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_BIT_5 37 /* bit only usable on 64-bit architectures */
323 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
324 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
325 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
326 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
327 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
328 #define VM_HIGH_ARCH_5 BIT(VM_HIGH_ARCH_BIT_5)
329 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
331 #ifdef CONFIG_ARCH_HAS_PKEYS
332 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
333 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
334 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
335 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
336 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
338 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
340 # define VM_PKEY_BIT4 0
342 #endif /* CONFIG_ARCH_HAS_PKEYS */
344 #ifdef CONFIG_X86_USER_SHADOW_STACK
346 * VM_SHADOW_STACK should not be set with VM_SHARED because of lack of
349 * These VMAs will get a single end guard page. This helps userspace protect
350 * itself from attacks. A single page is enough for current shadow stack archs
351 * (x86). See the comments near alloc_shstk() in arch/x86/kernel/shstk.c
352 * for more details on the guard size.
354 # define VM_SHADOW_STACK VM_HIGH_ARCH_5
356 # define VM_SHADOW_STACK VM_NONE
359 #if defined(CONFIG_X86)
360 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
361 #elif defined(CONFIG_PPC)
362 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
363 #elif defined(CONFIG_PARISC)
364 # define VM_GROWSUP VM_ARCH_1
365 #elif defined(CONFIG_SPARC64)
366 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
367 # define VM_ARCH_CLEAR VM_SPARC_ADI
368 #elif defined(CONFIG_ARM64)
369 # define VM_ARM64_BTI VM_ARCH_1 /* BTI guarded page, a.k.a. GP bit */
370 # define VM_ARCH_CLEAR VM_ARM64_BTI
371 #elif !defined(CONFIG_MMU)
372 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
375 #if defined(CONFIG_ARM64_MTE)
376 # define VM_MTE VM_HIGH_ARCH_0 /* Use Tagged memory for access control */
377 # define VM_MTE_ALLOWED VM_HIGH_ARCH_1 /* Tagged memory permitted */
379 # define VM_MTE VM_NONE
380 # define VM_MTE_ALLOWED VM_NONE
384 # define VM_GROWSUP VM_NONE
387 #ifdef CONFIG_HAVE_ARCH_USERFAULTFD_MINOR
388 # define VM_UFFD_MINOR_BIT 38
389 # define VM_UFFD_MINOR BIT(VM_UFFD_MINOR_BIT) /* UFFD minor faults */
390 #else /* !CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
391 # define VM_UFFD_MINOR VM_NONE
392 #endif /* CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
394 /* Bits set in the VMA until the stack is in its final location */
395 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ | VM_STACK_EARLY)
397 #define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0)
399 /* Common data flag combinations */
400 #define VM_DATA_FLAGS_TSK_EXEC (VM_READ | VM_WRITE | TASK_EXEC | \
401 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
402 #define VM_DATA_FLAGS_NON_EXEC (VM_READ | VM_WRITE | VM_MAYREAD | \
403 VM_MAYWRITE | VM_MAYEXEC)
404 #define VM_DATA_FLAGS_EXEC (VM_READ | VM_WRITE | VM_EXEC | \
405 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
407 #ifndef VM_DATA_DEFAULT_FLAGS /* arch can override this */
408 #define VM_DATA_DEFAULT_FLAGS VM_DATA_FLAGS_EXEC
411 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
412 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
415 #define VM_STARTGAP_FLAGS (VM_GROWSDOWN | VM_SHADOW_STACK)
417 #ifdef CONFIG_STACK_GROWSUP
418 #define VM_STACK VM_GROWSUP
419 #define VM_STACK_EARLY VM_GROWSDOWN
421 #define VM_STACK VM_GROWSDOWN
422 #define VM_STACK_EARLY 0
425 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
427 /* VMA basic access permission flags */
428 #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
432 * Special vmas that are non-mergable, non-mlock()able.
434 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
436 /* This mask prevents VMA from being scanned with khugepaged */
437 #define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB)
439 /* This mask defines which mm->def_flags a process can inherit its parent */
440 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
442 /* This mask represents all the VMA flag bits used by mlock */
443 #define VM_LOCKED_MASK (VM_LOCKED | VM_LOCKONFAULT)
445 /* Arch-specific flags to clear when updating VM flags on protection change */
446 #ifndef VM_ARCH_CLEAR
447 # define VM_ARCH_CLEAR VM_NONE
449 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
452 * mapping from the currently active vm_flags protection bits (the
453 * low four bits) to a page protection mask..
457 * The default fault flags that should be used by most of the
458 * arch-specific page fault handlers.
460 #define FAULT_FLAG_DEFAULT (FAULT_FLAG_ALLOW_RETRY | \
461 FAULT_FLAG_KILLABLE | \
462 FAULT_FLAG_INTERRUPTIBLE)
465 * fault_flag_allow_retry_first - check ALLOW_RETRY the first time
466 * @flags: Fault flags.
468 * This is mostly used for places where we want to try to avoid taking
469 * the mmap_lock for too long a time when waiting for another condition
470 * to change, in which case we can try to be polite to release the
471 * mmap_lock in the first round to avoid potential starvation of other
472 * processes that would also want the mmap_lock.
474 * Return: true if the page fault allows retry and this is the first
475 * attempt of the fault handling; false otherwise.
477 static inline bool fault_flag_allow_retry_first(enum fault_flag flags
)
479 return (flags
& FAULT_FLAG_ALLOW_RETRY
) &&
480 (!(flags
& FAULT_FLAG_TRIED
));
483 #define FAULT_FLAG_TRACE \
484 { FAULT_FLAG_WRITE, "WRITE" }, \
485 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
486 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
487 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
488 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
489 { FAULT_FLAG_TRIED, "TRIED" }, \
490 { FAULT_FLAG_USER, "USER" }, \
491 { FAULT_FLAG_REMOTE, "REMOTE" }, \
492 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }, \
493 { FAULT_FLAG_INTERRUPTIBLE, "INTERRUPTIBLE" }, \
494 { FAULT_FLAG_VMA_LOCK, "VMA_LOCK" }
497 * vm_fault is filled by the pagefault handler and passed to the vma's
498 * ->fault function. The vma's ->fault is responsible for returning a bitmask
499 * of VM_FAULT_xxx flags that give details about how the fault was handled.
501 * MM layer fills up gfp_mask for page allocations but fault handler might
502 * alter it if its implementation requires a different allocation context.
504 * pgoff should be used in favour of virtual_address, if possible.
508 struct vm_area_struct
*vma
; /* Target VMA */
509 gfp_t gfp_mask
; /* gfp mask to be used for allocations */
510 pgoff_t pgoff
; /* Logical page offset based on vma */
511 unsigned long address
; /* Faulting virtual address - masked */
512 unsigned long real_address
; /* Faulting virtual address - unmasked */
514 enum fault_flag flags
; /* FAULT_FLAG_xxx flags
515 * XXX: should really be 'const' */
516 pmd_t
*pmd
; /* Pointer to pmd entry matching
518 pud_t
*pud
; /* Pointer to pud entry matching
522 pte_t orig_pte
; /* Value of PTE at the time of fault */
523 pmd_t orig_pmd
; /* Value of PMD at the time of fault,
524 * used by PMD fault only.
528 struct page
*cow_page
; /* Page handler may use for COW fault */
529 struct page
*page
; /* ->fault handlers should return a
530 * page here, unless VM_FAULT_NOPAGE
531 * is set (which is also implied by
534 /* These three entries are valid only while holding ptl lock */
535 pte_t
*pte
; /* Pointer to pte entry matching
536 * the 'address'. NULL if the page
537 * table hasn't been allocated.
539 spinlock_t
*ptl
; /* Page table lock.
540 * Protects pte page table if 'pte'
541 * is not NULL, otherwise pmd.
543 pgtable_t prealloc_pte
; /* Pre-allocated pte page table.
544 * vm_ops->map_pages() sets up a page
545 * table from atomic context.
546 * do_fault_around() pre-allocates
547 * page table to avoid allocation from
553 * These are the virtual MM functions - opening of an area, closing and
554 * unmapping it (needed to keep files on disk up-to-date etc), pointer
555 * to the functions called when a no-page or a wp-page exception occurs.
557 struct vm_operations_struct
{
558 void (*open
)(struct vm_area_struct
* area
);
560 * @close: Called when the VMA is being removed from the MM.
561 * Context: User context. May sleep. Caller holds mmap_lock.
563 void (*close
)(struct vm_area_struct
* area
);
564 /* Called any time before splitting to check if it's allowed */
565 int (*may_split
)(struct vm_area_struct
*area
, unsigned long addr
);
566 int (*mremap
)(struct vm_area_struct
*area
);
568 * Called by mprotect() to make driver-specific permission
569 * checks before mprotect() is finalised. The VMA must not
570 * be modified. Returns 0 if mprotect() can proceed.
572 int (*mprotect
)(struct vm_area_struct
*vma
, unsigned long start
,
573 unsigned long end
, unsigned long newflags
);
574 vm_fault_t (*fault
)(struct vm_fault
*vmf
);
575 vm_fault_t (*huge_fault
)(struct vm_fault
*vmf
, unsigned int order
);
576 vm_fault_t (*map_pages
)(struct vm_fault
*vmf
,
577 pgoff_t start_pgoff
, pgoff_t end_pgoff
);
578 unsigned long (*pagesize
)(struct vm_area_struct
* area
);
580 /* notification that a previously read-only page is about to become
581 * writable, if an error is returned it will cause a SIGBUS */
582 vm_fault_t (*page_mkwrite
)(struct vm_fault
*vmf
);
584 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
585 vm_fault_t (*pfn_mkwrite
)(struct vm_fault
*vmf
);
587 /* called by access_process_vm when get_user_pages() fails, typically
588 * for use by special VMAs. See also generic_access_phys() for a generic
589 * implementation useful for any iomem mapping.
591 int (*access
)(struct vm_area_struct
*vma
, unsigned long addr
,
592 void *buf
, int len
, int write
);
594 /* Called by the /proc/PID/maps code to ask the vma whether it
595 * has a special name. Returning non-NULL will also cause this
596 * vma to be dumped unconditionally. */
597 const char *(*name
)(struct vm_area_struct
*vma
);
601 * set_policy() op must add a reference to any non-NULL @new mempolicy
602 * to hold the policy upon return. Caller should pass NULL @new to
603 * remove a policy and fall back to surrounding context--i.e. do not
604 * install a MPOL_DEFAULT policy, nor the task or system default
607 int (*set_policy
)(struct vm_area_struct
*vma
, struct mempolicy
*new);
610 * get_policy() op must add reference [mpol_get()] to any policy at
611 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
612 * in mm/mempolicy.c will do this automatically.
613 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
614 * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
615 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
616 * must return NULL--i.e., do not "fallback" to task or system default
619 struct mempolicy
*(*get_policy
)(struct vm_area_struct
*vma
,
620 unsigned long addr
, pgoff_t
*ilx
);
623 * Called by vm_normal_page() for special PTEs to find the
624 * page for @addr. This is useful if the default behavior
625 * (using pte_page()) would not find the correct page.
627 struct page
*(*find_special_page
)(struct vm_area_struct
*vma
,
631 #ifdef CONFIG_NUMA_BALANCING
632 static inline void vma_numab_state_init(struct vm_area_struct
*vma
)
634 vma
->numab_state
= NULL
;
636 static inline void vma_numab_state_free(struct vm_area_struct
*vma
)
638 kfree(vma
->numab_state
);
641 static inline void vma_numab_state_init(struct vm_area_struct
*vma
) {}
642 static inline void vma_numab_state_free(struct vm_area_struct
*vma
) {}
643 #endif /* CONFIG_NUMA_BALANCING */
645 #ifdef CONFIG_PER_VMA_LOCK
647 * Try to read-lock a vma. The function is allowed to occasionally yield false
648 * locked result to avoid performance overhead, in which case we fall back to
649 * using mmap_lock. The function should never yield false unlocked result.
651 static inline bool vma_start_read(struct vm_area_struct
*vma
)
654 * Check before locking. A race might cause false locked result.
655 * We can use READ_ONCE() for the mm_lock_seq here, and don't need
656 * ACQUIRE semantics, because this is just a lockless check whose result
657 * we don't rely on for anything - the mm_lock_seq read against which we
658 * need ordering is below.
660 if (READ_ONCE(vma
->vm_lock_seq
) == READ_ONCE(vma
->vm_mm
->mm_lock_seq
))
663 if (unlikely(down_read_trylock(&vma
->vm_lock
->lock
) == 0))
667 * Overflow might produce false locked result.
668 * False unlocked result is impossible because we modify and check
669 * vma->vm_lock_seq under vma->vm_lock protection and mm->mm_lock_seq
670 * modification invalidates all existing locks.
672 * We must use ACQUIRE semantics for the mm_lock_seq so that if we are
673 * racing with vma_end_write_all(), we only start reading from the VMA
674 * after it has been unlocked.
675 * This pairs with RELEASE semantics in vma_end_write_all().
677 if (unlikely(vma
->vm_lock_seq
== smp_load_acquire(&vma
->vm_mm
->mm_lock_seq
))) {
678 up_read(&vma
->vm_lock
->lock
);
684 static inline void vma_end_read(struct vm_area_struct
*vma
)
686 rcu_read_lock(); /* keeps vma alive till the end of up_read */
687 up_read(&vma
->vm_lock
->lock
);
691 /* WARNING! Can only be used if mmap_lock is expected to be write-locked */
692 static bool __is_vma_write_locked(struct vm_area_struct
*vma
, int *mm_lock_seq
)
694 mmap_assert_write_locked(vma
->vm_mm
);
697 * current task is holding mmap_write_lock, both vma->vm_lock_seq and
698 * mm->mm_lock_seq can't be concurrently modified.
700 *mm_lock_seq
= vma
->vm_mm
->mm_lock_seq
;
701 return (vma
->vm_lock_seq
== *mm_lock_seq
);
705 * Begin writing to a VMA.
706 * Exclude concurrent readers under the per-VMA lock until the currently
707 * write-locked mmap_lock is dropped or downgraded.
709 static inline void vma_start_write(struct vm_area_struct
*vma
)
713 if (__is_vma_write_locked(vma
, &mm_lock_seq
))
716 down_write(&vma
->vm_lock
->lock
);
718 * We should use WRITE_ONCE() here because we can have concurrent reads
719 * from the early lockless pessimistic check in vma_start_read().
720 * We don't really care about the correctness of that early check, but
721 * we should use WRITE_ONCE() for cleanliness and to keep KCSAN happy.
723 WRITE_ONCE(vma
->vm_lock_seq
, mm_lock_seq
);
724 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_assert_locked(struct vm_area_struct
*vma
)
736 if (!rwsem_is_locked(&vma
->vm_lock
->lock
))
737 vma_assert_write_locked(vma
);
740 static inline void vma_mark_detached(struct vm_area_struct
*vma
, bool detached
)
742 /* When detaching vma should be write-locked */
744 vma_assert_write_locked(vma
);
745 vma
->detached
= detached
;
748 static inline void release_fault_lock(struct vm_fault
*vmf
)
750 if (vmf
->flags
& FAULT_FLAG_VMA_LOCK
)
751 vma_end_read(vmf
->vma
);
753 mmap_read_unlock(vmf
->vma
->vm_mm
);
756 static inline void assert_fault_locked(struct vm_fault
*vmf
)
758 if (vmf
->flags
& FAULT_FLAG_VMA_LOCK
)
759 vma_assert_locked(vmf
->vma
);
761 mmap_assert_locked(vmf
->vma
->vm_mm
);
764 struct vm_area_struct
*lock_vma_under_rcu(struct mm_struct
*mm
,
765 unsigned long address
);
767 #else /* CONFIG_PER_VMA_LOCK */
769 static inline bool vma_start_read(struct vm_area_struct
*vma
)
771 static inline void vma_end_read(struct vm_area_struct
*vma
) {}
772 static inline void vma_start_write(struct vm_area_struct
*vma
) {}
773 static inline void vma_assert_write_locked(struct vm_area_struct
*vma
)
774 { mmap_assert_write_locked(vma
->vm_mm
); }
775 static inline void vma_mark_detached(struct vm_area_struct
*vma
,
778 static inline struct vm_area_struct
*lock_vma_under_rcu(struct mm_struct
*mm
,
779 unsigned long address
)
784 static inline void release_fault_lock(struct vm_fault
*vmf
)
786 mmap_read_unlock(vmf
->vma
->vm_mm
);
789 static inline void assert_fault_locked(struct vm_fault
*vmf
)
791 mmap_assert_locked(vmf
->vma
->vm_mm
);
794 #endif /* CONFIG_PER_VMA_LOCK */
796 extern const struct vm_operations_struct vma_dummy_vm_ops
;
799 * WARNING: vma_init does not initialize vma->vm_lock.
