * pfns_vector_pages() or pfns_vector_pfns()
1770 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1771 void frame_vector_destroy(struct frame_vector *vec);
1772 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1773 unsigned int gup_flags, struct frame_vector *vec);
1774 void put_vaddr_frames(struct frame_vector *vec);
1775 int frame_vector_to_pages(struct frame_vector *vec);
1776 void frame_vector_to_pfns(struct frame_vector *vec);
1778 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1780 return vec->nr_frames;
1783 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1786 int err = frame_vector_to_pages(vec);
1789 return ERR_PTR(err);
1791 return (struct page **)(vec->ptrs);
1794 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1797 frame_vector_to_pfns(vec);
1798 return (unsigned long *)(vec->ptrs);
1802 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1803 struct page **pages);
1804 int get_kernel_page(unsigned long start, int write, struct page **pages);
1805 struct page *get_dump_page(unsigned long addr);
1807 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1808 extern void do_invalidatepage(struct page *page, unsigned int offset,
1809 unsigned int length);
1811 void __set_page_dirty(struct page *, struct address_space *, int warn);
1812 int __set_page_dirty_nobuffers(struct page *page);
1813 int __set_page_dirty_no_writeback(struct page *page);
1814 int redirty_page_for_writepage(struct writeback_control *wbc,
1816 void account_page_dirtied(struct page *page, struct address_space *mapping);
1817 void account_page_cleaned(struct page *page, struct address_space *mapping,
1818 struct bdi_writeback *wb);
1819 int set_page_dirty(struct page *page);
1820 int set_page_dirty_lock(struct page *page);
1821 void __cancel_dirty_page(struct page *page);
1822 static inline void cancel_dirty_page(struct page *page)
1824 /* Avoid atomic ops, locking, etc. when not actually needed. */
1825 if (PageDirty(page))
1826 __cancel_dirty_page(page);
1828 int clear_page_dirty_for_io(struct page *page);
1830 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1832 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1833 unsigned long old_addr, struct vm_area_struct *new_vma,
1834 unsigned long new_addr, unsigned long len,
1835 bool need_rmap_locks);
1838 * Flags used by change_protection(). For now we make it a bitmap so
1839 * that we can pass in multiple flags just like parameters. However
1840 * for now all the callers are only use one of the flags at the same
1843 /* Whether we should allow dirty bit accounting */
1844 #define MM_CP_DIRTY_ACCT (1UL << 0)
1845 /* Whether this protection change is for NUMA hints */
1846 #define MM_CP_PROT_NUMA (1UL << 1)
1847 /* Whether this change is for write protecting */
1848 #define MM_CP_UFFD_WP (1UL << 2) /* do wp */
1849 #define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */
1850 #define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \
1851 MM_CP_UFFD_WP_RESOLVE)
1853 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1854 unsigned long end, pgprot_t newprot,
1855 unsigned long cp_flags);
1856 extern int mprotect_fixup(struct vm_area_struct *vma,
1857 struct vm_area_struct **pprev, unsigned long start,
1858 unsigned long end, unsigned long newflags);
1861 * doesn't attempt to fault and will return short.
1863 int get_user_pages_fast_only(unsigned long start, int nr_pages,
1864 unsigned int gup_flags, struct page **pages);
1865 int pin_user_pages_fast_only(unsigned long start, int nr_pages,
1866 unsigned int gup_flags, struct page **pages);
1868 static inline bool get_user_page_fast_only(unsigned long addr,
1869 unsigned int gup_flags, struct page **pagep)
1871 return get_user_pages_fast_only(addr, 1, gup_flags, pagep) == 1;
1874 * per-process(per-mm_struct) statistics.
1876 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1878 long val = atomic_long_read(&mm->rss_stat.count[member]);
1880 #ifdef SPLIT_RSS_COUNTING
1882 * counter is updated in asynchronous manner and may go to minus.
1883 * But it's never be expected number for users.
