1 /* SPDX-License-Identifier: GPL-2.0 */
5 #include <linux/errno.h>
6 #include <linux/mmdebug.h>
9 #include <linux/list.h>
10 #include <linux/mmzone.h>
11 #include <linux/rbtree.h>
12 #include <linux/atomic.h>
13 #include <linux/debug_locks.h>
14 #include <linux/mm_types.h>
15 #include <linux/mmap_lock.h>
16 #include <linux/range.h>
17 #include <linux/pfn.h>
18 #include <linux/percpu-refcount.h>
19 #include <linux/bit_spinlock.h>
20 #include <linux/shrinker.h>
21 #include <linux/resource.h>
22 #include <linux/page_ext.h>
23 #include <linux/err.h>
24 #include <linux/page-flags.h>
25 #include <linux/page_ref.h>
26 #include <linux/overflow.h>
27 #include <linux/sizes.h>
28 #include <linux/sched.h>
29 #include <linux/pgtable.h>
30 #include <linux/kasan.h>
31 #include <linux/memremap.h>
35 struct anon_vma_chain
;
39 extern int sysctl_page_lock_unfairness
;
41 void init_mm_internals(void);
43 #ifndef CONFIG_NUMA /* Don't use mapnrs, do it properly */
44 extern unsigned long max_mapnr
;
46 static inline void set_max_mapnr(unsigned long limit
)
51 static inline void set_max_mapnr(unsigned long limit
) { }
54 extern atomic_long_t _totalram_pages
;
55 static inline unsigned long totalram_pages(void)
57 return (unsigned long)atomic_long_read(&_totalram_pages
);
60 static inline void totalram_pages_inc(void)
62 atomic_long_inc(&_totalram_pages
);
65 static inline void totalram_pages_dec(void)
67 atomic_long_dec(&_totalram_pages
);
70 static inline void totalram_pages_add(long count
)
72 atomic_long_add(count
, &_totalram_pages
);
75 extern void * high_memory
;
76 extern int page_cluster
;
77 extern const int page_cluster_max
;
80 extern int sysctl_legacy_va_layout
;
82 #define sysctl_legacy_va_layout 0
85 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
86 extern const int mmap_rnd_bits_min
;
87 extern const int mmap_rnd_bits_max
;
88 extern int mmap_rnd_bits __read_mostly
;
90 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
91 extern const int mmap_rnd_compat_bits_min
;
92 extern const int mmap_rnd_compat_bits_max
;
93 extern int mmap_rnd_compat_bits __read_mostly
;
97 #include <asm/processor.h>
100 * Architectures that support memory tagging (assigning tags to memory regions,
101 * embedding these tags into addresses that point to these memory regions, and
102 * checking that the memory and the pointer tags match on memory accesses)
103 * redefine this macro to strip tags from pointers.
104 * It's defined as noop for architectures that don't support memory tagging.
106 #ifndef untagged_addr
107 #define untagged_addr(addr) (addr)
111 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
115 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
119 #define lm_alias(x) __va(__pa_symbol(x))
123 * To prevent common memory management code establishing
124 * a zero page mapping on a read fault.
125 * This macro should be defined within <asm/pgtable.h>.
126 * s390 does this to prevent multiplexing of hardware bits
127 * related to the physical page in case of virtualization.
129 #ifndef mm_forbids_zeropage
130 #define mm_forbids_zeropage(X) (0)
134 * On some architectures it is expensive to call memset() for small sizes.
135 * If an architecture decides to implement their own version of
136 * mm_zero_struct_page they should wrap the defines below in a #ifndef and
137 * define their own version of this macro in <asm/pgtable.h>
139 #if BITS_PER_LONG == 64
140 /* This function must be updated when the size of struct page grows above 96
141 * or reduces below 56. The idea that compiler optimizes out switch()
142 * statement, and only leaves move/store instructions. Also the compiler can
143 * combine write statements if they are both assignments and can be reordered,
144 * this can result in several of the writes here being dropped.
146 #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
147 static inline void __mm_zero_struct_page(struct page
*page
)
149 unsigned long *_pp
= (void *)page
;
151 /* Check that struct page is either 56, 64, 72, 80, 88 or 96 bytes */
152 BUILD_BUG_ON(sizeof(struct page
) & 7);
153 BUILD_BUG_ON(sizeof(struct page
) < 56);
154 BUILD_BUG_ON(sizeof(struct page
) > 96);
156 switch (sizeof(struct page
)) {
183 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
187 * Default maximum number of active map areas, this limits the number of vmas
188 * per mm struct. Users can overwrite this number by sysctl but there is a
191 * When a program's coredump is generated as ELF format, a section is created
192 * per a vma. In ELF, the number of sections is represented in unsigned short.
193 * This means the number of sections should be smaller than 65535 at coredump.
194 * Because the kernel adds some informative sections to a image of program at
195 * generating coredump, we need some margin. The number of extra sections is
196 * 1-3 now and depends on arch. We use "5" as safe margin, here.
198 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
199 * not a hard limit any more. Although some userspace tools can be surprised by
202 #define MAPCOUNT_ELF_CORE_MARGIN (5)
203 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
205 extern int sysctl_max_map_count
;
207 extern unsigned long sysctl_user_reserve_kbytes
;
208 extern unsigned long sysctl_admin_reserve_kbytes
;
210 extern int sysctl_overcommit_memory
;
211 extern int sysctl_overcommit_ratio
;
212 extern unsigned long sysctl_overcommit_kbytes
;
214 int overcommit_ratio_handler(struct ctl_table
*, int, void *, size_t *,
216 int overcommit_kbytes_handler(struct ctl_table
*, int, void *, size_t *,
218 int overcommit_policy_handler(struct ctl_table
*, int, void *, size_t *,
221 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
222 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
223 #define folio_page_idx(folio, p) (page_to_pfn(p) - folio_pfn(folio))
225 #define nth_page(page,n) ((page) + (n))
226 #define folio_page_idx(folio, p) ((p) - &(folio)->page)
229 /* to align the pointer to the (next) page boundary */
230 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
232 /* to align the pointer to the (prev) page boundary */
233 #define PAGE_ALIGN_DOWN(addr) ALIGN_DOWN(addr, PAGE_SIZE)
235 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
236 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
238 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
239 static inline struct folio
*lru_to_folio(struct list_head
*head
)
241 return list_entry((head
)->prev
, struct folio
, lru
);
244 void setup_initial_init_mm(void *start_code
, void *end_code
,
245 void *end_data
, void *brk
);
248 * Linux kernel virtual memory manager primitives.
249 * The idea being to have a "virtual" mm in the same way
250 * we have a virtual fs - giving a cleaner interface to the
251 * mm details, and allowing different kinds of memory mappings
252 * (from shared memory to executable loading to arbitrary
256 struct vm_area_struct
*vm_area_alloc(struct mm_struct
*);
257 struct vm_area_struct
*vm_area_dup(struct vm_area_struct
*);
258 void vm_area_free(struct vm_area_struct
*);
261 extern struct rb_root nommu_region_tree
;
262 extern struct rw_semaphore nommu_region_sem
;
264 extern unsigned int kobjsize(const void *objp
);
268 * vm_flags in vm_area_struct, see mm_types.h.
269 * When changing, update also include/trace/events/mmflags.h
271 #define VM_NONE 0x00000000
273 #define VM_READ 0x00000001 /* currently active flags */
274 #define VM_WRITE 0x00000002
275 #define VM_EXEC 0x00000004
276 #define VM_SHARED 0x00000008
278 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
279 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
280 #define VM_MAYWRITE 0x00000020
281 #define VM_MAYEXEC 0x00000040
282 #define VM_MAYSHARE 0x00000080
284 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
286 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
287 #else /* CONFIG_MMU */
288 #define VM_MAYOVERLAY 0x00000200 /* nommu: R/O MAP_PRIVATE mapping that might overlay a file mapping */
289 #define VM_UFFD_MISSING 0
290 #endif /* CONFIG_MMU */
291 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
292 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
294 #define VM_LOCKED 0x00002000
295 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
297 /* Used by sys_madvise() */
298 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
299 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
301 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
302 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
303 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
304 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
305 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
306 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
307 #define VM_SYNC 0x00800000 /* Synchronous page faults */
308 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
309 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
310 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
312 #ifdef CONFIG_MEM_SOFT_DIRTY
313 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
315 # define VM_SOFTDIRTY 0
318 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
319 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
320 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
321 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
323 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
324 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
325 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
326 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
327 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
328 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
329 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
330 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
331 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
332 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
333 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
334 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
336 #ifdef CONFIG_ARCH_HAS_PKEYS
337 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
338 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
339 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
340 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
341 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
343 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
345 # define VM_PKEY_BIT4 0
347 #endif /* CONFIG_ARCH_HAS_PKEYS */
349 #if defined(CONFIG_X86)
350 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
351 #elif defined(CONFIG_PPC)
352 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
353 #elif defined(CONFIG_PARISC)
354 # define VM_GROWSUP VM_ARCH_1
355 #elif defined(CONFIG_IA64)
356 # define VM_GROWSUP VM_ARCH_1
357 #elif defined(CONFIG_SPARC64)
358 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
359 # define VM_ARCH_CLEAR VM_SPARC_ADI
360 #elif defined(CONFIG_ARM64)
361 # define VM_ARM64_BTI VM_ARCH_1 /* BTI guarded page, a.k.a. GP bit */
362 # define VM_ARCH_CLEAR VM_ARM64_BTI
363 #elif !defined(CONFIG_MMU)
364 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
367 #if defined(CONFIG_ARM64_MTE)
368 # define VM_MTE VM_HIGH_ARCH_0 /* Use Tagged memory for access control */
369 # define VM_MTE_ALLOWED VM_HIGH_ARCH_1 /* Tagged memory permitted */
371 # define VM_MTE VM_NONE
372 # define VM_MTE_ALLOWED VM_NONE
376 # define VM_GROWSUP VM_NONE
379 #ifdef CONFIG_HAVE_ARCH_USERFAULTFD_MINOR
380 # define VM_UFFD_MINOR_BIT 37
381 # define VM_UFFD_MINOR BIT(VM_UFFD_MINOR_BIT) /* UFFD minor faults */
382 #else /* !CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
383 # define VM_UFFD_MINOR VM_NONE
384 #endif /* CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
386 /* Bits set in the VMA until the stack is in its final location */
387 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
389 #define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0)
391 /* Common data flag combinations */
392 #define VM_DATA_FLAGS_TSK_EXEC (VM_READ | VM_WRITE | TASK_EXEC | \
393 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
394 #define VM_DATA_FLAGS_NON_EXEC (VM_READ | VM_WRITE | VM_MAYREAD | \
395 VM_MAYWRITE | VM_MAYEXEC)
396 #define VM_DATA_FLAGS_EXEC (VM_READ | VM_WRITE | VM_EXEC | \
397 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
399 #ifndef VM_DATA_DEFAULT_FLAGS /* arch can override this */
400 #define VM_DATA_DEFAULT_FLAGS VM_DATA_FLAGS_EXEC
403 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
404 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
407 #ifdef CONFIG_STACK_GROWSUP
408 #define VM_STACK VM_GROWSUP
410 #define VM_STACK VM_GROWSDOWN
413 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
415 /* VMA basic access permission flags */
416 #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
420 * Special vmas that are non-mergable, non-mlock()able.
422 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
424 /* This mask prevents VMA from being scanned with khugepaged */
425 #define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB)
427 /* This mask defines which mm->def_flags a process can inherit its parent */
428 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
430 /* This mask represents all the VMA flag bits used by mlock */
431 #define VM_LOCKED_MASK (VM_LOCKED | VM_LOCKONFAULT)
433 /* Arch-specific flags to clear when updating VM flags on protection change */
434 #ifndef VM_ARCH_CLEAR
435 # define VM_ARCH_CLEAR VM_NONE
437 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
440 * mapping from the currently active vm_flags protection bits (the
441 * low four bits) to a page protection mask..
445 * The default fault flags that should be used by most of the
446 * arch-specific page fault handlers.
448 #define FAULT_FLAG_DEFAULT (FAULT_FLAG_ALLOW_RETRY | \
449 FAULT_FLAG_KILLABLE | \
450 FAULT_FLAG_INTERRUPTIBLE)
453 * fault_flag_allow_retry_first - check ALLOW_RETRY the first time
454 * @flags: Fault flags.
456 * This is mostly used for places where we want to try to avoid taking
457 * the mmap_lock for too long a time when waiting for another condition
458 * to change, in which case we can try to be polite to release the
459 * mmap_lock in the first round to avoid potential starvation of other
460 * processes that would also want the mmap_lock.
462 * Return: true if the page fault allows retry and this is the first
463 * attempt of the fault handling; false otherwise.
465 static inline bool fault_flag_allow_retry_first(enum fault_flag flags
)
467 return (flags
& FAULT_FLAG_ALLOW_RETRY
) &&
468 (!(flags
& FAULT_FLAG_TRIED
));
471 #define FAULT_FLAG_TRACE \
472 { FAULT_FLAG_WRITE, "WRITE" }, \
473 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
474 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
475 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
476 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
477 { FAULT_FLAG_TRIED, "TRIED" }, \
478 { FAULT_FLAG_USER, "USER" }, \
479 { FAULT_FLAG_REMOTE, "REMOTE" }, \
480 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }, \
481 { FAULT_FLAG_INTERRUPTIBLE, "INTERRUPTIBLE" }
484 * vm_fault is filled by the pagefault handler and passed to the vma's
485 * ->fault function. The vma's ->fault is responsible for returning a bitmask
486 * of VM_FAULT_xxx flags that give details about how the fault was handled.
488 * MM layer fills up gfp_mask for page allocations but fault handler might
489 * alter it if its implementation requires a different allocation context.
491 * pgoff should be used in favour of virtual_address, if possible.
495 struct vm_area_struct
*vma
; /* Target VMA */
496 gfp_t gfp_mask
; /* gfp mask to be used for allocations */
497 pgoff_t pgoff
; /* Logical page offset based on vma */
498 unsigned long address
; /* Faulting virtual address - masked */
499 unsigned long real_address
; /* Faulting virtual address - unmasked */
501 enum fault_flag flags
; /* FAULT_FLAG_xxx flags
502 * XXX: should really be 'const' */
503 pmd_t
*pmd
; /* Pointer to pmd entry matching
505 pud_t
*pud
; /* Pointer to pud entry matching
509 pte_t orig_pte
; /* Value of PTE at the time of fault */
510 pmd_t orig_pmd
; /* Value of PMD at the time of fault,
511 * used by PMD fault only.
515 struct page
*cow_page
; /* Page handler may use for COW fault */
516 struct page
*page
; /* ->fault handlers should return a
517 * page here, unless VM_FAULT_NOPAGE
518 * is set (which is also implied by
521 /* These three entries are valid only while holding ptl lock */
522 pte_t
*pte
; /* Pointer to pte entry matching
523 * the 'address'. NULL if the page
524 * table hasn't been allocated.
526 spinlock_t
*ptl
; /* Page table lock.
527 * Protects pte page table if 'pte'
528 * is not NULL, otherwise pmd.
530 pgtable_t prealloc_pte
; /* Pre-allocated pte page table.
531 * vm_ops->map_pages() sets up a page
532 * table from atomic context.
