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>
32 #include <linux/slab.h>
36 struct anon_vma_chain
;
40 extern int sysctl_page_lock_unfairness
;
42 void mm_core_init(void);
43 void init_mm_internals(void);
45 #ifndef CONFIG_NUMA /* Don't use mapnrs, do it properly */
46 extern unsigned long max_mapnr
;
48 static inline void set_max_mapnr(unsigned long limit
)
53 static inline void set_max_mapnr(unsigned long limit
) { }
56 extern atomic_long_t _totalram_pages
;
57 static inline unsigned long totalram_pages(void)
59 return (unsigned long)atomic_long_read(&_totalram_pages
);
62 static inline void totalram_pages_inc(void)
64 atomic_long_inc(&_totalram_pages
);
67 static inline void totalram_pages_dec(void)
69 atomic_long_dec(&_totalram_pages
);
72 static inline void totalram_pages_add(long count
)
74 atomic_long_add(count
, &_totalram_pages
);
77 extern void * high_memory
;
78 extern int page_cluster
;
79 extern const int page_cluster_max
;
82 extern int sysctl_legacy_va_layout
;
84 #define sysctl_legacy_va_layout 0
87 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
88 extern const int mmap_rnd_bits_min
;
89 extern const int mmap_rnd_bits_max
;
90 extern int mmap_rnd_bits __read_mostly
;
92 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
93 extern const int mmap_rnd_compat_bits_min
;
94 extern const int mmap_rnd_compat_bits_max
;
95 extern int mmap_rnd_compat_bits __read_mostly
;
99 #include <asm/processor.h>
102 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
106 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
110 #define lm_alias(x) __va(__pa_symbol(x))
114 * To prevent common memory management code establishing
115 * a zero page mapping on a read fault.
116 * This macro should be defined within <asm/pgtable.h>.
117 * s390 does this to prevent multiplexing of hardware bits
118 * related to the physical page in case of virtualization.
120 #ifndef mm_forbids_zeropage
121 #define mm_forbids_zeropage(X) (0)
125 * On some architectures it is expensive to call memset() for small sizes.
126 * If an architecture decides to implement their own version of
127 * mm_zero_struct_page they should wrap the defines below in a #ifndef and
128 * define their own version of this macro in <asm/pgtable.h>
130 #if BITS_PER_LONG == 64
131 /* This function must be updated when the size of struct page grows above 96
132 * or reduces below 56. The idea that compiler optimizes out switch()
133 * statement, and only leaves move/store instructions. Also the compiler can
134 * combine write statements if they are both assignments and can be reordered,
135 * this can result in several of the writes here being dropped.
137 #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
138 static inline void __mm_zero_struct_page(struct page
*page
)
140 unsigned long *_pp
= (void *)page
;
142 /* Check that struct page is either 56, 64, 72, 80, 88 or 96 bytes */
143 BUILD_BUG_ON(sizeof(struct page
) & 7);
144 BUILD_BUG_ON(sizeof(struct page
) < 56);
145 BUILD_BUG_ON(sizeof(struct page
) > 96);
147 switch (sizeof(struct page
)) {
174 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
178 * Default maximum number of active map areas, this limits the number of vmas
179 * per mm struct. Users can overwrite this number by sysctl but there is a
182 * When a program's coredump is generated as ELF format, a section is created
183 * per a vma. In ELF, the number of sections is represented in unsigned short.
184 * This means the number of sections should be smaller than 65535 at coredump.
185 * Because the kernel adds some informative sections to a image of program at
186 * generating coredump, we need some margin. The number of extra sections is
187 * 1-3 now and depends on arch. We use "5" as safe margin, here.
189 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
190 * not a hard limit any more. Although some userspace tools can be surprised by
193 #define MAPCOUNT_ELF_CORE_MARGIN (5)
194 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
196 extern int sysctl_max_map_count
;
198 extern unsigned long sysctl_user_reserve_kbytes
;
199 extern unsigned long sysctl_admin_reserve_kbytes
;
201 extern int sysctl_overcommit_memory
;
202 extern int sysctl_overcommit_ratio
;
203 extern unsigned long sysctl_overcommit_kbytes
;
205 int overcommit_ratio_handler(struct ctl_table
*, int, void *, size_t *,
207 int overcommit_kbytes_handler(struct ctl_table
*, int, void *, size_t *,
209 int overcommit_policy_handler(struct ctl_table
*, int, void *, size_t *,
212 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
213 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
214 #define folio_page_idx(folio, p) (page_to_pfn(p) - folio_pfn(folio))
216 #define nth_page(page,n) ((page) + (n))
217 #define folio_page_idx(folio, p) ((p) - &(folio)->page)
220 /* to align the pointer to the (next) page boundary */
221 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
223 /* to align the pointer to the (prev) page boundary */
224 #define PAGE_ALIGN_DOWN(addr) ALIGN_DOWN(addr, PAGE_SIZE)
226 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
227 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
229 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
230 static inline struct folio
*lru_to_folio(struct list_head
*head
)
232 return list_entry((head
)->prev
, struct folio
, lru
);
235 void setup_initial_init_mm(void *start_code
, void *end_code
,
236 void *end_data
, void *brk
);
239 * Linux kernel virtual memory manager primitives.
240 * The idea being to have a "virtual" mm in the same way
241 * we have a virtual fs - giving a cleaner interface to the
242 * mm details, and allowing different kinds of memory mappings
243 * (from shared memory to executable loading to arbitrary
247 struct vm_area_struct
*vm_area_alloc(struct mm_struct
*);
248 struct vm_area_struct
*vm_area_dup(struct vm_area_struct
*);
249 void vm_area_free(struct vm_area_struct
*);
250 /* Use only if VMA has no other users */
251 void __vm_area_free(struct vm_area_struct
*vma
);
254 extern struct rb_root nommu_region_tree
;
255 extern struct rw_semaphore nommu_region_sem
;
257 extern unsigned int kobjsize(const void *objp
);
261 * vm_flags in vm_area_struct, see mm_types.h.
262 * When changing, update also include/trace/events/mmflags.h
264 #define VM_NONE 0x00000000
266 #define VM_READ 0x00000001 /* currently active flags */
267 #define VM_WRITE 0x00000002
268 #define VM_EXEC 0x00000004
269 #define VM_SHARED 0x00000008
271 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
272 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
273 #define VM_MAYWRITE 0x00000020
274 #define VM_MAYEXEC 0x00000040
275 #define VM_MAYSHARE 0x00000080
277 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
279 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
280 #else /* CONFIG_MMU */
281 #define VM_MAYOVERLAY 0x00000200 /* nommu: R/O MAP_PRIVATE mapping that might overlay a file mapping */
282 #define VM_UFFD_MISSING 0
283 #endif /* CONFIG_MMU */
284 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
285 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
287 #define VM_LOCKED 0x00002000
288 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
290 /* Used by sys_madvise() */
291 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
292 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
294 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
295 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
296 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
297 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
298 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
299 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
300 #define VM_SYNC 0x00800000 /* Synchronous page faults */
301 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
302 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
303 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
305 #ifdef CONFIG_MEM_SOFT_DIRTY
306 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
308 # define VM_SOFTDIRTY 0
311 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
312 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
313 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
314 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
316 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
317 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
318 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
319 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
320 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
321 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
322 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
323 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
324 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
325 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
326 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
327 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
329 #ifdef CONFIG_ARCH_HAS_PKEYS
330 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
331 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
332 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
333 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
334 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
336 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
338 # define VM_PKEY_BIT4 0
340 #endif /* CONFIG_ARCH_HAS_PKEYS */
342 #if defined(CONFIG_X86)
343 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
344 #elif defined(CONFIG_PPC)
345 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
346 #elif defined(CONFIG_PARISC)
347 # define VM_GROWSUP VM_ARCH_1
348 #elif defined(CONFIG_IA64)
349 # define VM_GROWSUP VM_ARCH_1
350 #elif defined(CONFIG_SPARC64)
351 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
352 # define VM_ARCH_CLEAR VM_SPARC_ADI
353 #elif defined(CONFIG_ARM64)
354 # define VM_ARM64_BTI VM_ARCH_1 /* BTI guarded page, a.k.a. GP bit */
355 # define VM_ARCH_CLEAR VM_ARM64_BTI
356 #elif !defined(CONFIG_MMU)
357 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
360 #if defined(CONFIG_ARM64_MTE)
361 # define VM_MTE VM_HIGH_ARCH_0 /* Use Tagged memory for access control */
362 # define VM_MTE_ALLOWED VM_HIGH_ARCH_1 /* Tagged memory permitted */
364 # define VM_MTE VM_NONE
365 # define VM_MTE_ALLOWED VM_NONE
369 # define VM_GROWSUP VM_NONE
372 #ifdef CONFIG_HAVE_ARCH_USERFAULTFD_MINOR
373 # define VM_UFFD_MINOR_BIT 37
374 # define VM_UFFD_MINOR BIT(VM_UFFD_MINOR_BIT) /* UFFD minor faults */
375 #else /* !CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
376 # define VM_UFFD_MINOR VM_NONE
377 #endif /* CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
379 /* Bits set in the VMA until the stack is in its final location */
380 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ | VM_STACK_EARLY)
382 #define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0)
384 /* Common data flag combinations */
385 #define VM_DATA_FLAGS_TSK_EXEC (VM_READ | VM_WRITE | TASK_EXEC | \
386 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
387 #define VM_DATA_FLAGS_NON_EXEC (VM_READ | VM_WRITE | VM_MAYREAD | \
388 VM_MAYWRITE | VM_MAYEXEC)
389 #define VM_DATA_FLAGS_EXEC (VM_READ | VM_WRITE | VM_EXEC | \
390 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
392 #ifndef VM_DATA_DEFAULT_FLAGS /* arch can override this */
393 #define VM_DATA_DEFAULT_FLAGS VM_DATA_FLAGS_EXEC
396 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
397 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
400 #ifdef CONFIG_STACK_GROWSUP
401 #define VM_STACK VM_GROWSUP
402 #define VM_STACK_EARLY VM_GROWSDOWN
404 #define VM_STACK VM_GROWSDOWN
405 #define VM_STACK_EARLY 0
408 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
410 /* VMA basic access permission flags */
411 #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
415 * Special vmas that are non-mergable, non-mlock()able.
417 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
419 /* This mask prevents VMA from being scanned with khugepaged */
420 #define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB)
422 /* This mask defines which mm->def_flags a process can inherit its parent */
423 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
425 /* This mask represents all the VMA flag bits used by mlock */
426 #define VM_LOCKED_MASK (VM_LOCKED | VM_LOCKONFAULT)
428 /* Arch-specific flags to clear when updating VM flags on protection change */
429 #ifndef VM_ARCH_CLEAR
430 # define VM_ARCH_CLEAR VM_NONE
432 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
435 * mapping from the currently active vm_flags protection bits (the
436 * low four bits) to a page protection mask..
440 * The default fault flags that should be used by most of the
441 * arch-specific page fault handlers.
443 #define FAULT_FLAG_DEFAULT (FAULT_FLAG_ALLOW_RETRY | \
444 FAULT_FLAG_KILLABLE | \
445 FAULT_FLAG_INTERRUPTIBLE)
448 * fault_flag_allow_retry_first - check ALLOW_RETRY the first time
449 * @flags: Fault flags.
451 * This is mostly used for places where we want to try to avoid taking
452 * the mmap_lock for too long a time when waiting for another condition
453 * to change, in which case we can try to be polite to release the
454 * mmap_lock in the first round to avoid potential starvation of other
455 * processes that would also want the mmap_lock.
457 * Return: true if the page fault allows retry and this is the first
458 * attempt of the fault handling; false otherwise.
460 static inline bool fault_flag_allow_retry_first(enum fault_flag flags
)
462 return (flags
& FAULT_FLAG_ALLOW_RETRY
) &&
463 (!(flags
& FAULT_FLAG_TRIED
));
466 #define FAULT_FLAG_TRACE \
467 { FAULT_FLAG_WRITE, "WRITE" }, \
468 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
469 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
470 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
471 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
472 { FAULT_FLAG_TRIED, "TRIED" }, \
473 { FAULT_FLAG_USER, "USER" }, \
474 { FAULT_FLAG_REMOTE, "REMOTE" }, \
475 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }, \
476 { FAULT_FLAG_INTERRUPTIBLE, "INTERRUPTIBLE" }, \
477 { FAULT_FLAG_VMA_LOCK, "VMA_LOCK" }
480 * vm_fault is filled by the pagefault handler and passed to the vma's
481 * ->fault function. The vma's ->fault is responsible for returning a bitmask
482 * of VM_FAULT_xxx flags that give details about how the fault was handled.
484 * MM layer fills up gfp_mask for page allocations but fault handler might
485 * alter it if its implementation requires a different allocation context.
487 * pgoff should be used in favour of virtual_address, if possible.
491 struct vm_area_struct
*vma
; /* Target VMA */
492 gfp_t gfp_mask
; /* gfp mask to be used for allocations */
493 pgoff_t pgoff
; /* Logical page offset based on vma */
494 unsigned long address
; /* Faulting virtual address - masked */
495 unsigned long real_address
; /* Faulting virtual address - unmasked */
497 enum fault_flag flags
; /* FAULT_FLAG_xxx flags
498 * XXX: should really be 'const' */
499 pmd_t
*pmd
; /* Pointer to pmd entry matching
501 pud_t
*pud
; /* Pointer to pud entry matching
505 pte_t orig_pte
; /* Value of PTE at the time of fault */
506 pmd_t orig_pmd
; /* Value of PMD at the time of fault,
507 * used by PMD fault only.
511 struct page
*cow_page
; /* Page handler may use for COW fault */
512 struct page
*page
; /* ->fault handlers should return a
513 * page here, unless VM_FAULT_NOPAGE
514 * is set (which is also implied by
517 /* These three entries are valid only while holding ptl lock */
518 pte_t
*pte
; /* Pointer to pte entry matching
519 * the 'address'. NULL if the page
520 * table hasn't been allocated.
522 spinlock_t
*ptl
; /* Page table lock.
523 * Protects pte page table if 'pte'
524 * is not NULL, otherwise pmd.
526 pgtable_t prealloc_pte
; /* Pre-allocated pte page table.
527 * vm_ops->map_pages() sets up a page
528 * table from atomic context.
529 * do_fault_around() pre-allocates
530 * page table to avoid allocation from
535 /* page entry size for vm->huge_fault() */
536 enum page_entry_size
{
543 * These are the virtual MM functions - opening of an area, closing and
544 * unmapping it (needed to keep files on disk up-to-date etc), pointer
545 * to the functions called when a no-page or a wp-page exception occurs.
547 struct vm_operations_struct
{
548 void (*open
)(struct vm_area_struct
* area
);
550 * @close: Called when the VMA is being removed from the MM.
551 * Context: User context. May sleep. Caller holds mmap_lock.
553 void (*close
)(struct vm_area_struct
* area
);
554 /* Called any time before splitting to check if it's allowed */
555 int (*may_split
)(struct vm_area_struct
*area
, unsigned long addr
);
556 int (*mremap
)(struct vm_area_struct
*area
);
558 * Called by mprotect() to make driver-specific permission
559 * checks before mprotect() is finalised. The VMA must not
560 * be modified. Returns 0 if mprotect() can proceed.
562 int (*mprotect
)(struct vm_area_struct
*vma
, unsigned long start
,
563 unsigned long end
, unsigned long newflags
);
564 vm_fault_t (*fault
)(struct vm_fault
*vmf
);
565 vm_fault_t (*huge_fault
)(struct vm_fault
*vmf
,
566 enum page_entry_size pe_size
);
567 vm_fault_t (*map_pages
)(struct vm_fault
*vmf
,
568 pgoff_t start_pgoff
, pgoff_t end_pgoff
);
569 unsigned long (*pagesize
)(struct vm_area_struct
* area
);
571 /* notification that a previously read-only page is about to become
572 * writable, if an error is returned it will cause a SIGBUS */
573 vm_fault_t (*page_mkwrite
)(struct vm_fault
*vmf
);
575 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
576 vm_fault_t (*pfn_mkwrite
)(struct vm_fault
*vmf
);
578 /* called by access_process_vm when get_user_pages() fails, typically
579 * for use by special VMAs. See also generic_access_phys() for a generic
580 * implementation useful for any iomem mapping.
