1 /* SPDX-License-Identifier: GPL-2.0 */
5 #include <linux/errno.h>
6 #include <linux/mmdebug.h>
9 #include <linux/list.h>
10 #include <linux/mmzone.h>
11 #include <linux/rbtree.h>
12 #include <linux/atomic.h>
13 #include <linux/debug_locks.h>
14 #include <linux/mm_types.h>
15 #include <linux/mmap_lock.h>
16 #include <linux/range.h>
17 #include <linux/pfn.h>
18 #include <linux/percpu-refcount.h>
19 #include <linux/bit_spinlock.h>
20 #include <linux/shrinker.h>
21 #include <linux/resource.h>
22 #include <linux/page_ext.h>
23 #include <linux/err.h>
24 #include <linux/page-flags.h>
25 #include <linux/page_ref.h>
26 #include <linux/overflow.h>
27 #include <linux/sizes.h>
28 #include <linux/sched.h>
29 #include <linux/pgtable.h>
30 #include <linux/kasan.h>
34 struct anon_vma_chain;
38 extern int sysctl_page_lock_unfairness;
40 void init_mm_internals(void);
42 #ifndef CONFIG_NUMA /* Don't use mapnrs, do it properly */
43 extern unsigned long max_mapnr;
45 static inline void set_max_mapnr(unsigned long limit)
50 static inline void set_max_mapnr(unsigned long limit) { }
53 extern atomic_long_t _totalram_pages;
54 static inline unsigned long totalram_pages(void)
56 return (unsigned long)atomic_long_read(&_totalram_pages);
59 static inline void totalram_pages_inc(void)
61 atomic_long_inc(&_totalram_pages);
64 static inline void totalram_pages_dec(void)
66 atomic_long_dec(&_totalram_pages);
69 static inline void totalram_pages_add(long count)
71 atomic_long_add(count, &_totalram_pages);
74 extern void * high_memory;
75 extern int page_cluster;
78 extern int sysctl_legacy_va_layout;
80 #define sysctl_legacy_va_layout 0
83 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
84 extern const int mmap_rnd_bits_min;
85 extern const int mmap_rnd_bits_max;
86 extern int mmap_rnd_bits __read_mostly;
88 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
89 extern const int mmap_rnd_compat_bits_min;
90 extern const int mmap_rnd_compat_bits_max;
91 extern int mmap_rnd_compat_bits __read_mostly;
95 #include <asm/processor.h>
98 * Architectures that support memory tagging (assigning tags to memory regions,
99 * embedding these tags into addresses that point to these memory regions, and
100 * checking that the memory and the pointer tags match on memory accesses)
101 * redefine this macro to strip tags from pointers.
102 * It's defined as noop for architectures that don't support memory tagging.
104 #ifndef untagged_addr
105 #define untagged_addr(addr) (addr)
109 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
113 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
117 #define lm_alias(x) __va(__pa_symbol(x))
121 * To prevent common memory management code establishing
122 * a zero page mapping on a read fault.
123 * This macro should be defined within <asm/pgtable.h>.
124 * s390 does this to prevent multiplexing of hardware bits
125 * related to the physical page in case of virtualization.
127 #ifndef mm_forbids_zeropage
128 #define mm_forbids_zeropage(X) (0)
132 * On some architectures it is expensive to call memset() for small sizes.
133 * If an architecture decides to implement their own version of
134 * mm_zero_struct_page they should wrap the defines below in a #ifndef and
135 * define their own version of this macro in <asm/pgtable.h>
137 #if BITS_PER_LONG == 64
138 /* This function must be updated when the size of struct page grows above 80
139 * or reduces below 56. The idea that compiler optimizes out switch()
140 * statement, and only leaves move/store instructions. Also the compiler can
141 * combine write statements if they are both assignments and can be reordered,
142 * this can result in several of the writes here being dropped.
144 #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
145 static inline void __mm_zero_struct_page(struct page *page)
147 unsigned long *_pp = (void *)page;
149 /* Check that struct page is either 56, 64, 72, or 80 bytes */
150 BUILD_BUG_ON(sizeof(struct page) & 7);
151 BUILD_BUG_ON(sizeof(struct page) < 56);
152 BUILD_BUG_ON(sizeof(struct page) > 80);
154 switch (sizeof(struct page)) {
175 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
179 * Default maximum number of active map areas, this limits the number of vmas
180 * per mm struct. Users can overwrite this number by sysctl but there is a
183 * When a program's coredump is generated as ELF format, a section is created
184 * per a vma. In ELF, the number of sections is represented in unsigned short.
185 * This means the number of sections should be smaller than 65535 at coredump.
186 * Because the kernel adds some informative sections to a image of program at
187 * generating coredump, we need some margin. The number of extra sections is
188 * 1-3 now and depends on arch. We use "5" as safe margin, here.
190 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
191 * not a hard limit any more. Although some userspace tools can be surprised by
194 #define MAPCOUNT_ELF_CORE_MARGIN (5)
195 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
197 extern int sysctl_max_map_count;
199 extern unsigned long sysctl_user_reserve_kbytes;
200 extern unsigned long sysctl_admin_reserve_kbytes;
202 extern int sysctl_overcommit_memory;
203 extern int sysctl_overcommit_ratio;
204 extern unsigned long sysctl_overcommit_kbytes;
206 int overcommit_ratio_handler(struct ctl_table *, int, void *, size_t *,
208 int overcommit_kbytes_handler(struct ctl_table *, int, void *, size_t *,
210 int overcommit_policy_handler(struct ctl_table *, int, void *, size_t *,
213 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
214 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
215 #define folio_page_idx(folio, p) (page_to_pfn(p) - folio_pfn(folio))
217 #define nth_page(page,n) ((page) + (n))
218 #define folio_page_idx(folio, p) ((p) - &(folio)->page)
221 /* to align the pointer to the (next) page boundary */
222 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
224 /* to align the pointer to the (prev) page boundary */
225 #define PAGE_ALIGN_DOWN(addr) ALIGN_DOWN(addr, PAGE_SIZE)
227 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
228 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
230 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
231 static inline struct folio *lru_to_folio(struct list_head *head)
233 return list_entry((head)->prev, struct folio, lru);
236 void setup_initial_init_mm(void *start_code, void *end_code,
237 void *end_data, void *brk);
240 * Linux kernel virtual memory manager primitives.
241 * The idea being to have a "virtual" mm in the same way
242 * we have a virtual fs - giving a cleaner interface to the
243 * mm details, and allowing different kinds of memory mappings
244 * (from shared memory to executable loading to arbitrary
248 struct vm_area_struct *vm_area_alloc(struct mm_struct *);
249 struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
250 void vm_area_free(struct vm_area_struct *);
253 extern struct rb_root nommu_region_tree;
254 extern struct rw_semaphore nommu_region_sem;
256 extern unsigned int kobjsize(const void *objp);
260 * vm_flags in vm_area_struct, see mm_types.h.
261 * When changing, update also include/trace/events/mmflags.h
263 #define VM_NONE 0x00000000
265 #define VM_READ 0x00000001 /* currently active flags */
266 #define VM_WRITE 0x00000002
267 #define VM_EXEC 0x00000004
268 #define VM_SHARED 0x00000008
270 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
271 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
272 #define VM_MAYWRITE 0x00000020
273 #define VM_MAYEXEC 0x00000040
274 #define VM_MAYSHARE 0x00000080
276 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
277 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
278 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
279 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
281 #define VM_LOCKED 0x00002000
282 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
284 /* Used by sys_madvise() */
285 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
286 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
288 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
289 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
290 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
291 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
292 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
293 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
294 #define VM_SYNC 0x00800000 /* Synchronous page faults */
295 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
296 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
297 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
299 #ifdef CONFIG_MEM_SOFT_DIRTY
300 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
302 # define VM_SOFTDIRTY 0
305 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
306 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
307 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
308 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
310 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
311 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
312 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
313 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
314 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
315 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
316 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
317 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
318 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
319 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
320 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
321 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
323 #ifdef CONFIG_ARCH_HAS_PKEYS
324 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
325 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
326 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
327 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
328 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
330 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
332 # define VM_PKEY_BIT4 0
334 #endif /* CONFIG_ARCH_HAS_PKEYS */
336 #if defined(CONFIG_X86)
337 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
338 #elif defined(CONFIG_PPC)
339 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
340 #elif defined(CONFIG_PARISC)
341 # define VM_GROWSUP VM_ARCH_1
342 #elif defined(CONFIG_IA64)
343 # define VM_GROWSUP VM_ARCH_1
344 #elif defined(CONFIG_SPARC64)
345 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
346 # define VM_ARCH_CLEAR VM_SPARC_ADI
347 #elif defined(CONFIG_ARM64)
348 # define VM_ARM64_BTI VM_ARCH_1 /* BTI guarded page, a.k.a. GP bit */
349 # define VM_ARCH_CLEAR VM_ARM64_BTI
350 #elif !defined(CONFIG_MMU)
351 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
354 #if defined(CONFIG_ARM64_MTE)
355 # define VM_MTE VM_HIGH_ARCH_0 /* Use Tagged memory for access control */
356 # define VM_MTE_ALLOWED VM_HIGH_ARCH_1 /* Tagged memory permitted */
358 # define VM_MTE VM_NONE
359 # define VM_MTE_ALLOWED VM_NONE
363 # define VM_GROWSUP VM_NONE
366 #ifdef CONFIG_HAVE_ARCH_USERFAULTFD_MINOR
367 # define VM_UFFD_MINOR_BIT 37
368 # define VM_UFFD_MINOR BIT(VM_UFFD_MINOR_BIT) /* UFFD minor faults */
369 #else /* !CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
370 # define VM_UFFD_MINOR VM_NONE
371 #endif /* CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
373 /* Bits set in the VMA until the stack is in its final location */
374 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
376 #define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0)
378 /* Common data flag combinations */
379 #define VM_DATA_FLAGS_TSK_EXEC (VM_READ | VM_WRITE | TASK_EXEC | \
380 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
381 #define VM_DATA_FLAGS_NON_EXEC (VM_READ | VM_WRITE | VM_MAYREAD | \
382 VM_MAYWRITE | VM_MAYEXEC)
383 #define VM_DATA_FLAGS_EXEC (VM_READ | VM_WRITE | VM_EXEC | \
384 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
386 #ifndef VM_DATA_DEFAULT_FLAGS /* arch can override this */
387 #define VM_DATA_DEFAULT_FLAGS VM_DATA_FLAGS_EXEC
390 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
391 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
394 #ifdef CONFIG_STACK_GROWSUP
395 #define VM_STACK VM_GROWSUP
397 #define VM_STACK VM_GROWSDOWN
400 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
402 /* VMA basic access permission flags */
403 #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
407 * Special vmas that are non-mergable, non-mlock()able.
409 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
411 /* This mask prevents VMA from being scanned with khugepaged */
412 #define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB)
414 /* This mask defines which mm->def_flags a process can inherit its parent */
415 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
417 /* This mask is used to clear all the VMA flags used by mlock */
418 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
420 /* Arch-specific flags to clear when updating VM flags on protection change */
421 #ifndef VM_ARCH_CLEAR
422 # define VM_ARCH_CLEAR VM_NONE
424 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
427 * mapping from the currently active vm_flags protection bits (the
428 * low four bits) to a page protection mask..
430 extern pgprot_t protection_map[16];
433 * The default fault flags that should be used by most of the
434 * arch-specific page fault handlers.