533 * do_fault_around() pre-allocates
534 * page table to avoid allocation from
539 /* page entry size for vm->huge_fault() */
540 enum page_entry_size
{
547 * These are the virtual MM functions - opening of an area, closing and
548 * unmapping it (needed to keep files on disk up-to-date etc), pointer
549 * to the functions called when a no-page or a wp-page exception occurs.
551 struct vm_operations_struct
{
552 void (*open
)(struct vm_area_struct
* area
);
554 * @close: Called when the VMA is being removed from the MM.
555 * Context: User context. May sleep. Caller holds mmap_lock.
557 void (*close
)(struct vm_area_struct
* area
);
558 /* Called any time before splitting to check if it's allowed */
559 int (*may_split
)(struct vm_area_struct
*area
, unsigned long addr
);
560 int (*mremap
)(struct vm_area_struct
*area
);
562 * Called by mprotect() to make driver-specific permission
563 * checks before mprotect() is finalised. The VMA must not
564 * be modified. Returns 0 if mprotect() can proceed.
566 int (*mprotect
)(struct vm_area_struct
*vma
, unsigned long start
,
567 unsigned long end
, unsigned long newflags
);
568 vm_fault_t (*fault
)(struct vm_fault
*vmf
);
569 vm_fault_t (*huge_fault
)(struct vm_fault
*vmf
,
570 enum page_entry_size pe_size
);
571 vm_fault_t (*map_pages
)(struct vm_fault
*vmf
,
572 pgoff_t start_pgoff
, pgoff_t end_pgoff
);
573 unsigned long (*pagesize
)(struct vm_area_struct
* area
);
575 /* notification that a previously read-only page is about to become
576 * writable, if an error is returned it will cause a SIGBUS */
577 vm_fault_t (*page_mkwrite
)(struct vm_fault
*vmf
);
579 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
580 vm_fault_t (*pfn_mkwrite
)(struct vm_fault
*vmf
);
582 /* called by access_process_vm when get_user_pages() fails, typically
583 * for use by special VMAs. See also generic_access_phys() for a generic
584 * implementation useful for any iomem mapping.
586 int (*access
)(struct vm_area_struct
*vma
, unsigned long addr
,
587 void *buf
, int len
, int write
);
589 /* Called by the /proc/PID/maps code to ask the vma whether it
590 * has a special name. Returning non-NULL will also cause this
591 * vma to be dumped unconditionally. */
592 const char *(*name
)(struct vm_area_struct
*vma
);
596 * set_policy() op must add a reference to any non-NULL @new mempolicy
597 * to hold the policy upon return. Caller should pass NULL @new to
598 * remove a policy and fall back to surrounding context--i.e. do not
599 * install a MPOL_DEFAULT policy, nor the task or system default
602 int (*set_policy
)(struct vm_area_struct
*vma
, struct mempolicy
*new);
605 * get_policy() op must add reference [mpol_get()] to any policy at
606 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
607 * in mm/mempolicy.c will do this automatically.
608 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
609 * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
610 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
611 * must return NULL--i.e., do not "fallback" to task or system default
614 struct mempolicy
*(*get_policy
)(struct vm_area_struct
*vma
,
618 * Called by vm_normal_page() for special PTEs to find the
619 * page for @addr. This is useful if the default behavior
620 * (using pte_page()) would not find the correct page.
622 struct page
*(*find_special_page
)(struct vm_area_struct
*vma
,
626 static inline void vma_init(struct vm_area_struct
*vma
, struct mm_struct
*mm
)
628 static const struct vm_operations_struct dummy_vm_ops
= {};
630 memset(vma
, 0, sizeof(*vma
));
632 vma
->vm_ops
= &dummy_vm_ops
;
633 INIT_LIST_HEAD(&vma
->anon_vma_chain
);
636 /* Use when VMA is not part of the VMA tree and needs no locking */
637 static inline void vm_flags_init(struct vm_area_struct
*vma
,
640 ACCESS_PRIVATE(vma
, __vm_flags
) = flags
;
643 /* Use when VMA is part of the VMA tree and modifications need coordination */
644 static inline void vm_flags_reset(struct vm_area_struct
*vma
,
647 mmap_assert_write_locked(vma
->vm_mm
);
648 vm_flags_init(vma
, flags
);
651 static inline void vm_flags_reset_once(struct vm_area_struct
*vma
,
654 mmap_assert_write_locked(vma
->vm_mm
);
655 WRITE_ONCE(ACCESS_PRIVATE(vma
, __vm_flags
), flags
);
658 static inline void vm_flags_set(struct vm_area_struct
*vma
,
661 mmap_assert_write_locked(vma
->vm_mm
);
662 ACCESS_PRIVATE(vma
, __vm_flags
) |= flags
;
665 static inline void vm_flags_clear(struct vm_area_struct
*vma
,
668 mmap_assert_write_locked(vma
->vm_mm
);
669 ACCESS_PRIVATE(vma
, __vm_flags
) &= ~flags
;
673 * Use only if VMA is not part of the VMA tree or has no other users and
674 * therefore needs no locking.
676 static inline void __vm_flags_mod(struct vm_area_struct
*vma
,
677 vm_flags_t set
, vm_flags_t clear
)
679 vm_flags_init(vma
, (vma
->vm_flags
| set
) & ~clear
);
683 * Use only when the order of set/clear operations is unimportant, otherwise
684 * use vm_flags_{set|clear} explicitly.
686 static inline void vm_flags_mod(struct vm_area_struct
*vma
,
687 vm_flags_t set
, vm_flags_t clear
)
689 mmap_assert_write_locked(vma
->vm_mm
);
690 __vm_flags_mod(vma
, set
, clear
);
693 static inline void vma_set_anonymous(struct vm_area_struct
*vma
)
698 static inline bool vma_is_anonymous(struct vm_area_struct
*vma
)
703 static inline bool vma_is_temporary_stack(struct vm_area_struct
*vma
)
705 int maybe_stack
= vma
->vm_flags
& (VM_GROWSDOWN
| VM_GROWSUP
);
710 if ((vma
->vm_flags
& VM_STACK_INCOMPLETE_SETUP
) ==
711 VM_STACK_INCOMPLETE_SETUP
)
717 static inline bool vma_is_foreign(struct vm_area_struct
*vma
)
722 if (current
->mm
!= vma
->vm_mm
)
728 static inline bool vma_is_accessible(struct vm_area_struct
*vma
)
730 return vma
->vm_flags
& VM_ACCESS_FLAGS
;
734 struct vm_area_struct
*vma_find(struct vma_iterator
*vmi
, unsigned long max
)
736 return mas_find(&vmi
->mas
, max
- 1);
739 static inline struct vm_area_struct
*vma_next(struct vma_iterator
*vmi
)
742 * Uses mas_find() to get the first VMA when the iterator starts.
743 * Calling mas_next() could skip the first entry.
745 return mas_find(&vmi
->mas
, ULONG_MAX
);
748 static inline struct vm_area_struct
*vma_prev(struct vma_iterator
*vmi
)
750 return mas_prev(&vmi
->mas
, 0);
753 static inline unsigned long vma_iter_addr(struct vma_iterator
*vmi
)
755 return vmi
->mas
.index
;
758 static inline unsigned long vma_iter_end(struct vma_iterator
*vmi
)
760 return vmi
->mas
.last
+ 1;
762 static inline int vma_iter_bulk_alloc(struct vma_iterator
*vmi
,
765 return mas_expected_entries(&vmi
->mas
, count
);
768 /* Free any unused preallocations */
769 static inline void vma_iter_free(struct vma_iterator
*vmi
)
771 mas_destroy(&vmi
->mas
);
774 static inline int vma_iter_bulk_store(struct vma_iterator
*vmi
,
775 struct vm_area_struct
*vma
)
777 vmi
->mas
.index
= vma
->vm_start
;
778 vmi
->mas
.last
= vma
->vm_end
- 1;
779 mas_store(&vmi
->mas
, vma
);
780 if (unlikely(mas_is_err(&vmi
->mas
)))
786 static inline void vma_iter_invalidate(struct vma_iterator
*vmi
)
788 mas_pause(&vmi
->mas
);
791 static inline void vma_iter_set(struct vma_iterator
*vmi
, unsigned long addr
)
793 mas_set(&vmi
->mas
, addr
);
796 #define for_each_vma(__vmi, __vma) \
797 while (((__vma) = vma_next(&(__vmi))) != NULL)
799 /* The MM code likes to work with exclusive end addresses */
800 #define for_each_vma_range(__vmi, __vma, __end) \
801 while (((__vma) = vma_find(&(__vmi), (__end))) != NULL)
805 * The vma_is_shmem is not inline because it is used only by slow
806 * paths in userfault.
808 bool vma_is_shmem(struct vm_area_struct
*vma
);
809 bool vma_is_anon_shmem(struct vm_area_struct
*vma
);
811 static inline bool vma_is_shmem(struct vm_area_struct
*vma
) { return false; }
812 static inline bool vma_is_anon_shmem(struct vm_area_struct
*vma
) { return false; }
815 int vma_is_stack_for_current(struct vm_area_struct
*vma
);
817 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
818 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
824 * compound_order() can be called without holding a reference, which means
825 * that niceties like page_folio() don't work. These callers should be
826 * prepared to handle wild return values. For example, PG_head may be
827 * set before _folio_order is initialised, or this may be a tail page.
828 * See compaction.c for some good examples.
830 static inline unsigned int compound_order(struct page
*page
)
832 struct folio
*folio
= (struct folio
*)page
;
834 if (!test_bit(PG_head
, &folio
->flags
))
836 return folio
->_folio_order
;
840 * folio_order - The allocation order of a folio.
843 * A folio is composed of 2^order pages. See get_order() for the definition
846 * Return: The order of the folio.
848 static inline unsigned int folio_order(struct folio
*folio
)
850 if (!folio_test_large(folio
))
852 return folio
->_folio_order
;
855 #include <linux/huge_mm.h>
858 * Methods to modify the page usage count.
860 * What counts for a page usage:
861 * - cache mapping (page->mapping)
862 * - private data (page->private)
863 * - page mapped in a task's page tables, each mapping
864 * is counted separately
866 * Also, many kernel routines increase the page count before a critical
867 * routine so they can be sure the page doesn't go away from under them.
871 * Drop a ref, return true if the refcount fell to zero (the page has no users)
873 static inline int put_page_testzero(struct page
*page
)
875 VM_BUG_ON_PAGE(page_ref_count(page
) == 0, page
);
876 return page_ref_dec_and_test(page
);
879 static inline int folio_put_testzero(struct folio
*folio
)
881 return put_page_testzero(&folio
->page
);
885 * Try to grab a ref unless the page has a refcount of zero, return false if
887 * This can be called when MMU is off so it must not access
888 * any of the virtual mappings.
890 static inline bool get_page_unless_zero(struct page
*page
)
892 return page_ref_add_unless(page
, 1, 0);
895 static inline struct folio
*folio_get_nontail_page(struct page
*page
)
897 if (unlikely(!get_page_unless_zero(page
)))
899 return (struct folio
*)page
;
902 extern int page_is_ram(unsigned long pfn
);
910 int region_intersects(resource_size_t offset
, size_t size
, unsigned long flags
,
913 /* Support for virtually mapped pages */
914 struct page
*vmalloc_to_page(const void *addr
);
915 unsigned long vmalloc_to_pfn(const void *addr
);
918 * Determine if an address is within the vmalloc range
920 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
921 * is no special casing required.
924 #ifndef is_ioremap_addr
925 #define is_ioremap_addr(x) is_vmalloc_addr(x)
929 extern bool is_vmalloc_addr(const void *x
);
930 extern int is_vmalloc_or_module_addr(const void *x
);
932 static inline bool is_vmalloc_addr(const void *x
)
936 static inline int is_vmalloc_or_module_addr(const void *x
)
943 * How many times the entire folio is mapped as a single unit (eg by a
944 * PMD or PUD entry). This is probably not what you want, except for
945 * debugging purposes - it does not include PTE-mapped sub-pages; look
946 * at folio_mapcount() or page_mapcount() or total_mapcount() instead.
948 static inline int folio_entire_mapcount(struct folio
*folio
)
950 VM_BUG_ON_FOLIO(!folio_test_large(folio
), folio
);
951 return atomic_read(&folio
->_entire_mapcount
) + 1;
955 * The atomic page->_mapcount, starts from -1: so that transitions
956 * both from it and to it can be tracked, using atomic_inc_and_test
957 * and atomic_add_negative(-1).
959 static inline void page_mapcount_reset(struct page
*page
)
961 atomic_set(&(page
)->_mapcount
, -1);
965 * page_mapcount() - Number of times this precise page is mapped.
968 * The number of times this page is mapped. If this page is part of
969 * a large folio, it includes the number of times this page is mapped
970 * as part of that folio.
972 * The result is undefined for pages which cannot be mapped into userspace.
973 * For example SLAB or special types of pages. See function page_has_type().
974 * They use this field in struct page differently.
976 static inline int page_mapcount(struct page
*page
)
978 int mapcount
= atomic_read(&page
->_mapcount
) + 1;
980 if (unlikely(PageCompound(page
)))
981 mapcount
+= folio_entire_mapcount(page_folio(page
));
986 int folio_total_mapcount(struct folio
*folio
);
989 * folio_mapcount() - Calculate the number of mappings of this folio.
992 * A large folio tracks both how many times the entire folio is mapped,
993 * and how many times each individual page in the folio is mapped.
994 * This function calculates the total number of times the folio is
997 * Return: The number of times this folio is mapped.
999 static inline int folio_mapcount(struct folio
*folio
)
1001 if (likely(!folio_test_large(folio
)))
1002 return atomic_read(&folio
->_mapcount
) + 1;
1003 return folio_total_mapcount(folio
);
1006 static inline int total_mapcount(struct page
*page
)
1008 if (likely(!PageCompound(page
)))
1009 return atomic_read(&page
->_mapcount
) + 1;
1010 return folio_total_mapcount(page_folio(page
));
1013 static inline bool folio_large_is_mapped(struct folio
*folio
)
1016 * Reading _entire_mapcount below could be omitted if hugetlb
1017 * participated in incrementing nr_pages_mapped when compound mapped.
1019 return atomic_read(&folio
->_nr_pages_mapped
) > 0 ||
1020 atomic_read(&folio
->_entire_mapcount
) >= 0;
1024 * folio_mapped - Is this folio mapped into userspace?
1025 * @folio: The folio.
1027 * Return: True if any page in this folio is referenced by user page tables.
1029 static inline bool folio_mapped(struct folio
*folio
)
1031 if (likely(!folio_test_large(folio
)))
1032 return atomic_read(&folio
->_mapcount
) >= 0;
1033 return folio_large_is_mapped(folio
);
1037 * Return true if this page is mapped into pagetables.
1038 * For compound page it returns true if any sub-page of compound page is mapped,
1039 * even if this particular sub-page is not itself mapped by any PTE or PMD.
1041 static inline bool page_mapped(struct page
*page
)
1043 if (likely(!PageCompound(page
)))
1044 return atomic_read(&page
->_mapcount
) >= 0;
1045 return folio_large_is_mapped(page_folio(page
));
1048 static inline struct page
*virt_to_head_page(const void *x
)
1050 struct page
*page
= virt_to_page(x
);
1052 return compound_head(page
);
1055 static inline struct folio
*virt_to_folio(const void *x
)
1057 struct page
*page
= virt_to_page(x
);
1059 return page_folio(page
);
1062 void __folio_put(struct folio
*folio
);
1064 void put_pages_list(struct list_head
*pages
);
1066 void split_page(struct page
*page
, unsigned int order
);
1067 void folio_copy(struct folio
*dst
, struct folio
*src
);
1069 unsigned long nr_free_buffer_pages(void);
1072 * Compound pages have a destructor function. Provide a
1073 * prototype for that function and accessor functions.
