1 /* SPDX-License-Identifier: GPL-2.0 */
5 #include <linux/errno.h>
9 #include <linux/mmdebug.h>
10 #include <linux/gfp.h>
11 #include <linux/bug.h>
12 #include <linux/list.h>
13 #include <linux/mmzone.h>
14 #include <linux/rbtree.h>
15 #include <linux/atomic.h>
16 #include <linux/debug_locks.h>
17 #include <linux/mm_types.h>
18 #include <linux/mmap_lock.h>
19 #include <linux/range.h>
20 #include <linux/pfn.h>
21 #include <linux/percpu-refcount.h>
22 #include <linux/bit_spinlock.h>
23 #include <linux/shrinker.h>
24 #include <linux/resource.h>
25 #include <linux/page_ext.h>
26 #include <linux/err.h>
27 #include <linux/page_ref.h>
28 #include <linux/memremap.h>
29 #include <linux/overflow.h>
30 #include <linux/sizes.h>
34 struct anon_vma_chain;
37 struct writeback_control;
40 void init_mm_internals(void);
42 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
43 extern unsigned long max_mapnr;
45 static inline void set_max_mapnr(unsigned long limit)
50 static inline void set_max_mapnr(unsigned long limit) { }
53 extern atomic_long_t _totalram_pages;
54 static inline unsigned long totalram_pages(void)
56 return (unsigned long)atomic_long_read(&_totalram_pages);
59 static inline void totalram_pages_inc(void)
61 atomic_long_inc(&_totalram_pages);
64 static inline void totalram_pages_dec(void)
66 atomic_long_dec(&_totalram_pages);
69 static inline void totalram_pages_add(long count)
71 atomic_long_add(count, &_totalram_pages);
74 static inline void totalram_pages_set(long val)
76 atomic_long_set(&_totalram_pages, val);
79 extern void * high_memory;
80 extern int page_cluster;
83 extern int sysctl_legacy_va_layout;
85 #define sysctl_legacy_va_layout 0
88 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
89 extern const int mmap_rnd_bits_min;
90 extern const int mmap_rnd_bits_max;
91 extern int mmap_rnd_bits __read_mostly;
93 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
94 extern const int mmap_rnd_compat_bits_min;
95 extern const int mmap_rnd_compat_bits_max;
96 extern int mmap_rnd_compat_bits __read_mostly;
100 #include <asm/pgtable.h>
101 #include <asm/processor.h>
104 * Architectures that support memory tagging (assigning tags to memory regions,
105 * embedding these tags into addresses that point to these memory regions, and
106 * checking that the memory and the pointer tags match on memory accesses)
107 * redefine this macro to strip tags from pointers.
108 * It's defined as noop for arcitectures that don't support memory tagging.
110 #ifndef untagged_addr
111 #define untagged_addr(addr) (addr)
115 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
119 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
123 #define lm_alias(x) __va(__pa_symbol(x))
127 * To prevent common memory management code establishing
128 * a zero page mapping on a read fault.
129 * This macro should be defined within <asm/pgtable.h>.
130 * s390 does this to prevent multiplexing of hardware bits
131 * related to the physical page in case of virtualization.
133 #ifndef mm_forbids_zeropage
134 #define mm_forbids_zeropage(X) (0)
138 * On some architectures it is expensive to call memset() for small sizes.
139 * If an architecture decides to implement their own version of
140 * mm_zero_struct_page they should wrap the defines below in a #ifndef and
141 * define their own version of this macro in <asm/pgtable.h>
143 #if BITS_PER_LONG == 64
144 /* This function must be updated when the size of struct page grows above 80
145 * or reduces below 56. The idea that compiler optimizes out switch()
146 * statement, and only leaves move/store instructions. Also the compiler can
147 * combine write statments if they are both assignments and can be reordered,
148 * this can result in several of the writes here being dropped.
150 #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
151 static inline void __mm_zero_struct_page(struct page *page)
153 unsigned long *_pp = (void *)page;
155 /* Check that struct page is either 56, 64, 72, or 80 bytes */
156 BUILD_BUG_ON(sizeof(struct page) & 7);
157 BUILD_BUG_ON(sizeof(struct page) < 56);
158 BUILD_BUG_ON(sizeof(struct page) > 80);
160 switch (sizeof(struct page)) {
162 _pp[9] = 0; /* fallthrough */
164 _pp[8] = 0; /* fallthrough */
166 _pp[7] = 0; /* fallthrough */
178 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
182 * Default maximum number of active map areas, this limits the number of vmas
183 * per mm struct. Users can overwrite this number by sysctl but there is a
186 * When a program's coredump is generated as ELF format, a section is created
187 * per a vma. In ELF, the number of sections is represented in unsigned short.
188 * This means the number of sections should be smaller than 65535 at coredump.
189 * Because the kernel adds some informative sections to a image of program at
190 * generating coredump, we need some margin. The number of extra sections is
191 * 1-3 now and depends on arch. We use "5" as safe margin, here.
193 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
194 * not a hard limit any more. Although some userspace tools can be surprised by
197 #define MAPCOUNT_ELF_CORE_MARGIN (5)
198 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
200 extern int sysctl_max_map_count;
202 extern unsigned long sysctl_user_reserve_kbytes;
203 extern unsigned long sysctl_admin_reserve_kbytes;
205 extern int sysctl_overcommit_memory;
206 extern int sysctl_overcommit_ratio;
207 extern unsigned long sysctl_overcommit_kbytes;
209 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
211 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
214 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
216 /* to align the pointer to the (next) page boundary */
217 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
219 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
220 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
222 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
225 * Linux kernel virtual memory manager primitives.
226 * The idea being to have a "virtual" mm in the same way
227 * we have a virtual fs - giving a cleaner interface to the
228 * mm details, and allowing different kinds of memory mappings
229 * (from shared memory to executable loading to arbitrary
233 struct vm_area_struct *vm_area_alloc(struct mm_struct *);
234 struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
235 void vm_area_free(struct vm_area_struct *);
238 extern struct rb_root nommu_region_tree;
239 extern struct rw_semaphore nommu_region_sem;
241 extern unsigned int kobjsize(const void *objp);
245 * vm_flags in vm_area_struct, see mm_types.h.
246 * When changing, update also include/trace/events/mmflags.h
248 #define VM_NONE 0x00000000
250 #define VM_READ 0x00000001 /* currently active flags */
251 #define VM_WRITE 0x00000002
252 #define VM_EXEC 0x00000004
253 #define VM_SHARED 0x00000008
255 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
256 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
257 #define VM_MAYWRITE 0x00000020
258 #define VM_MAYEXEC 0x00000040
259 #define VM_MAYSHARE 0x00000080
261 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
262 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
263 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
264 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
265 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
267 #define VM_LOCKED 0x00002000
268 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
270 /* Used by sys_madvise() */
271 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
272 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
274 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
275 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
276 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
277 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
278 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
279 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
280 #define VM_SYNC 0x00800000 /* Synchronous page faults */
281 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
282 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
283 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
285 #ifdef CONFIG_MEM_SOFT_DIRTY
286 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
288 # define VM_SOFTDIRTY 0
291 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
292 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
293 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
294 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
296 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
297 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
298 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
299 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
300 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
301 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
302 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
303 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
304 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
305 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
306 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
307 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
309 #ifdef CONFIG_ARCH_HAS_PKEYS
310 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
311 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
312 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
313 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
314 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
316 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
318 # define VM_PKEY_BIT4 0
320 #endif /* CONFIG_ARCH_HAS_PKEYS */
322 #if defined(CONFIG_X86)
323 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
324 #elif defined(CONFIG_PPC)
325 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
326 #elif defined(CONFIG_PARISC)
327 # define VM_GROWSUP VM_ARCH_1
328 #elif defined(CONFIG_IA64)
329 # define VM_GROWSUP VM_ARCH_1
330 #elif defined(CONFIG_SPARC64)
331 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
332 # define VM_ARCH_CLEAR VM_SPARC_ADI
333 #elif !defined(CONFIG_MMU)
334 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
337 #if defined(CONFIG_X86_INTEL_MPX)
338 /* MPX specific bounds table or bounds directory */
339 # define VM_MPX VM_HIGH_ARCH_4
341 # define VM_MPX VM_NONE
345 # define VM_GROWSUP VM_NONE
348 /* Bits set in the VMA until the stack is in its final location */
349 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
351 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
352 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
355 #ifdef CONFIG_STACK_GROWSUP
356 #define VM_STACK VM_GROWSUP
358 #define VM_STACK VM_GROWSDOWN
361 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
364 * Special vmas that are non-mergable, non-mlock()able.
365 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
367 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
369 /* This mask defines which mm->def_flags a process can inherit its parent */
370 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
372 /* This mask is used to clear all the VMA flags used by mlock */
373 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
375 /* Arch-specific flags to clear when updating VM flags on protection change */
376 #ifndef VM_ARCH_CLEAR
377 # define VM_ARCH_CLEAR VM_NONE
379 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
382 * mapping from the currently active vm_flags protection bits (the
383 * low four bits) to a page protection mask..