436 #define FAULT_FLAG_DEFAULT (FAULT_FLAG_ALLOW_RETRY | \
437 FAULT_FLAG_KILLABLE | \
438 FAULT_FLAG_INTERRUPTIBLE)
441 * fault_flag_allow_retry_first - check ALLOW_RETRY the first time
442 * @flags: Fault flags.
444 * This is mostly used for places where we want to try to avoid taking
445 * the mmap_lock for too long a time when waiting for another condition
446 * to change, in which case we can try to be polite to release the
447 * mmap_lock in the first round to avoid potential starvation of other
448 * processes that would also want the mmap_lock.
450 * Return: true if the page fault allows retry and this is the first
451 * attempt of the fault handling; false otherwise.
453 static inline bool fault_flag_allow_retry_first(enum fault_flag flags)
455 return (flags & FAULT_FLAG_ALLOW_RETRY) &&
456 (!(flags & FAULT_FLAG_TRIED));
459 #define FAULT_FLAG_TRACE \
460 { FAULT_FLAG_WRITE, "WRITE" }, \
461 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
462 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
463 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
464 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
465 { FAULT_FLAG_TRIED, "TRIED" }, \
466 { FAULT_FLAG_USER, "USER" }, \
467 { FAULT_FLAG_REMOTE, "REMOTE" }, \
468 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }, \
469 { FAULT_FLAG_INTERRUPTIBLE, "INTERRUPTIBLE" }
472 * vm_fault is filled by the pagefault handler and passed to the vma's
473 * ->fault function. The vma's ->fault is responsible for returning a bitmask
474 * of VM_FAULT_xxx flags that give details about how the fault was handled.
476 * MM layer fills up gfp_mask for page allocations but fault handler might
477 * alter it if its implementation requires a different allocation context.
479 * pgoff should be used in favour of virtual_address, if possible.
483 struct vm_area_struct *vma; /* Target VMA */
484 gfp_t gfp_mask; /* gfp mask to be used for allocations */
485 pgoff_t pgoff; /* Logical page offset based on vma */
486 unsigned long address; /* Faulting virtual address - masked */
487 unsigned long real_address; /* Faulting virtual address - unmasked */
489 enum fault_flag flags; /* FAULT_FLAG_xxx flags
490 * XXX: should really be 'const' */
491 pmd_t *pmd; /* Pointer to pmd entry matching
493 pud_t *pud; /* Pointer to pud entry matching
497 pte_t orig_pte; /* Value of PTE at the time of fault */
498 pmd_t orig_pmd; /* Value of PMD at the time of fault,
499 * used by PMD fault only.
503 struct page *cow_page; /* Page handler may use for COW fault */
504 struct page *page; /* ->fault handlers should return a
505 * page here, unless VM_FAULT_NOPAGE
506 * is set (which is also implied by
509 /* These three entries are valid only while holding ptl lock */
510 pte_t *pte; /* Pointer to pte entry matching
511 * the 'address'. NULL if the page
512 * table hasn't been allocated.
514 spinlock_t *ptl; /* Page table lock.
515 * Protects pte page table if 'pte'
516 * is not NULL, otherwise pmd.
518 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
519 * vm_ops->map_pages() sets up a page
520 * table from atomic context.
521 * do_fault_around() pre-allocates
522 * page table to avoid allocation from
527 /* page entry size for vm->huge_fault() */
528 enum page_entry_size {
535 * These are the virtual MM functions - opening of an area, closing and
536 * unmapping it (needed to keep files on disk up-to-date etc), pointer
537 * to the functions called when a no-page or a wp-page exception occurs.
539 struct vm_operations_struct {
540 void (*open)(struct vm_area_struct * area);
542 * @close: Called when the VMA is being removed from the MM.
543 * Context: User context. May sleep. Caller holds mmap_lock.
545 void (*close)(struct vm_area_struct * area);
546 /* Called any time before splitting to check if it's allowed */
547 int (*may_split)(struct vm_area_struct *area, unsigned long addr);
548 int (*mremap)(struct vm_area_struct *area);
550 * Called by mprotect() to make driver-specific permission
551 * checks before mprotect() is finalised. The VMA must not
552 * be modified. Returns 0 if eprotect() can proceed.
554 int (*mprotect)(struct vm_area_struct *vma, unsigned long start,
555 unsigned long end, unsigned long newflags);
556 vm_fault_t (*fault)(struct vm_fault *vmf);
557 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
558 enum page_entry_size pe_size);
559 vm_fault_t (*map_pages)(struct vm_fault *vmf,
560 pgoff_t start_pgoff, pgoff_t end_pgoff);
561 unsigned long (*pagesize)(struct vm_area_struct * area);
563 /* notification that a previously read-only page is about to become
564 * writable, if an error is returned it will cause a SIGBUS */
565 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
567 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
568 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
570 /* called by access_process_vm when get_user_pages() fails, typically
571 * for use by special VMAs. See also generic_access_phys() for a generic
572 * implementation useful for any iomem mapping.
574 int (*access)(struct vm_area_struct *vma, unsigned long addr,
575 void *buf, int len, int write);
577 /* Called by the /proc/PID/maps code to ask the vma whether it
578 * has a special name. Returning non-NULL will also cause this
579 * vma to be dumped unconditionally. */
580 const char *(*name)(struct vm_area_struct *vma);
584 * set_policy() op must add a reference to any non-NULL @new mempolicy
585 * to hold the policy upon return. Caller should pass NULL @new to
586 * remove a policy and fall back to surrounding context--i.e. do not
587 * install a MPOL_DEFAULT policy, nor the task or system default
590 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
593 * get_policy() op must add reference [mpol_get()] to any policy at
594 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
595 * in mm/mempolicy.c will do this automatically.
596 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
597 * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
598 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
599 * must return NULL--i.e., do not "fallback" to task or system default
602 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
606 * Called by vm_normal_page() for special PTEs to find the
607 * page for @addr. This is useful if the default behavior
608 * (using pte_page()) would not find the correct page.
610 struct page *(*find_special_page)(struct vm_area_struct *vma,
614 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
616 static const struct vm_operations_struct dummy_vm_ops = {};
618 memset(vma, 0, sizeof(*vma));
620 vma->vm_ops = &dummy_vm_ops;
621 INIT_LIST_HEAD(&vma->anon_vma_chain);
624 static inline void vma_set_anonymous(struct vm_area_struct *vma)
629 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
634 static inline bool vma_is_temporary_stack(struct vm_area_struct *vma)
636 int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP);
641 if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) ==
642 VM_STACK_INCOMPLETE_SETUP)
648 static inline bool vma_is_foreign(struct vm_area_struct *vma)
653 if (current->mm != vma->vm_mm)
659 static inline bool vma_is_accessible(struct vm_area_struct *vma)
661 return vma->vm_flags & VM_ACCESS_FLAGS;
666 * The vma_is_shmem is not inline because it is used only by slow
667 * paths in userfault.
669 bool vma_is_shmem(struct vm_area_struct *vma);
671 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
674 int vma_is_stack_for_current(struct vm_area_struct *vma);
676 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
677 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
682 static inline unsigned int compound_order(struct page *page)
686 return page[1].compound_order;
690 * folio_order - The allocation order of a folio.
693 * A folio is composed of 2^order pages. See get_order() for the definition
696 * Return: The order of the folio.
698 static inline unsigned int folio_order(struct folio *folio)
700 return compound_order(&folio->page);
703 #include <linux/huge_mm.h>
706 * Methods to modify the page usage count.
708 * What counts for a page usage:
709 * - cache mapping (page->mapping)
710 * - private data (page->private)
711 * - page mapped in a task's page tables, each mapping
712 * is counted separately
714 * Also, many kernel routines increase the page count before a critical
715 * routine so they can be sure the page doesn't go away from under them.
719 * Drop a ref, return true if the refcount fell to zero (the page has no users)
721 static inline int put_page_testzero(struct page *page)
723 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
724 return page_ref_dec_and_test(page);
727 static inline int folio_put_testzero(struct folio *folio)
729 return put_page_testzero(&folio->page);
733 * Try to grab a ref unless the page has a refcount of zero, return false if
735 * This can be called when MMU is off so it must not access
736 * any of the virtual mappings.
738 static inline bool get_page_unless_zero(struct page *page)
740 return page_ref_add_unless(page, 1, 0);
743 extern int page_is_ram(unsigned long pfn);
751 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
754 /* Support for virtually mapped pages */
755 struct page *vmalloc_to_page(const void *addr);
756 unsigned long vmalloc_to_pfn(const void *addr);
759 * Determine if an address is within the vmalloc range
761 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
762 * is no special casing required.
765 #ifndef is_ioremap_addr
766 #define is_ioremap_addr(x) is_vmalloc_addr(x)
770 extern bool is_vmalloc_addr(const void *x);
771 extern int is_vmalloc_or_module_addr(const void *x);
773 static inline bool is_vmalloc_addr(const void *x)
777 static inline int is_vmalloc_or_module_addr(const void *x)
784 * How many times the entire folio is mapped as a single unit (eg by a
785 * PMD or PUD entry). This is probably not what you want, except for
786 * debugging purposes; look at folio_mapcount() or page_mapcount()
789 static inline int folio_entire_mapcount(struct folio *folio)
791 VM_BUG_ON_FOLIO(!folio_test_large(folio), folio);
792 return atomic_read(folio_mapcount_ptr(folio)) + 1;
796 * Mapcount of compound page as a whole, does not include mapped sub-pages.
798 * Must be called only for compound pages.
800 static inline int compound_mapcount(struct page *page)
802 return folio_entire_mapcount(page_folio(page));
806 * The atomic page->_mapcount, starts from -1: so that transitions
807 * both from it and to it can be tracked, using atomic_inc_and_test
808 * and atomic_add_negative(-1).
810 static inline void page_mapcount_reset(struct page *page)
812 atomic_set(&(page)->_mapcount, -1);
815 int __page_mapcount(struct page *page);
818 * Mapcount of 0-order page; when compound sub-page, includes
819 * compound_mapcount().
821 * Result is undefined for pages which cannot be mapped into userspace.
822 * For example SLAB or special types of pages. See function page_has_type().
823 * They use this place in struct page differently.
825 static inline int page_mapcount(struct page *page)
827 if (unlikely(PageCompound(page)))
828 return __page_mapcount(page);
829 return atomic_read(&page->_mapcount) + 1;
832 int folio_mapcount(struct folio *folio);
834 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
835 static inline int total_mapcount(struct page *page)
837 return folio_mapcount(page_folio(page));
841 static inline int total_mapcount(struct page *page)
843 return page_mapcount(page);
847 static inline struct page *virt_to_head_page(const void *x)
849 struct page *page = virt_to_page(x);
851 return compound_head(page);
854 static inline struct folio *virt_to_folio(const void *x)
856 struct page *page = virt_to_page(x);
858 return page_folio(page);
861 void __put_page(struct page *page);
863 void put_pages_list(struct list_head *pages);
865 void split_page(struct page *page, unsigned int order);
866 void folio_copy(struct folio *dst, struct folio *src);
868 unsigned long nr_free_buffer_pages(void);
871 * Compound pages have a destructor function. Provide a
872 * prototype for that function and accessor functions.
873 * These are _only_ valid on the head of a compound page.
