1 /* SPDX-License-Identifier: GPL-2.0 */
5 #include <linux/errno.h>
9 #include <linux/mmdebug.h>
10 #include <linux/gfp.h>
11 #include <linux/bug.h>
12 #include <linux/list.h>
13 #include <linux/mmzone.h>
14 #include <linux/rbtree.h>
15 #include <linux/atomic.h>
16 #include <linux/debug_locks.h>
17 #include <linux/mm_types.h>
18 #include <linux/mmap_lock.h>
19 #include <linux/range.h>
20 #include <linux/pfn.h>
21 #include <linux/percpu-refcount.h>
22 #include <linux/bit_spinlock.h>
23 #include <linux/shrinker.h>
24 #include <linux/resource.h>
25 #include <linux/page_ext.h>
26 #include <linux/err.h>
27 #include <linux/page-flags.h>
28 #include <linux/page_ref.h>
29 #include <linux/memremap.h>
30 #include <linux/overflow.h>
31 #include <linux/sizes.h>
32 #include <linux/sched.h>
33 #include <linux/pgtable.h>
34 #include <linux/kasan.h>
38 struct anon_vma_chain;
41 struct writeback_control;
45 extern int sysctl_page_lock_unfairness;
47 void init_mm_internals(void);
49 #ifndef CONFIG_NUMA /* Don't use mapnrs, do it properly */
50 extern unsigned long max_mapnr;
52 static inline void set_max_mapnr(unsigned long limit)
57 static inline void set_max_mapnr(unsigned long limit) { }
60 extern atomic_long_t _totalram_pages;
61 static inline unsigned long totalram_pages(void)
63 return (unsigned long)atomic_long_read(&_totalram_pages);
66 static inline void totalram_pages_inc(void)
68 atomic_long_inc(&_totalram_pages);
71 static inline void totalram_pages_dec(void)
73 atomic_long_dec(&_totalram_pages);
76 static inline void totalram_pages_add(long count)
78 atomic_long_add(count, &_totalram_pages);
81 extern void * high_memory;
82 extern int page_cluster;
85 extern int sysctl_legacy_va_layout;
87 #define sysctl_legacy_va_layout 0
90 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
91 extern const int mmap_rnd_bits_min;
92 extern const int mmap_rnd_bits_max;
93 extern int mmap_rnd_bits __read_mostly;
95 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
96 extern const int mmap_rnd_compat_bits_min;
97 extern const int mmap_rnd_compat_bits_max;
98 extern int mmap_rnd_compat_bits __read_mostly;
101 #include <asm/page.h>
102 #include <asm/processor.h>
105 * Architectures that support memory tagging (assigning tags to memory regions,
106 * embedding these tags into addresses that point to these memory regions, and
107 * checking that the memory and the pointer tags match on memory accesses)
108 * redefine this macro to strip tags from pointers.
109 * It's defined as noop for architectures that don't support memory tagging.
111 #ifndef untagged_addr
112 #define untagged_addr(addr) (addr)
116 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
120 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
124 #define lm_alias(x) __va(__pa_symbol(x))
128 * To prevent common memory management code establishing
129 * a zero page mapping on a read fault.
130 * This macro should be defined within <asm/pgtable.h>.
131 * s390 does this to prevent multiplexing of hardware bits
132 * related to the physical page in case of virtualization.
134 #ifndef mm_forbids_zeropage
135 #define mm_forbids_zeropage(X) (0)
139 * On some architectures it is expensive to call memset() for small sizes.
140 * If an architecture decides to implement their own version of
141 * mm_zero_struct_page they should wrap the defines below in a #ifndef and
142 * define their own version of this macro in <asm/pgtable.h>
144 #if BITS_PER_LONG == 64
145 /* This function must be updated when the size of struct page grows above 80
146 * or reduces below 56. The idea that compiler optimizes out switch()
147 * statement, and only leaves move/store instructions. Also the compiler can
148 * combine write statements if they are both assignments and can be reordered,
149 * this can result in several of the writes here being dropped.
151 #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
152 static inline void __mm_zero_struct_page(struct page *page)
154 unsigned long *_pp = (void *)page;
156 /* Check that struct page is either 56, 64, 72, or 80 bytes */
157 BUILD_BUG_ON(sizeof(struct page) & 7);
158 BUILD_BUG_ON(sizeof(struct page) < 56);
159 BUILD_BUG_ON(sizeof(struct page) > 80);
161 switch (sizeof(struct page)) {
182 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
186 * Default maximum number of active map areas, this limits the number of vmas
187 * per mm struct. Users can overwrite this number by sysctl but there is a
190 * When a program's coredump is generated as ELF format, a section is created
191 * per a vma. In ELF, the number of sections is represented in unsigned short.
192 * This means the number of sections should be smaller than 65535 at coredump.
193 * Because the kernel adds some informative sections to a image of program at
194 * generating coredump, we need some margin. The number of extra sections is
195 * 1-3 now and depends on arch. We use "5" as safe margin, here.
197 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
198 * not a hard limit any more. Although some userspace tools can be surprised by
201 #define MAPCOUNT_ELF_CORE_MARGIN (5)
202 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
204 extern int sysctl_max_map_count;
206 extern unsigned long sysctl_user_reserve_kbytes;
207 extern unsigned long sysctl_admin_reserve_kbytes;
209 extern int sysctl_overcommit_memory;
210 extern int sysctl_overcommit_ratio;
211 extern unsigned long sysctl_overcommit_kbytes;
213 int overcommit_ratio_handler(struct ctl_table *, int, void *, size_t *,
215 int overcommit_kbytes_handler(struct ctl_table *, int, void *, size_t *,
217 int overcommit_policy_handler(struct ctl_table *, int, void *, size_t *,
220 * Any attempt to mark this function as static leads to build failure
221 * when CONFIG_DEBUG_INFO_BTF is enabled because __add_to_page_cache_locked()
222 * is referred to by BPF code. This must be visible for error injection.
224 int __add_to_page_cache_locked(struct page *page, struct address_space *mapping,
225 pgoff_t index, gfp_t gfp, void **shadowp);
227 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
228 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
230 #define nth_page(page,n) ((page) + (n))
233 /* to align the pointer to the (next) page boundary */
234 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
236 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
237 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
239 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
241 void setup_initial_init_mm(void *start_code, void *end_code,
242 void *end_data, void *brk);
245 * Linux kernel virtual memory manager primitives.
246 * The idea being to have a "virtual" mm in the same way
247 * we have a virtual fs - giving a cleaner interface to the
248 * mm details, and allowing different kinds of memory mappings
249 * (from shared memory to executable loading to arbitrary
253 struct vm_area_struct *vm_area_alloc(struct mm_struct *);
254 struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
255 void vm_area_free(struct vm_area_struct *);
258 extern struct rb_root nommu_region_tree;
259 extern struct rw_semaphore nommu_region_sem;
261 extern unsigned int kobjsize(const void *objp);
265 * vm_flags in vm_area_struct, see mm_types.h.
266 * When changing, update also include/trace/events/mmflags.h
268 #define VM_NONE 0x00000000
270 #define VM_READ 0x00000001 /* currently active flags */
271 #define VM_WRITE 0x00000002
272 #define VM_EXEC 0x00000004
273 #define VM_SHARED 0x00000008
275 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
276 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
277 #define VM_MAYWRITE 0x00000020
278 #define VM_MAYEXEC 0x00000040
279 #define VM_MAYSHARE 0x00000080
281 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
282 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
283 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
284 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
285 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
287 #define VM_LOCKED 0x00002000
288 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
290 /* Used by sys_madvise() */
291 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
292 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
294 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
295 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
296 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
297 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
298 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
299 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
300 #define VM_SYNC 0x00800000 /* Synchronous page faults */
301 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
302 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
303 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
305 #ifdef CONFIG_MEM_SOFT_DIRTY
306 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
308 # define VM_SOFTDIRTY 0
311 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
312 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
313 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
314 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
316 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
317 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
318 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
319 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
320 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
321 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
322 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
323 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
324 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
325 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
326 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
327 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
329 #ifdef CONFIG_ARCH_HAS_PKEYS
330 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
331 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
332 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
333 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
334 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
336 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
338 # define VM_PKEY_BIT4 0
340 #endif /* CONFIG_ARCH_HAS_PKEYS */
342 #if defined(CONFIG_X86)
343 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
344 #elif defined(CONFIG_PPC)
345 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
346 #elif defined(CONFIG_PARISC)
347 # define VM_GROWSUP VM_ARCH_1
348 #elif defined(CONFIG_IA64)
349 # define VM_GROWSUP VM_ARCH_1
350 #elif defined(CONFIG_SPARC64)
351 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
352 # define VM_ARCH_CLEAR VM_SPARC_ADI
353 #elif defined(CONFIG_ARM64)
354 # define VM_ARM64_BTI VM_ARCH_1 /* BTI guarded page, a.k.a. GP bit */
355 # define VM_ARCH_CLEAR VM_ARM64_BTI
356 #elif !defined(CONFIG_MMU)
357 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
360 #if defined(CONFIG_ARM64_MTE)
361 # define VM_MTE VM_HIGH_ARCH_0 /* Use Tagged memory for access control */
362 # define VM_MTE_ALLOWED VM_HIGH_ARCH_1 /* Tagged memory permitted */
364 # define VM_MTE VM_NONE
365 # define VM_MTE_ALLOWED VM_NONE
369 # define VM_GROWSUP VM_NONE
372 #ifdef CONFIG_HAVE_ARCH_USERFAULTFD_MINOR
373 # define VM_UFFD_MINOR_BIT 37
374 # define VM_UFFD_MINOR BIT(VM_UFFD_MINOR_BIT) /* UFFD minor faults */
375 #else /* !CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
376 # define VM_UFFD_MINOR VM_NONE
377 #endif /* CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
379 /* Bits set in the VMA until the stack is in its final location */
380 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
382 #define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0)
384 /* Common data flag combinations */
385 #define VM_DATA_FLAGS_TSK_EXEC (VM_READ | VM_WRITE | TASK_EXEC | \
386 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
387 #define VM_DATA_FLAGS_NON_EXEC (VM_READ | VM_WRITE | VM_MAYREAD | \
388 VM_MAYWRITE | VM_MAYEXEC)
389 #define VM_DATA_FLAGS_EXEC (VM_READ | VM_WRITE | VM_EXEC | \
390 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
392 #ifndef VM_DATA_DEFAULT_FLAGS /* arch can override this */
393 #define VM_DATA_DEFAULT_FLAGS VM_DATA_FLAGS_EXEC
396 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
397 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
400 #ifdef CONFIG_STACK_GROWSUP
401 #define VM_STACK VM_GROWSUP
403 #define VM_STACK VM_GROWSDOWN
406 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
408 /* VMA basic access permission flags */
409 #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
413 * Special vmas that are non-mergable, non-mlock()able.
415 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
417 /* This mask prevents VMA from being scanned with khugepaged */
418 #define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB)
420 /* This mask defines which mm->def_flags a process can inherit its parent */
421 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
423 /* This mask is used to clear all the VMA flags used by mlock */
424 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
426 /* Arch-specific flags to clear when updating VM flags on protection change */
427 #ifndef VM_ARCH_CLEAR
428 # define VM_ARCH_CLEAR VM_NONE
430 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
433 * mapping from the currently active vm_flags protection bits (the
434 * low four bits) to a page protection mask..
