1 /* SPDX-License-Identifier: GPL-2.0 */
5 #include <linux/errno.h>
9 #include <linux/mmdebug.h>
10 #include <linux/gfp.h>
11 #include <linux/bug.h>
12 #include <linux/list.h>
13 #include <linux/mmzone.h>
14 #include <linux/rbtree.h>
15 #include <linux/atomic.h>
16 #include <linux/debug_locks.h>
17 #include <linux/mm_types.h>
18 #include <linux/range.h>
19 #include <linux/pfn.h>
20 #include <linux/percpu-refcount.h>
21 #include <linux/bit_spinlock.h>
22 #include <linux/shrinker.h>
23 #include <linux/resource.h>
24 #include <linux/page_ext.h>
25 #include <linux/err.h>
26 #include <linux/page_ref.h>
27 #include <linux/memremap.h>
28 #include <linux/overflow.h>
29 #include <linux/sizes.h>
33 struct anon_vma_chain;
36 struct writeback_control;
39 void init_mm_internals(void);
41 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
42 extern unsigned long max_mapnr;
44 static inline void set_max_mapnr(unsigned long limit)
49 static inline void set_max_mapnr(unsigned long limit) { }
52 extern atomic_long_t _totalram_pages;
53 static inline unsigned long totalram_pages(void)
55 return (unsigned long)atomic_long_read(&_totalram_pages);
58 static inline void totalram_pages_inc(void)
60 atomic_long_inc(&_totalram_pages);
63 static inline void totalram_pages_dec(void)
65 atomic_long_dec(&_totalram_pages);
68 static inline void totalram_pages_add(long count)
70 atomic_long_add(count, &_totalram_pages);
73 extern void * high_memory;
74 extern int page_cluster;
77 extern int sysctl_legacy_va_layout;
79 #define sysctl_legacy_va_layout 0
82 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
83 extern const int mmap_rnd_bits_min;
84 extern const int mmap_rnd_bits_max;
85 extern int mmap_rnd_bits __read_mostly;
87 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
88 extern const int mmap_rnd_compat_bits_min;
89 extern const int mmap_rnd_compat_bits_max;
90 extern int mmap_rnd_compat_bits __read_mostly;
94 #include <asm/pgtable.h>
95 #include <asm/processor.h>
98 * Architectures that support memory tagging (assigning tags to memory regions,
99 * embedding these tags into addresses that point to these memory regions, and
100 * checking that the memory and the pointer tags match on memory accesses)
101 * redefine this macro to strip tags from pointers.
102 * It's defined as noop for arcitectures that don't support memory tagging.
104 #ifndef untagged_addr
105 #define untagged_addr(addr) (addr)
109 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
113 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
117 #define lm_alias(x) __va(__pa_symbol(x))
121 * To prevent common memory management code establishing
122 * a zero page mapping on a read fault.
123 * This macro should be defined within <asm/pgtable.h>.
124 * s390 does this to prevent multiplexing of hardware bits
125 * related to the physical page in case of virtualization.
127 #ifndef mm_forbids_zeropage
128 #define mm_forbids_zeropage(X) (0)
132 * On some architectures it is expensive to call memset() for small sizes.
133 * If an architecture decides to implement their own version of
134 * mm_zero_struct_page they should wrap the defines below in a #ifndef and
135 * define their own version of this macro in <asm/pgtable.h>
137 #if BITS_PER_LONG == 64
138 /* This function must be updated when the size of struct page grows above 80
139 * or reduces below 56. The idea that compiler optimizes out switch()
140 * statement, and only leaves move/store instructions. Also the compiler can
141 * combine write statments if they are both assignments and can be reordered,
142 * this can result in several of the writes here being dropped.
144 #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
145 static inline void __mm_zero_struct_page(struct page *page)
147 unsigned long *_pp = (void *)page;
149 /* Check that struct page is either 56, 64, 72, or 80 bytes */
150 BUILD_BUG_ON(sizeof(struct page) & 7);
151 BUILD_BUG_ON(sizeof(struct page) < 56);
152 BUILD_BUG_ON(sizeof(struct page) > 80);
154 switch (sizeof(struct page)) {
156 _pp[9] = 0; /* fallthrough */
158 _pp[8] = 0; /* fallthrough */
160 _pp[7] = 0; /* fallthrough */
172 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
176 * Default maximum number of active map areas, this limits the number of vmas
177 * per mm struct. Users can overwrite this number by sysctl but there is a
180 * When a program's coredump is generated as ELF format, a section is created
181 * per a vma. In ELF, the number of sections is represented in unsigned short.
182 * This means the number of sections should be smaller than 65535 at coredump.
183 * Because the kernel adds some informative sections to a image of program at
184 * generating coredump, we need some margin. The number of extra sections is
185 * 1-3 now and depends on arch. We use "5" as safe margin, here.
187 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
188 * not a hard limit any more. Although some userspace tools can be surprised by
191 #define MAPCOUNT_ELF_CORE_MARGIN (5)
192 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
194 extern int sysctl_max_map_count;
196 extern unsigned long sysctl_user_reserve_kbytes;
197 extern unsigned long sysctl_admin_reserve_kbytes;
199 extern int sysctl_overcommit_memory;
200 extern int sysctl_overcommit_ratio;
201 extern unsigned long sysctl_overcommit_kbytes;
203 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
205 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
208 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
210 /* to align the pointer to the (next) page boundary */
211 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
213 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
214 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
216 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
219 * Linux kernel virtual memory manager primitives.
220 * The idea being to have a "virtual" mm in the same way
221 * we have a virtual fs - giving a cleaner interface to the
222 * mm details, and allowing different kinds of memory mappings
223 * (from shared memory to executable loading to arbitrary
227 struct vm_area_struct *vm_area_alloc(struct mm_struct *);
228 struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
229 void vm_area_free(struct vm_area_struct *);
232 extern struct rb_root nommu_region_tree;
233 extern struct rw_semaphore nommu_region_sem;
235 extern unsigned int kobjsize(const void *objp);
239 * vm_flags in vm_area_struct, see mm_types.h.
240 * When changing, update also include/trace/events/mmflags.h
242 #define VM_NONE 0x00000000
244 #define VM_READ 0x00000001 /* currently active flags */
245 #define VM_WRITE 0x00000002
246 #define VM_EXEC 0x00000004
247 #define VM_SHARED 0x00000008
249 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
250 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
251 #define VM_MAYWRITE 0x00000020
252 #define VM_MAYEXEC 0x00000040
253 #define VM_MAYSHARE 0x00000080
255 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
256 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
257 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
258 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
259 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
261 #define VM_LOCKED 0x00002000
262 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
264 /* Used by sys_madvise() */
265 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
266 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
268 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
269 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
270 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
271 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
272 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
273 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
274 #define VM_SYNC 0x00800000 /* Synchronous page faults */
275 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
276 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
277 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
279 #ifdef CONFIG_MEM_SOFT_DIRTY
280 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
282 # define VM_SOFTDIRTY 0
285 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
286 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
287 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
288 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
290 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
291 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
292 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
293 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
294 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
295 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
296 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
297 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
298 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
299 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
300 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
301 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
303 #ifdef CONFIG_ARCH_HAS_PKEYS
304 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
305 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
306 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
307 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
308 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
310 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
312 # define VM_PKEY_BIT4 0
314 #endif /* CONFIG_ARCH_HAS_PKEYS */
316 #if defined(CONFIG_X86)
317 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
318 #elif defined(CONFIG_PPC)
319 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
320 #elif defined(CONFIG_PARISC)
321 # define VM_GROWSUP VM_ARCH_1
322 #elif defined(CONFIG_IA64)
323 # define VM_GROWSUP VM_ARCH_1
324 #elif defined(CONFIG_SPARC64)
325 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
326 # define VM_ARCH_CLEAR VM_SPARC_ADI
327 #elif defined(CONFIG_ARM64)
328 # define VM_ARM64_BTI VM_ARCH_1 /* BTI guarded page, a.k.a. GP bit */
329 # define VM_ARCH_CLEAR VM_ARM64_BTI
330 #elif !defined(CONFIG_MMU)
331 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
334 #if defined(CONFIG_X86_INTEL_MPX)
335 /* MPX specific bounds table or bounds directory */
336 # define VM_MPX VM_HIGH_ARCH_4
338 # define VM_MPX VM_NONE
342 # define VM_GROWSUP VM_NONE
345 /* Bits set in the VMA until the stack is in its final location */
346 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
348 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
349 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
352 #ifdef CONFIG_STACK_GROWSUP
353 #define VM_STACK VM_GROWSUP
355 #define VM_STACK VM_GROWSDOWN
358 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
361 * Special vmas that are non-mergable, non-mlock()able.
362 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
364 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
366 /* This mask defines which mm->def_flags a process can inherit its parent */
367 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
369 /* This mask is used to clear all the VMA flags used by mlock */
370 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
372 /* Arch-specific flags to clear when updating VM flags on protection change */
373 #ifndef VM_ARCH_CLEAR
374 # define VM_ARCH_CLEAR VM_NONE
376 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
379 * mapping from the currently active vm_flags protection bits (the
380 * low four bits) to a page protection mask..