800 * Use vm_area_alloc()/vm_area_free() if vma needs locking.
802 static inline void vma_init(struct vm_area_struct
*vma
, struct mm_struct
*mm
)
804 memset(vma
, 0, sizeof(*vma
));
806 vma
->vm_ops
= &vma_dummy_vm_ops
;
807 INIT_LIST_HEAD(&vma
->anon_vma_chain
);
808 vma_mark_detached(vma
, false);
809 vma_numab_state_init(vma
);
812 /* Use when VMA is not part of the VMA tree and needs no locking */
813 static inline void vm_flags_init(struct vm_area_struct
*vma
,
816 ACCESS_PRIVATE(vma
, __vm_flags
) = flags
;
820 * Use when VMA is part of the VMA tree and modifications need coordination
821 * Note: vm_flags_reset and vm_flags_reset_once do not lock the vma and
822 * it should be locked explicitly beforehand.
824 static inline void vm_flags_reset(struct vm_area_struct
*vma
,
827 vma_assert_write_locked(vma
);
828 vm_flags_init(vma
, flags
);
831 static inline void vm_flags_reset_once(struct vm_area_struct
*vma
,
834 vma_assert_write_locked(vma
);
835 WRITE_ONCE(ACCESS_PRIVATE(vma
, __vm_flags
), flags
);
838 static inline void vm_flags_set(struct vm_area_struct
*vma
,
841 vma_start_write(vma
);
842 ACCESS_PRIVATE(vma
, __vm_flags
) |= flags
;
845 static inline void vm_flags_clear(struct vm_area_struct
*vma
,
848 vma_start_write(vma
);
849 ACCESS_PRIVATE(vma
, __vm_flags
) &= ~flags
;
853 * Use only if VMA is not part of the VMA tree or has no other users and
854 * therefore needs no locking.
856 static inline void __vm_flags_mod(struct vm_area_struct
*vma
,
857 vm_flags_t set
, vm_flags_t clear
)
859 vm_flags_init(vma
, (vma
->vm_flags
| set
) & ~clear
);
863 * Use only when the order of set/clear operations is unimportant, otherwise
864 * use vm_flags_{set|clear} explicitly.
866 static inline void vm_flags_mod(struct vm_area_struct
*vma
,
867 vm_flags_t set
, vm_flags_t clear
)
869 vma_start_write(vma
);
870 __vm_flags_mod(vma
, set
, clear
);
873 static inline void vma_set_anonymous(struct vm_area_struct
*vma
)
878 static inline bool vma_is_anonymous(struct vm_area_struct
*vma
)
884 * Indicate if the VMA is a heap for the given task; for
885 * /proc/PID/maps that is the heap of the main task.
887 static inline bool vma_is_initial_heap(const struct vm_area_struct
*vma
)
889 return vma
->vm_start
<= vma
->vm_mm
->brk
&&
890 vma
->vm_end
>= vma
->vm_mm
->start_brk
;
894 * Indicate if the VMA is a stack for the given task; for
895 * /proc/PID/maps that is the stack of the main task.
897 static inline bool vma_is_initial_stack(const struct vm_area_struct
*vma
)
900 * We make no effort to guess what a given thread considers to be
901 * its "stack". It's not even well-defined for programs written
904 return vma
->vm_start
<= vma
->vm_mm
->start_stack
&&
905 vma
->vm_end
>= vma
->vm_mm
->start_stack
;
908 static inline bool vma_is_temporary_stack(struct vm_area_struct
*vma
)
910 int maybe_stack
= vma
->vm_flags
& (VM_GROWSDOWN
| VM_GROWSUP
);
915 if ((vma
->vm_flags
& VM_STACK_INCOMPLETE_SETUP
) ==
916 VM_STACK_INCOMPLETE_SETUP
)
922 static inline bool vma_is_foreign(struct vm_area_struct
*vma
)
927 if (current
->mm
!= vma
->vm_mm
)
933 static inline bool vma_is_accessible(struct vm_area_struct
*vma
)
935 return vma
->vm_flags
& VM_ACCESS_FLAGS
;
938 static inline bool is_shared_maywrite(vm_flags_t vm_flags
)
940 return (vm_flags
& (VM_SHARED
| VM_MAYWRITE
)) ==
941 (VM_SHARED
| VM_MAYWRITE
);
944 static inline bool vma_is_shared_maywrite(struct vm_area_struct
*vma
)
946 return is_shared_maywrite(vma
->vm_flags
);
950 struct vm_area_struct
*vma_find(struct vma_iterator
*vmi
, unsigned long max
)
952 return mas_find(&vmi
->mas
, max
- 1);
955 static inline struct vm_area_struct
*vma_next(struct vma_iterator
*vmi
)
958 * Uses mas_find() to get the first VMA when the iterator starts.
959 * Calling mas_next() could skip the first entry.
961 return mas_find(&vmi
->mas
, ULONG_MAX
);
965 struct vm_area_struct
*vma_iter_next_range(struct vma_iterator
*vmi
)
967 return mas_next_range(&vmi
->mas
, ULONG_MAX
);
971 static inline struct vm_area_struct
*vma_prev(struct vma_iterator
*vmi
)
973 return mas_prev(&vmi
->mas
, 0);
977 struct vm_area_struct
*vma_iter_prev_range(struct vma_iterator
*vmi
)
979 return mas_prev_range(&vmi
->mas
, 0);
982 static inline unsigned long vma_iter_addr(struct vma_iterator
*vmi
)
984 return vmi
->mas
.index
;
987 static inline unsigned long vma_iter_end(struct vma_iterator
*vmi
)
989 return vmi
->mas
.last
+ 1;
991 static inline int vma_iter_bulk_alloc(struct vma_iterator
*vmi
,
994 return mas_expected_entries(&vmi
->mas
, count
);
997 /* Free any unused preallocations */
998 static inline void vma_iter_free(struct vma_iterator
*vmi
)
1000 mas_destroy(&vmi
->mas
);
1003 static inline int vma_iter_bulk_store(struct vma_iterator
*vmi
,
1004 struct vm_area_struct
*vma
)
1006 vmi
->mas
.index
= vma
->vm_start
;
1007 vmi
->mas
.last
= vma
->vm_end
- 1;
1008 mas_store(&vmi
->mas
, vma
);
1009 if (unlikely(mas_is_err(&vmi
->mas
)))
1015 static inline void vma_iter_invalidate(struct vma_iterator
*vmi
)
1017 mas_pause(&vmi
->mas
);
1020 static inline void vma_iter_set(struct vma_iterator
*vmi
, unsigned long addr
)
1022 mas_set(&vmi
->mas
, addr
);
1025 #define for_each_vma(__vmi, __vma) \
1026 while (((__vma) = vma_next(&(__vmi))) != NULL)
1028 /* The MM code likes to work with exclusive end addresses */
1029 #define for_each_vma_range(__vmi, __vma, __end) \
1030 while (((__vma) = vma_find(&(__vmi), (__end))) != NULL)
1034 * The vma_is_shmem is not inline because it is used only by slow
1035 * paths in userfault.
1037 bool vma_is_shmem(struct vm_area_struct
*vma
);
1038 bool vma_is_anon_shmem(struct vm_area_struct
*vma
);
1040 static inline bool vma_is_shmem(struct vm_area_struct
*vma
) { return false; }
1041 static inline bool vma_is_anon_shmem(struct vm_area_struct
*vma
) { return false; }
1044 int vma_is_stack_for_current(struct vm_area_struct
*vma
);
1046 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
1047 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
1053 * compound_order() can be called without holding a reference, which means
1054 * that niceties like page_folio() don't work. These callers should be
1055 * prepared to handle wild return values. For example, PG_head may be
1056 * set before the order is initialised, or this may be a tail page.
1057 * See compaction.c for some good examples.
1059 static inline unsigned int compound_order(struct page
*page
)
1061 struct folio
*folio
= (struct folio
*)page
;
1063 if (!test_bit(PG_head
, &folio
->flags
))
1065 return folio
->_flags_1
& 0xff;
1069 * folio_order - The allocation order of a folio.
1070 * @folio: The folio.
1072 * A folio is composed of 2^order pages. See get_order() for the definition
1075 * Return: The order of the folio.
1077 static inline unsigned int folio_order(struct folio
*folio
)
1079 if (!folio_test_large(folio
))
1081 return folio
->_flags_1
& 0xff;
1084 #include <linux/huge_mm.h>
1087 * Methods to modify the page usage count.
1089 * What counts for a page usage:
1090 * - cache mapping (page->mapping)
1091 * - private data (page->private)
1092 * - page mapped in a task's page tables, each mapping
1093 * is counted separately
1095 * Also, many kernel routines increase the page count before a critical
1096 * routine so they can be sure the page doesn't go away from under them.
1100 * Drop a ref, return true if the refcount fell to zero (the page has no users)
1102 static inline int put_page_testzero(struct page
*page
)
1104 VM_BUG_ON_PAGE(page_ref_count(page
) == 0, page
);
1105 return page_ref_dec_and_test(page
);
1108 static inline int folio_put_testzero(struct folio
*folio
)
1110 return put_page_testzero(&folio
->page
);
1114 * Try to grab a ref unless the page has a refcount of zero, return false if
1116 * This can be called when MMU is off so it must not access
1117 * any of the virtual mappings.
1119 static inline bool get_page_unless_zero(struct page
*page
)
1121 return page_ref_add_unless(page
, 1, 0);
1124 static inline struct folio
*folio_get_nontail_page(struct page
*page
)
1126 if (unlikely(!get_page_unless_zero(page
)))
1128 return (struct folio
*)page
;
1131 extern int page_is_ram(unsigned long pfn
);
1139 int region_intersects(resource_size_t offset
, size_t size
, unsigned long flags
,
1140 unsigned long desc
);
1142 /* Support for virtually mapped pages */
1143 struct page
*vmalloc_to_page(const void *addr
);
1144 unsigned long vmalloc_to_pfn(const void *addr
);
1147 * Determine if an address is within the vmalloc range
1149 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
1150 * is no special casing required.
1153 extern bool is_vmalloc_addr(const void *x
);
1154 extern int is_vmalloc_or_module_addr(const void *x
);
1156 static inline bool is_vmalloc_addr(const void *x
)
1160 static inline int is_vmalloc_or_module_addr(const void *x
)
1167 * How many times the entire folio is mapped as a single unit (eg by a
1168 * PMD or PUD entry). This is probably not what you want, except for
1169 * debugging purposes - it does not include PTE-mapped sub-pages; look
1170 * at folio_mapcount() or page_mapcount() or total_mapcount() instead.
1172 static inline int folio_entire_mapcount(struct folio
*folio
)
1174 VM_BUG_ON_FOLIO(!folio_test_large(folio
), folio
);
1175 return atomic_read(&folio
->_entire_mapcount
) + 1;
1179 * The atomic page->_mapcount, starts from -1: so that transitions
1180 * both from it and to it can be tracked, using atomic_inc_and_test
1181 * and atomic_add_negative(-1).
1183 static inline void page_mapcount_reset(struct page
*page
)
1185 atomic_set(&(page
)->_mapcount
, -1);
1189 * page_mapcount() - Number of times this precise page is mapped.
1192 * The number of times this page is mapped. If this page is part of
1193 * a large folio, it includes the number of times this page is mapped
1194 * as part of that folio.
1196 * The result is undefined for pages which cannot be mapped into userspace.
1197 * For example SLAB or special types of pages. See function page_has_type().
1198 * They use this field in struct page differently.
1200 static inline int page_mapcount(struct page
*page
)
1202 int mapcount
= atomic_read(&page
->_mapcount
) + 1;
1204 if (unlikely(PageCompound(page
)))
1205 mapcount
+= folio_entire_mapcount(page_folio(page
));
1210 int folio_total_mapcount(struct folio
*folio
);
1213 * folio_mapcount() - Calculate the number of mappings of this folio.
1214 * @folio: The folio.
1216 * A large folio tracks both how many times the entire folio is mapped,
1217 * and how many times each individual page in the folio is mapped.
1218 * This function calculates the total number of times the folio is
1221 * Return: The number of times this folio is mapped.
1223 static inline int folio_mapcount(struct folio
*folio
)
1225 if (likely(!folio_test_large(folio
)))
1226 return atomic_read(&folio
->_mapcount
) + 1;
1227 return folio_total_mapcount(folio
);
1230 static inline int total_mapcount(struct page
*page
)
1232 if (likely(!PageCompound(page
)))
1233 return atomic_read(&page
->_mapcount
) + 1;
1234 return folio_total_mapcount(page_folio(page
));
1237 static inline bool folio_large_is_mapped(struct folio
*folio
)
1240 * Reading _entire_mapcount below could be omitted if hugetlb
1241 * participated in incrementing nr_pages_mapped when compound mapped.
1243 return atomic_read(&folio
->_nr_pages_mapped
) > 0 ||
1244 atomic_read(&folio
->_entire_mapcount
) >= 0;
1248 * folio_mapped - Is this folio mapped into userspace?
1249 * @folio: The folio.
1251 * Return: True if any page in this folio is referenced by user page tables.
1253 static inline bool folio_mapped(struct folio
*folio
)
1255 if (likely(!folio_test_large(folio
)))
1256 return atomic_read(&folio
->_mapcount
) >= 0;
1257 return folio_large_is_mapped(folio
);
1261 * Return true if this page is mapped into pagetables.
1262 * For compound page it returns true if any sub-page of compound page is mapped,
1263 * even if this particular sub-page is not itself mapped by any PTE or PMD.
1265 static inline bool page_mapped(struct page
*page
)
1267 if (likely(!PageCompound(page
)))
1268 return atomic_read(&page
->_mapcount
) >= 0;
1269 return folio_large_is_mapped(page_folio(page
));
1272 static inline struct page
*virt_to_head_page(const void *x
)
1274 struct page
*page
= virt_to_page(x
);
1276 return compound_head(page
);
1279 static inline struct folio
*virt_to_folio(const void *x
)
1281 struct page
*page
= virt_to_page(x
);
1283 return page_folio(page
);
1286 void __folio_put(struct folio
*folio
);
1288 void put_pages_list(struct list_head
*pages
);
1290 void split_page(struct page
*page
, unsigned int order
);
1291 void folio_copy(struct folio
*dst
, struct folio
*src
);
1293 unsigned long nr_free_buffer_pages(void);
1295 void destroy_large_folio(struct folio
*folio
);
1297 /* Returns the number of bytes in this potentially compound page. */
1298 static inline unsigned long page_size(struct page
*page
)
1300 return PAGE_SIZE
<< compound_order(page
);
1303 /* Returns the number of bits needed for the number of bytes in a page */
1304 static inline unsigned int page_shift(struct page
*page
)
1306 return PAGE_SHIFT
+ compound_order(page
);
1310 * thp_order - Order of a transparent huge page.
1311 * @page: Head page of a transparent huge page.
1313 static inline unsigned int thp_order(struct page
*page
)
1315 VM_BUG_ON_PGFLAGS(PageTail(page
), page
);
1316 return compound_order(page
);
1320 * thp_size - Size of a transparent huge page.
1321 * @page: Head page of a transparent huge page.
1323 * Return: Number of bytes in this page.
1325 static inline unsigned long thp_size(struct page
*page
)
1327 return PAGE_SIZE
<< thp_order(page
);
1332 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
1333 * servicing faults for write access. In the normal case, do always want
1334 * pte_mkwrite. But get_user_pages can cause write faults for mappings
1335 * that do not have writing enabled, when used by access_process_vm.
1337 static inline pte_t
maybe_mkwrite(pte_t pte
, struct vm_area_struct
*vma
)
1339 if (likely(vma
->vm_flags
& VM_WRITE
))
1340 pte
= pte_mkwrite(pte
, vma
);
1344 vm_fault_t
do_set_pmd(struct vm_fault
*vmf
, struct page
*page
);
1345 void set_pte_range(struct vm_fault
*vmf
, struct folio
*folio
,
1346 struct page
*page
, unsigned int nr
, unsigned long addr
);
1348 vm_fault_t
finish_fault(struct vm_fault
*vmf
);
1352 * Multiple processes may "see" the same page. E.g. for untouched
1353 * mappings of /dev/null, all processes see the same page full of
1354 * zeroes, and text pages of executables and shared libraries have
1355 * only one copy in memory, at most, normally.