436 extern pgprot_t protection_map[16];
439 * enum fault_flag - Fault flag definitions.
440 * @FAULT_FLAG_WRITE: Fault was a write fault.
441 * @FAULT_FLAG_MKWRITE: Fault was mkwrite of existing PTE.
442 * @FAULT_FLAG_ALLOW_RETRY: Allow to retry the fault if blocked.
443 * @FAULT_FLAG_RETRY_NOWAIT: Don't drop mmap_lock and wait when retrying.
444 * @FAULT_FLAG_KILLABLE: The fault task is in SIGKILL killable region.
445 * @FAULT_FLAG_TRIED: The fault has been tried once.
446 * @FAULT_FLAG_USER: The fault originated in userspace.
447 * @FAULT_FLAG_REMOTE: The fault is not for current task/mm.
448 * @FAULT_FLAG_INSTRUCTION: The fault was during an instruction fetch.
449 * @FAULT_FLAG_INTERRUPTIBLE: The fault can be interrupted by non-fatal signals.
451 * About @FAULT_FLAG_ALLOW_RETRY and @FAULT_FLAG_TRIED: we can specify
452 * whether we would allow page faults to retry by specifying these two
453 * fault flags correctly. Currently there can be three legal combinations:
455 * (a) ALLOW_RETRY and !TRIED: this means the page fault allows retry, and
456 * this is the first try
458 * (b) ALLOW_RETRY and TRIED: this means the page fault allows retry, and
459 * we've already tried at least once
461 * (c) !ALLOW_RETRY and !TRIED: this means the page fault does not allow retry
463 * The unlisted combination (!ALLOW_RETRY && TRIED) is illegal and should never
464 * be used. Note that page faults can be allowed to retry for multiple times,
465 * in which case we'll have an initial fault with flags (a) then later on
466 * continuous faults with flags (b). We should always try to detect pending
467 * signals before a retry to make sure the continuous page faults can still be
468 * interrupted if necessary.
471 FAULT_FLAG_WRITE = 1 << 0,
472 FAULT_FLAG_MKWRITE = 1 << 1,
473 FAULT_FLAG_ALLOW_RETRY = 1 << 2,
474 FAULT_FLAG_RETRY_NOWAIT = 1 << 3,
475 FAULT_FLAG_KILLABLE = 1 << 4,
476 FAULT_FLAG_TRIED = 1 << 5,
477 FAULT_FLAG_USER = 1 << 6,
478 FAULT_FLAG_REMOTE = 1 << 7,
479 FAULT_FLAG_INSTRUCTION = 1 << 8,
480 FAULT_FLAG_INTERRUPTIBLE = 1 << 9,
484 * The default fault flags that should be used by most of the
485 * arch-specific page fault handlers.
487 #define FAULT_FLAG_DEFAULT (FAULT_FLAG_ALLOW_RETRY | \
488 FAULT_FLAG_KILLABLE | \
489 FAULT_FLAG_INTERRUPTIBLE)
492 * fault_flag_allow_retry_first - check ALLOW_RETRY the first time
493 * @flags: Fault flags.
495 * This is mostly used for places where we want to try to avoid taking
496 * the mmap_lock for too long a time when waiting for another condition
497 * to change, in which case we can try to be polite to release the
498 * mmap_lock in the first round to avoid potential starvation of other
499 * processes that would also want the mmap_lock.
501 * Return: true if the page fault allows retry and this is the first
502 * attempt of the fault handling; false otherwise.
504 static inline bool fault_flag_allow_retry_first(enum fault_flag flags)
506 return (flags & FAULT_FLAG_ALLOW_RETRY) &&
507 (!(flags & FAULT_FLAG_TRIED));
510 #define FAULT_FLAG_TRACE \
511 { FAULT_FLAG_WRITE, "WRITE" }, \
512 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
513 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
514 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
515 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
516 { FAULT_FLAG_TRIED, "TRIED" }, \
517 { FAULT_FLAG_USER, "USER" }, \
518 { FAULT_FLAG_REMOTE, "REMOTE" }, \
519 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }, \
520 { FAULT_FLAG_INTERRUPTIBLE, "INTERRUPTIBLE" }
523 * vm_fault is filled by the pagefault handler and passed to the vma's
524 * ->fault function. The vma's ->fault is responsible for returning a bitmask
525 * of VM_FAULT_xxx flags that give details about how the fault was handled.
527 * MM layer fills up gfp_mask for page allocations but fault handler might
528 * alter it if its implementation requires a different allocation context.
530 * pgoff should be used in favour of virtual_address, if possible.
534 struct vm_area_struct *vma; /* Target VMA */
535 gfp_t gfp_mask; /* gfp mask to be used for allocations */
536 pgoff_t pgoff; /* Logical page offset based on vma */
537 unsigned long address; /* Faulting virtual address */
539 enum fault_flag flags; /* FAULT_FLAG_xxx flags
540 * XXX: should really be 'const' */
541 pmd_t *pmd; /* Pointer to pmd entry matching
543 pud_t *pud; /* Pointer to pud entry matching
547 pte_t orig_pte; /* Value of PTE at the time of fault */
548 pmd_t orig_pmd; /* Value of PMD at the time of fault,
549 * used by PMD fault only.
553 struct page *cow_page; /* Page handler may use for COW fault */
554 struct page *page; /* ->fault handlers should return a
555 * page here, unless VM_FAULT_NOPAGE
556 * is set (which is also implied by
559 /* These three entries are valid only while holding ptl lock */
560 pte_t *pte; /* Pointer to pte entry matching
561 * the 'address'. NULL if the page
562 * table hasn't been allocated.
564 spinlock_t *ptl; /* Page table lock.
565 * Protects pte page table if 'pte'
566 * is not NULL, otherwise pmd.
568 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
569 * vm_ops->map_pages() sets up a page
570 * table from atomic context.
571 * do_fault_around() pre-allocates
572 * page table to avoid allocation from
577 /* page entry size for vm->huge_fault() */
578 enum page_entry_size {
585 * These are the virtual MM functions - opening of an area, closing and
586 * unmapping it (needed to keep files on disk up-to-date etc), pointer
587 * to the functions called when a no-page or a wp-page exception occurs.
589 struct vm_operations_struct {
590 void (*open)(struct vm_area_struct * area);
591 void (*close)(struct vm_area_struct * area);
592 /* Called any time before splitting to check if it's allowed */
593 int (*may_split)(struct vm_area_struct *area, unsigned long addr);
594 int (*mremap)(struct vm_area_struct *area);
596 * Called by mprotect() to make driver-specific permission
597 * checks before mprotect() is finalised. The VMA must not
598 * be modified. Returns 0 if eprotect() can proceed.
600 int (*mprotect)(struct vm_area_struct *vma, unsigned long start,
601 unsigned long end, unsigned long newflags);
602 vm_fault_t (*fault)(struct vm_fault *vmf);
603 vm_fault_t (*huge_fault)(struct vm_fault *vmf,
604 enum page_entry_size pe_size);
605 vm_fault_t (*map_pages)(struct vm_fault *vmf,
606 pgoff_t start_pgoff, pgoff_t end_pgoff);
607 unsigned long (*pagesize)(struct vm_area_struct * area);
609 /* notification that a previously read-only page is about to become
610 * writable, if an error is returned it will cause a SIGBUS */
611 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
613 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
614 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
616 /* called by access_process_vm when get_user_pages() fails, typically
617 * for use by special VMAs. See also generic_access_phys() for a generic
618 * implementation useful for any iomem mapping.
620 int (*access)(struct vm_area_struct *vma, unsigned long addr,
621 void *buf, int len, int write);
623 /* Called by the /proc/PID/maps code to ask the vma whether it
624 * has a special name. Returning non-NULL will also cause this
625 * vma to be dumped unconditionally. */
626 const char *(*name)(struct vm_area_struct *vma);
630 * set_policy() op must add a reference to any non-NULL @new mempolicy
631 * to hold the policy upon return. Caller should pass NULL @new to
632 * remove a policy and fall back to surrounding context--i.e. do not
633 * install a MPOL_DEFAULT policy, nor the task or system default
636 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
639 * get_policy() op must add reference [mpol_get()] to any policy at
640 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
641 * in mm/mempolicy.c will do this automatically.
642 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
643 * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
644 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
645 * must return NULL--i.e., do not "fallback" to task or system default
648 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
652 * Called by vm_normal_page() for special PTEs to find the
653 * page for @addr. This is useful if the default behavior
654 * (using pte_page()) would not find the correct page.
656 struct page *(*find_special_page)(struct vm_area_struct *vma,
660 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
662 static const struct vm_operations_struct dummy_vm_ops = {};
664 memset(vma, 0, sizeof(*vma));
666 vma->vm_ops = &dummy_vm_ops;
667 INIT_LIST_HEAD(&vma->anon_vma_chain);
670 static inline void vma_set_anonymous(struct vm_area_struct *vma)
675 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
680 static inline bool vma_is_temporary_stack(struct vm_area_struct *vma)
682 int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP);
687 if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) ==
688 VM_STACK_INCOMPLETE_SETUP)
694 static inline bool vma_is_foreign(struct vm_area_struct *vma)
699 if (current->mm != vma->vm_mm)
705 static inline bool vma_is_accessible(struct vm_area_struct *vma)
707 return vma->vm_flags & VM_ACCESS_FLAGS;
712 * The vma_is_shmem is not inline because it is used only by slow
713 * paths in userfault.
715 bool vma_is_shmem(struct vm_area_struct *vma);
717 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
720 int vma_is_stack_for_current(struct vm_area_struct *vma);
722 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
723 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
728 #include <linux/huge_mm.h>
731 * Methods to modify the page usage count.
733 * What counts for a page usage:
734 * - cache mapping (page->mapping)
735 * - private data (page->private)
736 * - page mapped in a task's page tables, each mapping
737 * is counted separately
739 * Also, many kernel routines increase the page count before a critical
740 * routine so they can be sure the page doesn't go away from under them.
744 * Drop a ref, return true if the refcount fell to zero (the page has no users)
746 static inline int put_page_testzero(struct page *page)
748 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
749 return page_ref_dec_and_test(page);
753 * Try to grab a ref unless the page has a refcount of zero, return false if
755 * This can be called when MMU is off so it must not access
756 * any of the virtual mappings.
758 static inline int get_page_unless_zero(struct page *page)
760 return page_ref_add_unless(page, 1, 0);
763 extern int page_is_ram(unsigned long pfn);
771 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
774 /* Support for virtually mapped pages */
775 struct page *vmalloc_to_page(const void *addr);
776 unsigned long vmalloc_to_pfn(const void *addr);
779 * Determine if an address is within the vmalloc range
781 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
782 * is no special casing required.