582 int (*access
)(struct vm_area_struct
*vma
, unsigned long addr
,
583 void *buf
, int len
, int write
);
585 /* Called by the /proc/PID/maps code to ask the vma whether it
586 * has a special name. Returning non-NULL will also cause this
587 * vma to be dumped unconditionally. */
588 const char *(*name
)(struct vm_area_struct
*vma
);
592 * set_policy() op must add a reference to any non-NULL @new mempolicy
593 * to hold the policy upon return. Caller should pass NULL @new to
594 * remove a policy and fall back to surrounding context--i.e. do not
595 * install a MPOL_DEFAULT policy, nor the task or system default
598 int (*set_policy
)(struct vm_area_struct
*vma
, struct mempolicy
*new);
601 * get_policy() op must add reference [mpol_get()] to any policy at
602 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
603 * in mm/mempolicy.c will do this automatically.
604 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
605 * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
606 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
607 * must return NULL--i.e., do not "fallback" to task or system default
610 struct mempolicy
*(*get_policy
)(struct vm_area_struct
*vma
,
614 * Called by vm_normal_page() for special PTEs to find the
615 * page for @addr. This is useful if the default behavior
616 * (using pte_page()) would not find the correct page.
618 struct page
*(*find_special_page
)(struct vm_area_struct
*vma
,
622 #ifdef CONFIG_NUMA_BALANCING
623 static inline void vma_numab_state_init(struct vm_area_struct
*vma
)
625 vma
->numab_state
= NULL
;
627 static inline void vma_numab_state_free(struct vm_area_struct
*vma
)
629 kfree(vma
->numab_state
);
632 static inline void vma_numab_state_init(struct vm_area_struct
*vma
) {}
633 static inline void vma_numab_state_free(struct vm_area_struct
*vma
) {}
634 #endif /* CONFIG_NUMA_BALANCING */
636 #ifdef CONFIG_PER_VMA_LOCK
638 * Try to read-lock a vma. The function is allowed to occasionally yield false
639 * locked result to avoid performance overhead, in which case we fall back to
640 * using mmap_lock. The function should never yield false unlocked result.
642 static inline bool vma_start_read(struct vm_area_struct
*vma
)
645 * Check before locking. A race might cause false locked result.
646 * We can use READ_ONCE() for the mm_lock_seq here, and don't need
647 * ACQUIRE semantics, because this is just a lockless check whose result
648 * we don't rely on for anything - the mm_lock_seq read against which we
649 * need ordering is below.
651 if (READ_ONCE(vma
->vm_lock_seq
) == READ_ONCE(vma
->vm_mm
->mm_lock_seq
))
654 if (unlikely(down_read_trylock(&vma
->vm_lock
->lock
) == 0))
658 * Overflow might produce false locked result.
659 * False unlocked result is impossible because we modify and check
660 * vma->vm_lock_seq under vma->vm_lock protection and mm->mm_lock_seq
661 * modification invalidates all existing locks.
663 * We must use ACQUIRE semantics for the mm_lock_seq so that if we are
664 * racing with vma_end_write_all(), we only start reading from the VMA
665 * after it has been unlocked.
666 * This pairs with RELEASE semantics in vma_end_write_all().
668 if (unlikely(vma
->vm_lock_seq
== smp_load_acquire(&vma
->vm_mm
->mm_lock_seq
))) {
669 up_read(&vma
->vm_lock
->lock
);
675 static inline void vma_end_read(struct vm_area_struct
*vma
)
677 rcu_read_lock(); /* keeps vma alive till the end of up_read */
678 up_read(&vma
->vm_lock
->lock
);
682 static bool __is_vma_write_locked(struct vm_area_struct
*vma
, int *mm_lock_seq
)
684 mmap_assert_write_locked(vma
->vm_mm
);
687 * current task is holding mmap_write_lock, both vma->vm_lock_seq and
688 * mm->mm_lock_seq can't be concurrently modified.
690 *mm_lock_seq
= vma
->vm_mm
->mm_lock_seq
;
691 return (vma
->vm_lock_seq
== *mm_lock_seq
);
694 static inline void vma_start_write(struct vm_area_struct
*vma
)
698 if (__is_vma_write_locked(vma
, &mm_lock_seq
))
701 down_write(&vma
->vm_lock
->lock
);
703 * We should use WRITE_ONCE() here because we can have concurrent reads
704 * from the early lockless pessimistic check in vma_start_read().
705 * We don't really care about the correctness of that early check, but
706 * we should use WRITE_ONCE() for cleanliness and to keep KCSAN happy.
708 WRITE_ONCE(vma
->vm_lock_seq
, mm_lock_seq
);
709 up_write(&vma
->vm_lock
->lock
);
712 static inline bool vma_try_start_write(struct vm_area_struct
*vma
)
716 if (__is_vma_write_locked(vma
, &mm_lock_seq
))
719 if (!down_write_trylock(&vma
->vm_lock
->lock
))
722 WRITE_ONCE(vma
->vm_lock_seq
, mm_lock_seq
);
723 up_write(&vma
->vm_lock
->lock
);
727 static inline void vma_assert_write_locked(struct vm_area_struct
*vma
)
731 VM_BUG_ON_VMA(!__is_vma_write_locked(vma
, &mm_lock_seq
), vma
);
734 static inline void vma_mark_detached(struct vm_area_struct
*vma
, bool detached
)
736 /* When detaching vma should be write-locked */
738 vma_assert_write_locked(vma
);
739 vma
->detached
= detached
;
742 struct vm_area_struct
*lock_vma_under_rcu(struct mm_struct
*mm
,
743 unsigned long address
);
745 #else /* CONFIG_PER_VMA_LOCK */
747 static inline bool vma_start_read(struct vm_area_struct
*vma
)
749 static inline void vma_end_read(struct vm_area_struct
*vma
) {}
750 static inline void vma_start_write(struct vm_area_struct
*vma
) {}
751 static inline bool vma_try_start_write(struct vm_area_struct
*vma
)
753 static inline void vma_assert_write_locked(struct vm_area_struct
*vma
) {}
754 static inline void vma_mark_detached(struct vm_area_struct
*vma
,
757 #endif /* CONFIG_PER_VMA_LOCK */
760 * WARNING: vma_init does not initialize vma->vm_lock.
761 * Use vm_area_alloc()/vm_area_free() if vma needs locking.
763 static inline void vma_init(struct vm_area_struct
*vma
, struct mm_struct
*mm
)
765 static const struct vm_operations_struct dummy_vm_ops
= {};
767 memset(vma
, 0, sizeof(*vma
));
769 vma
->vm_ops
= &dummy_vm_ops
;
770 INIT_LIST_HEAD(&vma
->anon_vma_chain
);
771 vma_mark_detached(vma
, false);
772 vma_numab_state_init(vma
);
775 /* Use when VMA is not part of the VMA tree and needs no locking */
776 static inline void vm_flags_init(struct vm_area_struct
*vma
,
779 ACCESS_PRIVATE(vma
, __vm_flags
) = flags
;
782 /* Use when VMA is part of the VMA tree and modifications need coordination */
783 static inline void vm_flags_reset(struct vm_area_struct
*vma
,
786 vma_start_write(vma
);
787 vm_flags_init(vma
, flags
);
790 static inline void vm_flags_reset_once(struct vm_area_struct
*vma
,
793 vma_start_write(vma
);
794 WRITE_ONCE(ACCESS_PRIVATE(vma
, __vm_flags
), flags
);
797 static inline void vm_flags_set(struct vm_area_struct
*vma
,
800 vma_start_write(vma
);
801 ACCESS_PRIVATE(vma
, __vm_flags
) |= flags
;
804 static inline void vm_flags_clear(struct vm_area_struct
*vma
,
807 vma_start_write(vma
);
808 ACCESS_PRIVATE(vma
, __vm_flags
) &= ~flags
;
812 * Use only if VMA is not part of the VMA tree or has no other users and
813 * therefore needs no locking.
815 static inline void __vm_flags_mod(struct vm_area_struct
*vma
,
816 vm_flags_t set
, vm_flags_t clear
)
818 vm_flags_init(vma
, (vma
->vm_flags
| set
) & ~clear
);
822 * Use only when the order of set/clear operations is unimportant, otherwise
823 * use vm_flags_{set|clear} explicitly.
825 static inline void vm_flags_mod(struct vm_area_struct
*vma
,
826 vm_flags_t set
, vm_flags_t clear
)
828 vma_start_write(vma
);
829 __vm_flags_mod(vma
, set
, clear
);
832 static inline void vma_set_anonymous(struct vm_area_struct
*vma
)
837 static inline bool vma_is_anonymous(struct vm_area_struct
*vma
)
842 static inline bool vma_is_temporary_stack(struct vm_area_struct
*vma
)
844 int maybe_stack
= vma
->vm_flags
& (VM_GROWSDOWN
| VM_GROWSUP
);
849 if ((vma
->vm_flags
& VM_STACK_INCOMPLETE_SETUP
) ==
850 VM_STACK_INCOMPLETE_SETUP
)
856 static inline bool vma_is_foreign(struct vm_area_struct
*vma
)
861 if (current
->mm
!= vma
->vm_mm
)
867 static inline bool vma_is_accessible(struct vm_area_struct
*vma
)
869 return vma
->vm_flags
& VM_ACCESS_FLAGS
;
873 struct vm_area_struct
*vma_find(struct vma_iterator
*vmi
, unsigned long max
)
875 return mas_find(&vmi
->mas
, max
- 1);
878 static inline struct vm_area_struct
*vma_next(struct vma_iterator
*vmi
)
881 * Uses mas_find() to get the first VMA when the iterator starts.
882 * Calling mas_next() could skip the first entry.
884 return mas_find(&vmi
->mas
, ULONG_MAX
);
888 struct vm_area_struct
*vma_iter_next_range(struct vma_iterator
*vmi
)
890 return mas_next_range(&vmi
->mas
, ULONG_MAX
);
894 static inline struct vm_area_struct
*vma_prev(struct vma_iterator
*vmi
)
896 return mas_prev(&vmi
->mas
, 0);
900 struct vm_area_struct
*vma_iter_prev_range(struct vma_iterator
*vmi
)
902 return mas_prev_range(&vmi
->mas
, 0);
905 static inline unsigned long vma_iter_addr(struct vma_iterator
*vmi
)
907 return vmi
->mas
.index
;
910 static inline unsigned long vma_iter_end(struct vma_iterator
*vmi
)
912 return vmi
->mas
.last
+ 1;
914 static inline int vma_iter_bulk_alloc(struct vma_iterator
*vmi
,
917 return mas_expected_entries(&vmi
->mas
, count
);
920 /* Free any unused preallocations */
921 static inline void vma_iter_free(struct vma_iterator
*vmi
)
923 mas_destroy(&vmi
->mas
);
926 static inline int vma_iter_bulk_store(struct vma_iterator
*vmi
,
927 struct vm_area_struct
*vma
)
929 vmi
->mas
.index
= vma
->vm_start
;
930 vmi
->mas
.last
= vma
->vm_end
- 1;
931 mas_store(&vmi
->mas
, vma
);
932 if (unlikely(mas_is_err(&vmi
->mas
)))
938 static inline void vma_iter_invalidate(struct vma_iterator
*vmi
)
940 mas_pause(&vmi
->mas
);
943 static inline void vma_iter_set(struct vma_iterator
*vmi
, unsigned long addr
)
945 mas_set(&vmi
->mas
, addr
);
948 #define for_each_vma(__vmi, __vma) \
949 while (((__vma) = vma_next(&(__vmi))) != NULL)
951 /* The MM code likes to work with exclusive end addresses */
952 #define for_each_vma_range(__vmi, __vma, __end) \
953 while (((__vma) = vma_find(&(__vmi), (__end))) != NULL)
957 * The vma_is_shmem is not inline because it is used only by slow
958 * paths in userfault.
960 bool vma_is_shmem(struct vm_area_struct
*vma
);
961 bool vma_is_anon_shmem(struct vm_area_struct
*vma
);
963 static inline bool vma_is_shmem(struct vm_area_struct
*vma
) { return false; }
964 static inline bool vma_is_anon_shmem(struct vm_area_struct
*vma
) { return false; }
967 int vma_is_stack_for_current(struct vm_area_struct
*vma
);
969 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
970 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
976 * compound_order() can be called without holding a reference, which means
977 * that niceties like page_folio() don't work. These callers should be
978 * prepared to handle wild return values. For example, PG_head may be
979 * set before _folio_order is initialised, or this may be a tail page.
980 * See compaction.c for some good examples.
982 static inline unsigned int compound_order(struct page
*page
)
984 struct folio
*folio
= (struct folio
*)page
;
986 if (!test_bit(PG_head
, &folio
->flags
))
988 return folio
->_folio_order
;
992 * folio_order - The allocation order of a folio.
995 * A folio is composed of 2^order pages. See get_order() for the definition
998 * Return: The order of the folio.
1000 static inline unsigned int folio_order(struct folio
*folio
)
1002 if (!folio_test_large(folio
))
1004 return folio
->_folio_order
;
1007 #include <linux/huge_mm.h>
1010 * Methods to modify the page usage count.
1012 * What counts for a page usage:
1013 * - cache mapping (page->mapping)
1014 * - private data (page->private)
1015 * - page mapped in a task's page tables, each mapping
1016 * is counted separately
1018 * Also, many kernel routines increase the page count before a critical
1019 * routine so they can be sure the page doesn't go away from under them.
1023 * Drop a ref, return true if the refcount fell to zero (the page has no users)
1025 static inline int put_page_testzero(struct page
*page
)
1027 VM_BUG_ON_PAGE(page_ref_count(page
) == 0, page
);
1028 return page_ref_dec_and_test(page
);
1031 static inline int folio_put_testzero(struct folio
*folio
)
1033 return put_page_testzero(&folio
->page
);
1037 * Try to grab a ref unless the page has a refcount of zero, return false if
1039 * This can be called when MMU is off so it must not access
1040 * any of the virtual mappings.
1042 static inline bool get_page_unless_zero(struct page
*page
)
1044 return page_ref_add_unless(page
, 1, 0);
1047 static inline struct folio
*folio_get_nontail_page(struct page
*page
)
1049 if (unlikely(!get_page_unless_zero(page
)))
1051 return (struct folio
*)page
;
1054 extern int page_is_ram(unsigned long pfn
);
1062 int region_intersects(resource_size_t offset
, size_t size
, unsigned long flags
,
1063 unsigned long desc
);
1065 /* Support for virtually mapped pages */
1066 struct page
*vmalloc_to_page(const void *addr
);
1067 unsigned long vmalloc_to_pfn(const void *addr
);
1070 * Determine if an address is within the vmalloc range
1072 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
1073 * is no special casing required.
1076 #ifndef is_ioremap_addr
1077 #define is_ioremap_addr(x) is_vmalloc_addr(x)
1081 extern bool is_vmalloc_addr(const void *x
);
1082 extern int is_vmalloc_or_module_addr(const void *x
);
1084 static inline bool is_vmalloc_addr(const void *x
)
1088 static inline int is_vmalloc_or_module_addr(const void *x
)
1095 * How many times the entire folio is mapped as a single unit (eg by a
1096 * PMD or PUD entry). This is probably not what you want, except for
1097 * debugging purposes - it does not include PTE-mapped sub-pages; look
1098 * at folio_mapcount() or page_mapcount() or total_mapcount() instead.
1100 static inline int folio_entire_mapcount(struct folio
*folio
)
1102 VM_BUG_ON_FOLIO(!folio_test_large(folio
), folio
);
1103 return atomic_read(&folio
->_entire_mapcount
) + 1;
1107 * The atomic page->_mapcount, starts from -1: so that transitions
1108 * both from it and to it can be tracked, using atomic_inc_and_test
1109 * and atomic_add_negative(-1).
1111 static inline void page_mapcount_reset(struct page
*page
)
1113 atomic_set(&(page
)->_mapcount
, -1);
1117 * page_mapcount() - Number of times this precise page is mapped.
1120 * The number of times this page is mapped. If this page is part of
1121 * a large folio, it includes the number of times this page is mapped
1122 * as part of that folio.
1124 * The result is undefined for pages which cannot be mapped into userspace.
1125 * For example SLAB or special types of pages. See function page_has_type().
1126 * They use this field in struct page differently.