1888 return (unsigned long)val;
1891 void mm_trace_rss_stat(struct mm_struct *mm, int member, long count);
1893 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1895 long count = atomic_long_add_return(value, &mm->rss_stat.count[member]);
1897 mm_trace_rss_stat(mm, member, count);
1900 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1902 long count = atomic_long_inc_return(&mm->rss_stat.count[member]);
1904 mm_trace_rss_stat(mm, member, count);
1907 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1909 long count = atomic_long_dec_return(&mm->rss_stat.count[member]);
1911 mm_trace_rss_stat(mm, member, count);
1914 /* Optimized variant when page is already known not to be PageAnon */
1915 static inline int mm_counter_file(struct page *page)
1917 if (PageSwapBacked(page))
1918 return MM_SHMEMPAGES;
1919 return MM_FILEPAGES;
1922 static inline int mm_counter(struct page *page)
1925 return MM_ANONPAGES;
1926 return mm_counter_file(page);
1929 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1931 return get_mm_counter(mm, MM_FILEPAGES) +
1932 get_mm_counter(mm, MM_ANONPAGES) +
1933 get_mm_counter(mm, MM_SHMEMPAGES);
1936 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1938 return max(mm->hiwater_rss, get_mm_rss(mm));
1941 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1943 return max(mm->hiwater_vm, mm->total_vm);
1946 static inline void update_hiwater_rss(struct mm_struct *mm)
1948 unsigned long _rss = get_mm_rss(mm);
1950 if ((mm)->hiwater_rss < _rss)
1951 (mm)->hiwater_rss = _rss;
1954 static inline void update_hiwater_vm(struct mm_struct *mm)
1956 if (mm->hiwater_vm < mm->total_vm)
1957 mm->hiwater_vm = mm->total_vm;
1960 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1962 mm->hiwater_rss = get_mm_rss(mm);
1965 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1966 struct mm_struct *mm)
1968 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1970 if (*maxrss < hiwater_rss)
1971 *maxrss = hiwater_rss;
1974 #if defined(SPLIT_RSS_COUNTING)
1975 void sync_mm_rss(struct mm_struct *mm);
1977 static inline void sync_mm_rss(struct mm_struct *mm)
1982 #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
1983 static inline int pte_special(pte_t pte)
1988 static inline pte_t pte_mkspecial(pte_t pte)
1994 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
1995 static inline int pte_devmap(pte_t pte)
2001 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
2003 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
2005 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
2009 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
2013 #ifdef __PAGETABLE_P4D_FOLDED
2014 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2015 unsigned long address)
2020 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
2023 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
2024 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2025 unsigned long address)
2029 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
2030 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
2033 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
2035 static inline void mm_inc_nr_puds(struct mm_struct *mm)
2037 if (mm_pud_folded(mm))
2039 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2042 static inline void mm_dec_nr_puds(struct mm_struct *mm)
2044 if (mm_pud_folded(mm))
2046 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2050 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
2051 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
2052 unsigned long address)
2057 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
2058 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
2061 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
2063 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
2065 if (mm_pmd_folded(mm))
2067 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2070 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
2072 if (mm_pmd_folded(mm))
2074 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2079 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
2081 atomic_long_set(&mm->pgtables_bytes, 0);
2084 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2086 return atomic_long_read(&mm->pgtables_bytes);
2089 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
2091 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2094 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
2096 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2100 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
2101 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2106 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
2107 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
2110 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
2111 int __pte_alloc_kernel(pmd_t *pmd);
2113 #if defined(CONFIG_MMU)
2115 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2116 unsigned long address)
2118 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
2119 NULL : p4d_offset(pgd, address);
2122 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2123 unsigned long address)
2125 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
2126 NULL : pud_offset(p4d, address);
2129 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
2131 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
2132 NULL: pmd_offset(pud, address);
2134 #endif /* CONFIG_MMU */
2136 #if USE_SPLIT_PTE_PTLOCKS
2137 #if ALLOC_SPLIT_PTLOCKS
2138 void __init ptlock_cache_init(void);
2139 extern bool ptlock_alloc(struct page *page);
2140 extern void ptlock_free(struct page *page);
2142 static inline spinlock_t *ptlock_ptr(struct page *page)
2146 #else /* ALLOC_SPLIT_PTLOCKS */
2147 static inline void ptlock_cache_init(void)
2151 static inline bool ptlock_alloc(struct page *page)
2156 static inline void ptlock_free(struct page *page)
2160 static inline spinlock_t *ptlock_ptr(struct page *page)
2164 #endif /* ALLOC_SPLIT_PTLOCKS */
2166 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2168 return ptlock_ptr(pmd_page(*pmd));
2171 static inline bool ptlock_init(struct page *page)
2174 * prep_new_page() initialize page->private (and therefore page->ptl)
2175 * with 0. Make sure nobody took it in use in between.
2177 * It can happen if arch try to use slab for page table allocation:
2178 * slab code uses page->slab_cache, which share storage with page->ptl.
2180 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
2181 if (!ptlock_alloc(page))
2183 spin_lock_init(ptlock_ptr(page));
2187 #else /* !USE_SPLIT_PTE_PTLOCKS */
2189 * We use mm->page_table_lock to guard all pagetable pages of the mm.