785 #ifndef is_ioremap_addr
786 #define is_ioremap_addr(x) is_vmalloc_addr(x)
790 extern bool is_vmalloc_addr(const void *x);
791 extern int is_vmalloc_or_module_addr(const void *x);
793 static inline bool is_vmalloc_addr(const void *x)
797 static inline int is_vmalloc_or_module_addr(const void *x)
803 extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
804 static inline void *kvmalloc(size_t size, gfp_t flags)
806 return kvmalloc_node(size, flags, NUMA_NO_NODE);
808 static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
810 return kvmalloc_node(size, flags | __GFP_ZERO, node);
812 static inline void *kvzalloc(size_t size, gfp_t flags)
814 return kvmalloc(size, flags | __GFP_ZERO);
817 static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
821 if (unlikely(check_mul_overflow(n, size, &bytes)))
824 return kvmalloc(bytes, flags);
827 static inline void *kvcalloc(size_t n, size_t size, gfp_t flags)
829 return kvmalloc_array(n, size, flags | __GFP_ZERO);
832 extern void *kvrealloc(const void *p, size_t oldsize, size_t newsize,
834 extern void kvfree(const void *addr);
835 extern void kvfree_sensitive(const void *addr, size_t len);
837 static inline int head_compound_mapcount(struct page *head)
839 return atomic_read(compound_mapcount_ptr(head)) + 1;
843 * Mapcount of compound page as a whole, does not include mapped sub-pages.
845 * Must be called only for compound pages or any their tail sub-pages.
847 static inline int compound_mapcount(struct page *page)
849 VM_BUG_ON_PAGE(!PageCompound(page), page);
850 page = compound_head(page);
851 return head_compound_mapcount(page);
855 * The atomic page->_mapcount, starts from -1: so that transitions
856 * both from it and to it can be tracked, using atomic_inc_and_test
857 * and atomic_add_negative(-1).
859 static inline void page_mapcount_reset(struct page *page)
861 atomic_set(&(page)->_mapcount, -1);
864 int __page_mapcount(struct page *page);
867 * Mapcount of 0-order page; when compound sub-page, includes
868 * compound_mapcount().
870 * Result is undefined for pages which cannot be mapped into userspace.
871 * For example SLAB or special types of pages. See function page_has_type().
872 * They use this place in struct page differently.
874 static inline int page_mapcount(struct page *page)
876 if (unlikely(PageCompound(page)))
877 return __page_mapcount(page);
878 return atomic_read(&page->_mapcount) + 1;
881 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
882 int total_mapcount(struct page *page);
883 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
885 static inline int total_mapcount(struct page *page)
887 return page_mapcount(page);
889 static inline int page_trans_huge_mapcount(struct page *page,
892 int mapcount = page_mapcount(page);
894 *total_mapcount = mapcount;
899 static inline struct page *virt_to_head_page(const void *x)
901 struct page *page = virt_to_page(x);
903 return compound_head(page);
906 void __put_page(struct page *page);
908 void put_pages_list(struct list_head *pages);
910 void split_page(struct page *page, unsigned int order);
911 void copy_huge_page(struct page *dst, struct page *src);
914 * Compound pages have a destructor function. Provide a
915 * prototype for that function and accessor functions.
916 * These are _only_ valid on the head of a compound page.
918 typedef void compound_page_dtor(struct page *);
920 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
921 enum compound_dtor_id {
924 #ifdef CONFIG_HUGETLB_PAGE
927 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
932 extern compound_page_dtor * const compound_page_dtors[NR_COMPOUND_DTORS];
934 static inline void set_compound_page_dtor(struct page *page,
935 enum compound_dtor_id compound_dtor)
937 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
938 page[1].compound_dtor = compound_dtor;
941 static inline void destroy_compound_page(struct page *page)
943 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
944 compound_page_dtors[page[1].compound_dtor](page);
947 static inline unsigned int compound_order(struct page *page)
951 return page[1].compound_order;
954 static inline bool hpage_pincount_available(struct page *page)
957 * Can the page->hpage_pinned_refcount field be used? That field is in
958 * the 3rd page of the compound page, so the smallest (2-page) compound
959 * pages cannot support it.
961 page = compound_head(page);
962 return PageCompound(page) && compound_order(page) > 1;
965 static inline int head_compound_pincount(struct page *head)
967 return atomic_read(compound_pincount_ptr(head));
970 static inline int compound_pincount(struct page *page)
972 VM_BUG_ON_PAGE(!hpage_pincount_available(page), page);
973 page = compound_head(page);
974 return head_compound_pincount(page);
977 static inline void set_compound_order(struct page *page, unsigned int order)
979 page[1].compound_order = order;
980 page[1].compound_nr = 1U << order;
983 /* Returns the number of pages in this potentially compound page. */
984 static inline unsigned long compound_nr(struct page *page)
988 return page[1].compound_nr;
991 /* Returns the number of bytes in this potentially compound page. */
992 static inline unsigned long page_size(struct page *page)
994 return PAGE_SIZE << compound_order(page);
997 /* Returns the number of bits needed for the number of bytes in a page */
998 static inline unsigned int page_shift(struct page *page)
1000 return PAGE_SHIFT + compound_order(page);
1003 void free_compound_page(struct page *page);
1007 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
1008 * servicing faults for write access. In the normal case, do always want
1009 * pte_mkwrite. But get_user_pages can cause write faults for mappings
1010 * that do not have writing enabled, when used by access_process_vm.
1012 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
1014 if (likely(vma->vm_flags & VM_WRITE))
1015 pte = pte_mkwrite(pte);
1019 vm_fault_t do_set_pmd(struct vm_fault *vmf, struct page *page);
1020 void do_set_pte(struct vm_fault *vmf, struct page *page, unsigned long addr);
1022 vm_fault_t finish_fault(struct vm_fault *vmf);
1023 vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
1027 * Multiple processes may "see" the same page. E.g. for untouched
1028 * mappings of /dev/null, all processes see the same page full of
1029 * zeroes, and text pages of executables and shared libraries have
1030 * only one copy in memory, at most, normally.
1032 * For the non-reserved pages, page_count(page) denotes a reference count.
1033 * page_count() == 0 means the page is free. page->lru is then used for
1034 * freelist management in the buddy allocator.
1035 * page_count() > 0 means the page has been allocated.
1037 * Pages are allocated by the slab allocator in order to provide memory
1038 * to kmalloc and kmem_cache_alloc. In this case, the management of the
1039 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
1040 * unless a particular usage is carefully commented. (the responsibility of
1041 * freeing the kmalloc memory is the caller's, of course).
1043 * A page may be used by anyone else who does a __get_free_page().
1044 * In this case, page_count still tracks the references, and should only
1045 * be used through the normal accessor functions. The top bits of page->flags
1046 * and page->virtual store page management information, but all other fields
1047 * are unused and could be used privately, carefully. The management of this
1048 * page is the responsibility of the one who allocated it, and those who have
1049 * subsequently been given references to it.
1051 * The other pages (we may call them "pagecache pages") are completely
1052 * managed by the Linux memory manager: I/O, buffers, swapping etc.
1053 * The following discussion applies only to them.
1055 * A pagecache page contains an opaque `private' member, which belongs to the
1056 * page's address_space. Usually, this is the address of a circular list of
1057 * the page's disk buffers. PG_private must be set to tell the VM to call
1058 * into the filesystem to release these pages.
1060 * A page may belong to an inode's memory mapping. In this case, page->mapping
1061 * is the pointer to the inode, and page->index is the file offset of the page,
1062 * in units of PAGE_SIZE.
1064 * If pagecache pages are not associated with an inode, they are said to be
1065 * anonymous pages. These may become associated with the swapcache, and in that
1066 * case PG_swapcache is set, and page->private is an offset into the swapcache.
1068 * In either case (swapcache or inode backed), the pagecache itself holds one
1069 * reference to the page. Setting PG_private should also increment the
1070 * refcount. The each user mapping also has a reference to the page.
1072 * The pagecache pages are stored in a per-mapping radix tree, which is
1073 * rooted at mapping->i_pages, and indexed by offset.
1074 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
1075 * lists, we instead now tag pages as dirty/writeback in the radix tree.
1077 * All pagecache pages may be subject to I/O:
1078 * - inode pages may need to be read from disk,
1079 * - inode pages which have been modified and are MAP_SHARED may need
1080 * to be written back to the inode on disk,
1081 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
1082 * modified may need to be swapped out to swap space and (later) to be read
1087 * The zone field is never updated after free_area_init_core()
1088 * sets it, so none of the operations on it need to be atomic.
1091 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
1092 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
1093 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
1094 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
1095 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
1096 #define KASAN_TAG_PGOFF (LAST_CPUPID_PGOFF - KASAN_TAG_WIDTH)
1099 * Define the bit shifts to access each section. For non-existent
1100 * sections we define the shift as 0; that plus a 0 mask ensures
1101 * the compiler will optimise away reference to them.
1103 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
1104 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
1105 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
1106 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
1107 #define KASAN_TAG_PGSHIFT (KASAN_TAG_PGOFF * (KASAN_TAG_WIDTH != 0))
1109 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
1110 #ifdef NODE_NOT_IN_PAGE_FLAGS
1111 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
1112 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
1113 SECTIONS_PGOFF : ZONES_PGOFF)
1115 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
1116 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
1117 NODES_PGOFF : ZONES_PGOFF)
1120 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
1122 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
1123 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
1124 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
1125 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
1126 #define KASAN_TAG_MASK ((1UL << KASAN_TAG_WIDTH) - 1)
1127 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
1129 static inline enum zone_type page_zonenum(const struct page *page)
1131 ASSERT_EXCLUSIVE_BITS(page->flags, ZONES_MASK << ZONES_PGSHIFT);
1132 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
1135 #ifdef CONFIG_ZONE_DEVICE
1136 static inline bool is_zone_device_page(const struct page *page)
1138 return page_zonenum(page) == ZONE_DEVICE;
1140 extern void memmap_init_zone_device(struct zone *, unsigned long,
1141 unsigned long, struct dev_pagemap *);
1143 static inline bool is_zone_device_page(const struct page *page)
1149 static inline bool is_zone_movable_page(const struct page *page)
1151 return page_zonenum(page) == ZONE_MOVABLE;
1154 #ifdef CONFIG_DEV_PAGEMAP_OPS
1155 void free_devmap_managed_page(struct page *page);
1156 DECLARE_STATIC_KEY_FALSE(devmap_managed_key);
1158 static inline bool page_is_devmap_managed(struct page *page)
1160 if (!static_branch_unlikely(&devmap_managed_key))
1162 if (!is_zone_device_page(page))
1164 switch (page->pgmap->type) {
1165 case MEMORY_DEVICE_PRIVATE:
1166 case MEMORY_DEVICE_FS_DAX:
1174 void put_devmap_managed_page(struct page *page);
1176 #else /* CONFIG_DEV_PAGEMAP_OPS */
1177 static inline bool page_is_devmap_managed(struct page *page)
1182 static inline void put_devmap_managed_page(struct page *page)
1185 #endif /* CONFIG_DEV_PAGEMAP_OPS */
1187 static inline bool is_device_private_page(const struct page *page)
1189 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
1190 IS_ENABLED(CONFIG_DEVICE_PRIVATE) &&
1191 is_zone_device_page(page) &&
1192 page->pgmap->type == MEMORY_DEVICE_PRIVATE;
1195 static inline bool is_pci_p2pdma_page(const struct page *page)
1197 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS) &&
1198 IS_ENABLED(CONFIG_PCI_P2PDMA) &&
1199 is_zone_device_page(page) &&
1200 page->pgmap->type == MEMORY_DEVICE_PCI_P2PDMA;
1203 /* 127: arbitrary random number, small enough to assemble well */
1204 #define page_ref_zero_or_close_to_overflow(page) \
1205 ((unsigned int) page_ref_count(page) + 127u <= 127u)
1207 static inline void get_page(struct page *page)
1209 page = compound_head(page);
1211 * Getting a normal page or the head of a compound page
1212 * requires to already have an elevated page->_refcount.
