1 /* SPDX-License-Identifier: GPL-2.0 */
5 #include <linux/errno.h>
9 #include <linux/mmdebug.h>
10 #include <linux/gfp.h>
11 #include <linux/bug.h>
12 #include <linux/list.h>
13 #include <linux/mmzone.h>
14 #include <linux/rbtree.h>
15 #include <linux/atomic.h>
16 #include <linux/debug_locks.h>
17 #include <linux/mm_types.h>
18 #include <linux/mmap_lock.h>
19 #include <linux/range.h>
20 #include <linux/pfn.h>
21 #include <linux/percpu-refcount.h>
22 #include <linux/bit_spinlock.h>
23 #include <linux/shrinker.h>
24 #include <linux/resource.h>
25 #include <linux/page_ext.h>
26 #include <linux/err.h>
27 #include <linux/page-flags.h>
28 #include <linux/page_ref.h>
29 #include <linux/memremap.h>
30 #include <linux/overflow.h>
31 #include <linux/sizes.h>
32 #include <linux/sched.h>
33 #include <linux/pgtable.h>
34 #include <linux/kasan.h>
38 struct anon_vma_chain
;
41 struct writeback_control
;
45 extern int sysctl_page_lock_unfairness
;
47 void init_mm_internals(void);
49 #ifndef CONFIG_NUMA /* Don't use mapnrs, do it properly */
50 extern unsigned long max_mapnr
;
52 static inline void set_max_mapnr(unsigned long limit
)
57 static inline void set_max_mapnr(unsigned long limit
) { }
60 extern atomic_long_t _totalram_pages
;
61 static inline unsigned long totalram_pages(void)
63 return (unsigned long)atomic_long_read(&_totalram_pages
);
66 static inline void totalram_pages_inc(void)
68 atomic_long_inc(&_totalram_pages
);
71 static inline void totalram_pages_dec(void)
73 atomic_long_dec(&_totalram_pages
);
76 static inline void totalram_pages_add(long count
)
78 atomic_long_add(count
, &_totalram_pages
);
81 extern void * high_memory
;
82 extern int page_cluster
;
85 extern int sysctl_legacy_va_layout
;
87 #define sysctl_legacy_va_layout 0
90 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
91 extern const int mmap_rnd_bits_min
;
92 extern const int mmap_rnd_bits_max
;
93 extern int mmap_rnd_bits __read_mostly
;
95 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
96 extern const int mmap_rnd_compat_bits_min
;
97 extern const int mmap_rnd_compat_bits_max
;
98 extern int mmap_rnd_compat_bits __read_mostly
;
101 #include <asm/page.h>
102 #include <asm/processor.h>
105 * Architectures that support memory tagging (assigning tags to memory regions,
106 * embedding these tags into addresses that point to these memory regions, and
107 * checking that the memory and the pointer tags match on memory accesses)
108 * redefine this macro to strip tags from pointers.
109 * It's defined as noop for architectures that don't support memory tagging.
111 #ifndef untagged_addr
112 #define untagged_addr(addr) (addr)
116 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
120 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
124 #define lm_alias(x) __va(__pa_symbol(x))
128 * To prevent common memory management code establishing
129 * a zero page mapping on a read fault.
130 * This macro should be defined within <asm/pgtable.h>.
131 * s390 does this to prevent multiplexing of hardware bits
132 * related to the physical page in case of virtualization.
134 #ifndef mm_forbids_zeropage
135 #define mm_forbids_zeropage(X) (0)
139 * On some architectures it is expensive to call memset() for small sizes.
140 * If an architecture decides to implement their own version of
141 * mm_zero_struct_page they should wrap the defines below in a #ifndef and
142 * define their own version of this macro in <asm/pgtable.h>
144 #if BITS_PER_LONG == 64
145 /* This function must be updated when the size of struct page grows above 80
146 * or reduces below 56. The idea that compiler optimizes out switch()
147 * statement, and only leaves move/store instructions. Also the compiler can
148 * combine write statements if they are both assignments and can be reordered,
149 * this can result in several of the writes here being dropped.
151 #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
152 static inline void __mm_zero_struct_page(struct page
*page
)
154 unsigned long *_pp
= (void *)page
;
156 /* Check that struct page is either 56, 64, 72, or 80 bytes */
157 BUILD_BUG_ON(sizeof(struct page
) & 7);
158 BUILD_BUG_ON(sizeof(struct page
) < 56);
159 BUILD_BUG_ON(sizeof(struct page
) > 80);
161 switch (sizeof(struct page
)) {
182 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
186 * Default maximum number of active map areas, this limits the number of vmas
187 * per mm struct. Users can overwrite this number by sysctl but there is a
190 * When a program's coredump is generated as ELF format, a section is created
191 * per a vma. In ELF, the number of sections is represented in unsigned short.
192 * This means the number of sections should be smaller than 65535 at coredump.
193 * Because the kernel adds some informative sections to a image of program at
194 * generating coredump, we need some margin. The number of extra sections is
195 * 1-3 now and depends on arch. We use "5" as safe margin, here.
197 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
198 * not a hard limit any more. Although some userspace tools can be surprised by
201 #define MAPCOUNT_ELF_CORE_MARGIN (5)
202 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
204 extern int sysctl_max_map_count
;
206 extern unsigned long sysctl_user_reserve_kbytes
;
207 extern unsigned long sysctl_admin_reserve_kbytes
;
209 extern int sysctl_overcommit_memory
;
210 extern int sysctl_overcommit_ratio
;
211 extern unsigned long sysctl_overcommit_kbytes
;
213 int overcommit_ratio_handler(struct ctl_table
*, int, void *, size_t *,
215 int overcommit_kbytes_handler(struct ctl_table
*, int, void *, size_t *,
217 int overcommit_policy_handler(struct ctl_table
*, int, void *, size_t *,
220 * Any attempt to mark this function as static leads to build failure
221 * when CONFIG_DEBUG_INFO_BTF is enabled because __add_to_page_cache_locked()
222 * is referred to by BPF code. This must be visible for error injection.
224 int __add_to_page_cache_locked(struct page
*page
, struct address_space
*mapping
,
225 pgoff_t index
, gfp_t gfp
, void **shadowp
);
227 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
228 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
230 #define nth_page(page,n) ((page) + (n))
233 /* to align the pointer to the (next) page boundary */
234 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
236 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
237 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
239 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
241 void setup_initial_init_mm(void *start_code
, void *end_code
,
242 void *end_data
, void *brk
);
245 * Linux kernel virtual memory manager primitives.
246 * The idea being to have a "virtual" mm in the same way
247 * we have a virtual fs - giving a cleaner interface to the
248 * mm details, and allowing different kinds of memory mappings
249 * (from shared memory to executable loading to arbitrary
253 struct vm_area_struct
*vm_area_alloc(struct mm_struct
*);
254 struct vm_area_struct
*vm_area_dup(struct vm_area_struct
*);
255 void vm_area_free(struct vm_area_struct
*);
258 extern struct rb_root nommu_region_tree
;
259 extern struct rw_semaphore nommu_region_sem
;
261 extern unsigned int kobjsize(const void *objp
);
265 * vm_flags in vm_area_struct, see mm_types.h.
266 * When changing, update also include/trace/events/mmflags.h
268 #define VM_NONE 0x00000000
270 #define VM_READ 0x00000001 /* currently active flags */
271 #define VM_WRITE 0x00000002
272 #define VM_EXEC 0x00000004
273 #define VM_SHARED 0x00000008
275 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
276 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
277 #define VM_MAYWRITE 0x00000020
278 #define VM_MAYEXEC 0x00000040
279 #define VM_MAYSHARE 0x00000080
281 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
282 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
283 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
284 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
286 #define VM_LOCKED 0x00002000
287 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
289 /* Used by sys_madvise() */
290 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
291 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
293 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
294 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
295 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
296 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
297 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
298 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
299 #define VM_SYNC 0x00800000 /* Synchronous page faults */
300 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
301 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
302 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
304 #ifdef CONFIG_MEM_SOFT_DIRTY
305 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
307 # define VM_SOFTDIRTY 0
310 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
311 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
312 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
313 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
315 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
316 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
317 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
318 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
319 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
320 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
321 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
322 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
323 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
324 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
325 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
326 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
328 #ifdef CONFIG_ARCH_HAS_PKEYS
329 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
330 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
331 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
332 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
333 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
335 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
337 # define VM_PKEY_BIT4 0
339 #endif /* CONFIG_ARCH_HAS_PKEYS */
341 #if defined(CONFIG_X86)
342 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
343 #elif defined(CONFIG_PPC)
344 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
345 #elif defined(CONFIG_PARISC)
346 # define VM_GROWSUP VM_ARCH_1
347 #elif defined(CONFIG_IA64)
348 # define VM_GROWSUP VM_ARCH_1
349 #elif defined(CONFIG_SPARC64)
350 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
351 # define VM_ARCH_CLEAR VM_SPARC_ADI
352 #elif defined(CONFIG_ARM64)
353 # define VM_ARM64_BTI VM_ARCH_1 /* BTI guarded page, a.k.a. GP bit */
354 # define VM_ARCH_CLEAR VM_ARM64_BTI
355 #elif !defined(CONFIG_MMU)
356 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
359 #if defined(CONFIG_ARM64_MTE)
360 # define VM_MTE VM_HIGH_ARCH_0 /* Use Tagged memory for access control */
361 # define VM_MTE_ALLOWED VM_HIGH_ARCH_1 /* Tagged memory permitted */
363 # define VM_MTE VM_NONE
364 # define VM_MTE_ALLOWED VM_NONE
368 # define VM_GROWSUP VM_NONE
371 #ifdef CONFIG_HAVE_ARCH_USERFAULTFD_MINOR
372 # define VM_UFFD_MINOR_BIT 37
373 # define VM_UFFD_MINOR BIT(VM_UFFD_MINOR_BIT) /* UFFD minor faults */
374 #else /* !CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
375 # define VM_UFFD_MINOR VM_NONE
376 #endif /* CONFIG_HAVE_ARCH_USERFAULTFD_MINOR */
378 /* Bits set in the VMA until the stack is in its final location */
379 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
381 #define TASK_EXEC ((current->personality & READ_IMPLIES_EXEC) ? VM_EXEC : 0)
383 /* Common data flag combinations */
384 #define VM_DATA_FLAGS_TSK_EXEC (VM_READ | VM_WRITE | TASK_EXEC | \
385 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
386 #define VM_DATA_FLAGS_NON_EXEC (VM_READ | VM_WRITE | VM_MAYREAD | \
387 VM_MAYWRITE | VM_MAYEXEC)
388 #define VM_DATA_FLAGS_EXEC (VM_READ | VM_WRITE | VM_EXEC | \
389 VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC)
391 #ifndef VM_DATA_DEFAULT_FLAGS /* arch can override this */
392 #define VM_DATA_DEFAULT_FLAGS VM_DATA_FLAGS_EXEC
395 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
396 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
399 #ifdef CONFIG_STACK_GROWSUP
400 #define VM_STACK VM_GROWSUP
402 #define VM_STACK VM_GROWSDOWN
405 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
407 /* VMA basic access permission flags */
408 #define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
412 * Special vmas that are non-mergable, non-mlock()able.