875 typedef void compound_page_dtor(struct page *);
877 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
878 enum compound_dtor_id {
881 #ifdef CONFIG_HUGETLB_PAGE
884 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
889 extern compound_page_dtor * const compound_page_dtors[NR_COMPOUND_DTORS];
891 static inline void set_compound_page_dtor(struct page *page,
892 enum compound_dtor_id compound_dtor)
894 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
895 page[1].compound_dtor = compound_dtor;
898 static inline void destroy_compound_page(struct page *page)
900 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
901 compound_page_dtors[page[1].compound_dtor](page);
904 static inline int head_compound_pincount(struct page *head)
906 return atomic_read(compound_pincount_ptr(head));
909 static inline void set_compound_order(struct page *page, unsigned int order)
911 page[1].compound_order = order;
913 page[1].compound_nr = 1U << order;
917 /* Returns the number of pages in this potentially compound page. */
918 static inline unsigned long compound_nr(struct page *page)
923 return page[1].compound_nr;
925 return 1UL << compound_order(page);
929 /* Returns the number of bytes in this potentially compound page. */
930 static inline unsigned long page_size(struct page *page)
932 return PAGE_SIZE << compound_order(page);
935 /* Returns the number of bits needed for the number of bytes in a page */
936 static inline unsigned int page_shift(struct page *page)
938 return PAGE_SHIFT + compound_order(page);
942 * thp_order - Order of a transparent huge page.
943 * @page: Head page of a transparent huge page.
945 static inline unsigned int thp_order(struct page *page)
947 VM_BUG_ON_PGFLAGS(PageTail(page), page);
948 return compound_order(page);
952 * thp_nr_pages - The number of regular pages in this huge page.
953 * @page: The head page of a huge page.
955 static inline int thp_nr_pages(struct page *page)
957 VM_BUG_ON_PGFLAGS(PageTail(page), page);
958 return compound_nr(page);
962 * thp_size - Size of a transparent huge page.
963 * @page: Head page of a transparent huge page.
965 * Return: Number of bytes in this page.
967 static inline unsigned long thp_size(struct page *page)
969 return PAGE_SIZE << thp_order(page);
972 void free_compound_page(struct page *page);
976 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
977 * servicing faults for write access. In the normal case, do always want
978 * pte_mkwrite. But get_user_pages can cause write faults for mappings
979 * that do not have writing enabled, when used by access_process_vm.
981 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
983 if (likely(vma->vm_flags & VM_WRITE))
984 pte = pte_mkwrite(pte);
988 vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page);
989 void do_set_pte(struct vm_fault *vmf, struct page *page, unsigned long addr);
991 vm_fault_t finish_fault(struct vm_fault *vmf);
992 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
996 * Multiple processes may "see" the same page. E.g. for untouched
997 * mappings of /dev/null, all processes see the same page full of
998 * zeroes, and text pages of executables and shared libraries have
999 * only one copy in memory, at most, normally.
1001 * For the non-reserved pages, page_count(page) denotes a reference count.
1002 * page_count() == 0 means the page is free. page->lru is then used for
1003 * freelist management in the buddy allocator.
1004 * page_count() > 0 means the page has been allocated.
1006 * Pages are allocated by the slab allocator in order to provide memory
1007 * to kmalloc and kmem_cache_alloc. In this case, the management of the
1008 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
1009 * unless a particular usage is carefully commented. (the responsibility of
1010 * freeing the kmalloc memory is the caller's, of course).
1012 * A page may be used by anyone else who does a __get_free_page().
1013 * In this case, page_count still tracks the references, and should only
1014 * be used through the normal accessor functions. The top bits of page->flags
1015 * and page->virtual store page management information, but all other fields
1016 * are unused and could be used privately, carefully. The management of this
1017 * page is the responsibility of the one who allocated it, and those who have
1018 * subsequently been given references to it.
1020 * The other pages (we may call them "pagecache pages") are completely
1021 * managed by the Linux memory manager: I/O, buffers, swapping etc.
1022 * The following discussion applies only to them.
1024 * A pagecache page contains an opaque `private' member, which belongs to the
1025 * page's address_space. Usually, this is the address of a circular list of
1026 * the page's disk buffers. PG_private must be set to tell the VM to call
1027 * into the filesystem to release these pages.
1029 * A page may belong to an inode's memory mapping. In this case, page->mapping
1030 * is the pointer to the inode, and page->index is the file offset of the page,
1031 * in units of PAGE_SIZE.
1033 * If pagecache pages are not associated with an inode, they are said to be
1034 * anonymous pages. These may become associated with the swapcache, and in that
1035 * case PG_swapcache is set, and page->private is an offset into the swapcache.
1037 * In either case (swapcache or inode backed), the pagecache itself holds one
1038 * reference to the page. Setting PG_private should also increment the
1039 * refcount. The each user mapping also has a reference to the page.
1041 * The pagecache pages are stored in a per-mapping radix tree, which is
1042 * rooted at mapping->i_pages, and indexed by offset.
1043 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
1044 * lists, we instead now tag pages as dirty/writeback in the radix tree.
1046 * All pagecache pages may be subject to I/O:
1047 * - inode pages may need to be read from disk,
1048 * - inode pages which have been modified and are MAP_SHARED may need
1049 * to be written back to the inode on disk,
1050 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
1051 * modified may need to be swapped out to swap space and (later) to be read
1056 * The zone field is never updated after free_area_init_core()
1057 * sets it, so none of the operations on it need to be atomic.
1060 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
1061 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
1062 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
1063 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
1064 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
1065 #define KASAN_TAG_PGOFF (LAST_CPUPID_PGOFF - KASAN_TAG_WIDTH)
1068 * Define the bit shifts to access each section. For non-existent
1069 * sections we define the shift as 0; that plus a 0 mask ensures
1070 * the compiler will optimise away reference to them.
1072 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
1073 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
1074 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
1075 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
1076 #define KASAN_TAG_PGSHIFT (KASAN_TAG_PGOFF * (KASAN_TAG_WIDTH != 0))
1078 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
1079 #ifdef NODE_NOT_IN_PAGE_FLAGS
1080 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
1081 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
1082 SECTIONS_PGOFF : ZONES_PGOFF)
1084 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
1085 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
1086 NODES_PGOFF : ZONES_PGOFF)
1089 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
1091 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
1092 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
1093 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
1094 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
1095 #define KASAN_TAG_MASK ((1UL << KASAN_TAG_WIDTH) - 1)
1096 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
1098 static inline enum zone_type page_zonenum(const struct page *page)
1100 ASSERT_EXCLUSIVE_BITS(page->flags, ZONES_MASK << ZONES_PGSHIFT);
1101 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
1104 static inline enum zone_type folio_zonenum(const struct folio *folio)
1106 return page_zonenum(&folio->page);
1109 #ifdef CONFIG_ZONE_DEVICE
1110 static inline bool is_zone_device_page(const struct page *page)
1112 return page_zonenum(page) == ZONE_DEVICE;
1114 extern void memmap_init_zone_device(struct zone *, unsigned long,
1115 unsigned long, struct dev_pagemap *);
1117 static inline bool is_zone_device_page(const struct page *page)
1123 static inline bool folio_is_zone_device(const struct folio *folio)
1125 return is_zone_device_page(&folio->page);
1128 static inline bool is_zone_movable_page(const struct page *page)
1130 return page_zonenum(page) == ZONE_MOVABLE;
1133 #if defined(CONFIG_ZONE_DEVICE) && defined(CONFIG_FS_DAX)
1134 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
1136 bool __put_devmap_managed_page_refs(struct page *page, int refs);
1137 static inline bool put_devmap_managed_page_refs(struct page *page, int refs)
1139 if (!static_branch_unlikely(&devmap_managed_key))
1141 if (!is_zone_device_page(page))
1143 return __put_devmap_managed_page_refs(page, refs);
1145 #else /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1146 static inline bool put_devmap_managed_page_refs(struct page *page, int refs)
1150 #endif /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1152 static inline bool put_devmap_managed_page(struct page *page)
1154 return put_devmap_managed_page_refs(page, 1);
1157 /* 127: arbitrary random number, small enough to assemble well */
1158 #define folio_ref_zero_or_close_to_overflow(folio) \
1159 ((unsigned int) folio_ref_count(folio) + 127u <= 127u)
1162 * folio_get - Increment the reference count on a folio.
1163 * @folio: The folio.
1165 * Context: May be called in any context, as long as you know that
1166 * you have a refcount on the folio. If you do not already have one,
1167 * folio_try_get() may be the right interface for you to use.
1169 static inline void folio_get(struct folio *folio)
1171 VM_BUG_ON_FOLIO(folio_ref_zero_or_close_to_overflow(folio), folio);
1172 folio_ref_inc(folio);
1175 static inline void get_page(struct page *page)
1177 folio_get(page_folio(page));
1180 bool __must_check try_grab_page(struct page *page, unsigned int flags);
1182 static inline __must_check bool try_get_page(struct page *page)
1184 page = compound_head(page);
1185 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1192 * folio_put - Decrement the reference count on a folio.
1193 * @folio: The folio.
1195 * If the folio's reference count reaches zero, the memory will be
1196 * released back to the page allocator and may be used by another
1197 * allocation immediately. Do not access the memory or the struct folio
1198 * after calling folio_put() unless you can be sure that it wasn't the
1201 * Context: May be called in process or interrupt context, but not in NMI
1202 * context. May be called while holding a spinlock.
1204 static inline void folio_put(struct folio *folio)
1206 if (folio_put_testzero(folio))
1207 __put_page(&folio->page);
1211 * folio_put_refs - Reduce the reference count on a folio.
1212 * @folio: The folio.
1213 * @refs: The amount to subtract from the folio's reference count.
1215 * If the folio's reference count reaches zero, the memory will be
1216 * released back to the page allocator and may be used by another
1217 * allocation immediately. Do not access the memory or the struct folio
1218 * after calling folio_put_refs() unless you can be sure that these weren't
1219 * the last references.
1221 * Context: May be called in process or interrupt context, but not in NMI
1222 * context. May be called while holding a spinlock.
1224 static inline void folio_put_refs(struct folio *folio, int refs)
1226 if (folio_ref_sub_and_test(folio, refs))
1227 __put_page(&folio->page);
1230 static inline void put_page(struct page *page)
1232 struct folio *folio = page_folio(page);
1235 * For some devmap managed pages we need to catch refcount transition
1238 if (put_devmap_managed_page(&folio->page))
1244 * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
1245 * the page's refcount so that two separate items are tracked: the original page
1246 * reference count, and also a new count of how many pin_user_pages() calls were
1247 * made against the page. ("gup-pinned" is another term for the latter).
1249 * With this scheme, pin_user_pages() becomes special: such pages are marked as
1250 * distinct from normal pages. As such, the unpin_user_page() call (and its
1251 * variants) must be used in order to release gup-pinned pages.
1255 * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
1256 * counts with respect to pin_user_pages() and unpin_user_page() becomes
1257 * simpler, due to the fact that adding an even power of two to the page
1258 * refcount has the effect of using only the upper N bits, for the code that
1259 * counts up using the bias value. This means that the lower bits are left for
1260 * the exclusive use of the original code that increments and decrements by one
1261 * (or at least, by much smaller values than the bias value).
1263 * Of course, once the lower bits overflow into the upper bits (and this is
1264 * OK, because subtraction recovers the original values), then visual inspection
1265 * no longer suffices to directly view the separate counts. However, for normal
1266 * applications that don't have huge page reference counts, this won't be an
1269 * Locking: the lockless algorithm described in folio_try_get_rcu()
1270 * provides safe operation for get_user_pages(), page_mkclean() and
1271 * other calls that race to set up page table entries.