382 extern pgprot_t protection_map[16];
384 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
385 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
386 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
387 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
388 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
389 #define FAULT_FLAG_TRIED 0x20 /* Second try */
390 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
391 #define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */
392 #define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */
394 #define FAULT_FLAG_TRACE \
395 { FAULT_FLAG_WRITE, "WRITE" }, \
396 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
397 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
398 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
399 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
400 { FAULT_FLAG_TRIED, "TRIED" }, \
401 { FAULT_FLAG_USER, "USER" }, \
402 { FAULT_FLAG_REMOTE, "REMOTE" }, \
403 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }
406 * vm_fault is filled by the the pagefault handler and passed to the vma's
407 * ->fault function. The vma's ->fault is responsible for returning a bitmask
408 * of VM_FAULT_xxx flags that give details about how the fault was handled.
410 * MM layer fills up gfp_mask for page allocations but fault handler might
411 * alter it if its implementation requires a different allocation context.
413 * pgoff should be used in favour of virtual_address, if possible.
416 struct vm_area_struct *vma; /* Target VMA */
417 unsigned int flags; /* FAULT_FLAG_xxx flags */
418 gfp_t gfp_mask; /* gfp mask to be used for allocations */
419 pgoff_t pgoff; /* Logical page offset based on vma */
420 unsigned long address; /* Faulting virtual address */
421 pmd_t *pmd; /* Pointer to pmd entry matching
423 pud_t *pud; /* Pointer to pud entry matching
426 pte_t orig_pte; /* Value of PTE at the time of fault */
428 struct page *cow_page; /* Page handler may use for COW fault */
429 struct mem_cgroup *memcg; /* Cgroup cow_page belongs to */
430 struct page *page; /* ->fault handlers should return a
431 * page here, unless VM_FAULT_NOPAGE
432 * is set (which is also implied by
435 /* These three entries are valid only while holding ptl lock */
436 pte_t *pte; /* Pointer to pte entry matching
437 * the 'address'. NULL if the page
438 * table hasn't been allocated.
440 spinlock_t *ptl; /* Page table lock.
441 * Protects pte page table if 'pte'
442 * is not NULL, otherwise pmd.
444 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
445 * vm_ops->map_pages() calls
446 * alloc_set_pte() from atomic context.
447 * do_fault_around() pre-allocates
448 * page table to avoid allocation from
453 /* page entry size for vm->huge_fault() */
454 enum page_entry_size {
461 * These are the virtual MM functions - opening of an area, closing and
462 * unmapping it (needed to keep files on disk up-to-date etc), pointer
463 * to the functions called when a no-page or a wp-page exception occurs.
465 struct vm_operations_struct {
466 void (*open)(struct vm_area_struct * area);
467 void (*close)(struct vm_area_struct * area);
468 int (*split)(struct vm_area_struct * area, unsigned long addr);
469 int (*mremap)(struct vm_area_struct * area);
470 vm_fault_t (*fault)(struct vm_fault *vmf);
471 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
472 enum page_entry_size pe_size);
473 void (*map_pages)(struct vm_fault *vmf,
474 pgoff_t start_pgoff, pgoff_t end_pgoff);
475 unsigned long (*pagesize)(struct vm_area_struct * area);
477 /* notification that a previously read-only page is about to become
478 * writable, if an error is returned it will cause a SIGBUS */
479 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
481 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
482 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
484 /* called by access_process_vm when get_user_pages() fails, typically
485 * for use by special VMAs that can switch between memory and hardware
487 int (*access)(struct vm_area_struct *vma, unsigned long addr,
488 void *buf, int len, int write);
490 /* Called by the /proc/PID/maps code to ask the vma whether it
491 * has a special name. Returning non-NULL will also cause this
492 * vma to be dumped unconditionally. */
493 const char *(*name)(struct vm_area_struct *vma);
497 * set_policy() op must add a reference to any non-NULL @new mempolicy
498 * to hold the policy upon return. Caller should pass NULL @new to
499 * remove a policy and fall back to surrounding context--i.e. do not
500 * install a MPOL_DEFAULT policy, nor the task or system default
503 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
506 * get_policy() op must add reference [mpol_get()] to any policy at
507 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
508 * in mm/mempolicy.c will do this automatically.
509 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
510 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
511 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
512 * must return NULL--i.e., do not "fallback" to task or system default
515 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
519 * Called by vm_normal_page() for special PTEs to find the
520 * page for @addr. This is useful if the default behavior
521 * (using pte_page()) would not find the correct page.
523 struct page *(*find_special_page)(struct vm_area_struct *vma,
527 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
529 static const struct vm_operations_struct dummy_vm_ops = {};
531 memset(vma, 0, sizeof(*vma));
533 vma->vm_ops = &dummy_vm_ops;
534 INIT_LIST_HEAD(&vma->anon_vma_chain);
537 static inline void vma_set_anonymous(struct vm_area_struct *vma)
542 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
549 * The vma_is_shmem is not inline because it is used only by slow
550 * paths in userfault.
552 bool vma_is_shmem(struct vm_area_struct *vma);
554 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
557 int vma_is_stack_for_current(struct vm_area_struct *vma);
559 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
560 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
566 * FIXME: take this include out, include page-flags.h in
567 * files which need it (119 of them)
569 #include <linux/page-flags.h>
570 #include <linux/huge_mm.h>
573 * Methods to modify the page usage count.
575 * What counts for a page usage:
576 * - cache mapping (page->mapping)
577 * - private data (page->private)
578 * - page mapped in a task's page tables, each mapping
579 * is counted separately
581 * Also, many kernel routines increase the page count before a critical
582 * routine so they can be sure the page doesn't go away from under them.
586 * Drop a ref, return true if the refcount fell to zero (the page has no users)
588 static inline int put_page_testzero(struct page *page)
590 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
591 return page_ref_dec_and_test(page);
595 * Try to grab a ref unless the page has a refcount of zero, return false if
597 * This can be called when MMU is off so it must not access
598 * any of the virtual mappings.
600 static inline int get_page_unless_zero(struct page *page)
602 return page_ref_add_unless(page, 1, 0);
605 extern int page_is_ram(unsigned long pfn);
613 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
616 /* Support for virtually mapped pages */
617 struct page *vmalloc_to_page(const void *addr);
618 unsigned long vmalloc_to_pfn(const void *addr);
621 * Determine if an address is within the vmalloc range
623 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
624 * is no special casing required.
627 #ifndef is_ioremap_addr
628 #define is_ioremap_addr(x) is_vmalloc_addr(x)
632 extern bool is_vmalloc_addr(const void *x);
633 extern int is_vmalloc_or_module_addr(const void *x);
635 static inline bool is_vmalloc_addr(const void *x)
639 static inline int is_vmalloc_or_module_addr(const void *x)
645 extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
646 static inline void *kvmalloc(size_t size, gfp_t flags)
648 return kvmalloc_node(size, flags, NUMA_NO_NODE);
650 static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
652 return kvmalloc_node(size, flags | __GFP_ZERO, node);
654 static inline void *kvzalloc(size_t size, gfp_t flags)
656 return kvmalloc(size, flags | __GFP_ZERO);
659 static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
663 if (unlikely(check_mul_overflow(n, size, &bytes)))
666 return kvmalloc(bytes, flags);
669 static inline void *kvcalloc(size_t n, size_t size, gfp_t flags)
671 return kvmalloc_array(n, size, flags | __GFP_ZERO);
674 extern void kvfree(const void *addr);
676 static inline int compound_mapcount(struct page *page)
678 VM_BUG_ON_PAGE(!PageCompound(page), page);
679 page = compound_head(page);
680 return atomic_read(compound_mapcount_ptr(page)) + 1;
684 * The atomic page->_mapcount, starts from -1: so that transitions
685 * both from it and to it can be tracked, using atomic_inc_and_test
686 * and atomic_add_negative(-1).
688 static inline void page_mapcount_reset(struct page *page)
690 atomic_set(&(page)->_mapcount, -1);
693 int __page_mapcount(struct page *page);
695 static inline int page_mapcount(struct page *page)
697 VM_BUG_ON_PAGE(PageSlab(page), page);
699 if (unlikely(PageCompound(page)))
700 return __page_mapcount(page);
701 return atomic_read(&page->_mapcount) + 1;
704 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
705 int total_mapcount(struct page *page);
706 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
708 static inline int total_mapcount(struct page *page)
710 return page_mapcount(page);
712 static inline int page_trans_huge_mapcount(struct page *page,
715 int mapcount = page_mapcount(page);
717 *total_mapcount = mapcount;
722 static inline struct page *virt_to_head_page(const void *x)
724 struct page *page = virt_to_page(x);
726 return compound_head(page);
729 void __put_page(struct page *page);
731 void put_pages_list(struct list_head *pages);
733 void split_page(struct page *page, unsigned int order);
736 * Compound pages have a destructor function. Provide a
737 * prototype for that function and accessor functions.
738 * These are _only_ valid on the head of a compound page.