1128 static inline int page_mapcount(struct page
*page
)
1130 int mapcount
= atomic_read(&page
->_mapcount
) + 1;
1132 if (unlikely(PageCompound(page
)))
1133 mapcount
+= folio_entire_mapcount(page_folio(page
));
1138 int folio_total_mapcount(struct folio
*folio
);
1141 * folio_mapcount() - Calculate the number of mappings of this folio.
1142 * @folio: The folio.
1144 * A large folio tracks both how many times the entire folio is mapped,
1145 * and how many times each individual page in the folio is mapped.
1146 * This function calculates the total number of times the folio is
1149 * Return: The number of times this folio is mapped.
1151 static inline int folio_mapcount(struct folio
*folio
)
1153 if (likely(!folio_test_large(folio
)))
1154 return atomic_read(&folio
->_mapcount
) + 1;
1155 return folio_total_mapcount(folio
);
1158 static inline int total_mapcount(struct page
*page
)
1160 if (likely(!PageCompound(page
)))
1161 return atomic_read(&page
->_mapcount
) + 1;
1162 return folio_total_mapcount(page_folio(page
));
1165 static inline bool folio_large_is_mapped(struct folio
*folio
)
1168 * Reading _entire_mapcount below could be omitted if hugetlb
1169 * participated in incrementing nr_pages_mapped when compound mapped.
1171 return atomic_read(&folio
->_nr_pages_mapped
) > 0 ||
1172 atomic_read(&folio
->_entire_mapcount
) >= 0;
1176 * folio_mapped - Is this folio mapped into userspace?
1177 * @folio: The folio.
1179 * Return: True if any page in this folio is referenced by user page tables.
1181 static inline bool folio_mapped(struct folio
*folio
)
1183 if (likely(!folio_test_large(folio
)))
1184 return atomic_read(&folio
->_mapcount
) >= 0;
1185 return folio_large_is_mapped(folio
);
1189 * Return true if this page is mapped into pagetables.
1190 * For compound page it returns true if any sub-page of compound page is mapped,
1191 * even if this particular sub-page is not itself mapped by any PTE or PMD.
1193 static inline bool page_mapped(struct page
*page
)
1195 if (likely(!PageCompound(page
)))
1196 return atomic_read(&page
->_mapcount
) >= 0;
1197 return folio_large_is_mapped(page_folio(page
));
1200 static inline struct page
*virt_to_head_page(const void *x
)
1202 struct page
*page
= virt_to_page(x
);
1204 return compound_head(page
);
1207 static inline struct folio
*virt_to_folio(const void *x
)
1209 struct page
*page
= virt_to_page(x
);
1211 return page_folio(page
);
1214 void __folio_put(struct folio
*folio
);
1216 void put_pages_list(struct list_head
*pages
);
1218 void split_page(struct page
*page
, unsigned int order
);
1219 void folio_copy(struct folio
*dst
, struct folio
*src
);
1221 unsigned long nr_free_buffer_pages(void);
1224 * Compound pages have a destructor function. Provide a
1225 * prototype for that function and accessor functions.
1226 * These are _only_ valid on the head of a compound page.
1228 typedef void compound_page_dtor(struct page
*);
1230 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
1231 enum compound_dtor_id
{
1234 #ifdef CONFIG_HUGETLB_PAGE
1237 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1238 TRANSHUGE_PAGE_DTOR
,
1243 static inline void folio_set_compound_dtor(struct folio
*folio
,
1244 enum compound_dtor_id compound_dtor
)
1246 VM_BUG_ON_FOLIO(compound_dtor
>= NR_COMPOUND_DTORS
, folio
);
1247 folio
->_folio_dtor
= compound_dtor
;
1250 void destroy_large_folio(struct folio
*folio
);
1252 /* Returns the number of bytes in this potentially compound page. */
1253 static inline unsigned long page_size(struct page
*page
)
1255 return PAGE_SIZE
<< compound_order(page
);
1258 /* Returns the number of bits needed for the number of bytes in a page */
1259 static inline unsigned int page_shift(struct page
*page
)
1261 return PAGE_SHIFT
+ compound_order(page
);
1265 * thp_order - Order of a transparent huge page.
1266 * @page: Head page of a transparent huge page.
1268 static inline unsigned int thp_order(struct page
*page
)
1270 VM_BUG_ON_PGFLAGS(PageTail(page
), page
);
1271 return compound_order(page
);
1275 * thp_size - Size of a transparent huge page.
1276 * @page: Head page of a transparent huge page.
1278 * Return: Number of bytes in this page.
1280 static inline unsigned long thp_size(struct page
*page
)
1282 return PAGE_SIZE
<< thp_order(page
);
1285 void free_compound_page(struct page
*page
);
1289 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
1290 * servicing faults for write access. In the normal case, do always want
1291 * pte_mkwrite. But get_user_pages can cause write faults for mappings
1292 * that do not have writing enabled, when used by access_process_vm.
1294 static inline pte_t
maybe_mkwrite(pte_t pte
, struct vm_area_struct
*vma
)
1296 if (likely(vma
->vm_flags
& VM_WRITE
))
1297 pte
= pte_mkwrite(pte
);
1301 vm_fault_t
do_set_pmd(struct vm_fault
*vmf
, struct page
*page
);
1302 void do_set_pte(struct vm_fault
*vmf
, struct page
*page
, unsigned long addr
);
1304 vm_fault_t
finish_fault(struct vm_fault
*vmf
);
1305 vm_fault_t
finish_mkwrite_fault(struct vm_fault
*vmf
);
1309 * Multiple processes may "see" the same page. E.g. for untouched
1310 * mappings of /dev/null, all processes see the same page full of
1311 * zeroes, and text pages of executables and shared libraries have
1312 * only one copy in memory, at most, normally.
1314 * For the non-reserved pages, page_count(page) denotes a reference count.
1315 * page_count() == 0 means the page is free. page->lru is then used for
1316 * freelist management in the buddy allocator.
1317 * page_count() > 0 means the page has been allocated.
1319 * Pages are allocated by the slab allocator in order to provide memory
1320 * to kmalloc and kmem_cache_alloc. In this case, the management of the
1321 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
1322 * unless a particular usage is carefully commented. (the responsibility of
1323 * freeing the kmalloc memory is the caller's, of course).
1325 * A page may be used by anyone else who does a __get_free_page().
1326 * In this case, page_count still tracks the references, and should only
1327 * be used through the normal accessor functions. The top bits of page->flags
1328 * and page->virtual store page management information, but all other fields
1329 * are unused and could be used privately, carefully. The management of this
1330 * page is the responsibility of the one who allocated it, and those who have
1331 * subsequently been given references to it.
1333 * The other pages (we may call them "pagecache pages") are completely
1334 * managed by the Linux memory manager: I/O, buffers, swapping etc.
1335 * The following discussion applies only to them.
1337 * A pagecache page contains an opaque `private' member, which belongs to the
1338 * page's address_space. Usually, this is the address of a circular list of
1339 * the page's disk buffers. PG_private must be set to tell the VM to call
1340 * into the filesystem to release these pages.
1342 * A page may belong to an inode's memory mapping. In this case, page->mapping
1343 * is the pointer to the inode, and page->index is the file offset of the page,
1344 * in units of PAGE_SIZE.
1346 * If pagecache pages are not associated with an inode, they are said to be
1347 * anonymous pages. These may become associated with the swapcache, and in that
1348 * case PG_swapcache is set, and page->private is an offset into the swapcache.
1350 * In either case (swapcache or inode backed), the pagecache itself holds one
1351 * reference to the page. Setting PG_private should also increment the
1352 * refcount. The each user mapping also has a reference to the page.
1354 * The pagecache pages are stored in a per-mapping radix tree, which is
1355 * rooted at mapping->i_pages, and indexed by offset.
1356 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
1357 * lists, we instead now tag pages as dirty/writeback in the radix tree.
1359 * All pagecache pages may be subject to I/O:
1360 * - inode pages may need to be read from disk,
1361 * - inode pages which have been modified and are MAP_SHARED may need
1362 * to be written back to the inode on disk,
1363 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
1364 * modified may need to be swapped out to swap space and (later) to be read
1368 #if defined(CONFIG_ZONE_DEVICE) && defined(CONFIG_FS_DAX)
1369 DECLARE_STATIC_KEY_FALSE(devmap_managed_key
);
1371 bool __put_devmap_managed_page_refs(struct page
*page
, int refs
);
1372 static inline bool put_devmap_managed_page_refs(struct page
*page
, int refs
)
1374 if (!static_branch_unlikely(&devmap_managed_key
))
1376 if (!is_zone_device_page(page
))
1378 return __put_devmap_managed_page_refs(page
, refs
);
1380 #else /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1381 static inline bool put_devmap_managed_page_refs(struct page
*page
, int refs
)
1385 #endif /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1387 static inline bool put_devmap_managed_page(struct page
*page
)
1389 return put_devmap_managed_page_refs(page
, 1);
1392 /* 127: arbitrary random number, small enough to assemble well */
1393 #define folio_ref_zero_or_close_to_overflow(folio) \
1394 ((unsigned int) folio_ref_count(folio) + 127u <= 127u)
1397 * folio_get - Increment the reference count on a folio.
1398 * @folio: The folio.
1400 * Context: May be called in any context, as long as you know that
1401 * you have a refcount on the folio. If you do not already have one,
1402 * folio_try_get() may be the right interface for you to use.
1404 static inline void folio_get(struct folio
*folio
)
1406 VM_BUG_ON_FOLIO(folio_ref_zero_or_close_to_overflow(folio
), folio
);
1407 folio_ref_inc(folio
);
1410 static inline void get_page(struct page
*page
)
1412 folio_get(page_folio(page
));
1415 static inline __must_check
bool try_get_page(struct page
*page
)
1417 page
= compound_head(page
);
1418 if (WARN_ON_ONCE(page_ref_count(page
) <= 0))
1425 * folio_put - Decrement the reference count on a folio.
1426 * @folio: The folio.
1428 * If the folio's reference count reaches zero, the memory will be
1429 * released back to the page allocator and may be used by another
1430 * allocation immediately. Do not access the memory or the struct folio
1431 * after calling folio_put() unless you can be sure that it wasn't the
1434 * Context: May be called in process or interrupt context, but not in NMI
1435 * context. May be called while holding a spinlock.
1437 static inline void folio_put(struct folio
*folio
)
1439 if (folio_put_testzero(folio
))
1444 * folio_put_refs - Reduce the reference count on a folio.
1445 * @folio: The folio.
1446 * @refs: The amount to subtract from the folio's reference count.
1448 * If the folio's reference count reaches zero, the memory will be
1449 * released back to the page allocator and may be used by another
1450 * allocation immediately. Do not access the memory or the struct folio
1451 * after calling folio_put_refs() unless you can be sure that these weren't
1452 * the last references.
1454 * Context: May be called in process or interrupt context, but not in NMI
1455 * context. May be called while holding a spinlock.
1457 static inline void folio_put_refs(struct folio
*folio
, int refs
)
1459 if (folio_ref_sub_and_test(folio
, refs
))
1464 * union release_pages_arg - an array of pages or folios
1466 * release_pages() releases a simple array of multiple pages, and
1467 * accepts various different forms of said page array: either
1468 * a regular old boring array of pages, an array of folios, or
1469 * an array of encoded page pointers.
1471 * The transparent union syntax for this kind of "any of these
1472 * argument types" is all kinds of ugly, so look away.
1475 struct page
**pages
;
1476 struct folio
**folios
;
1477 struct encoded_page
**encoded_pages
;
1478 } release_pages_arg
__attribute__ ((__transparent_union__
));
1480 void release_pages(release_pages_arg
, int nr
);
1483 * folios_put - Decrement the reference count on an array of folios.
1484 * @folios: The folios.
1485 * @nr: How many folios there are.
1487 * Like folio_put(), but for an array of folios. This is more efficient
1488 * than writing the loop yourself as it will optimise the locks which
1489 * need to be taken if the folios are freed.
1491 * Context: May be called in process or interrupt context, but not in NMI
1492 * context. May be called while holding a spinlock.
1494 static inline void folios_put(struct folio
**folios
, unsigned int nr
)
1496 release_pages(folios
, nr
);
1499 static inline void put_page(struct page
*page
)
1501 struct folio
*folio
= page_folio(page
);
1504 * For some devmap managed pages we need to catch refcount transition
1507 if (put_devmap_managed_page(&folio
->page
))
1513 * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
1514 * the page's refcount so that two separate items are tracked: the original page
1515 * reference count, and also a new count of how many pin_user_pages() calls were
1516 * made against the page. ("gup-pinned" is another term for the latter).
1518 * With this scheme, pin_user_pages() becomes special: such pages are marked as
1519 * distinct from normal pages. As such, the unpin_user_page() call (and its
1520 * variants) must be used in order to release gup-pinned pages.
1524 * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
1525 * counts with respect to pin_user_pages() and unpin_user_page() becomes
1526 * simpler, due to the fact that adding an even power of two to the page
1527 * refcount has the effect of using only the upper N bits, for the code that
1528 * counts up using the bias value. This means that the lower bits are left for
1529 * the exclusive use of the original code that increments and decrements by one
1530 * (or at least, by much smaller values than the bias value).
1532 * Of course, once the lower bits overflow into the upper bits (and this is
1533 * OK, because subtraction recovers the original values), then visual inspection
1534 * no longer suffices to directly view the separate counts. However, for normal
1535 * applications that don't have huge page reference counts, this won't be an
1538 * Locking: the lockless algorithm described in folio_try_get_rcu()
1539 * provides safe operation for get_user_pages(), page_mkclean() and
1540 * other calls that race to set up page table entries.
1542 #define GUP_PIN_COUNTING_BIAS (1U << 10)
1544 void unpin_user_page(struct page
*page
);
1545 void unpin_user_pages_dirty_lock(struct page
**pages
, unsigned long npages
,
1547 void unpin_user_page_range_dirty_lock(struct page
*page
, unsigned long npages
,
1549 void unpin_user_pages(struct page
**pages
, unsigned long npages
);
1551 static inline bool is_cow_mapping(vm_flags_t flags
)
1553 return (flags
& (VM_SHARED
| VM_MAYWRITE
)) == VM_MAYWRITE
;
1557 static inline bool is_nommu_shared_mapping(vm_flags_t flags
)
1560 * NOMMU shared mappings are ordinary MAP_SHARED mappings and selected
1561 * R/O MAP_PRIVATE file mappings that are an effective R/O overlay of
1562 * a file mapping. R/O MAP_PRIVATE mappings might still modify
1563 * underlying memory if ptrace is active, so this is only possible if
1564 * ptrace does not apply. Note that there is no mprotect() to upgrade
1565 * write permissions later.
1567 return flags
& (VM_MAYSHARE
| VM_MAYOVERLAY
);
1571 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1572 #define SECTION_IN_PAGE_FLAGS
1576 * The identification function is mainly used by the buddy allocator for
1577 * determining if two pages could be buddies. We are not really identifying
1578 * the zone since we could be using the section number id if we do not have
1579 * node id available in page flags.