2191 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2193 return &mm->page_table_lock;
2195 static inline void ptlock_cache_init(void) {}
2196 static inline bool ptlock_init(struct page *page) { return true; }
2197 static inline void ptlock_free(struct page *page) {}
2198 #endif /* USE_SPLIT_PTE_PTLOCKS */
2200 static inline void pgtable_init(void)
2202 ptlock_cache_init();
2203 pgtable_cache_init();
2206 static inline bool pgtable_pte_page_ctor(struct page *page)
2208 if (!ptlock_init(page))
2210 __SetPageTable(page);
2211 inc_lruvec_page_state(page, NR_PAGETABLE);
2215 static inline void pgtable_pte_page_dtor(struct page *page)
2218 __ClearPageTable(page);
2219 dec_lruvec_page_state(page, NR_PAGETABLE);
2222 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
2224 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
2225 pte_t *__pte = pte_offset_map(pmd, address); \
2231 #define pte_unmap_unlock(pte, ptl) do { \
2236 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2238 #define pte_alloc_map(mm, pmd, address) \
2239 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2241 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2242 (pte_alloc(mm, pmd) ? \
2243 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2245 #define pte_alloc_kernel(pmd, address) \
2246 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2247 NULL: pte_offset_kernel(pmd, address))
2249 #if USE_SPLIT_PMD_PTLOCKS
2251 static struct page *pmd_to_page(pmd_t *pmd)
2253 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2254 return virt_to_page((void *)((unsigned long) pmd & mask));
2257 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2259 return ptlock_ptr(pmd_to_page(pmd));
2262 static inline bool pmd_ptlock_init(struct page *page)
2264 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2265 page->pmd_huge_pte = NULL;
2267 return ptlock_init(page);
2270 static inline void pmd_ptlock_free(struct page *page)
2272 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2273 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2278 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2282 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2284 return &mm->page_table_lock;
2287 static inline bool pmd_ptlock_init(struct page *page) { return true; }
2288 static inline void pmd_ptlock_free(struct page *page) {}
2290 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2294 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2296 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2301 static inline bool pgtable_pmd_page_ctor(struct page *page)
2303 if (!pmd_ptlock_init(page))
2305 __SetPageTable(page);
2306 inc_lruvec_page_state(page, NR_PAGETABLE);
2310 static inline void pgtable_pmd_page_dtor(struct page *page)
2312 pmd_ptlock_free(page);
2313 __ClearPageTable(page);
2314 dec_lruvec_page_state(page, NR_PAGETABLE);
2318 * No scalability reason to split PUD locks yet, but follow the same pattern
2319 * as the PMD locks to make it easier if we decide to. The VM should not be
2320 * considered ready to switch to split PUD locks yet; there may be places
2321 * which need to be converted from page_table_lock.
2323 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2325 return &mm->page_table_lock;
2328 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2330 spinlock_t *ptl = pud_lockptr(mm, pud);
2336 extern void __init pagecache_init(void);
2337 extern void __init free_area_init_memoryless_node(int nid);
2338 extern void free_initmem(void);
2341 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2342 * into the buddy system. The freed pages will be poisoned with pattern
2343 * "poison" if it's within range [0, UCHAR_MAX].
2344 * Return pages freed into the buddy system.
2346 extern unsigned long free_reserved_area(void *start, void *end,
2347 int poison, const char *s);
2349 #ifdef CONFIG_HIGHMEM
2351 * Free a highmem page into the buddy system, adjusting totalhigh_pages
2352 * and totalram_pages.
2354 extern void free_highmem_page(struct page *page);
2357 extern void adjust_managed_page_count(struct page *page, long count);
2358 extern void mem_init_print_info(const char *str);
2360 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2362 /* Free the reserved page into the buddy system, so it gets managed. */
2363 static inline void __free_reserved_page(struct page *page)
2365 ClearPageReserved(page);
2366 init_page_count(page);
2370 static inline void free_reserved_page(struct page *page)
2372 __free_reserved_page(page);
2373 adjust_managed_page_count(page, 1);
2376 static inline void mark_page_reserved(struct page *page)
2378 SetPageReserved(page);
2379 adjust_managed_page_count(page, -1);
2383 * Default method to free all the __init memory into the buddy system.
2384 * The freed pages will be poisoned with pattern "poison" if it's within
2385 * range [0, UCHAR_MAX].
2386 * Return pages freed into the buddy system.
2388 static inline unsigned long free_initmem_default(int poison)
2390 extern char __init_begin[], __init_end[];
2392 return free_reserved_area(&__init_begin, &__init_end,
2393 poison, "unused kernel");
2396 static inline unsigned long get_num_physpages(void)
2399 unsigned long phys_pages = 0;
2401 for_each_online_node(nid)
2402 phys_pages += node_present_pages(nid);
2408 * Using memblock node mappings, an architecture may initialise its
2409 * zones, allocate the backing mem_map and account for memory holes in an
2410 * architecture independent manner.