1214 VM_BUG_ON_PAGE(page_ref_zero_or_close_to_overflow(page), page);
1218 bool __must_check try_grab_page(struct page *page, unsigned int flags);
1219 __maybe_unused struct page *try_grab_compound_head(struct page *page, int refs,
1220 unsigned int flags);
1223 static inline __must_check bool try_get_page(struct page *page)
1225 page = compound_head(page);
1226 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1232 static inline void put_page(struct page *page)
1234 page = compound_head(page);
1237 * For devmap managed pages we need to catch refcount transition from
1238 * 2 to 1, when refcount reach one it means the page is free and we
1239 * need to inform the device driver through callback. See
1240 * include/linux/memremap.h and HMM for details.
1242 if (page_is_devmap_managed(page)) {
1243 put_devmap_managed_page(page);
1247 if (put_page_testzero(page))
1252 * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
1253 * the page's refcount so that two separate items are tracked: the original page
1254 * reference count, and also a new count of how many pin_user_pages() calls were
1255 * made against the page. ("gup-pinned" is another term for the latter).
1257 * With this scheme, pin_user_pages() becomes special: such pages are marked as
1258 * distinct from normal pages. As such, the unpin_user_page() call (and its
1259 * variants) must be used in order to release gup-pinned pages.
1263 * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
1264 * counts with respect to pin_user_pages() and unpin_user_page() becomes
1265 * simpler, due to the fact that adding an even power of two to the page
1266 * refcount has the effect of using only the upper N bits, for the code that
1267 * counts up using the bias value. This means that the lower bits are left for
1268 * the exclusive use of the original code that increments and decrements by one
1269 * (or at least, by much smaller values than the bias value).
1271 * Of course, once the lower bits overflow into the upper bits (and this is
1272 * OK, because subtraction recovers the original values), then visual inspection
1273 * no longer suffices to directly view the separate counts. However, for normal
1274 * applications that don't have huge page reference counts, this won't be an
1277 * Locking: the lockless algorithm described in page_cache_get_speculative()
1278 * and page_cache_gup_pin_speculative() provides safe operation for
1279 * get_user_pages and page_mkclean and other calls that race to set up page
1282 #define GUP_PIN_COUNTING_BIAS (1U << 10)
1284 void unpin_user_page(struct page *page);
1285 void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1287 void unpin_user_page_range_dirty_lock(struct page *page, unsigned long npages,
1289 void unpin_user_pages(struct page **pages, unsigned long npages);
1292 * page_maybe_dma_pinned - Report if a page is pinned for DMA.
1295 * This function checks if a page has been pinned via a call to
1296 * a function in the pin_user_pages() family.
1298 * For non-huge pages, the return value is partially fuzzy: false is not fuzzy,
1299 * because it means "definitely not pinned for DMA", but true means "probably
1300 * pinned for DMA, but possibly a false positive due to having at least
1301 * GUP_PIN_COUNTING_BIAS worth of normal page references".
1303 * False positives are OK, because: a) it's unlikely for a page to get that many
1304 * refcounts, and b) all the callers of this routine are expected to be able to
1305 * deal gracefully with a false positive.
1307 * For huge pages, the result will be exactly correct. That's because we have
1308 * more tracking data available: the 3rd struct page in the compound page is
1309 * used to track the pincount (instead using of the GUP_PIN_COUNTING_BIAS
1312 * For more information, please see Documentation/core-api/pin_user_pages.rst.
1314 * Return: True, if it is likely that the page has been "dma-pinned".
1315 * False, if the page is definitely not dma-pinned.
1317 static inline bool page_maybe_dma_pinned(struct page *page)
1319 if (hpage_pincount_available(page))
1320 return compound_pincount(page) > 0;
1323 * page_ref_count() is signed. If that refcount overflows, then
1324 * page_ref_count() returns a negative value, and callers will avoid
1325 * further incrementing the refcount.
1327 * Here, for that overflow case, use the signed bit to count a little
1328 * bit higher via unsigned math, and thus still get an accurate result.
1330 return ((unsigned int)page_ref_count(compound_head(page))) >=
1331 GUP_PIN_COUNTING_BIAS;
1334 static inline bool is_cow_mapping(vm_flags_t flags)
1336 return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
1340 * This should most likely only be called during fork() to see whether we
1341 * should break the cow immediately for a page on the src mm.
1343 static inline bool page_needs_cow_for_dma(struct vm_area_struct *vma,
1346 if (!is_cow_mapping(vma->vm_flags))
1349 if (!test_bit(MMF_HAS_PINNED, &vma->vm_mm->flags))
1352 return page_maybe_dma_pinned(page);
1355 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1356 #define SECTION_IN_PAGE_FLAGS
1360 * The identification function is mainly used by the buddy allocator for
1361 * determining if two pages could be buddies. We are not really identifying
1362 * the zone since we could be using the section number id if we do not have
1363 * node id available in page flags.
1364 * We only guarantee that it will return the same value for two combinable
1367 static inline int page_zone_id(struct page *page)
1369 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1372 #ifdef NODE_NOT_IN_PAGE_FLAGS
1373 extern int page_to_nid(const struct page *page);
1375 static inline int page_to_nid(const struct page *page)
1377 struct page *p = (struct page *)page;
1379 return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1383 #ifdef CONFIG_NUMA_BALANCING
1384 static inline int cpu_pid_to_cpupid(int cpu, int pid)
1386 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1389 static inline int cpupid_to_pid(int cpupid)
1391 return cpupid & LAST__PID_MASK;
1394 static inline int cpupid_to_cpu(int cpupid)
1396 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1399 static inline int cpupid_to_nid(int cpupid)
1401 return cpu_to_node(cpupid_to_cpu(cpupid));
1404 static inline bool cpupid_pid_unset(int cpupid)
1406 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1409 static inline bool cpupid_cpu_unset(int cpupid)
1411 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1414 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1416 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1419 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1420 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1421 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1423 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1426 static inline int page_cpupid_last(struct page *page)
1428 return page->_last_cpupid;
1430 static inline void page_cpupid_reset_last(struct page *page)
1432 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1435 static inline int page_cpupid_last(struct page *page)
1437 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1440 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1442 static inline void page_cpupid_reset_last(struct page *page)
1444 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1446 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1447 #else /* !CONFIG_NUMA_BALANCING */
1448 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
1450 return page_to_nid(page); /* XXX */
1453 static inline int page_cpupid_last(struct page *page)
1455 return page_to_nid(page); /* XXX */
1458 static inline int cpupid_to_nid(int cpupid)
1463 static inline int cpupid_to_pid(int cpupid)
1468 static inline int cpupid_to_cpu(int cpupid)
1473 static inline int cpu_pid_to_cpupid(int nid, int pid)
1478 static inline bool cpupid_pid_unset(int cpupid)
1483 static inline void page_cpupid_reset_last(struct page *page)
1487 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1491 #endif /* CONFIG_NUMA_BALANCING */
1493 #if defined(CONFIG_KASAN_SW_TAGS) || defined(CONFIG_KASAN_HW_TAGS)
1496 * KASAN per-page tags are stored xor'ed with 0xff. This allows to avoid
1497 * setting tags for all pages to native kernel tag value 0xff, as the default
1498 * value 0x00 maps to 0xff.
1501 static inline u8 page_kasan_tag(const struct page *page)
1505 if (kasan_enabled()) {
1506 tag = (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1513 static inline void page_kasan_tag_set(struct page *page, u8 tag)
1515 if (kasan_enabled()) {
1517 page->flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1518 page->flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1522 static inline void page_kasan_tag_reset(struct page *page)
1524 if (kasan_enabled())
1525 page_kasan_tag_set(page, 0xff);
1528 #else /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1530 static inline u8 page_kasan_tag(const struct page *page)
1535 static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1536 static inline void page_kasan_tag_reset(struct page *page) { }
1538 #endif /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1540 static inline struct zone *page_zone(const struct page *page)
1542 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1545 static inline pg_data_t *page_pgdat(const struct page *page)
1547 return NODE_DATA(page_to_nid(page));
1550 #ifdef SECTION_IN_PAGE_FLAGS
1551 static inline void set_page_section(struct page *page, unsigned long section)
1553 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1554 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1557 static inline unsigned long page_to_section(const struct page *page)
1559 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1563 /* MIGRATE_CMA and ZONE_MOVABLE do not allow pin pages */
1564 #ifdef CONFIG_MIGRATION
1565 static inline bool is_pinnable_page(struct page *page)
1567 return !(is_zone_movable_page(page) || is_migrate_cma_page(page)) ||
1568 is_zero_pfn(page_to_pfn(page));
1571 static inline bool is_pinnable_page(struct page *page)
1577 static inline void set_page_zone(struct page *page, enum zone_type zone)
1579 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1580 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1583 static inline void set_page_node(struct page *page, unsigned long node)
1585 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1586 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1589 static inline void set_page_links(struct page *page, enum zone_type zone,
1590 unsigned long node, unsigned long pfn)
1592 set_page_zone(page, zone);
1593 set_page_node(page, node);
1594 #ifdef SECTION_IN_PAGE_FLAGS
1595 set_page_section(page, pfn_to_section_nr(pfn));
1600 * Some inline functions in vmstat.h depend on page_zone()
1602 #include <linux/vmstat.h>
1604 static __always_inline void *lowmem_page_address(const struct page *page)
1606 return page_to_virt(page);
1609 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1610 #define HASHED_PAGE_VIRTUAL
1613 #if defined(WANT_PAGE_VIRTUAL)
1614 static inline void *page_address(const struct page *page)
1616 return page->virtual;
1618 static inline void set_page_address(struct page *page, void *address)
1620 page->virtual = address;
1622 #define page_address_init() do { } while(0)
1625 #if defined(HASHED_PAGE_VIRTUAL)
1626 void *page_address(const struct page *page);
1627 void set_page_address(struct page *page, void *virtual);
1628 void page_address_init(void);
1631 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1632 #define page_address(page) lowmem_page_address(page)
1633 #define set_page_address(page, address) do { } while(0)
1634 #define page_address_init() do { } while(0)
1637 extern void *page_rmapping(struct page *page);
1638 extern struct anon_vma *page_anon_vma(struct page *page);
1639 extern struct address_space *page_mapping(struct page *page);
1641 extern struct address_space *__page_file_mapping(struct page *);
1644 struct address_space *page_file_mapping(struct page *page)
1646 if (unlikely(PageSwapCache(page)))
1647 return __page_file_mapping(page);
1649 return page->mapping;
1652 extern pgoff_t __page_file_index(struct page *page);
1655 * Return the pagecache index of the passed page. Regular pagecache pages
1656 * use ->index whereas swapcache pages use swp_offset(->private)
1658 static inline pgoff_t page_index(struct page *page)
1660 if (unlikely(PageSwapCache(page)))
1661 return __page_file_index(page);
1665 bool page_mapped(struct page *page);
1666 struct address_space *page_mapping(struct page *page);
1669 * Return true only if the page has been allocated with
1670 * ALLOC_NO_WATERMARKS and the low watermark was not
1671 * met implying that the system is under some pressure.