385 extern pgprot_t protection_map[16];
387 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
388 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
389 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
390 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
391 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
392 #define FAULT_FLAG_TRIED 0x20 /* Second try */
393 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
394 #define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */
395 #define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */
397 #define FAULT_FLAG_TRACE \
398 { FAULT_FLAG_WRITE, "WRITE" }, \
399 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
400 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
401 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
402 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
403 { FAULT_FLAG_TRIED, "TRIED" }, \
404 { FAULT_FLAG_USER, "USER" }, \
405 { FAULT_FLAG_REMOTE, "REMOTE" }, \
406 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }
409 * vm_fault is filled by the the pagefault handler and passed to the vma's
410 * ->fault function. The vma's ->fault is responsible for returning a bitmask
411 * of VM_FAULT_xxx flags that give details about how the fault was handled.
413 * MM layer fills up gfp_mask for page allocations but fault handler might
414 * alter it if its implementation requires a different allocation context.
416 * pgoff should be used in favour of virtual_address, if possible.
419 struct vm_area_struct *vma; /* Target VMA */
420 unsigned int flags; /* FAULT_FLAG_xxx flags */
421 gfp_t gfp_mask; /* gfp mask to be used for allocations */
422 pgoff_t pgoff; /* Logical page offset based on vma */
423 unsigned long address; /* Faulting virtual address */
424 pmd_t *pmd; /* Pointer to pmd entry matching
426 pud_t *pud; /* Pointer to pud entry matching
429 pte_t orig_pte; /* Value of PTE at the time of fault */
431 struct page *cow_page; /* Page handler may use for COW fault */
432 struct mem_cgroup *memcg; /* Cgroup cow_page belongs to */
433 struct page *page; /* ->fault handlers should return a
434 * page here, unless VM_FAULT_NOPAGE
435 * is set (which is also implied by
438 /* These three entries are valid only while holding ptl lock */
439 pte_t *pte; /* Pointer to pte entry matching
440 * the 'address'. NULL if the page
441 * table hasn't been allocated.
443 spinlock_t *ptl; /* Page table lock.
444 * Protects pte page table if 'pte'
445 * is not NULL, otherwise pmd.
447 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
448 * vm_ops->map_pages() calls
449 * alloc_set_pte() from atomic context.
450 * do_fault_around() pre-allocates
451 * page table to avoid allocation from
456 /* page entry size for vm->huge_fault() */
457 enum page_entry_size {
464 * These are the virtual MM functions - opening of an area, closing and
465 * unmapping it (needed to keep files on disk up-to-date etc), pointer
466 * to the functions called when a no-page or a wp-page exception occurs.
468 struct vm_operations_struct {
469 void (*open)(struct vm_area_struct * area);
470 void (*close)(struct vm_area_struct * area);
471 int (*split)(struct vm_area_struct * area, unsigned long addr);
472 int (*mremap)(struct vm_area_struct * area);
473 vm_fault_t (*fault)(struct vm_fault *vmf);
474 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
475 enum page_entry_size pe_size);
476 void (*map_pages)(struct vm_fault *vmf,
477 pgoff_t start_pgoff, pgoff_t end_pgoff);
478 unsigned long (*pagesize)(struct vm_area_struct * area);
480 /* notification that a previously read-only page is about to become
481 * writable, if an error is returned it will cause a SIGBUS */
482 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
484 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
485 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
487 /* called by access_process_vm when get_user_pages() fails, typically
488 * for use by special VMAs that can switch between memory and hardware
490 int (*access)(struct vm_area_struct *vma, unsigned long addr,
491 void *buf, int len, int write);
493 /* Called by the /proc/PID/maps code to ask the vma whether it
494 * has a special name. Returning non-NULL will also cause this
495 * vma to be dumped unconditionally. */
496 const char *(*name)(struct vm_area_struct *vma);
500 * set_policy() op must add a reference to any non-NULL @new mempolicy
501 * to hold the policy upon return. Caller should pass NULL @new to
502 * remove a policy and fall back to surrounding context--i.e. do not
503 * install a MPOL_DEFAULT policy, nor the task or system default
506 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
509 * get_policy() op must add reference [mpol_get()] to any policy at
510 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
511 * in mm/mempolicy.c will do this automatically.
512 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
513 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
514 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
515 * must return NULL--i.e., do not "fallback" to task or system default
518 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
522 * Called by vm_normal_page() for special PTEs to find the
523 * page for @addr. This is useful if the default behavior
524 * (using pte_page()) would not find the correct page.
526 struct page *(*find_special_page)(struct vm_area_struct *vma,
530 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
532 static const struct vm_operations_struct dummy_vm_ops = {};
534 memset(vma, 0, sizeof(*vma));
536 vma->vm_ops = &dummy_vm_ops;
537 INIT_LIST_HEAD(&vma->anon_vma_chain);
540 static inline void vma_set_anonymous(struct vm_area_struct *vma)
545 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
552 * The vma_is_shmem is not inline because it is used only by slow
553 * paths in userfault.
555 bool vma_is_shmem(struct vm_area_struct *vma);
557 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
560 int vma_is_stack_for_current(struct vm_area_struct *vma);
562 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
563 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
568 #if !defined(CONFIG_ARCH_HAS_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
569 static inline int pmd_devmap(pmd_t pmd)
573 static inline int pud_devmap(pud_t pud)
577 static inline int pgd_devmap(pgd_t pgd)
584 * FIXME: take this include out, include page-flags.h in
585 * files which need it (119 of them)
587 #include <linux/page-flags.h>
588 #include <linux/huge_mm.h>
591 * Methods to modify the page usage count.
593 * What counts for a page usage:
594 * - cache mapping (page->mapping)
595 * - private data (page->private)
596 * - page mapped in a task's page tables, each mapping
597 * is counted separately
599 * Also, many kernel routines increase the page count before a critical
600 * routine so they can be sure the page doesn't go away from under them.
604 * Drop a ref, return true if the refcount fell to zero (the page has no users)
606 static inline int put_page_testzero(struct page *page)
608 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
609 return page_ref_dec_and_test(page);
613 * Try to grab a ref unless the page has a refcount of zero, return false if
615 * This can be called when MMU is off so it must not access
616 * any of the virtual mappings.
618 static inline int get_page_unless_zero(struct page *page)
620 return page_ref_add_unless(page, 1, 0);
623 extern int page_is_ram(unsigned long pfn);
631 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
634 /* Support for virtually mapped pages */
635 struct page *vmalloc_to_page(const void *addr);
636 unsigned long vmalloc_to_pfn(const void *addr);
639 * Determine if an address is within the vmalloc range
641 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
642 * is no special casing required.
644 static inline bool is_vmalloc_addr(const void *x)
647 unsigned long addr = (unsigned long)x;
649 return addr >= VMALLOC_START && addr < VMALLOC_END;
655 #ifndef is_ioremap_addr
656 #define is_ioremap_addr(x) is_vmalloc_addr(x)
660 extern int is_vmalloc_or_module_addr(const void *x);
662 static inline int is_vmalloc_or_module_addr(const void *x)
668 extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
669 static inline void *kvmalloc(size_t size, gfp_t flags)
671 return kvmalloc_node(size, flags, NUMA_NO_NODE);
673 static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
675 return kvmalloc_node(size, flags | __GFP_ZERO, node);
677 static inline void *kvzalloc(size_t size, gfp_t flags)
679 return kvmalloc(size, flags | __GFP_ZERO);
682 static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
686 if (unlikely(check_mul_overflow(n, size, &bytes)))
689 return kvmalloc(bytes, flags);
692 static inline void *kvcalloc(size_t n, size_t size, gfp_t flags)
694 return kvmalloc_array(n, size, flags | __GFP_ZERO);
697 extern void kvfree(const void *addr);
698 extern void kvfree_sensitive(const void *addr, size_t len);
701 * Mapcount of compound page as a whole, does not include mapped sub-pages.
703 * Must be called only for compound pages or any their tail sub-pages.
705 static inline int compound_mapcount(struct page *page)
707 VM_BUG_ON_PAGE(!PageCompound(page), page);
708 page = compound_head(page);
709 return atomic_read(compound_mapcount_ptr(page)) + 1;
713 * The atomic page->_mapcount, starts from -1: so that transitions
714 * both from it and to it can be tracked, using atomic_inc_and_test
715 * and atomic_add_negative(-1).
717 static inline void page_mapcount_reset(struct page *page)
719 atomic_set(&(page)->_mapcount, -1);
722 int __page_mapcount(struct page *page);
725 * Mapcount of 0-order page; when compound sub-page, includes
726 * compound_mapcount().
728 * Result is undefined for pages which cannot be mapped into userspace.
729 * For example SLAB or special types of pages. See function page_has_type().
730 * They use this place in struct page differently.
732 static inline int page_mapcount(struct page *page)
734 if (unlikely(PageCompound(page)))
735 return __page_mapcount(page);
736 return atomic_read(&page->_mapcount) + 1;
739 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
740 int total_mapcount(struct page *page);
741 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
743 static inline int total_mapcount(struct page *page)
745 return page_mapcount(page);
747 static inline int page_trans_huge_mapcount(struct page *page,
750 int mapcount = page_mapcount(page);
752 *total_mapcount = mapcount;
757 static inline struct page *virt_to_head_page(const void *x)
759 struct page *page = virt_to_page(x);
761 return compound_head(page);
764 void __put_page(struct page *page);
766 void put_pages_list(struct list_head *pages);
768 void split_page(struct page *page, unsigned int order);
771 * Compound pages have a destructor function. Provide a
772 * prototype for that function and accessor functions.