414 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
416 /* This mask prevents VMA from being scanned with khugepaged */
417 #define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB)
419 /* This mask defines which mm->def_flags a process can inherit its parent */
420 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
422 /* This mask is used to clear all the VMA flags used by mlock */
423 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
425 /* Arch-specific flags to clear when updating VM flags on protection change */
426 #ifndef VM_ARCH_CLEAR
427 # define VM_ARCH_CLEAR VM_NONE
429 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
432 * mapping from the currently active vm_flags protection bits (the
433 * low four bits) to a page protection mask..
435 extern pgprot_t protection_map
[16];
438 * enum fault_flag - Fault flag definitions.
439 * @FAULT_FLAG_WRITE: Fault was a write fault.
440 * @FAULT_FLAG_MKWRITE: Fault was mkwrite of existing PTE.
441 * @FAULT_FLAG_ALLOW_RETRY: Allow to retry the fault if blocked.
442 * @FAULT_FLAG_RETRY_NOWAIT: Don't drop mmap_lock and wait when retrying.
443 * @FAULT_FLAG_KILLABLE: The fault task is in SIGKILL killable region.
444 * @FAULT_FLAG_TRIED: The fault has been tried once.
445 * @FAULT_FLAG_USER: The fault originated in userspace.
446 * @FAULT_FLAG_REMOTE: The fault is not for current task/mm.
447 * @FAULT_FLAG_INSTRUCTION: The fault was during an instruction fetch.
448 * @FAULT_FLAG_INTERRUPTIBLE: The fault can be interrupted by non-fatal signals.
450 * About @FAULT_FLAG_ALLOW_RETRY and @FAULT_FLAG_TRIED: we can specify
451 * whether we would allow page faults to retry by specifying these two
452 * fault flags correctly. Currently there can be three legal combinations:
454 * (a) ALLOW_RETRY and !TRIED: this means the page fault allows retry, and
455 * this is the first try
457 * (b) ALLOW_RETRY and TRIED: this means the page fault allows retry, and
458 * we've already tried at least once
460 * (c) !ALLOW_RETRY and !TRIED: this means the page fault does not allow retry
462 * The unlisted combination (!ALLOW_RETRY && TRIED) is illegal and should never
463 * be used. Note that page faults can be allowed to retry for multiple times,
464 * in which case we'll have an initial fault with flags (a) then later on
465 * continuous faults with flags (b). We should always try to detect pending
466 * signals before a retry to make sure the continuous page faults can still be
467 * interrupted if necessary.
470 FAULT_FLAG_WRITE
= 1 << 0,
471 FAULT_FLAG_MKWRITE
= 1 << 1,
472 FAULT_FLAG_ALLOW_RETRY
= 1 << 2,
473 FAULT_FLAG_RETRY_NOWAIT
= 1 << 3,
474 FAULT_FLAG_KILLABLE
= 1 << 4,
475 FAULT_FLAG_TRIED
= 1 << 5,
476 FAULT_FLAG_USER
= 1 << 6,
477 FAULT_FLAG_REMOTE
= 1 << 7,
478 FAULT_FLAG_INSTRUCTION
= 1 << 8,
479 FAULT_FLAG_INTERRUPTIBLE
= 1 << 9,
483 * The default fault flags that should be used by most of the
484 * arch-specific page fault handlers.
486 #define FAULT_FLAG_DEFAULT (FAULT_FLAG_ALLOW_RETRY | \
487 FAULT_FLAG_KILLABLE | \
488 FAULT_FLAG_INTERRUPTIBLE)
491 * fault_flag_allow_retry_first - check ALLOW_RETRY the first time
492 * @flags: Fault flags.
494 * This is mostly used for places where we want to try to avoid taking
495 * the mmap_lock for too long a time when waiting for another condition
496 * to change, in which case we can try to be polite to release the
497 * mmap_lock in the first round to avoid potential starvation of other
498 * processes that would also want the mmap_lock.
500 * Return: true if the page fault allows retry and this is the first
501 * attempt of the fault handling; false otherwise.
503 static inline bool fault_flag_allow_retry_first(enum fault_flag flags
)
505 return (flags
& FAULT_FLAG_ALLOW_RETRY
) &&
506 (!(flags
& FAULT_FLAG_TRIED
));
509 #define FAULT_FLAG_TRACE \
510 { FAULT_FLAG_WRITE, "WRITE" }, \
511 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
512 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
513 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
514 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
515 { FAULT_FLAG_TRIED, "TRIED" }, \
516 { FAULT_FLAG_USER, "USER" }, \
517 { FAULT_FLAG_REMOTE, "REMOTE" }, \
518 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }, \
519 { FAULT_FLAG_INTERRUPTIBLE, "INTERRUPTIBLE" }
522 * vm_fault is filled by the pagefault handler and passed to the vma's
523 * ->fault function. The vma's ->fault is responsible for returning a bitmask
524 * of VM_FAULT_xxx flags that give details about how the fault was handled.
526 * MM layer fills up gfp_mask for page allocations but fault handler might
527 * alter it if its implementation requires a different allocation context.
529 * pgoff should be used in favour of virtual_address, if possible.
533 struct vm_area_struct
*vma
; /* Target VMA */
534 gfp_t gfp_mask
; /* gfp mask to be used for allocations */
535 pgoff_t pgoff
; /* Logical page offset based on vma */
536 unsigned long address
; /* Faulting virtual address */
538 enum fault_flag flags
; /* FAULT_FLAG_xxx flags
539 * XXX: should really be 'const' */
540 pmd_t
*pmd
; /* Pointer to pmd entry matching
542 pud_t
*pud
; /* Pointer to pud entry matching
546 pte_t orig_pte
; /* Value of PTE at the time of fault */
547 pmd_t orig_pmd
; /* Value of PMD at the time of fault,
548 * used by PMD fault only.
552 struct page
*cow_page
; /* Page handler may use for COW fault */
553 struct page
*page
; /* ->fault handlers should return a
554 * page here, unless VM_FAULT_NOPAGE
555 * is set (which is also implied by
558 /* These three entries are valid only while holding ptl lock */
559 pte_t
*pte
; /* Pointer to pte entry matching
560 * the 'address'. NULL if the page
561 * table hasn't been allocated.
563 spinlock_t
*ptl
; /* Page table lock.
564 * Protects pte page table if 'pte'
565 * is not NULL, otherwise pmd.
567 pgtable_t prealloc_pte
; /* Pre-allocated pte page table.
568 * vm_ops->map_pages() sets up a page
569 * table from atomic context.
570 * do_fault_around() pre-allocates
571 * page table to avoid allocation from
576 /* page entry size for vm->huge_fault() */
577 enum page_entry_size
{
584 * These are the virtual MM functions - opening of an area, closing and
585 * unmapping it (needed to keep files on disk up-to-date etc), pointer
586 * to the functions called when a no-page or a wp-page exception occurs.
588 struct vm_operations_struct
{
589 void (*open
)(struct vm_area_struct
* area
);
590 void (*close
)(struct vm_area_struct
* area
);
591 /* Called any time before splitting to check if it's allowed */
592 int (*may_split
)(struct vm_area_struct
*area
, unsigned long addr
);
593 int (*mremap
)(struct vm_area_struct
*area
);
595 * Called by mprotect() to make driver-specific permission
596 * checks before mprotect() is finalised. The VMA must not
597 * be modified. Returns 0 if eprotect() can proceed.
599 int (*mprotect
)(struct vm_area_struct
*vma
, unsigned long start
,
600 unsigned long end
, unsigned long newflags
);
601 vm_fault_t (*fault
)(struct vm_fault
*vmf
);
602 vm_fault_t (*huge_fault
)(struct vm_fault
*vmf
,
603 enum page_entry_size pe_size
);
604 vm_fault_t (*map_pages
)(struct vm_fault
*vmf
,
605 pgoff_t start_pgoff
, pgoff_t end_pgoff
);
606 unsigned long (*pagesize
)(struct vm_area_struct
* area
);
608 /* notification that a previously read-only page is about to become
609 * writable, if an error is returned it will cause a SIGBUS */
610 vm_fault_t (*page_mkwrite
)(struct vm_fault
*vmf
);
612 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
613 vm_fault_t (*pfn_mkwrite
)(struct vm_fault
*vmf
);
615 /* called by access_process_vm when get_user_pages() fails, typically
616 * for use by special VMAs. See also generic_access_phys() for a generic
617 * implementation useful for any iomem mapping.
619 int (*access
)(struct vm_area_struct
*vma
, unsigned long addr
,
620 void *buf
, int len
, int write
);
622 /* Called by the /proc/PID/maps code to ask the vma whether it
623 * has a special name. Returning non-NULL will also cause this
624 * vma to be dumped unconditionally. */
625 const char *(*name
)(struct vm_area_struct
*vma
);
629 * set_policy() op must add a reference to any non-NULL @new mempolicy
630 * to hold the policy upon return. Caller should pass NULL @new to
631 * remove a policy and fall back to surrounding context--i.e. do not
632 * install a MPOL_DEFAULT policy, nor the task or system default
635 int (*set_policy
)(struct vm_area_struct
*vma
, struct mempolicy
*new);
638 * get_policy() op must add reference [mpol_get()] to any policy at
639 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
640 * in mm/mempolicy.c will do this automatically.