1074 * These are _only_ valid on the head of a compound page.
1076 typedef void compound_page_dtor(struct page
*);
1078 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
1079 enum compound_dtor_id
{
1082 #ifdef CONFIG_HUGETLB_PAGE
1085 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1086 TRANSHUGE_PAGE_DTOR
,
1090 extern compound_page_dtor
* const compound_page_dtors
[NR_COMPOUND_DTORS
];
1092 static inline void set_compound_page_dtor(struct page
*page
,
1093 enum compound_dtor_id compound_dtor
)
1095 struct folio
*folio
= (struct folio
*)page
;
1097 VM_BUG_ON_PAGE(compound_dtor
>= NR_COMPOUND_DTORS
, page
);
1098 VM_BUG_ON_PAGE(!PageHead(page
), page
);
1099 folio
->_folio_dtor
= compound_dtor
;
1102 static inline void folio_set_compound_dtor(struct folio
*folio
,
1103 enum compound_dtor_id compound_dtor
)
1105 VM_BUG_ON_FOLIO(compound_dtor
>= NR_COMPOUND_DTORS
, folio
);
1106 folio
->_folio_dtor
= compound_dtor
;
1109 void destroy_large_folio(struct folio
*folio
);
1111 static inline void set_compound_order(struct page
*page
, unsigned int order
)
1113 struct folio
*folio
= (struct folio
*)page
;
1115 folio
->_folio_order
= order
;
1117 folio
->_folio_nr_pages
= 1U << order
;
1121 /* Returns the number of bytes in this potentially compound page. */
1122 static inline unsigned long page_size(struct page
*page
)
1124 return PAGE_SIZE
<< compound_order(page
);
1127 /* Returns the number of bits needed for the number of bytes in a page */
1128 static inline unsigned int page_shift(struct page
*page
)
1130 return PAGE_SHIFT
+ compound_order(page
);
1134 * thp_order - Order of a transparent huge page.
1135 * @page: Head page of a transparent huge page.
1137 static inline unsigned int thp_order(struct page
*page
)
1139 VM_BUG_ON_PGFLAGS(PageTail(page
), page
);
1140 return compound_order(page
);
1144 * thp_size - Size of a transparent huge page.
1145 * @page: Head page of a transparent huge page.
1147 * Return: Number of bytes in this page.
1149 static inline unsigned long thp_size(struct page
*page
)
1151 return PAGE_SIZE
<< thp_order(page
);
1154 void free_compound_page(struct page
*page
);
1158 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
1159 * servicing faults for write access. In the normal case, do always want
1160 * pte_mkwrite. But get_user_pages can cause write faults for mappings
1161 * that do not have writing enabled, when used by access_process_vm.
1163 static inline pte_t
maybe_mkwrite(pte_t pte
, struct vm_area_struct
*vma
)
1165 if (likely(vma
->vm_flags
& VM_WRITE
))
1166 pte
= pte_mkwrite(pte
);
1170 vm_fault_t
do_set_pmd(struct vm_fault
*vmf
, struct page
*page
);
1171 void do_set_pte(struct vm_fault
*vmf
, struct page
*page
, unsigned long addr
);
1173 vm_fault_t
finish_fault(struct vm_fault
*vmf
);
1174 vm_fault_t
finish_mkwrite_fault(struct vm_fault
*vmf
);
1178 * Multiple processes may "see" the same page. E.g. for untouched
1179 * mappings of /dev/null, all processes see the same page full of
1180 * zeroes, and text pages of executables and shared libraries have
1181 * only one copy in memory, at most, normally.
1183 * For the non-reserved pages, page_count(page) denotes a reference count.
1184 * page_count() == 0 means the page is free. page->lru is then used for
1185 * freelist management in the buddy allocator.
1186 * page_count() > 0 means the page has been allocated.
1188 * Pages are allocated by the slab allocator in order to provide memory
1189 * to kmalloc and kmem_cache_alloc. In this case, the management of the
1190 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
1191 * unless a particular usage is carefully commented. (the responsibility of
1192 * freeing the kmalloc memory is the caller's, of course).
1194 * A page may be used by anyone else who does a __get_free_page().
1195 * In this case, page_count still tracks the references, and should only
1196 * be used through the normal accessor functions. The top bits of page->flags
1197 * and page->virtual store page management information, but all other fields
1198 * are unused and could be used privately, carefully. The management of this
1199 * page is the responsibility of the one who allocated it, and those who have
1200 * subsequently been given references to it.
1202 * The other pages (we may call them "pagecache pages") are completely
1203 * managed by the Linux memory manager: I/O, buffers, swapping etc.
1204 * The following discussion applies only to them.
1206 * A pagecache page contains an opaque `private' member, which belongs to the
1207 * page's address_space. Usually, this is the address of a circular list of
1208 * the page's disk buffers. PG_private must be set to tell the VM to call
1209 * into the filesystem to release these pages.
1211 * A page may belong to an inode's memory mapping. In this case, page->mapping
1212 * is the pointer to the inode, and page->index is the file offset of the page,
1213 * in units of PAGE_SIZE.
1215 * If pagecache pages are not associated with an inode, they are said to be
1216 * anonymous pages. These may become associated with the swapcache, and in that
1217 * case PG_swapcache is set, and page->private is an offset into the swapcache.
1219 * In either case (swapcache or inode backed), the pagecache itself holds one
1220 * reference to the page. Setting PG_private should also increment the
1221 * refcount. The each user mapping also has a reference to the page.
1223 * The pagecache pages are stored in a per-mapping radix tree, which is
1224 * rooted at mapping->i_pages, and indexed by offset.
1225 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
1226 * lists, we instead now tag pages as dirty/writeback in the radix tree.
1228 * All pagecache pages may be subject to I/O:
1229 * - inode pages may need to be read from disk,
1230 * - inode pages which have been modified and are MAP_SHARED may need
1231 * to be written back to the inode on disk,
1232 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
1233 * modified may need to be swapped out to swap space and (later) to be read
1237 #if defined(CONFIG_ZONE_DEVICE) && defined(CONFIG_FS_DAX)
1238 DECLARE_STATIC_KEY_FALSE(devmap_managed_key
);
1240 bool __put_devmap_managed_page_refs(struct page
*page
, int refs
);
1241 static inline bool put_devmap_managed_page_refs(struct page
*page
, int refs
)
1243 if (!static_branch_unlikely(&devmap_managed_key
))
1245 if (!is_zone_device_page(page
))
1247 return __put_devmap_managed_page_refs(page
, refs
);
1249 #else /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1250 static inline bool put_devmap_managed_page_refs(struct page
*page
, int refs
)
1254 #endif /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1256 static inline bool put_devmap_managed_page(struct page
*page
)
1258 return put_devmap_managed_page_refs(page
, 1);
1261 /* 127: arbitrary random number, small enough to assemble well */
1262 #define folio_ref_zero_or_close_to_overflow(folio) \
1263 ((unsigned int) folio_ref_count(folio) + 127u <= 127u)
1266 * folio_get - Increment the reference count on a folio.
1267 * @folio: The folio.
1269 * Context: May be called in any context, as long as you know that
1270 * you have a refcount on the folio. If you do not already have one,
1271 * folio_try_get() may be the right interface for you to use.
1273 static inline void folio_get(struct folio
*folio
)
1275 VM_BUG_ON_FOLIO(folio_ref_zero_or_close_to_overflow(folio
), folio
);
1276 folio_ref_inc(folio
);
1279 static inline void get_page(struct page
*page
)
1281 folio_get(page_folio(page
));
1284 static inline __must_check
bool try_get_page(struct page
*page
)
1286 page
= compound_head(page
);
1287 if (WARN_ON_ONCE(page_ref_count(page
) <= 0))
1294 * folio_put - Decrement the reference count on a folio.
1295 * @folio: The folio.
1297 * If the folio's reference count reaches zero, the memory will be
1298 * released back to the page allocator and may be used by another
1299 * allocation immediately. Do not access the memory or the struct folio
1300 * after calling folio_put() unless you can be sure that it wasn't the
1303 * Context: May be called in process or interrupt context, but not in NMI
1304 * context. May be called while holding a spinlock.
1306 static inline void folio_put(struct folio
*folio
)
1308 if (folio_put_testzero(folio
))
1313 * folio_put_refs - Reduce the reference count on a folio.
1314 * @folio: The folio.
1315 * @refs: The amount to subtract from the folio's reference count.
1317 * If the folio's reference count reaches zero, the memory will be
1318 * released back to the page allocator and may be used by another
1319 * allocation immediately. Do not access the memory or the struct folio
1320 * after calling folio_put_refs() unless you can be sure that these weren't
1321 * the last references.
1323 * Context: May be called in process or interrupt context, but not in NMI
1324 * context. May be called while holding a spinlock.
1326 static inline void folio_put_refs(struct folio
*folio
, int refs
)
1328 if (folio_ref_sub_and_test(folio
, refs
))
1333 * union release_pages_arg - an array of pages or folios
1335 * release_pages() releases a simple array of multiple pages, and
1336 * accepts various different forms of said page array: either
1337 * a regular old boring array of pages, an array of folios, or
1338 * an array of encoded page pointers.
1340 * The transparent union syntax for this kind of "any of these
1341 * argument types" is all kinds of ugly, so look away.
1344 struct page
**pages
;
1345 struct folio
**folios
;
1346 struct encoded_page
**encoded_pages
;
1347 } release_pages_arg
__attribute__ ((__transparent_union__
));
1349 void release_pages(release_pages_arg
, int nr
);
1352 * folios_put - Decrement the reference count on an array of folios.
1353 * @folios: The folios.
1354 * @nr: How many folios there are.
1356 * Like folio_put(), but for an array of folios. This is more efficient
1357 * than writing the loop yourself as it will optimise the locks which
1358 * need to be taken if the folios are freed.
1360 * Context: May be called in process or interrupt context, but not in NMI
1361 * context. May be called while holding a spinlock.
1363 static inline void folios_put(struct folio
**folios
, unsigned int nr
)
1365 release_pages(folios
, nr
);
1368 static inline void put_page(struct page
*page
)
1370 struct folio
*folio
= page_folio(page
);
1373 * For some devmap managed pages we need to catch refcount transition
1376 if (put_devmap_managed_page(&folio
->page
))
1382 * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
1383 * the page's refcount so that two separate items are tracked: the original page
1384 * reference count, and also a new count of how many pin_user_pages() calls were
1385 * made against the page. ("gup-pinned" is another term for the latter).
1387 * With this scheme, pin_user_pages() becomes special: such pages are marked as
1388 * distinct from normal pages. As such, the unpin_user_page() call (and its
1389 * variants) must be used in order to release gup-pinned pages.
1393 * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
1394 * counts with respect to pin_user_pages() and unpin_user_page() becomes
1395 * simpler, due to the fact that adding an even power of two to the page
1396 * refcount has the effect of using only the upper N bits, for the code that
1397 * counts up using the bias value. This means that the lower bits are left for
1398 * the exclusive use of the original code that increments and decrements by one
1399 * (or at least, by much smaller values than the bias value).
1401 * Of course, once the lower bits overflow into the upper bits (and this is
1402 * OK, because subtraction recovers the original values), then visual inspection
1403 * no longer suffices to directly view the separate counts. However, for normal
1404 * applications that don't have huge page reference counts, this won't be an
1407 * Locking: the lockless algorithm described in folio_try_get_rcu()
1408 * provides safe operation for get_user_pages(), page_mkclean() and
1409 * other calls that race to set up page table entries.
1411 #define GUP_PIN_COUNTING_BIAS (1U << 10)
1413 void unpin_user_page(struct page
*page
);
1414 void unpin_user_pages_dirty_lock(struct page
**pages
, unsigned long npages
,
1416 void unpin_user_page_range_dirty_lock(struct page
*page
, unsigned long npages
,
1418 void unpin_user_pages(struct page
**pages
, unsigned long npages
);
1420 static inline bool is_cow_mapping(vm_flags_t flags
)
1422 return (flags
& (VM_SHARED
| VM_MAYWRITE
)) == VM_MAYWRITE
;
1426 static inline bool is_nommu_shared_mapping(vm_flags_t flags
)
1429 * NOMMU shared mappings are ordinary MAP_SHARED mappings and selected
1430 * R/O MAP_PRIVATE file mappings that are an effective R/O overlay of
1431 * a file mapping. R/O MAP_PRIVATE mappings might still modify
1432 * underlying memory if ptrace is active, so this is only possible if
1433 * ptrace does not apply. Note that there is no mprotect() to upgrade
1434 * write permissions later.
1436 return flags
& (VM_MAYSHARE
| VM_MAYOVERLAY
);
1440 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1441 #define SECTION_IN_PAGE_FLAGS
1445 * The identification function is mainly used by the buddy allocator for
1446 * determining if two pages could be buddies. We are not really identifying
1447 * the zone since we could be using the section number id if we do not have
1448 * node id available in page flags.