1357 * For the non-reserved pages, page_count(page) denotes a reference count.
1358 * page_count() == 0 means the page is free. page->lru is then used for
1359 * freelist management in the buddy allocator.
1360 * page_count() > 0 means the page has been allocated.
1362 * Pages are allocated by the slab allocator in order to provide memory
1363 * to kmalloc and kmem_cache_alloc. In this case, the management of the
1364 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
1365 * unless a particular usage is carefully commented. (the responsibility of
1366 * freeing the kmalloc memory is the caller's, of course).
1368 * A page may be used by anyone else who does a __get_free_page().
1369 * In this case, page_count still tracks the references, and should only
1370 * be used through the normal accessor functions. The top bits of page->flags
1371 * and page->virtual store page management information, but all other fields
1372 * are unused and could be used privately, carefully. The management of this
1373 * page is the responsibility of the one who allocated it, and those who have
1374 * subsequently been given references to it.
1376 * The other pages (we may call them "pagecache pages") are completely
1377 * managed by the Linux memory manager: I/O, buffers, swapping etc.
1378 * The following discussion applies only to them.
1380 * A pagecache page contains an opaque `private' member, which belongs to the
1381 * page's address_space. Usually, this is the address of a circular list of
1382 * the page's disk buffers. PG_private must be set to tell the VM to call
1383 * into the filesystem to release these pages.
1385 * A page may belong to an inode's memory mapping. In this case, page->mapping
1386 * is the pointer to the inode, and page->index is the file offset of the page,
1387 * in units of PAGE_SIZE.
1389 * If pagecache pages are not associated with an inode, they are said to be
1390 * anonymous pages. These may become associated with the swapcache, and in that
1391 * case PG_swapcache is set, and page->private is an offset into the swapcache.
1393 * In either case (swapcache or inode backed), the pagecache itself holds one
1394 * reference to the page. Setting PG_private should also increment the
1395 * refcount. The each user mapping also has a reference to the page.
1397 * The pagecache pages are stored in a per-mapping radix tree, which is
1398 * rooted at mapping->i_pages, and indexed by offset.
1399 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
1400 * lists, we instead now tag pages as dirty/writeback in the radix tree.
1402 * All pagecache pages may be subject to I/O:
1403 * - inode pages may need to be read from disk,
1404 * - inode pages which have been modified and are MAP_SHARED may need
1405 * to be written back to the inode on disk,
1406 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
1407 * modified may need to be swapped out to swap space and (later) to be read
1411 #if defined(CONFIG_ZONE_DEVICE) && defined(CONFIG_FS_DAX)
1412 DECLARE_STATIC_KEY_FALSE(devmap_managed_key
);
1414 bool __put_devmap_managed_page_refs(struct page
*page
, int refs
);
1415 static inline bool put_devmap_managed_page_refs(struct page
*page
, int refs
)
1417 if (!static_branch_unlikely(&devmap_managed_key
))
1419 if (!is_zone_device_page(page
))
1421 return __put_devmap_managed_page_refs(page
, refs
);
1423 #else /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1424 static inline bool put_devmap_managed_page_refs(struct page
*page
, int refs
)
1428 #endif /* CONFIG_ZONE_DEVICE && CONFIG_FS_DAX */
1430 static inline bool put_devmap_managed_page(struct page
*page
)
1432 return put_devmap_managed_page_refs(page
, 1);
1435 /* 127: arbitrary random number, small enough to assemble well */
1436 #define folio_ref_zero_or_close_to_overflow(folio) \
1437 ((unsigned int) folio_ref_count(folio) + 127u <= 127u)
1440 * folio_get - Increment the reference count on a folio.
1441 * @folio: The folio.
1443 * Context: May be called in any context, as long as you know that
1444 * you have a refcount on the folio. If you do not already have one,
1445 * folio_try_get() may be the right interface for you to use.
1447 static inline void folio_get(struct folio
*folio
)
1449 VM_BUG_ON_FOLIO(folio_ref_zero_or_close_to_overflow(folio
), folio
);
1450 folio_ref_inc(folio
);
1453 static inline void get_page(struct page
*page
)
1455 folio_get(page_folio(page
));
1458 static inline __must_check
bool try_get_page(struct page
*page
)
1460 page
= compound_head(page
);
1461 if (WARN_ON_ONCE(page_ref_count(page
) <= 0))
1468 * folio_put - Decrement the reference count on a folio.
1469 * @folio: The folio.
1471 * If the folio's reference count reaches zero, the memory will be
1472 * released back to the page allocator and may be used by another
1473 * allocation immediately. Do not access the memory or the struct folio
1474 * after calling folio_put() unless you can be sure that it wasn't the
1477 * Context: May be called in process or interrupt context, but not in NMI
1478 * context. May be called while holding a spinlock.
1480 static inline void folio_put(struct folio
*folio
)
1482 if (folio_put_testzero(folio
))
1487 * folio_put_refs - Reduce the reference count on a folio.
1488 * @folio: The folio.
1489 * @refs: The amount to subtract from the folio's reference count.
1491 * If the folio's reference count reaches zero, the memory will be
1492 * released back to the page allocator and may be used by another
1493 * allocation immediately. Do not access the memory or the struct folio
1494 * after calling folio_put_refs() unless you can be sure that these weren't
1495 * the last references.
1497 * Context: May be called in process or interrupt context, but not in NMI
1498 * context. May be called while holding a spinlock.
1500 static inline void folio_put_refs(struct folio
*folio
, int refs
)
1502 if (folio_ref_sub_and_test(folio
, refs
))
1507 * union release_pages_arg - an array of pages or folios
1509 * release_pages() releases a simple array of multiple pages, and
1510 * accepts various different forms of said page array: either
1511 * a regular old boring array of pages, an array of folios, or
1512 * an array of encoded page pointers.
1514 * The transparent union syntax for this kind of "any of these
1515 * argument types" is all kinds of ugly, so look away.
1518 struct page
**pages
;
1519 struct folio
**folios
;
1520 struct encoded_page
**encoded_pages
;
1521 } release_pages_arg
__attribute__ ((__transparent_union__
));
1523 void release_pages(release_pages_arg
, int nr
);
1526 * folios_put - Decrement the reference count on an array of folios.
1527 * @folios: The folios.
1528 * @nr: How many folios there are.
1530 * Like folio_put(), but for an array of folios. This is more efficient
1531 * than writing the loop yourself as it will optimise the locks which
1532 * need to be taken if the folios are freed.
1534 * Context: May be called in process or interrupt context, but not in NMI
1535 * context. May be called while holding a spinlock.
1537 static inline void folios_put(struct folio
**folios
, unsigned int nr
)
1539 release_pages(folios
, nr
);
1542 static inline void put_page(struct page
*page
)
1544 struct folio
*folio
= page_folio(page
);
1547 * For some devmap managed pages we need to catch refcount transition
1550 if (put_devmap_managed_page(&folio
->page
))
1556 * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
1557 * the page's refcount so that two separate items are tracked: the original page
1558 * reference count, and also a new count of how many pin_user_pages() calls were
1559 * made against the page. ("gup-pinned" is another term for the latter).
1561 * With this scheme, pin_user_pages() becomes special: such pages are marked as
1562 * distinct from normal pages. As such, the unpin_user_page() call (and its
1563 * variants) must be used in order to release gup-pinned pages.
1567 * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
1568 * counts with respect to pin_user_pages() and unpin_user_page() becomes
1569 * simpler, due to the fact that adding an even power of two to the page
1570 * refcount has the effect of using only the upper N bits, for the code that
1571 * counts up using the bias value. This means that the lower bits are left for
1572 * the exclusive use of the original code that increments and decrements by one
1573 * (or at least, by much smaller values than the bias value).
1575 * Of course, once the lower bits overflow into the upper bits (and this is
1576 * OK, because subtraction recovers the original values), then visual inspection
1577 * no longer suffices to directly view the separate counts. However, for normal
1578 * applications that don't have huge page reference counts, this won't be an
1581 * Locking: the lockless algorithm described in folio_try_get_rcu()
1582 * provides safe operation for get_user_pages(), page_mkclean() and
1583 * other calls that race to set up page table entries.
1585 #define GUP_PIN_COUNTING_BIAS (1U << 10)
1587 void unpin_user_page(struct page
*page
);
1588 void unpin_user_pages_dirty_lock(struct page
**pages
, unsigned long npages
,
1590 void unpin_user_page_range_dirty_lock(struct page
*page
, unsigned long npages
,
1592 void unpin_user_pages(struct page
**pages
, unsigned long npages
);
1594 static inline bool is_cow_mapping(vm_flags_t flags
)
1596 return (flags
& (VM_SHARED
| VM_MAYWRITE
)) == VM_MAYWRITE
;
1600 static inline bool is_nommu_shared_mapping(vm_flags_t flags
)
1603 * NOMMU shared mappings are ordinary MAP_SHARED mappings and selected
1604 * R/O MAP_PRIVATE file mappings that are an effective R/O overlay of
1605 * a file mapping. R/O MAP_PRIVATE mappings might still modify
1606 * underlying memory if ptrace is active, so this is only possible if
1607 * ptrace does not apply. Note that there is no mprotect() to upgrade
1608 * write permissions later.
1610 return flags
& (VM_MAYSHARE
| VM_MAYOVERLAY
);
1614 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1615 #define SECTION_IN_PAGE_FLAGS
1619 * The identification function is mainly used by the buddy allocator for
1620 * determining if two pages could be buddies. We are not really identifying
1621 * the zone since we could be using the section number id if we do not have
1622 * node id available in page flags.
1623 * We only guarantee that it will return the same value for two combinable
1626 static inline int page_zone_id(struct page
*page
)
1628 return (page
->flags
>> ZONEID_PGSHIFT
) & ZONEID_MASK
;
1631 #ifdef NODE_NOT_IN_PAGE_FLAGS
1632 extern int page_to_nid(const struct page
*page
);
1634 static inline int page_to_nid(const struct page
*page
)
1636 struct page
*p
= (struct page
*)page
;
1638 return (PF_POISONED_CHECK(p
)->flags
>> NODES_PGSHIFT
) & NODES_MASK
;
1642 static inline int folio_nid(const struct folio
*folio
)
1644 return page_to_nid(&folio
->page
);
1647 #ifdef CONFIG_NUMA_BALANCING
1648 /* page access time bits needs to hold at least 4 seconds */
1649 #define PAGE_ACCESS_TIME_MIN_BITS 12
1650 #if LAST_CPUPID_SHIFT < PAGE_ACCESS_TIME_MIN_BITS
1651 #define PAGE_ACCESS_TIME_BUCKETS \
1652 (PAGE_ACCESS_TIME_MIN_BITS - LAST_CPUPID_SHIFT)
1654 #define PAGE_ACCESS_TIME_BUCKETS 0
1657 #define PAGE_ACCESS_TIME_MASK \
1658 (LAST_CPUPID_MASK << PAGE_ACCESS_TIME_BUCKETS)
1660 static inline int cpu_pid_to_cpupid(int cpu
, int pid
)
1662 return ((cpu
& LAST__CPU_MASK
) << LAST__PID_SHIFT
) | (pid
& LAST__PID_MASK
);
1665 static inline int cpupid_to_pid(int cpupid
)
1667 return cpupid
& LAST__PID_MASK
;
1670 static inline int cpupid_to_cpu(int cpupid
)
1672 return (cpupid
>> LAST__PID_SHIFT
) & LAST__CPU_MASK
;
1675 static inline int cpupid_to_nid(int cpupid
)
1677 return cpu_to_node(cpupid_to_cpu(cpupid
));
1680 static inline bool cpupid_pid_unset(int cpupid
)
1682 return cpupid_to_pid(cpupid
) == (-1 & LAST__PID_MASK
);
1685 static inline bool cpupid_cpu_unset(int cpupid
)
1687 return cpupid_to_cpu(cpupid
) == (-1 & LAST__CPU_MASK
);
1690 static inline bool __cpupid_match_pid(pid_t task_pid
, int cpupid
)
1692 return (task_pid
& LAST__PID_MASK
) == cpupid_to_pid(cpupid
);
1695 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1696 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1697 static inline int folio_xchg_last_cpupid(struct folio
*folio
, int cpupid
)
1699 return xchg(&folio
->_last_cpupid
, cpupid
& LAST_CPUPID_MASK
);
1702 static inline int folio_last_cpupid(struct folio
*folio
)
1704 return folio
->_last_cpupid
;
1706 static inline void page_cpupid_reset_last(struct page
*page
)
1708 page
->_last_cpupid
= -1 & LAST_CPUPID_MASK
;
1711 static inline int folio_last_cpupid(struct folio
*folio
)
1713 return (folio
->flags
>> LAST_CPUPID_PGSHIFT
) & LAST_CPUPID_MASK
;
1716 int folio_xchg_last_cpupid(struct folio
*folio
, int cpupid
);
1718 static inline void page_cpupid_reset_last(struct page
*page
)
1720 page
->flags
|= LAST_CPUPID_MASK
<< LAST_CPUPID_PGSHIFT
;
1722 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1724 static inline int folio_xchg_access_time(struct folio
*folio
, int time
)
1728 last_time
= folio_xchg_last_cpupid(folio
,
1729 time
>> PAGE_ACCESS_TIME_BUCKETS
);
1730 return last_time
<< PAGE_ACCESS_TIME_BUCKETS
;
1733 static inline void vma_set_access_pid_bit(struct vm_area_struct
*vma
)
1735 unsigned int pid_bit
;
1737 pid_bit
= hash_32(current
->pid
, ilog2(BITS_PER_LONG
));
1738 if (vma
->numab_state
&& !test_bit(pid_bit
, &vma
->numab_state
->pids_active
[1])) {
1739 __set_bit(pid_bit
, &vma
->numab_state
->pids_active
[1]);
1742 #else /* !CONFIG_NUMA_BALANCING */
1743 static inline int folio_xchg_last_cpupid(struct folio
*folio
, int cpupid
)
1745 return folio_nid(folio
); /* XXX */
1748 static inline int folio_xchg_access_time(struct folio
*folio
, int time
)
1753 static inline int folio_last_cpupid(struct folio
*folio
)
1755 return folio_nid(folio
); /* XXX */
1758 static inline int cpupid_to_nid(int cpupid
)
1763 static inline int cpupid_to_pid(int cpupid
)
1768 static inline int cpupid_to_cpu(int cpupid
)
1773 static inline int cpu_pid_to_cpupid(int nid
, int pid
)
1778 static inline bool cpupid_pid_unset(int cpupid
)
1783 static inline void page_cpupid_reset_last(struct page
*page
)
1787 static inline bool cpupid_match_pid(struct task_struct
*task
, int cpupid
)
1792 static inline void vma_set_access_pid_bit(struct vm_area_struct
*vma
)
1795 #endif /* CONFIG_NUMA_BALANCING */
1797 #if defined(CONFIG_KASAN_SW_TAGS) || defined(CONFIG_KASAN_HW_TAGS)
1800 * KASAN per-page tags are stored xor'ed with 0xff. This allows to avoid
1801 * setting tags for all pages to native kernel tag value 0xff, as the default
1802 * value 0x00 maps to 0xff.
1805 static inline u8
page_kasan_tag(const struct page
*page
)
1809 if (kasan_enabled()) {
1810 tag
= (page
->flags
>> KASAN_TAG_PGSHIFT
) & KASAN_TAG_MASK
;
1817 static inline void page_kasan_tag_set(struct page
*page
, u8 tag
)
1819 unsigned long old_flags
, flags
;
1821 if (!kasan_enabled())
1825 old_flags
= READ_ONCE(page
->flags
);
1828 flags
&= ~(KASAN_TAG_MASK
<< KASAN_TAG_PGSHIFT
);
1829 flags
|= (tag
& KASAN_TAG_MASK
) << KASAN_TAG_PGSHIFT
;
1830 } while (unlikely(!try_cmpxchg(&page
->flags
, &old_flags
, flags
)));
1833 static inline void page_kasan_tag_reset(struct page
*page
)
1835 if (kasan_enabled())
1836 page_kasan_tag_set(page
, 0xff);
1839 #else /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1841 static inline u8
page_kasan_tag(const struct page
*page
)
1846 static inline void page_kasan_tag_set(struct page
*page
, u8 tag
) { }
1847 static inline void page_kasan_tag_reset(struct page
*page
) { }
1849 #endif /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1851 static inline struct zone
*page_zone(const struct page
*page
)
1853 return &NODE_DATA(page_to_nid(page
))->node_zones
[page_zonenum(page
)];
1856 static inline pg_data_t
*page_pgdat(const struct page
*page
)
1858 return NODE_DATA(page_to_nid(page
));
1861 static inline struct zone
*folio_zone(const struct folio
*folio
)
1863 return page_zone(&folio
->page
);
1866 static inline pg_data_t
*folio_pgdat(const struct folio
*folio
)
1868 return page_pgdat(&folio
->page
);
1871 #ifdef SECTION_IN_PAGE_FLAGS
1872 static inline void set_page_section(struct page
*page
, unsigned long section
)
1874 page
->flags
&= ~(SECTIONS_MASK
<< SECTIONS_PGSHIFT
);
1875 page
->flags
|= (section
& SECTIONS_MASK
) << SECTIONS_PGSHIFT
;
1878 static inline unsigned long page_to_section(const struct page
*page
)
1880 return (page
->flags
>> SECTIONS_PGSHIFT
) & SECTIONS_MASK
;
1885 * folio_pfn - Return the Page Frame Number of a folio.