773 * These are _only_ valid on the head of a compound page.
775 typedef void compound_page_dtor(struct page *);
777 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
778 enum compound_dtor_id {
781 #ifdef CONFIG_HUGETLB_PAGE
784 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
789 extern compound_page_dtor * const compound_page_dtors[];
791 static inline void set_compound_page_dtor(struct page *page,
792 enum compound_dtor_id compound_dtor)
794 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
795 page[1].compound_dtor = compound_dtor;
798 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
800 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
801 return compound_page_dtors[page[1].compound_dtor];
804 static inline unsigned int compound_order(struct page *page)
808 return page[1].compound_order;
811 static inline void set_compound_order(struct page *page, unsigned int order)
813 page[1].compound_order = order;
816 /* Returns the number of pages in this potentially compound page. */
817 static inline unsigned long compound_nr(struct page *page)
819 return 1UL << compound_order(page);
822 /* Returns the number of bytes in this potentially compound page. */
823 static inline unsigned long page_size(struct page *page)
825 return PAGE_SIZE << compound_order(page);
828 /* Returns the number of bits needed for the number of bytes in a page */
829 static inline unsigned int page_shift(struct page *page)
831 return PAGE_SHIFT + compound_order(page);
834 void free_compound_page(struct page *page);
838 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
839 * servicing faults for write access. In the normal case, do always want
840 * pte_mkwrite. But get_user_pages can cause write faults for mappings
841 * that do not have writing enabled, when used by access_process_vm.
843 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
845 if (likely(vma->vm_flags & VM_WRITE))
846 pte = pte_mkwrite(pte);
850 vm_fault_t alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
852 vm_fault_t finish_fault(struct vm_fault *vmf);
853 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
857 * Multiple processes may "see" the same page. E.g. for untouched
858 * mappings of /dev/null, all processes see the same page full of
859 * zeroes, and text pages of executables and shared libraries have
860 * only one copy in memory, at most, normally.
862 * For the non-reserved pages, page_count(page) denotes a reference count.
863 * page_count() == 0 means the page is free. page->lru is then used for
864 * freelist management in the buddy allocator.
865 * page_count() > 0 means the page has been allocated.
867 * Pages are allocated by the slab allocator in order to provide memory
868 * to kmalloc and kmem_cache_alloc. In this case, the management of the
869 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
870 * unless a particular usage is carefully commented. (the responsibility of
871 * freeing the kmalloc memory is the caller's, of course).
873 * A page may be used by anyone else who does a __get_free_page().
874 * In this case, page_count still tracks the references, and should only
875 * be used through the normal accessor functions. The top bits of page->flags
876 * and page->virtual store page management information, but all other fields
877 * are unused and could be used privately, carefully. The management of this
878 * page is the responsibility of the one who allocated it, and those who have
879 * subsequently been given references to it.
881 * The other pages (we may call them "pagecache pages") are completely
882 * managed by the Linux memory manager: I/O, buffers, swapping etc.
883 * The following discussion applies only to them.
885 * A pagecache page contains an opaque `private' member, which belongs to the
886 * page's address_space. Usually, this is the address of a circular list of
887 * the page's disk buffers. PG_private must be set to tell the VM to call
888 * into the filesystem to release these pages.
890 * A page may belong to an inode's memory mapping. In this case, page->mapping
891 * is the pointer to the inode, and page->index is the file offset of the page,
892 * in units of PAGE_SIZE.
894 * If pagecache pages are not associated with an inode, they are said to be
895 * anonymous pages. These may become associated with the swapcache, and in that
896 * case PG_swapcache is set, and page->private is an offset into the swapcache.
898 * In either case (swapcache or inode backed), the pagecache itself holds one
899 * reference to the page. Setting PG_private should also increment the
900 * refcount. The each user mapping also has a reference to the page.
902 * The pagecache pages are stored in a per-mapping radix tree, which is
903 * rooted at mapping->i_pages, and indexed by offset.
904 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
905 * lists, we instead now tag pages as dirty/writeback in the radix tree.
907 * All pagecache pages may be subject to I/O:
908 * - inode pages may need to be read from disk,
909 * - inode pages which have been modified and are MAP_SHARED may need
910 * to be written back to the inode on disk,
911 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
912 * modified may need to be swapped out to swap space and (later) to be read
917 * The zone field is never updated after free_area_init_core()
918 * sets it, so none of the operations on it need to be atomic.
921 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
922 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
923 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
924 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
925 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
926 #define KASAN_TAG_PGOFF (LAST_CPUPID_PGOFF - KASAN_TAG_WIDTH)
929 * Define the bit shifts to access each section. For non-existent
930 * sections we define the shift as 0; that plus a 0 mask ensures
931 * the compiler will optimise away reference to them.
933 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
934 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
935 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
936 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
937 #define KASAN_TAG_PGSHIFT (KASAN_TAG_PGOFF * (KASAN_TAG_WIDTH != 0))
939 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
940 #ifdef NODE_NOT_IN_PAGE_FLAGS
941 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
942 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
943 SECTIONS_PGOFF : ZONES_PGOFF)
945 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
946 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
947 NODES_PGOFF : ZONES_PGOFF)
950 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
952 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
953 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
956 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
957 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
958 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
959 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
960 #define KASAN_TAG_MASK ((1UL << KASAN_TAG_WIDTH) - 1)
961 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
963 static inline enum zone_type page_zonenum(const struct page *page)
965 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
968 #ifdef CONFIG_ZONE_DEVICE
969 static inline bool is_zone_device_page(const struct page *page)
971 return page_zonenum(page) == ZONE_DEVICE;
973 extern void memmap_init_zone_device(struct zone *, unsigned long,
974 unsigned long, struct dev_pagemap *);
976 static inline bool is_zone_device_page(const struct page *page)
982 #ifdef CONFIG_DEV_PAGEMAP_OPS
983 void __put_devmap_managed_page(struct page *page);
984 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
985 static inline bool put_devmap_managed_page(struct page *page)
987 if (!static_branch_unlikely(&devmap_managed_key))
989 if (!is_zone_device_page(page))
991 switch (page->pgmap->type) {
992 case MEMORY_DEVICE_PRIVATE:
993 case MEMORY_DEVICE_FS_DAX:
994 __put_devmap_managed_page(page);
1002 #else /* CONFIG_DEV_PAGEMAP_OPS */
1003 static inline bool put_devmap_managed_page(struct page *page)
1007 #endif /* CONFIG_DEV_PAGEMAP_OPS */
1009 static inline bool is_device_private_page(const struct page *page)
1011 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
1012 IS_ENABLED(CONFIG_DEVICE_PRIVATE) &&
1013 is_zone_device_page(page) &&
1014 page->pgmap->type == MEMORY_DEVICE_PRIVATE;
1017 static inline bool is_pci_p2pdma_page(const struct page *page)
1019 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
1020 IS_ENABLED(CONFIG_PCI_P2PDMA) &&
1021 is_zone_device_page(page) &&
1022 page->pgmap->type == MEMORY_DEVICE_PCI_P2PDMA;
1025 /* 127: arbitrary random number, small enough to assemble well */
1026 #define page_ref_zero_or_close_to_overflow(page) \
1027 ((unsigned int) page_ref_count(page) + 127u <= 127u)
1029 static inline void get_page(struct page *page)
1031 page = compound_head(page);
1033 * Getting a normal page or the head of a compound page
1034 * requires to already have an elevated page->_refcount.
1036 VM_BUG_ON_PAGE(page_ref_zero_or_close_to_overflow(page), page);
1040 static inline __must_check bool try_get_page(struct page *page)
1042 page = compound_head(page);
1043 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1049 static inline void put_page(struct page *page)
1051 page = compound_head(page);
1054 * For devmap managed pages we need to catch refcount transition from
1055 * 2 to 1, when refcount reach one it means the page is free and we
1056 * need to inform the device driver through callback. See
1057 * include/linux/memremap.h and HMM for details.
1059 if (put_devmap_managed_page(page))
1062 if (put_page_testzero(page))
1067 * put_user_page() - release a gup-pinned page
1068 * @page: pointer to page to be released
1070 * Pages that were pinned via get_user_pages*() must be released via
1071 * either put_user_page(), or one of the put_user_pages*() routines
1072 * below. This is so that eventually, pages that are pinned via
1073 * get_user_pages*() can be separately tracked and uniquely handled. In
1074 * particular, interactions with RDMA and filesystems need special
1077 * put_user_page() and put_page() are not interchangeable, despite this early
1078 * implementation that makes them look the same. put_user_page() calls must
1079 * be perfectly matched up with get_user_page() calls.