740 typedef void compound_page_dtor(struct page *);
742 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
743 enum compound_dtor_id {
746 #ifdef CONFIG_HUGETLB_PAGE
749 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
754 extern compound_page_dtor * const compound_page_dtors[];
756 static inline void set_compound_page_dtor(struct page *page,
757 enum compound_dtor_id compound_dtor)
759 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
760 page[1].compound_dtor = compound_dtor;
763 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
765 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
766 return compound_page_dtors[page[1].compound_dtor];
769 static inline unsigned int compound_order(struct page *page)
773 return page[1].compound_order;
776 static inline void set_compound_order(struct page *page, unsigned int order)
778 page[1].compound_order = order;
781 /* Returns the number of pages in this potentially compound page. */
782 static inline unsigned long compound_nr(struct page *page)
784 return 1UL << compound_order(page);
787 /* Returns the number of bytes in this potentially compound page. */
788 static inline unsigned long page_size(struct page *page)
790 return PAGE_SIZE << compound_order(page);
793 /* Returns the number of bits needed for the number of bytes in a page */
794 static inline unsigned int page_shift(struct page *page)
796 return PAGE_SHIFT + compound_order(page);
799 void free_compound_page(struct page *page);
803 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
804 * servicing faults for write access. In the normal case, do always want
805 * pte_mkwrite. But get_user_pages can cause write faults for mappings
806 * that do not have writing enabled, when used by access_process_vm.
808 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
810 if (likely(vma->vm_flags & VM_WRITE))
811 pte = pte_mkwrite(pte);
815 vm_fault_t alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
817 vm_fault_t finish_fault(struct vm_fault *vmf);
818 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
822 * Multiple processes may "see" the same page. E.g. for untouched
823 * mappings of /dev/null, all processes see the same page full of
824 * zeroes, and text pages of executables and shared libraries have
825 * only one copy in memory, at most, normally.
827 * For the non-reserved pages, page_count(page) denotes a reference count.
828 * page_count() == 0 means the page is free. page->lru is then used for
829 * freelist management in the buddy allocator.
830 * page_count() > 0 means the page has been allocated.
832 * Pages are allocated by the slab allocator in order to provide memory
833 * to kmalloc and kmem_cache_alloc. In this case, the management of the
834 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
835 * unless a particular usage is carefully commented. (the responsibility of
836 * freeing the kmalloc memory is the caller's, of course).
838 * A page may be used by anyone else who does a __get_free_page().
839 * In this case, page_count still tracks the references, and should only
840 * be used through the normal accessor functions. The top bits of page->flags
841 * and page->virtual store page management information, but all other fields
842 * are unused and could be used privately, carefully. The management of this
843 * page is the responsibility of the one who allocated it, and those who have
844 * subsequently been given references to it.
846 * The other pages (we may call them "pagecache pages") are completely
847 * managed by the Linux memory manager: I/O, buffers, swapping etc.
848 * The following discussion applies only to them.
850 * A pagecache page contains an opaque `private' member, which belongs to the
851 * page's address_space. Usually, this is the address of a circular list of
852 * the page's disk buffers. PG_private must be set to tell the VM to call
853 * into the filesystem to release these pages.
855 * A page may belong to an inode's memory mapping. In this case, page->mapping
856 * is the pointer to the inode, and page->index is the file offset of the page,
857 * in units of PAGE_SIZE.
859 * If pagecache pages are not associated with an inode, they are said to be
860 * anonymous pages. These may become associated with the swapcache, and in that
861 * case PG_swapcache is set, and page->private is an offset into the swapcache.
863 * In either case (swapcache or inode backed), the pagecache itself holds one
864 * reference to the page. Setting PG_private should also increment the
865 * refcount. The each user mapping also has a reference to the page.
867 * The pagecache pages are stored in a per-mapping radix tree, which is
868 * rooted at mapping->i_pages, and indexed by offset.
869 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
870 * lists, we instead now tag pages as dirty/writeback in the radix tree.
872 * All pagecache pages may be subject to I/O:
873 * - inode pages may need to be read from disk,
874 * - inode pages which have been modified and are MAP_SHARED may need
875 * to be written back to the inode on disk,
876 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
877 * modified may need to be swapped out to swap space and (later) to be read
882 * The zone field is never updated after free_area_init_core()
883 * sets it, so none of the operations on it need to be atomic.
886 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
887 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
888 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
889 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
890 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
891 #define KASAN_TAG_PGOFF (LAST_CPUPID_PGOFF - KASAN_TAG_WIDTH)
894 * Define the bit shifts to access each section. For non-existent
895 * sections we define the shift as 0; that plus a 0 mask ensures
896 * the compiler will optimise away reference to them.
898 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
899 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
900 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
901 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
902 #define KASAN_TAG_PGSHIFT (KASAN_TAG_PGOFF * (KASAN_TAG_WIDTH != 0))
904 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
905 #ifdef NODE_NOT_IN_PAGE_FLAGS
906 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
907 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
908 SECTIONS_PGOFF : ZONES_PGOFF)
910 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
911 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
912 NODES_PGOFF : ZONES_PGOFF)
915 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
917 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
918 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
919 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
920 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
921 #define KASAN_TAG_MASK ((1UL << KASAN_TAG_WIDTH) - 1)
922 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
924 static inline enum zone_type page_zonenum(const struct page *page)
926 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
929 #ifdef CONFIG_ZONE_DEVICE
930 static inline bool is_zone_device_page(const struct page *page)
932 return page_zonenum(page) == ZONE_DEVICE;
934 extern void memmap_init_zone_device(struct zone *, unsigned long,
935 unsigned long, struct dev_pagemap *);
937 static inline bool is_zone_device_page(const struct page *page)
943 #ifdef CONFIG_DEV_PAGEMAP_OPS
944 void free_devmap_managed_page(struct page *page);
945 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
947 static inline bool page_is_devmap_managed(struct page *page)
949 if (!static_branch_unlikely(&devmap_managed_key))
951 if (!is_zone_device_page(page))
953 switch (page->pgmap->type) {
954 case MEMORY_DEVICE_PRIVATE:
955 case MEMORY_DEVICE_FS_DAX:
963 void put_devmap_managed_page(struct page *page);
965 #else /* CONFIG_DEV_PAGEMAP_OPS */
966 static inline bool page_is_devmap_managed(struct page *page)
971 static inline void put_devmap_managed_page(struct page *page)
974 #endif /* CONFIG_DEV_PAGEMAP_OPS */
976 static inline bool is_device_private_page(const struct page *page)
978 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
979 IS_ENABLED(CONFIG_DEVICE_PRIVATE) &&
980 is_zone_device_page(page) &&
981 page->pgmap->type == MEMORY_DEVICE_PRIVATE;
984 static inline bool is_pci_p2pdma_page(const struct page *page)
986 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
987 IS_ENABLED(CONFIG_PCI_P2PDMA) &&
988 is_zone_device_page(page) &&
989 page->pgmap->type == MEMORY_DEVICE_PCI_P2PDMA;
992 /* 127: arbitrary random number, small enough to assemble well */
993 #define page_ref_zero_or_close_to_overflow(page) \
994 ((unsigned int) page_ref_count(page) + 127u <= 127u)
996 static inline void get_page(struct page *page)
998 page = compound_head(page);
1000 * Getting a normal page or the head of a compound page
1001 * requires to already have an elevated page->_refcount.
1003 VM_BUG_ON_PAGE(page_ref_zero_or_close_to_overflow(page), page);
1007 static inline __must_check bool try_get_page(struct page *page)
1009 page = compound_head(page);
1010 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1016 static inline void put_page(struct page *page)
1018 page = compound_head(page);
1021 * For devmap managed pages we need to catch refcount transition from
1022 * 2 to 1, when refcount reach one it means the page is free and we
1023 * need to inform the device driver through callback. See
1024 * include/linux/memremap.h and HMM for details.
1026 if (page_is_devmap_managed(page)) {
1027 put_devmap_managed_page(page);
1031 if (put_page_testzero(page))
1036 * unpin_user_page() - release a gup-pinned page
1037 * @page: pointer to page to be released
1039 * Pages that were pinned via pin_user_pages*() must be released via either
1040 * unpin_user_page(), or one of the unpin_user_pages*() routines. This is so
1041 * that eventually such pages can be separately tracked and uniquely handled. In
1042 * particular, interactions with RDMA and filesystems need special handling.
1044 * unpin_user_page() and put_page() are not interchangeable, despite this early
1045 * implementation that makes them look the same. unpin_user_page() calls must
1046 * be perfectly matched up with pin*() calls.
1048 static inline void unpin_user_page(struct page *page)
1053 void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1056 void unpin_user_pages(struct page **pages, unsigned long npages);
1058 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1059 #define SECTION_IN_PAGE_FLAGS
1063 * The identification function is mainly used by the buddy allocator for
1064 * determining if two pages could be buddies. We are not really identifying
1065 * the zone since we could be using the section number id if we do not have
1066 * node id available in page flags.