2412 * An architecture is expected to register range of page frames backed by
2413 * physical memory with memblock_add[_node]() before calling
2414 * free_area_init() passing in the PFN each zone ends at. At a basic
2415 * usage, an architecture is expected to do something like
2417 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2419 * for_each_valid_physical_page_range()
2420 * memblock_add_node(base, size, nid)
2421 * free_area_init(max_zone_pfns);
2423 void free_area_init(unsigned long *max_zone_pfn);
2424 unsigned long node_map_pfn_alignment(void);
2425 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2426 unsigned long end_pfn);
2427 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2428 unsigned long end_pfn);
2429 extern void get_pfn_range_for_nid(unsigned int nid,
2430 unsigned long *start_pfn, unsigned long *end_pfn);
2431 extern unsigned long find_min_pfn_with_active_regions(void);
2433 #ifndef CONFIG_NEED_MULTIPLE_NODES
2434 static inline int early_pfn_to_nid(unsigned long pfn)
2439 /* please see mm/page_alloc.c */
2440 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2443 extern void set_dma_reserve(unsigned long new_dma_reserve);
2444 extern void memmap_init_zone(unsigned long, int, unsigned long,
2445 unsigned long, unsigned long, enum meminit_context,
2446 struct vmem_altmap *, int migratetype);
2447 extern void setup_per_zone_wmarks(void);
2448 extern int __meminit init_per_zone_wmark_min(void);
2449 extern void mem_init(void);
2450 extern void __init mmap_init(void);
2451 extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2452 extern long si_mem_available(void);
2453 extern void si_meminfo(struct sysinfo * val);
2454 extern void si_meminfo_node(struct sysinfo *val, int nid);
2455 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2456 extern unsigned long arch_reserved_kernel_pages(void);
2459 extern __printf(3, 4)
2460 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2462 extern void setup_per_cpu_pageset(void);
2465 extern int min_free_kbytes;
2466 extern int watermark_boost_factor;
2467 extern int watermark_scale_factor;
2468 extern bool arch_has_descending_max_zone_pfns(void);
2471 extern atomic_long_t mmap_pages_allocated;
2472 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2474 /* interval_tree.c */
2475 void vma_interval_tree_insert(struct vm_area_struct *node,
2476 struct rb_root_cached *root);
2477 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2478 struct vm_area_struct *prev,
2479 struct rb_root_cached *root);
2480 void vma_interval_tree_remove(struct vm_area_struct *node,
2481 struct rb_root_cached *root);
2482 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2483 unsigned long start, unsigned long last);
2484 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2485 unsigned long start, unsigned long last);
2487 #define vma_interval_tree_foreach(vma, root, start, last) \
2488 for (vma = vma_interval_tree_iter_first(root, start, last); \
2489 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2491 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2492 struct rb_root_cached *root);
2493 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2494 struct rb_root_cached *root);
2495 struct anon_vma_chain *
2496 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2497 unsigned long start, unsigned long last);
2498 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2499 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2500 #ifdef CONFIG_DEBUG_VM_RB
2501 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2504 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2505 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2506 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2509 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2510 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2511 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2512 struct vm_area_struct *expand);
2513 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2514 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2516 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2518 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2519 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2520 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2521 struct mempolicy *, struct vm_userfaultfd_ctx);
2522 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2523 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2524 unsigned long addr, int new_below);
2525 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2526 unsigned long addr, int new_below);
2527 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2528 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2529 struct rb_node **, struct rb_node *);
2530 extern void unlink_file_vma(struct vm_area_struct *);
2531 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2532 unsigned long addr, unsigned long len, pgoff_t pgoff,
2533 bool *need_rmap_locks);
2534 extern void exit_mmap(struct mm_struct *);
2536 static inline int check_data_rlimit(unsigned long rlim,
2538 unsigned long start,
2539 unsigned long end_data,
2540 unsigned long start_data)
2542 if (rlim < RLIM_INFINITY) {
2543 if (((new - start) + (end_data - start_data)) > rlim)
2550 extern int mm_take_all_locks(struct mm_struct *mm);
2551 extern void mm_drop_all_locks(struct mm_struct *mm);
2553 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2554 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2555 extern struct file *get_task_exe_file(struct task_struct *task);
2557 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2558 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2560 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2561 const struct vm_special_mapping *sm);