641 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
642 * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
643 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
644 * must return NULL--i.e., do not "fallback" to task or system default
647 struct mempolicy
*(*get_policy
)(struct vm_area_struct
*vma
,
651 * Called by vm_normal_page() for special PTEs to find the
652 * page for @addr. This is useful if the default behavior
653 * (using pte_page()) would not find the correct page.
655 struct page
*(*find_special_page
)(struct vm_area_struct
*vma
,
659 static inline void vma_init(struct vm_area_struct
*vma
, struct mm_struct
*mm
)
661 static const struct vm_operations_struct dummy_vm_ops
= {};
663 memset(vma
, 0, sizeof(*vma
));
665 vma
->vm_ops
= &dummy_vm_ops
;
666 INIT_LIST_HEAD(&vma
->anon_vma_chain
);
669 static inline void vma_set_anonymous(struct vm_area_struct
*vma
)
674 static inline bool vma_is_anonymous(struct vm_area_struct
*vma
)
679 static inline bool vma_is_temporary_stack(struct vm_area_struct
*vma
)
681 int maybe_stack
= vma
->vm_flags
& (VM_GROWSDOWN
| VM_GROWSUP
);
686 if ((vma
->vm_flags
& VM_STACK_INCOMPLETE_SETUP
) ==
687 VM_STACK_INCOMPLETE_SETUP
)
693 static inline bool vma_is_foreign(struct vm_area_struct
*vma
)
698 if (current
->mm
!= vma
->vm_mm
)
704 static inline bool vma_is_accessible(struct vm_area_struct
*vma
)
706 return vma
->vm_flags
& VM_ACCESS_FLAGS
;
711 * The vma_is_shmem is not inline because it is used only by slow
712 * paths in userfault.
714 bool vma_is_shmem(struct vm_area_struct
*vma
);
716 static inline bool vma_is_shmem(struct vm_area_struct
*vma
) { return false; }
719 int vma_is_stack_for_current(struct vm_area_struct
*vma
);
721 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
722 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
727 #include <linux/huge_mm.h>
730 * Methods to modify the page usage count.
732 * What counts for a page usage:
733 * - cache mapping (page->mapping)
734 * - private data (page->private)
735 * - page mapped in a task's page tables, each mapping
736 * is counted separately
738 * Also, many kernel routines increase the page count before a critical
739 * routine so they can be sure the page doesn't go away from under them.
743 * Drop a ref, return true if the refcount fell to zero (the page has no users)
745 static inline int put_page_testzero(struct page
*page
)
747 VM_BUG_ON_PAGE(page_ref_count(page
) == 0, page
);
748 return page_ref_dec_and_test(page
);
752 * Try to grab a ref unless the page has a refcount of zero, return false if
754 * This can be called when MMU is off so it must not access
755 * any of the virtual mappings.
757 static inline int get_page_unless_zero(struct page
*page
)
759 return page_ref_add_unless(page
, 1, 0);
762 extern int page_is_ram(unsigned long pfn
);
770 int region_intersects(resource_size_t offset
, size_t size
, unsigned long flags
,
773 /* Support for virtually mapped pages */
774 struct page
*vmalloc_to_page(const void *addr
);
775 unsigned long vmalloc_to_pfn(const void *addr
);
778 * Determine if an address is within the vmalloc range
780 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
781 * is no special casing required.
784 #ifndef is_ioremap_addr
785 #define is_ioremap_addr(x) is_vmalloc_addr(x)
789 extern bool is_vmalloc_addr(const void *x
);
790 extern int is_vmalloc_or_module_addr(const void *x
);
792 static inline bool is_vmalloc_addr(const void *x
)
796 static inline int is_vmalloc_or_module_addr(const void *x
)
802 extern void *kvmalloc_node(size_t size
, gfp_t flags
, int node
);
803 static inline void *kvmalloc(size_t size
, gfp_t flags
)
805 return kvmalloc_node(size
, flags
, NUMA_NO_NODE
);
807 static inline void *kvzalloc_node(size_t size
, gfp_t flags
, int node
)
809 return kvmalloc_node(size
, flags
| __GFP_ZERO
, node
);
811 static inline void *kvzalloc(size_t size
, gfp_t flags
)
813 return kvmalloc(size
, flags
| __GFP_ZERO
);
816 static inline void *kvmalloc_array(size_t n
, size_t size
, gfp_t flags
)
820 if (unlikely(check_mul_overflow(n
, size
, &bytes
)))
823 return kvmalloc(bytes
, flags
);
826 static inline void *kvcalloc(size_t n
, size_t size
, gfp_t flags
)
828 return kvmalloc_array(n
, size
, flags
| __GFP_ZERO
);
831 extern void *kvrealloc(const void *p
, size_t oldsize
, size_t newsize
,
833 extern void kvfree(const void *addr
);
834 extern void kvfree_sensitive(const void *addr
, size_t len
);
836 static inline int head_compound_mapcount(struct page
*head
)
838 return atomic_read(compound_mapcount_ptr(head
)) + 1;
842 * Mapcount of compound page as a whole, does not include mapped sub-pages.
844 * Must be called only for compound pages or any their tail sub-pages.
846 static inline int compound_mapcount(struct page
*page
)
848 VM_BUG_ON_PAGE(!PageCompound(page
), page
);
849 page
= compound_head(page
);
850 return head_compound_mapcount(page
);
854 * The atomic page->_mapcount, starts from -1: so that transitions
855 * both from it and to it can be tracked, using atomic_inc_and_test
856 * and atomic_add_negative(-1).
858 static inline void page_mapcount_reset(struct page
*page
)
860 atomic_set(&(page
)->_mapcount
, -1);
863 int __page_mapcount(struct page
*page
);
866 * Mapcount of 0-order page; when compound sub-page, includes
867 * compound_mapcount().
869 * Result is undefined for pages which cannot be mapped into userspace.
870 * For example SLAB or special types of pages. See function page_has_type().
871 * They use this place in struct page differently.
873 static inline int page_mapcount(struct page
*page
)
875 if (unlikely(PageCompound(page
)))
876 return __page_mapcount(page
);
877 return atomic_read(&page
->_mapcount
) + 1;
880 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
881 int total_mapcount(struct page
*page
);
882 int page_trans_huge_mapcount(struct page
*page
, int *total_mapcount
);
884 static inline int total_mapcount(struct page
*page
)
886 return page_mapcount(page
);
888 static inline int page_trans_huge_mapcount(struct page
*page
,
891 int mapcount
= page_mapcount(page
);
893 *total_mapcount
= mapcount
;
898 static inline struct page
*virt_to_head_page(const void *x
)
900 struct page
*page
= virt_to_page(x
);
902 return compound_head(page
);
905 void __put_page(struct page
*page
);
907 void put_pages_list(struct list_head
*pages
);
909 void split_page(struct page
*page
, unsigned int order
);
910 void copy_huge_page(struct page
*dst
, struct page
*src
);
913 * Compound pages have a destructor function. Provide a
914 * prototype for that function and accessor functions.
915 * These are _only_ valid on the head of a compound page.
917 typedef void compound_page_dtor(struct page
*);
919 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
920 enum compound_dtor_id
{
923 #ifdef CONFIG_HUGETLB_PAGE
926 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
931 extern compound_page_dtor
* const compound_page_dtors
[NR_COMPOUND_DTORS
];
933 static inline void set_compound_page_dtor(struct page
*page
,
934 enum compound_dtor_id compound_dtor
)
936 VM_BUG_ON_PAGE(compound_dtor
>= NR_COMPOUND_DTORS
, page
);
937 page
[1].compound_dtor
= compound_dtor
;
940 static inline void destroy_compound_page(struct page
*page
)
942 VM_BUG_ON_PAGE(page
[1].compound_dtor
>= NR_COMPOUND_DTORS
, page
);
943 compound_page_dtors
[page
[1].compound_dtor
](page
);
946 static inline unsigned int compound_order(struct page
*page
)
950 return page
[1].compound_order
;
953 static inline bool hpage_pincount_available(struct page
*page
)
956 * Can the page->hpage_pinned_refcount field be used? That field is in
957 * the 3rd page of the compound page, so the smallest (2-page) compound
958 * pages cannot support it.
960 page
= compound_head(page
);
961 return PageCompound(page
) && compound_order(page
) > 1;
964 static inline int head_compound_pincount(struct page
*head
)
966 return atomic_read(compound_pincount_ptr(head
));
969 static inline int compound_pincount(struct page
*page
)
971 VM_BUG_ON_PAGE(!hpage_pincount_available(page
), page
);
972 page
= compound_head(page
);
973 return head_compound_pincount(page
);
976 static inline void set_compound_order(struct page
*page
, unsigned int order
)
978 page
[1].compound_order
= order
;
979 page
[1].compound_nr
= 1U << order
;
982 /* Returns the number of pages in this potentially compound page. */
983 static inline unsigned long compound_nr(struct page
*page
)
987 return page
[1].compound_nr
;
990 /* Returns the number of bytes in this potentially compound page. */
991 static inline unsigned long page_size(struct page
*page
)
993 return PAGE_SIZE
<< compound_order(page
);
996 /* Returns the number of bits needed for the number of bytes in a page */
997 static inline unsigned int page_shift(struct page
*page
)
999 return PAGE_SHIFT
+ compound_order(page
);
1002 void free_compound_page(struct page
*page
);
1006 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
1007 * servicing faults for write access. In the normal case, do always want
1008 * pte_mkwrite. But get_user_pages can cause write faults for mappings
1009 * that do not have writing enabled, when used by access_process_vm.
1011 static inline pte_t
maybe_mkwrite(pte_t pte
, struct vm_area_struct
*vma
)
1013 if (likely(vma
->vm_flags
& VM_WRITE
))
1014 pte
= pte_mkwrite(pte
);
1018 vm_fault_t
do_set_pmd(struct vm_fault
*vmf
, struct page
*page
);
1019 void do_set_pte(struct vm_fault
*vmf
, struct page
*page
, unsigned long addr
);
1021 vm_fault_t
finish_fault(struct vm_fault
*vmf
);
1022 vm_fault_t
finish_mkwrite_fault(struct vm_fault
*vmf
);
1026 * Multiple processes may "see" the same page. E.g. for untouched
1027 * mappings of /dev/null, all processes see the same page full of
1028 * zeroes, and text pages of executables and shared libraries have
1029 * only one copy in memory, at most, normally.