1580 * We only guarantee that it will return the same value for two combinable
1583 static inline int page_zone_id(struct page
*page
)
1585 return (page
->flags
>> ZONEID_PGSHIFT
) & ZONEID_MASK
;
1588 #ifdef NODE_NOT_IN_PAGE_FLAGS
1589 extern int page_to_nid(const struct page
*page
);
1591 static inline int page_to_nid(const struct page
*page
)
1593 struct page
*p
= (struct page
*)page
;
1595 return (PF_POISONED_CHECK(p
)->flags
>> NODES_PGSHIFT
) & NODES_MASK
;
1599 static inline int folio_nid(const struct folio
*folio
)
1601 return page_to_nid(&folio
->page
);
1604 #ifdef CONFIG_NUMA_BALANCING
1605 /* page access time bits needs to hold at least 4 seconds */
1606 #define PAGE_ACCESS_TIME_MIN_BITS 12
1607 #if LAST_CPUPID_SHIFT < PAGE_ACCESS_TIME_MIN_BITS
1608 #define PAGE_ACCESS_TIME_BUCKETS \
1609 (PAGE_ACCESS_TIME_MIN_BITS - LAST_CPUPID_SHIFT)
1611 #define PAGE_ACCESS_TIME_BUCKETS 0
1614 #define PAGE_ACCESS_TIME_MASK \
1615 (LAST_CPUPID_MASK << PAGE_ACCESS_TIME_BUCKETS)
1617 static inline int cpu_pid_to_cpupid(int cpu
, int pid
)
1619 return ((cpu
& LAST__CPU_MASK
) << LAST__PID_SHIFT
) | (pid
& LAST__PID_MASK
);
1622 static inline int cpupid_to_pid(int cpupid
)
1624 return cpupid
& LAST__PID_MASK
;
1627 static inline int cpupid_to_cpu(int cpupid
)
1629 return (cpupid
>> LAST__PID_SHIFT
) & LAST__CPU_MASK
;
1632 static inline int cpupid_to_nid(int cpupid
)
1634 return cpu_to_node(cpupid_to_cpu(cpupid
));
1637 static inline bool cpupid_pid_unset(int cpupid
)
1639 return cpupid_to_pid(cpupid
) == (-1 & LAST__PID_MASK
);
1642 static inline bool cpupid_cpu_unset(int cpupid
)
1644 return cpupid_to_cpu(cpupid
) == (-1 & LAST__CPU_MASK
);
1647 static inline bool __cpupid_match_pid(pid_t task_pid
, int cpupid
)
1649 return (task_pid
& LAST__PID_MASK
) == cpupid_to_pid(cpupid
);
1652 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1653 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1654 static inline int page_cpupid_xchg_last(struct page
*page
, int cpupid
)
1656 return xchg(&page
->_last_cpupid
, cpupid
& LAST_CPUPID_MASK
);
1659 static inline int page_cpupid_last(struct page
*page
)
1661 return page
->_last_cpupid
;
1663 static inline void page_cpupid_reset_last(struct page
*page
)
1665 page
->_last_cpupid
= -1 & LAST_CPUPID_MASK
;
1668 static inline int page_cpupid_last(struct page
*page
)
1670 return (page
->flags
>> LAST_CPUPID_PGSHIFT
) & LAST_CPUPID_MASK
;
1673 extern int page_cpupid_xchg_last(struct page
*page
, int cpupid
);
1675 static inline void page_cpupid_reset_last(struct page
*page
)
1677 page
->flags
|= LAST_CPUPID_MASK
<< LAST_CPUPID_PGSHIFT
;
1679 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1681 static inline int xchg_page_access_time(struct page
*page
, int time
)
1685 last_time
= page_cpupid_xchg_last(page
, time
>> PAGE_ACCESS_TIME_BUCKETS
);
1686 return last_time
<< PAGE_ACCESS_TIME_BUCKETS
;
1689 static inline void vma_set_access_pid_bit(struct vm_area_struct
*vma
)
1691 unsigned int pid_bit
;
1693 pid_bit
= hash_32(current
->pid
, ilog2(BITS_PER_LONG
));
1694 if (vma
->numab_state
&& !test_bit(pid_bit
, &vma
->numab_state
->access_pids
[1])) {
1695 __set_bit(pid_bit
, &vma
->numab_state
->access_pids
[1]);
1698 #else /* !CONFIG_NUMA_BALANCING */
1699 static inline int page_cpupid_xchg_last(struct page
*page
, int cpupid
)
1701 return page_to_nid(page
); /* XXX */
1704 static inline int xchg_page_access_time(struct page
*page
, int time
)
1709 static inline int page_cpupid_last(struct page
*page
)
1711 return page_to_nid(page
); /* XXX */
1714 static inline int cpupid_to_nid(int cpupid
)
1719 static inline int cpupid_to_pid(int cpupid
)
1724 static inline int cpupid_to_cpu(int cpupid
)
1729 static inline int cpu_pid_to_cpupid(int nid
, int pid
)
1734 static inline bool cpupid_pid_unset(int cpupid
)
1739 static inline void page_cpupid_reset_last(struct page
*page
)
1743 static inline bool cpupid_match_pid(struct task_struct
*task
, int cpupid
)
1748 static inline void vma_set_access_pid_bit(struct vm_area_struct
*vma
)
1751 #endif /* CONFIG_NUMA_BALANCING */
1753 #if defined(CONFIG_KASAN_SW_TAGS) || defined(CONFIG_KASAN_HW_TAGS)
1756 * KASAN per-page tags are stored xor'ed with 0xff. This allows to avoid
1757 * setting tags for all pages to native kernel tag value 0xff, as the default
1758 * value 0x00 maps to 0xff.
1761 static inline u8
page_kasan_tag(const struct page
*page
)
1765 if (kasan_enabled()) {
1766 tag
= (page
->flags
>> KASAN_TAG_PGSHIFT
) & KASAN_TAG_MASK
;
1773 static inline void page_kasan_tag_set(struct page
*page
, u8 tag
)
1775 unsigned long old_flags
, flags
;
1777 if (!kasan_enabled())
1781 old_flags
= READ_ONCE(page
->flags
);
1784 flags
&= ~(KASAN_TAG_MASK
<< KASAN_TAG_PGSHIFT
);
1785 flags
|= (tag
& KASAN_TAG_MASK
) << KASAN_TAG_PGSHIFT
;
1786 } while (unlikely(!try_cmpxchg(&page
->flags
, &old_flags
, flags
)));
1789 static inline void page_kasan_tag_reset(struct page
*page
)
1791 if (kasan_enabled())
1792 page_kasan_tag_set(page
, 0xff);
1795 #else /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1797 static inline u8
page_kasan_tag(const struct page
*page
)
1802 static inline void page_kasan_tag_set(struct page
*page
, u8 tag
) { }
1803 static inline void page_kasan_tag_reset(struct page
*page
) { }
1805 #endif /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1807 static inline struct zone
*page_zone(const struct page
*page
)
1809 return &NODE_DATA(page_to_nid(page
))->node_zones
[page_zonenum(page
)];
1812 static inline pg_data_t
*page_pgdat(const struct page
*page
)
1814 return NODE_DATA(page_to_nid(page
));
1817 static inline struct zone
*folio_zone(const struct folio
*folio
)
1819 return page_zone(&folio
->page
);
1822 static inline pg_data_t
*folio_pgdat(const struct folio
*folio
)
1824 return page_pgdat(&folio
->page
);
1827 #ifdef SECTION_IN_PAGE_FLAGS
1828 static inline void set_page_section(struct page
*page
, unsigned long section
)
1830 page
->flags
&= ~(SECTIONS_MASK
<< SECTIONS_PGSHIFT
);
1831 page
->flags
|= (section
& SECTIONS_MASK
) << SECTIONS_PGSHIFT
;
1834 static inline unsigned long page_to_section(const struct page
*page
)
1836 return (page
->flags
>> SECTIONS_PGSHIFT
) & SECTIONS_MASK
;
1841 * folio_pfn - Return the Page Frame Number of a folio.
1842 * @folio: The folio.
1844 * A folio may contain multiple pages. The pages have consecutive
1845 * Page Frame Numbers.
1847 * Return: The Page Frame Number of the first page in the folio.
1849 static inline unsigned long folio_pfn(struct folio
*folio
)
1851 return page_to_pfn(&folio
->page
);
1854 static inline struct folio
*pfn_folio(unsigned long pfn
)
1856 return page_folio(pfn_to_page(pfn
));
1860 * folio_maybe_dma_pinned - Report if a folio may be pinned for DMA.
1861 * @folio: The folio.
1863 * This function checks if a folio has been pinned via a call to
1864 * a function in the pin_user_pages() family.
1866 * For small folios, the return value is partially fuzzy: false is not fuzzy,
1867 * because it means "definitely not pinned for DMA", but true means "probably
1868 * pinned for DMA, but possibly a false positive due to having at least
1869 * GUP_PIN_COUNTING_BIAS worth of normal folio references".
1871 * False positives are OK, because: a) it's unlikely for a folio to
1872 * get that many refcounts, and b) all the callers of this routine are
1873 * expected to be able to deal gracefully with a false positive.
1875 * For large folios, the result will be exactly correct. That's because
1876 * we have more tracking data available: the _pincount field is used
1877 * instead of the GUP_PIN_COUNTING_BIAS scheme.
1879 * For more information, please see Documentation/core-api/pin_user_pages.rst.
1881 * Return: True, if it is likely that the page has been "dma-pinned".
1882 * False, if the page is definitely not dma-pinned.
1884 static inline bool folio_maybe_dma_pinned(struct folio
*folio
)
1886 if (folio_test_large(folio
))
1887 return atomic_read(&folio
->_pincount
) > 0;
1890 * folio_ref_count() is signed. If that refcount overflows, then
1891 * folio_ref_count() returns a negative value, and callers will avoid
1892 * further incrementing the refcount.
1894 * Here, for that overflow case, use the sign bit to count a little
1895 * bit higher via unsigned math, and thus still get an accurate result.
1897 return ((unsigned int)folio_ref_count(folio
)) >=
1898 GUP_PIN_COUNTING_BIAS
;
1901 static inline bool page_maybe_dma_pinned(struct page
*page
)
1903 return folio_maybe_dma_pinned(page_folio(page
));
1907 * This should most likely only be called during fork() to see whether we
1908 * should break the cow immediately for an anon page on the src mm.
1910 * The caller has to hold the PT lock and the vma->vm_mm->->write_protect_seq.
1912 static inline bool page_needs_cow_for_dma(struct vm_area_struct
*vma
,
1915 VM_BUG_ON(!(raw_read_seqcount(&vma
->vm_mm
->write_protect_seq
) & 1));
1917 if (!test_bit(MMF_HAS_PINNED
, &vma
->vm_mm
->flags
))
1920 return page_maybe_dma_pinned(page
);
1924 * is_zero_page - Query if a page is a zero page
1925 * @page: The page to query
1927 * This returns true if @page is one of the permanent zero pages.
1929 static inline bool is_zero_page(const struct page
*page
)
1931 return is_zero_pfn(page_to_pfn(page
));
1935 * is_zero_folio - Query if a folio is a zero page
1936 * @folio: The folio to query
1938 * This returns true if @folio is one of the permanent zero pages.
1940 static inline bool is_zero_folio(const struct folio
*folio
)
1942 return is_zero_page(&folio
->page
);
1945 /* MIGRATE_CMA and ZONE_MOVABLE do not allow pin folios */
1946 #ifdef CONFIG_MIGRATION
1947 static inline bool folio_is_longterm_pinnable(struct folio
*folio
)
1950 int mt
= folio_migratetype(folio
);
1952 if (mt
== MIGRATE_CMA
|| mt
== MIGRATE_ISOLATE
)
1955 /* The zero page can be "pinned" but gets special handling. */
1956 if (is_zero_folio(folio
))
1959 /* Coherent device memory must always allow eviction. */
1960 if (folio_is_device_coherent(folio
))
1963 /* Otherwise, non-movable zone folios can be pinned. */
1964 return !folio_is_zone_movable(folio
);
1968 static inline bool folio_is_longterm_pinnable(struct folio
*folio
)
1974 static inline void set_page_zone(struct page
*page
, enum zone_type zone
)
1976 page
->flags
&= ~(ZONES_MASK
<< ZONES_PGSHIFT
);
1977 page
->flags
|= (zone
& ZONES_MASK
) << ZONES_PGSHIFT
;
1980 static inline void set_page_node(struct page
*page
, unsigned long node
)
1982 page
->flags
&= ~(NODES_MASK
<< NODES_PGSHIFT
);
1983 page
->flags
|= (node
& NODES_MASK
) << NODES_PGSHIFT
;
1986 static inline void set_page_links(struct page
*page
, enum zone_type zone
,
1987 unsigned long node
, unsigned long pfn
)
1989 set_page_zone(page
, zone
);
1990 set_page_node(page
, node
);
1991 #ifdef SECTION_IN_PAGE_FLAGS
1992 set_page_section(page
, pfn_to_section_nr(pfn
));
1997 * folio_nr_pages - The number of pages in the folio.
1998 * @folio: The folio.
2000 * Return: A positive power of two.
2002 static inline long folio_nr_pages(struct folio
*folio
)
2004 if (!folio_test_large(folio
))
2007 return folio
->_folio_nr_pages
;
2009 return 1L << folio
->_folio_order
;
2014 * compound_nr() returns the number of pages in this potentially compound
2015 * page. compound_nr() can be called on a tail page, and is defined to
2016 * return 1 in that case.
2018 static inline unsigned long compound_nr(struct page
*page
)
2020 struct folio
*folio
= (struct folio
*)page
;
2022 if (!test_bit(PG_head
, &folio
->flags
))
2025 return folio
->_folio_nr_pages
;
2027 return 1L << folio
->_folio_order
;
2032 * thp_nr_pages - The number of regular pages in this huge page.
2033 * @page: The head page of a huge page.
2035 static inline int thp_nr_pages(struct page
*page
)
2037 return folio_nr_pages((struct folio
*)page
);
2041 * folio_next - Move to the next physical folio.
2042 * @folio: The folio we're currently operating on.
2044 * If you have physically contiguous memory which may span more than
2045 * one folio (eg a &struct bio_vec), use this function to move from one
2046 * folio to the next. Do not use it if the memory is only virtually
2047 * contiguous as the folios are almost certainly not adjacent to each
2048 * other. This is the folio equivalent to writing ``page++``.
2050 * Context: We assume that the folios are refcounted and/or locked at a
2051 * higher level and do not adjust the reference counts.
2052 * Return: The next struct folio.
2054 static inline struct folio
*folio_next(struct folio
*folio
)
2056 return (struct folio
*)folio_page(folio
, folio_nr_pages(folio
));
2060 * folio_shift - The size of the memory described by this folio.
2061 * @folio: The folio.
2063 * A folio represents a number of bytes which is a power-of-two in size.
2064 * This function tells you which power-of-two the folio is. See also
2065 * folio_size() and folio_order().
2067 * Context: The caller should have a reference on the folio to prevent
2068 * it from being split. It is not necessary for the folio to be locked.
2069 * Return: The base-2 logarithm of the size of this folio.
2071 static inline unsigned int folio_shift(struct folio
*folio
)
2073 return PAGE_SHIFT
+ folio_order(folio
);
2077 * folio_size - The number of bytes in a folio.
2078 * @folio: The folio.
2080 * Context: The caller should have a reference on the folio to prevent
2081 * it from being split. It is not necessary for the folio to be locked.
2082 * Return: The number of bytes in this folio.
2084 static inline size_t folio_size(struct folio
*folio
)
2086 return PAGE_SIZE
<< folio_order(folio
);
2090 * folio_estimated_sharers - Estimate the number of sharers of a folio.
2091 * @folio: The folio.
2093 * folio_estimated_sharers() aims to serve as a function to efficiently
2094 * estimate the number of processes sharing a folio. This is done by
2095 * looking at the precise mapcount of the first subpage in the folio, and
2096 * assuming the other subpages are the same. This may not be true for large
2097 * folios. If you want exact mapcounts for exact calculations, look at
2098 * page_mapcount() or folio_total_mapcount().
2100 * Return: The estimated number of processes sharing a folio.
2102 static inline int folio_estimated_sharers(struct folio
*folio
)
2104 return page_mapcount(folio_page(folio
, 0));
2107 #ifndef HAVE_ARCH_MAKE_PAGE_ACCESSIBLE
2108 static inline int arch_make_page_accessible(struct page
*page
)
2114 #ifndef HAVE_ARCH_MAKE_FOLIO_ACCESSIBLE
2115 static inline int arch_make_folio_accessible(struct folio
*folio
)
2118 long i
, nr
= folio_nr_pages(folio
);
2120 for (i
= 0; i
< nr
; i
++) {
2121 ret
= arch_make_page_accessible(folio_page(folio
, i
));
2131 * Some inline functions in vmstat.h depend on page_zone()
2133 #include <linux/vmstat.h>
2135 static __always_inline
void *lowmem_page_address(const struct page
*page
)
2137 return page_to_virt(page
);
2140 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
2141 #define HASHED_PAGE_VIRTUAL
2144 #if defined(WANT_PAGE_VIRTUAL)
2145 static inline void *page_address(const struct page
*page
)
2147 return page
->virtual;
2149 static inline void set_page_address(struct page
*page
, void *address
)
2151 page
->virtual = address
;
2153 #define page_address_init() do { } while(0)
2156 #if defined(HASHED_PAGE_VIRTUAL)
2157 void *page_address(const struct page
*page
);
2158 void set_page_address(struct page
*page
, void *virtual);
2159 void page_address_init(void);
2162 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
2163 #define page_address(page) lowmem_page_address(page)
2164 #define set_page_address(page, address) do { } while(0)
2165 #define page_address_init() do { } while(0)
2168 static inline void *folio_address(const struct folio
*folio
)
2170 return page_address(&folio
->page
);
2173 extern void *page_rmapping(struct page
*page
);
2174 extern pgoff_t
__page_file_index(struct page
*page
);
2177 * Return the pagecache index of the passed page. Regular pagecache pages
2178 * use ->index whereas swapcache pages use swp_offset(->private)
2180 static inline pgoff_t
page_index(struct page
*page
)
2182 if (unlikely(PageSwapCache(page
)))
2183 return __page_file_index(page
);
2188 * Return true only if the page has been allocated with
2189 * ALLOC_NO_WATERMARKS and the low watermark was not
2190 * met implying that the system is under some pressure.