2562 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2563 unsigned long addr, unsigned long len,
2564 unsigned long flags,
2565 const struct vm_special_mapping *spec);
2566 /* This is an obsolete alternative to _install_special_mapping. */
2567 extern int install_special_mapping(struct mm_struct *mm,
2568 unsigned long addr, unsigned long len,
2569 unsigned long flags, struct page **pages);
2571 unsigned long randomize_stack_top(unsigned long stack_top);
2573 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2575 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2576 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2577 struct list_head *uf);
2578 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2579 unsigned long len, unsigned long prot, unsigned long flags,
2580 unsigned long pgoff, unsigned long *populate, struct list_head *uf);
2581 extern int __do_munmap(struct mm_struct *, unsigned long, size_t,
2582 struct list_head *uf, bool downgrade);
2583 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2584 struct list_head *uf);
2585 extern int do_madvise(struct mm_struct *mm, unsigned long start, size_t len_in, int behavior);
2588 extern int __mm_populate(unsigned long addr, unsigned long len,
2590 static inline void mm_populate(unsigned long addr, unsigned long len)
2593 (void) __mm_populate(addr, len, 1);
2596 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2599 /* These take the mm semaphore themselves */
2600 extern int __must_check vm_brk(unsigned long, unsigned long);
2601 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2602 extern int vm_munmap(unsigned long, size_t);
2603 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2604 unsigned long, unsigned long,
2605 unsigned long, unsigned long);
2607 struct vm_unmapped_area_info {
2608 #define VM_UNMAPPED_AREA_TOPDOWN 1
2609 unsigned long flags;
2610 unsigned long length;
2611 unsigned long low_limit;
2612 unsigned long high_limit;
2613 unsigned long align_mask;
2614 unsigned long align_offset;
2617 extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info);
2620 extern void truncate_inode_pages(struct address_space *, loff_t);
2621 extern void truncate_inode_pages_range(struct address_space *,
2622 loff_t lstart, loff_t lend);
2623 extern void truncate_inode_pages_final(struct address_space *);
2625 /* generic vm_area_ops exported for stackable file systems */
2626 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2627 extern vm_fault_t filemap_map_pages(struct vm_fault *vmf,
2628 pgoff_t start_pgoff, pgoff_t end_pgoff);
2629 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2631 /* mm/page-writeback.c */
2632 int __must_check write_one_page(struct page *page);
2633 void task_dirty_inc(struct task_struct *tsk);
2635 extern unsigned long stack_guard_gap;
2636 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2637 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2639 /* CONFIG_STACK_GROWSUP still needs to grow downwards at some places */
2640 extern int expand_downwards(struct vm_area_struct *vma,
2641 unsigned long address);
2643 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2645 #define expand_upwards(vma, address) (0)
2648 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2649 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2650 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2651 struct vm_area_struct **pprev);
2653 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2654 NULL if none. Assume start_addr < end_addr. */
2655 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2657 struct vm_area_struct * vma = find_vma(mm,start_addr);
2659 if (vma && end_addr <= vma->vm_start)
2664 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2666 unsigned long vm_start = vma->vm_start;
2668 if (vma->vm_flags & VM_GROWSDOWN) {
2669 vm_start -= stack_guard_gap;
2670 if (vm_start > vma->vm_start)
2676 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2678 unsigned long vm_end = vma->vm_end;
2680 if (vma->vm_flags & VM_GROWSUP) {
2681 vm_end += stack_guard_gap;
2682 if (vm_end < vma->vm_end)
2683 vm_end = -PAGE_SIZE;
2688 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2690 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2693 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2694 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2695 unsigned long vm_start, unsigned long vm_end)
2697 struct vm_area_struct *vma = find_vma(mm, vm_start);
2699 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2705 static inline bool range_in_vma(struct vm_area_struct *vma,
2706 unsigned long start, unsigned long end)
2708 return (vma && vma->vm_start <= start && end <= vma->vm_end);
2712 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2713 void vma_set_page_prot(struct vm_area_struct *vma);
2715 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2719 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2721 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2725 void vma_set_file(struct vm_area_struct *vma, struct file *file);
2727 #ifdef CONFIG_NUMA_BALANCING
2728 unsigned long change_prot_numa(struct vm_area_struct *vma,
2729 unsigned long start, unsigned long end);
2732 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2733 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2734 unsigned long pfn, unsigned long size, pgprot_t);
2735 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2736 int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
2737 struct page **pages, unsigned long *num);