1673 static inline bool page_is_pfmemalloc(const struct page *page)
1676 * lru.next has bit 1 set if the page is allocated from the
1677 * pfmemalloc reserves. Callers may simply overwrite it if
1678 * they do not need to preserve that information.
1680 return (uintptr_t)page->lru.next & BIT(1);
1684 * Only to be called by the page allocator on a freshly allocated
1687 static inline void set_page_pfmemalloc(struct page *page)
1689 page->lru.next = (void *)BIT(1);
1692 static inline void clear_page_pfmemalloc(struct page *page)
1694 page->lru.next = NULL;
1698 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1700 extern void pagefault_out_of_memory(void);
1702 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1703 #define offset_in_thp(page, p) ((unsigned long)(p) & (thp_size(page) - 1))
1706 * Flags passed to show_mem() and show_free_areas() to suppress output in
1709 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1711 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1714 extern bool can_do_mlock(void);
1716 static inline bool can_do_mlock(void) { return false; }
1718 extern int user_shm_lock(size_t, struct ucounts *);
1719 extern void user_shm_unlock(size_t, struct ucounts *);
1722 * Parameter block passed down to zap_pte_range in exceptional cases.
1724 struct zap_details {
1725 struct address_space *check_mapping; /* Check page->mapping if set */
1726 pgoff_t first_index; /* Lowest page->index to unmap */
1727 pgoff_t last_index; /* Highest page->index to unmap */
1728 struct page *single_page; /* Locked page to be unmapped */
1731 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1733 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1736 void zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1737 unsigned long size);
1738 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1739 unsigned long size);
1740 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1741 unsigned long start, unsigned long end);
1743 struct mmu_notifier_range;
1745 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1746 unsigned long end, unsigned long floor, unsigned long ceiling);
1748 copy_page_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma);
1749 int follow_invalidate_pte(struct mm_struct *mm, unsigned long address,
1750 struct mmu_notifier_range *range, pte_t **ptepp,
1751 pmd_t **pmdpp, spinlock_t **ptlp);
1752 int follow_pte(struct mm_struct *mm, unsigned long address,
1753 pte_t **ptepp, spinlock_t **ptlp);
1754 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1755 unsigned long *pfn);
1756 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1757 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1758 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1759 void *buf, int len, int write);
1761 extern void truncate_pagecache(struct inode *inode, loff_t new);
1762 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1763 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1764 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1765 int truncate_inode_page(struct address_space *mapping, struct page *page);
1766 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1767 int invalidate_inode_page(struct page *page);
1770 extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1771 unsigned long address, unsigned int flags,
1772 struct pt_regs *regs);
1773 extern int fixup_user_fault(struct mm_struct *mm,
1774 unsigned long address, unsigned int fault_flags,
1776 void unmap_mapping_page(struct page *page);
1777 void unmap_mapping_pages(struct address_space *mapping,
1778 pgoff_t start, pgoff_t nr, bool even_cows);
1779 void unmap_mapping_range(struct address_space *mapping,
1780 loff_t const holebegin, loff_t const holelen, int even_cows);
1782 static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1783 unsigned long address, unsigned int flags,
1784 struct pt_regs *regs)
1786 /* should never happen if there's no MMU */
1788 return VM_FAULT_SIGBUS;
1790 static inline int fixup_user_fault(struct mm_struct *mm, unsigned long address,
1791 unsigned int fault_flags, bool *unlocked)
1793 /* should never happen if there's no MMU */
1797 static inline void unmap_mapping_page(struct page *page) { }
1798 static inline void unmap_mapping_pages(struct address_space *mapping,
1799 pgoff_t start, pgoff_t nr, bool even_cows) { }
1800 static inline void unmap_mapping_range(struct address_space *mapping,
1801 loff_t const holebegin, loff_t const holelen, int even_cows) { }
1804 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1805 loff_t const holebegin, loff_t const holelen)
1807 unmap_mapping_range(mapping, holebegin, holelen, 0);
1810 extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1811 void *buf, int len, unsigned int gup_flags);
1812 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1813 void *buf, int len, unsigned int gup_flags);
1814 extern int __access_remote_vm(struct mm_struct *mm, unsigned long addr,
1815 void *buf, int len, unsigned int gup_flags);
1817 long get_user_pages_remote(struct mm_struct *mm,
1818 unsigned long start, unsigned long nr_pages,
1819 unsigned int gup_flags, struct page **pages,
1820 struct vm_area_struct **vmas, int *locked);
1821 long pin_user_pages_remote(struct mm_struct *mm,
1822 unsigned long start, unsigned long nr_pages,
1823 unsigned int gup_flags, struct page **pages,
1824 struct vm_area_struct **vmas, int *locked);
1825 long get_user_pages(unsigned long start, unsigned long nr_pages,
1826 unsigned int gup_flags, struct page **pages,
1827 struct vm_area_struct **vmas);
1828 long pin_user_pages(unsigned long start, unsigned long nr_pages,
1829 unsigned int gup_flags, struct page **pages,
1830 struct vm_area_struct **vmas);
1831 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1832 unsigned int gup_flags, struct page **pages, int *locked);
1833 long pin_user_pages_locked(unsigned long start, unsigned long nr_pages,
1834 unsigned int gup_flags, struct page **pages, int *locked);
1835 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1836 struct page **pages, unsigned int gup_flags);
1837 long pin_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1838 struct page **pages, unsigned int gup_flags);
1840 int get_user_pages_fast(unsigned long start, int nr_pages,
1841 unsigned int gup_flags, struct page **pages);
1842 int pin_user_pages_fast(unsigned long start, int nr_pages,
1843 unsigned int gup_flags, struct page **pages);
1845 int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
1846 int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
1847 struct task_struct *task, bool bypass_rlim);
1850 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1851 struct page **pages);
1852 int get_kernel_page(unsigned long start, int write, struct page **pages);
1853 struct page *get_dump_page(unsigned long addr);
1855 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1856 extern void do_invalidatepage(struct page *page, unsigned int offset,
1857 unsigned int length);
1859 int redirty_page_for_writepage(struct writeback_control *wbc,
1861 void account_page_cleaned(struct page *page, struct address_space *mapping,
1862 struct bdi_writeback *wb);
1863 int set_page_dirty(struct page *page);
1864 int set_page_dirty_lock(struct page *page);
1865 void __cancel_dirty_page(struct page *page);
1866 static inline void cancel_dirty_page(struct page *page)
1868 /* Avoid atomic ops, locking, etc. when not actually needed. */
1869 if (PageDirty(page))
1870 __cancel_dirty_page(page);
1872 int clear_page_dirty_for_io(struct page *page);
1874 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1876 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1877 unsigned long old_addr, struct vm_area_struct *new_vma,
1878 unsigned long new_addr, unsigned long len,
1879 bool need_rmap_locks);
1882 * Flags used by change_protection(). For now we make it a bitmap so
1883 * that we can pass in multiple flags just like parameters. However
1884 * for now all the callers are only use one of the flags at the same
1887 /* Whether we should allow dirty bit accounting */
1888 #define MM_CP_DIRTY_ACCT (1UL << 0)
1889 /* Whether this protection change is for NUMA hints */
1890 #define MM_CP_PROT_NUMA (1UL << 1)
1891 /* Whether this change is for write protecting */
1892 #define MM_CP_UFFD_WP (1UL << 2) /* do wp */
1893 #define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */
1894 #define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \
1895 MM_CP_UFFD_WP_RESOLVE)
1897 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1898 unsigned long end, pgprot_t newprot,
1899 unsigned long cp_flags);
1900 extern int mprotect_fixup(struct vm_area_struct *vma,
1901 struct vm_area_struct **pprev, unsigned long start,
1902 unsigned long end, unsigned long newflags);
1905 * doesn't attempt to fault and will return short.
1907 int get_user_pages_fast_only(unsigned long start, int nr_pages,
1908 unsigned int gup_flags, struct page **pages);
1909 int pin_user_pages_fast_only(unsigned long start, int nr_pages,
1910 unsigned int gup_flags, struct page **pages);
1912 static inline bool get_user_page_fast_only(unsigned long addr,
1913 unsigned int gup_flags, struct page **pagep)
1915 return get_user_pages_fast_only(addr, 1, gup_flags, pagep) == 1;
1918 * per-process(per-mm_struct) statistics.
1920 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1922 long val = atomic_long_read(&mm->rss_stat.count[member]);
1924 #ifdef SPLIT_RSS_COUNTING
1926 * counter is updated in asynchronous manner and may go to minus.
1927 * But it's never be expected number for users.
1932 return (unsigned long)val;
1935 void mm_trace_rss_stat(struct mm_struct *mm, int member, long count);
1937 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1939 long count = atomic_long_add_return(value, &mm->rss_stat.count[member]);
1941 mm_trace_rss_stat(mm, member, count);
1944 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1946 long count = atomic_long_inc_return(&mm->rss_stat.count[member]);
1948 mm_trace_rss_stat(mm, member, count);
1951 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1953 long count = atomic_long_dec_return(&mm->rss_stat.count[member]);
1955 mm_trace_rss_stat(mm, member, count);
1958 /* Optimized variant when page is already known not to be PageAnon */
1959 static inline int mm_counter_file(struct page *page)
1961 if (PageSwapBacked(page))
1962 return MM_SHMEMPAGES;
1963 return MM_FILEPAGES;
1966 static inline int mm_counter(struct page *page)
1969 return MM_ANONPAGES;
1970 return mm_counter_file(page);
1973 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1975 return get_mm_counter(mm, MM_FILEPAGES) +
1976 get_mm_counter(mm, MM_ANONPAGES) +
1977 get_mm_counter(mm, MM_SHMEMPAGES);
1980 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1982 return max(mm->hiwater_rss, get_mm_rss(mm));
1985 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1987 return max(mm->hiwater_vm, mm->total_vm);
1990 static inline void update_hiwater_rss(struct mm_struct *mm)
1992 unsigned long _rss = get_mm_rss(mm);
1994 if ((mm)->hiwater_rss < _rss)
1995 (mm)->hiwater_rss = _rss;
1998 static inline void update_hiwater_vm(struct mm_struct *mm)
2000 if (mm->hiwater_vm < mm->total_vm)
2001 mm->hiwater_vm = mm->total_vm;
2004 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
2006 mm->hiwater_rss = get_mm_rss(mm);
2009 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
2010 struct mm_struct *mm)
2012 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
2014 if (*maxrss < hiwater_rss)
2015 *maxrss = hiwater_rss;
2018 #if defined(SPLIT_RSS_COUNTING)
2019 void sync_mm_rss(struct mm_struct *mm);
2021 static inline void sync_mm_rss(struct mm_struct *mm)
2026 #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
2027 static inline int pte_special(pte_t pte)
2032 static inline pte_t pte_mkspecial(pte_t pte)
2038 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
2039 static inline int pte_devmap(pte_t pte)
2045 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
2047 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
2049 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
2053 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
2057 #ifdef __PAGETABLE_P4D_FOLDED
2058 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2059 unsigned long address)
2064 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
2067 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
2068 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2069 unsigned long address)
2073 static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
2074 static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
2077 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
2079 static inline void mm_inc_nr_puds(struct mm_struct *mm)
2081 if (mm_pud_folded(mm))
2083 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2086 static inline void mm_dec_nr_puds(struct mm_struct *mm)
2088 if (mm_pud_folded(mm))
2090 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
2094 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
2095 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
2096 unsigned long address)
2101 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
2102 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
2105 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
2107 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
2109 if (mm_pmd_folded(mm))
2111 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2114 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
2116 if (mm_pmd_folded(mm))
2118 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
2123 static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
2125 atomic_long_set(&mm->pgtables_bytes, 0);
2128 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2130 return atomic_long_read(&mm->pgtables_bytes);
2133 static inline void mm_inc_nr_ptes(struct mm_struct *mm)
2135 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2138 static inline void mm_dec_nr_ptes(struct mm_struct *mm)
2140 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
2144 static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
2145 static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
2150 static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
2151 static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
2154 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
2155 int __pte_alloc_kernel(pmd_t *pmd);
2157 #if defined(CONFIG_MMU)
2159 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
2160 unsigned long address)
2162 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
2163 NULL : p4d_offset(pgd, address);
2166 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
2167 unsigned long address)
2169 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
2170 NULL : pud_offset(p4d, address);
2173 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
2175 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
2176 NULL: pmd_offset(pud, address);
2178 #endif /* CONFIG_MMU */
2180 #if USE_SPLIT_PTE_PTLOCKS
2181 #if ALLOC_SPLIT_PTLOCKS
2182 void __init ptlock_cache_init(void);
2183 extern bool ptlock_alloc(struct page *page);
2184 extern void ptlock_free(struct page *page);
2186 static inline spinlock_t *ptlock_ptr(struct page *page)
2190 #else /* ALLOC_SPLIT_PTLOCKS */
2191 static inline void ptlock_cache_init(void)
2195 static inline bool ptlock_alloc(struct page *page)
2200 static inline void ptlock_free(struct page *page)
2204 static inline spinlock_t *ptlock_ptr(struct page *page)
2208 #endif /* ALLOC_SPLIT_PTLOCKS */
2210 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2212 return ptlock_ptr(pmd_page(*pmd));
2215 static inline bool ptlock_init(struct page *page)
2218 * prep_new_page() initialize page->private (and therefore page->ptl)
2219 * with 0. Make sure nobody took it in use in between.