1449 * We only guarantee that it will return the same value for two combinable
1452 static inline int page_zone_id(struct page
*page
)
1454 return (page
->flags
>> ZONEID_PGSHIFT
) & ZONEID_MASK
;
1457 #ifdef NODE_NOT_IN_PAGE_FLAGS
1458 extern int page_to_nid(const struct page
*page
);
1460 static inline int page_to_nid(const struct page
*page
)
1462 struct page
*p
= (struct page
*)page
;
1464 return (PF_POISONED_CHECK(p
)->flags
>> NODES_PGSHIFT
) & NODES_MASK
;
1468 static inline int folio_nid(const struct folio
*folio
)
1470 return page_to_nid(&folio
->page
);
1473 #ifdef CONFIG_NUMA_BALANCING
1474 /* page access time bits needs to hold at least 4 seconds */
1475 #define PAGE_ACCESS_TIME_MIN_BITS 12
1476 #if LAST_CPUPID_SHIFT < PAGE_ACCESS_TIME_MIN_BITS
1477 #define PAGE_ACCESS_TIME_BUCKETS \
1478 (PAGE_ACCESS_TIME_MIN_BITS - LAST_CPUPID_SHIFT)
1480 #define PAGE_ACCESS_TIME_BUCKETS 0
1483 #define PAGE_ACCESS_TIME_MASK \
1484 (LAST_CPUPID_MASK << PAGE_ACCESS_TIME_BUCKETS)
1486 static inline int cpu_pid_to_cpupid(int cpu
, int pid
)
1488 return ((cpu
& LAST__CPU_MASK
) << LAST__PID_SHIFT
) | (pid
& LAST__PID_MASK
);
1491 static inline int cpupid_to_pid(int cpupid
)
1493 return cpupid
& LAST__PID_MASK
;
1496 static inline int cpupid_to_cpu(int cpupid
)
1498 return (cpupid
>> LAST__PID_SHIFT
) & LAST__CPU_MASK
;
1501 static inline int cpupid_to_nid(int cpupid
)
1503 return cpu_to_node(cpupid_to_cpu(cpupid
));
1506 static inline bool cpupid_pid_unset(int cpupid
)
1508 return cpupid_to_pid(cpupid
) == (-1 & LAST__PID_MASK
);
1511 static inline bool cpupid_cpu_unset(int cpupid
)
1513 return cpupid_to_cpu(cpupid
) == (-1 & LAST__CPU_MASK
);
1516 static inline bool __cpupid_match_pid(pid_t task_pid
, int cpupid
)
1518 return (task_pid
& LAST__PID_MASK
) == cpupid_to_pid(cpupid
);
1521 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1522 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1523 static inline int page_cpupid_xchg_last(struct page
*page
, int cpupid
)
1525 return xchg(&page
->_last_cpupid
, cpupid
& LAST_CPUPID_MASK
);
1528 static inline int page_cpupid_last(struct page
*page
)
1530 return page
->_last_cpupid
;
1532 static inline void page_cpupid_reset_last(struct page
*page
)
1534 page
->_last_cpupid
= -1 & LAST_CPUPID_MASK
;
1537 static inline int page_cpupid_last(struct page
*page
)
1539 return (page
->flags
>> LAST_CPUPID_PGSHIFT
) & LAST_CPUPID_MASK
;
1542 extern int page_cpupid_xchg_last(struct page
*page
, int cpupid
);
1544 static inline void page_cpupid_reset_last(struct page
*page
)
1546 page
->flags
|= LAST_CPUPID_MASK
<< LAST_CPUPID_PGSHIFT
;
1548 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1550 static inline int xchg_page_access_time(struct page
*page
, int time
)
1554 last_time
= page_cpupid_xchg_last(page
, time
>> PAGE_ACCESS_TIME_BUCKETS
);
1555 return last_time
<< PAGE_ACCESS_TIME_BUCKETS
;
1557 #else /* !CONFIG_NUMA_BALANCING */
1558 static inline int page_cpupid_xchg_last(struct page
*page
, int cpupid
)
1560 return page_to_nid(page
); /* XXX */
1563 static inline int xchg_page_access_time(struct page
*page
, int time
)
1568 static inline int page_cpupid_last(struct page
*page
)
1570 return page_to_nid(page
); /* XXX */
1573 static inline int cpupid_to_nid(int cpupid
)
1578 static inline int cpupid_to_pid(int cpupid
)
1583 static inline int cpupid_to_cpu(int cpupid
)
1588 static inline int cpu_pid_to_cpupid(int nid
, int pid
)
1593 static inline bool cpupid_pid_unset(int cpupid
)
1598 static inline void page_cpupid_reset_last(struct page
*page
)
1602 static inline bool cpupid_match_pid(struct task_struct
*task
, int cpupid
)
1606 #endif /* CONFIG_NUMA_BALANCING */
1608 #if defined(CONFIG_KASAN_SW_TAGS) || defined(CONFIG_KASAN_HW_TAGS)
1611 * KASAN per-page tags are stored xor'ed with 0xff. This allows to avoid
1612 * setting tags for all pages to native kernel tag value 0xff, as the default
1613 * value 0x00 maps to 0xff.
1616 static inline u8
page_kasan_tag(const struct page
*page
)
1620 if (kasan_enabled()) {
1621 tag
= (page
->flags
>> KASAN_TAG_PGSHIFT
) & KASAN_TAG_MASK
;
1628 static inline void page_kasan_tag_set(struct page
*page
, u8 tag
)
1630 unsigned long old_flags
, flags
;
1632 if (!kasan_enabled())
1636 old_flags
= READ_ONCE(page
->flags
);
1639 flags
&= ~(KASAN_TAG_MASK
<< KASAN_TAG_PGSHIFT
);
1640 flags
|= (tag
& KASAN_TAG_MASK
) << KASAN_TAG_PGSHIFT
;
1641 } while (unlikely(!try_cmpxchg(&page
->flags
, &old_flags
, flags
)));
1644 static inline void page_kasan_tag_reset(struct page
*page
)
1646 if (kasan_enabled())
1647 page_kasan_tag_set(page
, 0xff);
1650 #else /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1652 static inline u8
page_kasan_tag(const struct page
*page
)
1657 static inline void page_kasan_tag_set(struct page
*page
, u8 tag
) { }
1658 static inline void page_kasan_tag_reset(struct page
*page
) { }
1660 #endif /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1662 static inline struct zone
*page_zone(const struct page
*page
)
1664 return &NODE_DATA(page_to_nid(page
))->node_zones
[page_zonenum(page
)];
1667 static inline pg_data_t
*page_pgdat(const struct page
*page
)
1669 return NODE_DATA(page_to_nid(page
));
1672 static inline struct zone
*folio_zone(const struct folio
*folio
)
1674 return page_zone(&folio
->page
);
1677 static inline pg_data_t
*folio_pgdat(const struct folio
*folio
)
1679 return page_pgdat(&folio
->page
);
1682 #ifdef SECTION_IN_PAGE_FLAGS
1683 static inline void set_page_section(struct page
*page
, unsigned long section
)
1685 page
->flags
&= ~(SECTIONS_MASK
<< SECTIONS_PGSHIFT
);
1686 page
->flags
|= (section
& SECTIONS_MASK
) << SECTIONS_PGSHIFT
;
1689 static inline unsigned long page_to_section(const struct page
*page
)
1691 return (page
->flags
>> SECTIONS_PGSHIFT
) & SECTIONS_MASK
;
1696 * folio_pfn - Return the Page Frame Number of a folio.
1697 * @folio: The folio.
1699 * A folio may contain multiple pages. The pages have consecutive
1700 * Page Frame Numbers.
1702 * Return: The Page Frame Number of the first page in the folio.
1704 static inline unsigned long folio_pfn(struct folio
*folio
)
1706 return page_to_pfn(&folio
->page
);
1709 static inline struct folio
*pfn_folio(unsigned long pfn
)
1711 return page_folio(pfn_to_page(pfn
));
1715 * folio_maybe_dma_pinned - Report if a folio may be pinned for DMA.
1716 * @folio: The folio.
1718 * This function checks if a folio has been pinned via a call to
1719 * a function in the pin_user_pages() family.
1721 * For small folios, the return value is partially fuzzy: false is not fuzzy,
1722 * because it means "definitely not pinned for DMA", but true means "probably
1723 * pinned for DMA, but possibly a false positive due to having at least
1724 * GUP_PIN_COUNTING_BIAS worth of normal folio references".
1726 * False positives are OK, because: a) it's unlikely for a folio to
1727 * get that many refcounts, and b) all the callers of this routine are
1728 * expected to be able to deal gracefully with a false positive.
1730 * For large folios, the result will be exactly correct. That's because
1731 * we have more tracking data available: the _pincount field is used
1732 * instead of the GUP_PIN_COUNTING_BIAS scheme.
1734 * For more information, please see Documentation/core-api/pin_user_pages.rst.
1736 * Return: True, if it is likely that the page has been "dma-pinned".
1737 * False, if the page is definitely not dma-pinned.
1739 static inline bool folio_maybe_dma_pinned(struct folio
*folio
)
1741 if (folio_test_large(folio
))
1742 return atomic_read(&folio
->_pincount
) > 0;
1745 * folio_ref_count() is signed. If that refcount overflows, then
1746 * folio_ref_count() returns a negative value, and callers will avoid
1747 * further incrementing the refcount.
1749 * Here, for that overflow case, use the sign bit to count a little
1750 * bit higher via unsigned math, and thus still get an accurate result.
1752 return ((unsigned int)folio_ref_count(folio
)) >=
1753 GUP_PIN_COUNTING_BIAS
;
1756 static inline bool page_maybe_dma_pinned(struct page
*page
)
1758 return folio_maybe_dma_pinned(page_folio(page
));
1762 * This should most likely only be called during fork() to see whether we
1763 * should break the cow immediately for an anon page on the src mm.
1765 * The caller has to hold the PT lock and the vma->vm_mm->->write_protect_seq.
1767 static inline bool page_needs_cow_for_dma(struct vm_area_struct
*vma
,
1770 VM_BUG_ON(!(raw_read_seqcount(&vma
->vm_mm
->write_protect_seq
) & 1));
1772 if (!test_bit(MMF_HAS_PINNED
, &vma
->vm_mm
->flags
))
1775 return page_maybe_dma_pinned(page
);
1778 /* MIGRATE_CMA and ZONE_MOVABLE do not allow pin pages */
1779 #ifdef CONFIG_MIGRATION
1780 static inline bool is_longterm_pinnable_page(struct page
*page
)
1783 int mt
= get_pageblock_migratetype(page
);
1785 if (mt
== MIGRATE_CMA
|| mt
== MIGRATE_ISOLATE
)
1788 /* The zero page may always be pinned */
1789 if (is_zero_pfn(page_to_pfn(page
)))
1792 /* Coherent device memory must always allow eviction. */
1793 if (is_device_coherent_page(page
))
1796 /* Otherwise, non-movable zone pages can be pinned. */
1797 return !is_zone_movable_page(page
);
1800 static inline bool is_longterm_pinnable_page(struct page
*page
)
1806 static inline bool folio_is_longterm_pinnable(struct folio
*folio
)
1808 return is_longterm_pinnable_page(&folio
->page
);
1811 static inline void set_page_zone(struct page
*page
, enum zone_type zone
)
1813 page
->flags
&= ~(ZONES_MASK
<< ZONES_PGSHIFT
);
1814 page
->flags
|= (zone
& ZONES_MASK
) << ZONES_PGSHIFT
;
1817 static inline void set_page_node(struct page
*page
, unsigned long node
)
1819 page
->flags
&= ~(NODES_MASK
<< NODES_PGSHIFT
);
1820 page
->flags
|= (node
& NODES_MASK
) << NODES_PGSHIFT
;
1823 static inline void set_page_links(struct page
*page
, enum zone_type zone
,
1824 unsigned long node
, unsigned long pfn
)
1826 set_page_zone(page
, zone
);
1827 set_page_node(page
, node
);
1828 #ifdef SECTION_IN_PAGE_FLAGS
1829 set_page_section(page
, pfn_to_section_nr(pfn
));
1834 * folio_nr_pages - The number of pages in the folio.
1835 * @folio: The folio.
1837 * Return: A positive power of two.
1839 static inline long folio_nr_pages(struct folio
*folio
)
1841 if (!folio_test_large(folio
))
1844 return folio
->_folio_nr_pages
;
1846 return 1L << folio
->_folio_order
;
1851 * compound_nr() returns the number of pages in this potentially compound
1852 * page. compound_nr() can be called on a tail page, and is defined to
1853 * return 1 in that case.
1855 static inline unsigned long compound_nr(struct page
*page
)
1857 struct folio
*folio
= (struct folio
*)page
;
1859 if (!test_bit(PG_head
, &folio
->flags
))
1862 return folio
->_folio_nr_pages
;
1864 return 1L << folio
->_folio_order
;
1869 * thp_nr_pages - The number of regular pages in this huge page.
1870 * @page: The head page of a huge page.
1872 static inline int thp_nr_pages(struct page
*page
)
1874 return folio_nr_pages((struct folio
*)page
);
1878 * folio_next - Move to the next physical folio.
1879 * @folio: The folio we're currently operating on.
1881 * If you have physically contiguous memory which may span more than
1882 * one folio (eg a &struct bio_vec), use this function to move from one
1883 * folio to the next. Do not use it if the memory is only virtually
1884 * contiguous as the folios are almost certainly not adjacent to each
1885 * other. This is the folio equivalent to writing ``page++``.
1887 * Context: We assume that the folios are refcounted and/or locked at a
1888 * higher level and do not adjust the reference counts.
1889 * Return: The next struct folio.
1891 static inline struct folio
*folio_next(struct folio
*folio
)
1893 return (struct folio
*)folio_page(folio
, folio_nr_pages(folio
));
1897 * folio_shift - The size of the memory described by this folio.
1898 * @folio: The folio.
1900 * A folio represents a number of bytes which is a power-of-two in size.
1901 * This function tells you which power-of-two the folio is. See also
1902 * folio_size() and folio_order().
1904 * Context: The caller should have a reference on the folio to prevent
1905 * it from being split. It is not necessary for the folio to be locked.
1906 * Return: The base-2 logarithm of the size of this folio.
1908 static inline unsigned int folio_shift(struct folio
*folio
)
1910 return PAGE_SHIFT
+ folio_order(folio
);
1914 * folio_size - The number of bytes in a folio.
1915 * @folio: The folio.
1917 * Context: The caller should have a reference on the folio to prevent
1918 * it from being split. It is not necessary for the folio to be locked.
1919 * Return: The number of bytes in this folio.
1921 static inline size_t folio_size(struct folio
*folio
)
1923 return PAGE_SIZE
<< folio_order(folio
);
1927 * folio_estimated_sharers - Estimate the number of sharers of a folio.
1928 * @folio: The folio.
1930 * folio_estimated_sharers() aims to serve as a function to efficiently
1931 * estimate the number of processes sharing a folio. This is done by
1932 * looking at the precise mapcount of the first subpage in the folio, and
1933 * assuming the other subpages are the same. This may not be true for large
1934 * folios. If you want exact mapcounts for exact calculations, look at
1935 * page_mapcount() or folio_total_mapcount().
1937 * Return: The estimated number of processes sharing a folio.
1939 static inline int folio_estimated_sharers(struct folio
*folio
)
1941 return page_mapcount(folio_page(folio
, 0));
1944 #ifndef HAVE_ARCH_MAKE_PAGE_ACCESSIBLE
1945 static inline int arch_make_page_accessible(struct page
*page
)
1951 #ifndef HAVE_ARCH_MAKE_FOLIO_ACCESSIBLE
1952 static inline int arch_make_folio_accessible(struct folio
*folio
)
1955 long i
, nr
= folio_nr_pages(folio
);
1957 for (i
= 0; i
< nr
; i
++) {
1958 ret
= arch_make_page_accessible(folio_page(folio
, i
));
1968 * Some inline functions in vmstat.h depend on page_zone()
1970 #include <linux/vmstat.h>
1972 static __always_inline
void *lowmem_page_address(const struct page
*page
)
1974 return page_to_virt(page
);
1977 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1978 #define HASHED_PAGE_VIRTUAL
1981 #if defined(WANT_PAGE_VIRTUAL)
1982 static inline void *page_address(const struct page
*page
)
1984 return page
->virtual;
1986 static inline void set_page_address(struct page
*page
, void *address
)
1988 page
->virtual = address
;
1990 #define page_address_init() do { } while(0)
1993 #if defined(HASHED_PAGE_VIRTUAL)
1994 void *page_address(const struct page
*page
);
1995 void set_page_address(struct page
*page
, void *virtual);
1996 void page_address_init(void);
1999 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
2000 #define page_address(page) lowmem_page_address(page)
2001 #define set_page_address(page, address) do { } while(0)
2002 #define page_address_init() do { } while(0)
2005 static inline void *folio_address(const struct folio
*folio
)
2007 return page_address(&folio
->page
);
2010 extern void *page_rmapping(struct page
*page
);
2011 extern pgoff_t
__page_file_index(struct page
*page
);
2014 * Return the pagecache index of the passed page. Regular pagecache pages
2015 * use ->index whereas swapcache pages use swp_offset(->private)
2017 static inline pgoff_t
page_index(struct page
*page
)
2019 if (unlikely(PageSwapCache(page
)))
2020 return __page_file_index(page
);
2025 * Return true only if the page has been allocated with
2026 * ALLOC_NO_WATERMARKS and the low watermark was not
2027 * met implying that the system is under some pressure.