2738 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
2740 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
2742 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2744 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2745 unsigned long pfn, pgprot_t pgprot);
2746 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2748 vm_fault_t vmf_insert_mixed_prot(struct vm_area_struct *vma, unsigned long addr,
2749 pfn_t pfn, pgprot_t pgprot);
2750 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
2751 unsigned long addr, pfn_t pfn);
2752 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2754 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2755 unsigned long addr, struct page *page)
2757 int err = vm_insert_page(vma, addr, page);
2760 return VM_FAULT_OOM;
2761 if (err < 0 && err != -EBUSY)
2762 return VM_FAULT_SIGBUS;
2764 return VM_FAULT_NOPAGE;
2767 #ifndef io_remap_pfn_range
2768 static inline int io_remap_pfn_range(struct vm_area_struct *vma,
2769 unsigned long addr, unsigned long pfn,
2770 unsigned long size, pgprot_t prot)
2772 return remap_pfn_range(vma, addr, pfn, size, pgprot_decrypted(prot));
2776 static inline vm_fault_t vmf_error(int err)
2779 return VM_FAULT_OOM;
2780 return VM_FAULT_SIGBUS;
2783 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
2784 unsigned int foll_flags);
2786 #define FOLL_WRITE 0x01 /* check pte is writable */
2787 #define FOLL_TOUCH 0x02 /* mark page accessed */
2788 #define FOLL_GET 0x04 /* do get_page on page */
2789 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2790 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2791 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2792 * and return without waiting upon it */
2793 #define FOLL_POPULATE 0x40 /* fault in page */
2794 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2795 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2796 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2797 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2798 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2799 #define FOLL_MLOCK 0x1000 /* lock present pages */
2800 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2801 #define FOLL_COW 0x4000 /* internal GUP flag */
2802 #define FOLL_ANON 0x8000 /* don't do file mappings */
2803 #define FOLL_LONGTERM 0x10000 /* mapping lifetime is indefinite: see below */
2804 #define FOLL_SPLIT_PMD 0x20000 /* split huge pmd before returning */
2805 #define FOLL_PIN 0x40000 /* pages must be released via unpin_user_page */
2806 #define FOLL_FAST_ONLY 0x80000 /* gup_fast: prevent fall-back to slow gup */
2809 * FOLL_PIN and FOLL_LONGTERM may be used in various combinations with each
2810 * other. Here is what they mean, and how to use them:
2812 * FOLL_LONGTERM indicates that the page will be held for an indefinite time
2813 * period _often_ under userspace control. This is in contrast to
2814 * iov_iter_get_pages(), whose usages are transient.
2816 * FIXME: For pages which are part of a filesystem, mappings are subject to the
2817 * lifetime enforced by the filesystem and we need guarantees that longterm
2818 * users like RDMA and V4L2 only establish mappings which coordinate usage with
2819 * the filesystem. Ideas for this coordination include revoking the longterm
2820 * pin, delaying writeback, bounce buffer page writeback, etc. As FS DAX was
2821 * added after the problem with filesystems was found FS DAX VMAs are
2822 * specifically failed. Filesystem pages are still subject to bugs and use of
2823 * FOLL_LONGTERM should be avoided on those pages.
2825 * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call.
2826 * Currently only get_user_pages() and get_user_pages_fast() support this flag
2827 * and calls to get_user_pages_[un]locked are specifically not allowed. This
2828 * is due to an incompatibility with the FS DAX check and
2829 * FAULT_FLAG_ALLOW_RETRY.
2831 * In the CMA case: long term pins in a CMA region would unnecessarily fragment
2832 * that region. And so, CMA attempts to migrate the page before pinning, when
2833 * FOLL_LONGTERM is specified.
2835 * FOLL_PIN indicates that a special kind of tracking (not just page->_refcount,
2836 * but an additional pin counting system) will be invoked. This is intended for
2837 * anything that gets a page reference and then touches page data (for example,
2838 * Direct IO). This lets the filesystem know that some non-file-system entity is
2839 * potentially changing the pages' data. In contrast to FOLL_GET (whose pages
2840 * are released via put_page()), FOLL_PIN pages must be released, ultimately, by
2841 * a call to unpin_user_page().
2843 * FOLL_PIN is similar to FOLL_GET: both of these pin pages. They use different
2844 * and separate refcounting mechanisms, however, and that means that each has
2845 * its own acquire and release mechanisms:
2847 * FOLL_GET: get_user_pages*() to acquire, and put_page() to release.
2849 * FOLL_PIN: pin_user_pages*() to acquire, and unpin_user_pages to release.
2851 * FOLL_PIN and FOLL_GET are mutually exclusive for a given function call.
2852 * (The underlying pages may experience both FOLL_GET-based and FOLL_PIN-based
2853 * calls applied to them, and that's perfectly OK. This is a constraint on the
2854 * callers, not on the pages.)
2856 * FOLL_PIN should be set internally by the pin_user_pages*() APIs, never
2857 * directly by the caller. That's in order to help avoid mismatches when
2858 * releasing pages: get_user_pages*() pages must be released via put_page(),
2859 * while pin_user_pages*() pages must be released via unpin_user_page().
2861 * Please see Documentation/core-api/pin_user_pages.rst for more information.