1273 #define GUP_PIN_COUNTING_BIAS (1U << 10)
1275 void unpin_user_page(struct page *page);
1276 void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1278 void unpin_user_page_range_dirty_lock(struct page *page, unsigned long npages,
1280 void unpin_user_pages(struct page **pages, unsigned long npages);
1282 static inline bool is_cow_mapping(vm_flags_t flags)
1284 return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
1287 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1288 #define SECTION_IN_PAGE_FLAGS
1292 * The identification function is mainly used by the buddy allocator for
1293 * determining if two pages could be buddies. We are not really identifying
1294 * the zone since we could be using the section number id if we do not have
1295 * node id available in page flags.
1296 * We only guarantee that it will return the same value for two combinable
1299 static inline int page_zone_id(struct page *page)
1301 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1304 #ifdef NODE_NOT_IN_PAGE_FLAGS
1305 extern int page_to_nid(const struct page *page);
1307 static inline int page_to_nid(const struct page *page)
1309 struct page *p = (struct page *)page;
1311 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1315 static inline int folio_nid(const struct folio *folio)
1317 return page_to_nid(&folio->page);
1320 #ifdef CONFIG_NUMA_BALANCING
1321 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1323 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1326 static inline int cpupid_to_pid(int cpupid)
1328 return cpupid & LAST__PID_MASK;
1331 static inline int cpupid_to_cpu(int cpupid)
1333 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1336 static inline int cpupid_to_nid(int cpupid)
1338 return cpu_to_node(cpupid_to_cpu(cpupid));
1341 static inline bool cpupid_pid_unset(int cpupid)
1343 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1346 static inline bool cpupid_cpu_unset(int cpupid)
1348 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1351 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1353 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1356 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1357 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1358 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1360 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1363 static inline int page_cpupid_last(struct page *page)
1365 return page->_last_cpupid;
1367 static inline void page_cpupid_reset_last(struct page *page)
1369 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1372 static inline int page_cpupid_last(struct page *page)
1374 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1377 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1379 static inline void page_cpupid_reset_last(struct page *page)
1381 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1383 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1384 #else /* !CONFIG_NUMA_BALANCING */
1385 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1387 return page_to_nid(page); /* XXX */
1390 static inline int page_cpupid_last(struct page *page)
1392 return page_to_nid(page); /* XXX */
1395 static inline int cpupid_to_nid(int cpupid)
1400 static inline int cpupid_to_pid(int cpupid)
1405 static inline int cpupid_to_cpu(int cpupid)
1410 static inline int cpu_pid_to_cpupid(int nid, int pid)
1415 static inline bool cpupid_pid_unset(int cpupid)
1420 static inline void page_cpupid_reset_last(struct page *page)
1424 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1428 #endif /* CONFIG_NUMA_BALANCING */
1430 #if defined(CONFIG_KASAN_SW_TAGS) || defined(CONFIG_KASAN_HW_TAGS)
1433 * KASAN per-page tags are stored xor'ed with 0xff. This allows to avoid
1434 * setting tags for all pages to native kernel tag value 0xff, as the default
1435 * value 0x00 maps to 0xff.
1438 static inline u8 page_kasan_tag(const struct page *page)
1442 if (kasan_enabled()) {
1443 tag = (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1450 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1452 unsigned long old_flags, flags;
1454 if (!kasan_enabled())
1458 old_flags = READ_ONCE(page->flags);
1461 flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1462 flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1463 } while (unlikely(!try_cmpxchg(&page->flags, &old_flags, flags)));
1466 static inline void page_kasan_tag_reset(struct page *page)
1468 if (kasan_enabled())
1469 page_kasan_tag_set(page, 0xff);
1472 #else /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1474 static inline u8 page_kasan_tag(const struct page *page)
1479 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1480 static inline void page_kasan_tag_reset(struct page *page) { }
1482 #endif /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1484 static inline struct zone *page_zone(const struct page *page)
1486 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1489 static inline pg_data_t *page_pgdat(const struct page *page)
1491 return NODE_DATA(page_to_nid(page));
1494 static inline struct zone *folio_zone(const struct folio *folio)
1496 return page_zone(&folio->page);
1499 static inline pg_data_t *folio_pgdat(const struct folio *folio)
1501 return page_pgdat(&folio->page);
1504 #ifdef SECTION_IN_PAGE_FLAGS
1505 static inline void set_page_section(struct page *page, unsigned long section)
1507 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1508 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1511 static inline unsigned long page_to_section(const struct page *page)
1513 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1518 * folio_pfn - Return the Page Frame Number of a folio.
1519 * @folio: The folio.
1521 * A folio may contain multiple pages. The pages have consecutive
1522 * Page Frame Numbers.
1524 * Return: The Page Frame Number of the first page in the folio.
1526 static inline unsigned long folio_pfn(struct folio *folio)
1528 return page_to_pfn(&folio->page);
1531 static inline atomic_t *folio_pincount_ptr(struct folio *folio)
1533 return &folio_page(folio, 1)->compound_pincount;
1537 * folio_maybe_dma_pinned - Report if a folio may be pinned for DMA.
1538 * @folio: The folio.
1540 * This function checks if a folio has been pinned via a call to
1541 * a function in the pin_user_pages() family.
1543 * For small folios, the return value is partially fuzzy: false is not fuzzy,
1544 * because it means "definitely not pinned for DMA", but true means "probably
1545 * pinned for DMA, but possibly a false positive due to having at least
1546 * GUP_PIN_COUNTING_BIAS worth of normal folio references".
1548 * False positives are OK, because: a) it's unlikely for a folio to
1549 * get that many refcounts, and b) all the callers of this routine are
1550 * expected to be able to deal gracefully with a false positive.
1552 * For large folios, the result will be exactly correct. That's because
1553 * we have more tracking data available: the compound_pincount is used
1554 * instead of the GUP_PIN_COUNTING_BIAS scheme.
1556 * For more information, please see Documentation/core-api/pin_user_pages.rst.
1558 * Return: True, if it is likely that the page has been "dma-pinned".
1559 * False, if the page is definitely not dma-pinned.
1561 static inline bool folio_maybe_dma_pinned(struct folio *folio)
1563 if (folio_test_large(folio))
1564 return atomic_read(folio_pincount_ptr(folio)) > 0;
1567 * folio_ref_count() is signed. If that refcount overflows, then
1568 * folio_ref_count() returns a negative value, and callers will avoid
1569 * further incrementing the refcount.
1571 * Here, for that overflow case, use the sign bit to count a little
1572 * bit higher via unsigned math, and thus still get an accurate result.
1574 return ((unsigned int)folio_ref_count(folio)) >=
1575 GUP_PIN_COUNTING_BIAS;
1578 static inline bool page_maybe_dma_pinned(struct page *page)
1580 return folio_maybe_dma_pinned(page_folio(page));
1584 * This should most likely only be called during fork() to see whether we
1585 * should break the cow immediately for an anon page on the src mm.
1587 * The caller has to hold the PT lock and the vma->vm_mm->->write_protect_seq.
1589 static inline bool page_needs_cow_for_dma(struct vm_area_struct *vma,
1592 VM_BUG_ON(!(raw_read_seqcount(&vma->vm_mm->write_protect_seq) & 1));
1594 if (!test_bit(MMF_HAS_PINNED, &vma->vm_mm->flags))
1597 return page_maybe_dma_pinned(page);
1600 /* MIGRATE_CMA and ZONE_MOVABLE do not allow pin pages */
1601 #ifdef CONFIG_MIGRATION
1602 static inline bool is_pinnable_page(struct page *page)
1605 int mt = get_pageblock_migratetype(page);
1607 if (mt == MIGRATE_CMA || mt == MIGRATE_ISOLATE)
1610 return !is_zone_movable_page(page) || is_zero_pfn(page_to_pfn(page));
1613 static inline bool is_pinnable_page(struct page *page)
1619 static inline bool folio_is_pinnable(struct folio *folio)
1621 return is_pinnable_page(&folio->page);
1624 static inline void set_page_zone(struct page *page, enum zone_type zone)
1626 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1627 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1630 static inline void set_page_node(struct page *page, unsigned long node)
1632 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1633 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1636 static inline void set_page_links(struct page *page, enum zone_type zone,
1637 unsigned long node, unsigned long pfn)
1639 set_page_zone(page, zone);
1640 set_page_node(page, node);
1641 #ifdef SECTION_IN_PAGE_FLAGS
1642 set_page_section(page, pfn_to_section_nr(pfn));
1647 * folio_nr_pages - The number of pages in the folio.
1648 * @folio: The folio.
1650 * Return: A positive power of two.
1652 static inline long folio_nr_pages(struct folio *folio)
1654 return compound_nr(&folio->page);
1658 * folio_next - Move to the next physical folio.
1659 * @folio: The folio we're currently operating on.
1661 * If you have physically contiguous memory which may span more than
1662 * one folio (eg a &struct bio_vec), use this function to move from one
1663 * folio to the next. Do not use it if the memory is only virtually
1664 * contiguous as the folios are almost certainly not adjacent to each
1665 * other. This is the folio equivalent to writing ``page++``.
1667 * Context: We assume that the folios are refcounted and/or locked at a
1668 * higher level and do not adjust the reference counts.
1669 * Return: The next struct folio.
1671 static inline struct folio *folio_next(struct folio *folio)
1673 return (struct folio *)folio_page(folio, folio_nr_pages(folio));
1677 * folio_shift - The size of the memory described by this folio.
1678 * @folio: The folio.
1680 * A folio represents a number of bytes which is a power-of-two in size.
1681 * This function tells you which power-of-two the folio is. See also
1682 * folio_size() and folio_order().
1684 * Context: The caller should have a reference on the folio to prevent
1685 * it from being split. It is not necessary for the folio to be locked.
1686 * Return: The base-2 logarithm of the size of this folio.
1688 static inline unsigned int folio_shift(struct folio *folio)
1690 return PAGE_SHIFT + folio_order(folio);
1694 * folio_size - The number of bytes in a folio.
1695 * @folio: The folio.
1697 * Context: The caller should have a reference on the folio to prevent
1698 * it from being split. It is not necessary for the folio to be locked.
1699 * Return: The number of bytes in this folio.
1701 static inline size_t folio_size(struct folio *folio)
1703 return PAGE_SIZE << folio_order(folio);
1706 #ifndef HAVE_ARCH_MAKE_PAGE_ACCESSIBLE
1707 static inline int arch_make_page_accessible(struct page *page)
1713 #ifndef HAVE_ARCH_MAKE_FOLIO_ACCESSIBLE
1714 static inline int arch_make_folio_accessible(struct folio *folio)
1717 long i, nr = folio_nr_pages(folio);
1719 for (i = 0; i < nr; i++) {
1720 ret = arch_make_page_accessible(folio_page(folio, i));
1730 * Some inline functions in vmstat.h depend on page_zone()
1732 #include <linux/vmstat.h>
1734 static __always_inline void *lowmem_page_address(const struct page *page)
1736 return page_to_virt(page);
1739 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1740 #define HASHED_PAGE_VIRTUAL
1743 #if defined(WANT_PAGE_VIRTUAL)
1744 static inline void *page_address(const struct page *page)
1746 return page->virtual;
1748 static inline void set_page_address(struct page *page, void *address)
1750 page->virtual = address;
1752 #define page_address_init() do { } while(0)
1755 #if defined(HASHED_PAGE_VIRTUAL)
1756 void *page_address(const struct page *page);
1757 void set_page_address(struct page *page, void *virtual);
1758 void page_address_init(void);
1761 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1762 #define page_address(page) lowmem_page_address(page)
1763 #define set_page_address(page, address) do { } while(0)
1764 #define page_address_init() do { } while(0)
1767 static inline void *folio_address(const struct folio *folio)
1769 return page_address(&folio->page);
1772 extern void *page_rmapping(struct page *page);
1773 extern pgoff_t __page_file_index(struct page *page);
1776 * Return the pagecache index of the passed page. Regular pagecache pages
1777 * use ->index whereas swapcache pages use swp_offset(->private)
1779 static inline pgoff_t page_index(struct page *page)
1781 if (unlikely(PageSwapCache(page)))
1782 return __page_file_index(page);
1786 bool page_mapped(struct page *page);
1787 bool folio_mapped(struct folio *folio);
1790 * Return true only if the page has been allocated with
1791 * ALLOC_NO_WATERMARKS and the low watermark was not
1792 * met implying that the system is under some pressure.