1081 static inline void put_user_page(struct page *page)
1086 void put_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1089 void put_user_pages(struct page **pages, unsigned long npages);
1091 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1092 #define SECTION_IN_PAGE_FLAGS
1096 * The identification function is mainly used by the buddy allocator for
1097 * determining if two pages could be buddies. We are not really identifying
1098 * the zone since we could be using the section number id if we do not have
1099 * node id available in page flags.
1100 * We only guarantee that it will return the same value for two combinable
1103 static inline int page_zone_id(struct page *page)
1105 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1108 #ifdef NODE_NOT_IN_PAGE_FLAGS
1109 extern int page_to_nid(const struct page *page);
1111 static inline int page_to_nid(const struct page *page)
1113 struct page *p = (struct page *)page;
1115 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1119 #ifdef CONFIG_NUMA_BALANCING
1120 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1122 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1125 static inline int cpupid_to_pid(int cpupid)
1127 return cpupid & LAST__PID_MASK;
1130 static inline int cpupid_to_cpu(int cpupid)
1132 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1135 static inline int cpupid_to_nid(int cpupid)
1137 return cpu_to_node(cpupid_to_cpu(cpupid));
1140 static inline bool cpupid_pid_unset(int cpupid)
1142 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1145 static inline bool cpupid_cpu_unset(int cpupid)
1147 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1150 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1152 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1155 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1156 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1157 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1159 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1162 static inline int page_cpupid_last(struct page *page)
1164 return page->_last_cpupid;
1166 static inline void page_cpupid_reset_last(struct page *page)
1168 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1171 static inline int page_cpupid_last(struct page *page)
1173 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1176 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1178 static inline void page_cpupid_reset_last(struct page *page)
1180 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1182 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1183 #else /* !CONFIG_NUMA_BALANCING */
1184 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1186 return page_to_nid(page); /* XXX */
1189 static inline int page_cpupid_last(struct page *page)
1191 return page_to_nid(page); /* XXX */
1194 static inline int cpupid_to_nid(int cpupid)
1199 static inline int cpupid_to_pid(int cpupid)
1204 static inline int cpupid_to_cpu(int cpupid)
1209 static inline int cpu_pid_to_cpupid(int nid, int pid)
1214 static inline bool cpupid_pid_unset(int cpupid)
1219 static inline void page_cpupid_reset_last(struct page *page)
1223 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1227 #endif /* CONFIG_NUMA_BALANCING */
1229 #ifdef CONFIG_KASAN_SW_TAGS
1232 * KASAN per-page tags are stored xor'ed with 0xff. This allows to avoid
1233 * setting tags for all pages to native kernel tag value 0xff, as the default
1234 * value 0x00 maps to 0xff.
1237 static inline u8 page_kasan_tag(const struct page *page)
1241 tag = (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1247 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1250 page->flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1251 page->flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1254 static inline void page_kasan_tag_reset(struct page *page)
1256 page_kasan_tag_set(page, 0xff);
1259 static inline u8 page_kasan_tag(const struct page *page)
1264 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1265 static inline void page_kasan_tag_reset(struct page *page) { }
1268 static inline struct zone *page_zone(const struct page *page)
1270 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1273 static inline pg_data_t *page_pgdat(const struct page *page)
1275 return NODE_DATA(page_to_nid(page));
1278 #ifdef SECTION_IN_PAGE_FLAGS
1279 static inline void set_page_section(struct page *page, unsigned long section)
1281 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1282 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1285 static inline unsigned long page_to_section(const struct page *page)
1287 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1291 static inline void set_page_zone(struct page *page, enum zone_type zone)
1293 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1294 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1297 static inline void set_page_node(struct page *page, unsigned long node)
1299 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1300 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1303 static inline void set_page_links(struct page *page, enum zone_type zone,
1304 unsigned long node, unsigned long pfn)
1306 set_page_zone(page, zone);
1307 set_page_node(page, node);
1308 #ifdef SECTION_IN_PAGE_FLAGS
1309 set_page_section(page, pfn_to_section_nr(pfn));
1314 static inline struct mem_cgroup *page_memcg(struct page *page)
1316 return page->mem_cgroup;
1318 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1320 WARN_ON_ONCE(!rcu_read_lock_held());
1321 return READ_ONCE(page->mem_cgroup);
1324 static inline struct mem_cgroup *page_memcg(struct page *page)
1328 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1330 WARN_ON_ONCE(!rcu_read_lock_held());
1336 * Some inline functions in vmstat.h depend on page_zone()
1338 #include <linux/vmstat.h>
1340 static __always_inline void *lowmem_page_address(const struct page *page)
1342 return page_to_virt(page);
1345 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1346 #define HASHED_PAGE_VIRTUAL
1349 #if defined(WANT_PAGE_VIRTUAL)
1350 static inline void *page_address(const struct page *page)
1352 return page->virtual;
1354 static inline void set_page_address(struct page *page, void *address)
1356 page->virtual = address;
1358 #define page_address_init() do { } while(0)
1361 #if defined(HASHED_PAGE_VIRTUAL)
1362 void *page_address(const struct page *page);
1363 void set_page_address(struct page *page, void *virtual);
1364 void page_address_init(void);
1367 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1368 #define page_address(page) lowmem_page_address(page)
1369 #define set_page_address(page, address) do { } while(0)
1370 #define page_address_init() do { } while(0)
1373 extern void *page_rmapping(struct page *page);
1374 extern struct anon_vma *page_anon_vma(struct page *page);
1375 extern struct address_space *page_mapping(struct page *page);
1377 extern struct address_space *__page_file_mapping(struct page *);
1380 struct address_space *page_file_mapping(struct page *page)
1382 if (unlikely(PageSwapCache(page)))
1383 return __page_file_mapping(page);
1385 return page->mapping;
1388 extern pgoff_t __page_file_index(struct page *page);
1391 * Return the pagecache index of the passed page. Regular pagecache pages
1392 * use ->index whereas swapcache pages use swp_offset(->private)
1394 static inline pgoff_t page_index(struct page *page)
1396 if (unlikely(PageSwapCache(page)))
1397 return __page_file_index(page);
1401 bool page_mapped(struct page *page);
1402 struct address_space *page_mapping(struct page *page);
1403 struct address_space *page_mapping_file(struct page *page);
1406 * Return true only if the page has been allocated with
1407 * ALLOC_NO_WATERMARKS and the low watermark was not
1408 * met implying that the system is under some pressure.
1410 static inline bool page_is_pfmemalloc(struct page *page)
1413 * Page index cannot be this large so this must be
1414 * a pfmemalloc page.
1416 return page->index == -1UL;
1420 * Only to be called by the page allocator on a freshly allocated
1423 static inline void set_page_pfmemalloc(struct page *page)
1428 static inline void clear_page_pfmemalloc(struct page *page)
1434 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1436 extern void pagefault_out_of_memory(void);
1438 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1441 * Flags passed to show_mem() and show_free_areas() to suppress output in
1444 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1446 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1449 extern bool can_do_mlock(void);
1451 static inline bool can_do_mlock(void) { return false; }
1453 extern int user_shm_lock(size_t, struct user_struct *);
1454 extern void user_shm_unlock(size_t, struct user_struct *);
1457 * Parameter block passed down to zap_pte_range in exceptional cases.