1031 * For the non-reserved pages, page_count(page) denotes a reference count.
1032 * page_count() == 0 means the page is free. page->lru is then used for
1033 * freelist management in the buddy allocator.
1034 * page_count() > 0 means the page has been allocated.
1036 * Pages are allocated by the slab allocator in order to provide memory
1037 * to kmalloc and kmem_cache_alloc. In this case, the management of the
1038 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
1039 * unless a particular usage is carefully commented. (the responsibility of
1040 * freeing the kmalloc memory is the caller's, of course).
1042 * A page may be used by anyone else who does a __get_free_page().
1043 * In this case, page_count still tracks the references, and should only
1044 * be used through the normal accessor functions. The top bits of page->flags
1045 * and page->virtual store page management information, but all other fields
1046 * are unused and could be used privately, carefully. The management of this
1047 * page is the responsibility of the one who allocated it, and those who have
1048 * subsequently been given references to it.
1050 * The other pages (we may call them "pagecache pages") are completely
1051 * managed by the Linux memory manager: I/O, buffers, swapping etc.
1052 * The following discussion applies only to them.
1054 * A pagecache page contains an opaque `private' member, which belongs to the
1055 * page's address_space. Usually, this is the address of a circular list of
1056 * the page's disk buffers. PG_private must be set to tell the VM to call
1057 * into the filesystem to release these pages.
1059 * A page may belong to an inode's memory mapping. In this case, page->mapping
1060 * is the pointer to the inode, and page->index is the file offset of the page,
1061 * in units of PAGE_SIZE.
1063 * If pagecache pages are not associated with an inode, they are said to be
1064 * anonymous pages. These may become associated with the swapcache, and in that
1065 * case PG_swapcache is set, and page->private is an offset into the swapcache.
1067 * In either case (swapcache or inode backed), the pagecache itself holds one
1068 * reference to the page. Setting PG_private should also increment the
1069 * refcount. The each user mapping also has a reference to the page.
1071 * The pagecache pages are stored in a per-mapping radix tree, which is
1072 * rooted at mapping->i_pages, and indexed by offset.
1073 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
1074 * lists, we instead now tag pages as dirty/writeback in the radix tree.
1076 * All pagecache pages may be subject to I/O:
1077 * - inode pages may need to be read from disk,
1078 * - inode pages which have been modified and are MAP_SHARED may need
1079 * to be written back to the inode on disk,
1080 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
1081 * modified may need to be swapped out to swap space and (later) to be read
1086 * The zone field is never updated after free_area_init_core()
1087 * sets it, so none of the operations on it need to be atomic.
1090 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
1091 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
1092 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
1093 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
1094 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
1095 #define KASAN_TAG_PGOFF (LAST_CPUPID_PGOFF - KASAN_TAG_WIDTH)
1098 * Define the bit shifts to access each section. For non-existent
1099 * sections we define the shift as 0; that plus a 0 mask ensures
1100 * the compiler will optimise away reference to them.
1102 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
1103 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
1104 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
1105 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
1106 #define KASAN_TAG_PGSHIFT (KASAN_TAG_PGOFF * (KASAN_TAG_WIDTH != 0))
1108 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
1109 #ifdef NODE_NOT_IN_PAGE_FLAGS
1110 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
1111 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
1112 SECTIONS_PGOFF : ZONES_PGOFF)
1114 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
1115 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
1116 NODES_PGOFF : ZONES_PGOFF)
1119 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
1121 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
1122 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
1123 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
1124 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
1125 #define KASAN_TAG_MASK ((1UL << KASAN_TAG_WIDTH) - 1)
1126 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
1128 static inline enum zone_type
page_zonenum(const struct page
*page
)
1130 ASSERT_EXCLUSIVE_BITS(page
->flags
, ZONES_MASK
<< ZONES_PGSHIFT
);
1131 return (page
->flags
>> ZONES_PGSHIFT
) & ZONES_MASK
;
1134 #ifdef CONFIG_ZONE_DEVICE
1135 static inline bool is_zone_device_page(const struct page
*page
)
1137 return page_zonenum(page
) == ZONE_DEVICE
;
1139 extern void memmap_init_zone_device(struct zone
*, unsigned long,
1140 unsigned long, struct dev_pagemap
*);
1142 static inline bool is_zone_device_page(const struct page
*page
)
1148 static inline bool is_zone_movable_page(const struct page
*page
)
1150 return page_zonenum(page
) == ZONE_MOVABLE
;
1153 #ifdef CONFIG_DEV_PAGEMAP_OPS
1154 void free_devmap_managed_page(struct page
*page
);
1155 DECLARE_STATIC_KEY_FALSE(devmap_managed_key
);
1157 static inline bool page_is_devmap_managed(struct page
*page
)
1159 if (!static_branch_unlikely(&devmap_managed_key
))
1161 if (!is_zone_device_page(page
))
1163 switch (page
->pgmap
->type
) {
1164 case MEMORY_DEVICE_PRIVATE
:
1165 case MEMORY_DEVICE_FS_DAX
:
1173 void put_devmap_managed_page(struct page
*page
);
1175 #else /* CONFIG_DEV_PAGEMAP_OPS */
1176 static inline bool page_is_devmap_managed(struct page
*page
)
1181 static inline void put_devmap_managed_page(struct page
*page
)
1184 #endif /* CONFIG_DEV_PAGEMAP_OPS */
1186 static inline bool is_device_private_page(const struct page
*page
)
1188 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS
) &&
1189 IS_ENABLED(CONFIG_DEVICE_PRIVATE
) &&
1190 is_zone_device_page(page
) &&
1191 page
->pgmap
->type
== MEMORY_DEVICE_PRIVATE
;
1194 static inline bool is_pci_p2pdma_page(const struct page
*page
)
1196 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS
) &&
1197 IS_ENABLED(CONFIG_PCI_P2PDMA
) &&
1198 is_zone_device_page(page
) &&
1199 page
->pgmap
->type
== MEMORY_DEVICE_PCI_P2PDMA
;
1202 /* 127: arbitrary random number, small enough to assemble well */
1203 #define page_ref_zero_or_close_to_overflow(page) \
1204 ((unsigned int) page_ref_count(page) + 127u <= 127u)
1206 static inline void get_page(struct page
*page
)
1208 page
= compound_head(page
);
1210 * Getting a normal page or the head of a compound page
1211 * requires to already have an elevated page->_refcount.
1213 VM_BUG_ON_PAGE(page_ref_zero_or_close_to_overflow(page
), page
);
1217 bool __must_check
try_grab_page(struct page
*page
, unsigned int flags
);
1218 struct page
*try_grab_compound_head(struct page
*page
, int refs
,
1219 unsigned int flags
);
1222 static inline __must_check
bool try_get_page(struct page
*page
)
1224 page
= compound_head(page
);
1225 if (WARN_ON_ONCE(page_ref_count(page
) <= 0))
1231 static inline void put_page(struct page
*page
)
1233 page
= compound_head(page
);
1236 * For devmap managed pages we need to catch refcount transition from
1237 * 2 to 1, when refcount reach one it means the page is free and we
1238 * need to inform the device driver through callback. See
1239 * include/linux/memremap.h and HMM for details.
1241 if (page_is_devmap_managed(page
)) {
1242 put_devmap_managed_page(page
);
1246 if (put_page_testzero(page
))
1251 * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
1252 * the page's refcount so that two separate items are tracked: the original page
1253 * reference count, and also a new count of how many pin_user_pages() calls were
1254 * made against the page. ("gup-pinned" is another term for the latter).
1256 * With this scheme, pin_user_pages() becomes special: such pages are marked as
1257 * distinct from normal pages. As such, the unpin_user_page() call (and its
1258 * variants) must be used in order to release gup-pinned pages.
1262 * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
1263 * counts with respect to pin_user_pages() and unpin_user_page() becomes
1264 * simpler, due to the fact that adding an even power of two to the page
1265 * refcount has the effect of using only the upper N bits, for the code that
1266 * counts up using the bias value. This means that the lower bits are left for
1267 * the exclusive use of the original code that increments and decrements by one
1268 * (or at least, by much smaller values than the bias value).
1270 * Of course, once the lower bits overflow into the upper bits (and this is
1271 * OK, because subtraction recovers the original values), then visual inspection
1272 * no longer suffices to directly view the separate counts. However, for normal
1273 * applications that don't have huge page reference counts, this won't be an
1276 * Locking: the lockless algorithm described in page_cache_get_speculative()
1277 * and page_cache_gup_pin_speculative() provides safe operation for
1278 * get_user_pages and page_mkclean and other calls that race to set up page
1281 #define GUP_PIN_COUNTING_BIAS (1U << 10)
1283 void unpin_user_page(struct page
*page
);
1284 void unpin_user_pages_dirty_lock(struct page
**pages
, unsigned long npages
,
1286 void unpin_user_page_range_dirty_lock(struct page
*page
, unsigned long npages
,
1288 void unpin_user_pages(struct page
**pages
, unsigned long npages
);
1291 * page_maybe_dma_pinned - Report if a page is pinned for DMA.
1294 * This function checks if a page has been pinned via a call to
1295 * a function in the pin_user_pages() family.
1297 * For non-huge pages, the return value is partially fuzzy: false is not fuzzy,
1298 * because it means "definitely not pinned for DMA", but true means "probably
1299 * pinned for DMA, but possibly a false positive due to having at least
1300 * GUP_PIN_COUNTING_BIAS worth of normal page references".
1302 * False positives are OK, because: a) it's unlikely for a page to get that many
1303 * refcounts, and b) all the callers of this routine are expected to be able to
1304 * deal gracefully with a false positive.
1306 * For huge pages, the result will be exactly correct. That's because we have
1307 * more tracking data available: the 3rd struct page in the compound page is
1308 * used to track the pincount (instead using of the GUP_PIN_COUNTING_BIAS
1311 * For more information, please see Documentation/core-api/pin_user_pages.rst.
1313 * Return: True, if it is likely that the page has been "dma-pinned".
1314 * False, if the page is definitely not dma-pinned.