2864 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
2866 if (vm_fault & VM_FAULT_OOM)
2868 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2869 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2870 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2875 typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
2876 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2877 unsigned long size, pte_fn_t fn, void *data);
2878 extern int apply_to_existing_page_range(struct mm_struct *mm,
2879 unsigned long address, unsigned long size,
2880 pte_fn_t fn, void *data);
2882 extern void init_mem_debugging_and_hardening(void);
2883 #ifdef CONFIG_PAGE_POISONING
2884 extern void __kernel_poison_pages(struct page *page, int numpages);
2885 extern void __kernel_unpoison_pages(struct page *page, int numpages);
2886 extern bool _page_poisoning_enabled_early;
2887 DECLARE_STATIC_KEY_FALSE(_page_poisoning_enabled);
2888 static inline bool page_poisoning_enabled(void)
2890 return _page_poisoning_enabled_early;
2893 * For use in fast paths after init_mem_debugging() has run, or when a
2894 * false negative result is not harmful when called too early.
2896 static inline bool page_poisoning_enabled_static(void)
2898 return static_branch_unlikely(&_page_poisoning_enabled);
2900 static inline void kernel_poison_pages(struct page *page, int numpages)
2902 if (page_poisoning_enabled_static())
2903 __kernel_poison_pages(page, numpages);
2905 static inline void kernel_unpoison_pages(struct page *page, int numpages)
2907 if (page_poisoning_enabled_static())
2908 __kernel_unpoison_pages(page, numpages);
2911 static inline bool page_poisoning_enabled(void) { return false; }
2912 static inline bool page_poisoning_enabled_static(void) { return false; }
2913 static inline void __kernel_poison_pages(struct page *page, int nunmpages) { }
2914 static inline void kernel_poison_pages(struct page *page, int numpages) { }
2915 static inline void kernel_unpoison_pages(struct page *page, int numpages) { }
2918 DECLARE_STATIC_KEY_FALSE(init_on_alloc);
2919 static inline bool want_init_on_alloc(gfp_t flags)
2921 if (static_branch_unlikely(&init_on_alloc))
2923 return flags & __GFP_ZERO;
2926 DECLARE_STATIC_KEY_FALSE(init_on_free);
2927 static inline bool want_init_on_free(void)
2929 return static_branch_unlikely(&init_on_free);
2932 extern bool _debug_pagealloc_enabled_early;
2933 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
2935 static inline bool debug_pagealloc_enabled(void)
2937 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
2938 _debug_pagealloc_enabled_early;
2942 * For use in fast paths after init_debug_pagealloc() has run, or when a
2943 * false negative result is not harmful when called too early.
2945 static inline bool debug_pagealloc_enabled_static(void)
2947 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
2950 return static_branch_unlikely(&_debug_pagealloc_enabled);
2953 #ifdef CONFIG_DEBUG_PAGEALLOC
2955 * To support DEBUG_PAGEALLOC architecture must ensure that
2956 * __kernel_map_pages() never fails
2958 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2960 static inline void debug_pagealloc_map_pages(struct page *page, int numpages)
2962 if (debug_pagealloc_enabled_static())
2963 __kernel_map_pages(page, numpages, 1);
2966 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages)
2968 if (debug_pagealloc_enabled_static())
2969 __kernel_map_pages(page, numpages, 0);
2971 #else /* CONFIG_DEBUG_PAGEALLOC */
2972 static inline void debug_pagealloc_map_pages(struct page *page, int numpages) {}
2973 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages) {}
2974 #endif /* CONFIG_DEBUG_PAGEALLOC */
2976 #ifdef __HAVE_ARCH_GATE_AREA
2977 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2978 extern int in_gate_area_no_mm(unsigned long addr);
2979 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2981 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2985 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2986 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2990 #endif /* __HAVE_ARCH_GATE_AREA */
2992 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2994 #ifdef CONFIG_SYSCTL
2995 extern int sysctl_drop_caches;
2996 int drop_caches_sysctl_handler(struct ctl_table *, int, void *, size_t *,
3000 void drop_slab(void);
3001 void drop_slab_node(int nid);
3004 #define randomize_va_space 0
3006 extern int randomize_va_space;
3009 const char * arch_vma_name(struct vm_area_struct *vma);
3011 void print_vma_addr(char *prefix, unsigned long rip);
3013 static inline void print_vma_addr(char *prefix, unsigned long rip)
3018 void *sparse_buffer_alloc(unsigned long size);
3019 struct page * __populate_section_memmap(unsigned long pfn,
3020 unsigned long nr_pages, int nid, struct vmem_altmap *altmap);
3021 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
3022 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
3023 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
3024 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
3025 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node,
3026 struct vmem_altmap *altmap);
3027 void *vmemmap_alloc_block(unsigned long size, int node);
3029 void *vmemmap_alloc_block_buf(unsigned long size, int node,
3030 struct vmem_altmap *altmap);
3031 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
3032 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
3033 int node, struct vmem_altmap *altmap);
3034 int vmemmap_populate(unsigned long start, unsigned long end, int node,
3035 struct vmem_altmap *altmap);
3036 void vmemmap_populate_print_last(void);
3037 #ifdef CONFIG_MEMORY_HOTPLUG
3038 void vmemmap_free(unsigned long start, unsigned long end,
3039 struct vmem_altmap *altmap);
3041 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
3042 unsigned long nr_pages);
3045 MF_COUNT_INCREASED = 1 << 0,
3046 MF_ACTION_REQUIRED = 1 << 1,
3047 MF_MUST_KILL = 1 << 2,
3048 MF_SOFT_OFFLINE = 1 << 3,
3050 extern int memory_failure(unsigned long pfn, int flags);
3051 extern void memory_failure_queue(unsigned long pfn, int flags);
3052 extern void memory_failure_queue_kick(int cpu);
3053 extern int unpoison_memory(unsigned long pfn);
3054 extern int sysctl_memory_failure_early_kill;
3055 extern int sysctl_memory_failure_recovery;
3056 extern void shake_page(struct page *p, int access);
3057 extern atomic_long_t num_poisoned_pages __read_mostly;
3058 extern int soft_offline_page(unsigned long pfn, int flags);
3062 * Error handlers for various types of pages.