2029 static inline bool page_is_pfmemalloc(const struct page
*page
)
2032 * lru.next has bit 1 set if the page is allocated from the
2033 * pfmemalloc reserves. Callers may simply overwrite it if
2034 * they do not need to preserve that information.
2036 return (uintptr_t)page
->lru
.next
& BIT(1);
2040 * Return true only if the folio has been allocated with
2041 * ALLOC_NO_WATERMARKS and the low watermark was not
2042 * met implying that the system is under some pressure.
2044 static inline bool folio_is_pfmemalloc(const struct folio
*folio
)
2047 * lru.next has bit 1 set if the page is allocated from the
2048 * pfmemalloc reserves. Callers may simply overwrite it if
2049 * they do not need to preserve that information.
2051 return (uintptr_t)folio
->lru
.next
& BIT(1);
2055 * Only to be called by the page allocator on a freshly allocated
2058 static inline void set_page_pfmemalloc(struct page
*page
)
2060 page
->lru
.next
= (void *)BIT(1);
2063 static inline void clear_page_pfmemalloc(struct page
*page
)
2065 page
->lru
.next
= NULL
;
2069 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
2071 extern void pagefault_out_of_memory(void);
2073 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
2074 #define offset_in_thp(page, p) ((unsigned long)(p) & (thp_size(page) - 1))
2075 #define offset_in_folio(folio, p) ((unsigned long)(p) & (folio_size(folio) - 1))
2078 * Flags passed to show_mem() and show_free_areas() to suppress output in
2081 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
2083 extern void __show_free_areas(unsigned int flags
, nodemask_t
*nodemask
, int max_zone_idx
);
2084 static void __maybe_unused
show_free_areas(unsigned int flags
, nodemask_t
*nodemask
)
2086 __show_free_areas(flags
, nodemask
, MAX_NR_ZONES
- 1);
2090 * Parameter block passed down to zap_pte_range in exceptional cases.
2092 struct zap_details
{
2093 struct folio
*single_folio
; /* Locked folio to be unmapped */
2094 bool even_cows
; /* Zap COWed private pages too? */
2095 zap_flags_t zap_flags
; /* Extra flags for zapping */
2099 * Whether to drop the pte markers, for example, the uffd-wp information for
2100 * file-backed memory. This should only be specified when we will completely
2101 * drop the page in the mm, either by truncation or unmapping of the vma. By
2102 * default, the flag is not set.
2104 #define ZAP_FLAG_DROP_MARKER ((__force zap_flags_t) BIT(0))
2105 /* Set in unmap_vmas() to indicate a final unmap call. Only used by hugetlb */
2106 #define ZAP_FLAG_UNMAP ((__force zap_flags_t) BIT(1))
2108 #ifdef CONFIG_SCHED_MM_CID
2109 void sched_mm_cid_before_execve(struct task_struct
*t
);
2110 void sched_mm_cid_after_execve(struct task_struct
*t
);
2111 void sched_mm_cid_fork(struct task_struct
*t
);
2112 void sched_mm_cid_exit_signals(struct task_struct
*t
);
2113 static inline int task_mm_cid(struct task_struct
*t
)
2118 static inline void sched_mm_cid_before_execve(struct task_struct
*t
) { }
2119 static inline void sched_mm_cid_after_execve(struct task_struct
*t
) { }
2120 static inline void sched_mm_cid_fork(struct task_struct
*t
) { }
2121 static inline void sched_mm_cid_exit_signals(struct task_struct
*t
) { }
2122 static inline int task_mm_cid(struct task_struct
*t
)
2125 * Use the processor id as a fall-back when the mm cid feature is
2126 * disabled. This provides functional per-cpu data structure accesses
2127 * in user-space, althrough it won't provide the memory usage benefits.
2129 return raw_smp_processor_id();
2134 extern bool can_do_mlock(void);
2136 static inline bool can_do_mlock(void) { return false; }
2138 extern int user_shm_lock(size_t, struct ucounts
*);
2139 extern void user_shm_unlock(size_t, struct ucounts
*);
2141 struct folio
*vm_normal_folio(struct vm_area_struct
*vma
, unsigned long addr
,
2143 struct page
*vm_normal_page(struct vm_area_struct
*vma
, unsigned long addr
,
2145 struct page
*vm_normal_page_pmd(struct vm_area_struct
*vma
, unsigned long addr
,
2148 void zap_vma_ptes(struct vm_area_struct
*vma
, unsigned long address
,
2149 unsigned long size
);
2150 void zap_page_range_single(struct vm_area_struct
*vma
, unsigned long address
,
2151 unsigned long size
, struct zap_details
*details
);
2152 static inline void zap_vma_pages(struct vm_area_struct
*vma
)
2154 zap_page_range_single(vma
, vma
->vm_start
,
2155 vma
->vm_end
- vma
->vm_start
, NULL
);
2157 void unmap_vmas(struct mmu_gather
*tlb
, struct maple_tree
*mt
,
2158 struct vm_area_struct
*start_vma
, unsigned long start
,
2159 unsigned long end
, bool mm_wr_locked
);
2161 struct mmu_notifier_range
;
2163 void free_pgd_range(struct mmu_gather
*tlb
, unsigned long addr
,
2164 unsigned long end
, unsigned long floor
, unsigned long ceiling
);
2166 copy_page_range(struct vm_area_struct
*dst_vma
, struct vm_area_struct
*src_vma
);
2167 int follow_pte(struct mm_struct
*mm
, unsigned long address
,
2168 pte_t
**ptepp
, spinlock_t
**ptlp
);
2169 int follow_pfn(struct vm_area_struct
*vma
, unsigned long address
,
2170 unsigned long *pfn
);
2171 int follow_phys(struct vm_area_struct
*vma
, unsigned long address
,
2172 unsigned int flags
, unsigned long *prot
, resource_size_t
*phys
);
2173 int generic_access_phys(struct vm_area_struct
*vma
, unsigned long addr
,
2174 void *buf
, int len
, int write
);
2176 extern void truncate_pagecache(struct inode
*inode
, loff_t
new);
2177 extern void truncate_setsize(struct inode
*inode
, loff_t newsize
);
2178 void pagecache_isize_extended(struct inode
*inode
, loff_t from
, loff_t to
);
2179 void truncate_pagecache_range(struct inode
*inode
, loff_t offset
, loff_t end
);
2180 int generic_error_remove_page(struct address_space
*mapping
, struct page
*page
);
2183 extern vm_fault_t
handle_mm_fault(struct vm_area_struct
*vma
,
2184 unsigned long address
, unsigned int flags
,
2185 struct pt_regs
*regs
);
2186 extern int fixup_user_fault(struct mm_struct
*mm
,
2187 unsigned long address
, unsigned int fault_flags
,
2189 void unmap_mapping_pages(struct address_space
*mapping
,
2190 pgoff_t start
, pgoff_t nr
, bool even_cows
);
2191 void unmap_mapping_range(struct address_space
*mapping
,
2192 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
);
2194 static inline vm_fault_t
handle_mm_fault(struct vm_area_struct
*vma
,
2195 unsigned long address
, unsigned int flags
,
2196 struct pt_regs
*regs
)
2198 /* should never happen if there's no MMU */
2200 return VM_FAULT_SIGBUS
;
2202 static inline int fixup_user_fault(struct mm_struct
*mm
, unsigned long address
,
2203 unsigned int fault_flags
, bool *unlocked
)
2205 /* should never happen if there's no MMU */
2209 static inline void unmap_mapping_pages(struct address_space
*mapping
,
2210 pgoff_t start
, pgoff_t nr
, bool even_cows
) { }
2211 static inline void unmap_mapping_range(struct address_space
*mapping
,
2212 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
) { }
2215 static inline void unmap_shared_mapping_range(struct address_space
*mapping
,
2216 loff_t
const holebegin
, loff_t
const holelen
)
2218 unmap_mapping_range(mapping
, holebegin
, holelen
, 0);
2221 extern int access_process_vm(struct task_struct
*tsk
, unsigned long addr
,
2222 void *buf
, int len
, unsigned int gup_flags
);
2223 extern int access_remote_vm(struct mm_struct
*mm
, unsigned long addr
,
2224 void *buf
, int len
, unsigned int gup_flags
);
2225 extern int __access_remote_vm(struct mm_struct
*mm
, unsigned long addr
,
2226 void *buf
, int len
, unsigned int gup_flags
);
2228 long get_user_pages_remote(struct mm_struct
*mm
,
2229 unsigned long start
, unsigned long nr_pages
,
2230 unsigned int gup_flags
, struct page
**pages
,
2231 struct vm_area_struct
**vmas
, int *locked
);
2232 long pin_user_pages_remote(struct mm_struct
*mm
,
2233 unsigned long start
, unsigned long nr_pages
,
2234 unsigned int gup_flags
, struct page
**pages
,
2235 struct vm_area_struct
**vmas
, int *locked
);
2236 long get_user_pages(unsigned long start
, unsigned long nr_pages
,
2237 unsigned int gup_flags
, struct page
**pages
,
2238 struct vm_area_struct
**vmas
);
2239 long pin_user_pages(unsigned long start
, unsigned long nr_pages
,
2240 unsigned int gup_flags
, struct page
**pages
,
2241 struct vm_area_struct
**vmas
);
2242 long get_user_pages_unlocked(unsigned long start
, unsigned long nr_pages
,
2243 struct page
**pages
, unsigned int gup_flags
);
2244 long pin_user_pages_unlocked(unsigned long start
, unsigned long nr_pages
,
2245 struct page
**pages
, unsigned int gup_flags
);
2247 int get_user_pages_fast(unsigned long start
, int nr_pages
,
2248 unsigned int gup_flags
, struct page
**pages
);
2249 int pin_user_pages_fast(unsigned long start
, int nr_pages
,
2250 unsigned int gup_flags
, struct page
**pages
);
2252 int account_locked_vm(struct mm_struct
*mm
, unsigned long pages
, bool inc
);
2253 int __account_locked_vm(struct mm_struct
*mm
, unsigned long pages
, bool inc
,
2254 struct task_struct
*task
, bool bypass_rlim
);
2257 struct page
*get_dump_page(unsigned long addr
);
2259 bool folio_mark_dirty(struct folio
*folio
);
2260 bool set_page_dirty(struct page
*page
);
2261 int set_page_dirty_lock(struct page
*page
);
2263 int get_cmdline(struct task_struct
*task
, char *buffer
, int buflen
);
2265 extern unsigned long move_page_tables(struct vm_area_struct
*vma
,
2266 unsigned long old_addr
, struct vm_area_struct
*new_vma
,
2267 unsigned long new_addr
, unsigned long len
,
2268 bool need_rmap_locks
);
2271 * Flags used by change_protection(). For now we make it a bitmap so
2272 * that we can pass in multiple flags just like parameters. However
2273 * for now all the callers are only use one of the flags at the same
2277 * Whether we should manually check if we can map individual PTEs writable,
2278 * because something (e.g., COW, uffd-wp) blocks that from happening for all
2279 * PTEs automatically in a writable mapping.
2281 #define MM_CP_TRY_CHANGE_WRITABLE (1UL << 0)
2282 /* Whether this protection change is for NUMA hints */
2283 #define MM_CP_PROT_NUMA (1UL << 1)
2284 /* Whether this change is for write protecting */
2285 #define MM_CP_UFFD_WP (1UL << 2) /* do wp */
2286 #define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */
2287 #define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \
2288 MM_CP_UFFD_WP_RESOLVE)
2290 int vma_wants_writenotify(struct vm_area_struct
*vma
, pgprot_t vm_page_prot
);
2291 static inline bool vma_wants_manual_pte_write_upgrade(struct vm_area_struct
*vma
)
2294 * We want to check manually if we can change individual PTEs writable
2295 * if we can't do that automatically for all PTEs in a mapping. For
2296 * private mappings, that's always the case when we have write
2297 * permissions as we properly have to handle COW.
2299 if (vma
->vm_flags
& VM_SHARED
)
2300 return vma_wants_writenotify(vma
, vma
->vm_page_prot
);
2301 return !!(vma
->vm_flags
& VM_WRITE
);
2304 bool can_change_pte_writable(struct vm_area_struct
*vma
, unsigned long addr
,
2306 extern long change_protection(struct mmu_gather
*tlb
,
2307 struct vm_area_struct
*vma
, unsigned long start
,
2308 unsigned long end
, unsigned long cp_flags
);
2309 extern int mprotect_fixup(struct vma_iterator
*vmi
, struct mmu_gather
*tlb
,
2310 struct vm_area_struct
*vma
, struct vm_area_struct
**pprev
,
2311 unsigned long start
, unsigned long end
, unsigned long newflags
);
2314 * doesn't attempt to fault and will return short.
2316 int get_user_pages_fast_only(unsigned long start
, int nr_pages
,
2317 unsigned int gup_flags
, struct page
**pages
);
2319 static inline bool get_user_page_fast_only(unsigned long addr
,
2320 unsigned int gup_flags
, struct page
**pagep
)
2322 return get_user_pages_fast_only(addr
, 1, gup_flags
, pagep
) == 1;
2325 * per-process(per-mm_struct) statistics.