1886 * @folio: The folio.
1888 * A folio may contain multiple pages. The pages have consecutive
1889 * Page Frame Numbers.
1891 * Return: The Page Frame Number of the first page in the folio.
1893 static inline unsigned long folio_pfn(struct folio
*folio
)
1895 return page_to_pfn(&folio
->page
);
1898 static inline struct folio
*pfn_folio(unsigned long pfn
)
1900 return page_folio(pfn_to_page(pfn
));
1904 * folio_maybe_dma_pinned - Report if a folio may be pinned for DMA.
1905 * @folio: The folio.
1907 * This function checks if a folio has been pinned via a call to
1908 * a function in the pin_user_pages() family.
1910 * For small folios, the return value is partially fuzzy: false is not fuzzy,
1911 * because it means "definitely not pinned for DMA", but true means "probably
1912 * pinned for DMA, but possibly a false positive due to having at least
1913 * GUP_PIN_COUNTING_BIAS worth of normal folio references".
1915 * False positives are OK, because: a) it's unlikely for a folio to
1916 * get that many refcounts, and b) all the callers of this routine are
1917 * expected to be able to deal gracefully with a false positive.
1919 * For large folios, the result will be exactly correct. That's because
1920 * we have more tracking data available: the _pincount field is used
1921 * instead of the GUP_PIN_COUNTING_BIAS scheme.
1923 * For more information, please see Documentation/core-api/pin_user_pages.rst.
1925 * Return: True, if it is likely that the page has been "dma-pinned".
1926 * False, if the page is definitely not dma-pinned.
1928 static inline bool folio_maybe_dma_pinned(struct folio
*folio
)
1930 if (folio_test_large(folio
))
1931 return atomic_read(&folio
->_pincount
) > 0;
1934 * folio_ref_count() is signed. If that refcount overflows, then
1935 * folio_ref_count() returns a negative value, and callers will avoid
1936 * further incrementing the refcount.
1938 * Here, for that overflow case, use the sign bit to count a little
1939 * bit higher via unsigned math, and thus still get an accurate result.
1941 return ((unsigned int)folio_ref_count(folio
)) >=
1942 GUP_PIN_COUNTING_BIAS
;
1945 static inline bool page_maybe_dma_pinned(struct page
*page
)
1947 return folio_maybe_dma_pinned(page_folio(page
));
1951 * This should most likely only be called during fork() to see whether we
1952 * should break the cow immediately for an anon page on the src mm.
1954 * The caller has to hold the PT lock and the vma->vm_mm->->write_protect_seq.
1956 static inline bool page_needs_cow_for_dma(struct vm_area_struct
*vma
,
1959 VM_BUG_ON(!(raw_read_seqcount(&vma
->vm_mm
->write_protect_seq
) & 1));
1961 if (!test_bit(MMF_HAS_PINNED
, &vma
->vm_mm
->flags
))
1964 return page_maybe_dma_pinned(page
);
1968 * is_zero_page - Query if a page is a zero page
1969 * @page: The page to query
1971 * This returns true if @page is one of the permanent zero pages.
1973 static inline bool is_zero_page(const struct page
*page
)
1975 return is_zero_pfn(page_to_pfn(page
));
1979 * is_zero_folio - Query if a folio is a zero page
1980 * @folio: The folio to query
1982 * This returns true if @folio is one of the permanent zero pages.
1984 static inline bool is_zero_folio(const struct folio
*folio
)
1986 return is_zero_page(&folio
->page
);
1989 /* MIGRATE_CMA and ZONE_MOVABLE do not allow pin folios */
1990 #ifdef CONFIG_MIGRATION
1991 static inline bool folio_is_longterm_pinnable(struct folio
*folio
)
1994 int mt
= folio_migratetype(folio
);
1996 if (mt
== MIGRATE_CMA
|| mt
== MIGRATE_ISOLATE
)
1999 /* The zero page can be "pinned" but gets special handling. */
2000 if (is_zero_folio(folio
))
2003 /* Coherent device memory must always allow eviction. */
2004 if (folio_is_device_coherent(folio
))
2007 /* Otherwise, non-movable zone folios can be pinned. */
2008 return !folio_is_zone_movable(folio
);
2012 static inline bool folio_is_longterm_pinnable(struct folio
*folio
)
2018 static inline void set_page_zone(struct page
*page
, enum zone_type zone
)
2020 page
->flags
&= ~(ZONES_MASK
<< ZONES_PGSHIFT
);
2021 page
->flags
|= (zone
& ZONES_MASK
) << ZONES_PGSHIFT
;
2024 static inline void set_page_node(struct page
*page
, unsigned long node
)
2026 page
->flags
&= ~(NODES_MASK
<< NODES_PGSHIFT
);
2027 page
->flags
|= (node
& NODES_MASK
) << NODES_PGSHIFT
;
2030 static inline void set_page_links(struct page
*page
, enum zone_type zone
,
2031 unsigned long node
, unsigned long pfn
)
2033 set_page_zone(page
, zone
);
2034 set_page_node(page
, node
);
2035 #ifdef SECTION_IN_PAGE_FLAGS
2036 set_page_section(page
, pfn_to_section_nr(pfn
));
2041 * folio_nr_pages - The number of pages in the folio.
2042 * @folio: The folio.
2044 * Return: A positive power of two.
2046 static inline long folio_nr_pages(struct folio
*folio
)
2048 if (!folio_test_large(folio
))
2051 return folio
->_folio_nr_pages
;
2053 return 1L << (folio
->_flags_1
& 0xff);
2058 * compound_nr() returns the number of pages in this potentially compound
2059 * page. compound_nr() can be called on a tail page, and is defined to
2060 * return 1 in that case.
2062 static inline unsigned long compound_nr(struct page
*page
)
2064 struct folio
*folio
= (struct folio
*)page
;
2066 if (!test_bit(PG_head
, &folio
->flags
))
2069 return folio
->_folio_nr_pages
;
2071 return 1L << (folio
->_flags_1
& 0xff);
2076 * thp_nr_pages - The number of regular pages in this huge page.
2077 * @page: The head page of a huge page.
2079 static inline int thp_nr_pages(struct page
*page
)
2081 return folio_nr_pages((struct folio
*)page
);
2085 * folio_next - Move to the next physical folio.
2086 * @folio: The folio we're currently operating on.
2088 * If you have physically contiguous memory which may span more than
2089 * one folio (eg a &struct bio_vec), use this function to move from one
2090 * folio to the next. Do not use it if the memory is only virtually
2091 * contiguous as the folios are almost certainly not adjacent to each
2092 * other. This is the folio equivalent to writing ``page++``.
2094 * Context: We assume that the folios are refcounted and/or locked at a
2095 * higher level and do not adjust the reference counts.
2096 * Return: The next struct folio.
2098 static inline struct folio
*folio_next(struct folio
*folio
)
2100 return (struct folio
*)folio_page(folio
, folio_nr_pages(folio
));
2104 * folio_shift - The size of the memory described by this folio.
2105 * @folio: The folio.
2107 * A folio represents a number of bytes which is a power-of-two in size.
2108 * This function tells you which power-of-two the folio is. See also
2109 * folio_size() and folio_order().
2111 * Context: The caller should have a reference on the folio to prevent
2112 * it from being split. It is not necessary for the folio to be locked.
2113 * Return: The base-2 logarithm of the size of this folio.
2115 static inline unsigned int folio_shift(struct folio
*folio
)
2117 return PAGE_SHIFT
+ folio_order(folio
);
2121 * folio_size - The number of bytes in a folio.
2122 * @folio: The folio.
2124 * Context: The caller should have a reference on the folio to prevent
2125 * it from being split. It is not necessary for the folio to be locked.
2126 * Return: The number of bytes in this folio.
2128 static inline size_t folio_size(struct folio
*folio
)
2130 return PAGE_SIZE
<< folio_order(folio
);
2134 * folio_estimated_sharers - Estimate the number of sharers of a folio.
2135 * @folio: The folio.
2137 * folio_estimated_sharers() aims to serve as a function to efficiently
2138 * estimate the number of processes sharing a folio. This is done by
2139 * looking at the precise mapcount of the first subpage in the folio, and
2140 * assuming the other subpages are the same. This may not be true for large
2141 * folios. If you want exact mapcounts for exact calculations, look at
2142 * page_mapcount() or folio_total_mapcount().
2144 * Return: The estimated number of processes sharing a folio.
2146 static inline int folio_estimated_sharers(struct folio
*folio
)
2148 return page_mapcount(folio_page(folio
, 0));
2151 #ifndef HAVE_ARCH_MAKE_PAGE_ACCESSIBLE
2152 static inline int arch_make_page_accessible(struct page
*page
)
2158 #ifndef HAVE_ARCH_MAKE_FOLIO_ACCESSIBLE
2159 static inline int arch_make_folio_accessible(struct folio
*folio
)
2162 long i
, nr
= folio_nr_pages(folio
);
2164 for (i
= 0; i
< nr
; i
++) {
2165 ret
= arch_make_page_accessible(folio_page(folio
, i
));
2175 * Some inline functions in vmstat.h depend on page_zone()
2177 #include <linux/vmstat.h>
2179 static __always_inline
void *lowmem_page_address(const struct page
*page
)
2181 return page_to_virt(page
);
2184 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
2185 #define HASHED_PAGE_VIRTUAL
2188 #if defined(WANT_PAGE_VIRTUAL)
2189 static inline void *page_address(const struct page
*page
)
2191 return page
->virtual;
2193 static inline void set_page_address(struct page
*page
, void *address
)
2195 page
->virtual = address
;
2197 #define page_address_init() do { } while(0)
2200 #if defined(HASHED_PAGE_VIRTUAL)
2201 void *page_address(const struct page
*page
);
2202 void set_page_address(struct page
*page
, void *virtual);
2203 void page_address_init(void);
2206 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
2207 #define page_address(page) lowmem_page_address(page)
2208 #define set_page_address(page, address) do { } while(0)
2209 #define page_address_init() do { } while(0)
2212 static inline void *folio_address(const struct folio
*folio
)
2214 return page_address(&folio
->page
);
2217 extern pgoff_t
__page_file_index(struct page
*page
);
2220 * Return the pagecache index of the passed page. Regular pagecache pages
2221 * use ->index whereas swapcache pages use swp_offset(->private)
2223 static inline pgoff_t
page_index(struct page
*page
)
2225 if (unlikely(PageSwapCache(page
)))
2226 return __page_file_index(page
);
2231 * Return true only if the page has been allocated with
2232 * ALLOC_NO_WATERMARKS and the low watermark was not
2233 * met implying that the system is under some pressure.
2235 static inline bool page_is_pfmemalloc(const struct page
*page
)
2238 * lru.next has bit 1 set if the page is allocated from the
2239 * pfmemalloc reserves. Callers may simply overwrite it if
2240 * they do not need to preserve that information.
2242 return (uintptr_t)page
->lru
.next
& BIT(1);
2246 * Return true only if the folio has been allocated with
2247 * ALLOC_NO_WATERMARKS and the low watermark was not
2248 * met implying that the system is under some pressure.
2250 static inline bool folio_is_pfmemalloc(const struct folio
*folio
)
2253 * lru.next has bit 1 set if the page is allocated from the
2254 * pfmemalloc reserves. Callers may simply overwrite it if
2255 * they do not need to preserve that information.
2257 return (uintptr_t)folio
->lru
.next
& BIT(1);
2261 * Only to be called by the page allocator on a freshly allocated
2264 static inline void set_page_pfmemalloc(struct page
*page
)
2266 page
->lru
.next
= (void *)BIT(1);
2269 static inline void clear_page_pfmemalloc(struct page
*page
)
2271 page
->lru
.next
= NULL
;
2275 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
2277 extern void pagefault_out_of_memory(void);
2279 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
2280 #define offset_in_thp(page, p) ((unsigned long)(p) & (thp_size(page) - 1))
2281 #define offset_in_folio(folio, p) ((unsigned long)(p) & (folio_size(folio) - 1))
2284 * Parameter block passed down to zap_pte_range in exceptional cases.
2286 struct zap_details
{
2287 struct folio
*single_folio
; /* Locked folio to be unmapped */
2288 bool even_cows
; /* Zap COWed private pages too? */
2289 zap_flags_t zap_flags
; /* Extra flags for zapping */
2293 * Whether to drop the pte markers, for example, the uffd-wp information for
2294 * file-backed memory. This should only be specified when we will completely
2295 * drop the page in the mm, either by truncation or unmapping of the vma. By
2296 * default, the flag is not set.
2298 #define ZAP_FLAG_DROP_MARKER ((__force zap_flags_t) BIT(0))
2299 /* Set in unmap_vmas() to indicate a final unmap call. Only used by hugetlb */
2300 #define ZAP_FLAG_UNMAP ((__force zap_flags_t) BIT(1))
2302 #ifdef CONFIG_SCHED_MM_CID
2303 void sched_mm_cid_before_execve(struct task_struct
*t
);
2304 void sched_mm_cid_after_execve(struct task_struct
*t
);
2305 void sched_mm_cid_fork(struct task_struct
*t
);
2306 void sched_mm_cid_exit_signals(struct task_struct
*t
);
2307 static inline int task_mm_cid(struct task_struct
*t
)
2312 static inline void sched_mm_cid_before_execve(struct task_struct
*t
) { }
2313 static inline void sched_mm_cid_after_execve(struct task_struct
*t
) { }
2314 static inline void sched_mm_cid_fork(struct task_struct
*t
) { }
2315 static inline void sched_mm_cid_exit_signals(struct task_struct
*t
) { }
2316 static inline int task_mm_cid(struct task_struct
*t
)
2319 * Use the processor id as a fall-back when the mm cid feature is
2320 * disabled. This provides functional per-cpu data structure accesses
2321 * in user-space, althrough it won't provide the memory usage benefits.