1459 struct zap_details {
1460 struct address_space *check_mapping; /* Check page->mapping if set */
1461 pgoff_t first_index; /* Lowest page->index to unmap */
1462 pgoff_t last_index; /* Highest page->index to unmap */
1463 struct page *single_page; /* Locked page to be unmapped */
1466 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1468 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1471 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1472 unsigned long size);
1473 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1474 unsigned long size);
1475 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1476 unsigned long start, unsigned long end);
1478 struct mmu_notifier_range;
1480 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1481 unsigned long end, unsigned long floor, unsigned long ceiling);
1482 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1483 struct vm_area_struct *vma);
1484 int follow_invalidate_pte(struct mm_struct *mm, unsigned long address,
1485 struct mmu_notifier_range *range, pte_t **ptepp,
1486 pmd_t **pmdpp, spinlock_t **ptlp);
1487 int follow_pte(struct mm_struct *mm, unsigned long address,
1488 pte_t **ptepp, spinlock_t **ptlp);
1489 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1490 unsigned long *pfn);
1491 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1492 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1493 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1494 void *buf, int len, int write);
1496 extern void truncate_pagecache(struct inode *inode, loff_t new);
1497 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1498 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1499 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1500 int truncate_inode_page(struct address_space *mapping, struct page *page);
1501 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1502 int invalidate_inode_page(struct page *page);
1505 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1506 unsigned long address, unsigned int flags);
1507 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1508 unsigned long address, unsigned int fault_flags,
1510 void unmap_mapping_page(struct page *page);
1511 void unmap_mapping_pages(struct address_space *mapping,
1512 pgoff_t start, pgoff_t nr, bool even_cows);
1513 void unmap_mapping_range(struct address_space *mapping,
1514 loff_t const holebegin, loff_t const holelen, int even_cows);
1516 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1517 unsigned long address, unsigned int flags)
1519 /* should never happen if there's no MMU */
1521 return VM_FAULT_SIGBUS;
1523 static inline int fixup_user_fault(struct task_struct *tsk,
1524 struct mm_struct *mm, unsigned long address,
1525 unsigned int fault_flags, bool *unlocked)
1527 /* should never happen if there's no MMU */
1531 static inline void unmap_mapping_page(struct page *page) { }
1532 static inline void unmap_mapping_pages(struct address_space *mapping,
1533 pgoff_t start, pgoff_t nr, bool even_cows) { }
1534 static inline void unmap_mapping_range(struct address_space *mapping,
1535 loff_t const holebegin, loff_t const holelen, int even_cows) { }
1538 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1539 loff_t const holebegin, loff_t const holelen)
1541 unmap_mapping_range(mapping, holebegin, holelen, 0);
1544 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1545 void *buf, int len, unsigned int gup_flags);
1546 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1547 void *buf, int len, unsigned int gup_flags);
1548 extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1549 unsigned long addr, void *buf, int len, unsigned int gup_flags);
1551 long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1552 unsigned long start, unsigned long nr_pages,
1553 unsigned int gup_flags, struct page **pages,
1554 struct vm_area_struct **vmas, int *locked);
1555 long get_user_pages(unsigned long start, unsigned long nr_pages,
1556 unsigned int gup_flags, struct page **pages,
1557 struct vm_area_struct **vmas);
1558 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1559 unsigned int gup_flags, struct page **pages, int *locked);
1560 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1561 struct page **pages, unsigned int gup_flags);
1563 int get_user_pages_fast(unsigned long start, int nr_pages,
1564 unsigned int gup_flags, struct page **pages);
1566 int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
1567 int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
1568 struct task_struct *task, bool bypass_rlim);
1570 /* Container for pinned pfns / pages */
1571 struct frame_vector {
1572 unsigned int nr_allocated; /* Number of frames we have space for */
1573 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1574 bool got_ref; /* Did we pin pages by getting page ref? */
1575 bool is_pfns; /* Does array contain pages or pfns? */
1576 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1577 * pfns_vector_pages() or pfns_vector_pfns()
1581 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1582 void frame_vector_destroy(struct frame_vector *vec);
1583 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1584 unsigned int gup_flags, struct frame_vector *vec);
1585 void put_vaddr_frames(struct frame_vector *vec);
1586 int frame_vector_to_pages(struct frame_vector *vec);
1587 void frame_vector_to_pfns(struct frame_vector *vec);
1589 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1591 return vec->nr_frames;
1594 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1597 int err = frame_vector_to_pages(vec);
1600 return ERR_PTR(err);
1602 return (struct page **)(vec->ptrs);
1605 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1608 frame_vector_to_pfns(vec);
1609 return (unsigned long *)(vec->ptrs);
1613 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1614 struct page **pages);
1615 int get_kernel_page(unsigned long start, int write, struct page **pages);
1616 struct page *get_dump_page(unsigned long addr);
1618 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1619 extern void do_invalidatepage(struct page *page, unsigned int offset,
1620 unsigned int length);
1622 void __set_page_dirty(struct page *, struct address_space *, int warn);
1623 int __set_page_dirty_nobuffers(struct page *page);
1624 int __set_page_dirty_no_writeback(struct page *page);
1625 int redirty_page_for_writepage(struct writeback_control *wbc,
1627 void account_page_dirtied(struct page *page, struct address_space *mapping);
1628 void account_page_cleaned(struct page *page, struct address_space *mapping,
1629 struct bdi_writeback *wb);
1630 int set_page_dirty(struct page *page);
1631 int set_page_dirty_lock(struct page *page);
1632 void __cancel_dirty_page(struct page *page);
1633 static inline void cancel_dirty_page(struct page *page)
1635 /* Avoid atomic ops, locking, etc. when not actually needed. */
1636 if (PageDirty(page))
1637 __cancel_dirty_page(page);
1639 int clear_page_dirty_for_io(struct page *page);
1641 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1643 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1644 unsigned long old_addr, struct vm_area_struct *new_vma,
1645 unsigned long new_addr, unsigned long len,
1646 bool need_rmap_locks);
1647 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1648 unsigned long end, pgprot_t newprot,
1649 int dirty_accountable, int prot_numa);
1650 extern int mprotect_fixup(struct vm_area_struct *vma,
1651 struct vm_area_struct **pprev, unsigned long start,
1652 unsigned long end, unsigned long newflags);
1655 * doesn't attempt to fault and will return short.
1657 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1658 struct page **pages);
1660 * per-process(per-mm_struct) statistics.
1662 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1664 long val = atomic_long_read(&mm->rss_stat.count[member]);
1666 #ifdef SPLIT_RSS_COUNTING
1668 * counter is updated in asynchronous manner and may go to minus.
1669 * But it's never be expected number for users.
1674 return (unsigned long)val;
1677 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1679 atomic_long_add(value, &mm->rss_stat.count[member]);
1682 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1684 atomic_long_inc(&mm->rss_stat.count[member]);
1687 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1689 atomic_long_dec(&mm->rss_stat.count[member]);
1692 /* Optimized variant when page is already known not to be PageAnon */
1693 static inline int mm_counter_file(struct page *page)
1695 if (PageSwapBacked(page))
1696 return MM_SHMEMPAGES;
1697 return MM_FILEPAGES;
1700 static inline int mm_counter(struct page *page)
1703 return MM_ANONPAGES;
1704 return mm_counter_file(page);
1707 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1709 return get_mm_counter(mm, MM_FILEPAGES) +
1710 get_mm_counter(mm, MM_ANONPAGES) +
1711 get_mm_counter(mm, MM_SHMEMPAGES);
1714 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1716 return max(mm->hiwater_rss, get_mm_rss(mm));
1719 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1721 return max(mm->hiwater_vm, mm->total_vm);
1724 static inline void update_hiwater_rss(struct mm_struct *mm)
1726 unsigned long _rss = get_mm_rss(mm);
1728 if ((mm)->hiwater_rss < _rss)
1729 (mm)->hiwater_rss = _rss;
1732 static inline void update_hiwater_vm(struct mm_struct *mm)
1734 if (mm->hiwater_vm < mm->total_vm)
1735 mm->hiwater_vm = mm->total_vm;
1738 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1740 mm->hiwater_rss = get_mm_rss(mm);
1743 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1744 struct mm_struct *mm)
1746 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1748 if (*maxrss < hiwater_rss)
1749 *maxrss = hiwater_rss;
1752 #if defined(SPLIT_RSS_COUNTING)
1753 void sync_mm_rss(struct mm_struct *mm);
1755 static inline void sync_mm_rss(struct mm_struct *mm)
1760 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
1761 static inline int pte_devmap(pte_t pte)
1767 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1769 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1771 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1775 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1779 #ifdef __PAGETABLE_P4D_FOLDED
1780 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1781 unsigned long address)
1786 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1789 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
1790 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1791 unsigned long address)
1795 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
1796 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
1799 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
1801 static inline void mm_inc_nr_puds(struct mm_struct *mm)
1803 if (mm_pud_folded(mm))
1805 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1808 static inline void mm_dec_nr_puds(struct mm_struct *mm)
1810 if (mm_pud_folded(mm))
1812 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1816 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1817 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1818 unsigned long address)
1823 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1824 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1827 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1829 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1831 if (mm_pmd_folded(mm))
1833 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1836 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1838 if (mm_pmd_folded(mm))
1840 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1845 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
1847 atomic_long_set(&mm->pgtables_bytes, 0);
1850 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1852 return atomic_long_read(&mm->pgtables_bytes);
1855 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
1857 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1860 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
1862 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1866 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
1867 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1872 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
1873 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
1876 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
1877 int __pte_alloc_kernel(pmd_t *pmd);
1880 * The following ifdef needed to get the 4level-fixup.h header to work.
1881 * Remove it when 4level-fixup.h has been removed.
1883 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1885 #ifndef __ARCH_HAS_5LEVEL_HACK
1886 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1887 unsigned long address)
1889 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
1890 NULL : p4d_offset(pgd, address);
1893 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1894 unsigned long address)
1896 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
1897 NULL : pud_offset(p4d, address);
1899 #endif /* !__ARCH_HAS_5LEVEL_HACK */
1901 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1903 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1904 NULL: pmd_offset(pud, address);
1906 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1908 #if USE_SPLIT_PTE_PTLOCKS
1909 #if ALLOC_SPLIT_PTLOCKS
1910 void __init ptlock_cache_init(void);
1911 extern bool ptlock_alloc(struct page *page);
1912 extern void ptlock_free(struct page *page);
1914 static inline spinlock_t *ptlock_ptr(struct page *page)
1918 #else /* ALLOC_SPLIT_PTLOCKS */
1919 static inline void ptlock_cache_init(void)
1923 static inline bool ptlock_alloc(struct page *page)
1928 static inline void ptlock_free(struct page *page)
1932 static inline spinlock_t *ptlock_ptr(struct page *page)
1936 #endif /* ALLOC_SPLIT_PTLOCKS */
1938 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1940 return ptlock_ptr(pmd_page(*pmd));
1943 static inline bool ptlock_init(struct page *page)
1946 * prep_new_page() initialize page->private (and therefore page->ptl)
1947 * with 0. Make sure nobody took it in use in between.
1949 * It can happen if arch try to use slab for page table allocation:
1950 * slab code uses page->slab_cache, which share storage with page->ptl.