2327 static inline unsigned long get_mm_counter(struct mm_struct
*mm
, int member
)
2329 return percpu_counter_read_positive(&mm
->rss_stat
[member
]);
2332 void mm_trace_rss_stat(struct mm_struct
*mm
, int member
);
2334 static inline void add_mm_counter(struct mm_struct
*mm
, int member
, long value
)
2336 percpu_counter_add(&mm
->rss_stat
[member
], value
);
2338 mm_trace_rss_stat(mm
, member
);
2341 static inline void inc_mm_counter(struct mm_struct
*mm
, int member
)
2343 percpu_counter_inc(&mm
->rss_stat
[member
]);
2345 mm_trace_rss_stat(mm
, member
);
2348 static inline void dec_mm_counter(struct mm_struct
*mm
, int member
)
2350 percpu_counter_dec(&mm
->rss_stat
[member
]);
2352 mm_trace_rss_stat(mm
, member
);
2355 /* Optimized variant when page is already known not to be PageAnon */
2356 static inline int mm_counter_file(struct page
*page
)
2358 if (PageSwapBacked(page
))
2359 return MM_SHMEMPAGES
;
2360 return MM_FILEPAGES
;
2363 static inline int mm_counter(struct page
*page
)
2366 return MM_ANONPAGES
;
2367 return mm_counter_file(page
);
2370 static inline unsigned long get_mm_rss(struct mm_struct
*mm
)
2372 return get_mm_counter(mm
, MM_FILEPAGES
) +
2373 get_mm_counter(mm
, MM_ANONPAGES
) +
2374 get_mm_counter(mm
, MM_SHMEMPAGES
);
2377 static inline unsigned long get_mm_hiwater_rss(struct mm_struct
*mm
)
2379 return max(mm
->hiwater_rss
, get_mm_rss(mm
));
2382 static inline unsigned long get_mm_hiwater_vm(struct mm_struct
*mm
)
2384 return max(mm
->hiwater_vm
, mm
->total_vm
);
2387 static inline void update_hiwater_rss(struct mm_struct
*mm
)
2389 unsigned long _rss
= get_mm_rss(mm
);
2391 if ((mm
)->hiwater_rss
< _rss
)
2392 (mm
)->hiwater_rss
= _rss
;
2395 static inline void update_hiwater_vm(struct mm_struct
*mm
)
2397 if (mm
->hiwater_vm
< mm
->total_vm
)
2398 mm
->hiwater_vm
= mm
->total_vm
;
2401 static inline void reset_mm_hiwater_rss(struct mm_struct
*mm
)
2403 mm
->hiwater_rss
= get_mm_rss(mm
);
2406 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss
,
2407 struct mm_struct
*mm
)
2409 unsigned long hiwater_rss
= get_mm_hiwater_rss(mm
);
2411 if (*maxrss
< hiwater_rss
)
2412 *maxrss
= hiwater_rss
;
2415 #if defined(SPLIT_RSS_COUNTING)
2416 void sync_mm_rss(struct mm_struct
*mm
);
2418 static inline void sync_mm_rss(struct mm_struct
*mm
)
2423 #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
2424 static inline int pte_special(pte_t pte
)
2429 static inline pte_t
pte_mkspecial(pte_t pte
)
2435 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
2436 static inline int pte_devmap(pte_t pte
)
2442 extern pte_t
*__get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
2444 static inline pte_t
*get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
2448 __cond_lock(*ptl
, ptep
= __get_locked_pte(mm
, addr
, ptl
));
2452 #ifdef __PAGETABLE_P4D_FOLDED
2453 static inline int __p4d_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
2454 unsigned long address
)
2459 int __p4d_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
);
2462 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
2463 static inline int __pud_alloc(struct mm_struct
*mm
, p4d_t
*p4d
,
2464 unsigned long address
)
2468 static inline void mm_inc_nr_puds(struct mm_struct
*mm
) {}
2469 static inline void mm_dec_nr_puds(struct mm_struct
*mm
) {}
2472 int __pud_alloc(struct mm_struct
*mm
, p4d_t
*p4d
, unsigned long address
);
2474 static inline void mm_inc_nr_puds(struct mm_struct
*mm
)
2476 if (mm_pud_folded(mm
))
2478 atomic_long_add(PTRS_PER_PUD
* sizeof(pud_t
), &mm
->pgtables_bytes
);
2481 static inline void mm_dec_nr_puds(struct mm_struct
*mm
)
2483 if (mm_pud_folded(mm
))
2485 atomic_long_sub(PTRS_PER_PUD
* sizeof(pud_t
), &mm
->pgtables_bytes
);
2489 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
2490 static inline int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
,
2491 unsigned long address
)
2496 static inline void mm_inc_nr_pmds(struct mm_struct
*mm
) {}
2497 static inline void mm_dec_nr_pmds(struct mm_struct
*mm
) {}
2500 int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
);
2502 static inline void mm_inc_nr_pmds(struct mm_struct
*mm
)
2504 if (mm_pmd_folded(mm
))
2506 atomic_long_add(PTRS_PER_PMD
* sizeof(pmd_t
), &mm
->pgtables_bytes
);
2509 static inline void mm_dec_nr_pmds(struct mm_struct
*mm
)
2511 if (mm_pmd_folded(mm
))
2513 atomic_long_sub(PTRS_PER_PMD
* sizeof(pmd_t
), &mm
->pgtables_bytes
);
2518 static inline void mm_pgtables_bytes_init(struct mm_struct
*mm
)
2520 atomic_long_set(&mm
->pgtables_bytes
, 0);
2523 static inline unsigned long mm_pgtables_bytes(const struct mm_struct
*mm
)
2525 return atomic_long_read(&mm
->pgtables_bytes
);
2528 static inline void mm_inc_nr_ptes(struct mm_struct
*mm
)
2530 atomic_long_add(PTRS_PER_PTE
* sizeof(pte_t
), &mm
->pgtables_bytes
);
2533 static inline void mm_dec_nr_ptes(struct mm_struct
*mm
)
2535 atomic_long_sub(PTRS_PER_PTE
* sizeof(pte_t
), &mm
->pgtables_bytes
);
2539 static inline void mm_pgtables_bytes_init(struct mm_struct
*mm
) {}
2540 static inline unsigned long mm_pgtables_bytes(const struct mm_struct
*mm
)
2545 static inline void mm_inc_nr_ptes(struct mm_struct
*mm
) {}
2546 static inline void mm_dec_nr_ptes(struct mm_struct
*mm
) {}
2549 int __pte_alloc(struct mm_struct
*mm
, pmd_t
*pmd
);
2550 int __pte_alloc_kernel(pmd_t
*pmd
);
2552 #if defined(CONFIG_MMU)
2554 static inline p4d_t
*p4d_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
2555 unsigned long address
)
2557 return (unlikely(pgd_none(*pgd
)) && __p4d_alloc(mm
, pgd
, address
)) ?
2558 NULL
: p4d_offset(pgd
, address
);
2561 static inline pud_t
*pud_alloc(struct mm_struct
*mm
, p4d_t
*p4d
,
2562 unsigned long address
)
2564 return (unlikely(p4d_none(*p4d
)) && __pud_alloc(mm
, p4d
, address
)) ?
2565 NULL
: pud_offset(p4d
, address
);
2568 static inline pmd_t
*pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
)
2570 return (unlikely(pud_none(*pud
)) && __pmd_alloc(mm
, pud
, address
))?
2571 NULL
: pmd_offset(pud
, address
);
2573 #endif /* CONFIG_MMU */
2575 #if USE_SPLIT_PTE_PTLOCKS
2576 #if ALLOC_SPLIT_PTLOCKS
2577 void __init
ptlock_cache_init(void);
2578 extern bool ptlock_alloc(struct page
*page
);
2579 extern void ptlock_free(struct page
*page
);
2581 static inline spinlock_t
*ptlock_ptr(struct page
*page
)
2585 #else /* ALLOC_SPLIT_PTLOCKS */
2586 static inline void ptlock_cache_init(void)
2590 static inline bool ptlock_alloc(struct page
*page
)
2595 static inline void ptlock_free(struct page
*page
)
2599 static inline spinlock_t
*ptlock_ptr(struct page
*page
)
2603 #endif /* ALLOC_SPLIT_PTLOCKS */
2605 static inline spinlock_t
*pte_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
2607 return ptlock_ptr(pmd_page(*pmd
));
2610 static inline bool ptlock_init(struct page
*page
)
2613 * prep_new_page() initialize page->private (and therefore page->ptl)
2614 * with 0. Make sure nobody took it in use in between.
2616 * It can happen if arch try to use slab for page table allocation:
2617 * slab code uses page->slab_cache, which share storage with page->ptl.
2619 VM_BUG_ON_PAGE(*(unsigned long *)&page
->ptl
, page
);
2620 if (!ptlock_alloc(page
))
2622 spin_lock_init(ptlock_ptr(page
));
2626 #else /* !USE_SPLIT_PTE_PTLOCKS */
2628 * We use mm->page_table_lock to guard all pagetable pages of the mm.
2630 static inline spinlock_t
*pte_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
2632 return &mm
->page_table_lock
;
2634 static inline void ptlock_cache_init(void) {}
2635 static inline bool ptlock_init(struct page
*page
) { return true; }
2636 static inline void ptlock_free(struct page
*page
) {}
2637 #endif /* USE_SPLIT_PTE_PTLOCKS */
2639 static inline void pgtable_init(void)
2641 ptlock_cache_init();
2642 pgtable_cache_init();
2645 static inline bool pgtable_pte_page_ctor(struct page
*page
)
2647 if (!ptlock_init(page
))
2649 __SetPageTable(page
);
2650 inc_lruvec_page_state(page
, NR_PAGETABLE
);
2654 static inline void pgtable_pte_page_dtor(struct page
*page
)
2657 __ClearPageTable(page
);
2658 dec_lruvec_page_state(page
, NR_PAGETABLE
);
2661 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
2663 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
2664 pte_t *__pte = pte_offset_map(pmd, address); \
2670 #define pte_unmap_unlock(pte, ptl) do { \
2675 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2677 #define pte_alloc_map(mm, pmd, address) \
2678 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2680 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2681 (pte_alloc(mm, pmd) ? \
2682 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2684 #define pte_alloc_kernel(pmd, address) \
2685 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2686 NULL: pte_offset_kernel(pmd, address))
2688 #if USE_SPLIT_PMD_PTLOCKS
2690 static inline struct page
*pmd_pgtable_page(pmd_t
*pmd
)
2692 unsigned long mask
= ~(PTRS_PER_PMD
* sizeof(pmd_t
) - 1);
2693 return virt_to_page((void *)((unsigned long) pmd
& mask
));
2696 static inline spinlock_t
*pmd_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
2698 return ptlock_ptr(pmd_pgtable_page(pmd
));
2701 static inline bool pmd_ptlock_init(struct page
*page
)
2703 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2704 page
->pmd_huge_pte
= NULL
;
2706 return ptlock_init(page
);
2709 static inline void pmd_ptlock_free(struct page
*page
)
2711 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2712 VM_BUG_ON_PAGE(page
->pmd_huge_pte
, page
);
2717 #define pmd_huge_pte(mm, pmd) (pmd_pgtable_page(pmd)->pmd_huge_pte)
2721 static inline spinlock_t
*pmd_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
2723 return &mm
->page_table_lock
;
2726 static inline bool pmd_ptlock_init(struct page
*page
) { return true; }
2727 static inline void pmd_ptlock_free(struct page
*page
) {}
2729 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2733 static inline spinlock_t
*pmd_lock(struct mm_struct
*mm
, pmd_t
*pmd
)
2735 spinlock_t
*ptl
= pmd_lockptr(mm
, pmd
);
2740 static inline bool pgtable_pmd_page_ctor(struct page
*page
)
2742 if (!pmd_ptlock_init(page
))
2744 __SetPageTable(page
);
2745 inc_lruvec_page_state(page
, NR_PAGETABLE
);
2749 static inline void pgtable_pmd_page_dtor(struct page
*page
)
2751 pmd_ptlock_free(page
);
2752 __ClearPageTable(page
);
2753 dec_lruvec_page_state(page
, NR_PAGETABLE
);
2757 * No scalability reason to split PUD locks yet, but follow the same pattern
2758 * as the PMD locks to make it easier if we decide to. The VM should not be
2759 * considered ready to switch to split PUD locks yet; there may be places
2760 * which need to be converted from page_table_lock.
2762 static inline spinlock_t
*pud_lockptr(struct mm_struct
*mm
, pud_t
*pud
)
2764 return &mm
->page_table_lock
;
2767 static inline spinlock_t
*pud_lock(struct mm_struct
*mm
, pud_t
*pud
)
2769 spinlock_t
*ptl
= pud_lockptr(mm
, pud
);
2775 extern void __init
pagecache_init(void);
2776 extern void free_initmem(void);
2779 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2780 * into the buddy system. The freed pages will be poisoned with pattern
2781 * "poison" if it's within range [0, UCHAR_MAX].
2782 * Return pages freed into the buddy system.
2784 extern unsigned long free_reserved_area(void *start
, void *end
,
2785 int poison
, const char *s
);
2787 extern void adjust_managed_page_count(struct page
*page
, long count
);
2788 extern void mem_init_print_info(void);
2790 extern void reserve_bootmem_region(phys_addr_t start
, phys_addr_t end
);
2792 /* Free the reserved page into the buddy system, so it gets managed. */
2793 static inline void free_reserved_page(struct page
*page
)
2795 ClearPageReserved(page
);
2796 init_page_count(page
);
2798 adjust_managed_page_count(page
, 1);
2800 #define free_highmem_page(page) free_reserved_page(page)
2802 static inline void mark_page_reserved(struct page
*page
)
2804 SetPageReserved(page
);
2805 adjust_managed_page_count(page
, -1);
2809 * Default method to free all the __init memory into the buddy system.
2810 * The freed pages will be poisoned with pattern "poison" if it's within
2811 * range [0, UCHAR_MAX].
2812 * Return pages freed into the buddy system.
2814 static inline unsigned long free_initmem_default(int poison
)
2816 extern char __init_begin
[], __init_end
[];
2818 return free_reserved_area(&__init_begin
, &__init_end
,
2819 poison
, "unused kernel image (initmem)");
2822 static inline unsigned long get_num_physpages(void)
2825 unsigned long phys_pages
= 0;
2827 for_each_online_node(nid
)
2828 phys_pages
+= node_present_pages(nid
);
2834 * Using memblock node mappings, an architecture may initialise its
2835 * zones, allocate the backing mem_map and account for memory holes in an
2836 * architecture independent manner.