1316 static inline bool page_maybe_dma_pinned(struct page
*page
)
1318 if (hpage_pincount_available(page
))
1319 return compound_pincount(page
) > 0;
1322 * page_ref_count() is signed. If that refcount overflows, then
1323 * page_ref_count() returns a negative value, and callers will avoid
1324 * further incrementing the refcount.
1326 * Here, for that overflow case, use the signed bit to count a little
1327 * bit higher via unsigned math, and thus still get an accurate result.
1329 return ((unsigned int)page_ref_count(compound_head(page
))) >=
1330 GUP_PIN_COUNTING_BIAS
;
1333 static inline bool is_cow_mapping(vm_flags_t flags
)
1335 return (flags
& (VM_SHARED
| VM_MAYWRITE
)) == VM_MAYWRITE
;
1339 * This should most likely only be called during fork() to see whether we
1340 * should break the cow immediately for a page on the src mm.
1342 static inline bool page_needs_cow_for_dma(struct vm_area_struct
*vma
,
1345 if (!is_cow_mapping(vma
->vm_flags
))
1348 if (!test_bit(MMF_HAS_PINNED
, &vma
->vm_mm
->flags
))
1351 return page_maybe_dma_pinned(page
);
1354 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1355 #define SECTION_IN_PAGE_FLAGS
1359 * The identification function is mainly used by the buddy allocator for
1360 * determining if two pages could be buddies. We are not really identifying
1361 * the zone since we could be using the section number id if we do not have
1362 * node id available in page flags.
1363 * We only guarantee that it will return the same value for two combinable
1366 static inline int page_zone_id(struct page
*page
)
1368 return (page
->flags
>> ZONEID_PGSHIFT
) & ZONEID_MASK
;
1371 #ifdef NODE_NOT_IN_PAGE_FLAGS
1372 extern int page_to_nid(const struct page
*page
);
1374 static inline int page_to_nid(const struct page
*page
)
1376 struct page
*p
= (struct page
*)page
;
1378 return (PF_POISONED_CHECK(p
)->flags
>> NODES_PGSHIFT
) & NODES_MASK
;
1382 #ifdef CONFIG_NUMA_BALANCING
1383 static inline int cpu_pid_to_cpupid(int cpu
, int pid
)
1385 return ((cpu
& LAST__CPU_MASK
) << LAST__PID_SHIFT
) | (pid
& LAST__PID_MASK
);
1388 static inline int cpupid_to_pid(int cpupid
)
1390 return cpupid
& LAST__PID_MASK
;
1393 static inline int cpupid_to_cpu(int cpupid
)
1395 return (cpupid
>> LAST__PID_SHIFT
) & LAST__CPU_MASK
;
1398 static inline int cpupid_to_nid(int cpupid
)
1400 return cpu_to_node(cpupid_to_cpu(cpupid
));
1403 static inline bool cpupid_pid_unset(int cpupid
)
1405 return cpupid_to_pid(cpupid
) == (-1 & LAST__PID_MASK
);
1408 static inline bool cpupid_cpu_unset(int cpupid
)
1410 return cpupid_to_cpu(cpupid
) == (-1 & LAST__CPU_MASK
);
1413 static inline bool __cpupid_match_pid(pid_t task_pid
, int cpupid
)
1415 return (task_pid
& LAST__PID_MASK
) == cpupid_to_pid(cpupid
);
1418 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1419 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1420 static inline int page_cpupid_xchg_last(struct page
*page
, int cpupid
)
1422 return xchg(&page
->_last_cpupid
, cpupid
& LAST_CPUPID_MASK
);
1425 static inline int page_cpupid_last(struct page
*page
)
1427 return page
->_last_cpupid
;
1429 static inline void page_cpupid_reset_last(struct page
*page
)
1431 page
->_last_cpupid
= -1 & LAST_CPUPID_MASK
;
1434 static inline int page_cpupid_last(struct page
*page
)
1436 return (page
->flags
>> LAST_CPUPID_PGSHIFT
) & LAST_CPUPID_MASK
;
1439 extern int page_cpupid_xchg_last(struct page
*page
, int cpupid
);
1441 static inline void page_cpupid_reset_last(struct page
*page
)
1443 page
->flags
|= LAST_CPUPID_MASK
<< LAST_CPUPID_PGSHIFT
;
1445 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1446 #else /* !CONFIG_NUMA_BALANCING */
1447 static inline int page_cpupid_xchg_last(struct page
*page
, int cpupid
)
1449 return page_to_nid(page
); /* XXX */
1452 static inline int page_cpupid_last(struct page
*page
)
1454 return page_to_nid(page
); /* XXX */
1457 static inline int cpupid_to_nid(int cpupid
)
1462 static inline int cpupid_to_pid(int cpupid
)
1467 static inline int cpupid_to_cpu(int cpupid
)
1472 static inline int cpu_pid_to_cpupid(int nid
, int pid
)
1477 static inline bool cpupid_pid_unset(int cpupid
)
1482 static inline void page_cpupid_reset_last(struct page
*page
)
1486 static inline bool cpupid_match_pid(struct task_struct
*task
, int cpupid
)
1490 #endif /* CONFIG_NUMA_BALANCING */
1492 #if defined(CONFIG_KASAN_SW_TAGS) || defined(CONFIG_KASAN_HW_TAGS)
1495 * KASAN per-page tags are stored xor'ed with 0xff. This allows to avoid
1496 * setting tags for all pages to native kernel tag value 0xff, as the default
1497 * value 0x00 maps to 0xff.
1500 static inline u8
page_kasan_tag(const struct page
*page
)
1504 if (kasan_enabled()) {
1505 tag
= (page
->flags
>> KASAN_TAG_PGSHIFT
) & KASAN_TAG_MASK
;
1512 static inline void page_kasan_tag_set(struct page
*page
, u8 tag
)
1514 unsigned long old_flags
, flags
;
1516 if (!kasan_enabled())
1520 old_flags
= READ_ONCE(page
->flags
);
1523 flags
&= ~(KASAN_TAG_MASK
<< KASAN_TAG_PGSHIFT
);
1524 flags
|= (tag
& KASAN_TAG_MASK
) << KASAN_TAG_PGSHIFT
;
1525 } while (unlikely(!try_cmpxchg(&page
->flags
, &old_flags
, flags
)));
1528 static inline void page_kasan_tag_reset(struct page
*page
)
1530 if (kasan_enabled())
1531 page_kasan_tag_set(page
, 0xff);
1534 #else /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1536 static inline u8
page_kasan_tag(const struct page
*page
)
1541 static inline void page_kasan_tag_set(struct page
*page
, u8 tag
) { }
1542 static inline void page_kasan_tag_reset(struct page
*page
) { }
1544 #endif /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
1546 static inline struct zone
*page_zone(const struct page
*page
)
1548 return &NODE_DATA(page_to_nid(page
))->node_zones
[page_zonenum(page
)];
1551 static inline pg_data_t
*page_pgdat(const struct page
*page
)
1553 return NODE_DATA(page_to_nid(page
));
1556 #ifdef SECTION_IN_PAGE_FLAGS
1557 static inline void set_page_section(struct page
*page
, unsigned long section
)
1559 page
->flags
&= ~(SECTIONS_MASK
<< SECTIONS_PGSHIFT
);
1560 page
->flags
|= (section
& SECTIONS_MASK
) << SECTIONS_PGSHIFT
;
1563 static inline unsigned long page_to_section(const struct page
*page
)
1565 return (page
->flags
>> SECTIONS_PGSHIFT
) & SECTIONS_MASK
;
1569 /* MIGRATE_CMA and ZONE_MOVABLE do not allow pin pages */
1570 #ifdef CONFIG_MIGRATION
1571 static inline bool is_pinnable_page(struct page
*page
)
1573 return !(is_zone_movable_page(page
) || is_migrate_cma_page(page
)) ||
1574 is_zero_pfn(page_to_pfn(page
));
1577 static inline bool is_pinnable_page(struct page
*page
)
1583 static inline void set_page_zone(struct page
*page
, enum zone_type zone
)
1585 page
->flags
&= ~(ZONES_MASK
<< ZONES_PGSHIFT
);
1586 page
->flags
|= (zone
& ZONES_MASK
) << ZONES_PGSHIFT
;
1589 static inline void set_page_node(struct page
*page
, unsigned long node
)
1591 page
->flags
&= ~(NODES_MASK
<< NODES_PGSHIFT
);
1592 page
->flags
|= (node
& NODES_MASK
) << NODES_PGSHIFT
;
1595 static inline void set_page_links(struct page
*page
, enum zone_type zone
,
1596 unsigned long node
, unsigned long pfn
)
1598 set_page_zone(page
, zone
);
1599 set_page_node(page
, node
);
1600 #ifdef SECTION_IN_PAGE_FLAGS
1601 set_page_section(page
, pfn_to_section_nr(pfn
));
1606 * Some inline functions in vmstat.h depend on page_zone()
1608 #include <linux/vmstat.h>
1610 static __always_inline
void *lowmem_page_address(const struct page
*page
)
1612 return page_to_virt(page
);
1615 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1616 #define HASHED_PAGE_VIRTUAL
1619 #if defined(WANT_PAGE_VIRTUAL)
1620 static inline void *page_address(const struct page
*page
)
1622 return page
->virtual;
1624 static inline void set_page_address(struct page
*page
, void *address
)
1626 page
->virtual = address
;
1628 #define page_address_init() do { } while(0)
1631 #if defined(HASHED_PAGE_VIRTUAL)
1632 void *page_address(const struct page
*page
);
1633 void set_page_address(struct page
*page
, void *virtual);
1634 void page_address_init(void);
1637 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1638 #define page_address(page) lowmem_page_address(page)
1639 #define set_page_address(page, address) do { } while(0)
1640 #define page_address_init() do { } while(0)
1643 extern void *page_rmapping(struct page
*page
);
1644 extern struct anon_vma
*page_anon_vma(struct page
*page
);
1645 extern struct address_space
*page_mapping(struct page
*page
);
1647 extern struct address_space
*__page_file_mapping(struct page
*);
1650 struct address_space
*page_file_mapping(struct page
*page
)
1652 if (unlikely(PageSwapCache(page
)))
1653 return __page_file_mapping(page
);
1655 return page
->mapping
;
1658 extern pgoff_t
__page_file_index(struct page
*page
);
1661 * Return the pagecache index of the passed page. Regular pagecache pages
1662 * use ->index whereas swapcache pages use swp_offset(->private)
1664 static inline pgoff_t
page_index(struct page
*page
)
1666 if (unlikely(PageSwapCache(page
)))
1667 return __page_file_index(page
);
1671 bool page_mapped(struct page
*page
);
1672 struct address_space
*page_mapping(struct page
*page
);
1675 * Return true only if the page has been allocated with
1676 * ALLOC_NO_WATERMARKS and the low watermark was not
1677 * met implying that the system is under some pressure.