2323 return raw_smp_processor_id();
2328 extern bool can_do_mlock(void);
2330 static inline bool can_do_mlock(void) { return false; }
2332 extern int user_shm_lock(size_t, struct ucounts
*);
2333 extern void user_shm_unlock(size_t, struct ucounts
*);
2335 struct folio
*vm_normal_folio(struct vm_area_struct
*vma
, unsigned long addr
,
2337 struct page
*vm_normal_page(struct vm_area_struct
*vma
, unsigned long addr
,
2339 struct folio
*vm_normal_folio_pmd(struct vm_area_struct
*vma
,
2340 unsigned long addr
, pmd_t pmd
);
2341 struct page
*vm_normal_page_pmd(struct vm_area_struct
*vma
, unsigned long addr
,
2344 void zap_vma_ptes(struct vm_area_struct
*vma
, unsigned long address
,
2345 unsigned long size
);
2346 void zap_page_range_single(struct vm_area_struct
*vma
, unsigned long address
,
2347 unsigned long size
, struct zap_details
*details
);
2348 static inline void zap_vma_pages(struct vm_area_struct
*vma
)
2350 zap_page_range_single(vma
, vma
->vm_start
,
2351 vma
->vm_end
- vma
->vm_start
, NULL
);
2353 void unmap_vmas(struct mmu_gather
*tlb
, struct ma_state
*mas
,
2354 struct vm_area_struct
*start_vma
, unsigned long start
,
2355 unsigned long end
, unsigned long tree_end
, bool mm_wr_locked
);
2357 struct mmu_notifier_range
;
2359 void free_pgd_range(struct mmu_gather
*tlb
, unsigned long addr
,
2360 unsigned long end
, unsigned long floor
, unsigned long ceiling
);
2362 copy_page_range(struct vm_area_struct
*dst_vma
, struct vm_area_struct
*src_vma
);
2363 int follow_pte(struct mm_struct
*mm
, unsigned long address
,
2364 pte_t
**ptepp
, spinlock_t
**ptlp
);
2365 int follow_pfn(struct vm_area_struct
*vma
, unsigned long address
,
2366 unsigned long *pfn
);
2367 int follow_phys(struct vm_area_struct
*vma
, unsigned long address
,
2368 unsigned int flags
, unsigned long *prot
, resource_size_t
*phys
);
2369 int generic_access_phys(struct vm_area_struct
*vma
, unsigned long addr
,
2370 void *buf
, int len
, int write
);
2372 extern void truncate_pagecache(struct inode
*inode
, loff_t
new);
2373 extern void truncate_setsize(struct inode
*inode
, loff_t newsize
);
2374 void pagecache_isize_extended(struct inode
*inode
, loff_t from
, loff_t to
);
2375 void truncate_pagecache_range(struct inode
*inode
, loff_t offset
, loff_t end
);
2376 int generic_error_remove_page(struct address_space
*mapping
, struct page
*page
);
2378 struct vm_area_struct
*lock_mm_and_find_vma(struct mm_struct
*mm
,
2379 unsigned long address
, struct pt_regs
*regs
);
2382 extern vm_fault_t
handle_mm_fault(struct vm_area_struct
*vma
,
2383 unsigned long address
, unsigned int flags
,
2384 struct pt_regs
*regs
);
2385 extern int fixup_user_fault(struct mm_struct
*mm
,
2386 unsigned long address
, unsigned int fault_flags
,
2388 void unmap_mapping_pages(struct address_space
*mapping
,
2389 pgoff_t start
, pgoff_t nr
, bool even_cows
);
2390 void unmap_mapping_range(struct address_space
*mapping
,
2391 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
);
2393 static inline vm_fault_t
handle_mm_fault(struct vm_area_struct
*vma
,
2394 unsigned long address
, unsigned int flags
,
2395 struct pt_regs
*regs
)
2397 /* should never happen if there's no MMU */
2399 return VM_FAULT_SIGBUS
;
2401 static inline int fixup_user_fault(struct mm_struct
*mm
, unsigned long address
,
2402 unsigned int fault_flags
, bool *unlocked
)
2404 /* should never happen if there's no MMU */
2408 static inline void unmap_mapping_pages(struct address_space
*mapping
,
2409 pgoff_t start
, pgoff_t nr
, bool even_cows
) { }
2410 static inline void unmap_mapping_range(struct address_space
*mapping
,
2411 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
) { }
2414 static inline void unmap_shared_mapping_range(struct address_space
*mapping
,
2415 loff_t
const holebegin
, loff_t
const holelen
)
2417 unmap_mapping_range(mapping
, holebegin
, holelen
, 0);
2420 static inline struct vm_area_struct
*vma_lookup(struct mm_struct
*mm
,
2421 unsigned long addr
);
2423 extern int access_process_vm(struct task_struct
*tsk
, unsigned long addr
,
2424 void *buf
, int len
, unsigned int gup_flags
);
2425 extern int access_remote_vm(struct mm_struct
*mm
, unsigned long addr
,
2426 void *buf
, int len
, unsigned int gup_flags
);
2428 long get_user_pages_remote(struct mm_struct
*mm
,
2429 unsigned long start
, unsigned long nr_pages
,
2430 unsigned int gup_flags
, struct page
**pages
,
2432 long pin_user_pages_remote(struct mm_struct
*mm
,
2433 unsigned long start
, unsigned long nr_pages
,
2434 unsigned int gup_flags
, struct page
**pages
,
2438 * Retrieves a single page alongside its VMA. Does not support FOLL_NOWAIT.
2440 static inline struct page
*get_user_page_vma_remote(struct mm_struct
*mm
,
2443 struct vm_area_struct
**vmap
)
2446 struct vm_area_struct
*vma
;
2449 if (WARN_ON_ONCE(unlikely(gup_flags
& FOLL_NOWAIT
)))
2450 return ERR_PTR(-EINVAL
);
2452 got
= get_user_pages_remote(mm
, addr
, 1, gup_flags
, &page
, NULL
);
2455 return ERR_PTR(got
);
2457 vma
= vma_lookup(mm
, addr
);
2458 if (WARN_ON_ONCE(!vma
)) {
2460 return ERR_PTR(-EINVAL
);
2467 long get_user_pages(unsigned long start
, unsigned long nr_pages
,
2468 unsigned int gup_flags
, struct page
**pages
);
2469 long pin_user_pages(unsigned long start
, unsigned long nr_pages
,
2470 unsigned int gup_flags
, struct page
**pages
);
2471 long get_user_pages_unlocked(unsigned long start
, unsigned long nr_pages
,
2472 struct page
**pages
, unsigned int gup_flags
);
2473 long pin_user_pages_unlocked(unsigned long start
, unsigned long nr_pages
,
2474 struct page
**pages
, unsigned int gup_flags
);
2476 int get_user_pages_fast(unsigned long start
, int nr_pages
,
2477 unsigned int gup_flags
, struct page
**pages
);
2478 int pin_user_pages_fast(unsigned long start
, int nr_pages
,
2479 unsigned int gup_flags
, struct page
**pages
);
2480 void folio_add_pin(struct folio
*folio
);
2482 int account_locked_vm(struct mm_struct
*mm
, unsigned long pages
, bool inc
);
2483 int __account_locked_vm(struct mm_struct
*mm
, unsigned long pages
, bool inc
,
2484 struct task_struct
*task
, bool bypass_rlim
);
2487 struct page
*get_dump_page(unsigned long addr
);
2489 bool folio_mark_dirty(struct folio
*folio
);
2490 bool set_page_dirty(struct page
*page
);
2491 int set_page_dirty_lock(struct page
*page
);
2493 int get_cmdline(struct task_struct
*task
, char *buffer
, int buflen
);
2495 extern unsigned long move_page_tables(struct vm_area_struct
*vma
,
2496 unsigned long old_addr
, struct vm_area_struct
*new_vma
,
2497 unsigned long new_addr
, unsigned long len
,
2498 bool need_rmap_locks
, bool for_stack
);
2501 * Flags used by change_protection(). For now we make it a bitmap so
2502 * that we can pass in multiple flags just like parameters. However
2503 * for now all the callers are only use one of the flags at the same
2507 * Whether we should manually check if we can map individual PTEs writable,
2508 * because something (e.g., COW, uffd-wp) blocks that from happening for all
2509 * PTEs automatically in a writable mapping.
2511 #define MM_CP_TRY_CHANGE_WRITABLE (1UL << 0)
2512 /* Whether this protection change is for NUMA hints */
2513 #define MM_CP_PROT_NUMA (1UL << 1)
2514 /* Whether this change is for write protecting */
2515 #define MM_CP_UFFD_WP (1UL << 2) /* do wp */
2516 #define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */
2517 #define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \
2518 MM_CP_UFFD_WP_RESOLVE)
2520 bool vma_needs_dirty_tracking(struct vm_area_struct
*vma
);
2521 int vma_wants_writenotify(struct vm_area_struct
*vma
, pgprot_t vm_page_prot
);
2522 static inline bool vma_wants_manual_pte_write_upgrade(struct vm_area_struct
*vma
)
2525 * We want to check manually if we can change individual PTEs writable
2526 * if we can't do that automatically for all PTEs in a mapping. For
2527 * private mappings, that's always the case when we have write
2528 * permissions as we properly have to handle COW.
2530 if (vma
->vm_flags
& VM_SHARED
)
2531 return vma_wants_writenotify(vma
, vma
->vm_page_prot
);
2532 return !!(vma
->vm_flags
& VM_WRITE
);
2535 bool can_change_pte_writable(struct vm_area_struct
*vma
, unsigned long addr
,
2537 extern long change_protection(struct mmu_gather
*tlb
,
2538 struct vm_area_struct
*vma
, unsigned long start
,
2539 unsigned long end
, unsigned long cp_flags
);
2540 extern int mprotect_fixup(struct vma_iterator
*vmi
, struct mmu_gather
*tlb
,
2541 struct vm_area_struct
*vma
, struct vm_area_struct
**pprev
,
2542 unsigned long start
, unsigned long end
, unsigned long newflags
);
2545 * doesn't attempt to fault and will return short.
2547 int get_user_pages_fast_only(unsigned long start
, int nr_pages
,
2548 unsigned int gup_flags
, struct page
**pages
);
2550 static inline bool get_user_page_fast_only(unsigned long addr
,
2551 unsigned int gup_flags
, struct page
**pagep
)
2553 return get_user_pages_fast_only(addr
, 1, gup_flags
, pagep
) == 1;
2556 * per-process(per-mm_struct) statistics.
2558 static inline unsigned long get_mm_counter(struct mm_struct
*mm
, int member
)
2560 return percpu_counter_read_positive(&mm
->rss_stat
[member
]);
2563 void mm_trace_rss_stat(struct mm_struct
*mm
, int member
);
2565 static inline void add_mm_counter(struct mm_struct
*mm
, int member
, long value
)
2567 percpu_counter_add(&mm
->rss_stat
[member
], value
);
2569 mm_trace_rss_stat(mm
, member
);
2572 static inline void inc_mm_counter(struct mm_struct
*mm
, int member
)
2574 percpu_counter_inc(&mm
->rss_stat
[member
]);
2576 mm_trace_rss_stat(mm
, member
);
2579 static inline void dec_mm_counter(struct mm_struct
*mm
, int member
)
2581 percpu_counter_dec(&mm
->rss_stat
[member
]);
2583 mm_trace_rss_stat(mm
, member
);
2586 /* Optimized variant when page is already known not to be PageAnon */
2587 static inline int mm_counter_file(struct page
*page
)
2589 if (PageSwapBacked(page
))
2590 return MM_SHMEMPAGES
;
2591 return MM_FILEPAGES
;
2594 static inline int mm_counter(struct page
*page
)
2597 return MM_ANONPAGES
;
2598 return mm_counter_file(page
);
2601 static inline unsigned long get_mm_rss(struct mm_struct
*mm
)
2603 return get_mm_counter(mm
, MM_FILEPAGES
) +
2604 get_mm_counter(mm
, MM_ANONPAGES
) +
2605 get_mm_counter(mm
, MM_SHMEMPAGES
);
2608 static inline unsigned long get_mm_hiwater_rss(struct mm_struct
*mm
)
2610 return max(mm
->hiwater_rss
, get_mm_rss(mm
));
2613 static inline unsigned long get_mm_hiwater_vm(struct mm_struct
*mm
)
2615 return max(mm
->hiwater_vm
, mm
->total_vm
);
2618 static inline void update_hiwater_rss(struct mm_struct
*mm
)
2620 unsigned long _rss
= get_mm_rss(mm
);
2622 if ((mm
)->hiwater_rss
< _rss
)
2623 (mm
)->hiwater_rss
= _rss
;
2626 static inline void update_hiwater_vm(struct mm_struct
*mm
)
2628 if (mm
->hiwater_vm
< mm
->total_vm
)
2629 mm
->hiwater_vm
= mm
->total_vm
;
2632 static inline void reset_mm_hiwater_rss(struct mm_struct
*mm
)
2634 mm
->hiwater_rss
= get_mm_rss(mm
);
2637 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss
,
2638 struct mm_struct
*mm
)
2640 unsigned long hiwater_rss
= get_mm_hiwater_rss(mm
);
2642 if (*maxrss
< hiwater_rss
)
2643 *maxrss
= hiwater_rss
;
2646 #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
2647 static inline int pte_special(pte_t pte
)
2652 static inline pte_t
pte_mkspecial(pte_t pte
)
2658 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
2659 static inline int pte_devmap(pte_t pte
)
2665 extern pte_t
*__get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
2667 static inline pte_t
*get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
2671 __cond_lock(*ptl
, ptep
= __get_locked_pte(mm
, addr
, ptl
));
2675 #ifdef __PAGETABLE_P4D_FOLDED
2676 static inline int __p4d_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
2677 unsigned long address
)
2682 int __p4d_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
);
2685 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
2686 static inline int __pud_alloc(struct mm_struct
*mm
, p4d_t
*p4d
,
2687 unsigned long address
)
2691 static inline void mm_inc_nr_puds(struct mm_struct
*mm
) {}
2692 static inline void mm_dec_nr_puds(struct mm_struct
*mm
) {}
2695 int __pud_alloc(struct mm_struct
*mm
, p4d_t
*p4d
, unsigned long address
);
2697 static inline void mm_inc_nr_puds(struct mm_struct
*mm
)
2699 if (mm_pud_folded(mm
))
2701 atomic_long_add(PTRS_PER_PUD
* sizeof(pud_t
), &mm
->pgtables_bytes
);
2704 static inline void mm_dec_nr_puds(struct mm_struct
*mm
)
2706 if (mm_pud_folded(mm
))
2708 atomic_long_sub(PTRS_PER_PUD
* sizeof(pud_t
), &mm
->pgtables_bytes
);
2712 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
2713 static inline int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
,
2714 unsigned long address
)
2719 static inline void mm_inc_nr_pmds(struct mm_struct
*mm
) {}
2720 static inline void mm_dec_nr_pmds(struct mm_struct
*mm
) {}
2723 int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
);
2725 static inline void mm_inc_nr_pmds(struct mm_struct
*mm
)
2727 if (mm_pmd_folded(mm
))
2729 atomic_long_add(PTRS_PER_PMD
* sizeof(pmd_t
), &mm
->pgtables_bytes
);
2732 static inline void mm_dec_nr_pmds(struct mm_struct
*mm
)
2734 if (mm_pmd_folded(mm
))
2736 atomic_long_sub(PTRS_PER_PMD
* sizeof(pmd_t
), &mm
->pgtables_bytes
);
2741 static inline void mm_pgtables_bytes_init(struct mm_struct
*mm
)
2743 atomic_long_set(&mm
->pgtables_bytes
, 0);
2746 static inline unsigned long mm_pgtables_bytes(const struct mm_struct
*mm
)
2748 return atomic_long_read(&mm
->pgtables_bytes
);
2751 static inline void mm_inc_nr_ptes(struct mm_struct
*mm
)
2753 atomic_long_add(PTRS_PER_PTE
* sizeof(pte_t
), &mm
->pgtables_bytes
);
2756 static inline void mm_dec_nr_ptes(struct mm_struct
*mm
)
2758 atomic_long_sub(PTRS_PER_PTE
* sizeof(pte_t
), &mm
->pgtables_bytes
);
2762 static inline void mm_pgtables_bytes_init(struct mm_struct
*mm
) {}
2763 static inline unsigned long mm_pgtables_bytes(const struct mm_struct
*mm
)
2768 static inline void mm_inc_nr_ptes(struct mm_struct
*mm
) {}
2769 static inline void mm_dec_nr_ptes(struct mm_struct
*mm
) {}
2772 int __pte_alloc(struct mm_struct
*mm
, pmd_t
*pmd
);
2773 int __pte_alloc_kernel(pmd_t
*pmd
);
2775 #if defined(CONFIG_MMU)
2777 static inline p4d_t
*p4d_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
2778 unsigned long address
)
2780 return (unlikely(pgd_none(*pgd
)) && __p4d_alloc(mm
, pgd
, address
)) ?
2781 NULL
: p4d_offset(pgd
, address
);
2784 static inline pud_t
*pud_alloc(struct mm_struct
*mm
, p4d_t
*p4d
,
2785 unsigned long address
)
2787 return (unlikely(p4d_none(*p4d
)) && __pud_alloc(mm
, p4d
, address
)) ?
2788 NULL
: pud_offset(p4d
, address
);
2791 static inline pmd_t
*pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
)
2793 return (unlikely(pud_none(*pud
)) && __pmd_alloc(mm
, pud
, address
))?
2794 NULL
: pmd_offset(pud
, address
);
2796 #endif /* CONFIG_MMU */
2798 static inline struct ptdesc
*virt_to_ptdesc(const void *x
)
2800 return page_ptdesc(virt_to_page(x
));
2803 static inline void *ptdesc_to_virt(const struct ptdesc
*pt
)
2805 return page_to_virt(ptdesc_page(pt
));
2808 static inline void *ptdesc_address(const struct ptdesc
*pt
)
2810 return folio_address(ptdesc_folio(pt
));
2813 static inline bool pagetable_is_reserved(struct ptdesc
*pt
)
2815 return folio_test_reserved(ptdesc_folio(pt
));
2819 * pagetable_alloc - Allocate pagetables
2821 * @order: desired pagetable order
2823 * pagetable_alloc allocates memory for page tables as well as a page table
2824 * descriptor to describe that memory.
2826 * Return: The ptdesc describing the allocated page tables.