1067 * We only guarantee that it will return the same value for two combinable
1070 static inline int page_zone_id(struct page *page)
1072 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1075 #ifdef NODE_NOT_IN_PAGE_FLAGS
1076 extern int page_to_nid(const struct page *page);
1078 static inline int page_to_nid(const struct page *page)
1080 struct page *p = (struct page *)page;
1082 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1086 #ifdef CONFIG_NUMA_BALANCING
1087 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1089 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1092 static inline int cpupid_to_pid(int cpupid)
1094 return cpupid & LAST__PID_MASK;
1097 static inline int cpupid_to_cpu(int cpupid)
1099 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1102 static inline int cpupid_to_nid(int cpupid)
1104 return cpu_to_node(cpupid_to_cpu(cpupid));
1107 static inline bool cpupid_pid_unset(int cpupid)
1109 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1112 static inline bool cpupid_cpu_unset(int cpupid)
1114 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1117 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1119 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1122 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1123 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1124 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1126 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1129 static inline int page_cpupid_last(struct page *page)
1131 return page->_last_cpupid;
1133 static inline void page_cpupid_reset_last(struct page *page)
1135 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1138 static inline int page_cpupid_last(struct page *page)
1140 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1143 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1145 static inline void page_cpupid_reset_last(struct page *page)
1147 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1149 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1150 #else /* !CONFIG_NUMA_BALANCING */
1151 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1153 return page_to_nid(page); /* XXX */
1156 static inline int page_cpupid_last(struct page *page)
1158 return page_to_nid(page); /* XXX */
1161 static inline int cpupid_to_nid(int cpupid)
1166 static inline int cpupid_to_pid(int cpupid)
1171 static inline int cpupid_to_cpu(int cpupid)
1176 static inline int cpu_pid_to_cpupid(int nid, int pid)
1181 static inline bool cpupid_pid_unset(int cpupid)
1186 static inline void page_cpupid_reset_last(struct page *page)
1190 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1194 #endif /* CONFIG_NUMA_BALANCING */
1196 #ifdef CONFIG_KASAN_SW_TAGS
1197 static inline u8 page_kasan_tag(const struct page *page)
1199 return (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1202 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1204 page->flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1205 page->flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1208 static inline void page_kasan_tag_reset(struct page *page)
1210 page_kasan_tag_set(page, 0xff);
1213 static inline u8 page_kasan_tag(const struct page *page)
1218 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1219 static inline void page_kasan_tag_reset(struct page *page) { }
1222 static inline struct zone *page_zone(const struct page *page)
1224 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1227 static inline pg_data_t *page_pgdat(const struct page *page)
1229 return NODE_DATA(page_to_nid(page));
1232 #ifdef SECTION_IN_PAGE_FLAGS
1233 static inline void set_page_section(struct page *page, unsigned long section)
1235 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1236 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1239 static inline unsigned long page_to_section(const struct page *page)
1241 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1245 static inline void set_page_zone(struct page *page, enum zone_type zone)
1247 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1248 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1251 static inline void set_page_node(struct page *page, unsigned long node)
1253 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1254 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1257 static inline void set_page_links(struct page *page, enum zone_type zone,
1258 unsigned long node, unsigned long pfn)
1260 set_page_zone(page, zone);
1261 set_page_node(page, node);
1262 #ifdef SECTION_IN_PAGE_FLAGS
1263 set_page_section(page, pfn_to_section_nr(pfn));
1268 static inline struct mem_cgroup *page_memcg(struct page *page)
1270 return page->mem_cgroup;
1272 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1274 WARN_ON_ONCE(!rcu_read_lock_held());
1275 return READ_ONCE(page->mem_cgroup);
1278 static inline struct mem_cgroup *page_memcg(struct page *page)
1282 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1284 WARN_ON_ONCE(!rcu_read_lock_held());
1290 * Some inline functions in vmstat.h depend on page_zone()
1292 #include <linux/vmstat.h>
1294 static __always_inline void *lowmem_page_address(const struct page *page)
1296 return page_to_virt(page);
1299 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1300 #define HASHED_PAGE_VIRTUAL
1303 #if defined(WANT_PAGE_VIRTUAL)
1304 static inline void *page_address(const struct page *page)
1306 return page->virtual;
1308 static inline void set_page_address(struct page *page, void *address)
1310 page->virtual = address;
1312 #define page_address_init() do { } while(0)
1315 #if defined(HASHED_PAGE_VIRTUAL)
1316 void *page_address(const struct page *page);
1317 void set_page_address(struct page *page, void *virtual);
1318 void page_address_init(void);
1321 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1322 #define page_address(page) lowmem_page_address(page)
1323 #define set_page_address(page, address) do { } while(0)
1324 #define page_address_init() do { } while(0)
1327 extern void *page_rmapping(struct page *page);
1328 extern struct anon_vma *page_anon_vma(struct page *page);
1329 extern struct address_space *page_mapping(struct page *page);
1331 extern struct address_space *__page_file_mapping(struct page *);
1334 struct address_space *page_file_mapping(struct page *page)
1336 if (unlikely(PageSwapCache(page)))
1337 return __page_file_mapping(page);
1339 return page->mapping;
1342 extern pgoff_t __page_file_index(struct page *page);
1345 * Return the pagecache index of the passed page. Regular pagecache pages
1346 * use ->index whereas swapcache pages use swp_offset(->private)
1348 static inline pgoff_t page_index(struct page *page)
1350 if (unlikely(PageSwapCache(page)))
1351 return __page_file_index(page);
1355 bool page_mapped(struct page *page);
1356 struct address_space *page_mapping(struct page *page);
1357 struct address_space *page_mapping_file(struct page *page);
1360 * Return true only if the page has been allocated with
1361 * ALLOC_NO_WATERMARKS and the low watermark was not
1362 * met implying that the system is under some pressure.
1364 static inline bool page_is_pfmemalloc(struct page *page)
1367 * Page index cannot be this large so this must be
1368 * a pfmemalloc page.
1370 return page->index == -1UL;
1374 * Only to be called by the page allocator on a freshly allocated
1377 static inline void set_page_pfmemalloc(struct page *page)
1382 static inline void clear_page_pfmemalloc(struct page *page)
1388 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1390 extern void pagefault_out_of_memory(void);
1392 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1395 * Flags passed to show_mem() and show_free_areas() to suppress output in
1398 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1400 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1403 extern bool can_do_mlock(void);
1405 static inline bool can_do_mlock(void) { return false; }
1407 extern int user_shm_lock(size_t, struct user_struct *);
1408 extern void user_shm_unlock(size_t, struct user_struct *);
1411 * Parameter block passed down to zap_pte_range in exceptional cases.
1413 struct zap_details {
1414 struct address_space *check_mapping; /* Check page->mapping if set */
1415 pgoff_t first_index; /* Lowest page->index to unmap */
1416 pgoff_t last_index; /* Highest page->index to unmap */
1419 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1421 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1424 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1425 unsigned long size);
1426 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1427 unsigned long size);
1428 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1429 unsigned long start, unsigned long end);
1431 struct mmu_notifier_range;
1433 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1434 unsigned long end, unsigned long floor, unsigned long ceiling);
1435 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1436 struct vm_area_struct *vma);
1437 int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
1438 struct mmu_notifier_range *range,
1439 pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp);
1440 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1441 unsigned long *pfn);
1442 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1443 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1444 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1445 void *buf, int len, int write);
1447 extern void truncate_pagecache(struct inode *inode, loff_t new);
1448 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1449 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1450 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1451 int truncate_inode_page(struct address_space *mapping, struct page *page);
1452 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1453 int invalidate_inode_page(struct page *page);
1456 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1457 unsigned long address, unsigned int flags);
1458 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1459 unsigned long address, unsigned int fault_flags,
1461 void unmap_mapping_pages(struct address_space *mapping,
1462 pgoff_t start, pgoff_t nr, bool even_cows);
1463 void unmap_mapping_range(struct address_space *mapping,
1464 loff_t const holebegin, loff_t const holelen, int even_cows);
1466 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1467 unsigned long address, unsigned int flags)
1469 /* should never happen if there's no MMU */
1471 return VM_FAULT_SIGBUS;
1473 static inline int fixup_user_fault(struct task_struct *tsk,
1474 struct mm_struct *mm, unsigned long address,
1475 unsigned int fault_flags, bool *unlocked)
1477 /* should never happen if there's no MMU */
1481 static inline void unmap_mapping_pages(struct address_space *mapping,
1482 pgoff_t start, pgoff_t nr, bool even_cows) { }
1483 static inline void unmap_mapping_range(struct address_space *mapping,
1484 loff_t const holebegin, loff_t const holelen, int even_cows) { }
1487 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1488 loff_t const holebegin, loff_t const holelen)
1490 unmap_mapping_range(mapping, holebegin, holelen, 0);
1493 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1494 void *buf, int len, unsigned int gup_flags);
1495 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1496 void *buf, int len, unsigned int gup_flags);