2192 static inline bool page_is_pfmemalloc(const struct page
*page
)
2195 * lru.next has bit 1 set if the page is allocated from the
2196 * pfmemalloc reserves. Callers may simply overwrite it if
2197 * they do not need to preserve that information.
2199 return (uintptr_t)page
->lru
.next
& BIT(1);
2203 * Return true only if the folio has been allocated with
2204 * ALLOC_NO_WATERMARKS and the low watermark was not
2205 * met implying that the system is under some pressure.
2207 static inline bool folio_is_pfmemalloc(const struct folio
*folio
)
2210 * lru.next has bit 1 set if the page is allocated from the
2211 * pfmemalloc reserves. Callers may simply overwrite it if
2212 * they do not need to preserve that information.
2214 return (uintptr_t)folio
->lru
.next
& BIT(1);
2218 * Only to be called by the page allocator on a freshly allocated
2221 static inline void set_page_pfmemalloc(struct page
*page
)
2223 page
->lru
.next
= (void *)BIT(1);
2226 static inline void clear_page_pfmemalloc(struct page
*page
)
2228 page
->lru
.next
= NULL
;
2232 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
2234 extern void pagefault_out_of_memory(void);
2236 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
2237 #define offset_in_thp(page, p) ((unsigned long)(p) & (thp_size(page) - 1))
2238 #define offset_in_folio(folio, p) ((unsigned long)(p) & (folio_size(folio) - 1))
2241 * Flags passed to show_mem() and show_free_areas() to suppress output in
2244 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
2246 extern void __show_free_areas(unsigned int flags
, nodemask_t
*nodemask
, int max_zone_idx
);
2247 static void __maybe_unused
show_free_areas(unsigned int flags
, nodemask_t
*nodemask
)
2249 __show_free_areas(flags
, nodemask
, MAX_NR_ZONES
- 1);
2253 * Parameter block passed down to zap_pte_range in exceptional cases.
2255 struct zap_details
{
2256 struct folio
*single_folio
; /* Locked folio to be unmapped */
2257 bool even_cows
; /* Zap COWed private pages too? */
2258 zap_flags_t zap_flags
; /* Extra flags for zapping */
2262 * Whether to drop the pte markers, for example, the uffd-wp information for
2263 * file-backed memory. This should only be specified when we will completely
2264 * drop the page in the mm, either by truncation or unmapping of the vma. By
2265 * default, the flag is not set.
2267 #define ZAP_FLAG_DROP_MARKER ((__force zap_flags_t) BIT(0))
2268 /* Set in unmap_vmas() to indicate a final unmap call. Only used by hugetlb */
2269 #define ZAP_FLAG_UNMAP ((__force zap_flags_t) BIT(1))
2271 #ifdef CONFIG_SCHED_MM_CID
2272 void sched_mm_cid_before_execve(struct task_struct
*t
);
2273 void sched_mm_cid_after_execve(struct task_struct
*t
);
2274 void sched_mm_cid_fork(struct task_struct
*t
);
2275 void sched_mm_cid_exit_signals(struct task_struct
*t
);
2276 static inline int task_mm_cid(struct task_struct
*t
)
2281 static inline void sched_mm_cid_before_execve(struct task_struct
*t
) { }
2282 static inline void sched_mm_cid_after_execve(struct task_struct
*t
) { }
2283 static inline void sched_mm_cid_fork(struct task_struct
*t
) { }
2284 static inline void sched_mm_cid_exit_signals(struct task_struct
*t
) { }
2285 static inline int task_mm_cid(struct task_struct
*t
)
2288 * Use the processor id as a fall-back when the mm cid feature is
2289 * disabled. This provides functional per-cpu data structure accesses
2290 * in user-space, althrough it won't provide the memory usage benefits.
2292 return raw_smp_processor_id();
2297 extern bool can_do_mlock(void);
2299 static inline bool can_do_mlock(void) { return false; }
2301 extern int user_shm_lock(size_t, struct ucounts
*);
2302 extern void user_shm_unlock(size_t, struct ucounts
*);
2304 struct folio
*vm_normal_folio(struct vm_area_struct
*vma
, unsigned long addr
,
2306 struct page
*vm_normal_page(struct vm_area_struct
*vma
, unsigned long addr
,
2308 struct page
*vm_normal_page_pmd(struct vm_area_struct
*vma
, unsigned long addr
,
2311 void zap_vma_ptes(struct vm_area_struct
*vma
, unsigned long address
,
2312 unsigned long size
);
2313 void zap_page_range_single(struct vm_area_struct
*vma
, unsigned long address
,
2314 unsigned long size
, struct zap_details
*details
);
2315 static inline void zap_vma_pages(struct vm_area_struct
*vma
)
2317 zap_page_range_single(vma
, vma
->vm_start
,
2318 vma
->vm_end
- vma
->vm_start
, NULL
);
2320 void unmap_vmas(struct mmu_gather
*tlb
, struct maple_tree
*mt
,
2321 struct vm_area_struct
*start_vma
, unsigned long start
,
2322 unsigned long end
, bool mm_wr_locked
);
2324 struct mmu_notifier_range
;
2326 void free_pgd_range(struct mmu_gather
*tlb
, unsigned long addr
,
2327 unsigned long end
, unsigned long floor
, unsigned long ceiling
);
2329 copy_page_range(struct vm_area_struct
*dst_vma
, struct vm_area_struct
*src_vma
);
2330 int follow_pte(struct mm_struct
*mm
, unsigned long address
,
2331 pte_t
**ptepp
, spinlock_t
**ptlp
);
2332 int follow_pfn(struct vm_area_struct
*vma
, unsigned long address
,
2333 unsigned long *pfn
);
2334 int follow_phys(struct vm_area_struct
*vma
, unsigned long address
,
2335 unsigned int flags
, unsigned long *prot
, resource_size_t
*phys
);
2336 int generic_access_phys(struct vm_area_struct
*vma
, unsigned long addr
,
2337 void *buf
, int len
, int write
);
2339 extern void truncate_pagecache(struct inode
*inode
, loff_t
new);
2340 extern void truncate_setsize(struct inode
*inode
, loff_t newsize
);
2341 void pagecache_isize_extended(struct inode
*inode
, loff_t from
, loff_t to
);
2342 void truncate_pagecache_range(struct inode
*inode
, loff_t offset
, loff_t end
);
2343 int generic_error_remove_page(struct address_space
*mapping
, struct page
*page
);
2345 struct vm_area_struct
*lock_mm_and_find_vma(struct mm_struct
*mm
,
2346 unsigned long address
, struct pt_regs
*regs
);
2349 extern vm_fault_t
handle_mm_fault(struct vm_area_struct
*vma
,
2350 unsigned long address
, unsigned int flags
,
2351 struct pt_regs
*regs
);
2352 extern int fixup_user_fault(struct mm_struct
*mm
,
2353 unsigned long address
, unsigned int fault_flags
,
2355 void unmap_mapping_pages(struct address_space
*mapping
,
2356 pgoff_t start
, pgoff_t nr
, bool even_cows
);
2357 void unmap_mapping_range(struct address_space
*mapping
,
2358 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
);
2360 static inline vm_fault_t
handle_mm_fault(struct vm_area_struct
*vma
,
2361 unsigned long address
, unsigned int flags
,
2362 struct pt_regs
*regs
)
2364 /* should never happen if there's no MMU */
2366 return VM_FAULT_SIGBUS
;
2368 static inline int fixup_user_fault(struct mm_struct
*mm
, unsigned long address
,
2369 unsigned int fault_flags
, bool *unlocked
)
2371 /* should never happen if there's no MMU */
2375 static inline void unmap_mapping_pages(struct address_space
*mapping
,
2376 pgoff_t start
, pgoff_t nr
, bool even_cows
) { }
2377 static inline void unmap_mapping_range(struct address_space
*mapping
,
2378 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
) { }
2381 static inline void unmap_shared_mapping_range(struct address_space
*mapping
,
2382 loff_t
const holebegin
, loff_t
const holelen
)
2384 unmap_mapping_range(mapping
, holebegin
, holelen
, 0);
2387 static inline struct vm_area_struct
*vma_lookup(struct mm_struct
*mm
,
2388 unsigned long addr
);
2390 extern int access_process_vm(struct task_struct
*tsk
, unsigned long addr
,
2391 void *buf
, int len
, unsigned int gup_flags
);
2392 extern int access_remote_vm(struct mm_struct
*mm
, unsigned long addr
,
2393 void *buf
, int len
, unsigned int gup_flags
);
2394 extern int __access_remote_vm(struct mm_struct
*mm
, unsigned long addr
,
2395 void *buf
, int len
, unsigned int gup_flags
);
2397 long get_user_pages_remote(struct mm_struct
*mm
,
2398 unsigned long start
, unsigned long nr_pages
,
2399 unsigned int gup_flags
, struct page
**pages
,
2401 long pin_user_pages_remote(struct mm_struct
*mm
,
2402 unsigned long start
, unsigned long nr_pages
,
2403 unsigned int gup_flags
, struct page
**pages
,
2406 static inline struct page
*get_user_page_vma_remote(struct mm_struct
*mm
,
2409 struct vm_area_struct
**vmap
)
2412 struct vm_area_struct
*vma
;
2413 int got
= get_user_pages_remote(mm
, addr
, 1, gup_flags
, &page
, NULL
);
2416 return ERR_PTR(got
);
2420 vma
= vma_lookup(mm
, addr
);
2421 if (WARN_ON_ONCE(!vma
)) {
2423 return ERR_PTR(-EINVAL
);
2430 long get_user_pages(unsigned long start
, unsigned long nr_pages
,
2431 unsigned int gup_flags
, struct page
**pages
);
2432 long pin_user_pages(unsigned long start
, unsigned long nr_pages
,
2433 unsigned int gup_flags
, struct page
**pages
);
2434 long get_user_pages_unlocked(unsigned long start
, unsigned long nr_pages
,
2435 struct page
**pages
, unsigned int gup_flags
);
2436 long pin_user_pages_unlocked(unsigned long start
, unsigned long nr_pages
,
2437 struct page
**pages
, unsigned int gup_flags
);
2439 int get_user_pages_fast(unsigned long start
, int nr_pages
,
2440 unsigned int gup_flags
, struct page
**pages
);
2441 int pin_user_pages_fast(unsigned long start
, int nr_pages
,
2442 unsigned int gup_flags
, struct page
**pages
);
2443 void folio_add_pin(struct folio
*folio
);
2445 int account_locked_vm(struct mm_struct
*mm
, unsigned long pages
, bool inc
);
2446 int __account_locked_vm(struct mm_struct
*mm
, unsigned long pages
, bool inc
,
2447 struct task_struct
*task
, bool bypass_rlim
);
2450 struct page
*get_dump_page(unsigned long addr
);
2452 bool folio_mark_dirty(struct folio
*folio
);
2453 bool set_page_dirty(struct page
*page
);
2454 int set_page_dirty_lock(struct page
*page
);
2456 int get_cmdline(struct task_struct
*task
, char *buffer
, int buflen
);
2458 extern unsigned long move_page_tables(struct vm_area_struct
*vma
,
2459 unsigned long old_addr
, struct vm_area_struct
*new_vma
,
2460 unsigned long new_addr
, unsigned long len
,
2461 bool need_rmap_locks
);
2464 * Flags used by change_protection(). For now we make it a bitmap so
2465 * that we can pass in multiple flags just like parameters. However
2466 * for now all the callers are only use one of the flags at the same
2470 * Whether we should manually check if we can map individual PTEs writable,
2471 * because something (e.g., COW, uffd-wp) blocks that from happening for all
2472 * PTEs automatically in a writable mapping.
2474 #define MM_CP_TRY_CHANGE_WRITABLE (1UL << 0)
2475 /* Whether this protection change is for NUMA hints */
2476 #define MM_CP_PROT_NUMA (1UL << 1)
2477 /* Whether this change is for write protecting */
2478 #define MM_CP_UFFD_WP (1UL << 2) /* do wp */
2479 #define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */
2480 #define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \
2481 MM_CP_UFFD_WP_RESOLVE)
2483 bool vma_needs_dirty_tracking(struct vm_area_struct
*vma
);
2484 int vma_wants_writenotify(struct vm_area_struct
*vma
, pgprot_t vm_page_prot
);
2485 static inline bool vma_wants_manual_pte_write_upgrade(struct vm_area_struct
*vma
)
2488 * We want to check manually if we can change individual PTEs writable
2489 * if we can't do that automatically for all PTEs in a mapping. For
2490 * private mappings, that's always the case when we have write
2491 * permissions as we properly have to handle COW.
2493 if (vma
->vm_flags
& VM_SHARED
)
2494 return vma_wants_writenotify(vma
, vma
->vm_page_prot
);
2495 return !!(vma
->vm_flags
& VM_WRITE
);
2498 bool can_change_pte_writable(struct vm_area_struct
*vma
, unsigned long addr
,
2500 extern long change_protection(struct mmu_gather
*tlb
,
2501 struct vm_area_struct
*vma
, unsigned long start
,
2502 unsigned long end
, unsigned long cp_flags
);
2503 extern int mprotect_fixup(struct vma_iterator
*vmi
, struct mmu_gather
*tlb
,
2504 struct vm_area_struct
*vma
, struct vm_area_struct
**pprev
,
2505 unsigned long start
, unsigned long end
, unsigned long newflags
);
2508 * doesn't attempt to fault and will return short.
2510 int get_user_pages_fast_only(unsigned long start
, int nr_pages
,
2511 unsigned int gup_flags
, struct page
**pages
);
2513 static inline bool get_user_page_fast_only(unsigned long addr
,
2514 unsigned int gup_flags
, struct page
**pagep
)
2516 return get_user_pages_fast_only(addr
, 1, gup_flags
, pagep
) == 1;
2519 * per-process(per-mm_struct) statistics.