2828 static inline struct ptdesc
*pagetable_alloc(gfp_t gfp
, unsigned int order
)
2830 struct page
*page
= alloc_pages(gfp
| __GFP_COMP
, order
);
2832 return page_ptdesc(page
);
2836 * pagetable_free - Free pagetables
2837 * @pt: The page table descriptor
2839 * pagetable_free frees the memory of all page tables described by a page
2840 * table descriptor and the memory for the descriptor itself.
2842 static inline void pagetable_free(struct ptdesc
*pt
)
2844 struct page
*page
= ptdesc_page(pt
);
2846 __free_pages(page
, compound_order(page
));
2849 #if USE_SPLIT_PTE_PTLOCKS
2850 #if ALLOC_SPLIT_PTLOCKS
2851 void __init
ptlock_cache_init(void);
2852 bool ptlock_alloc(struct ptdesc
*ptdesc
);
2853 void ptlock_free(struct ptdesc
*ptdesc
);
2855 static inline spinlock_t
*ptlock_ptr(struct ptdesc
*ptdesc
)
2859 #else /* ALLOC_SPLIT_PTLOCKS */
2860 static inline void ptlock_cache_init(void)
2864 static inline bool ptlock_alloc(struct ptdesc
*ptdesc
)
2869 static inline void ptlock_free(struct ptdesc
*ptdesc
)
2873 static inline spinlock_t
*ptlock_ptr(struct ptdesc
*ptdesc
)
2875 return &ptdesc
->ptl
;
2877 #endif /* ALLOC_SPLIT_PTLOCKS */
2879 static inline spinlock_t
*pte_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
2881 return ptlock_ptr(page_ptdesc(pmd_page(*pmd
)));
2884 static inline bool ptlock_init(struct ptdesc
*ptdesc
)
2887 * prep_new_page() initialize page->private (and therefore page->ptl)
2888 * with 0. Make sure nobody took it in use in between.
2890 * It can happen if arch try to use slab for page table allocation:
2891 * slab code uses page->slab_cache, which share storage with page->ptl.
2893 VM_BUG_ON_PAGE(*(unsigned long *)&ptdesc
->ptl
, ptdesc_page(ptdesc
));
2894 if (!ptlock_alloc(ptdesc
))
2896 spin_lock_init(ptlock_ptr(ptdesc
));
2900 #else /* !USE_SPLIT_PTE_PTLOCKS */
2902 * We use mm->page_table_lock to guard all pagetable pages of the mm.
2904 static inline spinlock_t
*pte_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
2906 return &mm
->page_table_lock
;
2908 static inline void ptlock_cache_init(void) {}
2909 static inline bool ptlock_init(struct ptdesc
*ptdesc
) { return true; }
2910 static inline void ptlock_free(struct ptdesc
*ptdesc
) {}
2911 #endif /* USE_SPLIT_PTE_PTLOCKS */
2913 static inline bool pagetable_pte_ctor(struct ptdesc
*ptdesc
)
2915 struct folio
*folio
= ptdesc_folio(ptdesc
);
2917 if (!ptlock_init(ptdesc
))
2919 __folio_set_pgtable(folio
);
2920 lruvec_stat_add_folio(folio
, NR_PAGETABLE
);
2924 static inline void pagetable_pte_dtor(struct ptdesc
*ptdesc
)
2926 struct folio
*folio
= ptdesc_folio(ptdesc
);
2928 ptlock_free(ptdesc
);
2929 __folio_clear_pgtable(folio
);
2930 lruvec_stat_sub_folio(folio
, NR_PAGETABLE
);
2933 pte_t
*__pte_offset_map(pmd_t
*pmd
, unsigned long addr
, pmd_t
*pmdvalp
);
2934 static inline pte_t
*pte_offset_map(pmd_t
*pmd
, unsigned long addr
)
2936 return __pte_offset_map(pmd
, addr
, NULL
);
2939 pte_t
*__pte_offset_map_lock(struct mm_struct
*mm
, pmd_t
*pmd
,
2940 unsigned long addr
, spinlock_t
**ptlp
);
2941 static inline pte_t
*pte_offset_map_lock(struct mm_struct
*mm
, pmd_t
*pmd
,
2942 unsigned long addr
, spinlock_t
**ptlp
)
2946 __cond_lock(*ptlp
, pte
= __pte_offset_map_lock(mm
, pmd
, addr
, ptlp
));
2950 pte_t
*pte_offset_map_nolock(struct mm_struct
*mm
, pmd_t
*pmd
,
2951 unsigned long addr
, spinlock_t
**ptlp
);
2953 #define pte_unmap_unlock(pte, ptl) do { \
2958 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2960 #define pte_alloc_map(mm, pmd, address) \
2961 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2963 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2964 (pte_alloc(mm, pmd) ? \
2965 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2967 #define pte_alloc_kernel(pmd, address) \
2968 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2969 NULL: pte_offset_kernel(pmd, address))
2971 #if USE_SPLIT_PMD_PTLOCKS
2973 static inline struct page
*pmd_pgtable_page(pmd_t
*pmd
)
2975 unsigned long mask
= ~(PTRS_PER_PMD
* sizeof(pmd_t
) - 1);
2976 return virt_to_page((void *)((unsigned long) pmd
& mask
));
2979 static inline struct ptdesc
*pmd_ptdesc(pmd_t
*pmd
)
2981 return page_ptdesc(pmd_pgtable_page(pmd
));
2984 static inline spinlock_t
*pmd_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
2986 return ptlock_ptr(pmd_ptdesc(pmd
));
2989 static inline bool pmd_ptlock_init(struct ptdesc
*ptdesc
)
2991 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2992 ptdesc
->pmd_huge_pte
= NULL
;
2994 return ptlock_init(ptdesc
);
2997 static inline void pmd_ptlock_free(struct ptdesc
*ptdesc
)
2999 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
3000 VM_BUG_ON_PAGE(ptdesc
->pmd_huge_pte
, ptdesc_page(ptdesc
));
3002 ptlock_free(ptdesc
);
3005 #define pmd_huge_pte(mm, pmd) (pmd_ptdesc(pmd)->pmd_huge_pte)
3009 static inline spinlock_t
*pmd_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
3011 return &mm
->page_table_lock
;
3014 static inline bool pmd_ptlock_init(struct ptdesc
*ptdesc
) { return true; }
3015 static inline void pmd_ptlock_free(struct ptdesc
*ptdesc
) {}
3017 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
3021 static inline spinlock_t
*pmd_lock(struct mm_struct
*mm
, pmd_t
*pmd
)
3023 spinlock_t
*ptl
= pmd_lockptr(mm
, pmd
);
3028 static inline bool pagetable_pmd_ctor(struct ptdesc
*ptdesc
)
3030 struct folio
*folio
= ptdesc_folio(ptdesc
);
3032 if (!pmd_ptlock_init(ptdesc
))
3034 __folio_set_pgtable(folio
);
3035 lruvec_stat_add_folio(folio
, NR_PAGETABLE
);
3039 static inline void pagetable_pmd_dtor(struct ptdesc
*ptdesc
)
3041 struct folio
*folio
= ptdesc_folio(ptdesc
);
3043 pmd_ptlock_free(ptdesc
);
3044 __folio_clear_pgtable(folio
);
3045 lruvec_stat_sub_folio(folio
, NR_PAGETABLE
);
3049 * No scalability reason to split PUD locks yet, but follow the same pattern
3050 * as the PMD locks to make it easier if we decide to. The VM should not be
3051 * considered ready to switch to split PUD locks yet; there may be places
3052 * which need to be converted from page_table_lock.
3054 static inline spinlock_t
*pud_lockptr(struct mm_struct
*mm
, pud_t
*pud
)
3056 return &mm
->page_table_lock
;
3059 static inline spinlock_t
*pud_lock(struct mm_struct
*mm
, pud_t
*pud
)
3061 spinlock_t
*ptl
= pud_lockptr(mm
, pud
);
3067 static inline void pagetable_pud_ctor(struct ptdesc
*ptdesc
)
3069 struct folio
*folio
= ptdesc_folio(ptdesc
);
3071 __folio_set_pgtable(folio
);
3072 lruvec_stat_add_folio(folio
, NR_PAGETABLE
);
3075 static inline void pagetable_pud_dtor(struct ptdesc
*ptdesc
)
3077 struct folio
*folio
= ptdesc_folio(ptdesc
);
3079 __folio_clear_pgtable(folio
);
3080 lruvec_stat_sub_folio(folio
, NR_PAGETABLE
);
3083 extern void __init
pagecache_init(void);
3084 extern void free_initmem(void);
3087 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
3088 * into the buddy system. The freed pages will be poisoned with pattern
3089 * "poison" if it's within range [0, UCHAR_MAX].
3090 * Return pages freed into the buddy system.
3092 extern unsigned long free_reserved_area(void *start
, void *end
,
3093 int poison
, const char *s
);
3095 extern void adjust_managed_page_count(struct page
*page
, long count
);
3097 extern void reserve_bootmem_region(phys_addr_t start
,
3098 phys_addr_t end
, int nid
);
3100 /* Free the reserved page into the buddy system, so it gets managed. */
3101 static inline void free_reserved_page(struct page
*page
)
3103 ClearPageReserved(page
);
3104 init_page_count(page
);
3106 adjust_managed_page_count(page
, 1);
3108 #define free_highmem_page(page) free_reserved_page(page)
3110 static inline void mark_page_reserved(struct page
*page
)
3112 SetPageReserved(page
);
3113 adjust_managed_page_count(page
, -1);
3116 static inline void free_reserved_ptdesc(struct ptdesc
*pt
)
3118 free_reserved_page(ptdesc_page(pt
));
3122 * Default method to free all the __init memory into the buddy system.
3123 * The freed pages will be poisoned with pattern "poison" if it's within
3124 * range [0, UCHAR_MAX].
3125 * Return pages freed into the buddy system.
3127 static inline unsigned long free_initmem_default(int poison
)
3129 extern char __init_begin
[], __init_end
[];
3131 return free_reserved_area(&__init_begin
, &__init_end
,
3132 poison
, "unused kernel image (initmem)");
3135 static inline unsigned long get_num_physpages(void)
3138 unsigned long phys_pages
= 0;
3140 for_each_online_node(nid
)
3141 phys_pages
+= node_present_pages(nid
);
3147 * Using memblock node mappings, an architecture may initialise its
3148 * zones, allocate the backing mem_map and account for memory holes in an
3149 * architecture independent manner.
3151 * An architecture is expected to register range of page frames backed by
3152 * physical memory with memblock_add[_node]() before calling
3153 * free_area_init() passing in the PFN each zone ends at. At a basic
3154 * usage, an architecture is expected to do something like
3156 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
3158 * for_each_valid_physical_page_range()
3159 * memblock_add_node(base, size, nid, MEMBLOCK_NONE)
3160 * free_area_init(max_zone_pfns);
3162 void free_area_init(unsigned long *max_zone_pfn
);
3163 unsigned long node_map_pfn_alignment(void);
3164 unsigned long __absent_pages_in_range(int nid
, unsigned long start_pfn
,
3165 unsigned long end_pfn
);
3166 extern unsigned long absent_pages_in_range(unsigned long start_pfn
,
3167 unsigned long end_pfn
);
3168 extern void get_pfn_range_for_nid(unsigned int nid
,
3169 unsigned long *start_pfn
, unsigned long *end_pfn
);
3172 static inline int early_pfn_to_nid(unsigned long pfn
)
3177 /* please see mm/page_alloc.c */
3178 extern int __meminit
early_pfn_to_nid(unsigned long pfn
);
3181 extern void set_dma_reserve(unsigned long new_dma_reserve
);
3182 extern void mem_init(void);
3183 extern void __init
mmap_init(void);
3185 extern void __show_mem(unsigned int flags
, nodemask_t
*nodemask
, int max_zone_idx
);
3186 static inline void show_mem(void)
3188 __show_mem(0, NULL
, MAX_NR_ZONES
- 1);
3190 extern long si_mem_available(void);
3191 extern void si_meminfo(struct sysinfo
* val
);
3192 extern void si_meminfo_node(struct sysinfo
*val
, int nid
);
3193 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
3194 extern unsigned long arch_reserved_kernel_pages(void);
3197 extern __printf(3, 4)
3198 void warn_alloc(gfp_t gfp_mask
, nodemask_t
*nodemask
, const char *fmt
, ...);
3200 extern void setup_per_cpu_pageset(void);
3203 extern atomic_long_t mmap_pages_allocated
;
3204 extern int nommu_shrink_inode_mappings(struct inode
*, size_t, size_t);
3206 /* interval_tree.c */
3207 void vma_interval_tree_insert(struct vm_area_struct
*node
,
3208 struct rb_root_cached
*root
);
3209 void vma_interval_tree_insert_after(struct vm_area_struct
*node
,
3210 struct vm_area_struct
*prev
,
3211 struct rb_root_cached
*root
);
3212 void vma_interval_tree_remove(struct vm_area_struct
*node
,
3213 struct rb_root_cached
*root
);
3214 struct vm_area_struct
*vma_interval_tree_iter_first(struct rb_root_cached
*root
,
3215 unsigned long start
, unsigned long last
);
3216 struct vm_area_struct
*vma_interval_tree_iter_next(struct vm_area_struct
*node
,
3217 unsigned long start
, unsigned long last
);
3219 #define vma_interval_tree_foreach(vma, root, start, last) \
3220 for (vma = vma_interval_tree_iter_first(root, start, last); \
3221 vma; vma = vma_interval_tree_iter_next(vma, start, last))
3223 void anon_vma_interval_tree_insert(struct anon_vma_chain
*node
,
3224 struct rb_root_cached
*root
);
3225 void anon_vma_interval_tree_remove(struct anon_vma_chain
*node
,
3226 struct rb_root_cached
*root
);
3227 struct anon_vma_chain
*
3228 anon_vma_interval_tree_iter_first(struct rb_root_cached
*root
,
3229 unsigned long start
, unsigned long last
);
3230 struct anon_vma_chain
*anon_vma_interval_tree_iter_next(
3231 struct anon_vma_chain
*node
, unsigned long start
, unsigned long last
);
3232 #ifdef CONFIG_DEBUG_VM_RB
3233 void anon_vma_interval_tree_verify(struct anon_vma_chain
*node
);
3236 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
3237 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
3238 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
3241 extern int __vm_enough_memory(struct mm_struct
*mm
, long pages
, int cap_sys_admin
);
3242 extern int vma_expand(struct vma_iterator
*vmi
, struct vm_area_struct
*vma
,
3243 unsigned long start
, unsigned long end
, pgoff_t pgoff
,
3244 struct vm_area_struct
*next
);
3245 extern int vma_shrink(struct vma_iterator
*vmi
, struct vm_area_struct
*vma
,
3246 unsigned long start
, unsigned long end
, pgoff_t pgoff
);
3247 extern struct anon_vma
*find_mergeable_anon_vma(struct vm_area_struct
*);
3248 extern int insert_vm_struct(struct mm_struct
*, struct vm_area_struct
*);
3249 extern void unlink_file_vma(struct vm_area_struct
*);
3250 extern struct vm_area_struct
*copy_vma(struct vm_area_struct
**,
3251 unsigned long addr
, unsigned long len
, pgoff_t pgoff
,
3252 bool *need_rmap_locks
);
3253 extern void exit_mmap(struct mm_struct
*);
3254 struct vm_area_struct
*vma_modify(struct vma_iterator
*vmi
,
3255 struct vm_area_struct
*prev
,
3256 struct vm_area_struct
*vma
,
3257 unsigned long start
, unsigned long end
,
3258 unsigned long vm_flags
,
3259 struct mempolicy
*policy
,
3260 struct vm_userfaultfd_ctx uffd_ctx
,
3261 struct anon_vma_name
*anon_name
);
3263 /* We are about to modify the VMA's flags. */
3264 static inline struct vm_area_struct
3265 *vma_modify_flags(struct vma_iterator
*vmi
,
3266 struct vm_area_struct
*prev
,
3267 struct vm_area_struct
*vma
,
3268 unsigned long start
, unsigned long end
,
3269 unsigned long new_flags
)
3271 return vma_modify(vmi
, prev
, vma
, start
, end
, new_flags
,
3272 vma_policy(vma
), vma
->vm_userfaultfd_ctx
,
3273 anon_vma_name(vma
));
3276 /* We are about to modify the VMA's flags and/or anon_name. */
3277 static inline struct vm_area_struct
3278 *vma_modify_flags_name(struct vma_iterator
*vmi
,
3279 struct vm_area_struct
*prev
,
3280 struct vm_area_struct
*vma
,
3281 unsigned long start
,
3283 unsigned long new_flags
,
3284 struct anon_vma_name
*new_name
)
3286 return vma_modify(vmi
, prev
, vma
, start
, end
, new_flags
,
3287 vma_policy(vma
), vma
->vm_userfaultfd_ctx
, new_name
);
3290 /* We are about to modify the VMA's memory policy. */