1952 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1953 if (!ptlock_alloc(page))
1955 spin_lock_init(ptlock_ptr(page));
1959 #else /* !USE_SPLIT_PTE_PTLOCKS */
1961 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1963 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1965 return &mm->page_table_lock;
1967 static inline void ptlock_cache_init(void) {}
1968 static inline bool ptlock_init(struct page *page) { return true; }
1969 static inline void ptlock_free(struct page *page) {}
1970 #endif /* USE_SPLIT_PTE_PTLOCKS */
1972 static inline void pgtable_init(void)
1974 ptlock_cache_init();
1975 pgtable_cache_init();
1978 static inline bool pgtable_pte_page_ctor(struct page *page)
1980 if (!ptlock_init(page))
1982 __SetPageTable(page);
1983 inc_zone_page_state(page, NR_PAGETABLE);
1987 static inline void pgtable_pte_page_dtor(struct page *page)
1990 __ClearPageTable(page);
1991 dec_zone_page_state(page, NR_PAGETABLE);
1994 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1996 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1997 pte_t *__pte = pte_offset_map(pmd, address); \
2003 #define pte_unmap_unlock(pte, ptl) do { \
2008 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2010 #define pte_alloc_map(mm, pmd, address) \
2011 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2013 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2014 (pte_alloc(mm, pmd) ? \
2015 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2017 #define pte_alloc_kernel(pmd, address) \
2018 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2019 NULL: pte_offset_kernel(pmd, address))
2021 #if USE_SPLIT_PMD_PTLOCKS
2023 static struct page *pmd_to_page(pmd_t *pmd)
2025 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2026 return virt_to_page((void *)((unsigned long) pmd & mask));
2029 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2031 return ptlock_ptr(pmd_to_page(pmd));
2034 static inline bool pgtable_pmd_page_ctor(struct page *page)
2036 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2037 page->pmd_huge_pte = NULL;
2039 return ptlock_init(page);
2042 static inline void pgtable_pmd_page_dtor(struct page *page)
2044 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2045 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2050 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2054 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2056 return &mm->page_table_lock;
2059 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
2060 static inline void pgtable_pmd_page_dtor(struct page *page) {}
2062 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2066 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2068 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2074 * No scalability reason to split PUD locks yet, but follow the same pattern
2075 * as the PMD locks to make it easier if we decide to. The VM should not be
2076 * considered ready to switch to split PUD locks yet; there may be places
2077 * which need to be converted from page_table_lock.
2079 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2081 return &mm->page_table_lock;
2084 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2086 spinlock_t *ptl = pud_lockptr(mm, pud);
2092 extern void __init pagecache_init(void);
2093 extern void free_area_init(unsigned long * zones_size);
2094 extern void __init free_area_init_node(int nid, unsigned long * zones_size,
2095 unsigned long zone_start_pfn, unsigned long *zholes_size);
2096 extern void free_initmem(void);
2099 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2100 * into the buddy system. The freed pages will be poisoned with pattern
2101 * "poison" if it's within range [0, UCHAR_MAX].
2102 * Return pages freed into the buddy system.
2104 extern unsigned long free_reserved_area(void *start, void *end,
2105 int poison, const char *s);
2107 #ifdef CONFIG_HIGHMEM
2109 * Free a highmem page into the buddy system, adjusting totalhigh_pages
2110 * and totalram_pages.
2112 extern void free_highmem_page(struct page *page);
2115 extern void adjust_managed_page_count(struct page *page, long count);
2116 extern void mem_init_print_info(const char *str);
2118 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2120 /* Free the reserved page into the buddy system, so it gets managed. */
2121 static inline void __free_reserved_page(struct page *page)
2123 ClearPageReserved(page);
2124 init_page_count(page);
2128 static inline void free_reserved_page(struct page *page)
2130 __free_reserved_page(page);
2131 adjust_managed_page_count(page, 1);
2134 static inline void mark_page_reserved(struct page *page)
2136 SetPageReserved(page);
2137 adjust_managed_page_count(page, -1);
2141 * Default method to free all the __init memory into the buddy system.
2142 * The freed pages will be poisoned with pattern "poison" if it's within
2143 * range [0, UCHAR_MAX].
2144 * Return pages freed into the buddy system.
2146 static inline unsigned long free_initmem_default(int poison)
2148 extern char __init_begin[], __init_end[];
2150 return free_reserved_area(&__init_begin, &__init_end,
2151 poison, "unused kernel");
2154 static inline unsigned long get_num_physpages(void)
2157 unsigned long phys_pages = 0;
2159 for_each_online_node(nid)
2160 phys_pages += node_present_pages(nid);
2165 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
2167 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
2168 * zones, allocate the backing mem_map and account for memory holes in a more
2169 * architecture independent manner. This is a substitute for creating the
2170 * zone_sizes[] and zholes_size[] arrays and passing them to
2171 * free_area_init_node()
2173 * An architecture is expected to register range of page frames backed by
2174 * physical memory with memblock_add[_node]() before calling
2175 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
2176 * usage, an architecture is expected to do something like
2178 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2180 * for_each_valid_physical_page_range()
2181 * memblock_add_node(base, size, nid)
2182 * free_area_init_nodes(max_zone_pfns);
2184 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
2185 * registered physical page range. Similarly
2186 * sparse_memory_present_with_active_regions() calls memory_present() for
2187 * each range when SPARSEMEM is enabled.
2189 * See mm/page_alloc.c for more information on each function exposed by
2190 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
2192 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
2193 unsigned long node_map_pfn_alignment(void);
2194 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2195 unsigned long end_pfn);
2196 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2197 unsigned long end_pfn);
2198 extern void get_pfn_range_for_nid(unsigned int nid,
2199 unsigned long *start_pfn, unsigned long *end_pfn);
2200 extern unsigned long find_min_pfn_with_active_regions(void);
2201 extern void free_bootmem_with_active_regions(int nid,
2202 unsigned long max_low_pfn);
2203 extern void sparse_memory_present_with_active_regions(int nid);
2205 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
2207 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
2208 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
2209 static inline int __early_pfn_to_nid(unsigned long pfn,
2210 struct mminit_pfnnid_cache *state)
2215 /* please see mm/page_alloc.c */
2216 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2217 /* there is a per-arch backend function. */
2218 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
2219 struct mminit_pfnnid_cache *state);
2222 #if !defined(CONFIG_FLAT_NODE_MEM_MAP)
2223 void zero_resv_unavail(void);
2225 static inline void zero_resv_unavail(void) {}
2228 extern void set_dma_reserve(unsigned long new_dma_reserve);
2229 extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long,
2230 enum meminit_context, struct vmem_altmap *);
2231 extern void setup_per_zone_wmarks(void);
2232 extern int __meminit init_per_zone_wmark_min(void);
2233 extern void mem_init(void);
2234 extern void __init mmap_init(void);
2235 extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2236 extern long si_mem_available(void);
2237 extern void si_meminfo(struct sysinfo * val);
2238 extern void si_meminfo_node(struct sysinfo *val, int nid);
2239 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2240 extern unsigned long arch_reserved_kernel_pages(void);
2243 extern __printf(3, 4)
2244 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2246 extern void setup_per_cpu_pageset(void);
2248 extern void zone_pcp_update(struct zone *zone);
2249 extern void zone_pcp_reset(struct zone *zone);
2252 extern int min_free_kbytes;
2253 extern int watermark_boost_factor;
2254 extern int watermark_scale_factor;
2257 extern atomic_long_t mmap_pages_allocated;
2258 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2260 /* interval_tree.c */
2261 void vma_interval_tree_insert(struct vm_area_struct *node,
2262 struct rb_root_cached *root);
2263 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2264 struct vm_area_struct *prev,
2265 struct rb_root_cached *root);
2266 void vma_interval_tree_remove(struct vm_area_struct *node,
2267 struct rb_root_cached *root);
2268 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2269 unsigned long start, unsigned long last);
2270 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2271 unsigned long start, unsigned long last);
2273 #define vma_interval_tree_foreach(vma, root, start, last) \
2274 for (vma = vma_interval_tree_iter_first(root, start, last); \
2275 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2277 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2278 struct rb_root_cached *root);
2279 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2280 struct rb_root_cached *root);
2281 struct anon_vma_chain *
2282 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2283 unsigned long start, unsigned long last);
2284 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2285 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2286 #ifdef CONFIG_DEBUG_VM_RB
2287 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2290 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2291 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2292 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2295 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2296 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2297 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2298 struct vm_area_struct *expand);
2299 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2300 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2302 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2304 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2305 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2306 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2307 struct mempolicy *, struct vm_userfaultfd_ctx);
2308 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2309 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2310 unsigned long addr, int new_below);
2311 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2312 unsigned long addr, int new_below);
2313 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2314 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2315 struct rb_node **, struct rb_node *);
2316 extern void unlink_file_vma(struct vm_area_struct *);
2317 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2318 unsigned long addr, unsigned long len, pgoff_t pgoff,
2319 bool *need_rmap_locks);
2320 extern void exit_mmap(struct mm_struct *);
2322 static inline int check_data_rlimit(unsigned long rlim,
2324 unsigned long start,
2325 unsigned long end_data,
2326 unsigned long start_data)
2328 if (rlim < RLIM_INFINITY) {
2329 if (((new - start) + (end_data - start_data)) > rlim)
2336 extern int mm_take_all_locks(struct mm_struct *mm);
2337 extern void mm_drop_all_locks(struct mm_struct *mm);
2339 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2340 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2341 extern struct file *get_task_exe_file(struct task_struct *task);
2343 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2344 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2346 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2347 const struct vm_special_mapping *sm);
2348 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2349 unsigned long addr, unsigned long len,
2350 unsigned long flags,
2351 const struct vm_special_mapping *spec);
2352 /* This is an obsolete alternative to _install_special_mapping. */
2353 extern int install_special_mapping(struct mm_struct *mm,
2354 unsigned long addr, unsigned long len,
2355 unsigned long flags, struct page **pages);
2357 unsigned long randomize_stack_top(unsigned long stack_top);
2358 unsigned long randomize_page(unsigned long start, unsigned long range);
2360 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2362 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2363 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2364 struct list_head *uf);
2365 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2366 unsigned long len, unsigned long prot, unsigned long flags,
2367 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate,
2368 struct list_head *uf);
2369 extern int __do_munmap(struct mm_struct *, unsigned long, size_t,
2370 struct list_head *uf, bool downgrade);
2371 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2372 struct list_head *uf);
2374 static inline unsigned long
2375 do_mmap_pgoff(struct file *file, unsigned long addr,
2376 unsigned long len, unsigned long prot, unsigned long flags,
2377 unsigned long pgoff, unsigned long *populate,
2378 struct list_head *uf)
2380 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate, uf);
2384 extern int __mm_populate(unsigned long addr, unsigned long len,
2386 static inline void mm_populate(unsigned long addr, unsigned long len)
2389 (void) __mm_populate(addr, len, 1);
2392 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2395 /* These take the mm semaphore themselves */
2396 extern int __must_check vm_brk(unsigned long, unsigned long);
2397 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2398 extern int vm_munmap(unsigned long, size_t);
2399 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2400 unsigned long, unsigned long,
2401 unsigned long, unsigned long);
2403 struct vm_unmapped_area_info {
2404 #define VM_UNMAPPED_AREA_TOPDOWN 1
2405 unsigned long flags;
2406 unsigned long length;
2407 unsigned long low_limit;
2408 unsigned long high_limit;
2409 unsigned long align_mask;
2410 unsigned long align_offset;
2413 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2414 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2417 * Search for an unmapped address range.