1794 static inline bool page_is_pfmemalloc(const struct page *page)
1797 * lru.next has bit 1 set if the page is allocated from the
1798 * pfmemalloc reserves. Callers may simply overwrite it if
1799 * they do not need to preserve that information.
1801 return (uintptr_t)page->lru.next & BIT(1);
1805 * Only to be called by the page allocator on a freshly allocated
1808 static inline void set_page_pfmemalloc(struct page *page)
1810 page->lru.next = (void *)BIT(1);
1813 static inline void clear_page_pfmemalloc(struct page *page)
1815 page->lru.next = NULL;
1819 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1821 extern void pagefault_out_of_memory(void);
1823 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1824 #define offset_in_thp(page, p) ((unsigned long)(p) & (thp_size(page) - 1))
1825 #define offset_in_folio(folio, p) ((unsigned long)(p) & (folio_size(folio) - 1))
1828 * Flags passed to show_mem() and show_free_areas() to suppress output in
1831 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1833 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1836 extern bool can_do_mlock(void);
1838 static inline bool can_do_mlock(void) { return false; }
1840 extern int user_shm_lock(size_t, struct ucounts *);
1841 extern void user_shm_unlock(size_t, struct ucounts *);
1843 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1845 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1848 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1849 unsigned long size);
1850 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1851 unsigned long size);
1852 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1853 unsigned long start, unsigned long end);
1855 struct mmu_notifier_range;
1857 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1858 unsigned long end, unsigned long floor, unsigned long ceiling);
1860 copy_page_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma);
1861 int follow_pte(struct mm_struct *mm, unsigned long address,
1862 pte_t **ptepp, spinlock_t **ptlp);
1863 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1864 unsigned long *pfn);
1865 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1866 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1867 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1868 void *buf, int len, int write);
1870 extern void truncate_pagecache(struct inode *inode, loff_t new);
1871 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1872 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1873 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1874 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1877 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1878 unsigned long address, unsigned int flags,
1879 struct pt_regs *regs);
1880 extern int fixup_user_fault(struct mm_struct *mm,
1881 unsigned long address, unsigned int fault_flags,
1883 void unmap_mapping_pages(struct address_space *mapping,
1884 pgoff_t start, pgoff_t nr, bool even_cows);
1885 void unmap_mapping_range(struct address_space *mapping,
1886 loff_t const holebegin, loff_t const holelen, int even_cows);
1888 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1889 unsigned long address, unsigned int flags,
1890 struct pt_regs *regs)
1892 /* should never happen if there's no MMU */
1894 return VM_FAULT_SIGBUS;
1896 static inline int fixup_user_fault(struct mm_struct *mm, unsigned long address,
1897 unsigned int fault_flags, bool *unlocked)
1899 /* should never happen if there's no MMU */
1903 static inline void unmap_mapping_pages(struct address_space *mapping,
1904 pgoff_t start, pgoff_t nr, bool even_cows) { }
1905 static inline void unmap_mapping_range(struct address_space *mapping,
1906 loff_t const holebegin, loff_t const holelen, int even_cows) { }
1909 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1910 loff_t const holebegin, loff_t const holelen)
1912 unmap_mapping_range(mapping, holebegin, holelen, 0);
1915 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1916 void *buf, int len, unsigned int gup_flags);
1917 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1918 void *buf, int len, unsigned int gup_flags);
1919 extern int __access_remote_vm(struct mm_struct *mm, unsigned long addr,
1920 void *buf, int len, unsigned int gup_flags);
1922 long get_user_pages_remote(struct mm_struct *mm,
1923 unsigned long start, unsigned long nr_pages,
1924 unsigned int gup_flags, struct page **pages,
1925 struct vm_area_struct **vmas, int *locked);
1926 long pin_user_pages_remote(struct mm_struct *mm,
1927 unsigned long start, unsigned long nr_pages,
1928 unsigned int gup_flags, struct page **pages,
1929 struct vm_area_struct **vmas, int *locked);
1930 long get_user_pages(unsigned long start, unsigned long nr_pages,
1931 unsigned int gup_flags, struct page **pages,
1932 struct vm_area_struct **vmas);
1933 long pin_user_pages(unsigned long start, unsigned long nr_pages,
1934 unsigned int gup_flags, struct page **pages,
1935 struct vm_area_struct **vmas);
1936 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1937 struct page **pages, unsigned int gup_flags);
1938 long pin_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1939 struct page **pages, unsigned int gup_flags);
1941 int get_user_pages_fast(unsigned long start, int nr_pages,
1942 unsigned int gup_flags, struct page **pages);
1943 int pin_user_pages_fast(unsigned long start, int nr_pages,
1944 unsigned int gup_flags, struct page **pages);
1946 int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
1947 int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
1948 struct task_struct *task, bool bypass_rlim);
1951 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1952 struct page **pages);
1953 struct page *get_dump_page(unsigned long addr);
1955 bool folio_mark_dirty(struct folio *folio);
1956 bool set_page_dirty(struct page *page);
1957 int set_page_dirty_lock(struct page *page);
1959 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1961 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1962 unsigned long old_addr, struct vm_area_struct *new_vma,
1963 unsigned long new_addr, unsigned long len,
1964 bool need_rmap_locks);
1967 * Flags used by change_protection(). For now we make it a bitmap so
1968 * that we can pass in multiple flags just like parameters. However
1969 * for now all the callers are only use one of the flags at the same
1972 /* Whether we should allow dirty bit accounting */
1973 #define MM_CP_DIRTY_ACCT (1UL << 0)
1974 /* Whether this protection change is for NUMA hints */
1975 #define MM_CP_PROT_NUMA (1UL << 1)
1976 /* Whether this change is for write protecting */
1977 #define MM_CP_UFFD_WP (1UL << 2) /* do wp */
1978 #define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */
1979 #define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \
1980 MM_CP_UFFD_WP_RESOLVE)
1982 extern unsigned long change_protection(struct mmu_gather *tlb,
1983 struct vm_area_struct *vma, unsigned long start,
1984 unsigned long end, pgprot_t newprot,
1985 unsigned long cp_flags);
1986 extern int mprotect_fixup(struct mmu_gather *tlb, struct vm_area_struct *vma,
1987 struct vm_area_struct **pprev, unsigned long start,
1988 unsigned long end, unsigned long newflags);
1991 * doesn't attempt to fault and will return short.
1993 int get_user_pages_fast_only(unsigned long start, int nr_pages,
1994 unsigned int gup_flags, struct page **pages);
1995 int pin_user_pages_fast_only(unsigned long start, int nr_pages,
1996 unsigned int gup_flags, struct page **pages);
1998 static inline bool get_user_page_fast_only(unsigned long addr,
1999 unsigned int gup_flags, struct page **pagep)
2001 return get_user_pages_fast_only(addr, 1, gup_flags, pagep) == 1;
2004 * per-process(per-mm_struct) statistics.
2006 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
2008 long val = atomic_long_read(&mm->rss_stat.count[member]);
2010 #ifdef SPLIT_RSS_COUNTING
2012 * counter is updated in asynchronous manner and may go to minus.
2013 * But it's never be expected number for users.
2018 return (unsigned long)val;
2021 void mm_trace_rss_stat(struct mm_struct *mm, int member, long count);
2023 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
2025 long count = atomic_long_add_return(value, &mm->rss_stat.count[member]);
2027 mm_trace_rss_stat(mm, member, count);
2030 static inline void inc_mm_counter(struct mm_struct *mm, int member)
2032 long count = atomic_long_inc_return(&mm->rss_stat.count[member]);
2034 mm_trace_rss_stat(mm, member, count);
2037 static inline void dec_mm_counter(struct mm_struct *mm, int member)
2039 long count = atomic_long_dec_return(&mm->rss_stat.count[member]);
2041 mm_trace_rss_stat(mm, member, count);
2044 /* Optimized variant when page is already known not to be PageAnon */
2045 static inline int mm_counter_file(struct page *page)
2047 if (PageSwapBacked(page))
2048 return MM_SHMEMPAGES;
2049 return MM_FILEPAGES;
2052 static inline int mm_counter(struct page *page)
2055 return MM_ANONPAGES;
2056 return mm_counter_file(page);
2059 static inline unsigned long get_mm_rss(struct mm_struct *mm)
2061 return get_mm_counter(mm, MM_FILEPAGES) +
2062 get_mm_counter(mm, MM_ANONPAGES) +
2063 get_mm_counter(mm, MM_SHMEMPAGES);
2066 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
2068 return max(mm->hiwater_rss, get_mm_rss(mm));
2071 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
2073 return max(mm->hiwater_vm, mm->total_vm);
2076 static inline void update_hiwater_rss(struct mm_struct *mm)
2078 unsigned long _rss = get_mm_rss(mm);
2080 if ((mm)->hiwater_rss < _rss)
2081 (mm)->hiwater_rss = _rss;
2084 static inline void update_hiwater_vm(struct mm_struct *mm)
2086 if (mm->hiwater_vm < mm->total_vm)
2087 mm->hiwater_vm = mm->total_vm;
2090 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
2092 mm->hiwater_rss = get_mm_rss(mm);
2095 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
2096 struct mm_struct *mm)
2098 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
2100 if (*maxrss < hiwater_rss)
2101 *maxrss = hiwater_rss;
2104 #if defined(SPLIT_RSS_COUNTING)
2105 void sync_mm_rss(struct mm_struct *mm);
2107 static inline void sync_mm_rss(struct mm_struct *mm)
2112 #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
2113 static inline int pte_special(pte_t pte)
2118 static inline pte_t pte_mkspecial(pte_t pte)
2124 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
2125 static inline int pte_devmap(pte_t pte)
2131 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
2133 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
2135 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
2139 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
2143 #ifdef __PAGETABLE_P4D_FOLDED
2144 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2145 unsigned long address)
2150 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
2153 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
2154 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2155 unsigned long address)
2159 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
2160 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
2163 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
2165 static inline void mm_inc_nr_puds(struct mm_struct *mm)
2167 if (mm_pud_folded(mm))
2169 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2172 static inline void mm_dec_nr_puds(struct mm_struct *mm)
2174 if (mm_pud_folded(mm))
2176 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2180 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
2181 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
2182 unsigned long address)
2187 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
2188 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
2191 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
2193 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
2195 if (mm_pmd_folded(mm))
2197 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2200 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
2202 if (mm_pmd_folded(mm))
2204 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2209 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
2211 atomic_long_set(&mm->pgtables_bytes, 0);
2214 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2216 return atomic_long_read(&mm->pgtables_bytes);
2219 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
2221 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2224 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
2226 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2230 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
2231 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2236 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
2237 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
2240 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
2241 int __pte_alloc_kernel(pmd_t *pmd);
2243 #if defined(CONFIG_MMU)
2245 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2246 unsigned long address)
2248 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
2249 NULL : p4d_offset(pgd, address);
2252 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2253 unsigned long address)
2255 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
2256 NULL : pud_offset(p4d, address);
2259 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
2261 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
2262 NULL: pmd_offset(pud, address);
2264 #endif /* CONFIG_MMU */
2266 #if USE_SPLIT_PTE_PTLOCKS
2267 #if ALLOC_SPLIT_PTLOCKS
2268 void __init ptlock_cache_init(void);
2269 extern bool ptlock_alloc(struct page *page);
2270 extern void ptlock_free(struct page *page);
2272 static inline spinlock_t *ptlock_ptr(struct page *page)
2276 #else /* ALLOC_SPLIT_PTLOCKS */
2277 static inline void ptlock_cache_init(void)
2281 static inline bool ptlock_alloc(struct page *page)
2286 static inline void ptlock_free(struct page *page)
2290 static inline spinlock_t *ptlock_ptr(struct page *page)
2294 #endif /* ALLOC_SPLIT_PTLOCKS */
2296 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2298 return ptlock_ptr(pmd_page(*pmd));
2301 static inline bool ptlock_init(struct page *page)
2304 * prep_new_page() initialize page->private (and therefore page->ptl)
2305 * with 0. Make sure nobody took it in use in between.