1497 extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1498 unsigned long addr, void *buf, int len, unsigned int gup_flags);
1500 long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1501 unsigned long start, unsigned long nr_pages,
1502 unsigned int gup_flags, struct page **pages,
1503 struct vm_area_struct **vmas, int *locked);
1504 long pin_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1505 unsigned long start, unsigned long nr_pages,
1506 unsigned int gup_flags, struct page **pages,
1507 struct vm_area_struct **vmas, int *locked);
1508 long get_user_pages(unsigned long start, unsigned long nr_pages,
1509 unsigned int gup_flags, struct page **pages,
1510 struct vm_area_struct **vmas);
1511 long pin_user_pages(unsigned long start, unsigned long nr_pages,
1512 unsigned int gup_flags, struct page **pages,
1513 struct vm_area_struct **vmas);
1514 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1515 unsigned int gup_flags, struct page **pages, int *locked);
1516 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1517 struct page **pages, unsigned int gup_flags);
1519 int get_user_pages_fast(unsigned long start, int nr_pages,
1520 unsigned int gup_flags, struct page **pages);
1521 int pin_user_pages_fast(unsigned long start, int nr_pages,
1522 unsigned int gup_flags, struct page **pages);
1524 int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
1525 int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
1526 struct task_struct *task, bool bypass_rlim);
1528 /* Container for pinned pfns / pages */
1529 struct frame_vector {
1530 unsigned int nr_allocated; /* Number of frames we have space for */
1531 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1532 bool got_ref; /* Did we pin pages by getting page ref? */
1533 bool is_pfns; /* Does array contain pages or pfns? */
1534 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1535 * pfns_vector_pages() or pfns_vector_pfns()
1539 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1540 void frame_vector_destroy(struct frame_vector *vec);
1541 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1542 unsigned int gup_flags, struct frame_vector *vec);
1543 void put_vaddr_frames(struct frame_vector *vec);
1544 int frame_vector_to_pages(struct frame_vector *vec);
1545 void frame_vector_to_pfns(struct frame_vector *vec);
1547 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1549 return vec->nr_frames;
1552 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1555 int err = frame_vector_to_pages(vec);
1558 return ERR_PTR(err);
1560 return (struct page **)(vec->ptrs);
1563 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1566 frame_vector_to_pfns(vec);
1567 return (unsigned long *)(vec->ptrs);
1571 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1572 struct page **pages);
1573 int get_kernel_page(unsigned long start, int write, struct page **pages);
1574 struct page *get_dump_page(unsigned long addr);
1576 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1577 extern void do_invalidatepage(struct page *page, unsigned int offset,
1578 unsigned int length);
1580 void __set_page_dirty(struct page *, struct address_space *, int warn);
1581 int __set_page_dirty_nobuffers(struct page *page);
1582 int __set_page_dirty_no_writeback(struct page *page);
1583 int redirty_page_for_writepage(struct writeback_control *wbc,
1585 void account_page_dirtied(struct page *page, struct address_space *mapping);
1586 void account_page_cleaned(struct page *page, struct address_space *mapping,
1587 struct bdi_writeback *wb);
1588 int set_page_dirty(struct page *page);
1589 int set_page_dirty_lock(struct page *page);
1590 void __cancel_dirty_page(struct page *page);
1591 static inline void cancel_dirty_page(struct page *page)
1593 /* Avoid atomic ops, locking, etc. when not actually needed. */
1594 if (PageDirty(page))
1595 __cancel_dirty_page(page);
1597 int clear_page_dirty_for_io(struct page *page);
1599 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1601 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1602 unsigned long old_addr, struct vm_area_struct *new_vma,
1603 unsigned long new_addr, unsigned long len,
1604 bool need_rmap_locks);
1605 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1606 unsigned long end, pgprot_t newprot,
1607 int dirty_accountable, int prot_numa);
1608 extern int mprotect_fixup(struct vm_area_struct *vma,
1609 struct vm_area_struct **pprev, unsigned long start,
1610 unsigned long end, unsigned long newflags);
1613 * doesn't attempt to fault and will return short.
1615 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1616 struct page **pages);
1618 * per-process(per-mm_struct) statistics.
1620 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1622 long val = atomic_long_read(&mm->rss_stat.count[member]);
1624 #ifdef SPLIT_RSS_COUNTING
1626 * counter is updated in asynchronous manner and may go to minus.
1627 * But it's never be expected number for users.
1632 return (unsigned long)val;
1635 void mm_trace_rss_stat(struct mm_struct *mm, int member, long count);
1637 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1639 long count = atomic_long_add_return(value, &mm->rss_stat.count[member]);
1641 mm_trace_rss_stat(mm, member, count);
1644 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1646 long count = atomic_long_inc_return(&mm->rss_stat.count[member]);
1648 mm_trace_rss_stat(mm, member, count);
1651 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1653 long count = atomic_long_dec_return(&mm->rss_stat.count[member]);
1655 mm_trace_rss_stat(mm, member, count);
1658 /* Optimized variant when page is already known not to be PageAnon */
1659 static inline int mm_counter_file(struct page *page)
1661 if (PageSwapBacked(page))
1662 return MM_SHMEMPAGES;
1663 return MM_FILEPAGES;
1666 static inline int mm_counter(struct page *page)
1669 return MM_ANONPAGES;
1670 return mm_counter_file(page);
1673 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1675 return get_mm_counter(mm, MM_FILEPAGES) +
1676 get_mm_counter(mm, MM_ANONPAGES) +
1677 get_mm_counter(mm, MM_SHMEMPAGES);
1680 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1682 return max(mm->hiwater_rss, get_mm_rss(mm));
1685 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1687 return max(mm->hiwater_vm, mm->total_vm);
1690 static inline void update_hiwater_rss(struct mm_struct *mm)
1692 unsigned long _rss = get_mm_rss(mm);
1694 if ((mm)->hiwater_rss < _rss)
1695 (mm)->hiwater_rss = _rss;
1698 static inline void update_hiwater_vm(struct mm_struct *mm)
1700 if (mm->hiwater_vm < mm->total_vm)
1701 mm->hiwater_vm = mm->total_vm;
1704 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1706 mm->hiwater_rss = get_mm_rss(mm);
1709 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1710 struct mm_struct *mm)
1712 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1714 if (*maxrss < hiwater_rss)
1715 *maxrss = hiwater_rss;
1718 #if defined(SPLIT_RSS_COUNTING)
1719 void sync_mm_rss(struct mm_struct *mm);
1721 static inline void sync_mm_rss(struct mm_struct *mm)
1726 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
1727 static inline int pte_devmap(pte_t pte)
1733 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1735 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1737 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1741 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1745 #ifdef __PAGETABLE_P4D_FOLDED
1746 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1747 unsigned long address)
1752 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1755 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
1756 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1757 unsigned long address)
1761 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
1762 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
1765 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
1767 static inline void mm_inc_nr_puds(struct mm_struct *mm)
1769 if (mm_pud_folded(mm))
1771 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1774 static inline void mm_dec_nr_puds(struct mm_struct *mm)
1776 if (mm_pud_folded(mm))
1778 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1782 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1783 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1784 unsigned long address)
1789 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1790 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1793 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1795 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1797 if (mm_pmd_folded(mm))
1799 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1802 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1804 if (mm_pmd_folded(mm))
1806 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1811 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
1813 atomic_long_set(&mm->pgtables_bytes, 0);
1816 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1818 return atomic_long_read(&mm->pgtables_bytes);
1821 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
1823 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1826 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
1828 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1832 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
1833 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1838 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
1839 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
1842 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
1843 int __pte_alloc_kernel(pmd_t *pmd);
1845 #if defined(CONFIG_MMU)
1848 * The following ifdef needed to get the 5level-fixup.h header to work.
1849 * Remove it when 5level-fixup.h has been removed.
1851 #ifndef __ARCH_HAS_5LEVEL_HACK
1852 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1853 unsigned long address)
1855 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
1856 NULL : p4d_offset(pgd, address);
1859 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1860 unsigned long address)
1862 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
1863 NULL : pud_offset(p4d, address);
1865 #endif /* !__ARCH_HAS_5LEVEL_HACK */
1867 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1869 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1870 NULL: pmd_offset(pud, address);
1872 #endif /* CONFIG_MMU */
1874 #if USE_SPLIT_PTE_PTLOCKS
1875 #if ALLOC_SPLIT_PTLOCKS
1876 void __init ptlock_cache_init(void);
1877 extern bool ptlock_alloc(struct page *page);
1878 extern void ptlock_free(struct page *page);
1880 static inline spinlock_t *ptlock_ptr(struct page *page)
1884 #else /* ALLOC_SPLIT_PTLOCKS */
1885 static inline void ptlock_cache_init(void)
1889 static inline bool ptlock_alloc(struct page *page)
1894 static inline void ptlock_free(struct page *page)
1898 static inline spinlock_t *ptlock_ptr(struct page *page)
1902 #endif /* ALLOC_SPLIT_PTLOCKS */
1904 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1906 return ptlock_ptr(pmd_page(*pmd));
1909 static inline bool ptlock_init(struct page *page)
1912 * prep_new_page() initialize page->private (and therefore page->ptl)
1913 * with 0. Make sure nobody took it in use in between.
1915 * It can happen if arch try to use slab for page table allocation:
1916 * slab code uses page->slab_cache, which share storage with page->ptl.