2419 * We are looking for a range that:
2420 * - does not intersect with any VMA;
2421 * - is contained within the [low_limit, high_limit) interval;
2422 * - is at least the desired size.
2423 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2425 static inline unsigned long
2426 vm_unmapped_area(struct vm_unmapped_area_info *info)
2428 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2429 return unmapped_area_topdown(info);
2431 return unmapped_area(info);
2435 extern void truncate_inode_pages(struct address_space *, loff_t);
2436 extern void truncate_inode_pages_range(struct address_space *,
2437 loff_t lstart, loff_t lend);
2438 extern void truncate_inode_pages_final(struct address_space *);
2440 /* generic vm_area_ops exported for stackable file systems */
2441 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2442 extern void filemap_map_pages(struct vm_fault *vmf,
2443 pgoff_t start_pgoff, pgoff_t end_pgoff);
2444 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2446 /* mm/page-writeback.c */
2447 int __must_check write_one_page(struct page *page);
2448 void task_dirty_inc(struct task_struct *tsk);
2451 #define VM_READAHEAD_PAGES (SZ_128K / PAGE_SIZE)
2453 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2454 pgoff_t offset, unsigned long nr_to_read);
2456 void page_cache_sync_readahead(struct address_space *mapping,
2457 struct file_ra_state *ra,
2460 unsigned long size);
2462 void page_cache_async_readahead(struct address_space *mapping,
2463 struct file_ra_state *ra,
2467 unsigned long size);
2469 extern unsigned long stack_guard_gap;
2470 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2471 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2473 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2474 extern int expand_downwards(struct vm_area_struct *vma,
2475 unsigned long address);
2477 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2479 #define expand_upwards(vma, address) (0)
2482 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2483 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2484 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2485 struct vm_area_struct **pprev);
2487 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2488 NULL if none. Assume start_addr < end_addr. */
2489 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2491 struct vm_area_struct * vma = find_vma(mm,start_addr);
2493 if (vma && end_addr <= vma->vm_start)
2498 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2500 unsigned long vm_start = vma->vm_start;
2502 if (vma->vm_flags & VM_GROWSDOWN) {
2503 vm_start -= stack_guard_gap;
2504 if (vm_start > vma->vm_start)
2510 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2512 unsigned long vm_end = vma->vm_end;
2514 if (vma->vm_flags & VM_GROWSUP) {
2515 vm_end += stack_guard_gap;
2516 if (vm_end < vma->vm_end)
2517 vm_end = -PAGE_SIZE;
2522 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2524 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2527 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2528 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2529 unsigned long vm_start, unsigned long vm_end)
2531 struct vm_area_struct *vma = find_vma(mm, vm_start);
2533 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2539 static inline bool range_in_vma(struct vm_area_struct *vma,
2540 unsigned long start, unsigned long end)
2542 return (vma && vma->vm_start <= start && end <= vma->vm_end);
2546 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2547 void vma_set_page_prot(struct vm_area_struct *vma);
2549 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2553 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2555 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2559 #ifdef CONFIG_NUMA_BALANCING
2560 unsigned long change_prot_numa(struct vm_area_struct *vma,
2561 unsigned long start, unsigned long end);
2564 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2565 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2566 unsigned long pfn, unsigned long size, pgprot_t);
2567 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2568 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
2570 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
2572 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2574 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2575 unsigned long pfn, pgprot_t pgprot);
2576 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2578 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
2579 unsigned long addr, pfn_t pfn);
2580 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2582 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2583 unsigned long addr, struct page *page)
2585 int err = vm_insert_page(vma, addr, page);
2588 return VM_FAULT_OOM;
2589 if (err < 0 && err != -EBUSY)
2590 return VM_FAULT_SIGBUS;
2592 return VM_FAULT_NOPAGE;
2595 #ifndef io_remap_pfn_range
2596 static inline int io_remap_pfn_range(struct vm_area_struct *vma,
2597 unsigned long addr, unsigned long pfn,
2598 unsigned long size, pgprot_t prot)
2600 return remap_pfn_range(vma, addr, pfn, size, pgprot_decrypted(prot));
2604 static inline vm_fault_t vmf_error(int err)
2607 return VM_FAULT_OOM;
2608 return VM_FAULT_SIGBUS;
2611 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
2612 unsigned int foll_flags);
2614 #define FOLL_WRITE 0x01 /* check pte is writable */
2615 #define FOLL_TOUCH 0x02 /* mark page accessed */
2616 #define FOLL_GET 0x04 /* do get_page on page */
2617 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2618 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2619 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2620 * and return without waiting upon it */
2621 #define FOLL_POPULATE 0x40 /* fault in page */
2622 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2623 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2624 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2625 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2626 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2627 #define FOLL_MLOCK 0x1000 /* lock present pages */
2628 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2629 #define FOLL_COW 0x4000 /* internal GUP flag */
2630 #define FOLL_ANON 0x8000 /* don't do file mappings */
2631 #define FOLL_LONGTERM 0x10000 /* mapping lifetime is indefinite: see below */
2632 #define FOLL_SPLIT_PMD 0x20000 /* split huge pmd before returning */
2635 * NOTE on FOLL_LONGTERM:
2637 * FOLL_LONGTERM indicates that the page will be held for an indefinite time
2638 * period _often_ under userspace control. This is contrasted with
2639 * iov_iter_get_pages() where usages which are transient.
2641 * FIXME: For pages which are part of a filesystem, mappings are subject to the
2642 * lifetime enforced by the filesystem and we need guarantees that longterm
2643 * users like RDMA and V4L2 only establish mappings which coordinate usage with
2644 * the filesystem. Ideas for this coordination include revoking the longterm
2645 * pin, delaying writeback, bounce buffer page writeback, etc. As FS DAX was
2646 * added after the problem with filesystems was found FS DAX VMAs are
2647 * specifically failed. Filesystem pages are still subject to bugs and use of
2648 * FOLL_LONGTERM should be avoided on those pages.
2650 * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call.
2651 * Currently only get_user_pages() and get_user_pages_fast() support this flag
2652 * and calls to get_user_pages_[un]locked are specifically not allowed. This
2653 * is due to an incompatibility with the FS DAX check and
2654 * FAULT_FLAG_ALLOW_RETRY
2656 * In the CMA case: longterm pins in a CMA region would unnecessarily fragment
2657 * that region. And so CMA attempts to migrate the page before pinning when
2658 * FOLL_LONGTERM is specified.