2307 * It can happen if arch try to use slab for page table allocation:
2308 * slab code uses page->slab_cache, which share storage with page->ptl.
2310 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
2311 if (!ptlock_alloc(page))
2313 spin_lock_init(ptlock_ptr(page));
2317 #else /* !USE_SPLIT_PTE_PTLOCKS */
2319 * We use mm->page_table_lock to guard all pagetable pages of the mm.
2321 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2323 return &mm->page_table_lock;
2325 static inline void ptlock_cache_init(void) {}
2326 static inline bool ptlock_init(struct page *page) { return true; }
2327 static inline void ptlock_free(struct page *page) {}
2328 #endif /* USE_SPLIT_PTE_PTLOCKS */
2330 static inline void pgtable_init(void)
2332 ptlock_cache_init();
2333 pgtable_cache_init();
2336 static inline bool pgtable_pte_page_ctor(struct page *page)
2338 if (!ptlock_init(page))
2340 __SetPageTable(page);
2341 inc_lruvec_page_state(page, NR_PAGETABLE);
2345 static inline void pgtable_pte_page_dtor(struct page *page)
2348 __ClearPageTable(page);
2349 dec_lruvec_page_state(page, NR_PAGETABLE);
2352 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
2354 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
2355 pte_t *__pte = pte_offset_map(pmd, address); \
2361 #define pte_unmap_unlock(pte, ptl) do { \
2366 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2368 #define pte_alloc_map(mm, pmd, address) \
2369 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2371 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2372 (pte_alloc(mm, pmd) ? \
2373 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2375 #define pte_alloc_kernel(pmd, address) \
2376 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2377 NULL: pte_offset_kernel(pmd, address))
2379 #if USE_SPLIT_PMD_PTLOCKS
2381 static struct page *pmd_to_page(pmd_t *pmd)
2383 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2384 return virt_to_page((void *)((unsigned long) pmd & mask));
2387 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2389 return ptlock_ptr(pmd_to_page(pmd));
2392 static inline bool pmd_ptlock_init(struct page *page)
2394 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2395 page->pmd_huge_pte = NULL;
2397 return ptlock_init(page);
2400 static inline void pmd_ptlock_free(struct page *page)
2402 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2403 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2408 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2412 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2414 return &mm->page_table_lock;
2417 static inline bool pmd_ptlock_init(struct page *page) { return true; }
2418 static inline void pmd_ptlock_free(struct page *page) {}
2420 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2424 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2426 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2431 static inline bool pgtable_pmd_page_ctor(struct page *page)
2433 if (!pmd_ptlock_init(page))
2435 __SetPageTable(page);
2436 inc_lruvec_page_state(page, NR_PAGETABLE);
2440 static inline void pgtable_pmd_page_dtor(struct page *page)
2442 pmd_ptlock_free(page);
2443 __ClearPageTable(page);
2444 dec_lruvec_page_state(page, NR_PAGETABLE);
2448 * No scalability reason to split PUD locks yet, but follow the same pattern
2449 * as the PMD locks to make it easier if we decide to. The VM should not be
2450 * considered ready to switch to split PUD locks yet; there may be places
2451 * which need to be converted from page_table_lock.
2453 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2455 return &mm->page_table_lock;
2458 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2460 spinlock_t *ptl = pud_lockptr(mm, pud);
2466 extern void __init pagecache_init(void);
2467 extern void free_initmem(void);
2470 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2471 * into the buddy system. The freed pages will be poisoned with pattern
2472 * "poison" if it's within range [0, UCHAR_MAX].
2473 * Return pages freed into the buddy system.
2475 extern unsigned long free_reserved_area(void *start, void *end,
2476 int poison, const char *s);
2478 extern void adjust_managed_page_count(struct page *page, long count);
2479 extern void mem_init_print_info(void);
2481 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2483 /* Free the reserved page into the buddy system, so it gets managed. */
2484 static inline void free_reserved_page(struct page *page)
2486 ClearPageReserved(page);
2487 init_page_count(page);
2489 adjust_managed_page_count(page, 1);
2491 #define free_highmem_page(page) free_reserved_page(page)
2493 static inline void mark_page_reserved(struct page *page)
2495 SetPageReserved(page);
2496 adjust_managed_page_count(page, -1);
2500 * Default method to free all the __init memory into the buddy system.
2501 * The freed pages will be poisoned with pattern "poison" if it's within
2502 * range [0, UCHAR_MAX].
2503 * Return pages freed into the buddy system.
2505 static inline unsigned long free_initmem_default(int poison)
2507 extern char __init_begin[], __init_end[];
2509 return free_reserved_area(&__init_begin, &__init_end,
2510 poison, "unused kernel image (initmem)");
2513 static inline unsigned long get_num_physpages(void)
2516 unsigned long phys_pages = 0;
2518 for_each_online_node(nid)
2519 phys_pages += node_present_pages(nid);
2525 * Using memblock node mappings, an architecture may initialise its
2526 * zones, allocate the backing mem_map and account for memory holes in an
2527 * architecture independent manner.
2529 * An architecture is expected to register range of page frames backed by
2530 * physical memory with memblock_add[_node]() before calling
2531 * free_area_init() passing in the PFN each zone ends at. At a basic
2532 * usage, an architecture is expected to do something like
2534 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2536 * for_each_valid_physical_page_range()
2537 * memblock_add_node(base, size, nid, MEMBLOCK_NONE)
2538 * free_area_init(max_zone_pfns);
2540 void free_area_init(unsigned long *max_zone_pfn);
2541 unsigned long node_map_pfn_alignment(void);
2542 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2543 unsigned long end_pfn);
2544 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2545 unsigned long end_pfn);
2546 extern void get_pfn_range_for_nid(unsigned int nid,
2547 unsigned long *start_pfn, unsigned long *end_pfn);
2548 extern unsigned long find_min_pfn_with_active_regions(void);
2551 static inline int early_pfn_to_nid(unsigned long pfn)
2556 /* please see mm/page_alloc.c */
2557 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2560 extern void set_dma_reserve(unsigned long new_dma_reserve);
2561 extern void memmap_init_range(unsigned long, int, unsigned long,
2562 unsigned long, unsigned long, enum meminit_context,
2563 struct vmem_altmap *, int migratetype);
2564 extern void setup_per_zone_wmarks(void);
2565 extern void calculate_min_free_kbytes(void);
2566 extern int __meminit init_per_zone_wmark_min(void);
2567 extern void mem_init(void);
2568 extern void __init mmap_init(void);
2569 extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2570 extern long si_mem_available(void);
2571 extern void si_meminfo(struct sysinfo * val);
2572 extern void si_meminfo_node(struct sysinfo *val, int nid);
2573 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2574 extern unsigned long arch_reserved_kernel_pages(void);
2577 extern __printf(3, 4)
2578 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2580 extern void setup_per_cpu_pageset(void);
2583 extern int min_free_kbytes;
2584 extern int watermark_boost_factor;
2585 extern int watermark_scale_factor;
2586 extern bool arch_has_descending_max_zone_pfns(void);
2589 extern atomic_long_t mmap_pages_allocated;
2590 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2592 /* interval_tree.c */
2593 void vma_interval_tree_insert(struct vm_area_struct *node,
2594 struct rb_root_cached *root);
2595 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2596 struct vm_area_struct *prev,
2597 struct rb_root_cached *root);
2598 void vma_interval_tree_remove(struct vm_area_struct *node,
2599 struct rb_root_cached *root);
2600 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2601 unsigned long start, unsigned long last);
2602 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2603 unsigned long start, unsigned long last);
2605 #define vma_interval_tree_foreach(vma, root, start, last) \
2606 for (vma = vma_interval_tree_iter_first(root, start, last); \
2607 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2609 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2610 struct rb_root_cached *root);
2611 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2612 struct rb_root_cached *root);
2613 struct anon_vma_chain *
2614 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2615 unsigned long start, unsigned long last);
2616 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2617 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2618 #ifdef CONFIG_DEBUG_VM_RB
2619 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2622 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2623 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2624 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2627 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2628 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2629 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2630 struct vm_area_struct *expand);
2631 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2632 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2634 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2636 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2637 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2638 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2639 struct mempolicy *, struct vm_userfaultfd_ctx, struct anon_vma_name *);
2640 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2641 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2642 unsigned long addr, int new_below);
2643 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2644 unsigned long addr, int new_below);
2645 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2646 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2647 struct rb_node **, struct rb_node *);
2648 extern void unlink_file_vma(struct vm_area_struct *);
2649 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2650 unsigned long addr, unsigned long len, pgoff_t pgoff,
2651 bool *need_rmap_locks);
2652 extern void exit_mmap(struct mm_struct *);
2654 static inline int check_data_rlimit(unsigned long rlim,
2656 unsigned long start,
2657 unsigned long end_data,
2658 unsigned long start_data)
2660 if (rlim < RLIM_INFINITY) {
2661 if (((new - start) + (end_data - start_data)) > rlim)
2668 extern int mm_take_all_locks(struct mm_struct *mm);
2669 extern void mm_drop_all_locks(struct mm_struct *mm);
2671 extern int set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2672 extern int replace_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2673 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2674 extern struct file *get_task_exe_file(struct task_struct *task);
2676 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2677 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2679 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2680 const struct vm_special_mapping *sm);
2681 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2682 unsigned long addr, unsigned long len,
2683 unsigned long flags,
2684 const struct vm_special_mapping *spec);
2685 /* This is an obsolete alternative to _install_special_mapping. */
2686 extern int install_special_mapping(struct mm_struct *mm,
2687 unsigned long addr, unsigned long len,
2688 unsigned long flags, struct page **pages);
2690 unsigned long randomize_stack_top(unsigned long stack_top);
2691 unsigned long randomize_page(unsigned long start, unsigned long range);
2693 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2695 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2696 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2697 struct list_head *uf);
2698 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2699 unsigned long len, unsigned long prot, unsigned long flags,
2700 unsigned long pgoff, unsigned long *populate, struct list_head *uf);
2701 extern int __do_munmap(struct mm_struct *, unsigned long, size_t,
2702 struct list_head *uf, bool downgrade);
2703 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2704 struct list_head *uf);
2705 extern int do_madvise(struct mm_struct *mm, unsigned long start, size_t len_in, int behavior);
2708 extern int __mm_populate(unsigned long addr, unsigned long len,
2710 static inline void mm_populate(unsigned long addr, unsigned long len)
2713 (void) __mm_populate(addr, len, 1);
2716 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2719 /* These take the mm semaphore themselves */
2720 extern int __must_check vm_brk(unsigned long, unsigned long);
2721 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2722 extern int vm_munmap(unsigned long, size_t);
2723 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2724 unsigned long, unsigned long,
2725 unsigned long, unsigned long);
2727 struct vm_unmapped_area_info {
2728 #define VM_UNMAPPED_AREA_TOPDOWN 1
2729 unsigned long flags;
2730 unsigned long length;
2731 unsigned long low_limit;
2732 unsigned long high_limit;
2733 unsigned long align_mask;
2734 unsigned long align_offset;
2737 extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info);
2740 extern void truncate_inode_pages(struct address_space *, loff_t);
2741 extern void truncate_inode_pages_range(struct address_space *,
2742 loff_t lstart, loff_t lend);
2743 extern void truncate_inode_pages_final(struct address_space *);
2745 /* generic vm_area_ops exported for stackable file systems */
2746 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2747 extern vm_fault_t filemap_map_pages(struct vm_fault *vmf,
2748 pgoff_t start_pgoff, pgoff_t end_pgoff);
2749 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2751 extern unsigned long stack_guard_gap;
2752 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2753 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2755 /* CONFIG_STACK_GROWSUP still needs to grow downwards at some places */
2756 extern int expand_downwards(struct vm_area_struct *vma,
2757 unsigned long address);
2759 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2761 #define expand_upwards(vma, address) (0)
2764 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2765 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2766 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2767 struct vm_area_struct **pprev);
2770 * find_vma_intersection() - Look up the first VMA which intersects the interval
2771 * @mm: The process address space.