2838 * An architecture is expected to register range of page frames backed by
2839 * physical memory with memblock_add[_node]() before calling
2840 * free_area_init() passing in the PFN each zone ends at. At a basic
2841 * usage, an architecture is expected to do something like
2843 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2845 * for_each_valid_physical_page_range()
2846 * memblock_add_node(base, size, nid, MEMBLOCK_NONE)
2847 * free_area_init(max_zone_pfns);
2849 void free_area_init(unsigned long *max_zone_pfn
);
2850 unsigned long node_map_pfn_alignment(void);
2851 unsigned long __absent_pages_in_range(int nid
, unsigned long start_pfn
,
2852 unsigned long end_pfn
);
2853 extern unsigned long absent_pages_in_range(unsigned long start_pfn
,
2854 unsigned long end_pfn
);
2855 extern void get_pfn_range_for_nid(unsigned int nid
,
2856 unsigned long *start_pfn
, unsigned long *end_pfn
);
2859 static inline int early_pfn_to_nid(unsigned long pfn
)
2864 /* please see mm/page_alloc.c */
2865 extern int __meminit
early_pfn_to_nid(unsigned long pfn
);
2868 extern void set_dma_reserve(unsigned long new_dma_reserve
);
2869 extern void memmap_init_range(unsigned long, int, unsigned long,
2870 unsigned long, unsigned long, enum meminit_context
,
2871 struct vmem_altmap
*, int migratetype
);
2872 extern void setup_per_zone_wmarks(void);
2873 extern void calculate_min_free_kbytes(void);
2874 extern int __meminit
init_per_zone_wmark_min(void);
2875 extern void mem_init(void);
2876 extern void __init
mmap_init(void);
2878 extern void __show_mem(unsigned int flags
, nodemask_t
*nodemask
, int max_zone_idx
);
2879 static inline void show_mem(unsigned int flags
, nodemask_t
*nodemask
)
2881 __show_mem(flags
, nodemask
, MAX_NR_ZONES
- 1);
2883 extern long si_mem_available(void);
2884 extern void si_meminfo(struct sysinfo
* val
);
2885 extern void si_meminfo_node(struct sysinfo
*val
, int nid
);
2886 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2887 extern unsigned long arch_reserved_kernel_pages(void);
2890 extern __printf(3, 4)
2891 void warn_alloc(gfp_t gfp_mask
, nodemask_t
*nodemask
, const char *fmt
, ...);
2893 extern void setup_per_cpu_pageset(void);
2896 extern int min_free_kbytes
;
2897 extern int watermark_boost_factor
;
2898 extern int watermark_scale_factor
;
2899 extern bool arch_has_descending_max_zone_pfns(void);
2902 extern atomic_long_t mmap_pages_allocated
;
2903 extern int nommu_shrink_inode_mappings(struct inode
*, size_t, size_t);
2905 /* interval_tree.c */
2906 void vma_interval_tree_insert(struct vm_area_struct
*node
,
2907 struct rb_root_cached
*root
);
2908 void vma_interval_tree_insert_after(struct vm_area_struct
*node
,
2909 struct vm_area_struct
*prev
,
2910 struct rb_root_cached
*root
);
2911 void vma_interval_tree_remove(struct vm_area_struct
*node
,
2912 struct rb_root_cached
*root
);
2913 struct vm_area_struct
*vma_interval_tree_iter_first(struct rb_root_cached
*root
,
2914 unsigned long start
, unsigned long last
);
2915 struct vm_area_struct
*vma_interval_tree_iter_next(struct vm_area_struct
*node
,
2916 unsigned long start
, unsigned long last
);
2918 #define vma_interval_tree_foreach(vma, root, start, last) \
2919 for (vma = vma_interval_tree_iter_first(root, start, last); \
2920 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2922 void anon_vma_interval_tree_insert(struct anon_vma_chain
*node
,
2923 struct rb_root_cached
*root
);
2924 void anon_vma_interval_tree_remove(struct anon_vma_chain
*node
,
2925 struct rb_root_cached
*root
);
2926 struct anon_vma_chain
*
2927 anon_vma_interval_tree_iter_first(struct rb_root_cached
*root
,
2928 unsigned long start
, unsigned long last
);
2929 struct anon_vma_chain
*anon_vma_interval_tree_iter_next(
2930 struct anon_vma_chain
*node
, unsigned long start
, unsigned long last
);
2931 #ifdef CONFIG_DEBUG_VM_RB
2932 void anon_vma_interval_tree_verify(struct anon_vma_chain
*node
);
2935 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2936 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2937 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2940 extern int __vm_enough_memory(struct mm_struct
*mm
, long pages
, int cap_sys_admin
);
2941 extern int vma_expand(struct vma_iterator
*vmi
, struct vm_area_struct
*vma
,
2942 unsigned long start
, unsigned long end
, pgoff_t pgoff
,
2943 struct vm_area_struct
*next
);
2944 extern int vma_shrink(struct vma_iterator
*vmi
, struct vm_area_struct
*vma
,
2945 unsigned long start
, unsigned long end
, pgoff_t pgoff
);
2946 extern struct vm_area_struct
*vma_merge(struct vma_iterator
*vmi
,
2947 struct mm_struct
*, struct vm_area_struct
*prev
, unsigned long addr
,
2948 unsigned long end
, unsigned long vm_flags
, struct anon_vma
*,
2949 struct file
*, pgoff_t
, struct mempolicy
*, struct vm_userfaultfd_ctx
,
2950 struct anon_vma_name
*);
2951 extern struct anon_vma
*find_mergeable_anon_vma(struct vm_area_struct
*);
2952 extern int __split_vma(struct vma_iterator
*vmi
, struct vm_area_struct
*,
2953 unsigned long addr
, int new_below
);
2954 extern int split_vma(struct vma_iterator
*vmi
, struct vm_area_struct
*,
2955 unsigned long addr
, int new_below
);
2956 extern int insert_vm_struct(struct mm_struct
*, struct vm_area_struct
*);
2957 extern void unlink_file_vma(struct vm_area_struct
*);
2958 extern struct vm_area_struct
*copy_vma(struct vm_area_struct
**,
2959 unsigned long addr
, unsigned long len
, pgoff_t pgoff
,
2960 bool *need_rmap_locks
);
2961 extern void exit_mmap(struct mm_struct
*);
2963 static inline int check_data_rlimit(unsigned long rlim
,
2965 unsigned long start
,
2966 unsigned long end_data
,
2967 unsigned long start_data
)
2969 if (rlim
< RLIM_INFINITY
) {
2970 if (((new - start
) + (end_data
- start_data
)) > rlim
)
2977 extern int mm_take_all_locks(struct mm_struct
*mm
);
2978 extern void mm_drop_all_locks(struct mm_struct
*mm
);
2980 extern int set_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
);
2981 extern int replace_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
);
2982 extern struct file
*get_mm_exe_file(struct mm_struct
*mm
);
2983 extern struct file
*get_task_exe_file(struct task_struct
*task
);
2985 extern bool may_expand_vm(struct mm_struct
*, vm_flags_t
, unsigned long npages
);
2986 extern void vm_stat_account(struct mm_struct
*, vm_flags_t
, long npages
);
2988 extern bool vma_is_special_mapping(const struct vm_area_struct
*vma
,
2989 const struct vm_special_mapping
*sm
);
2990 extern struct vm_area_struct
*_install_special_mapping(struct mm_struct
*mm
,
2991 unsigned long addr
, unsigned long len
,
2992 unsigned long flags
,
2993 const struct vm_special_mapping
*spec
);
2994 /* This is an obsolete alternative to _install_special_mapping. */
2995 extern int install_special_mapping(struct mm_struct
*mm
,
2996 unsigned long addr
, unsigned long len
,
2997 unsigned long flags
, struct page
**pages
);
2999 unsigned long randomize_stack_top(unsigned long stack_top
);
3000 unsigned long randomize_page(unsigned long start
, unsigned long range
);
3002 extern unsigned long get_unmapped_area(struct file
*, unsigned long, unsigned long, unsigned long, unsigned long);
3004 extern unsigned long mmap_region(struct file
*file
, unsigned long addr
,
3005 unsigned long len
, vm_flags_t vm_flags
, unsigned long pgoff
,
3006 struct list_head
*uf
);
3007 extern unsigned long do_mmap(struct file
*file
, unsigned long addr
,
3008 unsigned long len
, unsigned long prot
, unsigned long flags
,
3009 unsigned long pgoff
, unsigned long *populate
, struct list_head
*uf
);
3010 extern int do_vmi_munmap(struct vma_iterator
*vmi
, struct mm_struct
*mm
,
3011 unsigned long start
, size_t len
, struct list_head
*uf
,
3013 extern int do_munmap(struct mm_struct
*, unsigned long, size_t,
3014 struct list_head
*uf
);
3015 extern int do_madvise(struct mm_struct
*mm
, unsigned long start
, size_t len_in
, int behavior
);
3018 extern int do_vma_munmap(struct vma_iterator
*vmi
, struct vm_area_struct
*vma
,
3019 unsigned long start
, unsigned long end
,
3020 struct list_head
*uf
, bool downgrade
);
3021 extern int __mm_populate(unsigned long addr
, unsigned long len
,
3023 static inline void mm_populate(unsigned long addr
, unsigned long len
)
3026 (void) __mm_populate(addr
, len
, 1);
3029 static inline void mm_populate(unsigned long addr
, unsigned long len
) {}
3032 /* These take the mm semaphore themselves */
3033 extern int __must_check
vm_brk(unsigned long, unsigned long);
3034 extern int __must_check
vm_brk_flags(unsigned long, unsigned long, unsigned long);
3035 extern int vm_munmap(unsigned long, size_t);
3036 extern unsigned long __must_check
vm_mmap(struct file
*, unsigned long,
3037 unsigned long, unsigned long,
3038 unsigned long, unsigned long);
3040 struct vm_unmapped_area_info
{
3041 #define VM_UNMAPPED_AREA_TOPDOWN 1
3042 unsigned long flags
;
3043 unsigned long length
;
3044 unsigned long low_limit
;
3045 unsigned long high_limit
;
3046 unsigned long align_mask
;
3047 unsigned long align_offset
;
3050 extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info
*info
);
3053 extern void truncate_inode_pages(struct address_space
*, loff_t
);
3054 extern void truncate_inode_pages_range(struct address_space
*,
3055 loff_t lstart
, loff_t lend
);
3056 extern void truncate_inode_pages_final(struct address_space
*);
3058 /* generic vm_area_ops exported for stackable file systems */
3059 extern vm_fault_t
filemap_fault(struct vm_fault
*vmf
);
3060 extern vm_fault_t
filemap_map_pages(struct vm_fault
*vmf
,
3061 pgoff_t start_pgoff
, pgoff_t end_pgoff
);
3062 extern vm_fault_t
filemap_page_mkwrite(struct vm_fault
*vmf
);
3064 extern unsigned long stack_guard_gap
;
3065 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
3066 extern int expand_stack(struct vm_area_struct
*vma
, unsigned long address
);
3068 /* CONFIG_STACK_GROWSUP still needs to grow downwards at some places */
3069 extern int expand_downwards(struct vm_area_struct
*vma
,
3070 unsigned long address
);
3072 extern int expand_upwards(struct vm_area_struct
*vma
, unsigned long address
);
3074 #define expand_upwards(vma, address) (0)
3077 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
3078 extern struct vm_area_struct
* find_vma(struct mm_struct
* mm
, unsigned long addr
);
3079 extern struct vm_area_struct
* find_vma_prev(struct mm_struct
* mm
, unsigned long addr
,
3080 struct vm_area_struct
**pprev
);
3083 * Look up the first VMA which intersects the interval [start_addr, end_addr)
3084 * NULL if none. Assume start_addr < end_addr.
3086 struct vm_area_struct
*find_vma_intersection(struct mm_struct
*mm
,
3087 unsigned long start_addr
, unsigned long end_addr
);
3090 * vma_lookup() - Find a VMA at a specific address
3091 * @mm: The process address space.
3092 * @addr: The user address.
3094 * Return: The vm_area_struct at the given address, %NULL otherwise.
3097 struct vm_area_struct
*vma_lookup(struct mm_struct
*mm
, unsigned long addr
)
3099 return mtree_load(&mm
->mm_mt
, addr
);
3102 static inline unsigned long vm_start_gap(struct vm_area_struct
*vma
)
3104 unsigned long vm_start
= vma
->vm_start
;
3106 if (vma
->vm_flags
& VM_GROWSDOWN
) {
3107 vm_start
-= stack_guard_gap
;
3108 if (vm_start
> vma
->vm_start
)
3114 static inline unsigned long vm_end_gap(struct vm_area_struct
*vma
)
3116 unsigned long vm_end
= vma
->vm_end
;
3118 if (vma
->vm_flags
& VM_GROWSUP
) {
3119 vm_end
+= stack_guard_gap
;
3120 if (vm_end
< vma
->vm_end
)
3121 vm_end
= -PAGE_SIZE
;
3126 static inline unsigned long vma_pages(struct vm_area_struct
*vma
)
3128 return (vma
->vm_end
- vma
->vm_start
) >> PAGE_SHIFT
;
3131 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
3132 static inline struct vm_area_struct
*find_exact_vma(struct mm_struct
*mm
,
3133 unsigned long vm_start
, unsigned long vm_end
)
3135 struct vm_area_struct
*vma
= vma_lookup(mm
, vm_start
);
3137 if (vma
&& (vma
->vm_start
!= vm_start
|| vma
->vm_end
!= vm_end
))
3143 static inline bool range_in_vma(struct vm_area_struct
*vma
,
3144 unsigned long start
, unsigned long end
)
3146 return (vma
&& vma
->vm_start
<= start
&& end
<= vma
->vm_end
);
3150 pgprot_t
vm_get_page_prot(unsigned long vm_flags
);
3151 void vma_set_page_prot(struct vm_area_struct
*vma
);
3153 static inline pgprot_t
vm_get_page_prot(unsigned long vm_flags
)
3157 static inline void vma_set_page_prot(struct vm_area_struct
*vma
)
3159 vma
->vm_page_prot
= vm_get_page_prot(vma
->vm_flags
);
3163 void vma_set_file(struct vm_area_struct
*vma
, struct file
*file
);
3165 #ifdef CONFIG_NUMA_BALANCING
3166 unsigned long change_prot_numa(struct vm_area_struct
*vma
,
3167 unsigned long start
, unsigned long end
);
3170 struct vm_area_struct
*find_extend_vma(struct mm_struct
*, unsigned long addr
);
3171 int remap_pfn_range(struct vm_area_struct
*, unsigned long addr
,
3172 unsigned long pfn
, unsigned long size
, pgprot_t
);
3173 int remap_pfn_range_notrack(struct vm_area_struct
*vma
, unsigned long addr
,
3174 unsigned long pfn
, unsigned long size
, pgprot_t prot
);
3175 int vm_insert_page(struct vm_area_struct
*, unsigned long addr
, struct page
*);
3176 int vm_insert_pages(struct vm_area_struct
*vma
, unsigned long addr
,
3177 struct page
**pages
, unsigned long *num
);
3178 int vm_map_pages(struct vm_area_struct
*vma
, struct page
**pages
,
3180 int vm_map_pages_zero(struct vm_area_struct
*vma
, struct page
**pages
,
3182 vm_fault_t
vmf_insert_pfn(struct vm_area_struct
*vma
, unsigned long addr
,
3184 vm_fault_t
vmf_insert_pfn_prot(struct vm_area_struct
*vma
, unsigned long addr
,
3185 unsigned long pfn
, pgprot_t pgprot
);
3186 vm_fault_t
vmf_insert_mixed(struct vm_area_struct
*vma
, unsigned long addr
,
3188 vm_fault_t
vmf_insert_mixed_prot(struct vm_area_struct
*vma
, unsigned long addr
,
3189 pfn_t pfn
, pgprot_t pgprot
);
3190 vm_fault_t
vmf_insert_mixed_mkwrite(struct vm_area_struct
*vma
,
3191 unsigned long addr
, pfn_t pfn
);
3192 int vm_iomap_memory(struct vm_area_struct
*vma
, phys_addr_t start
, unsigned long len
);
3194 static inline vm_fault_t
vmf_insert_page(struct vm_area_struct
*vma
,
3195 unsigned long addr
, struct page
*page
)
3197 int err
= vm_insert_page(vma
, addr
, page
);
3200 return VM_FAULT_OOM
;
3201 if (err
< 0 && err
!= -EBUSY
)
3202 return VM_FAULT_SIGBUS
;
3204 return VM_FAULT_NOPAGE
;
3207 #ifndef io_remap_pfn_range
3208 static inline int io_remap_pfn_range(struct vm_area_struct
*vma
,
3209 unsigned long addr
, unsigned long pfn
,
3210 unsigned long size
, pgprot_t prot
)
3212 return remap_pfn_range(vma
, addr
, pfn
, size
, pgprot_decrypted(prot
));
3216 static inline vm_fault_t
vmf_error(int err
)
3219 return VM_FAULT_OOM
;
3220 return VM_FAULT_SIGBUS
;
3223 struct page
*follow_page(struct vm_area_struct
*vma
, unsigned long address
,
3224 unsigned int foll_flags
);
3226 static inline int vm_fault_to_errno(vm_fault_t vm_fault
, int foll_flags
)
3228 if (vm_fault
& VM_FAULT_OOM
)
3230 if (vm_fault
& (VM_FAULT_HWPOISON
| VM_FAULT_HWPOISON_LARGE
))
3231 return (foll_flags
& FOLL_HWPOISON
) ? -EHWPOISON
: -EFAULT
;
3232 if (vm_fault
& (VM_FAULT_SIGBUS
| VM_FAULT_SIGSEGV
))
3238 * Indicates whether GUP can follow a PROT_NONE mapped page, or whether
3239 * a (NUMA hinting) fault is required.