2521 static inline unsigned long get_mm_counter(struct mm_struct
*mm
, int member
)
2523 return percpu_counter_read_positive(&mm
->rss_stat
[member
]);
2526 void mm_trace_rss_stat(struct mm_struct
*mm
, int member
);
2528 static inline void add_mm_counter(struct mm_struct
*mm
, int member
, long value
)
2530 percpu_counter_add(&mm
->rss_stat
[member
], value
);
2532 mm_trace_rss_stat(mm
, member
);
2535 static inline void inc_mm_counter(struct mm_struct
*mm
, int member
)
2537 percpu_counter_inc(&mm
->rss_stat
[member
]);
2539 mm_trace_rss_stat(mm
, member
);
2542 static inline void dec_mm_counter(struct mm_struct
*mm
, int member
)
2544 percpu_counter_dec(&mm
->rss_stat
[member
]);
2546 mm_trace_rss_stat(mm
, member
);
2549 /* Optimized variant when page is already known not to be PageAnon */
2550 static inline int mm_counter_file(struct page
*page
)
2552 if (PageSwapBacked(page
))
2553 return MM_SHMEMPAGES
;
2554 return MM_FILEPAGES
;
2557 static inline int mm_counter(struct page
*page
)
2560 return MM_ANONPAGES
;
2561 return mm_counter_file(page
);
2564 static inline unsigned long get_mm_rss(struct mm_struct
*mm
)
2566 return get_mm_counter(mm
, MM_FILEPAGES
) +
2567 get_mm_counter(mm
, MM_ANONPAGES
) +
2568 get_mm_counter(mm
, MM_SHMEMPAGES
);
2571 static inline unsigned long get_mm_hiwater_rss(struct mm_struct
*mm
)
2573 return max(mm
->hiwater_rss
, get_mm_rss(mm
));
2576 static inline unsigned long get_mm_hiwater_vm(struct mm_struct
*mm
)
2578 return max(mm
->hiwater_vm
, mm
->total_vm
);
2581 static inline void update_hiwater_rss(struct mm_struct
*mm
)
2583 unsigned long _rss
= get_mm_rss(mm
);
2585 if ((mm
)->hiwater_rss
< _rss
)
2586 (mm
)->hiwater_rss
= _rss
;
2589 static inline void update_hiwater_vm(struct mm_struct
*mm
)
2591 if (mm
->hiwater_vm
< mm
->total_vm
)
2592 mm
->hiwater_vm
= mm
->total_vm
;
2595 static inline void reset_mm_hiwater_rss(struct mm_struct
*mm
)
2597 mm
->hiwater_rss
= get_mm_rss(mm
);
2600 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss
,
2601 struct mm_struct
*mm
)
2603 unsigned long hiwater_rss
= get_mm_hiwater_rss(mm
);
2605 if (*maxrss
< hiwater_rss
)
2606 *maxrss
= hiwater_rss
;
2609 #if defined(SPLIT_RSS_COUNTING)
2610 void sync_mm_rss(struct mm_struct
*mm
);
2612 static inline void sync_mm_rss(struct mm_struct
*mm
)
2617 #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
2618 static inline int pte_special(pte_t pte
)
2623 static inline pte_t
pte_mkspecial(pte_t pte
)
2629 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
2630 static inline int pte_devmap(pte_t pte
)
2636 extern pte_t
*__get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
2638 static inline pte_t
*get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
2642 __cond_lock(*ptl
, ptep
= __get_locked_pte(mm
, addr
, ptl
));
2646 #ifdef __PAGETABLE_P4D_FOLDED
2647 static inline int __p4d_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
2648 unsigned long address
)
2653 int __p4d_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
);
2656 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
2657 static inline int __pud_alloc(struct mm_struct
*mm
, p4d_t
*p4d
,
2658 unsigned long address
)
2662 static inline void mm_inc_nr_puds(struct mm_struct
*mm
) {}
2663 static inline void mm_dec_nr_puds(struct mm_struct
*mm
) {}
2666 int __pud_alloc(struct mm_struct
*mm
, p4d_t
*p4d
, unsigned long address
);
2668 static inline void mm_inc_nr_puds(struct mm_struct
*mm
)
2670 if (mm_pud_folded(mm
))
2672 atomic_long_add(PTRS_PER_PUD
* sizeof(pud_t
), &mm
->pgtables_bytes
);
2675 static inline void mm_dec_nr_puds(struct mm_struct
*mm
)
2677 if (mm_pud_folded(mm
))
2679 atomic_long_sub(PTRS_PER_PUD
* sizeof(pud_t
), &mm
->pgtables_bytes
);
2683 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
2684 static inline int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
,
2685 unsigned long address
)
2690 static inline void mm_inc_nr_pmds(struct mm_struct
*mm
) {}
2691 static inline void mm_dec_nr_pmds(struct mm_struct
*mm
) {}
2694 int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
);
2696 static inline void mm_inc_nr_pmds(struct mm_struct
*mm
)
2698 if (mm_pmd_folded(mm
))
2700 atomic_long_add(PTRS_PER_PMD
* sizeof(pmd_t
), &mm
->pgtables_bytes
);
2703 static inline void mm_dec_nr_pmds(struct mm_struct
*mm
)
2705 if (mm_pmd_folded(mm
))
2707 atomic_long_sub(PTRS_PER_PMD
* sizeof(pmd_t
), &mm
->pgtables_bytes
);
2712 static inline void mm_pgtables_bytes_init(struct mm_struct
*mm
)
2714 atomic_long_set(&mm
->pgtables_bytes
, 0);
2717 static inline unsigned long mm_pgtables_bytes(const struct mm_struct
*mm
)
2719 return atomic_long_read(&mm
->pgtables_bytes
);
2722 static inline void mm_inc_nr_ptes(struct mm_struct
*mm
)
2724 atomic_long_add(PTRS_PER_PTE
* sizeof(pte_t
), &mm
->pgtables_bytes
);
2727 static inline void mm_dec_nr_ptes(struct mm_struct
*mm
)
2729 atomic_long_sub(PTRS_PER_PTE
* sizeof(pte_t
), &mm
->pgtables_bytes
);
2733 static inline void mm_pgtables_bytes_init(struct mm_struct
*mm
) {}
2734 static inline unsigned long mm_pgtables_bytes(const struct mm_struct
*mm
)
2739 static inline void mm_inc_nr_ptes(struct mm_struct
*mm
) {}
2740 static inline void mm_dec_nr_ptes(struct mm_struct
*mm
) {}
2743 int __pte_alloc(struct mm_struct
*mm
, pmd_t
*pmd
);
2744 int __pte_alloc_kernel(pmd_t
*pmd
);
2746 #if defined(CONFIG_MMU)
2748 static inline p4d_t
*p4d_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
2749 unsigned long address
)
2751 return (unlikely(pgd_none(*pgd
)) && __p4d_alloc(mm
, pgd
, address
)) ?
2752 NULL
: p4d_offset(pgd
, address
);
2755 static inline pud_t
*pud_alloc(struct mm_struct
*mm
, p4d_t
*p4d
,
2756 unsigned long address
)
2758 return (unlikely(p4d_none(*p4d
)) && __pud_alloc(mm
, p4d
, address
)) ?
2759 NULL
: pud_offset(p4d
, address
);
2762 static inline pmd_t
*pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
)
2764 return (unlikely(pud_none(*pud
)) && __pmd_alloc(mm
, pud
, address
))?
2765 NULL
: pmd_offset(pud
, address
);
2767 #endif /* CONFIG_MMU */
2769 #if USE_SPLIT_PTE_PTLOCKS
2770 #if ALLOC_SPLIT_PTLOCKS
2771 void __init
ptlock_cache_init(void);
2772 extern bool ptlock_alloc(struct page
*page
);
2773 extern void ptlock_free(struct page
*page
);
2775 static inline spinlock_t
*ptlock_ptr(struct page
*page
)
2779 #else /* ALLOC_SPLIT_PTLOCKS */
2780 static inline void ptlock_cache_init(void)
2784 static inline bool ptlock_alloc(struct page
*page
)
2789 static inline void ptlock_free(struct page
*page
)
2793 static inline spinlock_t
*ptlock_ptr(struct page
*page
)
2797 #endif /* ALLOC_SPLIT_PTLOCKS */
2799 static inline spinlock_t
*pte_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
2801 return ptlock_ptr(pmd_page(*pmd
));
2804 static inline bool ptlock_init(struct page
*page
)
2807 * prep_new_page() initialize page->private (and therefore page->ptl)
2808 * with 0. Make sure nobody took it in use in between.
2810 * It can happen if arch try to use slab for page table allocation:
2811 * slab code uses page->slab_cache, which share storage with page->ptl.
2813 VM_BUG_ON_PAGE(*(unsigned long *)&page
->ptl
, page
);
2814 if (!ptlock_alloc(page
))
2816 spin_lock_init(ptlock_ptr(page
));
2820 #else /* !USE_SPLIT_PTE_PTLOCKS */
2822 * We use mm->page_table_lock to guard all pagetable pages of the mm.
2824 static inline spinlock_t
*pte_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
2826 return &mm
->page_table_lock
;
2828 static inline void ptlock_cache_init(void) {}
2829 static inline bool ptlock_init(struct page
*page
) { return true; }
2830 static inline void ptlock_free(struct page
*page
) {}
2831 #endif /* USE_SPLIT_PTE_PTLOCKS */
2833 static inline bool pgtable_pte_page_ctor(struct page
*page
)
2835 if (!ptlock_init(page
))
2837 __SetPageTable(page
);
2838 inc_lruvec_page_state(page
, NR_PAGETABLE
);
2842 static inline void pgtable_pte_page_dtor(struct page
*page
)
2845 __ClearPageTable(page
);
2846 dec_lruvec_page_state(page
, NR_PAGETABLE
);
2849 pte_t
*__pte_offset_map(pmd_t
*pmd
, unsigned long addr
, pmd_t
*pmdvalp
);
2850 static inline pte_t
*pte_offset_map(pmd_t
*pmd
, unsigned long addr
)
2852 return __pte_offset_map(pmd
, addr
, NULL
);
2855 pte_t
*__pte_offset_map_lock(struct mm_struct
*mm
, pmd_t
*pmd
,
2856 unsigned long addr
, spinlock_t
**ptlp
);
2857 static inline pte_t
*pte_offset_map_lock(struct mm_struct
*mm
, pmd_t
*pmd
,
2858 unsigned long addr
, spinlock_t
**ptlp
)
2862 __cond_lock(*ptlp
, pte
= __pte_offset_map_lock(mm
, pmd
, addr
, ptlp
));
2866 pte_t
*pte_offset_map_nolock(struct mm_struct
*mm
, pmd_t
*pmd
,
2867 unsigned long addr
, spinlock_t
**ptlp
);
2869 #define pte_unmap_unlock(pte, ptl) do { \
2874 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2876 #define pte_alloc_map(mm, pmd, address) \
2877 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2879 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2880 (pte_alloc(mm, pmd) ? \
2881 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2883 #define pte_alloc_kernel(pmd, address) \
2884 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2885 NULL: pte_offset_kernel(pmd, address))
2887 #if USE_SPLIT_PMD_PTLOCKS
2889 static inline struct page
*pmd_pgtable_page(pmd_t
*pmd
)
2891 unsigned long mask
= ~(PTRS_PER_PMD
* sizeof(pmd_t
) - 1);
2892 return virt_to_page((void *)((unsigned long) pmd
& mask
));
2895 static inline spinlock_t
*pmd_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
2897 return ptlock_ptr(pmd_pgtable_page(pmd
));
2900 static inline bool pmd_ptlock_init(struct page
*page
)
2902 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2903 page
->pmd_huge_pte
= NULL
;
2905 return ptlock_init(page
);
2908 static inline void pmd_ptlock_free(struct page
*page
)
2910 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2911 VM_BUG_ON_PAGE(page
->pmd_huge_pte
, page
);
2916 #define pmd_huge_pte(mm, pmd) (pmd_pgtable_page(pmd)->pmd_huge_pte)
2920 static inline spinlock_t
*pmd_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
2922 return &mm
->page_table_lock
;
2925 static inline bool pmd_ptlock_init(struct page
*page
) { return true; }
2926 static inline void pmd_ptlock_free(struct page
*page
) {}
2928 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2932 static inline spinlock_t
*pmd_lock(struct mm_struct
*mm
, pmd_t
*pmd
)
2934 spinlock_t
*ptl
= pmd_lockptr(mm
, pmd
);
2939 static inline bool pgtable_pmd_page_ctor(struct page
*page
)
2941 if (!pmd_ptlock_init(page
))
2943 __SetPageTable(page
);
2944 inc_lruvec_page_state(page
, NR_PAGETABLE
);
2948 static inline void pgtable_pmd_page_dtor(struct page
*page
)
2950 pmd_ptlock_free(page
);
2951 __ClearPageTable(page
);
2952 dec_lruvec_page_state(page
, NR_PAGETABLE
);
2956 * No scalability reason to split PUD locks yet, but follow the same pattern
2957 * as the PMD locks to make it easier if we decide to. The VM should not be
2958 * considered ready to switch to split PUD locks yet; there may be places
2959 * which need to be converted from page_table_lock.
2961 static inline spinlock_t
*pud_lockptr(struct mm_struct
*mm
, pud_t
*pud
)
2963 return &mm
->page_table_lock
;
2966 static inline spinlock_t
*pud_lock(struct mm_struct
*mm
, pud_t
*pud
)
2968 spinlock_t
*ptl
= pud_lockptr(mm
, pud
);
2974 extern void __init
pagecache_init(void);
2975 extern void free_initmem(void);
2978 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2979 * into the buddy system. The freed pages will be poisoned with pattern
2980 * "poison" if it's within range [0, UCHAR_MAX].
2981 * Return pages freed into the buddy system.
2983 extern unsigned long free_reserved_area(void *start
, void *end
,
2984 int poison
, const char *s
);
2986 extern void adjust_managed_page_count(struct page
*page
, long count
);
2988 extern void reserve_bootmem_region(phys_addr_t start
,
2989 phys_addr_t end
, int nid
);
2991 /* Free the reserved page into the buddy system, so it gets managed. */
2992 static inline void free_reserved_page(struct page
*page
)
2994 ClearPageReserved(page
);
2995 init_page_count(page
);
2997 adjust_managed_page_count(page
, 1);
2999 #define free_highmem_page(page) free_reserved_page(page)
3001 static inline void mark_page_reserved(struct page
*page
)
3003 SetPageReserved(page
);
3004 adjust_managed_page_count(page
, -1);
3008 * Default method to free all the __init memory into the buddy system.
3009 * The freed pages will be poisoned with pattern "poison" if it's within
3010 * range [0, UCHAR_MAX].
3011 * Return pages freed into the buddy system.
3013 static inline unsigned long free_initmem_default(int poison
)
3015 extern char __init_begin
[], __init_end
[];
3017 return free_reserved_area(&__init_begin
, &__init_end
,
3018 poison
, "unused kernel image (initmem)");
3021 static inline unsigned long get_num_physpages(void)
3024 unsigned long phys_pages
= 0;
3026 for_each_online_node(nid
)
3027 phys_pages
+= node_present_pages(nid
);
3033 * Using memblock node mappings, an architecture may initialise its
3034 * zones, allocate the backing mem_map and account for memory holes in an
3035 * architecture independent manner.