2772 * @start_addr: The inclusive start user address.
2773 * @end_addr: The exclusive end user address.
2775 * Returns: The first VMA within the provided range, %NULL otherwise. Assumes
2776 * start_addr < end_addr.
2779 struct vm_area_struct *find_vma_intersection(struct mm_struct *mm,
2780 unsigned long start_addr,
2781 unsigned long end_addr)
2783 struct vm_area_struct *vma = find_vma(mm, start_addr);
2785 if (vma && end_addr <= vma->vm_start)
2791 * vma_lookup() - Find a VMA at a specific address
2792 * @mm: The process address space.
2793 * @addr: The user address.
2795 * Return: The vm_area_struct at the given address, %NULL otherwise.
2798 struct vm_area_struct *vma_lookup(struct mm_struct *mm, unsigned long addr)
2800 struct vm_area_struct *vma = find_vma(mm, addr);
2802 if (vma && addr < vma->vm_start)
2808 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2810 unsigned long vm_start = vma->vm_start;
2812 if (vma->vm_flags & VM_GROWSDOWN) {
2813 vm_start -= stack_guard_gap;
2814 if (vm_start > vma->vm_start)
2820 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2822 unsigned long vm_end = vma->vm_end;
2824 if (vma->vm_flags & VM_GROWSUP) {
2825 vm_end += stack_guard_gap;
2826 if (vm_end < vma->vm_end)
2827 vm_end = -PAGE_SIZE;
2832 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2834 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2837 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2838 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2839 unsigned long vm_start, unsigned long vm_end)
2841 struct vm_area_struct *vma = find_vma(mm, vm_start);
2843 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2849 static inline bool range_in_vma(struct vm_area_struct *vma,
2850 unsigned long start, unsigned long end)
2852 return (vma && vma->vm_start <= start && end <= vma->vm_end);
2856 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2857 void vma_set_page_prot(struct vm_area_struct *vma);
2859 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2863 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2865 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2869 void vma_set_file(struct vm_area_struct *vma, struct file *file);
2871 #ifdef CONFIG_NUMA_BALANCING
2872 unsigned long change_prot_numa(struct vm_area_struct *vma,
2873 unsigned long start, unsigned long end);
2876 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2877 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2878 unsigned long pfn, unsigned long size, pgprot_t);
2879 int remap_pfn_range_notrack(struct vm_area_struct *vma, unsigned long addr,
2880 unsigned long pfn, unsigned long size, pgprot_t prot);
2881 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2882 int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
2883 struct page **pages, unsigned long *num);
2884 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
2886 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
2888 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2890 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2891 unsigned long pfn, pgprot_t pgprot);
2892 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2894 vm_fault_t vmf_insert_mixed_prot(struct vm_area_struct *vma, unsigned long addr,
2895 pfn_t pfn, pgprot_t pgprot);
2896 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
2897 unsigned long addr, pfn_t pfn);
2898 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2900 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2901 unsigned long addr, struct page *page)
2903 int err = vm_insert_page(vma, addr, page);
2906 return VM_FAULT_OOM;
2907 if (err < 0 && err != -EBUSY)
2908 return VM_FAULT_SIGBUS;
2910 return VM_FAULT_NOPAGE;
2913 #ifndef io_remap_pfn_range
2914 static inline int io_remap_pfn_range(struct vm_area_struct *vma,
2915 unsigned long addr, unsigned long pfn,
2916 unsigned long size, pgprot_t prot)
2918 return remap_pfn_range(vma, addr, pfn, size, pgprot_decrypted(prot));
2922 static inline vm_fault_t vmf_error(int err)
2925 return VM_FAULT_OOM;
2926 return VM_FAULT_SIGBUS;
2929 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
2930 unsigned int foll_flags);
2932 #define FOLL_WRITE 0x01 /* check pte is writable */
2933 #define FOLL_TOUCH 0x02 /* mark page accessed */
2934 #define FOLL_GET 0x04 /* do get_page on page */
2935 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2936 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2937 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2938 * and return without waiting upon it */
2939 #define FOLL_NOFAULT 0x80 /* do not fault in pages */
2940 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2941 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2942 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2943 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2944 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2945 #define FOLL_COW 0x4000 /* internal GUP flag */
2946 #define FOLL_ANON 0x8000 /* don't do file mappings */
2947 #define FOLL_LONGTERM 0x10000 /* mapping lifetime is indefinite: see below */
2948 #define FOLL_SPLIT_PMD 0x20000 /* split huge pmd before returning */
2949 #define FOLL_PIN 0x40000 /* pages must be released via unpin_user_page */
2950 #define FOLL_FAST_ONLY 0x80000 /* gup_fast: prevent fall-back to slow gup */
2953 * FOLL_PIN and FOLL_LONGTERM may be used in various combinations with each
2954 * other. Here is what they mean, and how to use them:
2956 * FOLL_LONGTERM indicates that the page will be held for an indefinite time
2957 * period _often_ under userspace control. This is in contrast to
2958 * iov_iter_get_pages(), whose usages are transient.
2960 * FIXME: For pages which are part of a filesystem, mappings are subject to the
2961 * lifetime enforced by the filesystem and we need guarantees that longterm
2962 * users like RDMA and V4L2 only establish mappings which coordinate usage with
2963 * the filesystem. Ideas for this coordination include revoking the longterm
2964 * pin, delaying writeback, bounce buffer page writeback, etc. As FS DAX was
2965 * added after the problem with filesystems was found FS DAX VMAs are
2966 * specifically failed. Filesystem pages are still subject to bugs and use of
2967 * FOLL_LONGTERM should be avoided on those pages.
2969 * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call.
2970 * Currently only get_user_pages() and get_user_pages_fast() support this flag
2971 * and calls to get_user_pages_[un]locked are specifically not allowed. This
2972 * is due to an incompatibility with the FS DAX check and
2973 * FAULT_FLAG_ALLOW_RETRY.
2975 * In the CMA case: long term pins in a CMA region would unnecessarily fragment
2976 * that region. And so, CMA attempts to migrate the page before pinning, when
2977 * FOLL_LONGTERM is specified.
2979 * FOLL_PIN indicates that a special kind of tracking (not just page->_refcount,
2980 * but an additional pin counting system) will be invoked. This is intended for
2981 * anything that gets a page reference and then touches page data (for example,
2982 * Direct IO). This lets the filesystem know that some non-file-system entity is
2983 * potentially changing the pages' data. In contrast to FOLL_GET (whose pages
2984 * are released via put_page()), FOLL_PIN pages must be released, ultimately, by
2985 * a call to unpin_user_page().
2987 * FOLL_PIN is similar to FOLL_GET: both of these pin pages. They use different
2988 * and separate refcounting mechanisms, however, and that means that each has
2989 * its own acquire and release mechanisms:
2991 * FOLL_GET: get_user_pages*() to acquire, and put_page() to release.
2993 * FOLL_PIN: pin_user_pages*() to acquire, and unpin_user_pages to release.
2995 * FOLL_PIN and FOLL_GET are mutually exclusive for a given function call.
2996 * (The underlying pages may experience both FOLL_GET-based and FOLL_PIN-based
2997 * calls applied to them, and that's perfectly OK. This is a constraint on the
2998 * callers, not on the pages.)
3000 * FOLL_PIN should be set internally by the pin_user_pages*() APIs, never
3001 * directly by the caller. That's in order to help avoid mismatches when
3002 * releasing pages: get_user_pages*() pages must be released via put_page(),
3003 * while pin_user_pages*() pages must be released via unpin_user_page().
3005 * Please see Documentation/core-api/pin_user_pages.rst for more information.
3008 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
3010 if (vm_fault & VM_FAULT_OOM)
3012 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
3013 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
3014 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
3020 * Indicates for which pages that are write-protected in the page table,
3021 * whether GUP has to trigger unsharing via FAULT_FLAG_UNSHARE such that the
3022 * GUP pin will remain consistent with the pages mapped into the page tables
3025 * Temporary unmapping of PageAnonExclusive() pages or clearing of
3026 * PageAnonExclusive() has to protect against concurrent GUP:
3027 * * Ordinary GUP: Using the PT lock
3028 * * GUP-fast and fork(): mm->write_protect_seq
3029 * * GUP-fast and KSM or temporary unmapping (swap, migration):
3030 * clear/invalidate+flush of the page table entry
3032 * Must be called with the (sub)page that's actually referenced via the
3033 * page table entry, which might not necessarily be the head page for a
3036 static inline bool gup_must_unshare(unsigned int flags, struct page *page)
3039 * FOLL_WRITE is implicitly handled correctly as the page table entry
3040 * has to be writable -- and if it references (part of) an anonymous
3041 * folio, that part is required to be marked exclusive.
3043 if ((flags & (FOLL_WRITE | FOLL_PIN)) != FOLL_PIN)
3046 * Note: PageAnon(page) is stable until the page is actually getting
3049 if (!PageAnon(page))
3052 * Note that PageKsm() pages cannot be exclusive, and consequently,
3053 * cannot get pinned.
3055 return !PageAnonExclusive(page);
3058 typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
3059 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
3060 unsigned long size, pte_fn_t fn, void *data);
3061 extern int apply_to_existing_page_range(struct mm_struct *mm,
3062 unsigned long address, unsigned long size,
3063 pte_fn_t fn, void *data);
3065 extern void init_mem_debugging_and_hardening(void);
3066 #ifdef CONFIG_PAGE_POISONING
3067 extern void __kernel_poison_pages(struct page *page, int numpages);
3068 extern void __kernel_unpoison_pages(struct page *page, int numpages);
3069 extern bool _page_poisoning_enabled_early;
3070 DECLARE_STATIC_KEY_FALSE(_page_poisoning_enabled);
3071 static inline bool page_poisoning_enabled(void)
3073 return _page_poisoning_enabled_early;
3076 * For use in fast paths after init_mem_debugging() has run, or when a
3077 * false negative result is not harmful when called too early.