3241 static inline bool gup_can_follow_protnone(unsigned int flags
)
3244 * FOLL_FORCE has to be able to make progress even if the VMA is
3245 * inaccessible. Further, FOLL_FORCE access usually does not represent
3246 * application behaviour and we should avoid triggering NUMA hinting
3249 return flags
& FOLL_FORCE
;
3252 typedef int (*pte_fn_t
)(pte_t
*pte
, unsigned long addr
, void *data
);
3253 extern int apply_to_page_range(struct mm_struct
*mm
, unsigned long address
,
3254 unsigned long size
, pte_fn_t fn
, void *data
);
3255 extern int apply_to_existing_page_range(struct mm_struct
*mm
,
3256 unsigned long address
, unsigned long size
,
3257 pte_fn_t fn
, void *data
);
3259 extern void __init
init_mem_debugging_and_hardening(void);
3260 #ifdef CONFIG_PAGE_POISONING
3261 extern void __kernel_poison_pages(struct page
*page
, int numpages
);
3262 extern void __kernel_unpoison_pages(struct page
*page
, int numpages
);
3263 extern bool _page_poisoning_enabled_early
;
3264 DECLARE_STATIC_KEY_FALSE(_page_poisoning_enabled
);
3265 static inline bool page_poisoning_enabled(void)
3267 return _page_poisoning_enabled_early
;
3270 * For use in fast paths after init_mem_debugging() has run, or when a
3271 * false negative result is not harmful when called too early.
3273 static inline bool page_poisoning_enabled_static(void)
3275 return static_branch_unlikely(&_page_poisoning_enabled
);
3277 static inline void kernel_poison_pages(struct page
*page
, int numpages
)
3279 if (page_poisoning_enabled_static())
3280 __kernel_poison_pages(page
, numpages
);
3282 static inline void kernel_unpoison_pages(struct page
*page
, int numpages
)
3284 if (page_poisoning_enabled_static())
3285 __kernel_unpoison_pages(page
, numpages
);
3288 static inline bool page_poisoning_enabled(void) { return false; }
3289 static inline bool page_poisoning_enabled_static(void) { return false; }
3290 static inline void __kernel_poison_pages(struct page
*page
, int nunmpages
) { }
3291 static inline void kernel_poison_pages(struct page
*page
, int numpages
) { }
3292 static inline void kernel_unpoison_pages(struct page
*page
, int numpages
) { }
3295 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_ALLOC_DEFAULT_ON
, init_on_alloc
);
3296 static inline bool want_init_on_alloc(gfp_t flags
)
3298 if (static_branch_maybe(CONFIG_INIT_ON_ALLOC_DEFAULT_ON
,
3301 return flags
& __GFP_ZERO
;
3304 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_FREE_DEFAULT_ON
, init_on_free
);
3305 static inline bool want_init_on_free(void)
3307 return static_branch_maybe(CONFIG_INIT_ON_FREE_DEFAULT_ON
,
3311 extern bool _debug_pagealloc_enabled_early
;
3312 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled
);
3314 static inline bool debug_pagealloc_enabled(void)
3316 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC
) &&
3317 _debug_pagealloc_enabled_early
;
3321 * For use in fast paths after init_debug_pagealloc() has run, or when a
3322 * false negative result is not harmful when called too early.
3324 static inline bool debug_pagealloc_enabled_static(void)
3326 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC
))
3329 return static_branch_unlikely(&_debug_pagealloc_enabled
);
3332 #ifdef CONFIG_DEBUG_PAGEALLOC
3334 * To support DEBUG_PAGEALLOC architecture must ensure that
3335 * __kernel_map_pages() never fails
3337 extern void __kernel_map_pages(struct page
*page
, int numpages
, int enable
);
3339 static inline void debug_pagealloc_map_pages(struct page
*page
, int numpages
)
3341 if (debug_pagealloc_enabled_static())
3342 __kernel_map_pages(page
, numpages
, 1);
3345 static inline void debug_pagealloc_unmap_pages(struct page
*page
, int numpages
)
3347 if (debug_pagealloc_enabled_static())
3348 __kernel_map_pages(page
, numpages
, 0);
3350 #else /* CONFIG_DEBUG_PAGEALLOC */
3351 static inline void debug_pagealloc_map_pages(struct page
*page
, int numpages
) {}
3352 static inline void debug_pagealloc_unmap_pages(struct page
*page
, int numpages
) {}
3353 #endif /* CONFIG_DEBUG_PAGEALLOC */
3355 #ifdef __HAVE_ARCH_GATE_AREA
3356 extern struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
);
3357 extern int in_gate_area_no_mm(unsigned long addr
);
3358 extern int in_gate_area(struct mm_struct
*mm
, unsigned long addr
);
3360 static inline struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
)
3364 static inline int in_gate_area_no_mm(unsigned long addr
) { return 0; }
3365 static inline int in_gate_area(struct mm_struct
*mm
, unsigned long addr
)
3369 #endif /* __HAVE_ARCH_GATE_AREA */
3371 extern bool process_shares_mm(struct task_struct
*p
, struct mm_struct
*mm
);
3373 #ifdef CONFIG_SYSCTL
3374 extern int sysctl_drop_caches
;
3375 int drop_caches_sysctl_handler(struct ctl_table
*, int, void *, size_t *,
3379 void drop_slab(void);
3382 #define randomize_va_space 0
3384 extern int randomize_va_space
;
3387 const char * arch_vma_name(struct vm_area_struct
*vma
);
3389 void print_vma_addr(char *prefix
, unsigned long rip
);
3391 static inline void print_vma_addr(char *prefix
, unsigned long rip
)
3396 void *sparse_buffer_alloc(unsigned long size
);
3397 struct page
* __populate_section_memmap(unsigned long pfn
,
3398 unsigned long nr_pages
, int nid
, struct vmem_altmap
*altmap
,
3399 struct dev_pagemap
*pgmap
);
3400 void pmd_init(void *addr
);
3401 void pud_init(void *addr
);
3402 pgd_t
*vmemmap_pgd_populate(unsigned long addr
, int node
);
3403 p4d_t
*vmemmap_p4d_populate(pgd_t
*pgd
, unsigned long addr
, int node
);
3404 pud_t
*vmemmap_pud_populate(p4d_t
*p4d
, unsigned long addr
, int node
);
3405 pmd_t
*vmemmap_pmd_populate(pud_t
*pud
, unsigned long addr
, int node
);
3406 pte_t
*vmemmap_pte_populate(pmd_t
*pmd
, unsigned long addr
, int node
,
3407 struct vmem_altmap
*altmap
, struct page
*reuse
);
3408 void *vmemmap_alloc_block(unsigned long size
, int node
);
3410 void *vmemmap_alloc_block_buf(unsigned long size
, int node
,
3411 struct vmem_altmap
*altmap
);
3412 void vmemmap_verify(pte_t
*, int, unsigned long, unsigned long);
3413 void vmemmap_set_pmd(pmd_t
*pmd
, void *p
, int node
,
3414 unsigned long addr
, unsigned long next
);
3415 int vmemmap_check_pmd(pmd_t
*pmd
, int node
,
3416 unsigned long addr
, unsigned long next
);
3417 int vmemmap_populate_basepages(unsigned long start
, unsigned long end
,
3418 int node
, struct vmem_altmap
*altmap
);
3419 int vmemmap_populate_hugepages(unsigned long start
, unsigned long end
,
3420 int node
, struct vmem_altmap
*altmap
);
3421 int vmemmap_populate(unsigned long start
, unsigned long end
, int node
,
3422 struct vmem_altmap
*altmap
);
3423 void vmemmap_populate_print_last(void);
3424 #ifdef CONFIG_MEMORY_HOTPLUG
3425 void vmemmap_free(unsigned long start
, unsigned long end
,
3426 struct vmem_altmap
*altmap
);
3428 void register_page_bootmem_memmap(unsigned long section_nr
, struct page
*map
,
3429 unsigned long nr_pages
);
3432 MF_COUNT_INCREASED
= 1 << 0,
3433 MF_ACTION_REQUIRED
= 1 << 1,
3434 MF_MUST_KILL
= 1 << 2,
3435 MF_SOFT_OFFLINE
= 1 << 3,
3436 MF_UNPOISON
= 1 << 4,
3437 MF_SW_SIMULATED
= 1 << 5,
3438 MF_NO_RETRY
= 1 << 6,
3440 int mf_dax_kill_procs(struct address_space
*mapping
, pgoff_t index
,
3441 unsigned long count
, int mf_flags
);
3442 extern int memory_failure(unsigned long pfn
, int flags
);
3443 extern void memory_failure_queue_kick(int cpu
);
3444 extern int unpoison_memory(unsigned long pfn
);
3445 extern int sysctl_memory_failure_early_kill
;
3446 extern int sysctl_memory_failure_recovery
;
3447 extern void shake_page(struct page
*p
);
3448 extern atomic_long_t num_poisoned_pages __read_mostly
;
3449 extern int soft_offline_page(unsigned long pfn
, int flags
);
3450 #ifdef CONFIG_MEMORY_FAILURE
3451 extern void memory_failure_queue(unsigned long pfn
, int flags
);
3452 extern int __get_huge_page_for_hwpoison(unsigned long pfn
, int flags
,
3453 bool *migratable_cleared
);
3454 void num_poisoned_pages_inc(unsigned long pfn
);
3455 void num_poisoned_pages_sub(unsigned long pfn
, long i
);
3457 static inline void memory_failure_queue(unsigned long pfn
, int flags
)
3461 static inline int __get_huge_page_for_hwpoison(unsigned long pfn
, int flags
,
3462 bool *migratable_cleared
)
3467 static inline void num_poisoned_pages_inc(unsigned long pfn
)
3471 static inline void num_poisoned_pages_sub(unsigned long pfn
, long i
)
3476 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_MEMORY_HOTPLUG)
3477 extern void memblk_nr_poison_inc(unsigned long pfn
);
3478 extern void memblk_nr_poison_sub(unsigned long pfn
, long i
);
3480 static inline void memblk_nr_poison_inc(unsigned long pfn
)
3484 static inline void memblk_nr_poison_sub(unsigned long pfn
, long i
)
3489 #ifndef arch_memory_failure
3490 static inline int arch_memory_failure(unsigned long pfn
, int flags
)
3496 #ifndef arch_is_platform_page
3497 static inline bool arch_is_platform_page(u64 paddr
)
3504 * Error handlers for various types of pages.
3507 MF_IGNORED
, /* Error: cannot be handled */
3508 MF_FAILED
, /* Error: handling failed */
3509 MF_DELAYED
, /* Will be handled later */
3510 MF_RECOVERED
, /* Successfully recovered */
3513 enum mf_action_page_type
{
3515 MF_MSG_KERNEL_HIGH_ORDER
,
3517 MF_MSG_DIFFERENT_COMPOUND
,
3520 MF_MSG_UNMAP_FAILED
,
3521 MF_MSG_DIRTY_SWAPCACHE
,
3522 MF_MSG_CLEAN_SWAPCACHE
,
3523 MF_MSG_DIRTY_MLOCKED_LRU
,
3524 MF_MSG_CLEAN_MLOCKED_LRU
,
3525 MF_MSG_DIRTY_UNEVICTABLE_LRU
,
3526 MF_MSG_CLEAN_UNEVICTABLE_LRU
,
3529 MF_MSG_TRUNCATED_LRU
,
3537 * Sysfs entries for memory failure handling statistics.
3539 extern const struct attribute_group memory_failure_attr_group
;
3541 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
3542 extern void clear_huge_page(struct page
*page
,
3543 unsigned long addr_hint
,
3544 unsigned int pages_per_huge_page
);
3545 extern void copy_user_huge_page(struct page
*dst
, struct page
*src
,
3546 unsigned long addr_hint
,
3547 struct vm_area_struct
*vma
,
3548 unsigned int pages_per_huge_page
);
3549 extern long copy_huge_page_from_user(struct page
*dst_page
,
3550 const void __user
*usr_src
,
3551 unsigned int pages_per_huge_page
,
3552 bool allow_pagefault
);
3555 * vma_is_special_huge - Are transhuge page-table entries considered special?
3556 * @vma: Pointer to the struct vm_area_struct to consider
3558 * Whether transhuge page-table entries are considered "special" following
3559 * the definition in vm_normal_page().
3561 * Return: true if transhuge page-table entries should be considered special,
3564 static inline bool vma_is_special_huge(const struct vm_area_struct
*vma
)
3566 return vma_is_dax(vma
) || (vma
->vm_file
&&
3567 (vma
->vm_flags
& (VM_PFNMAP
| VM_MIXEDMAP
)));
3570 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
3572 #ifdef CONFIG_DEBUG_PAGEALLOC
3573 extern unsigned int _debug_guardpage_minorder
;
3574 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled
);
3576 static inline unsigned int debug_guardpage_minorder(void)
3578 return _debug_guardpage_minorder
;
3581 static inline bool debug_guardpage_enabled(void)
3583 return static_branch_unlikely(&_debug_guardpage_enabled
);
3586 static inline bool page_is_guard(struct page
*page
)
3588 if (!debug_guardpage_enabled())
3591 return PageGuard(page
);
3594 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
3595 static inline bool debug_guardpage_enabled(void) { return false; }
3596 static inline bool page_is_guard(struct page
*page
) { return false; }
3597 #endif /* CONFIG_DEBUG_PAGEALLOC */
3599 #if MAX_NUMNODES > 1
3600 void __init
setup_nr_node_ids(void);
3602 static inline void setup_nr_node_ids(void) {}
3605 extern int memcmp_pages(struct page
*page1
, struct page
*page2
);
3607 static inline int pages_identical(struct page
*page1
, struct page
*page2
)
3609 return !memcmp_pages(page1
, page2
);
3612 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
3613 unsigned long clean_record_shared_mapping_range(struct address_space
*mapping
,
3614 pgoff_t first_index
, pgoff_t nr
,
3615 pgoff_t bitmap_pgoff
,
3616 unsigned long *bitmap
,
3620 unsigned long wp_shared_mapping_range(struct address_space
*mapping
,
3621 pgoff_t first_index
, pgoff_t nr
);
3624 extern int sysctl_nr_trim_pages
;
3626 #ifdef CONFIG_PRINTK
3627 void mem_dump_obj(void *object
);
3629 static inline void mem_dump_obj(void *object
) {}
3633 * seal_check_future_write - Check for F_SEAL_FUTURE_WRITE flag and handle it
3634 * @seals: the seals to check
3635 * @vma: the vma to operate on
3637 * Check whether F_SEAL_FUTURE_WRITE is set; if so, do proper check/handling on
3638 * the vma flags. Return 0 if check pass, or <0 for errors.
3640 static inline int seal_check_future_write(int seals
, struct vm_area_struct
*vma
)
3642 if (seals
& F_SEAL_FUTURE_WRITE
) {
3644 * New PROT_WRITE and MAP_SHARED mmaps are not allowed when
3645 * "future write" seal active.
3647 if ((vma
->vm_flags
& VM_SHARED
) && (vma
->vm_flags
& VM_WRITE
))
3651 * Since an F_SEAL_FUTURE_WRITE sealed memfd can be mapped as
3652 * MAP_SHARED and read-only, take care to not allow mprotect to
3653 * revert protections on such mappings. Do this only for shared
3654 * mappings. For private mappings, don't need to mask
3655 * VM_MAYWRITE as we still want them to be COW-writable.
3657 if (vma
->vm_flags
& VM_SHARED
)
3658 vm_flags_clear(vma
, VM_MAYWRITE
);
3664 #ifdef CONFIG_ANON_VMA_NAME
3665 int madvise_set_anon_name(struct mm_struct
*mm
, unsigned long start
,
3666 unsigned long len_in
,
3667 struct anon_vma_name
*anon_name
);
3670 madvise_set_anon_name(struct mm_struct
*mm
, unsigned long start
,
3671 unsigned long len_in
, struct anon_vma_name
*anon_name
) {
3676 #endif /* _LINUX_MM_H */