3291 static inline struct vm_area_struct
3292 *vma_modify_policy(struct vma_iterator
*vmi
,
3293 struct vm_area_struct
*prev
,
3294 struct vm_area_struct
*vma
,
3295 unsigned long start
, unsigned long end
,
3296 struct mempolicy
*new_pol
)
3298 return vma_modify(vmi
, prev
, vma
, start
, end
, vma
->vm_flags
,
3299 new_pol
, vma
->vm_userfaultfd_ctx
, anon_vma_name(vma
));
3302 /* We are about to modify the VMA's flags and/or uffd context. */
3303 static inline struct vm_area_struct
3304 *vma_modify_flags_uffd(struct vma_iterator
*vmi
,
3305 struct vm_area_struct
*prev
,
3306 struct vm_area_struct
*vma
,
3307 unsigned long start
, unsigned long end
,
3308 unsigned long new_flags
,
3309 struct vm_userfaultfd_ctx new_ctx
)
3311 return vma_modify(vmi
, prev
, vma
, start
, end
, new_flags
,
3312 vma_policy(vma
), new_ctx
, anon_vma_name(vma
));
3315 static inline int check_data_rlimit(unsigned long rlim
,
3317 unsigned long start
,
3318 unsigned long end_data
,
3319 unsigned long start_data
)
3321 if (rlim
< RLIM_INFINITY
) {
3322 if (((new - start
) + (end_data
- start_data
)) > rlim
)
3329 extern int mm_take_all_locks(struct mm_struct
*mm
);
3330 extern void mm_drop_all_locks(struct mm_struct
*mm
);
3332 extern int set_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
);
3333 extern int replace_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
);
3334 extern struct file
*get_mm_exe_file(struct mm_struct
*mm
);
3335 extern struct file
*get_task_exe_file(struct task_struct
*task
);
3337 extern bool may_expand_vm(struct mm_struct
*, vm_flags_t
, unsigned long npages
);
3338 extern void vm_stat_account(struct mm_struct
*, vm_flags_t
, long npages
);
3340 extern bool vma_is_special_mapping(const struct vm_area_struct
*vma
,
3341 const struct vm_special_mapping
*sm
);
3342 extern struct vm_area_struct
*_install_special_mapping(struct mm_struct
*mm
,
3343 unsigned long addr
, unsigned long len
,
3344 unsigned long flags
,
3345 const struct vm_special_mapping
*spec
);
3346 /* This is an obsolete alternative to _install_special_mapping. */
3347 extern int install_special_mapping(struct mm_struct
*mm
,
3348 unsigned long addr
, unsigned long len
,
3349 unsigned long flags
, struct page
**pages
);
3351 unsigned long randomize_stack_top(unsigned long stack_top
);
3352 unsigned long randomize_page(unsigned long start
, unsigned long range
);
3354 extern unsigned long get_unmapped_area(struct file
*, unsigned long, unsigned long, unsigned long, unsigned long);
3356 extern unsigned long mmap_region(struct file
*file
, unsigned long addr
,
3357 unsigned long len
, vm_flags_t vm_flags
, unsigned long pgoff
,
3358 struct list_head
*uf
);
3359 extern unsigned long do_mmap(struct file
*file
, unsigned long addr
,
3360 unsigned long len
, unsigned long prot
, unsigned long flags
,
3361 vm_flags_t vm_flags
, unsigned long pgoff
, unsigned long *populate
,
3362 struct list_head
*uf
);
3363 extern int do_vmi_munmap(struct vma_iterator
*vmi
, struct mm_struct
*mm
,
3364 unsigned long start
, size_t len
, struct list_head
*uf
,
3366 extern int do_munmap(struct mm_struct
*, unsigned long, size_t,
3367 struct list_head
*uf
);
3368 extern int do_madvise(struct mm_struct
*mm
, unsigned long start
, size_t len_in
, int behavior
);
3371 extern int do_vma_munmap(struct vma_iterator
*vmi
, struct vm_area_struct
*vma
,
3372 unsigned long start
, unsigned long end
,
3373 struct list_head
*uf
, bool unlock
);
3374 extern int __mm_populate(unsigned long addr
, unsigned long len
,
3376 static inline void mm_populate(unsigned long addr
, unsigned long len
)
3379 (void) __mm_populate(addr
, len
, 1);
3382 static inline void mm_populate(unsigned long addr
, unsigned long len
) {}
3385 /* This takes the mm semaphore itself */
3386 extern int __must_check
vm_brk_flags(unsigned long, unsigned long, unsigned long);
3387 extern int vm_munmap(unsigned long, size_t);
3388 extern unsigned long __must_check
vm_mmap(struct file
*, unsigned long,
3389 unsigned long, unsigned long,
3390 unsigned long, unsigned long);
3392 struct vm_unmapped_area_info
{
3393 #define VM_UNMAPPED_AREA_TOPDOWN 1
3394 unsigned long flags
;
3395 unsigned long length
;
3396 unsigned long low_limit
;
3397 unsigned long high_limit
;
3398 unsigned long align_mask
;
3399 unsigned long align_offset
;
3402 extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info
*info
);
3405 extern void truncate_inode_pages(struct address_space
*, loff_t
);
3406 extern void truncate_inode_pages_range(struct address_space
*,
3407 loff_t lstart
, loff_t lend
);
3408 extern void truncate_inode_pages_final(struct address_space
*);
3410 /* generic vm_area_ops exported for stackable file systems */
3411 extern vm_fault_t
filemap_fault(struct vm_fault
*vmf
);
3412 extern vm_fault_t
filemap_map_pages(struct vm_fault
*vmf
,
3413 pgoff_t start_pgoff
, pgoff_t end_pgoff
);
3414 extern vm_fault_t
filemap_page_mkwrite(struct vm_fault
*vmf
);
3416 extern unsigned long stack_guard_gap
;
3417 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
3418 int expand_stack_locked(struct vm_area_struct
*vma
, unsigned long address
);
3419 struct vm_area_struct
*expand_stack(struct mm_struct
* mm
, unsigned long addr
);
3421 /* CONFIG_STACK_GROWSUP still needs to grow downwards at some places */
3422 int expand_downwards(struct vm_area_struct
*vma
, unsigned long address
);
3424 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
3425 extern struct vm_area_struct
* find_vma(struct mm_struct
* mm
, unsigned long addr
);
3426 extern struct vm_area_struct
* find_vma_prev(struct mm_struct
* mm
, unsigned long addr
,
3427 struct vm_area_struct
**pprev
);
3430 * Look up the first VMA which intersects the interval [start_addr, end_addr)
3431 * NULL if none. Assume start_addr < end_addr.
3433 struct vm_area_struct
*find_vma_intersection(struct mm_struct
*mm
,
3434 unsigned long start_addr
, unsigned long end_addr
);
3437 * vma_lookup() - Find a VMA at a specific address
3438 * @mm: The process address space.
3439 * @addr: The user address.
3441 * Return: The vm_area_struct at the given address, %NULL otherwise.
3444 struct vm_area_struct
*vma_lookup(struct mm_struct
*mm
, unsigned long addr
)
3446 return mtree_load(&mm
->mm_mt
, addr
);
3449 static inline unsigned long stack_guard_start_gap(struct vm_area_struct
*vma
)
3451 if (vma
->vm_flags
& VM_GROWSDOWN
)
3452 return stack_guard_gap
;
3454 /* See reasoning around the VM_SHADOW_STACK definition */
3455 if (vma
->vm_flags
& VM_SHADOW_STACK
)
3461 static inline unsigned long vm_start_gap(struct vm_area_struct
*vma
)
3463 unsigned long gap
= stack_guard_start_gap(vma
);
3464 unsigned long vm_start
= vma
->vm_start
;
3467 if (vm_start
> vma
->vm_start
)
3472 static inline unsigned long vm_end_gap(struct vm_area_struct
*vma
)
3474 unsigned long vm_end
= vma
->vm_end
;
3476 if (vma
->vm_flags
& VM_GROWSUP
) {
3477 vm_end
+= stack_guard_gap
;
3478 if (vm_end
< vma
->vm_end
)
3479 vm_end
= -PAGE_SIZE
;
3484 static inline unsigned long vma_pages(struct vm_area_struct
*vma
)
3486 return (vma
->vm_end
- vma
->vm_start
) >> PAGE_SHIFT
;
3489 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
3490 static inline struct vm_area_struct
*find_exact_vma(struct mm_struct
*mm
,
3491 unsigned long vm_start
, unsigned long vm_end
)
3493 struct vm_area_struct
*vma
= vma_lookup(mm
, vm_start
);
3495 if (vma
&& (vma
->vm_start
!= vm_start
|| vma
->vm_end
!= vm_end
))
3501 static inline bool range_in_vma(struct vm_area_struct
*vma
,
3502 unsigned long start
, unsigned long end
)
3504 return (vma
&& vma
->vm_start
<= start
&& end
<= vma
->vm_end
);
3508 pgprot_t
vm_get_page_prot(unsigned long vm_flags
);
3509 void vma_set_page_prot(struct vm_area_struct
*vma
);
3511 static inline pgprot_t
vm_get_page_prot(unsigned long vm_flags
)
3515 static inline void vma_set_page_prot(struct vm_area_struct
*vma
)
3517 vma
->vm_page_prot
= vm_get_page_prot(vma
->vm_flags
);
3521 void vma_set_file(struct vm_area_struct
*vma
, struct file
*file
);
3523 #ifdef CONFIG_NUMA_BALANCING
3524 unsigned long change_prot_numa(struct vm_area_struct
*vma
,
3525 unsigned long start
, unsigned long end
);
3528 struct vm_area_struct
*find_extend_vma_locked(struct mm_struct
*,
3529 unsigned long addr
);
3530 int remap_pfn_range(struct vm_area_struct
*, unsigned long addr
,
3531 unsigned long pfn
, unsigned long size
, pgprot_t
);
3532 int remap_pfn_range_notrack(struct vm_area_struct
*vma
, unsigned long addr
,
3533 unsigned long pfn
, unsigned long size
, pgprot_t prot
);
3534 int vm_insert_page(struct vm_area_struct
*, unsigned long addr
, struct page
*);
3535 int vm_insert_pages(struct vm_area_struct
*vma
, unsigned long addr
,
3536 struct page
**pages
, unsigned long *num
);
3537 int vm_map_pages(struct vm_area_struct
*vma
, struct page
**pages
,
3539 int vm_map_pages_zero(struct vm_area_struct
*vma
, struct page
**pages
,
3541 vm_fault_t
vmf_insert_pfn(struct vm_area_struct
*vma
, unsigned long addr
,
3543 vm_fault_t
vmf_insert_pfn_prot(struct vm_area_struct
*vma
, unsigned long addr
,
3544 unsigned long pfn
, pgprot_t pgprot
);
3545 vm_fault_t
vmf_insert_mixed(struct vm_area_struct
*vma
, unsigned long addr
,
3547 vm_fault_t
vmf_insert_mixed_mkwrite(struct vm_area_struct
*vma
,
3548 unsigned long addr
, pfn_t pfn
);
3549 int vm_iomap_memory(struct vm_area_struct
*vma
, phys_addr_t start
, unsigned long len
);
3551 static inline vm_fault_t
vmf_insert_page(struct vm_area_struct
*vma
,
3552 unsigned long addr
, struct page
*page
)
3554 int err
= vm_insert_page(vma
, addr
, page
);
3557 return VM_FAULT_OOM
;
3558 if (err
< 0 && err
!= -EBUSY
)
3559 return VM_FAULT_SIGBUS
;
3561 return VM_FAULT_NOPAGE
;
3564 #ifndef io_remap_pfn_range
3565 static inline int io_remap_pfn_range(struct vm_area_struct
*vma
,
3566 unsigned long addr
, unsigned long pfn
,
3567 unsigned long size
, pgprot_t prot
)
3569 return remap_pfn_range(vma
, addr
, pfn
, size
, pgprot_decrypted(prot
));
3573 static inline vm_fault_t
vmf_error(int err
)
3576 return VM_FAULT_OOM
;
3577 else if (err
== -EHWPOISON
)
3578 return VM_FAULT_HWPOISON
;
3579 return VM_FAULT_SIGBUS
;
3583 * Convert errno to return value for ->page_mkwrite() calls.
3585 * This should eventually be merged with vmf_error() above, but will need a
3586 * careful audit of all vmf_error() callers.
3588 static inline vm_fault_t
vmf_fs_error(int err
)
3591 return VM_FAULT_LOCKED
;
3592 if (err
== -EFAULT
|| err
== -EAGAIN
)
3593 return VM_FAULT_NOPAGE
;
3595 return VM_FAULT_OOM
;
3596 /* -ENOSPC, -EDQUOT, -EIO ... */
3597 return VM_FAULT_SIGBUS
;
3600 struct page
*follow_page(struct vm_area_struct
*vma
, unsigned long address
,
3601 unsigned int foll_flags
);
3603 static inline int vm_fault_to_errno(vm_fault_t vm_fault
, int foll_flags
)
3605 if (vm_fault
& VM_FAULT_OOM
)
3607 if (vm_fault
& (VM_FAULT_HWPOISON
| VM_FAULT_HWPOISON_LARGE
))
3608 return (foll_flags
& FOLL_HWPOISON
) ? -EHWPOISON
: -EFAULT
;
3609 if (vm_fault
& (VM_FAULT_SIGBUS
| VM_FAULT_SIGSEGV
))
3615 * Indicates whether GUP can follow a PROT_NONE mapped page, or whether
3616 * a (NUMA hinting) fault is required.
3618 static inline bool gup_can_follow_protnone(struct vm_area_struct
*vma
,
3622 * If callers don't want to honor NUMA hinting faults, no need to
3623 * determine if we would actually have to trigger a NUMA hinting fault.
3625 if (!(flags
& FOLL_HONOR_NUMA_FAULT
))
3629 * NUMA hinting faults don't apply in inaccessible (PROT_NONE) VMAs.
3631 * Requiring a fault here even for inaccessible VMAs would mean that
3632 * FOLL_FORCE cannot make any progress, because handle_mm_fault()
3633 * refuses to process NUMA hinting faults in inaccessible VMAs.
3635 return !vma_is_accessible(vma
);
3638 typedef int (*pte_fn_t
)(pte_t
*pte
, unsigned long addr
, void *data
);
3639 extern int apply_to_page_range(struct mm_struct
*mm
, unsigned long address
,
3640 unsigned long size
, pte_fn_t fn
, void *data
);
3641 extern int apply_to_existing_page_range(struct mm_struct
*mm
,
3642 unsigned long address
, unsigned long size
,
3643 pte_fn_t fn
, void *data
);
3645 #ifdef CONFIG_PAGE_POISONING
3646 extern void __kernel_poison_pages(struct page
*page
, int numpages
);
3647 extern void __kernel_unpoison_pages(struct page
*page
, int numpages
);
3648 extern bool _page_poisoning_enabled_early
;
3649 DECLARE_STATIC_KEY_FALSE(_page_poisoning_enabled
);
3650 static inline bool page_poisoning_enabled(void)
3652 return _page_poisoning_enabled_early
;
3655 * For use in fast paths after init_mem_debugging() has run, or when a
3656 * false negative result is not harmful when called too early.
3658 static inline bool page_poisoning_enabled_static(void)
3660 return static_branch_unlikely(&_page_poisoning_enabled
);
3662 static inline void kernel_poison_pages(struct page
*page
, int numpages
)
3664 if (page_poisoning_enabled_static())
3665 __kernel_poison_pages(page
, numpages
);
3667 static inline void kernel_unpoison_pages(struct page
*page
, int numpages
)
3669 if (page_poisoning_enabled_static())
3670 __kernel_unpoison_pages(page
, numpages
);
3673 static inline bool page_poisoning_enabled(void) { return false; }
3674 static inline bool page_poisoning_enabled_static(void) { return false; }
3675 static inline void __kernel_poison_pages(struct page
*page
, int nunmpages
) { }
3676 static inline void kernel_poison_pages(struct page
*page
, int numpages
) { }
3677 static inline void kernel_unpoison_pages(struct page
*page
, int numpages
) { }
3680 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_ALLOC_DEFAULT_ON
, init_on_alloc
);
3681 static inline bool want_init_on_alloc(gfp_t flags
)
3683 if (static_branch_maybe(CONFIG_INIT_ON_ALLOC_DEFAULT_ON
,
3686 return flags
& __GFP_ZERO
;
3689 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_FREE_DEFAULT_ON
, init_on_free
);
3690 static inline bool want_init_on_free(void)
3692 return static_branch_maybe(CONFIG_INIT_ON_FREE_DEFAULT_ON
,
3696 extern bool _debug_pagealloc_enabled_early
;
3697 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled
);
3699 static inline bool debug_pagealloc_enabled(void)
3701 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC
) &&
3702 _debug_pagealloc_enabled_early
;
3706 * For use in fast paths after mem_debugging_and_hardening_init() has run,
3707 * or when a false negative result is not harmful when called too early.