2221 * It can happen if arch try to use slab for page table allocation:
2222 * slab code uses page->slab_cache, which share storage with page->ptl.
2224 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
2225 if (!ptlock_alloc(page))
2227 spin_lock_init(ptlock_ptr(page));
2231 #else /* !USE_SPLIT_PTE_PTLOCKS */
2233 * We use mm->page_table_lock to guard all pagetable pages of the mm.
2235 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
2237 return &mm->page_table_lock;
2239 static inline void ptlock_cache_init(void) {}
2240 static inline bool ptlock_init(struct page *page) { return true; }
2241 static inline void ptlock_free(struct page *page) {}
2242 #endif /* USE_SPLIT_PTE_PTLOCKS */
2244 static inline void pgtable_init(void)
2246 ptlock_cache_init();
2247 pgtable_cache_init();
2250 static inline bool pgtable_pte_page_ctor(struct page *page)
2252 if (!ptlock_init(page))
2254 __SetPageTable(page);
2255 inc_lruvec_page_state(page, NR_PAGETABLE);
2259 static inline void pgtable_pte_page_dtor(struct page *page)
2262 __ClearPageTable(page);
2263 dec_lruvec_page_state(page, NR_PAGETABLE);
2266 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
2268 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
2269 pte_t *__pte = pte_offset_map(pmd, address); \
2275 #define pte_unmap_unlock(pte, ptl) do { \
2280 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2282 #define pte_alloc_map(mm, pmd, address) \
2283 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2285 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2286 (pte_alloc(mm, pmd) ? \
2287 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2289 #define pte_alloc_kernel(pmd, address) \
2290 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2291 NULL: pte_offset_kernel(pmd, address))
2293 #if USE_SPLIT_PMD_PTLOCKS
2295 static struct page *pmd_to_page(pmd_t *pmd)
2297 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2298 return virt_to_page((void *)((unsigned long) pmd & mask));
2301 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2303 return ptlock_ptr(pmd_to_page(pmd));
2306 static inline bool pmd_ptlock_init(struct page *page)
2308 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2309 page->pmd_huge_pte = NULL;
2311 return ptlock_init(page);
2314 static inline void pmd_ptlock_free(struct page *page)
2316 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2317 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2322 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2326 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2328 return &mm->page_table_lock;
2331 static inline bool pmd_ptlock_init(struct page *page) { return true; }
2332 static inline void pmd_ptlock_free(struct page *page) {}
2334 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2338 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2340 spinlock_t *ptl = pmd_lockptr(mm, pmd);
2345 static inline bool pgtable_pmd_page_ctor(struct page *page)
2347 if (!pmd_ptlock_init(page))
2349 __SetPageTable(page);
2350 inc_lruvec_page_state(page, NR_PAGETABLE);
2354 static inline void pgtable_pmd_page_dtor(struct page *page)
2356 pmd_ptlock_free(page);
2357 __ClearPageTable(page);
2358 dec_lruvec_page_state(page, NR_PAGETABLE);
2362 * No scalability reason to split PUD locks yet, but follow the same pattern
2363 * as the PMD locks to make it easier if we decide to. The VM should not be
2364 * considered ready to switch to split PUD locks yet; there may be places
2365 * which need to be converted from page_table_lock.
2367 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2369 return &mm->page_table_lock;
2372 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2374 spinlock_t *ptl = pud_lockptr(mm, pud);
2380 extern void __init pagecache_init(void);
2381 extern void __init free_area_init_memoryless_node(int nid);
2382 extern void free_initmem(void);
2385 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2386 * into the buddy system. The freed pages will be poisoned with pattern
2387 * "poison" if it's within range [0, UCHAR_MAX].
2388 * Return pages freed into the buddy system.
2390 extern unsigned long free_reserved_area(void *start, void *end,
2391 int poison, const char *s);
2393 extern void adjust_managed_page_count(struct page *page, long count);
2394 extern void mem_init_print_info(void);
2396 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2398 /* Free the reserved page into the buddy system, so it gets managed. */
2399 static inline void free_reserved_page(struct page *page)
2401 ClearPageReserved(page);
2402 init_page_count(page);
2404 adjust_managed_page_count(page, 1);
2406 #define free_highmem_page(page) free_reserved_page(page)
2408 static inline void mark_page_reserved(struct page *page)
2410 SetPageReserved(page);
2411 adjust_managed_page_count(page, -1);
2415 * Default method to free all the __init memory into the buddy system.
2416 * The freed pages will be poisoned with pattern "poison" if it's within
2417 * range [0, UCHAR_MAX].
2418 * Return pages freed into the buddy system.
2420 static inline unsigned long free_initmem_default(int poison)
2422 extern char __init_begin[], __init_end[];
2424 return free_reserved_area(&__init_begin, &__init_end,
2425 poison, "unused kernel image (initmem)");
2428 static inline unsigned long get_num_physpages(void)
2431 unsigned long phys_pages = 0;
2433 for_each_online_node(nid)
2434 phys_pages += node_present_pages(nid);
2440 * Using memblock node mappings, an architecture may initialise its
2441 * zones, allocate the backing mem_map and account for memory holes in an
2442 * architecture independent manner.
2444 * An architecture is expected to register range of page frames backed by
2445 * physical memory with memblock_add[_node]() before calling
2446 * free_area_init() passing in the PFN each zone ends at. At a basic
2447 * usage, an architecture is expected to do something like
2449 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2451 * for_each_valid_physical_page_range()
2452 * memblock_add_node(base, size, nid)
2453 * free_area_init(max_zone_pfns);
2455 void free_area_init(unsigned long *max_zone_pfn);
2456 unsigned long node_map_pfn_alignment(void);
2457 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2458 unsigned long end_pfn);
2459 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2460 unsigned long end_pfn);
2461 extern void get_pfn_range_for_nid(unsigned int nid,
2462 unsigned long *start_pfn, unsigned long *end_pfn);
2463 extern unsigned long find_min_pfn_with_active_regions(void);
2466 static inline int early_pfn_to_nid(unsigned long pfn)
2471 /* please see mm/page_alloc.c */
2472 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2475 extern void set_dma_reserve(unsigned long new_dma_reserve);
2476 extern void memmap_init_range(unsigned long, int, unsigned long,
2477 unsigned long, unsigned long, enum meminit_context,
2478 struct vmem_altmap *, int migratetype);
2479 extern void setup_per_zone_wmarks(void);
2480 extern int __meminit init_per_zone_wmark_min(void);
2481 extern void mem_init(void);
2482 extern void __init mmap_init(void);
2483 extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2484 extern long si_mem_available(void);
2485 extern void si_meminfo(struct sysinfo * val);
2486 extern void si_meminfo_node(struct sysinfo *val, int nid);
2487 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2488 extern unsigned long arch_reserved_kernel_pages(void);
2491 extern __printf(3, 4)
2492 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2494 extern void setup_per_cpu_pageset(void);
2497 extern int min_free_kbytes;
2498 extern int watermark_boost_factor;
2499 extern int watermark_scale_factor;
2500 extern bool arch_has_descending_max_zone_pfns(void);
2503 extern atomic_long_t mmap_pages_allocated;
2504 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2506 /* interval_tree.c */
2507 void vma_interval_tree_insert(struct vm_area_struct *node,
2508 struct rb_root_cached *root);
2509 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2510 struct vm_area_struct *prev,
2511 struct rb_root_cached *root);
2512 void vma_interval_tree_remove(struct vm_area_struct *node,
2513 struct rb_root_cached *root);
2514 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2515 unsigned long start, unsigned long last);
2516 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2517 unsigned long start, unsigned long last);
2519 #define vma_interval_tree_foreach(vma, root, start, last) \
2520 for (vma = vma_interval_tree_iter_first(root, start, last); \
2521 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2523 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2524 struct rb_root_cached *root);
2525 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2526 struct rb_root_cached *root);
2527 struct anon_vma_chain *
2528 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2529 unsigned long start, unsigned long last);
2530 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2531 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2532 #ifdef CONFIG_DEBUG_VM_RB
2533 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2536 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2537 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2538 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2541 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2542 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2543 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2544 struct vm_area_struct *expand);
2545 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2546 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2548 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2550 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2551 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2552 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2553 struct mempolicy *, struct vm_userfaultfd_ctx);
2554 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2555 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2556 unsigned long addr, int new_below);
2557 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2558 unsigned long addr, int new_below);
2559 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2560 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2561 struct rb_node **, struct rb_node *);
2562 extern void unlink_file_vma(struct vm_area_struct *);
2563 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2564 unsigned long addr, unsigned long len, pgoff_t pgoff,
2565 bool *need_rmap_locks);
2566 extern void exit_mmap(struct mm_struct *);
2568 static inline int check_data_rlimit(unsigned long rlim,
2570 unsigned long start,
2571 unsigned long end_data,
2572 unsigned long start_data)
2574 if (rlim < RLIM_INFINITY) {
2575 if (((new - start) + (end_data - start_data)) > rlim)
2582 extern int mm_take_all_locks(struct mm_struct *mm);
2583 extern void mm_drop_all_locks(struct mm_struct *mm);
2585 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2586 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2587 extern struct file *get_task_exe_file(struct task_struct *task);
2589 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2590 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2592 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2593 const struct vm_special_mapping *sm);
2594 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2595 unsigned long addr, unsigned long len,
2596 unsigned long flags,
2597 const struct vm_special_mapping *spec);
2598 /* This is an obsolete alternative to _install_special_mapping. */
2599 extern int install_special_mapping(struct mm_struct *mm,
2600 unsigned long addr, unsigned long len,
2601 unsigned long flags, struct page **pages);
2603 unsigned long randomize_stack_top(unsigned long stack_top);
2605 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2607 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2608 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2609 struct list_head *uf);
2610 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2611 unsigned long len, unsigned long prot, unsigned long flags,
2612 unsigned long pgoff, unsigned long *populate, struct list_head *uf);
2613 extern int __do_munmap(struct mm_struct *, unsigned long, size_t,
2614 struct list_head *uf, bool downgrade);
2615 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2616 struct list_head *uf);
2617 extern int do_madvise(struct mm_struct *mm, unsigned long start, size_t len_in, int behavior);
2620 extern int __mm_populate(unsigned long addr, unsigned long len,
2622 static inline void mm_populate(unsigned long addr, unsigned long len)
2625 (void) __mm_populate(addr, len, 1);
2628 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2631 /* These take the mm semaphore themselves */
2632 extern int __must_check vm_brk(unsigned long, unsigned long);
2633 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2634 extern int vm_munmap(unsigned long, size_t);
2635 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2636 unsigned long, unsigned long,
2637 unsigned long, unsigned long);
2639 struct vm_unmapped_area_info {
2640 #define VM_UNMAPPED_AREA_TOPDOWN 1
2641 unsigned long flags;
2642 unsigned long length;
2643 unsigned long low_limit;
2644 unsigned long high_limit;
2645 unsigned long align_mask;
2646 unsigned long align_offset;
2649 extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info);
2652 extern void truncate_inode_pages(struct address_space *, loff_t);
2653 extern void truncate_inode_pages_range(struct address_space *,
2654 loff_t lstart, loff_t lend);
2655 extern void truncate_inode_pages_final(struct address_space *);
2657 /* generic vm_area_ops exported for stackable file systems */
2658 extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2659 extern vm_fault_t filemap_map_pages(struct vm_fault *vmf,
2660 pgoff_t start_pgoff, pgoff_t end_pgoff);
2661 extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2663 /* mm/page-writeback.c */
2664 int __must_check write_one_page(struct page *page);
2665 void task_dirty_inc(struct task_struct *tsk);
2667 extern unsigned long stack_guard_gap;
2668 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2669 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2671 /* CONFIG_STACK_GROWSUP still needs to grow downwards at some places */
2672 extern int expand_downwards(struct vm_area_struct *vma,
2673 unsigned long address);
2675 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2677 #define expand_upwards(vma, address) (0)
2680 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2681 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2682 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2683 struct vm_area_struct **pprev);
2686 * find_vma_intersection() - Look up the first VMA which intersects the interval
2687 * @mm: The process address space.