1679 static inline bool page_is_pfmemalloc(const struct page
*page
)
1682 * lru.next has bit 1 set if the page is allocated from the
1683 * pfmemalloc reserves. Callers may simply overwrite it if
1684 * they do not need to preserve that information.
1686 return (uintptr_t)page
->lru
.next
& BIT(1);
1690 * Only to be called by the page allocator on a freshly allocated
1693 static inline void set_page_pfmemalloc(struct page
*page
)
1695 page
->lru
.next
= (void *)BIT(1);
1698 static inline void clear_page_pfmemalloc(struct page
*page
)
1700 page
->lru
.next
= NULL
;
1704 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1706 extern void pagefault_out_of_memory(void);
1708 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1709 #define offset_in_thp(page, p) ((unsigned long)(p) & (thp_size(page) - 1))
1712 * Flags passed to show_mem() and show_free_areas() to suppress output in
1715 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1717 extern void show_free_areas(unsigned int flags
, nodemask_t
*nodemask
);
1720 extern bool can_do_mlock(void);
1722 static inline bool can_do_mlock(void) { return false; }
1724 extern int user_shm_lock(size_t, struct ucounts
*);
1725 extern void user_shm_unlock(size_t, struct ucounts
*);
1728 * Parameter block passed down to zap_pte_range in exceptional cases.
1730 struct zap_details
{
1731 struct address_space
*check_mapping
; /* Check page->mapping if set */
1732 pgoff_t first_index
; /* Lowest page->index to unmap */
1733 pgoff_t last_index
; /* Highest page->index to unmap */
1734 struct page
*single_page
; /* Locked page to be unmapped */
1737 struct page
*vm_normal_page(struct vm_area_struct
*vma
, unsigned long addr
,
1739 struct page
*vm_normal_page_pmd(struct vm_area_struct
*vma
, unsigned long addr
,
1742 void zap_vma_ptes(struct vm_area_struct
*vma
, unsigned long address
,
1743 unsigned long size
);
1744 void zap_page_range(struct vm_area_struct
*vma
, unsigned long address
,
1745 unsigned long size
);
1746 void unmap_vmas(struct mmu_gather
*tlb
, struct vm_area_struct
*start_vma
,
1747 unsigned long start
, unsigned long end
);
1749 struct mmu_notifier_range
;
1751 void free_pgd_range(struct mmu_gather
*tlb
, unsigned long addr
,
1752 unsigned long end
, unsigned long floor
, unsigned long ceiling
);
1754 copy_page_range(struct vm_area_struct
*dst_vma
, struct vm_area_struct
*src_vma
);
1755 int follow_invalidate_pte(struct mm_struct
*mm
, unsigned long address
,
1756 struct mmu_notifier_range
*range
, pte_t
**ptepp
,
1757 pmd_t
**pmdpp
, spinlock_t
**ptlp
);
1758 int follow_pte(struct mm_struct
*mm
, unsigned long address
,
1759 pte_t
**ptepp
, spinlock_t
**ptlp
);
1760 int follow_pfn(struct vm_area_struct
*vma
, unsigned long address
,
1761 unsigned long *pfn
);
1762 int follow_phys(struct vm_area_struct
*vma
, unsigned long address
,
1763 unsigned int flags
, unsigned long *prot
, resource_size_t
*phys
);
1764 int generic_access_phys(struct vm_area_struct
*vma
, unsigned long addr
,
1765 void *buf
, int len
, int write
);
1767 extern void truncate_pagecache(struct inode
*inode
, loff_t
new);
1768 extern void truncate_setsize(struct inode
*inode
, loff_t newsize
);
1769 void pagecache_isize_extended(struct inode
*inode
, loff_t from
, loff_t to
);
1770 void truncate_pagecache_range(struct inode
*inode
, loff_t offset
, loff_t end
);
1771 int truncate_inode_page(struct address_space
*mapping
, struct page
*page
);
1772 int generic_error_remove_page(struct address_space
*mapping
, struct page
*page
);
1773 int invalidate_inode_page(struct page
*page
);
1776 extern vm_fault_t
handle_mm_fault(struct vm_area_struct
*vma
,
1777 unsigned long address
, unsigned int flags
,
1778 struct pt_regs
*regs
);
1779 extern int fixup_user_fault(struct mm_struct
*mm
,
1780 unsigned long address
, unsigned int fault_flags
,
1782 void unmap_mapping_page(struct page
*page
);
1783 void unmap_mapping_pages(struct address_space
*mapping
,
1784 pgoff_t start
, pgoff_t nr
, bool even_cows
);
1785 void unmap_mapping_range(struct address_space
*mapping
,
1786 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
);
1788 static inline vm_fault_t
handle_mm_fault(struct vm_area_struct
*vma
,
1789 unsigned long address
, unsigned int flags
,
1790 struct pt_regs
*regs
)
1792 /* should never happen if there's no MMU */
1794 return VM_FAULT_SIGBUS
;
1796 static inline int fixup_user_fault(struct mm_struct
*mm
, unsigned long address
,
1797 unsigned int fault_flags
, bool *unlocked
)
1799 /* should never happen if there's no MMU */
1803 static inline void unmap_mapping_page(struct page
*page
) { }
1804 static inline void unmap_mapping_pages(struct address_space
*mapping
,
1805 pgoff_t start
, pgoff_t nr
, bool even_cows
) { }
1806 static inline void unmap_mapping_range(struct address_space
*mapping
,
1807 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
) { }
1810 static inline void unmap_shared_mapping_range(struct address_space
*mapping
,
1811 loff_t
const holebegin
, loff_t
const holelen
)
1813 unmap_mapping_range(mapping
, holebegin
, holelen
, 0);
1816 extern void vma_do_file_update_time(struct vm_area_struct
*, const char[], int);
1817 extern struct file
*vma_do_pr_or_file(struct vm_area_struct
*, const char[],
1819 extern void vma_do_get_file(struct vm_area_struct
*, const char[], int);
1820 extern void vma_do_fput(struct vm_area_struct
*, const char[], int);
1822 #define vma_file_update_time(vma) vma_do_file_update_time(vma, __func__, \
1824 #define vma_pr_or_file(vma) vma_do_pr_or_file(vma, __func__, \
1826 #define vma_get_file(vma) vma_do_get_file(vma, __func__, __LINE__)
1827 #define vma_fput(vma) vma_do_fput(vma, __func__, __LINE__)
1830 extern struct file
*vmr_do_pr_or_file(struct vm_region
*, const char[], int);
1831 extern void vmr_do_fput(struct vm_region
*, const char[], int);
1833 #define vmr_pr_or_file(region) vmr_do_pr_or_file(region, __func__, \
1835 #define vmr_fput(region) vmr_do_fput(region, __func__, __LINE__)
1836 #endif /* !CONFIG_MMU */
1838 extern int access_process_vm(struct task_struct
*tsk
, unsigned long addr
,
1839 void *buf
, int len
, unsigned int gup_flags
);
1840 extern int access_remote_vm(struct mm_struct
*mm
, unsigned long addr
,
1841 void *buf
, int len
, unsigned int gup_flags
);
1842 extern int __access_remote_vm(struct mm_struct
*mm
, unsigned long addr
,
1843 void *buf
, int len
, unsigned int gup_flags
);
1845 long get_user_pages_remote(struct mm_struct
*mm
,
1846 unsigned long start
, unsigned long nr_pages
,
1847 unsigned int gup_flags
, struct page
**pages
,
1848 struct vm_area_struct
**vmas
, int *locked
);
1849 long pin_user_pages_remote(struct mm_struct
*mm
,
1850 unsigned long start
, unsigned long nr_pages
,
1851 unsigned int gup_flags
, struct page
**pages
,
1852 struct vm_area_struct
**vmas
, int *locked
);
1853 long get_user_pages(unsigned long start
, unsigned long nr_pages
,
1854 unsigned int gup_flags
, struct page
**pages
,
1855 struct vm_area_struct
**vmas
);
1856 long pin_user_pages(unsigned long start
, unsigned long nr_pages
,
1857 unsigned int gup_flags
, struct page
**pages
,
1858 struct vm_area_struct
**vmas
);
1859 long get_user_pages_locked(unsigned long start
, unsigned long nr_pages
,
1860 unsigned int gup_flags
, struct page
**pages
, int *locked
);
1861 long pin_user_pages_locked(unsigned long start
, unsigned long nr_pages
,
1862 unsigned int gup_flags
, struct page
**pages
, int *locked
);
1863 long get_user_pages_unlocked(unsigned long start
, unsigned long nr_pages
,
1864 struct page
**pages
, unsigned int gup_flags
);
1865 long pin_user_pages_unlocked(unsigned long start
, unsigned long nr_pages
,
1866 struct page
**pages
, unsigned int gup_flags
);
1868 int get_user_pages_fast(unsigned long start
, int nr_pages
,
1869 unsigned int gup_flags
, struct page
**pages
);
1870 int pin_user_pages_fast(unsigned long start
, int nr_pages
,
1871 unsigned int gup_flags
, struct page
**pages
);
1873 int account_locked_vm(struct mm_struct
*mm
, unsigned long pages
, bool inc
);
1874 int __account_locked_vm(struct mm_struct
*mm
, unsigned long pages
, bool inc
,
1875 struct task_struct
*task
, bool bypass_rlim
);
1878 int get_kernel_pages(const struct kvec
*iov
, int nr_pages
, int write
,
1879 struct page
**pages
);
1880 struct page
*get_dump_page(unsigned long addr
);
1882 extern int try_to_release_page(struct page
* page
, gfp_t gfp_mask
);
1883 extern void do_invalidatepage(struct page
*page
, unsigned int offset
,
1884 unsigned int length
);
1886 int redirty_page_for_writepage(struct writeback_control
*wbc
,
1888 void account_page_cleaned(struct page
*page
, struct address_space
*mapping
,
1889 struct bdi_writeback
*wb
);
1890 int set_page_dirty(struct page
*page
);
1891 int set_page_dirty_lock(struct page
*page
);
1892 void __cancel_dirty_page(struct page
*page
);
1893 static inline void cancel_dirty_page(struct page
*page
)
1895 /* Avoid atomic ops, locking, etc. when not actually needed. */
1896 if (PageDirty(page
))
1897 __cancel_dirty_page(page
);
1899 int clear_page_dirty_for_io(struct page
*page
);
1901 int get_cmdline(struct task_struct
*task
, char *buffer
, int buflen
);
1903 extern unsigned long move_page_tables(struct vm_area_struct
*vma
,
1904 unsigned long old_addr
, struct vm_area_struct
*new_vma
,
1905 unsigned long new_addr
, unsigned long len
,
1906 bool need_rmap_locks
);
1909 * Flags used by change_protection(). For now we make it a bitmap so
1910 * that we can pass in multiple flags just like parameters. However
1911 * for now all the callers are only use one of the flags at the same
1914 /* Whether we should allow dirty bit accounting */
1915 #define MM_CP_DIRTY_ACCT (1UL << 0)
1916 /* Whether this protection change is for NUMA hints */
1917 #define MM_CP_PROT_NUMA (1UL << 1)
1918 /* Whether this change is for write protecting */
1919 #define MM_CP_UFFD_WP (1UL << 2) /* do wp */
1920 #define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */
1921 #define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \
1922 MM_CP_UFFD_WP_RESOLVE)
1924 extern unsigned long change_protection(struct vm_area_struct
*vma
, unsigned long start
,
1925 unsigned long end
, pgprot_t newprot
,
1926 unsigned long cp_flags
);
1927 extern int mprotect_fixup(struct vm_area_struct
*vma
,
1928 struct vm_area_struct
**pprev
, unsigned long start
,
1929 unsigned long end
, unsigned long newflags
);
1932 * doesn't attempt to fault and will return short.