1918 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1919 if (!ptlock_alloc(page))
1921 spin_lock_init(ptlock_ptr(page));
1925 #else /* !USE_SPLIT_PTE_PTLOCKS */
1927 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1929 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1931 return &mm->page_table_lock;
1933 static inline void ptlock_cache_init(void) {}
1934 static inline bool ptlock_init(struct page *page) { return true; }
1935 static inline void ptlock_free(struct page *page) {}
1936 #endif /* USE_SPLIT_PTE_PTLOCKS */
1938 static inline void pgtable_init(void)
1940 ptlock_cache_init();
1941 pgtable_cache_init();
1944 static inline bool pgtable_pte_page_ctor(struct page *page)
1946 if (!ptlock_init(page))
1948 __SetPageTable(page);
1949 inc_zone_page_state(page, NR_PAGETABLE);
1953 static inline void pgtable_pte_page_dtor(struct page *page)
1956 __ClearPageTable(page);
1957 dec_zone_page_state(page, NR_PAGETABLE);
1960 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1962 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1963 pte_t *__pte = pte_offset_map(pmd, address); \
1969 #define pte_unmap_unlock(pte, ptl) do { \
1974 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
1976 #define pte_alloc_map(mm, pmd, address) \
1977 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
1979 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1980 (pte_alloc(mm, pmd) ? \
1981 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
1983 #define pte_alloc_kernel(pmd, address) \
1984 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
1985 NULL: pte_offset_kernel(pmd, address))
1987 #if USE_SPLIT_PMD_PTLOCKS
1989 static struct page *pmd_to_page(pmd_t *pmd)
1991 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1992 return virt_to_page((void *)((unsigned long) pmd & mask));
1995 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1997 return ptlock_ptr(pmd_to_page(pmd));
2000 static inline bool pgtable_pmd_page_ctor(struct page *page)
2002 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2003 page->pmd_huge_pte = NULL;
2005 return ptlock_init(page);
2008 static inline void pgtable_pmd_page_dtor(struct page *page)
2010 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2011 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2016 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2020 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2022 return &mm->page_table_lock;
2025 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
2026 static inline void pgtable_pmd_page_dtor(struct page *page) {}
2028 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2032 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2034 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2040 * No scalability reason to split PUD locks yet, but follow the same pattern
2041 * as the PMD locks to make it easier if we decide to. The VM should not be
2042 * considered ready to switch to split PUD locks yet; there may be places
2043 * which need to be converted from page_table_lock.
2045 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2047 return &mm->page_table_lock;
2050 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2052 spinlock_t *ptl = pud_lockptr(mm, pud);
2058 extern void __init pagecache_init(void);
2059 extern void free_area_init(unsigned long * zones_size);
2060 extern void __init free_area_init_node(int nid, unsigned long * zones_size,
2061 unsigned long zone_start_pfn, unsigned long *zholes_size);
2062 extern void free_initmem(void);
2065 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2066 * into the buddy system. The freed pages will be poisoned with pattern
2067 * "poison" if it's within range [0, UCHAR_MAX].
2068 * Return pages freed into the buddy system.
2070 extern unsigned long free_reserved_area(void *start, void *end,
2071 int poison, const char *s);
2073 #ifdef CONFIG_HIGHMEM
2075 * Free a highmem page into the buddy system, adjusting totalhigh_pages
2076 * and totalram_pages.
2078 extern void free_highmem_page(struct page *page);
2081 extern void adjust_managed_page_count(struct page *page, long count);
2082 extern void mem_init_print_info(const char *str);
2084 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2086 /* Free the reserved page into the buddy system, so it gets managed. */
2087 static inline void __free_reserved_page(struct page *page)
2089 ClearPageReserved(page);
2090 init_page_count(page);
2094 static inline void free_reserved_page(struct page *page)
2096 __free_reserved_page(page);
2097 adjust_managed_page_count(page, 1);
2100 static inline void mark_page_reserved(struct page *page)
2102 SetPageReserved(page);
2103 adjust_managed_page_count(page, -1);
2107 * Default method to free all the __init memory into the buddy system.
2108 * The freed pages will be poisoned with pattern "poison" if it's within
2109 * range [0, UCHAR_MAX].
2110 * Return pages freed into the buddy system.
2112 static inline unsigned long free_initmem_default(int poison)
2114 extern char __init_begin[], __init_end[];
2116 return free_reserved_area(&__init_begin, &__init_end,
2117 poison, "unused kernel");
2120 static inline unsigned long get_num_physpages(void)
2123 unsigned long phys_pages = 0;
2125 for_each_online_node(nid)
2126 phys_pages += node_present_pages(nid);
2131 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
2133 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
2134 * zones, allocate the backing mem_map and account for memory holes in a more
2135 * architecture independent manner. This is a substitute for creating the
2136 * zone_sizes[] and zholes_size[] arrays and passing them to
2137 * free_area_init_node()
2139 * An architecture is expected to register range of page frames backed by
2140 * physical memory with memblock_add[_node]() before calling
2141 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
2142 * usage, an architecture is expected to do something like
2144 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2146 * for_each_valid_physical_page_range()
2147 * memblock_add_node(base, size, nid)
2148 * free_area_init_nodes(max_zone_pfns);
2150 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
2151 * registered physical page range. Similarly
2152 * sparse_memory_present_with_active_regions() calls memory_present() for
2153 * each range when SPARSEMEM is enabled.
2155 * See mm/page_alloc.c for more information on each function exposed by
2156 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
2158 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
2159 unsigned long node_map_pfn_alignment(void);
2160 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2161 unsigned long end_pfn);
2162 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2163 unsigned long end_pfn);
2164 extern void get_pfn_range_for_nid(unsigned int nid,
2165 unsigned long *start_pfn, unsigned long *end_pfn);
2166 extern unsigned long find_min_pfn_with_active_regions(void);
2167 extern void free_bootmem_with_active_regions(int nid,
2168 unsigned long max_low_pfn);
2169 extern void sparse_memory_present_with_active_regions(int nid);
2171 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
2173 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
2174 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
2175 static inline int __early_pfn_to_nid(unsigned long pfn,
2176 struct mminit_pfnnid_cache *state)
2181 /* please see mm/page_alloc.c */
2182 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2183 /* there is a per-arch backend function. */
2184 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
2185 struct mminit_pfnnid_cache *state);
2188 extern void set_dma_reserve(unsigned long new_dma_reserve);
2189 extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long,
2190 enum memmap_context, struct vmem_altmap *);
2191 extern void setup_per_zone_wmarks(void);
2192 extern int __meminit init_per_zone_wmark_min(void);
2193 extern void mem_init(void);
2194 extern void __init mmap_init(void);
2195 extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2196 extern long si_mem_available(void);
2197 extern void si_meminfo(struct sysinfo * val);
2198 extern void si_meminfo_node(struct sysinfo *val, int nid);
2199 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2200 extern unsigned long arch_reserved_kernel_pages(void);
2203 extern __printf(3, 4)
2204 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2206 extern void setup_per_cpu_pageset(void);
2209 extern int min_free_kbytes;
2210 extern int watermark_boost_factor;
2211 extern int watermark_scale_factor;
2214 extern atomic_long_t mmap_pages_allocated;
2215 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2217 /* interval_tree.c */
2218 void vma_interval_tree_insert(struct vm_area_struct *node,
2219 struct rb_root_cached *root);
2220 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2221 struct vm_area_struct *prev,
2222 struct rb_root_cached *root);
2223 void vma_interval_tree_remove(struct vm_area_struct *node,
2224 struct rb_root_cached *root);
2225 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2226 unsigned long start, unsigned long last);
2227 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2228 unsigned long start, unsigned long last);
2230 #define vma_interval_tree_foreach(vma, root, start, last) \
2231 for (vma = vma_interval_tree_iter_first(root, start, last); \
2232 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2234 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2235 struct rb_root_cached *root);
2236 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2237 struct rb_root_cached *root);
2238 struct anon_vma_chain *
2239 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2240 unsigned long start, unsigned long last);
2241 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2242 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2243 #ifdef CONFIG_DEBUG_VM_RB
2244 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2247 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2248 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2249 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2252 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2253 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2254 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2255 struct vm_area_struct *expand);
2256 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2257 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2259 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2261 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2262 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2263 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2264 struct mempolicy *, struct vm_userfaultfd_ctx);
2265 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2266 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2267 unsigned long addr, int new_below);
2268 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2269 unsigned long addr, int new_below);
2270 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2271 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2272 struct rb_node **, struct rb_node *);
2273 extern void unlink_file_vma(struct vm_area_struct *);
2274 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2275 unsigned long addr, unsigned long len, pgoff_t pgoff,
2276 bool *need_rmap_locks);
2277 extern void exit_mmap(struct mm_struct *);
2279 static inline int check_data_rlimit(unsigned long rlim,
2281 unsigned long start,
2282 unsigned long end_data,
2283 unsigned long start_data)
2285 if (rlim < RLIM_INFINITY) {
2286 if (((new - start) + (end_data - start_data)) > rlim)
2293 extern int mm_take_all_locks(struct mm_struct *mm);
2294 extern void mm_drop_all_locks(struct mm_struct *mm);
2296 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2297 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2298 extern struct file *get_task_exe_file(struct task_struct *task);
2300 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2301 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2303 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2304 const struct vm_special_mapping *sm);
2305 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2306 unsigned long addr, unsigned long len,
2307 unsigned long flags,
2308 const struct vm_special_mapping *spec);
2309 /* This is an obsolete alternative to _install_special_mapping. */
2310 extern int install_special_mapping(struct mm_struct *mm,
2311 unsigned long addr, unsigned long len,
2312 unsigned long flags, struct page **pages);
2314 unsigned long randomize_stack_top(unsigned long stack_top);
2316 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2318 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2319 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2320 struct list_head *uf);
2321 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2322 unsigned long len, unsigned long prot, unsigned long flags,
2323 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate,
2324 struct list_head *uf);
2325 extern int __do_munmap(struct mm_struct *, unsigned long, size_t,
2326 struct list_head *uf, bool downgrade);
2327 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2328 struct list_head *uf);
2329 extern int do_madvise(unsigned long start, size_t len_in, int behavior);
2331 static inline unsigned long
2332 do_mmap_pgoff(struct file *file, unsigned long addr,
2333 unsigned long len, unsigned long prot, unsigned long flags,
2334 unsigned long pgoff, unsigned long *populate,
2335 struct list_head *uf)
2337 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate, uf);
2341 extern int __mm_populate(unsigned long addr, unsigned long len,
2343 static inline void mm_populate(unsigned long addr, unsigned long len)
2346 (void) __mm_populate(addr, len, 1);
2349 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2352 /* These take the mm semaphore themselves */
2353 extern int __must_check vm_brk(unsigned long, unsigned long);
2354 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2355 extern int vm_munmap(unsigned long, size_t);
2356 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2357 unsigned long, unsigned long,
2358 unsigned long, unsigned long);
2360 struct vm_unmapped_area_info {
2361 #define VM_UNMAPPED_AREA_TOPDOWN 1
2362 unsigned long flags;
2363 unsigned long length;
2364 unsigned long low_limit;
2365 unsigned long high_limit;
2366 unsigned long align_mask;
2367 unsigned long align_offset;
2370 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2371 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2374 * Search for an unmapped address range.