2688 * @start_addr: The inclusive start user address.
2689 * @end_addr: The exclusive end user address.
2691 * Returns: The first VMA within the provided range, %NULL otherwise. Assumes
2692 * start_addr < end_addr.
2695 struct vm_area_struct *find_vma_intersection(struct mm_struct *mm,
2696 unsigned long start_addr,
2697 unsigned long end_addr)
2699 struct vm_area_struct *vma = find_vma(mm, start_addr);
2701 if (vma && end_addr <= vma->vm_start)
2707 * vma_lookup() - Find a VMA at a specific address
2708 * @mm: The process address space.
2709 * @addr: The user address.
2711 * Return: The vm_area_struct at the given address, %NULL otherwise.
2714 struct vm_area_struct *vma_lookup(struct mm_struct *mm, unsigned long addr)
2716 struct vm_area_struct *vma = find_vma(mm, addr);
2718 if (vma && addr < vma->vm_start)
2724 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2726 unsigned long vm_start = vma->vm_start;
2728 if (vma->vm_flags & VM_GROWSDOWN) {
2729 vm_start -= stack_guard_gap;
2730 if (vm_start > vma->vm_start)
2736 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2738 unsigned long vm_end = vma->vm_end;
2740 if (vma->vm_flags & VM_GROWSUP) {
2741 vm_end += stack_guard_gap;
2742 if (vm_end < vma->vm_end)
2743 vm_end = -PAGE_SIZE;
2748 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2750 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2753 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2754 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2755 unsigned long vm_start, unsigned long vm_end)
2757 struct vm_area_struct *vma = find_vma(mm, vm_start);
2759 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2765 static inline bool range_in_vma(struct vm_area_struct *vma,
2766 unsigned long start, unsigned long end)
2768 return (vma && vma->vm_start <= start && end <= vma->vm_end);
2772 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2773 void vma_set_page_prot(struct vm_area_struct *vma);
2775 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2779 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2781 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2785 void vma_set_file(struct vm_area_struct *vma, struct file *file);
2787 #ifdef CONFIG_NUMA_BALANCING
2788 unsigned long change_prot_numa(struct vm_area_struct *vma,
2789 unsigned long start, unsigned long end);
2792 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2793 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2794 unsigned long pfn, unsigned long size, pgprot_t);
2795 int remap_pfn_range_notrack(struct vm_area_struct *vma, unsigned long addr,
2796 unsigned long pfn, unsigned long size, pgprot_t prot);
2797 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2798 int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
2799 struct page **pages, unsigned long *num);
2800 int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
2802 int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
2804 vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2806 vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2807 unsigned long pfn, pgprot_t pgprot);
2808 vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2810 vm_fault_t vmf_insert_mixed_prot(struct vm_area_struct *vma, unsigned long addr,
2811 pfn_t pfn, pgprot_t pgprot);
2812 vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
2813 unsigned long addr, pfn_t pfn);
2814 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2816 static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2817 unsigned long addr, struct page *page)
2819 int err = vm_insert_page(vma, addr, page);
2822 return VM_FAULT_OOM;
2823 if (err < 0 && err != -EBUSY)
2824 return VM_FAULT_SIGBUS;
2826 return VM_FAULT_NOPAGE;
2829 #ifndef io_remap_pfn_range
2830 static inline int io_remap_pfn_range(struct vm_area_struct *vma,
2831 unsigned long addr, unsigned long pfn,
2832 unsigned long size, pgprot_t prot)
2834 return remap_pfn_range(vma, addr, pfn, size, pgprot_decrypted(prot));
2838 static inline vm_fault_t vmf_error(int err)
2841 return VM_FAULT_OOM;
2842 return VM_FAULT_SIGBUS;
2845 struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
2846 unsigned int foll_flags);
2848 #define FOLL_WRITE 0x01 /* check pte is writable */
2849 #define FOLL_TOUCH 0x02 /* mark page accessed */
2850 #define FOLL_GET 0x04 /* do get_page on page */
2851 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2852 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2853 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2854 * and return without waiting upon it */
2855 #define FOLL_POPULATE 0x40 /* fault in page */
2856 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2857 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2858 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2859 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2860 #define FOLL_MLOCK 0x1000 /* lock present pages */
2861 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2862 #define FOLL_COW 0x4000 /* internal GUP flag */
2863 #define FOLL_ANON 0x8000 /* don't do file mappings */
2864 #define FOLL_LONGTERM 0x10000 /* mapping lifetime is indefinite: see below */
2865 #define FOLL_SPLIT_PMD 0x20000 /* split huge pmd before returning */
2866 #define FOLL_PIN 0x40000 /* pages must be released via unpin_user_page */
2867 #define FOLL_FAST_ONLY 0x80000 /* gup_fast: prevent fall-back to slow gup */
2870 * FOLL_PIN and FOLL_LONGTERM may be used in various combinations with each
2871 * other. Here is what they mean, and how to use them:
2873 * FOLL_LONGTERM indicates that the page will be held for an indefinite time
2874 * period _often_ under userspace control. This is in contrast to
2875 * iov_iter_get_pages(), whose usages are transient.
2877 * FIXME: For pages which are part of a filesystem, mappings are subject to the
2878 * lifetime enforced by the filesystem and we need guarantees that longterm
2879 * users like RDMA and V4L2 only establish mappings which coordinate usage with
2880 * the filesystem. Ideas for this coordination include revoking the longterm
2881 * pin, delaying writeback, bounce buffer page writeback, etc. As FS DAX was
2882 * added after the problem with filesystems was found FS DAX VMAs are
2883 * specifically failed. Filesystem pages are still subject to bugs and use of
2884 * FOLL_LONGTERM should be avoided on those pages.
2886 * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call.
2887 * Currently only get_user_pages() and get_user_pages_fast() support this flag
2888 * and calls to get_user_pages_[un]locked are specifically not allowed. This
2889 * is due to an incompatibility with the FS DAX check and
2890 * FAULT_FLAG_ALLOW_RETRY.
2892 * In the CMA case: long term pins in a CMA region would unnecessarily fragment
2893 * that region. And so, CMA attempts to migrate the page before pinning, when
2894 * FOLL_LONGTERM is specified.
2896 * FOLL_PIN indicates that a special kind of tracking (not just page->_refcount,
2897 * but an additional pin counting system) will be invoked. This is intended for
2898 * anything that gets a page reference and then touches page data (for example,
2899 * Direct IO). This lets the filesystem know that some non-file-system entity is
2900 * potentially changing the pages' data. In contrast to FOLL_GET (whose pages
2901 * are released via put_page()), FOLL_PIN pages must be released, ultimately, by
2902 * a call to unpin_user_page().
2904 * FOLL_PIN is similar to FOLL_GET: both of these pin pages. They use different
2905 * and separate refcounting mechanisms, however, and that means that each has
2906 * its own acquire and release mechanisms:
2908 * FOLL_GET: get_user_pages*() to acquire, and put_page() to release.
2910 * FOLL_PIN: pin_user_pages*() to acquire, and unpin_user_pages to release.
2912 * FOLL_PIN and FOLL_GET are mutually exclusive for a given function call.
2913 * (The underlying pages may experience both FOLL_GET-based and FOLL_PIN-based
2914 * calls applied to them, and that's perfectly OK. This is a constraint on the
2915 * callers, not on the pages.)
2917 * FOLL_PIN should be set internally by the pin_user_pages*() APIs, never
2918 * directly by the caller. That's in order to help avoid mismatches when
2919 * releasing pages: get_user_pages*() pages must be released via put_page(),
2920 * while pin_user_pages*() pages must be released via unpin_user_page().
2922 * Please see Documentation/core-api/pin_user_pages.rst for more information.