3037 * An architecture is expected to register range of page frames backed by
3038 * physical memory with memblock_add[_node]() before calling
3039 * free_area_init() passing in the PFN each zone ends at. At a basic
3040 * usage, an architecture is expected to do something like
3042 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
3044 * for_each_valid_physical_page_range()
3045 * memblock_add_node(base, size, nid, MEMBLOCK_NONE)
3046 * free_area_init(max_zone_pfns);
3048 void free_area_init(unsigned long *max_zone_pfn
);
3049 unsigned long node_map_pfn_alignment(void);
3050 unsigned long __absent_pages_in_range(int nid
, unsigned long start_pfn
,
3051 unsigned long end_pfn
);
3052 extern unsigned long absent_pages_in_range(unsigned long start_pfn
,
3053 unsigned long end_pfn
);
3054 extern void get_pfn_range_for_nid(unsigned int nid
,
3055 unsigned long *start_pfn
, unsigned long *end_pfn
);
3058 static inline int early_pfn_to_nid(unsigned long pfn
)
3063 /* please see mm/page_alloc.c */
3064 extern int __meminit
early_pfn_to_nid(unsigned long pfn
);
3067 extern void set_dma_reserve(unsigned long new_dma_reserve
);
3068 extern void mem_init(void);
3069 extern void __init
mmap_init(void);
3071 extern void __show_mem(unsigned int flags
, nodemask_t
*nodemask
, int max_zone_idx
);
3072 static inline void show_mem(unsigned int flags
, nodemask_t
*nodemask
)
3074 __show_mem(flags
, nodemask
, MAX_NR_ZONES
- 1);
3076 extern long si_mem_available(void);
3077 extern void si_meminfo(struct sysinfo
* val
);
3078 extern void si_meminfo_node(struct sysinfo
*val
, int nid
);
3079 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
3080 extern unsigned long arch_reserved_kernel_pages(void);
3083 extern __printf(3, 4)
3084 void warn_alloc(gfp_t gfp_mask
, nodemask_t
*nodemask
, const char *fmt
, ...);
3086 extern void setup_per_cpu_pageset(void);
3089 extern atomic_long_t mmap_pages_allocated
;
3090 extern int nommu_shrink_inode_mappings(struct inode
*, size_t, size_t);
3092 /* interval_tree.c */
3093 void vma_interval_tree_insert(struct vm_area_struct
*node
,
3094 struct rb_root_cached
*root
);
3095 void vma_interval_tree_insert_after(struct vm_area_struct
*node
,
3096 struct vm_area_struct
*prev
,
3097 struct rb_root_cached
*root
);
3098 void vma_interval_tree_remove(struct vm_area_struct
*node
,
3099 struct rb_root_cached
*root
);
3100 struct vm_area_struct
*vma_interval_tree_iter_first(struct rb_root_cached
*root
,
3101 unsigned long start
, unsigned long last
);
3102 struct vm_area_struct
*vma_interval_tree_iter_next(struct vm_area_struct
*node
,
3103 unsigned long start
, unsigned long last
);
3105 #define vma_interval_tree_foreach(vma, root, start, last) \
3106 for (vma = vma_interval_tree_iter_first(root, start, last); \
3107 vma; vma = vma_interval_tree_iter_next(vma, start, last))
3109 void anon_vma_interval_tree_insert(struct anon_vma_chain
*node
,
3110 struct rb_root_cached
*root
);
3111 void anon_vma_interval_tree_remove(struct anon_vma_chain
*node
,
3112 struct rb_root_cached
*root
);
3113 struct anon_vma_chain
*
3114 anon_vma_interval_tree_iter_first(struct rb_root_cached
*root
,
3115 unsigned long start
, unsigned long last
);
3116 struct anon_vma_chain
*anon_vma_interval_tree_iter_next(
3117 struct anon_vma_chain
*node
, unsigned long start
, unsigned long last
);
3118 #ifdef CONFIG_DEBUG_VM_RB
3119 void anon_vma_interval_tree_verify(struct anon_vma_chain
*node
);
3122 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
3123 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
3124 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
3127 extern int __vm_enough_memory(struct mm_struct
*mm
, long pages
, int cap_sys_admin
);
3128 extern int vma_expand(struct vma_iterator
*vmi
, struct vm_area_struct
*vma
,
3129 unsigned long start
, unsigned long end
, pgoff_t pgoff
,
3130 struct vm_area_struct
*next
);
3131 extern int vma_shrink(struct vma_iterator
*vmi
, struct vm_area_struct
*vma
,
3132 unsigned long start
, unsigned long end
, pgoff_t pgoff
);
3133 extern struct vm_area_struct
*vma_merge(struct vma_iterator
*vmi
,
3134 struct mm_struct
*, struct vm_area_struct
*prev
, unsigned long addr
,
3135 unsigned long end
, unsigned long vm_flags
, struct anon_vma
*,
3136 struct file
*, pgoff_t
, struct mempolicy
*, struct vm_userfaultfd_ctx
,
3137 struct anon_vma_name
*);
3138 extern struct anon_vma
*find_mergeable_anon_vma(struct vm_area_struct
*);
3139 extern int __split_vma(struct vma_iterator
*vmi
, struct vm_area_struct
*,
3140 unsigned long addr
, int new_below
);
3141 extern int split_vma(struct vma_iterator
*vmi
, struct vm_area_struct
*,
3142 unsigned long addr
, int new_below
);
3143 extern int insert_vm_struct(struct mm_struct
*, struct vm_area_struct
*);
3144 extern void unlink_file_vma(struct vm_area_struct
*);
3145 extern struct vm_area_struct
*copy_vma(struct vm_area_struct
**,
3146 unsigned long addr
, unsigned long len
, pgoff_t pgoff
,
3147 bool *need_rmap_locks
);
3148 extern void exit_mmap(struct mm_struct
*);
3150 static inline int check_data_rlimit(unsigned long rlim
,
3152 unsigned long start
,
3153 unsigned long end_data
,
3154 unsigned long start_data
)
3156 if (rlim
< RLIM_INFINITY
) {
3157 if (((new - start
) + (end_data
- start_data
)) > rlim
)
3164 extern int mm_take_all_locks(struct mm_struct
*mm
);
3165 extern void mm_drop_all_locks(struct mm_struct
*mm
);
3167 extern int set_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
);
3168 extern int replace_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
);
3169 extern struct file
*get_mm_exe_file(struct mm_struct
*mm
);
3170 extern struct file
*get_task_exe_file(struct task_struct
*task
);
3172 extern bool may_expand_vm(struct mm_struct
*, vm_flags_t
, unsigned long npages
);
3173 extern void vm_stat_account(struct mm_struct
*, vm_flags_t
, long npages
);
3175 extern bool vma_is_special_mapping(const struct vm_area_struct
*vma
,
3176 const struct vm_special_mapping
*sm
);
3177 extern struct vm_area_struct
*_install_special_mapping(struct mm_struct
*mm
,
3178 unsigned long addr
, unsigned long len
,
3179 unsigned long flags
,
3180 const struct vm_special_mapping
*spec
);
3181 /* This is an obsolete alternative to _install_special_mapping. */
3182 extern int install_special_mapping(struct mm_struct
*mm
,
3183 unsigned long addr
, unsigned long len
,
3184 unsigned long flags
, struct page
**pages
);
3186 unsigned long randomize_stack_top(unsigned long stack_top
);
3187 unsigned long randomize_page(unsigned long start
, unsigned long range
);
3189 extern unsigned long get_unmapped_area(struct file
*, unsigned long, unsigned long, unsigned long, unsigned long);
3191 extern unsigned long mmap_region(struct file
*file
, unsigned long addr
,
3192 unsigned long len
, vm_flags_t vm_flags
, unsigned long pgoff
,
3193 struct list_head
*uf
);
3194 extern unsigned long do_mmap(struct file
*file
, unsigned long addr
,
3195 unsigned long len
, unsigned long prot
, unsigned long flags
,
3196 unsigned long pgoff
, unsigned long *populate
, struct list_head
*uf
);
3197 extern int do_vmi_munmap(struct vma_iterator
*vmi
, struct mm_struct
*mm
,
3198 unsigned long start
, size_t len
, struct list_head
*uf
,
3200 extern int do_munmap(struct mm_struct
*, unsigned long, size_t,
3201 struct list_head
*uf
);
3202 extern int do_madvise(struct mm_struct
*mm
, unsigned long start
, size_t len_in
, int behavior
);
3205 extern int do_vma_munmap(struct vma_iterator
*vmi
, struct vm_area_struct
*vma
,
3206 unsigned long start
, unsigned long end
,
3207 struct list_head
*uf
, bool unlock
);
3208 extern int __mm_populate(unsigned long addr
, unsigned long len
,
3210 static inline void mm_populate(unsigned long addr
, unsigned long len
)
3213 (void) __mm_populate(addr
, len
, 1);
3216 static inline void mm_populate(unsigned long addr
, unsigned long len
) {}
3219 /* These take the mm semaphore themselves */
3220 extern int __must_check
vm_brk(unsigned long, unsigned long);
3221 extern int __must_check
vm_brk_flags(unsigned long, unsigned long, unsigned long);
3222 extern int vm_munmap(unsigned long, size_t);
3223 extern unsigned long __must_check
vm_mmap(struct file
*, unsigned long,
3224 unsigned long, unsigned long,
3225 unsigned long, unsigned long);
3227 struct vm_unmapped_area_info
{
3228 #define VM_UNMAPPED_AREA_TOPDOWN 1
3229 unsigned long flags
;
3230 unsigned long length
;
3231 unsigned long low_limit
;
3232 unsigned long high_limit
;
3233 unsigned long align_mask
;
3234 unsigned long align_offset
;
3237 extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info
*info
);
3240 extern void truncate_inode_pages(struct address_space
*, loff_t
);
3241 extern void truncate_inode_pages_range(struct address_space
*,
3242 loff_t lstart
, loff_t lend
);
3243 extern void truncate_inode_pages_final(struct address_space
*);
3245 /* generic vm_area_ops exported for stackable file systems */
3246 extern vm_fault_t
filemap_fault(struct vm_fault
*vmf
);
3247 extern vm_fault_t
filemap_map_pages(struct vm_fault
*vmf
,
3248 pgoff_t start_pgoff
, pgoff_t end_pgoff
);
3249 extern vm_fault_t
filemap_page_mkwrite(struct vm_fault
*vmf
);
3251 extern unsigned long stack_guard_gap
;
3252 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
3253 int expand_stack_locked(struct vm_area_struct
*vma
, unsigned long address
);
3254 struct vm_area_struct
*expand_stack(struct mm_struct
* mm
, unsigned long addr
);
3256 /* CONFIG_STACK_GROWSUP still needs to grow downwards at some places */
3257 int expand_downwards(struct vm_area_struct
*vma
, unsigned long address
);
3259 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
3260 extern struct vm_area_struct
* find_vma(struct mm_struct
* mm
, unsigned long addr
);
3261 extern struct vm_area_struct
* find_vma_prev(struct mm_struct
* mm
, unsigned long addr
,
3262 struct vm_area_struct
**pprev
);
3265 * Look up the first VMA which intersects the interval [start_addr, end_addr)
3266 * NULL if none. Assume start_addr < end_addr.
3268 struct vm_area_struct
*find_vma_intersection(struct mm_struct
*mm
,
3269 unsigned long start_addr
, unsigned long end_addr
);
3272 * vma_lookup() - Find a VMA at a specific address
3273 * @mm: The process address space.
3274 * @addr: The user address.
3276 * Return: The vm_area_struct at the given address, %NULL otherwise.
3279 struct vm_area_struct
*vma_lookup(struct mm_struct
*mm
, unsigned long addr
)
3281 return mtree_load(&mm
->mm_mt
, addr
);
3284 static inline unsigned long vm_start_gap(struct vm_area_struct
*vma
)
3286 unsigned long vm_start
= vma
->vm_start
;
3288 if (vma
->vm_flags
& VM_GROWSDOWN
) {
3289 vm_start
-= stack_guard_gap
;
3290 if (vm_start
> vma
->vm_start
)
3296 static inline unsigned long vm_end_gap(struct vm_area_struct
*vma
)
3298 unsigned long vm_end
= vma
->vm_end
;
3300 if (vma
->vm_flags
& VM_GROWSUP
) {
3301 vm_end
+= stack_guard_gap
;
3302 if (vm_end
< vma
->vm_end
)
3303 vm_end
= -PAGE_SIZE
;
3308 static inline unsigned long vma_pages(struct vm_area_struct
*vma
)
3310 return (vma
->vm_end
- vma
->vm_start
) >> PAGE_SHIFT
;
3313 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
3314 static inline struct vm_area_struct
*find_exact_vma(struct mm_struct
*mm
,
3315 unsigned long vm_start
, unsigned long vm_end
)
3317 struct vm_area_struct
*vma
= vma_lookup(mm
, vm_start
);
3319 if (vma
&& (vma
->vm_start
!= vm_start
|| vma
->vm_end
!= vm_end
))
3325 static inline bool range_in_vma(struct vm_area_struct
*vma
,
3326 unsigned long start
, unsigned long end
)
3328 return (vma
&& vma
->vm_start
<= start
&& end
<= vma
->vm_end
);
3332 pgprot_t
vm_get_page_prot(unsigned long vm_flags
);
3333 void vma_set_page_prot(struct vm_area_struct
*vma
);
3335 static inline pgprot_t
vm_get_page_prot(unsigned long vm_flags
)
3339 static inline void vma_set_page_prot(struct vm_area_struct
*vma
)
3341 vma
->vm_page_prot
= vm_get_page_prot(vma
->vm_flags
);
3345 void vma_set_file(struct vm_area_struct
*vma
, struct file
*file
);
3347 #ifdef CONFIG_NUMA_BALANCING
3348 unsigned long change_prot_numa(struct vm_area_struct
*vma
,
3349 unsigned long start
, unsigned long end
);
3352 struct vm_area_struct
*find_extend_vma_locked(struct mm_struct
*,
3353 unsigned long addr
);
3354 int remap_pfn_range(struct vm_area_struct
*, unsigned long addr
,
3355 unsigned long pfn
, unsigned long size
, pgprot_t
);
3356 int remap_pfn_range_notrack(struct vm_area_struct
*vma
, unsigned long addr
,
3357 unsigned long pfn
, unsigned long size
, pgprot_t prot
);
3358 int vm_insert_page(struct vm_area_struct
*, unsigned long addr
, struct page
*);
3359 int vm_insert_pages(struct vm_area_struct
*vma
, unsigned long addr
,
3360 struct page
**pages
, unsigned long *num
);
3361 int vm_map_pages(struct vm_area_struct
*vma
, struct page
**pages
,
3363 int vm_map_pages_zero(struct vm_area_struct
*vma
, struct page
**pages
,
3365 vm_fault_t
vmf_insert_pfn(struct vm_area_struct
*vma
, unsigned long addr
,
3367 vm_fault_t
vmf_insert_pfn_prot(struct vm_area_struct
*vma
, unsigned long addr
,
3368 unsigned long pfn
, pgprot_t pgprot
);
3369 vm_fault_t
vmf_insert_mixed(struct vm_area_struct
*vma
, unsigned long addr
,
3371 vm_fault_t
vmf_insert_mixed_mkwrite(struct vm_area_struct
*vma
,
3372 unsigned long addr
, pfn_t pfn
);
3373 int vm_iomap_memory(struct vm_area_struct
*vma
, phys_addr_t start
, unsigned long len
);
3375 static inline vm_fault_t
vmf_insert_page(struct vm_area_struct
*vma
,
3376 unsigned long addr
, struct page
*page
)
3378 int err
= vm_insert_page(vma
, addr
, page
);
3381 return VM_FAULT_OOM
;
3382 if (err
< 0 && err
!= -EBUSY
)
3383 return VM_FAULT_SIGBUS
;
3385 return VM_FAULT_NOPAGE
;
3388 #ifndef io_remap_pfn_range
3389 static inline int io_remap_pfn_range(struct vm_area_struct
*vma
,
3390 unsigned long addr
, unsigned long pfn
,
3391 unsigned long size
, pgprot_t prot
)
3393 return remap_pfn_range(vma
, addr
, pfn
, size
, pgprot_decrypted(prot
));
3397 static inline vm_fault_t
vmf_error(int err
)
3400 return VM_FAULT_OOM
;
3401 else if (err
== -EHWPOISON
)
3402 return VM_FAULT_HWPOISON
;
3403 return VM_FAULT_SIGBUS
;
3406 struct page
*follow_page(struct vm_area_struct
*vma
, unsigned long address
,
3407 unsigned int foll_flags
);
3409 static inline int vm_fault_to_errno(vm_fault_t vm_fault
, int foll_flags
)
3411 if (vm_fault
& VM_FAULT_OOM
)
3413 if (vm_fault
& (VM_FAULT_HWPOISON
| VM_FAULT_HWPOISON_LARGE
))
3414 return (foll_flags
& FOLL_HWPOISON
) ? -EHWPOISON
: -EFAULT
;
3415 if (vm_fault
& (VM_FAULT_SIGBUS
| VM_FAULT_SIGSEGV
))
3421 * Indicates whether GUP can follow a PROT_NONE mapped page, or whether
3422 * a (NUMA hinting) fault is required.
3424 static inline bool gup_can_follow_protnone(struct vm_area_struct
*vma
,
3428 * If callers don't want to honor NUMA hinting faults, no need to
3429 * determine if we would actually have to trigger a NUMA hinting fault.
3431 if (!(flags
& FOLL_HONOR_NUMA_FAULT
))
3435 * NUMA hinting faults don't apply in inaccessible (PROT_NONE) VMAs.
3437 * Requiring a fault here even for inaccessible VMAs would mean that
3438 * FOLL_FORCE cannot make any progress, because handle_mm_fault()
3439 * refuses to process NUMA hinting faults in inaccessible VMAs.
3441 return !vma_is_accessible(vma
);
3444 typedef int (*pte_fn_t
)(pte_t
*pte
, unsigned long addr
, void *data
);
3445 extern int apply_to_page_range(struct mm_struct
*mm
, unsigned long address
,
3446 unsigned long size
, pte_fn_t fn
, void *data
);
3447 extern int apply_to_existing_page_range(struct mm_struct
*mm
,
3448 unsigned long address
, unsigned long size
,
3449 pte_fn_t fn
, void *data
);
3451 #ifdef CONFIG_PAGE_POISONING
3452 extern void __kernel_poison_pages(struct page
*page
, int numpages
);
3453 extern void __kernel_unpoison_pages(struct page
*page
, int numpages
);
3454 extern bool _page_poisoning_enabled_early
;
3455 DECLARE_STATIC_KEY_FALSE(_page_poisoning_enabled
);
3456 static inline bool page_poisoning_enabled(void)
3458 return _page_poisoning_enabled_early
;
3461 * For use in fast paths after init_mem_debugging() has run, or when a
3462 * false negative result is not harmful when called too early.