2376 * We are looking for a range that:
2377 * - does not intersect with any VMA;
2378 * - is contained within the [low_limit, high_limit) interval;
2379 * - is at least the desired size.
2380 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2382 static inline unsigned long
2383 vm_unmapped_area(struct vm_unmapped_area_info *info)
2385 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2386 return unmapped_area_topdown(info);
2388 return unmapped_area(info);
2392 extern void truncate_inode_pages(struct address_space *, loff_t);
2393 extern void truncate_inode_pages_range(struct address_space *,
2394 loff_t lstart, loff_t lend);
2395 extern void truncate_inode_pages_final(struct address_space *);
2397 /* generic vm_area_ops exported for stackable file systems */
2398 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2399 extern void filemap_map_pages(struct vm_fault *vmf,
2400 pgoff_t start_pgoff, pgoff_t end_pgoff);
2401 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2403 /* mm/page-writeback.c */
2404 int __must_check write_one_page(struct page *page);
2405 void task_dirty_inc(struct task_struct *tsk);
2408 #define VM_READAHEAD_PAGES (SZ_128K / PAGE_SIZE)
2410 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2411 pgoff_t offset, unsigned long nr_to_read);
2413 void page_cache_sync_readahead(struct address_space *mapping,
2414 struct file_ra_state *ra,
2417 unsigned long size);
2419 void page_cache_async_readahead(struct address_space *mapping,
2420 struct file_ra_state *ra,
2424 unsigned long size);
2426 extern unsigned long stack_guard_gap;
2427 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2428 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2430 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2431 extern int expand_downwards(struct vm_area_struct *vma,
2432 unsigned long address);
2434 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2436 #define expand_upwards(vma, address) (0)
2439 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2440 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2441 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2442 struct vm_area_struct **pprev);
2444 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2445 NULL if none. Assume start_addr < end_addr. */
2446 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2448 struct vm_area_struct * vma = find_vma(mm,start_addr);
2450 if (vma && end_addr <= vma->vm_start)
2455 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2457 unsigned long vm_start = vma->vm_start;
2459 if (vma->vm_flags & VM_GROWSDOWN) {
2460 vm_start -= stack_guard_gap;
2461 if (vm_start > vma->vm_start)
2467 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2469 unsigned long vm_end = vma->vm_end;
2471 if (vma->vm_flags & VM_GROWSUP) {
2472 vm_end += stack_guard_gap;
2473 if (vm_end < vma->vm_end)
2474 vm_end = -PAGE_SIZE;
2479 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2481 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2484 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2485 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2486 unsigned long vm_start, unsigned long vm_end)
2488 struct vm_area_struct *vma = find_vma(mm, vm_start);
2490 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2496 static inline bool range_in_vma(struct vm_area_struct *vma,
2497 unsigned long start, unsigned long end)
2499 return (vma && vma->vm_start <= start && end <= vma->vm_end);
2503 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2504 void vma_set_page_prot(struct vm_area_struct *vma);
2506 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2510 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2512 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2516 #ifdef CONFIG_NUMA_BALANCING
2517 unsigned long change_prot_numa(struct vm_area_struct *vma,
2518 unsigned long start, unsigned long end);
2521 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2522 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2523 unsigned long pfn, unsigned long size, pgprot_t);
2524 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2525 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
2527 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
2529 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2531 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2532 unsigned long pfn, pgprot_t pgprot);
2533 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2535 vm_fault_t vmf_insert_mixed_prot(struct vm_area_struct *vma, unsigned long addr,
2536 pfn_t pfn, pgprot_t pgprot);
2537 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
2538 unsigned long addr, pfn_t pfn);
2539 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2541 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2542 unsigned long addr, struct page *page)
2544 int err = vm_insert_page(vma, addr, page);
2547 return VM_FAULT_OOM;
2548 if (err < 0 && err != -EBUSY)
2549 return VM_FAULT_SIGBUS;
2551 return VM_FAULT_NOPAGE;
2554 static inline vm_fault_t vmf_error(int err)
2557 return VM_FAULT_OOM;
2558 return VM_FAULT_SIGBUS;
2561 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
2562 unsigned int foll_flags);
2564 #define FOLL_WRITE 0x01 /* check pte is writable */
2565 #define FOLL_TOUCH 0x02 /* mark page accessed */
2566 #define FOLL_GET 0x04 /* do get_page on page */
2567 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2568 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2569 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2570 * and return without waiting upon it */
2571 #define FOLL_POPULATE 0x40 /* fault in page */
2572 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2573 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2574 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2575 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2576 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2577 #define FOLL_MLOCK 0x1000 /* lock present pages */
2578 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2579 #define FOLL_COW 0x4000 /* internal GUP flag */
2580 #define FOLL_ANON 0x8000 /* don't do file mappings */
2581 #define FOLL_LONGTERM 0x10000 /* mapping lifetime is indefinite: see below */
2582 #define FOLL_SPLIT_PMD 0x20000 /* split huge pmd before returning */
2583 #define FOLL_PIN 0x40000 /* pages must be released via unpin_user_page */
2586 * FOLL_PIN and FOLL_LONGTERM may be used in various combinations with each
2587 * other. Here is what they mean, and how to use them:
2589 * FOLL_LONGTERM indicates that the page will be held for an indefinite time
2590 * period _often_ under userspace control. This is in contrast to
2591 * iov_iter_get_pages(), whose usages are transient.
2593 * FIXME: For pages which are part of a filesystem, mappings are subject to the
2594 * lifetime enforced by the filesystem and we need guarantees that longterm
2595 * users like RDMA and V4L2 only establish mappings which coordinate usage with
2596 * the filesystem. Ideas for this coordination include revoking the longterm
2597 * pin, delaying writeback, bounce buffer page writeback, etc. As FS DAX was
2598 * added after the problem with filesystems was found FS DAX VMAs are
2599 * specifically failed. Filesystem pages are still subject to bugs and use of
2600 * FOLL_LONGTERM should be avoided on those pages.
2602 * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call.
2603 * Currently only get_user_pages() and get_user_pages_fast() support this flag
2604 * and calls to get_user_pages_[un]locked are specifically not allowed. This
2605 * is due to an incompatibility with the FS DAX check and
2606 * FAULT_FLAG_ALLOW_RETRY.
2608 * In the CMA case: long term pins in a CMA region would unnecessarily fragment
2609 * that region. And so, CMA attempts to migrate the page before pinning, when
2610 * FOLL_LONGTERM is specified.
2612 * FOLL_PIN indicates that a special kind of tracking (not just page->_refcount,
2613 * but an additional pin counting system) will be invoked. This is intended for
2614 * anything that gets a page reference and then touches page data (for example,
2615 * Direct IO). This lets the filesystem know that some non-file-system entity is
2616 * potentially changing the pages' data. In contrast to FOLL_GET (whose pages
2617 * are released via put_page()), FOLL_PIN pages must be released, ultimately, by
2618 * a call to unpin_user_page().
2620 * FOLL_PIN is similar to FOLL_GET: both of these pin pages. They use different
2621 * and separate refcounting mechanisms, however, and that means that each has
2622 * its own acquire and release mechanisms:
2624 * FOLL_GET: get_user_pages*() to acquire, and put_page() to release.
2626 * FOLL_PIN: pin_user_pages*() to acquire, and unpin_user_pages to release.