2925 static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
2927 if (vm_fault & VM_FAULT_OOM)
2929 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2930 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2931 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2936 typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
2937 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2938 unsigned long size, pte_fn_t fn, void *data);
2939 extern int apply_to_existing_page_range(struct mm_struct *mm,
2940 unsigned long address, unsigned long size,
2941 pte_fn_t fn, void *data);
2943 extern void init_mem_debugging_and_hardening(void);
2944 #ifdef CONFIG_PAGE_POISONING
2945 extern void __kernel_poison_pages(struct page *page, int numpages);
2946 extern void __kernel_unpoison_pages(struct page *page, int numpages);
2947 extern bool _page_poisoning_enabled_early;
2948 DECLARE_STATIC_KEY_FALSE(_page_poisoning_enabled);
2949 static inline bool page_poisoning_enabled(void)
2951 return _page_poisoning_enabled_early;
2954 * For use in fast paths after init_mem_debugging() has run, or when a
2955 * false negative result is not harmful when called too early.
2957 static inline bool page_poisoning_enabled_static(void)
2959 return static_branch_unlikely(&_page_poisoning_enabled);
2961 static inline void kernel_poison_pages(struct page *page, int numpages)
2963 if (page_poisoning_enabled_static())
2964 __kernel_poison_pages(page, numpages);
2966 static inline void kernel_unpoison_pages(struct page *page, int numpages)
2968 if (page_poisoning_enabled_static())
2969 __kernel_unpoison_pages(page, numpages);
2972 static inline bool page_poisoning_enabled(void) { return false; }
2973 static inline bool page_poisoning_enabled_static(void) { return false; }
2974 static inline void __kernel_poison_pages(struct page *page, int nunmpages) { }
2975 static inline void kernel_poison_pages(struct page *page, int numpages) { }
2976 static inline void kernel_unpoison_pages(struct page *page, int numpages) { }
2979 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_ALLOC_DEFAULT_ON, init_on_alloc);
2980 static inline bool want_init_on_alloc(gfp_t flags)
2982 if (static_branch_maybe(CONFIG_INIT_ON_ALLOC_DEFAULT_ON,
2985 return flags & __GFP_ZERO;
2988 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_FREE_DEFAULT_ON, init_on_free);
2989 static inline bool want_init_on_free(void)
2991 return static_branch_maybe(CONFIG_INIT_ON_FREE_DEFAULT_ON,
2995 extern bool _debug_pagealloc_enabled_early;
2996 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
2998 static inline bool debug_pagealloc_enabled(void)
3000 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
3001 _debug_pagealloc_enabled_early;
3005 * For use in fast paths after init_debug_pagealloc() has run, or when a
3006 * false negative result is not harmful when called too early.
3008 static inline bool debug_pagealloc_enabled_static(void)
3010 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
3013 return static_branch_unlikely(&_debug_pagealloc_enabled);
3016 #ifdef CONFIG_DEBUG_PAGEALLOC
3018 * To support DEBUG_PAGEALLOC architecture must ensure that
3019 * __kernel_map_pages() never fails
3021 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
3023 static inline void debug_pagealloc_map_pages(struct page *page, int numpages)
3025 if (debug_pagealloc_enabled_static())
3026 __kernel_map_pages(page, numpages, 1);
3029 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages)
3031 if (debug_pagealloc_enabled_static())
3032 __kernel_map_pages(page, numpages, 0);
3034 #else /* CONFIG_DEBUG_PAGEALLOC */
3035 static inline void debug_pagealloc_map_pages(struct page *page, int numpages) {}
3036 static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages) {}
3037 #endif /* CONFIG_DEBUG_PAGEALLOC */
3039 #ifdef __HAVE_ARCH_GATE_AREA
3040 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
3041 extern int in_gate_area_no_mm(unsigned long addr);
3042 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
3044 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
3048 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
3049 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
3053 #endif /* __HAVE_ARCH_GATE_AREA */
3055 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
3057 #ifdef CONFIG_SYSCTL
3058 extern int sysctl_drop_caches;
3059 int drop_caches_sysctl_handler(struct ctl_table *, int, void *, size_t *,
3063 void drop_slab(void);
3064 void drop_slab_node(int nid);
3067 #define randomize_va_space 0
3069 extern int randomize_va_space;
3072 const char * arch_vma_name(struct vm_area_struct *vma);
3074 void print_vma_addr(char *prefix, unsigned long rip);
3076 static inline void print_vma_addr(char *prefix, unsigned long rip)
3081 int vmemmap_remap_free(unsigned long start, unsigned long end,
3082 unsigned long reuse);
3083 int vmemmap_remap_alloc(unsigned long start, unsigned long end,
3084 unsigned long reuse, gfp_t gfp_mask);
3086 void *sparse_buffer_alloc(unsigned long size);
3087 struct page * __populate_section_memmap(unsigned long pfn,
3088 unsigned long nr_pages, int nid, struct vmem_altmap *altmap);
3089 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
3090 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
3091 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
3092 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
3093 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node,
3094 struct vmem_altmap *altmap);
3095 void *vmemmap_alloc_block(unsigned long size, int node);
3097 void *vmemmap_alloc_block_buf(unsigned long size, int node,
3098 struct vmem_altmap *altmap);
3099 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
3100 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
3101 int node, struct vmem_altmap *altmap);
3102 int vmemmap_populate(unsigned long start, unsigned long end, int node,
3103 struct vmem_altmap *altmap);
3104 void vmemmap_populate_print_last(void);
3105 #ifdef CONFIG_MEMORY_HOTPLUG
3106 void vmemmap_free(unsigned long start, unsigned long end,
3107 struct vmem_altmap *altmap);
3109 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
3110 unsigned long nr_pages);
3113 MF_COUNT_INCREASED = 1 << 0,
3114 MF_ACTION_REQUIRED = 1 << 1,
3115 MF_MUST_KILL = 1 << 2,
3116 MF_SOFT_OFFLINE = 1 << 3,
3118 extern int memory_failure(unsigned long pfn, int flags);
3119 extern void memory_failure_queue(unsigned long pfn, int flags);
3120 extern void memory_failure_queue_kick(int cpu);
3121 extern int unpoison_memory(unsigned long pfn);
3122 extern int sysctl_memory_failure_early_kill;
3123 extern int sysctl_memory_failure_recovery;
3124 extern void shake_page(struct page *p, int access);
3125 extern atomic_long_t num_poisoned_pages __read_mostly;
3126 extern int soft_offline_page(unsigned long pfn, int flags);
3130 * Error handlers for various types of pages.
3133 MF_IGNORED, /* Error: cannot be handled */
3134 MF_FAILED, /* Error: handling failed */
3135 MF_DELAYED, /* Will be handled later */
3136 MF_RECOVERED, /* Successfully recovered */
3139 enum mf_action_page_type {
3141 MF_MSG_KERNEL_HIGH_ORDER,
3143 MF_MSG_DIFFERENT_COMPOUND,
3144 MF_MSG_POISONED_HUGE,
3147 MF_MSG_NON_PMD_HUGE,
3148 MF_MSG_UNMAP_FAILED,
3149 MF_MSG_DIRTY_SWAPCACHE,
3150 MF_MSG_CLEAN_SWAPCACHE,
3151 MF_MSG_DIRTY_MLOCKED_LRU,
3152 MF_MSG_CLEAN_MLOCKED_LRU,
3153 MF_MSG_DIRTY_UNEVICTABLE_LRU,
3154 MF_MSG_CLEAN_UNEVICTABLE_LRU,
3157 MF_MSG_TRUNCATED_LRU,
3165 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
3166 extern void clear_huge_page(struct page *page,
3167 unsigned long addr_hint,
3168 unsigned int pages_per_huge_page);
3169 extern void copy_user_huge_page(struct page *dst, struct page *src,
3170 unsigned long addr_hint,
3171 struct vm_area_struct *vma,
3172 unsigned int pages_per_huge_page);
3173 extern long copy_huge_page_from_user(struct page *dst_page,
3174 const void __user *usr_src,
3175 unsigned int pages_per_huge_page,
3176 bool allow_pagefault);
3179 * vma_is_special_huge - Are transhuge page-table entries considered special?
3180 * @vma: Pointer to the struct vm_area_struct to consider
3182 * Whether transhuge page-table entries are considered "special" following
3183 * the definition in vm_normal_page().
3185 * Return: true if transhuge page-table entries should be considered special,
3188 static inline bool vma_is_special_huge(const struct vm_area_struct *vma)
3190 return vma_is_dax(vma) || (vma->vm_file &&
3191 (vma->vm_flags & (VM_PFNMAP | VM_MIXEDMAP)));
3194 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
3196 #ifdef CONFIG_DEBUG_PAGEALLOC
3197 extern unsigned int _debug_guardpage_minorder;
3198 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
3200 static inline unsigned int debug_guardpage_minorder(void)
3202 return _debug_guardpage_minorder;
3205 static inline bool debug_guardpage_enabled(void)
3207 return static_branch_unlikely(&_debug_guardpage_enabled);
3210 static inline bool page_is_guard(struct page *page)
3212 if (!debug_guardpage_enabled())
3215 return PageGuard(page);
3218 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
3219 static inline bool debug_guardpage_enabled(void) { return false; }
3220 static inline bool page_is_guard(struct page *page) { return false; }
3221 #endif /* CONFIG_DEBUG_PAGEALLOC */
3223 #if MAX_NUMNODES > 1
3224 void __init setup_nr_node_ids(void);
3226 static inline void setup_nr_node_ids(void) {}
3229 extern int memcmp_pages(struct page *page1, struct page *page2);
3231 static inline int pages_identical(struct page *page1, struct page *page2)
3233 return !memcmp_pages(page1, page2);
3236 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
3237 unsigned long clean_record_shared_mapping_range(struct address_space *mapping,
3238 pgoff_t first_index, pgoff_t nr,
3239 pgoff_t bitmap_pgoff,
3240 unsigned long *bitmap,
3244 unsigned long wp_shared_mapping_range(struct address_space *mapping,
3245 pgoff_t first_index, pgoff_t nr);
3248 extern int sysctl_nr_trim_pages;
3250 #ifdef CONFIG_PRINTK
3251 void mem_dump_obj(void *object);
3253 static inline void mem_dump_obj(void *object) {}
3257 * seal_check_future_write - Check for F_SEAL_FUTURE_WRITE flag and handle it
3258 * @seals: the seals to check
3259 * @vma: the vma to operate on
3261 * Check whether F_SEAL_FUTURE_WRITE is set; if so, do proper check/handling on
3262 * the vma flags. Return 0 if check pass, or <0 for errors.
3264 static inline int seal_check_future_write(int seals, struct vm_area_struct *vma)
3266 if (seals & F_SEAL_FUTURE_WRITE) {
3268 * New PROT_WRITE and MAP_SHARED mmaps are not allowed when
3269 * "future write" seal active.
3271 if ((vma->vm_flags & VM_SHARED) && (vma->vm_flags & VM_WRITE))
3275 * Since an F_SEAL_FUTURE_WRITE sealed memfd can be mapped as
3276 * MAP_SHARED and read-only, take care to not allow mprotect to
3277 * revert protections on such mappings. Do this only for shared
3278 * mappings. For private mappings, don't need to mask
3279 * VM_MAYWRITE as we still want them to be COW-writable.
3281 if (vma->vm_flags & VM_SHARED)
3282 vma->vm_flags &= ~(VM_MAYWRITE);
3288 #endif /* __KERNEL__ */
3289 #endif /* _LINUX_MM_H */