3709 static inline bool debug_pagealloc_enabled_static(void)
3711 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC
))
3714 return static_branch_unlikely(&_debug_pagealloc_enabled
);
3718 * To support DEBUG_PAGEALLOC architecture must ensure that
3719 * __kernel_map_pages() never fails
3721 extern void __kernel_map_pages(struct page
*page
, int numpages
, int enable
);
3722 #ifdef CONFIG_DEBUG_PAGEALLOC
3723 static inline void debug_pagealloc_map_pages(struct page
*page
, int numpages
)
3725 if (debug_pagealloc_enabled_static())
3726 __kernel_map_pages(page
, numpages
, 1);
3729 static inline void debug_pagealloc_unmap_pages(struct page
*page
, int numpages
)
3731 if (debug_pagealloc_enabled_static())
3732 __kernel_map_pages(page
, numpages
, 0);
3735 extern unsigned int _debug_guardpage_minorder
;
3736 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled
);
3738 static inline unsigned int debug_guardpage_minorder(void)
3740 return _debug_guardpage_minorder
;
3743 static inline bool debug_guardpage_enabled(void)
3745 return static_branch_unlikely(&_debug_guardpage_enabled
);
3748 static inline bool page_is_guard(struct page
*page
)
3750 if (!debug_guardpage_enabled())
3753 return PageGuard(page
);
3756 bool __set_page_guard(struct zone
*zone
, struct page
*page
, unsigned int order
,
3758 static inline bool set_page_guard(struct zone
*zone
, struct page
*page
,
3759 unsigned int order
, int migratetype
)
3761 if (!debug_guardpage_enabled())
3763 return __set_page_guard(zone
, page
, order
, migratetype
);
3766 void __clear_page_guard(struct zone
*zone
, struct page
*page
, unsigned int order
,
3768 static inline void clear_page_guard(struct zone
*zone
, struct page
*page
,
3769 unsigned int order
, int migratetype
)
3771 if (!debug_guardpage_enabled())
3773 __clear_page_guard(zone
, page
, order
, migratetype
);
3776 #else /* CONFIG_DEBUG_PAGEALLOC */
3777 static inline void debug_pagealloc_map_pages(struct page
*page
, int numpages
) {}
3778 static inline void debug_pagealloc_unmap_pages(struct page
*page
, int numpages
) {}
3779 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
3780 static inline bool debug_guardpage_enabled(void) { return false; }
3781 static inline bool page_is_guard(struct page
*page
) { return false; }
3782 static inline bool set_page_guard(struct zone
*zone
, struct page
*page
,
3783 unsigned int order
, int migratetype
) { return false; }
3784 static inline void clear_page_guard(struct zone
*zone
, struct page
*page
,
3785 unsigned int order
, int migratetype
) {}
3786 #endif /* CONFIG_DEBUG_PAGEALLOC */
3788 #ifdef __HAVE_ARCH_GATE_AREA
3789 extern struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
);
3790 extern int in_gate_area_no_mm(unsigned long addr
);
3791 extern int in_gate_area(struct mm_struct
*mm
, unsigned long addr
);
3793 static inline struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
)
3797 static inline int in_gate_area_no_mm(unsigned long addr
) { return 0; }
3798 static inline int in_gate_area(struct mm_struct
*mm
, unsigned long addr
)
3802 #endif /* __HAVE_ARCH_GATE_AREA */
3804 extern bool process_shares_mm(struct task_struct
*p
, struct mm_struct
*mm
);
3806 #ifdef CONFIG_SYSCTL
3807 extern int sysctl_drop_caches
;
3808 int drop_caches_sysctl_handler(struct ctl_table
*, int, void *, size_t *,
3812 void drop_slab(void);
3815 #define randomize_va_space 0
3817 extern int randomize_va_space
;
3820 const char * arch_vma_name(struct vm_area_struct
*vma
);
3822 void print_vma_addr(char *prefix
, unsigned long rip
);
3824 static inline void print_vma_addr(char *prefix
, unsigned long rip
)
3829 void *sparse_buffer_alloc(unsigned long size
);
3830 struct page
* __populate_section_memmap(unsigned long pfn
,
3831 unsigned long nr_pages
, int nid
, struct vmem_altmap
*altmap
,
3832 struct dev_pagemap
*pgmap
);
3833 void pmd_init(void *addr
);
3834 void pud_init(void *addr
);
3835 pgd_t
*vmemmap_pgd_populate(unsigned long addr
, int node
);
3836 p4d_t
*vmemmap_p4d_populate(pgd_t
*pgd
, unsigned long addr
, int node
);
3837 pud_t
*vmemmap_pud_populate(p4d_t
*p4d
, unsigned long addr
, int node
);
3838 pmd_t
*vmemmap_pmd_populate(pud_t
*pud
, unsigned long addr
, int node
);
3839 pte_t
*vmemmap_pte_populate(pmd_t
*pmd
, unsigned long addr
, int node
,
3840 struct vmem_altmap
*altmap
, struct page
*reuse
);
3841 void *vmemmap_alloc_block(unsigned long size
, int node
);
3843 void *vmemmap_alloc_block_buf(unsigned long size
, int node
,
3844 struct vmem_altmap
*altmap
);
3845 void vmemmap_verify(pte_t
*, int, unsigned long, unsigned long);
3846 void vmemmap_set_pmd(pmd_t
*pmd
, void *p
, int node
,
3847 unsigned long addr
, unsigned long next
);
3848 int vmemmap_check_pmd(pmd_t
*pmd
, int node
,
3849 unsigned long addr
, unsigned long next
);
3850 int vmemmap_populate_basepages(unsigned long start
, unsigned long end
,
3851 int node
, struct vmem_altmap
*altmap
);
3852 int vmemmap_populate_hugepages(unsigned long start
, unsigned long end
,
3853 int node
, struct vmem_altmap
*altmap
);
3854 int vmemmap_populate(unsigned long start
, unsigned long end
, int node
,
3855 struct vmem_altmap
*altmap
);
3856 void vmemmap_populate_print_last(void);
3857 #ifdef CONFIG_MEMORY_HOTPLUG
3858 void vmemmap_free(unsigned long start
, unsigned long end
,
3859 struct vmem_altmap
*altmap
);
3862 #define VMEMMAP_RESERVE_NR 2
3863 #ifdef CONFIG_ARCH_WANT_OPTIMIZE_DAX_VMEMMAP
3864 static inline bool __vmemmap_can_optimize(struct vmem_altmap
*altmap
,
3865 struct dev_pagemap
*pgmap
)
3867 unsigned long nr_pages
;
3868 unsigned long nr_vmemmap_pages
;
3870 if (!pgmap
|| !is_power_of_2(sizeof(struct page
)))
3873 nr_pages
= pgmap_vmemmap_nr(pgmap
);
3874 nr_vmemmap_pages
= ((nr_pages
* sizeof(struct page
)) >> PAGE_SHIFT
);
3876 * For vmemmap optimization with DAX we need minimum 2 vmemmap
3877 * pages. See layout diagram in Documentation/mm/vmemmap_dedup.rst
3879 return !altmap
&& (nr_vmemmap_pages
> VMEMMAP_RESERVE_NR
);
3882 * If we don't have an architecture override, use the generic rule
3884 #ifndef vmemmap_can_optimize
3885 #define vmemmap_can_optimize __vmemmap_can_optimize
3889 static inline bool vmemmap_can_optimize(struct vmem_altmap
*altmap
,
3890 struct dev_pagemap
*pgmap
)
3896 void register_page_bootmem_memmap(unsigned long section_nr
, struct page
*map
,
3897 unsigned long nr_pages
);
3900 MF_COUNT_INCREASED
= 1 << 0,
3901 MF_ACTION_REQUIRED
= 1 << 1,
3902 MF_MUST_KILL
= 1 << 2,
3903 MF_SOFT_OFFLINE
= 1 << 3,
3904 MF_UNPOISON
= 1 << 4,
3905 MF_SW_SIMULATED
= 1 << 5,
3906 MF_NO_RETRY
= 1 << 6,
3908 int mf_dax_kill_procs(struct address_space
*mapping
, pgoff_t index
,
3909 unsigned long count
, int mf_flags
);
3910 extern int memory_failure(unsigned long pfn
, int flags
);
3911 extern void memory_failure_queue_kick(int cpu
);
3912 extern int unpoison_memory(unsigned long pfn
);
3913 extern void shake_page(struct page
*p
);
3914 extern atomic_long_t num_poisoned_pages __read_mostly
;
3915 extern int soft_offline_page(unsigned long pfn
, int flags
);
3916 #ifdef CONFIG_MEMORY_FAILURE
3918 * Sysfs entries for memory failure handling statistics.
3920 extern const struct attribute_group memory_failure_attr_group
;
3921 extern void memory_failure_queue(unsigned long pfn
, int flags
);
3922 extern int __get_huge_page_for_hwpoison(unsigned long pfn
, int flags
,
3923 bool *migratable_cleared
);
3924 void num_poisoned_pages_inc(unsigned long pfn
);
3925 void num_poisoned_pages_sub(unsigned long pfn
, long i
);
3926 struct task_struct
*task_early_kill(struct task_struct
*tsk
, int force_early
);
3928 static inline void memory_failure_queue(unsigned long pfn
, int flags
)
3932 static inline int __get_huge_page_for_hwpoison(unsigned long pfn
, int flags
,
3933 bool *migratable_cleared
)
3938 static inline void num_poisoned_pages_inc(unsigned long pfn
)
3942 static inline void num_poisoned_pages_sub(unsigned long pfn
, long i
)
3947 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_KSM)
3948 void add_to_kill_ksm(struct task_struct
*tsk
, struct page
*p
,
3949 struct vm_area_struct
*vma
, struct list_head
*to_kill
,
3950 unsigned long ksm_addr
);
3953 #if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_MEMORY_HOTPLUG)
3954 extern void memblk_nr_poison_inc(unsigned long pfn
);
3955 extern void memblk_nr_poison_sub(unsigned long pfn
, long i
);
3957 static inline void memblk_nr_poison_inc(unsigned long pfn
)
3961 static inline void memblk_nr_poison_sub(unsigned long pfn
, long i
)
3966 #ifndef arch_memory_failure
3967 static inline int arch_memory_failure(unsigned long pfn
, int flags
)
3973 #ifndef arch_is_platform_page
3974 static inline bool arch_is_platform_page(u64 paddr
)
3981 * Error handlers for various types of pages.
3984 MF_IGNORED
, /* Error: cannot be handled */
3985 MF_FAILED
, /* Error: handling failed */
3986 MF_DELAYED
, /* Will be handled later */
3987 MF_RECOVERED
, /* Successfully recovered */
3990 enum mf_action_page_type
{
3992 MF_MSG_KERNEL_HIGH_ORDER
,
3994 MF_MSG_DIFFERENT_COMPOUND
,
3997 MF_MSG_UNMAP_FAILED
,
3998 MF_MSG_DIRTY_SWAPCACHE
,
3999 MF_MSG_CLEAN_SWAPCACHE
,
4000 MF_MSG_DIRTY_MLOCKED_LRU
,
4001 MF_MSG_CLEAN_MLOCKED_LRU
,
4002 MF_MSG_DIRTY_UNEVICTABLE_LRU
,
4003 MF_MSG_CLEAN_UNEVICTABLE_LRU
,
4006 MF_MSG_TRUNCATED_LRU
,
4013 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
4014 extern void clear_huge_page(struct page
*page
,
4015 unsigned long addr_hint
,
4016 unsigned int pages_per_huge_page
);
4017 int copy_user_large_folio(struct folio
*dst
, struct folio
*src
,
4018 unsigned long addr_hint
,
4019 struct vm_area_struct
*vma
);
4020 long copy_folio_from_user(struct folio
*dst_folio
,
4021 const void __user
*usr_src
,
4022 bool allow_pagefault
);
4025 * vma_is_special_huge - Are transhuge page-table entries considered special?
4026 * @vma: Pointer to the struct vm_area_struct to consider
4028 * Whether transhuge page-table entries are considered "special" following
4029 * the definition in vm_normal_page().
4031 * Return: true if transhuge page-table entries should be considered special,
4034 static inline bool vma_is_special_huge(const struct vm_area_struct
*vma
)
4036 return vma_is_dax(vma
) || (vma
->vm_file
&&
4037 (vma
->vm_flags
& (VM_PFNMAP
| VM_MIXEDMAP
)));
4040 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
4042 #if MAX_NUMNODES > 1
4043 void __init
setup_nr_node_ids(void);
4045 static inline void setup_nr_node_ids(void) {}
4048 extern int memcmp_pages(struct page
*page1
, struct page
*page2
);
4050 static inline int pages_identical(struct page
*page1
, struct page
*page2
)
4052 return !memcmp_pages(page1
, page2
);
4055 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
4056 unsigned long clean_record_shared_mapping_range(struct address_space
*mapping
,
4057 pgoff_t first_index
, pgoff_t nr
,
4058 pgoff_t bitmap_pgoff
,
4059 unsigned long *bitmap
,
4063 unsigned long wp_shared_mapping_range(struct address_space
*mapping
,
4064 pgoff_t first_index
, pgoff_t nr
);
4067 extern int sysctl_nr_trim_pages
;
4069 #ifdef CONFIG_PRINTK
4070 void mem_dump_obj(void *object
);
4072 static inline void mem_dump_obj(void *object
) {}
4076 * seal_check_write - Check for F_SEAL_WRITE or F_SEAL_FUTURE_WRITE flags and
4078 * @seals: the seals to check
4079 * @vma: the vma to operate on
4081 * Check whether F_SEAL_WRITE or F_SEAL_FUTURE_WRITE are set; if so, do proper
4082 * check/handling on the vma flags. Return 0 if check pass, or <0 for errors.
4084 static inline int seal_check_write(int seals
, struct vm_area_struct
*vma
)
4086 if (seals
& (F_SEAL_WRITE
| F_SEAL_FUTURE_WRITE
)) {
4088 * New PROT_WRITE and MAP_SHARED mmaps are not allowed when
4089 * write seals are active.
4091 if ((vma
->vm_flags
& VM_SHARED
) && (vma
->vm_flags
& VM_WRITE
))
4095 * Since an F_SEAL_[FUTURE_]WRITE sealed memfd can be mapped as
4096 * MAP_SHARED and read-only, take care to not allow mprotect to
4097 * revert protections on such mappings. Do this only for shared
4098 * mappings. For private mappings, don't need to mask
4099 * VM_MAYWRITE as we still want them to be COW-writable.
4101 if (vma
->vm_flags
& VM_SHARED
)
4102 vm_flags_clear(vma
, VM_MAYWRITE
);
4108 #ifdef CONFIG_ANON_VMA_NAME
4109 int madvise_set_anon_name(struct mm_struct
*mm
, unsigned long start
,
4110 unsigned long len_in
,
4111 struct anon_vma_name
*anon_name
);
4114 madvise_set_anon_name(struct mm_struct
*mm
, unsigned long start
,
4115 unsigned long len_in
, struct anon_vma_name
*anon_name
) {
4120 #ifdef CONFIG_UNACCEPTED_MEMORY
4122 bool range_contains_unaccepted_memory(phys_addr_t start
, phys_addr_t end
);
4123 void accept_memory(phys_addr_t start
, phys_addr_t end
);
4127 static inline bool range_contains_unaccepted_memory(phys_addr_t start
,
4133 static inline void accept_memory(phys_addr_t start
, phys_addr_t end
)
4139 static inline bool pfn_is_unaccepted_memory(unsigned long pfn
)
4141 phys_addr_t paddr
= pfn
<< PAGE_SHIFT
;
4143 return range_contains_unaccepted_memory(paddr
, paddr
+ PAGE_SIZE
);
4146 #endif /* _LINUX_MM_H */