1934 int get_user_pages_fast_only(unsigned long start
, int nr_pages
,
1935 unsigned int gup_flags
, struct page
**pages
);
1936 int pin_user_pages_fast_only(unsigned long start
, int nr_pages
,
1937 unsigned int gup_flags
, struct page
**pages
);
1939 static inline bool get_user_page_fast_only(unsigned long addr
,
1940 unsigned int gup_flags
, struct page
**pagep
)
1942 return get_user_pages_fast_only(addr
, 1, gup_flags
, pagep
) == 1;
1945 * per-process(per-mm_struct) statistics.
1947 static inline unsigned long get_mm_counter(struct mm_struct
*mm
, int member
)
1949 long val
= atomic_long_read(&mm
->rss_stat
.count
[member
]);
1951 #ifdef SPLIT_RSS_COUNTING
1953 * counter is updated in asynchronous manner and may go to minus.
1954 * But it's never be expected number for users.
1959 return (unsigned long)val
;
1962 void mm_trace_rss_stat(struct mm_struct
*mm
, int member
, long count
);
1964 static inline void add_mm_counter(struct mm_struct
*mm
, int member
, long value
)
1966 long count
= atomic_long_add_return(value
, &mm
->rss_stat
.count
[member
]);
1968 mm_trace_rss_stat(mm
, member
, count
);
1971 static inline void inc_mm_counter(struct mm_struct
*mm
, int member
)
1973 long count
= atomic_long_inc_return(&mm
->rss_stat
.count
[member
]);
1975 mm_trace_rss_stat(mm
, member
, count
);
1978 static inline void dec_mm_counter(struct mm_struct
*mm
, int member
)
1980 long count
= atomic_long_dec_return(&mm
->rss_stat
.count
[member
]);
1982 mm_trace_rss_stat(mm
, member
, count
);
1985 /* Optimized variant when page is already known not to be PageAnon */
1986 static inline int mm_counter_file(struct page
*page
)
1988 if (PageSwapBacked(page
))
1989 return MM_SHMEMPAGES
;
1990 return MM_FILEPAGES
;
1993 static inline int mm_counter(struct page
*page
)
1996 return MM_ANONPAGES
;
1997 return mm_counter_file(page
);
2000 static inline unsigned long get_mm_rss(struct mm_struct
*mm
)
2002 return get_mm_counter(mm
, MM_FILEPAGES
) +
2003 get_mm_counter(mm
, MM_ANONPAGES
) +
2004 get_mm_counter(mm
, MM_SHMEMPAGES
);
2007 static inline unsigned long get_mm_hiwater_rss(struct mm_struct
*mm
)
2009 return max(mm
->hiwater_rss
, get_mm_rss(mm
));
2012 static inline unsigned long get_mm_hiwater_vm(struct mm_struct
*mm
)
2014 return max(mm
->hiwater_vm
, mm
->total_vm
);
2017 static inline void update_hiwater_rss(struct mm_struct
*mm
)
2019 unsigned long _rss
= get_mm_rss(mm
);
2021 if ((mm
)->hiwater_rss
< _rss
)
2022 (mm
)->hiwater_rss
= _rss
;
2025 static inline void update_hiwater_vm(struct mm_struct
*mm
)
2027 if (mm
->hiwater_vm
< mm
->total_vm
)
2028 mm
->hiwater_vm
= mm
->total_vm
;
2031 static inline void reset_mm_hiwater_rss(struct mm_struct
*mm
)
2033 mm
->hiwater_rss
= get_mm_rss(mm
);
2036 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss
,
2037 struct mm_struct
*mm
)
2039 unsigned long hiwater_rss
= get_mm_hiwater_rss(mm
);
2041 if (*maxrss
< hiwater_rss
)
2042 *maxrss
= hiwater_rss
;
2045 #if defined(SPLIT_RSS_COUNTING)
2046 void sync_mm_rss(struct mm_struct
*mm
);
2048 static inline void sync_mm_rss(struct mm_struct
*mm
)
2053 #ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
2054 static inline int pte_special(pte_t pte
)
2059 static inline pte_t
pte_mkspecial(pte_t pte
)
2065 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
2066 static inline int pte_devmap(pte_t pte
)
2072 int vma_wants_writenotify(struct vm_area_struct
*vma
, pgprot_t vm_page_prot
);
2074 extern pte_t
*__get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
2076 static inline pte_t
*get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
2080 __cond_lock(*ptl
, ptep
= __get_locked_pte(mm
, addr
, ptl
));
2084 #ifdef __PAGETABLE_P4D_FOLDED
2085 static inline int __p4d_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
2086 unsigned long address
)
2091 int __p4d_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
);
2094 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
2095 static inline int __pud_alloc(struct mm_struct
*mm
, p4d_t
*p4d
,
2096 unsigned long address
)
2100 static inline void mm_inc_nr_puds(struct mm_struct
*mm
) {}
2101 static inline void mm_dec_nr_puds(struct mm_struct
*mm
) {}
2104 int __pud_alloc(struct mm_struct
*mm
, p4d_t
*p4d
, unsigned long address
);
2106 static inline void mm_inc_nr_puds(struct mm_struct
*mm
)
2108 if (mm_pud_folded(mm
))
2110 atomic_long_add(PTRS_PER_PUD
* sizeof(pud_t
), &mm
->pgtables_bytes
);
2113 static inline void mm_dec_nr_puds(struct mm_struct
*mm
)
2115 if (mm_pud_folded(mm
))
2117 atomic_long_sub(PTRS_PER_PUD
* sizeof(pud_t
), &mm
->pgtables_bytes
);
2121 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
2122 static inline int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
,
2123 unsigned long address
)
2128 static inline void mm_inc_nr_pmds(struct mm_struct
*mm
) {}
2129 static inline void mm_dec_nr_pmds(struct mm_struct
*mm
) {}
2132 int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
);
2134 static inline void mm_inc_nr_pmds(struct mm_struct
*mm
)
2136 if (mm_pmd_folded(mm
))
2138 atomic_long_add(PTRS_PER_PMD
* sizeof(pmd_t
), &mm
->pgtables_bytes
);
2141 static inline void mm_dec_nr_pmds(struct mm_struct
*mm
)
2143 if (mm_pmd_folded(mm
))
2145 atomic_long_sub(PTRS_PER_PMD
* sizeof(pmd_t
), &mm
->pgtables_bytes
);
2150 static inline void mm_pgtables_bytes_init(struct mm_struct
*mm
)
2152 atomic_long_set(&mm
->pgtables_bytes
, 0);
2155 static inline unsigned long mm_pgtables_bytes(const struct mm_struct
*mm
)
2157 return atomic_long_read(&mm
->pgtables_bytes
);
2160 static inline void mm_inc_nr_ptes(struct mm_struct
*mm
)
2162 atomic_long_add(PTRS_PER_PTE
* sizeof(pte_t
), &mm
->pgtables_bytes
);
2165 static inline void mm_dec_nr_ptes(struct mm_struct
*mm
)
2167 atomic_long_sub(PTRS_PER_PTE
* sizeof(pte_t
), &mm
->pgtables_bytes
);
2171 static inline void mm_pgtables_bytes_init(struct mm_struct
*mm
) {}
2172 static inline unsigned long mm_pgtables_bytes(const struct mm_struct
*mm
)
2177 static inline void mm_inc_nr_ptes(struct mm_struct
*mm
) {}
2178 static inline void mm_dec_nr_ptes(struct mm_struct
*mm
) {}
2181 int __pte_alloc(struct mm_struct
*mm
, pmd_t
*pmd
);
2182 int __pte_alloc_kernel(pmd_t
*pmd
);
2184 #if defined(CONFIG_MMU)
2186 static inline p4d_t
*p4d_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
2187 unsigned long address
)
2189 return (unlikely(pgd_none(*pgd
)) && __p4d_alloc(mm
, pgd
, address
)) ?
2190 NULL
: p4d_offset(pgd
, address
);
2193 static inline pud_t
*pud_alloc(struct mm_struct
*mm
, p4d_t
*p4d
,
2194 unsigned long address
)
2196 return (unlikely(p4d_none(*p4d
)) && __pud_alloc(mm
, p4d
, address
)) ?