3464 static inline bool page_poisoning_enabled_static(void)
3466 return static_branch_unlikely(&_page_poisoning_enabled
);
3468 static inline void kernel_poison_pages(struct page
*page
, int numpages
)
3470 if (page_poisoning_enabled_static())
3471 __kernel_poison_pages(page
, numpages
);
3473 static inline void kernel_unpoison_pages(struct page
*page
, int numpages
)
3475 if (page_poisoning_enabled_static())
3476 __kernel_unpoison_pages(page
, numpages
);
3479 static inline bool page_poisoning_enabled(void) { return false; }
3480 static inline bool page_poisoning_enabled_static(void) { return false; }
3481 static inline void __kernel_poison_pages(struct page
*page
, int nunmpages
) { }
3482 static inline void kernel_poison_pages(struct page
*page
, int numpages
) { }
3483 static inline void kernel_unpoison_pages(struct page
*page
, int numpages
) { }
3486 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_ALLOC_DEFAULT_ON
, init_on_alloc
);
3487 static inline bool want_init_on_alloc(gfp_t flags
)
3489 if (static_branch_maybe(CONFIG_INIT_ON_ALLOC_DEFAULT_ON
,
3492 return flags
& __GFP_ZERO
;
3495 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_FREE_DEFAULT_ON
, init_on_free
);
3496 static inline bool want_init_on_free(void)
3498 return static_branch_maybe(CONFIG_INIT_ON_FREE_DEFAULT_ON
,
3502 extern bool _debug_pagealloc_enabled_early
;
3503 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled
);
3505 static inline bool debug_pagealloc_enabled(void)
3507 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC
) &&
3508 _debug_pagealloc_enabled_early
;
3512 * For use in fast paths after init_debug_pagealloc() has run, or when a
3513 * false negative result is not harmful when called too early.
3515 static inline bool debug_pagealloc_enabled_static(void)
3517 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC
))
3520 return static_branch_unlikely(&_debug_pagealloc_enabled
);
3524 * To support DEBUG_PAGEALLOC architecture must ensure that
3525 * __kernel_map_pages() never fails
3527 extern void __kernel_map_pages(struct page
*page
, int numpages
, int enable
);
3528 #ifdef CONFIG_DEBUG_PAGEALLOC
3529 static inline void debug_pagealloc_map_pages(struct page
*page
, int numpages
)
3531 if (debug_pagealloc_enabled_static())
3532 __kernel_map_pages(page
, numpages
, 1);
3535 static inline void debug_pagealloc_unmap_pages(struct page
*page
, int numpages
)
3537 if (debug_pagealloc_enabled_static())
3538 __kernel_map_pages(page
, numpages
, 0);
3541 extern unsigned int _debug_guardpage_minorder
;
3542 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled
);
3544 static inline unsigned int debug_guardpage_minorder(void)
3546 return _debug_guardpage_minorder
;
3549 static inline bool debug_guardpage_enabled(void)
3551 return static_branch_unlikely(&_debug_guardpage_enabled
);
3554 static inline bool page_is_guard(struct page
*page
)
3556 if (!debug_guardpage_enabled())
3559 return PageGuard(page
);
3562 bool __set_page_guard(struct zone
*zone
, struct page
*page
, unsigned int order
,
3564 static inline bool set_page_guard(struct zone
*zone
, struct page
*page
,
3565 unsigned int order
, int migratetype
)
3567 if (!debug_guardpage_enabled())
3569 return __set_page_guard(zone
, page
, order
, migratetype
);
3572 void __clear_page_guard(struct zone
*zone
, struct page
*page
, unsigned int order
,
3574 static inline void clear_page_guard(struct zone
*zone
, struct page
*page
,
3575 unsigned int order
, int migratetype
)
3577 if (!debug_guardpage_enabled())
3579 __clear_page_guard(zone
, page
, order
, migratetype
);
3582 #else /* CONFIG_DEBUG_PAGEALLOC */
3583 static inline void debug_pagealloc_map_pages(struct page
*page
, int numpages
) {}
3584 static inline void debug_pagealloc_unmap_pages(struct page
*page
, int numpages
) {}
3585 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
3586 static inline bool debug_guardpage_enabled(void) { return false; }
3587 static inline bool page_is_guard(struct page
*page
) { return false; }
3588 static inline bool set_page_guard(struct zone
*zone
, struct page
*page
,
3589 unsigned int order
, int migratetype
) { return false; }
3590 static inline void clear_page_guard(struct zone
*zone
, struct page
*page
,
3591 unsigned int order
, int migratetype
) {}
3592 #endif /* CONFIG_DEBUG_PAGEALLOC */
3594 #ifdef __HAVE_ARCH_GATE_AREA
3595 extern struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
);
3596 extern int in_gate_area_no_mm(unsigned long addr
);
3597 extern int in_gate_area(struct mm_struct
*mm
, unsigned long addr
);
3599 static inline struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
)
3603 static inline int in_gate_area_no_mm(unsigned long addr
) { return 0; }
3604 static inline int in_gate_area(struct mm_struct
*mm
, unsigned long addr
)
3608 #endif /* __HAVE_ARCH_GATE_AREA */
3610 extern bool process_shares_mm(struct task_struct
*p
, struct mm_struct
*mm
);
3612 #ifdef CONFIG_SYSCTL
3613 extern int sysctl_drop_caches
;
3614 int drop_caches_sysctl_handler(struct ctl_table
*, int, void *, size_t *,
3618 void drop_slab(void);
3621 #define randomize_va_space 0
3623 extern int randomize_va_space
;
3626 const char * arch_vma_name(struct vm_area_struct
*vma
);
3628 void print_vma_addr(char *prefix
, unsigned long rip
);
3630 static inline void print_vma_addr(char *prefix
, unsigned long rip
)
3635 void *sparse_buffer_alloc(unsigned long size
);
3636 struct page
* __populate_section_memmap(unsigned long pfn
,
3637 unsigned long nr_pages
, int nid
, struct vmem_altmap
*altmap
,
3638 struct dev_pagemap
*pgmap
);
3639 void pmd_init(void *addr
);
3640 void pud_init(void *addr
);
3641 pgd_t
*vmemmap_pgd_populate(unsigned long addr
, int node
);
3642 p4d_t
*vmemmap_p4d_populate(pgd_t
*pgd
, unsigned long addr
, int node
);
3643 pud_t
*vmemmap_pud_populate(p4d_t
*p4d
, unsigned long addr
, int node
);
3644 pmd_t
*vmemmap_pmd_populate(pud_t
*pud
, unsigned long addr
, int node
);
3645 pte_t
*vmemmap_pte_populate(pmd_t
*pmd
, unsigned long addr
, int node
,
3646 struct vmem_altmap
*altmap
, struct page
*reuse
);
3647 void *vmemmap_alloc_block(unsigned long size
, int node
);
3649 void *vmemmap_alloc_block_buf(unsigned long size
, int node
,
3650 struct vmem_altmap
*altmap
);
3651 void vmemmap_verify(pte_t
*, int, unsigned long, unsigned long);
3652 void vmemmap_set_pmd(pmd_t
*pmd
, void *p
, int node
,
3653 unsigned long addr
, unsigned long next
);
3654 int vmemmap_check_pmd(pmd_t
*pmd
, int node
,
3655 unsigned long addr
, unsigned long next
);
3656 int vmemmap_populate_basepages(unsigned long start
, unsigned long end
,
3657 int node
, struct vmem_altmap
*altmap
);
3658 int vmemmap_populate_hugepages(unsigned long start
, unsigned long end
,
3659 int node
, struct vmem_altmap
*altmap
);
3660 int vmemmap_populate(unsigned long start
, unsigned long end
, int node
,
3661 struct vmem_altmap
*altmap
);
3662 void vmemmap_populate_print_last(void);
3663 #ifdef CONFIG_MEMORY_HOTPLUG
3664 void vmemmap_free(unsigned long start
, unsigned long end
,
3665 struct vmem_altmap
*altmap
);
3668 #ifdef CONFIG_ARCH_WANT_OPTIMIZE_VMEMMAP
3669 static inline bool vmemmap_can_optimize(struct vmem_altmap
*altmap
,
3670 struct dev_pagemap
*pgmap
)
3672 return is_power_of_2(sizeof(struct page
)) &&
3673 pgmap
&& (pgmap_vmemmap_nr(pgmap
) > 1) && !altmap
;
3676 static inline bool vmemmap_can_optimize(struct vmem_altmap
*altmap
,
3677 struct dev_pagemap
*pgmap
)
3683 void register_page_bootmem_memmap(unsigned long section_nr
, struct page
*map
,
3684 unsigned long nr_pages
);
3687 MF_COUNT_INCREASED
= 1 << 0,
3688 MF_ACTION_REQUIRED
= 1 << 1,
3689 MF_MUST_KILL
= 1 << 2,
3690 MF_SOFT_OFFLINE
= 1 << 3,
3691 MF_UNPOISON
= 1 << 4,
3692 MF_SW_SIMULATED
= 1 << 5,
3693 MF_NO_RETRY
= 1 << 6,
3695 int mf_dax_kill_procs(struct address_space
*mapping
, pgoff_t index
,
3696 unsigned long count
, int mf_flags
);
3697 extern int memory_failure(unsigned long pfn
, int flags
);
3698 extern void memory_failure_queue_kick(int cpu
);
3699 extern int unpoison_memory(unsigned long pfn
);
3700 extern void shake_page(struct page
*p
);
3701 extern atomic_long_t num_poisoned_pages __read_mostly
;
3702 extern int soft_offline_page(unsigned long pfn
, int flags
);
3703 #ifdef CONFIG_MEMORY_FAILURE
3705 * Sysfs entries for memory failure handling statistics.
3707 extern const struct attribute_group memory_failure_attr_group
;
3708 extern void memory_failure_queue(unsigned long pfn
, int flags
);
3709 extern int __get_huge_page_for_hwpoison(unsigned long pfn
, int flags
,
3710 bool *migratable_cleared
);
3711 void num_poisoned_pages_inc(unsigned long pfn
);
3712 void num_poisoned_pages_sub(unsigned long pfn
, long i
);
3713 struct task_struct
*task_early_kill(struct task_struct
*tsk
, int force_early
);
3715 static inline void memory_failure_queue(unsigned long pfn
, int flags
)
3719 static inline int __get_huge_page_for_hwpoison(unsigned long pfn
, int flags
,
3720 bool *migratable_cleared
)
3725 static inline void num_poisoned_pages_inc(unsigned long pfn
)
3729 static inline void num_poisoned_pages_sub(unsigned long pfn
, long i
)
3734 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_KSM)
3735 void add_to_kill_ksm(struct task_struct
*tsk
, struct page
*p
,
3736 struct vm_area_struct
*vma
, struct list_head
*to_kill
,
3737 unsigned long ksm_addr
);
3740 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_MEMORY_HOTPLUG)
3741 extern void memblk_nr_poison_inc(unsigned long pfn
);
3742 extern void memblk_nr_poison_sub(unsigned long pfn
, long i
);
3744 static inline void memblk_nr_poison_inc(unsigned long pfn
)
3748 static inline void memblk_nr_poison_sub(unsigned long pfn
, long i
)
3753 #ifndef arch_memory_failure
3754 static inline int arch_memory_failure(unsigned long pfn
, int flags
)
3760 #ifndef arch_is_platform_page
3761 static inline bool arch_is_platform_page(u64 paddr
)
3768 * Error handlers for various types of pages.
3771 MF_IGNORED
, /* Error: cannot be handled */
3772 MF_FAILED
, /* Error: handling failed */
3773 MF_DELAYED
, /* Will be handled later */
3774 MF_RECOVERED
, /* Successfully recovered */
3777 enum mf_action_page_type
{
3779 MF_MSG_KERNEL_HIGH_ORDER
,
3781 MF_MSG_DIFFERENT_COMPOUND
,
3784 MF_MSG_UNMAP_FAILED
,
3785 MF_MSG_DIRTY_SWAPCACHE
,
3786 MF_MSG_CLEAN_SWAPCACHE
,
3787 MF_MSG_DIRTY_MLOCKED_LRU
,
3788 MF_MSG_CLEAN_MLOCKED_LRU
,
3789 MF_MSG_DIRTY_UNEVICTABLE_LRU
,
3790 MF_MSG_CLEAN_UNEVICTABLE_LRU
,
3793 MF_MSG_TRUNCATED_LRU
,
3800 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
3801 extern void clear_huge_page(struct page
*page
,
3802 unsigned long addr_hint
,
3803 unsigned int pages_per_huge_page
);
3804 int copy_user_large_folio(struct folio
*dst
, struct folio
*src
,
3805 unsigned long addr_hint
,
3806 struct vm_area_struct
*vma
);
3807 long copy_folio_from_user(struct folio
*dst_folio
,
3808 const void __user
*usr_src
,
3809 bool allow_pagefault
);
3812 * vma_is_special_huge - Are transhuge page-table entries considered special?
3813 * @vma: Pointer to the struct vm_area_struct to consider
3815 * Whether transhuge page-table entries are considered "special" following
3816 * the definition in vm_normal_page().
3818 * Return: true if transhuge page-table entries should be considered special,
3821 static inline bool vma_is_special_huge(const struct vm_area_struct
*vma
)
3823 return vma_is_dax(vma
) || (vma
->vm_file
&&
3824 (vma
->vm_flags
& (VM_PFNMAP
| VM_MIXEDMAP
)));
3827 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
3829 #if MAX_NUMNODES > 1
3830 void __init
setup_nr_node_ids(void);
3832 static inline void setup_nr_node_ids(void) {}
3835 extern int memcmp_pages(struct page
*page1
, struct page
*page2
);
3837 static inline int pages_identical(struct page
*page1
, struct page
*page2
)
3839 return !memcmp_pages(page1
, page2
);
3842 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
3843 unsigned long clean_record_shared_mapping_range(struct address_space
*mapping
,
3844 pgoff_t first_index
, pgoff_t nr
,
3845 pgoff_t bitmap_pgoff
,
3846 unsigned long *bitmap
,
3850 unsigned long wp_shared_mapping_range(struct address_space
*mapping
,
3851 pgoff_t first_index
, pgoff_t nr
);
3854 extern int sysctl_nr_trim_pages
;
3856 #ifdef CONFIG_PRINTK
3857 void mem_dump_obj(void *object
);
3859 static inline void mem_dump_obj(void *object
) {}
3863 * seal_check_future_write - Check for F_SEAL_FUTURE_WRITE flag and handle it
3864 * @seals: the seals to check
3865 * @vma: the vma to operate on
3867 * Check whether F_SEAL_FUTURE_WRITE is set; if so, do proper check/handling on
3868 * the vma flags. Return 0 if check pass, or <0 for errors.
3870 static inline int seal_check_future_write(int seals
, struct vm_area_struct
*vma
)
3872 if (seals
& F_SEAL_FUTURE_WRITE
) {
3874 * New PROT_WRITE and MAP_SHARED mmaps are not allowed when
3875 * "future write" seal active.
3877 if ((vma
->vm_flags
& VM_SHARED
) && (vma
->vm_flags
& VM_WRITE
))
3881 * Since an F_SEAL_FUTURE_WRITE sealed memfd can be mapped as
3882 * MAP_SHARED and read-only, take care to not allow mprotect to
3883 * revert protections on such mappings. Do this only for shared
3884 * mappings. For private mappings, don't need to mask
3885 * VM_MAYWRITE as we still want them to be COW-writable.
3887 if (vma
->vm_flags
& VM_SHARED
)
3888 vm_flags_clear(vma
, VM_MAYWRITE
);
3894 #ifdef CONFIG_ANON_VMA_NAME
3895 int madvise_set_anon_name(struct mm_struct
*mm
, unsigned long start
,
3896 unsigned long len_in
,
3897 struct anon_vma_name
*anon_name
);
3900 madvise_set_anon_name(struct mm_struct
*mm
, unsigned long start
,
3901 unsigned long len_in
, struct anon_vma_name
*anon_name
) {
3906 #ifdef CONFIG_UNACCEPTED_MEMORY
3908 bool range_contains_unaccepted_memory(phys_addr_t start
, phys_addr_t end
);
3909 void accept_memory(phys_addr_t start
, phys_addr_t end
);
3913 static inline bool range_contains_unaccepted_memory(phys_addr_t start
,
3919 static inline void accept_memory(phys_addr_t start
, phys_addr_t end
)
3925 #endif /* _LINUX_MM_H */