3079 static inline bool page_poisoning_enabled_static(void)
3081 return static_branch_unlikely(&_page_poisoning_enabled);
3083 static inline void kernel_poison_pages(struct page *page, int numpages)
3085 if (page_poisoning_enabled_static())
3086 __kernel_poison_pages(page, numpages);
3088 static inline void kernel_unpoison_pages(struct page *page, int numpages)
3090 if (page_poisoning_enabled_static())
3091 __kernel_unpoison_pages(page, numpages);
3094 static inline bool page_poisoning_enabled(void) { return false; }
3095 static inline bool page_poisoning_enabled_static(void) { return false; }
3096 static inline void __kernel_poison_pages(struct page *page, int nunmpages) { }
3097 static inline void kernel_poison_pages(struct page *page, int numpages) { }
3098 static inline void kernel_unpoison_pages(struct page *page, int numpages) { }
3101 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_ALLOC_DEFAULT_ON, init_on_alloc);
3102 static inline bool want_init_on_alloc(gfp_t flags)
3104 if (static_branch_maybe(CONFIG_INIT_ON_ALLOC_DEFAULT_ON,
3107 return flags & __GFP_ZERO;
3110 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_FREE_DEFAULT_ON, init_on_free);
3111 static inline bool want_init_on_free(void)
3113 return static_branch_maybe(CONFIG_INIT_ON_FREE_DEFAULT_ON,
3117 extern bool _debug_pagealloc_enabled_early;
3118 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
3120 static inline bool debug_pagealloc_enabled(void)
3122 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
3123 _debug_pagealloc_enabled_early;
3127 * For use in fast paths after init_debug_pagealloc() has run, or when a
3128 * false negative result is not harmful when called too early.
3130 static inline bool debug_pagealloc_enabled_static(void)
3132 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
3135 return static_branch_unlikely(&_debug_pagealloc_enabled);
3138 #ifdef CONFIG_DEBUG_PAGEALLOC
3140 * To support DEBUG_PAGEALLOC architecture must ensure that
3141 * __kernel_map_pages() never fails
3143 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
3145 static inline void debug_pagealloc_map_pages(struct page *page, int numpages)
3147 if (debug_pagealloc_enabled_static())
3148 __kernel_map_pages(page, numpages, 1);
3151 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages)
3153 if (debug_pagealloc_enabled_static())
3154 __kernel_map_pages(page, numpages, 0);
3156 #else /* CONFIG_DEBUG_PAGEALLOC */
3157 static inline void debug_pagealloc_map_pages(struct page *page, int numpages) {}
3158 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages) {}
3159 #endif /* CONFIG_DEBUG_PAGEALLOC */
3161 #ifdef __HAVE_ARCH_GATE_AREA
3162 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
3163 extern int in_gate_area_no_mm(unsigned long addr);
3164 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
3166 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
3170 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
3171 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
3175 #endif /* __HAVE_ARCH_GATE_AREA */
3177 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
3179 #ifdef CONFIG_SYSCTL
3180 extern int sysctl_drop_caches;
3181 int drop_caches_sysctl_handler(struct ctl_table *, int, void *, size_t *,
3185 void drop_slab(void);
3188 #define randomize_va_space 0
3190 extern int randomize_va_space;
3193 const char * arch_vma_name(struct vm_area_struct *vma);
3195 void print_vma_addr(char *prefix, unsigned long rip);
3197 static inline void print_vma_addr(char *prefix, unsigned long rip)
3202 #ifdef CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP
3203 int vmemmap_remap_free(unsigned long start, unsigned long end,
3204 unsigned long reuse);
3205 int vmemmap_remap_alloc(unsigned long start, unsigned long end,
3206 unsigned long reuse, gfp_t gfp_mask);
3209 void *sparse_buffer_alloc(unsigned long size);
3210 struct page * __populate_section_memmap(unsigned long pfn,
3211 unsigned long nr_pages, int nid, struct vmem_altmap *altmap,
3212 struct dev_pagemap *pgmap);
3213 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
3214 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
3215 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
3216 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
3217 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node,
3218 struct vmem_altmap *altmap, struct page *reuse);
3219 void *vmemmap_alloc_block(unsigned long size, int node);
3221 void *vmemmap_alloc_block_buf(unsigned long size, int node,
3222 struct vmem_altmap *altmap);
3223 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
3224 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
3225 int node, struct vmem_altmap *altmap);
3226 int vmemmap_populate(unsigned long start, unsigned long end, int node,
3227 struct vmem_altmap *altmap);
3228 void vmemmap_populate_print_last(void);
3229 #ifdef CONFIG_MEMORY_HOTPLUG
3230 void vmemmap_free(unsigned long start, unsigned long end,
3231 struct vmem_altmap *altmap);
3233 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
3234 unsigned long nr_pages);
3237 MF_COUNT_INCREASED = 1 << 0,
3238 MF_ACTION_REQUIRED = 1 << 1,
3239 MF_MUST_KILL = 1 << 2,
3240 MF_SOFT_OFFLINE = 1 << 3,
3241 MF_UNPOISON = 1 << 4,
3242 MF_SW_SIMULATED = 1 << 5,
3244 extern int memory_failure(unsigned long pfn, int flags);
3245 extern void memory_failure_queue(unsigned long pfn, int flags);
3246 extern void memory_failure_queue_kick(int cpu);
3247 extern int unpoison_memory(unsigned long pfn);
3248 extern int sysctl_memory_failure_early_kill;
3249 extern int sysctl_memory_failure_recovery;
3250 extern void shake_page(struct page *p);
3251 extern atomic_long_t num_poisoned_pages __read_mostly;
3252 extern int soft_offline_page(unsigned long pfn, int flags);
3253 #ifdef CONFIG_MEMORY_FAILURE
3254 extern int __get_huge_page_for_hwpoison(unsigned long pfn, int flags);
3256 static inline int __get_huge_page_for_hwpoison(unsigned long pfn, int flags)
3262 #ifndef arch_memory_failure
3263 static inline int arch_memory_failure(unsigned long pfn, int flags)
3269 #ifndef arch_is_platform_page
3270 static inline bool arch_is_platform_page(u64 paddr)
3277 * Error handlers for various types of pages.
3280 MF_IGNORED, /* Error: cannot be handled */
3281 MF_FAILED, /* Error: handling failed */
3282 MF_DELAYED, /* Will be handled later */
3283 MF_RECOVERED, /* Successfully recovered */
3286 enum mf_action_page_type {
3288 MF_MSG_KERNEL_HIGH_ORDER,
3290 MF_MSG_DIFFERENT_COMPOUND,
3293 MF_MSG_NON_PMD_HUGE,
3294 MF_MSG_UNMAP_FAILED,
3295 MF_MSG_DIRTY_SWAPCACHE,
3296 MF_MSG_CLEAN_SWAPCACHE,
3297 MF_MSG_DIRTY_MLOCKED_LRU,
3298 MF_MSG_CLEAN_MLOCKED_LRU,
3299 MF_MSG_DIRTY_UNEVICTABLE_LRU,
3300 MF_MSG_CLEAN_UNEVICTABLE_LRU,
3303 MF_MSG_TRUNCATED_LRU,
3310 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
3311 extern void clear_huge_page(struct page *page,
3312 unsigned long addr_hint,
3313 unsigned int pages_per_huge_page);
3314 extern void copy_user_huge_page(struct page *dst, struct page *src,
3315 unsigned long addr_hint,
3316 struct vm_area_struct *vma,
3317 unsigned int pages_per_huge_page);
3318 extern long copy_huge_page_from_user(struct page *dst_page,
3319 const void __user *usr_src,
3320 unsigned int pages_per_huge_page,
3321 bool allow_pagefault);
3324 * vma_is_special_huge - Are transhuge page-table entries considered special?
3325 * @vma: Pointer to the struct vm_area_struct to consider
3327 * Whether transhuge page-table entries are considered "special" following
3328 * the definition in vm_normal_page().
3330 * Return: true if transhuge page-table entries should be considered special,
3333 static inline bool vma_is_special_huge(const struct vm_area_struct *vma)
3335 return vma_is_dax(vma) || (vma->vm_file &&
3336 (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP)));
3339 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
3341 #ifdef CONFIG_DEBUG_PAGEALLOC
3342 extern unsigned int _debug_guardpage_minorder;
3343 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
3345 static inline unsigned int debug_guardpage_minorder(void)
3347 return _debug_guardpage_minorder;
3350 static inline bool debug_guardpage_enabled(void)
3352 return static_branch_unlikely(&_debug_guardpage_enabled);
3355 static inline bool page_is_guard(struct page *page)
3357 if (!debug_guardpage_enabled())
3360 return PageGuard(page);
3363 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
3364 static inline bool debug_guardpage_enabled(void) { return false; }
3365 static inline bool page_is_guard(struct page *page) { return false; }
3366 #endif /* CONFIG_DEBUG_PAGEALLOC */
3368 #if MAX_NUMNODES > 1
3369 void __init setup_nr_node_ids(void);
3371 static inline void setup_nr_node_ids(void) {}
3374 extern int memcmp_pages(struct page *page1, struct page *page2);
3376 static inline int pages_identical(struct page *page1, struct page *page2)
3378 return !memcmp_pages(page1, page2);
3381 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
3382 unsigned long clean_record_shared_mapping_range(struct address_space *mapping,
3383 pgoff_t first_index, pgoff_t nr,
3384 pgoff_t bitmap_pgoff,
3385 unsigned long *bitmap,
3389 unsigned long wp_shared_mapping_range(struct address_space *mapping,
3390 pgoff_t first_index, pgoff_t nr);
3393 extern int sysctl_nr_trim_pages;
3395 #ifdef CONFIG_PRINTK
3396 void mem_dump_obj(void *object);
3398 static inline void mem_dump_obj(void *object) {}
3402 * seal_check_future_write - Check for F_SEAL_FUTURE_WRITE flag and handle it
3403 * @seals: the seals to check
3404 * @vma: the vma to operate on
3406 * Check whether F_SEAL_FUTURE_WRITE is set; if so, do proper check/handling on
3407 * the vma flags. Return 0 if check pass, or <0 for errors.
3409 static inline int seal_check_future_write(int seals, struct vm_area_struct *vma)
3411 if (seals & F_SEAL_FUTURE_WRITE) {
3413 * New PROT_WRITE and MAP_SHARED mmaps are not allowed when
3414 * "future write" seal active.
3416 if ((vma->vm_flags & VM_SHARED) && (vma->vm_flags & VM_WRITE))
3420 * Since an F_SEAL_FUTURE_WRITE sealed memfd can be mapped as
3421 * MAP_SHARED and read-only, take care to not allow mprotect to
3422 * revert protections on such mappings. Do this only for shared
3423 * mappings. For private mappings, don't need to mask
3424 * VM_MAYWRITE as we still want them to be COW-writable.
3426 if (vma->vm_flags & VM_SHARED)
3427 vma->vm_flags &= ~(VM_MAYWRITE);
3433 #ifdef CONFIG_ANON_VMA_NAME
3434 int madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
3435 unsigned long len_in,
3436 struct anon_vma_name *anon_name);
3439 madvise_set_anon_name(struct mm_struct *mm, unsigned long start,
3440 unsigned long len_in, struct anon_vma_name *anon_name) {
3446 * Whether to drop the pte markers, for example, the uffd-wp information for
3447 * file-backed memory. This should only be specified when we will completely
3448 * drop the page in the mm, either by truncation or unmapping of the vma. By
3449 * default, the flag is not set.
3451 #define ZAP_FLAG_DROP_MARKER ((__force zap_flags_t) BIT(0))
3453 #endif /* _LINUX_MM_H */