2628 * FOLL_PIN and FOLL_GET are mutually exclusive for a given function call.
2629 * (The underlying pages may experience both FOLL_GET-based and FOLL_PIN-based
2630 * calls applied to them, and that's perfectly OK. This is a constraint on the
2631 * callers, not on the pages.)
2633 * FOLL_PIN should be set internally by the pin_user_pages*() APIs, never
2634 * directly by the caller. That's in order to help avoid mismatches when
2635 * releasing pages: get_user_pages*() pages must be released via put_page(),
2636 * while pin_user_pages*() pages must be released via unpin_user_page().
2638 * Please see Documentation/vm/pin_user_pages.rst for more information.
2641 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
2643 if (vm_fault & VM_FAULT_OOM)
2645 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2646 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2647 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2652 typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
2653 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2654 unsigned long size, pte_fn_t fn, void *data);
2655 extern int apply_to_existing_page_range(struct mm_struct *mm,
2656 unsigned long address, unsigned long size,
2657 pte_fn_t fn, void *data);
2659 #ifdef CONFIG_PAGE_POISONING
2660 extern bool page_poisoning_enabled(void);
2661 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2663 static inline bool page_poisoning_enabled(void) { return false; }
2664 static inline void kernel_poison_pages(struct page *page, int numpages,
2668 #ifdef CONFIG_INIT_ON_ALLOC_DEFAULT_ON
2669 DECLARE_STATIC_KEY_TRUE(init_on_alloc);
2671 DECLARE_STATIC_KEY_FALSE(init_on_alloc);
2673 static inline bool want_init_on_alloc(gfp_t flags)
2675 if (static_branch_unlikely(&init_on_alloc) &&
2676 !page_poisoning_enabled())
2678 return flags & __GFP_ZERO;
2681 #ifdef CONFIG_INIT_ON_FREE_DEFAULT_ON
2682 DECLARE_STATIC_KEY_TRUE(init_on_free);
2684 DECLARE_STATIC_KEY_FALSE(init_on_free);
2686 static inline bool want_init_on_free(void)
2688 return static_branch_unlikely(&init_on_free) &&
2689 !page_poisoning_enabled();
2692 #ifdef CONFIG_DEBUG_PAGEALLOC
2693 extern void init_debug_pagealloc(void);
2695 static inline void init_debug_pagealloc(void) {}
2697 extern bool _debug_pagealloc_enabled_early;
2698 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
2700 static inline bool debug_pagealloc_enabled(void)
2702 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
2703 _debug_pagealloc_enabled_early;
2707 * For use in fast paths after init_debug_pagealloc() has run, or when a
2708 * false negative result is not harmful when called too early.
2710 static inline bool debug_pagealloc_enabled_static(void)
2712 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
2715 return static_branch_unlikely(&_debug_pagealloc_enabled);
2718 #if defined(CONFIG_DEBUG_PAGEALLOC) || defined(CONFIG_ARCH_HAS_SET_DIRECT_MAP)
2719 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2722 kernel_map_pages(struct page *page, int numpages, int enable)
2724 __kernel_map_pages(page, numpages, enable);
2726 #ifdef CONFIG_HIBERNATION
2727 extern bool kernel_page_present(struct page *page);
2728 #endif /* CONFIG_HIBERNATION */
2729 #else /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */
2731 kernel_map_pages(struct page *page, int numpages, int enable) {}
2732 #ifdef CONFIG_HIBERNATION
2733 static inline bool kernel_page_present(struct page *page) { return true; }
2734 #endif /* CONFIG_HIBERNATION */
2735 #endif /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */
2737 #ifdef __HAVE_ARCH_GATE_AREA
2738 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2739 extern int in_gate_area_no_mm(unsigned long addr);
2740 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2742 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2746 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2747 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2751 #endif /* __HAVE_ARCH_GATE_AREA */
2753 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2755 #ifdef CONFIG_SYSCTL
2756 extern int sysctl_drop_caches;
2757 int drop_caches_sysctl_handler(struct ctl_table *, int,
2758 void __user *, size_t *, loff_t *);
2761 void drop_slab(void);
2762 void drop_slab_node(int nid);
2765 #define randomize_va_space 0
2767 extern int randomize_va_space;
2770 const char * arch_vma_name(struct vm_area_struct *vma);
2772 void print_vma_addr(char *prefix, unsigned long rip);
2774 static inline void print_vma_addr(char *prefix, unsigned long rip)
2779 void *sparse_buffer_alloc(unsigned long size);
2780 struct page * __populate_section_memmap(unsigned long pfn,
2781 unsigned long nr_pages, int nid, struct vmem_altmap *altmap);
2782 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2783 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
2784 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
2785 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2786 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2787 void *vmemmap_alloc_block(unsigned long size, int node);
2789 void *vmemmap_alloc_block_buf(unsigned long size, int node);
2790 void *altmap_alloc_block_buf(unsigned long size, struct vmem_altmap *altmap);
2791 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2792 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2794 int vmemmap_populate(unsigned long start, unsigned long end, int node,
2795 struct vmem_altmap *altmap);
2796 void vmemmap_populate_print_last(void);
2797 #ifdef CONFIG_MEMORY_HOTPLUG
2798 void vmemmap_free(unsigned long start, unsigned long end,
2799 struct vmem_altmap *altmap);
2801 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2802 unsigned long nr_pages);
2805 MF_COUNT_INCREASED = 1 << 0,
2806 MF_ACTION_REQUIRED = 1 << 1,
2807 MF_MUST_KILL = 1 << 2,
2808 MF_SOFT_OFFLINE = 1 << 3,
2810 extern int memory_failure(unsigned long pfn, int flags);
2811 extern void memory_failure_queue(unsigned long pfn, int flags);
2812 extern int unpoison_memory(unsigned long pfn);
2813 extern int get_hwpoison_page(struct page *page);
2814 #define put_hwpoison_page(page) put_page(page)
2815 extern int sysctl_memory_failure_early_kill;
2816 extern int sysctl_memory_failure_recovery;
2817 extern void shake_page(struct page *p, int access);
2818 extern atomic_long_t num_poisoned_pages __read_mostly;
2819 extern int soft_offline_page(unsigned long pfn, int flags);
2823 * Error handlers for various types of pages.
2826 MF_IGNORED, /* Error: cannot be handled */
2827 MF_FAILED, /* Error: handling failed */
2828 MF_DELAYED, /* Will be handled later */
2829 MF_RECOVERED, /* Successfully recovered */
2832 enum mf_action_page_type {
2834 MF_MSG_KERNEL_HIGH_ORDER,
2836 MF_MSG_DIFFERENT_COMPOUND,
2837 MF_MSG_POISONED_HUGE,
2840 MF_MSG_NON_PMD_HUGE,
2841 MF_MSG_UNMAP_FAILED,
2842 MF_MSG_DIRTY_SWAPCACHE,
2843 MF_MSG_CLEAN_SWAPCACHE,
2844 MF_MSG_DIRTY_MLOCKED_LRU,
2845 MF_MSG_CLEAN_MLOCKED_LRU,
2846 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2847 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2850 MF_MSG_TRUNCATED_LRU,
2857 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2858 extern void clear_huge_page(struct page *page,
2859 unsigned long addr_hint,
2860 unsigned int pages_per_huge_page);
2861 extern void copy_user_huge_page(struct page *dst, struct page *src,
2862 unsigned long addr_hint,
2863 struct vm_area_struct *vma,
2864 unsigned int pages_per_huge_page);
2865 extern long copy_huge_page_from_user(struct page *dst_page,
2866 const void __user *usr_src,
2867 unsigned int pages_per_huge_page,
2868 bool allow_pagefault);
2869 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2871 #ifdef CONFIG_DEBUG_PAGEALLOC
2872 extern unsigned int _debug_guardpage_minorder;
2873 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
2875 static inline unsigned int debug_guardpage_minorder(void)
2877 return _debug_guardpage_minorder;
2880 static inline bool debug_guardpage_enabled(void)
2882 return static_branch_unlikely(&_debug_guardpage_enabled);
2885 static inline bool page_is_guard(struct page *page)
2887 if (!debug_guardpage_enabled())
2890 return PageGuard(page);
2893 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2894 static inline bool debug_guardpage_enabled(void) { return false; }
2895 static inline bool page_is_guard(struct page *page) { return false; }
2896 #endif /* CONFIG_DEBUG_PAGEALLOC */
2898 #if MAX_NUMNODES > 1
2899 void __init setup_nr_node_ids(void);
2901 static inline void setup_nr_node_ids(void) {}
2904 extern int memcmp_pages(struct page *page1, struct page *page2);
2906 static inline int pages_identical(struct page *page1, struct page *page2)
2908 return !memcmp_pages(page1, page2);
2911 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
2912 unsigned long clean_record_shared_mapping_range(struct address_space *mapping,
2913 pgoff_t first_index, pgoff_t nr,
2914 pgoff_t bitmap_pgoff,
2915 unsigned long *bitmap,
2919 unsigned long wp_shared_mapping_range(struct address_space *mapping,
2920 pgoff_t first_index, pgoff_t nr);
2923 #endif /* __KERNEL__ */
2924 #endif /* _LINUX_MM_H */