2661 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
2663 if (vm_fault & VM_FAULT_OOM)
2665 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2666 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2667 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2672 typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
2673 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2674 unsigned long size, pte_fn_t fn, void *data);
2677 #ifdef CONFIG_PAGE_POISONING
2678 extern bool page_poisoning_enabled(void);
2679 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2681 static inline bool page_poisoning_enabled(void) { return false; }
2682 static inline void kernel_poison_pages(struct page *page, int numpages,
2686 #ifdef CONFIG_INIT_ON_ALLOC_DEFAULT_ON
2687 DECLARE_STATIC_KEY_TRUE(init_on_alloc);
2689 DECLARE_STATIC_KEY_FALSE(init_on_alloc);
2691 static inline bool want_init_on_alloc(gfp_t flags)
2693 if (static_branch_unlikely(&init_on_alloc) &&
2694 !page_poisoning_enabled())
2696 return flags & __GFP_ZERO;
2699 #ifdef CONFIG_INIT_ON_FREE_DEFAULT_ON
2700 DECLARE_STATIC_KEY_TRUE(init_on_free);
2702 DECLARE_STATIC_KEY_FALSE(init_on_free);
2704 static inline bool want_init_on_free(void)
2706 return static_branch_unlikely(&init_on_free) &&
2707 !page_poisoning_enabled();
2710 #ifdef CONFIG_DEBUG_PAGEALLOC
2711 extern void init_debug_pagealloc(void);
2713 static inline void init_debug_pagealloc(void) {}
2715 extern bool _debug_pagealloc_enabled_early;
2716 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
2718 static inline bool debug_pagealloc_enabled(void)
2720 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
2721 _debug_pagealloc_enabled_early;
2725 * For use in fast paths after init_debug_pagealloc() has run, or when a
2726 * false negative result is not harmful when called too early.
2728 static inline bool debug_pagealloc_enabled_static(void)
2730 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
2733 return static_branch_unlikely(&_debug_pagealloc_enabled);
2736 #if defined(CONFIG_DEBUG_PAGEALLOC) || defined(CONFIG_ARCH_HAS_SET_DIRECT_MAP)
2737 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2740 * When called in DEBUG_PAGEALLOC context, the call should most likely be
2741 * guarded by debug_pagealloc_enabled() or debug_pagealloc_enabled_static()
2744 kernel_map_pages(struct page *page, int numpages, int enable)
2746 __kernel_map_pages(page, numpages, enable);
2748 #ifdef CONFIG_HIBERNATION
2749 extern bool kernel_page_present(struct page *page);
2750 #endif /* CONFIG_HIBERNATION */
2751 #else /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */
2753 kernel_map_pages(struct page *page, int numpages, int enable) {}
2754 #ifdef CONFIG_HIBERNATION
2755 static inline bool kernel_page_present(struct page *page) { return true; }
2756 #endif /* CONFIG_HIBERNATION */
2757 #endif /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */
2759 #ifdef __HAVE_ARCH_GATE_AREA
2760 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2761 extern int in_gate_area_no_mm(unsigned long addr);
2762 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2764 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2768 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2769 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2773 #endif /* __HAVE_ARCH_GATE_AREA */
2775 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2777 #ifdef CONFIG_SYSCTL
2778 extern int sysctl_drop_caches;
2779 int drop_caches_sysctl_handler(struct ctl_table *, int,
2780 void __user *, size_t *, loff_t *);
2783 void drop_slab(void);
2784 void drop_slab_node(int nid);
2787 #define randomize_va_space 0
2789 extern int randomize_va_space;
2792 const char * arch_vma_name(struct vm_area_struct *vma);
2794 void print_vma_addr(char *prefix, unsigned long rip);
2796 static inline void print_vma_addr(char *prefix, unsigned long rip)
2801 void *sparse_buffer_alloc(unsigned long size);
2802 struct page * __populate_section_memmap(unsigned long pfn,
2803 unsigned long nr_pages, int nid, struct vmem_altmap *altmap);
2804 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2805 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
2806 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
2807 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2808 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2809 void *vmemmap_alloc_block(unsigned long size, int node);
2811 void *vmemmap_alloc_block_buf(unsigned long size, int node);
2812 void *altmap_alloc_block_buf(unsigned long size, struct vmem_altmap *altmap);
2813 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2814 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2816 int vmemmap_populate(unsigned long start, unsigned long end, int node,
2817 struct vmem_altmap *altmap);
2818 void vmemmap_populate_print_last(void);
2819 #ifdef CONFIG_MEMORY_HOTPLUG
2820 void vmemmap_free(unsigned long start, unsigned long end,
2821 struct vmem_altmap *altmap);
2823 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2824 unsigned long nr_pages);
2827 MF_COUNT_INCREASED = 1 << 0,
2828 MF_ACTION_REQUIRED = 1 << 1,
2829 MF_MUST_KILL = 1 << 2,
2830 MF_SOFT_OFFLINE = 1 << 3,
2832 extern int memory_failure(unsigned long pfn, int flags);
2833 extern void memory_failure_queue(unsigned long pfn, int flags);
2834 extern int unpoison_memory(unsigned long pfn);
2835 extern int get_hwpoison_page(struct page *page);
2836 #define put_hwpoison_page(page) put_page(page)
2837 extern int sysctl_memory_failure_early_kill;
2838 extern int sysctl_memory_failure_recovery;
2839 extern void shake_page(struct page *p, int access);
2840 extern atomic_long_t num_poisoned_pages __read_mostly;
2841 extern int soft_offline_page(struct page *page, int flags);
2845 * Error handlers for various types of pages.
2848 MF_IGNORED, /* Error: cannot be handled */
2849 MF_FAILED, /* Error: handling failed */
2850 MF_DELAYED, /* Will be handled later */
2851 MF_RECOVERED, /* Successfully recovered */
2854 enum mf_action_page_type {
2856 MF_MSG_KERNEL_HIGH_ORDER,
2858 MF_MSG_DIFFERENT_COMPOUND,
2859 MF_MSG_POISONED_HUGE,
2862 MF_MSG_NON_PMD_HUGE,
2863 MF_MSG_UNMAP_FAILED,
2864 MF_MSG_DIRTY_SWAPCACHE,
2865 MF_MSG_CLEAN_SWAPCACHE,
2866 MF_MSG_DIRTY_MLOCKED_LRU,
2867 MF_MSG_CLEAN_MLOCKED_LRU,
2868 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2869 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2872 MF_MSG_TRUNCATED_LRU,
2879 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2880 extern void clear_huge_page(struct page *page,
2881 unsigned long addr_hint,
2882 unsigned int pages_per_huge_page);
2883 extern void copy_user_huge_page(struct page *dst, struct page *src,
2884 unsigned long addr_hint,
2885 struct vm_area_struct *vma,
2886 unsigned int pages_per_huge_page);
2887 extern long copy_huge_page_from_user(struct page *dst_page,
2888 const void __user *usr_src,
2889 unsigned int pages_per_huge_page,
2890 bool allow_pagefault);
2891 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2893 #ifdef CONFIG_DEBUG_PAGEALLOC
2894 extern unsigned int _debug_guardpage_minorder;
2895 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
2897 static inline unsigned int debug_guardpage_minorder(void)
2899 return _debug_guardpage_minorder;
2902 static inline bool debug_guardpage_enabled(void)
2904 return static_branch_unlikely(&_debug_guardpage_enabled);
2907 static inline bool page_is_guard(struct page *page)
2909 if (!debug_guardpage_enabled())
2912 return PageGuard(page);
2915 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2916 static inline bool debug_guardpage_enabled(void) { return false; }
2917 static inline bool page_is_guard(struct page *page) { return false; }
2918 #endif /* CONFIG_DEBUG_PAGEALLOC */
2920 #if MAX_NUMNODES > 1
2921 void __init setup_nr_node_ids(void);
2923 static inline void setup_nr_node_ids(void) {}
2926 extern int memcmp_pages(struct page *page1, struct page *page2);
2928 static inline int pages_identical(struct page *page1, struct page *page2)
2930 return !memcmp_pages(page1, page2);
2934 * seal_check_future_write - Check for F_SEAL_FUTURE_WRITE flag and handle it
2935 * @seals: the seals to check
2936 * @vma: the vma to operate on
2938 * Check whether F_SEAL_FUTURE_WRITE is set; if so, do proper check/handling on
2939 * the vma flags. Return 0 if check pass, or <0 for errors.
2941 static inline int seal_check_future_write(int seals, struct vm_area_struct *vma)
2943 if (seals & F_SEAL_FUTURE_WRITE) {
2945 * New PROT_WRITE and MAP_SHARED mmaps are not allowed when
2946 * "future write" seal active.
2948 if ((vma->vm_flags & VM_SHARED) && (vma->vm_flags & VM_WRITE))
2952 * Since an F_SEAL_FUTURE_WRITE sealed memfd can be mapped as
2953 * MAP_SHARED and read-only, take care to not allow mprotect to
2954 * revert protections on such mappings. Do this only for shared
2955 * mappings. For private mappings, don't need to mask
2956 * VM_MAYWRITE as we still want them to be COW-writable.
2958 if (vma->vm_flags & VM_SHARED)
2959 vma->vm_flags &= ~(VM_MAYWRITE);
2965 #endif /* __KERNEL__ */
2966 #endif /* _LINUX_MM_H */