3065 MF_IGNORED, /* Error: cannot be handled */
3066 MF_FAILED, /* Error: handling failed */
3067 MF_DELAYED, /* Will be handled later */
3068 MF_RECOVERED, /* Successfully recovered */
3071 enum mf_action_page_type {
3073 MF_MSG_KERNEL_HIGH_ORDER,
3075 MF_MSG_DIFFERENT_COMPOUND,
3076 MF_MSG_POISONED_HUGE,
3079 MF_MSG_NON_PMD_HUGE,
3080 MF_MSG_UNMAP_FAILED,
3081 MF_MSG_DIRTY_SWAPCACHE,
3082 MF_MSG_CLEAN_SWAPCACHE,
3083 MF_MSG_DIRTY_MLOCKED_LRU,
3084 MF_MSG_CLEAN_MLOCKED_LRU,
3085 MF_MSG_DIRTY_UNEVICTABLE_LRU,
3086 MF_MSG_CLEAN_UNEVICTABLE_LRU,
3089 MF_MSG_TRUNCATED_LRU,
3097 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
3098 extern void clear_huge_page(struct page *page,
3099 unsigned long addr_hint,
3100 unsigned int pages_per_huge_page);
3101 extern void copy_user_huge_page(struct page *dst, struct page *src,
3102 unsigned long addr_hint,
3103 struct vm_area_struct *vma,
3104 unsigned int pages_per_huge_page);
3105 extern long copy_huge_page_from_user(struct page *dst_page,
3106 const void __user *usr_src,
3107 unsigned int pages_per_huge_page,
3108 bool allow_pagefault);
3111 * vma_is_special_huge - Are transhuge page-table entries considered special?
3112 * @vma: Pointer to the struct vm_area_struct to consider
3114 * Whether transhuge page-table entries are considered "special" following
3115 * the definition in vm_normal_page().
3117 * Return: true if transhuge page-table entries should be considered special,
3120 static inline bool vma_is_special_huge(const struct vm_area_struct *vma)
3122 return vma_is_dax(vma) || (vma->vm_file &&
3123 (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP)));
3126 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
3128 #ifdef CONFIG_DEBUG_PAGEALLOC
3129 extern unsigned int _debug_guardpage_minorder;
3130 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
3132 static inline unsigned int debug_guardpage_minorder(void)
3134 return _debug_guardpage_minorder;
3137 static inline bool debug_guardpage_enabled(void)
3139 return static_branch_unlikely(&_debug_guardpage_enabled);
3142 static inline bool page_is_guard(struct page *page)
3144 if (!debug_guardpage_enabled())
3147 return PageGuard(page);
3150 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
3151 static inline bool debug_guardpage_enabled(void) { return false; }
3152 static inline bool page_is_guard(struct page *page) { return false; }
3153 #endif /* CONFIG_DEBUG_PAGEALLOC */
3155 #if MAX_NUMNODES > 1
3156 void __init setup_nr_node_ids(void);
3158 static inline void setup_nr_node_ids(void) {}
3161 extern int memcmp_pages(struct page *page1, struct page *page2);
3163 static inline int pages_identical(struct page *page1, struct page *page2)
3165 return !memcmp_pages(page1, page2);
3168 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
3169 unsigned long clean_record_shared_mapping_range(struct address_space *mapping,
3170 pgoff_t first_index, pgoff_t nr,
3171 pgoff_t bitmap_pgoff,
3172 unsigned long *bitmap,
3176 unsigned long wp_shared_mapping_range(struct address_space *mapping,
3177 pgoff_t first_index, pgoff_t nr);
3180 extern int sysctl_nr_trim_pages;
3182 #endif /* __KERNEL__ */
3183 #endif /* _LINUX_MM_H */