2197 NULL
: pud_offset(p4d
, address
);
2200 static inline pmd_t
*pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
)
2202 return (unlikely(pud_none(*pud
)) && __pmd_alloc(mm
, pud
, address
))?
2203 NULL
: pmd_offset(pud
, address
);
2205 #endif /* CONFIG_MMU */
2207 #if USE_SPLIT_PTE_PTLOCKS
2208 #if ALLOC_SPLIT_PTLOCKS
2209 void __init
ptlock_cache_init(void);
2210 extern bool ptlock_alloc(struct page
*page
);
2211 extern void ptlock_free(struct page
*page
);
2213 static inline spinlock_t
*ptlock_ptr(struct page
*page
)
2217 #else /* ALLOC_SPLIT_PTLOCKS */
2218 static inline void ptlock_cache_init(void)
2222 static inline bool ptlock_alloc(struct page
*page
)
2227 static inline void ptlock_free(struct page
*page
)
2231 static inline spinlock_t
*ptlock_ptr(struct page
*page
)
2235 #endif /* ALLOC_SPLIT_PTLOCKS */
2237 static inline spinlock_t
*pte_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
2239 return ptlock_ptr(pmd_page(*pmd
));
2242 static inline bool ptlock_init(struct page
*page
)
2245 * prep_new_page() initialize page->private (and therefore page->ptl)
2246 * with 0. Make sure nobody took it in use in between.
2248 * It can happen if arch try to use slab for page table allocation:
2249 * slab code uses page->slab_cache, which share storage with page->ptl.
2251 VM_BUG_ON_PAGE(*(unsigned long *)&page
->ptl
, page
);
2252 if (!ptlock_alloc(page
))
2254 spin_lock_init(ptlock_ptr(page
));
2258 #else /* !USE_SPLIT_PTE_PTLOCKS */
2260 * We use mm->page_table_lock to guard all pagetable pages of the mm.
2262 static inline spinlock_t
*pte_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
2264 return &mm
->page_table_lock
;
2266 static inline void ptlock_cache_init(void) {}
2267 static inline bool ptlock_init(struct page
*page
) { return true; }
2268 static inline void ptlock_free(struct page
*page
) {}
2269 #endif /* USE_SPLIT_PTE_PTLOCKS */
2271 static inline void pgtable_init(void)
2273 ptlock_cache_init();
2274 pgtable_cache_init();
2277 static inline bool pgtable_pte_page_ctor(struct page
*page
)
2279 if (!ptlock_init(page
))
2281 __SetPageTable(page
);
2282 inc_lruvec_page_state(page
, NR_PAGETABLE
);
2286 static inline void pgtable_pte_page_dtor(struct page
*page
)
2289 __ClearPageTable(page
);
2290 dec_lruvec_page_state(page
, NR_PAGETABLE
);
2293 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
2295 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
2296 pte_t *__pte = pte_offset_map(pmd, address); \
2302 #define pte_unmap_unlock(pte, ptl) do { \
2307 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2309 #define pte_alloc_map(mm, pmd, address) \
2310 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2312 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2313 (pte_alloc(mm, pmd) ? \
2314 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2316 #define pte_alloc_kernel(pmd, address) \
2317 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2318 NULL: pte_offset_kernel(pmd, address))
2320 #if USE_SPLIT_PMD_PTLOCKS
2322 static struct page
*pmd_to_page(pmd_t
*pmd
)
2324 unsigned long mask
= ~(PTRS_PER_PMD
* sizeof(pmd_t
) - 1);
2325 return virt_to_page((void *)((unsigned long) pmd
& mask
));
2328 static inline spinlock_t
*pmd_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
2330 return ptlock_ptr(pmd_to_page(pmd
));
2333 static inline bool pmd_ptlock_init(struct page
*page
)
2335 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2336 page
->pmd_huge_pte
= NULL
;
2338 return ptlock_init(page
);
2341 static inline void pmd_ptlock_free(struct page
*page
)
2343 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2344 VM_BUG_ON_PAGE(page
->pmd_huge_pte
, page
);
2349 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2353 static inline spinlock_t
*pmd_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
2355 return &mm
->page_table_lock
;
2358 static inline bool pmd_ptlock_init(struct page
*page
) { return true; }
2359 static inline void pmd_ptlock_free(struct page
*page
) {}
2361 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2365 static inline spinlock_t
*pmd_lock(struct mm_struct
*mm
, pmd_t
*pmd
)
2367 spinlock_t
*ptl
= pmd_lockptr(mm
, pmd
);
2372 static inline bool pgtable_pmd_page_ctor(struct page
*page
)
2374 if (!pmd_ptlock_init(page
))
2376 __SetPageTable(page
);
2377 inc_lruvec_page_state(page
, NR_PAGETABLE
);
2381 static inline void pgtable_pmd_page_dtor(struct page
*page
)
2383 pmd_ptlock_free(page
);
2384 __ClearPageTable(page
);
2385 dec_lruvec_page_state(page
, NR_PAGETABLE
);
2389 * No scalability reason to split PUD locks yet, but follow the same pattern
2390 * as the PMD locks to make it easier if we decide to. The VM should not be
2391 * considered ready to switch to split PUD locks yet; there may be places
2392 * which need to be converted from page_table_lock.
2394 static inline spinlock_t
*pud_lockptr(struct mm_struct
*mm
, pud_t
*pud
)
2396 return &mm
->page_table_lock
;
2399 static inline spinlock_t
*pud_lock(struct mm_struct
*mm
, pud_t
*pud
)
2401 spinlock_t
*ptl
= pud_lockptr(mm
, pud
);
2407 extern void __init
pagecache_init(void);
2408 extern void __init
free_area_init_memoryless_node(int nid
);
2409 extern void free_initmem(void);
2412 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2413 * into the buddy system. The freed pages will be poisoned with pattern
2414 * "poison" if it's within range [0, UCHAR_MAX].
2415 * Return pages freed into the buddy system.
2417 extern unsigned long free_reserved_area(void *start
, void *end
,
2418 int poison
, const char *s
);
2420 extern void adjust_managed_page_count(struct page
*page
, long count
);
2421 extern void mem_init_print_info(void);
2423 extern void reserve_bootmem_region(phys_addr_t start
, phys_addr_t end
);
2425 /* Free the reserved page into the buddy system, so it gets managed. */
2426 static inline void free_reserved_page(struct page
*page
)
2428 ClearPageReserved(page
);
2429 init_page_count(page
);
2431 adjust_managed_page_count(page
, 1);
2433 #define free_highmem_page(page) free_reserved_page(page)
2435 static inline void mark_page_reserved(struct page
*page
)
2437 SetPageReserved(page
);
2438 adjust_managed_page_count(page
, -1);
2442 * Default method to free all the __init memory into the buddy system.
2443 * The freed pages will be poisoned with pattern "poison" if it's within
2444 * range [0, UCHAR_MAX].
2445 * Return pages freed into the buddy system.
2447 static inline unsigned long free_initmem_default(int poison
)
2449 extern char __init_begin
[], __init_end
[];
2451 return free_reserved_area(&__init_begin
, &__init_end
,
2452 poison
, "unused kernel image (initmem)");
2455 static inline unsigned long get_num_physpages(void)
2458 unsigned long phys_pages
= 0;
2460 for_each_online_node(nid
)
2461 phys_pages
+= node_present_pages(nid
);
2467 * Using memblock node mappings, an architecture may initialise its
2468 * zones, allocate the backing mem_map and account for memory holes in an
2469 * architecture independent manner.
2471 * An architecture is expected to register range of page frames backed by
2472 * physical memory with memblock_add[_node]() before calling
2473 * free_area_init() passing in the PFN each zone ends at. At a basic
2474 * usage, an architecture is expected to do something like
2476 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2478 * for_each_valid_physical_page_range()
2479 * memblock_add_node(base, size, nid)
2480 * free_area_init(max_zone_pfns);
2482 void free_area_init(unsigned long *max_zone_pfn
);
2483 unsigned long node_map_pfn_alignment(void);
2484 unsigned long __absent_pages_in_range(int nid
, unsigned long start_pfn
,
2485 unsigned long end_pfn
);
2486 extern unsigned long absent_pages_in_range(unsigned long start_pfn
,
2487 unsigned long end_pfn
);
2488 extern void get_pfn_range_for_nid(unsigned int nid
,
2489 unsigned long *start_pfn
, unsigned long *end_pfn
);
2490 extern unsigned long find_min_pfn_with_active_regions(void);
2493 static inline int early_pfn_to_nid(unsigned long pfn
)
2498 /* please see mm/page_alloc.c */
2499 extern int __meminit
early_pfn_to_nid(unsigned long pfn
);
2502 extern void set_dma_reserve(unsigned long new_dma_reserve
);
2503 extern void memmap_init_range(unsigned long, int, unsigned long,
2504 unsigned long, unsigned long, enum meminit_context
,
2505 struct vmem_altmap
*, int migratetype
);
2506 extern void setup_per_zone_wmarks(void);
2507 extern int __meminit
init_per_zone_wmark_min(void);
2508 extern void mem_init(void);
2509 extern void __init
mmap_init(void);
2510 extern void show_mem(unsigned int flags
, nodemask_t
*nodemask
);
2511 extern long si_mem_available(void);
2512 extern void si_meminfo(struct sysinfo
* val
);
2513 extern void si_meminfo_node(struct sysinfo
*val
, int nid
);
2514 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2515 extern unsigned long arch_reserved_kernel_pages(void);
2518 extern __printf(3, 4)
2519 void warn_alloc(gfp_t gfp_mask
, nodemask_t
*nodemask
, const char *fmt
, ...);
2521 extern void setup_per_cpu_pageset(void);
2524 extern int min_free_kbytes
;
2525 extern int watermark_boost_factor
;
2526 extern int watermark_scale_factor
;
2527 extern bool arch_has_descending_max_zone_pfns(void);
2530 extern atomic_long_t mmap_pages_allocated
;
2531 extern int nommu_shrink_inode_mappings(struct inode
*, size_t, size_t);
2533 /* interval_tree.c */
2534 void vma_interval_tree_insert(struct vm_area_struct
*node
,
2535 struct rb_root_cached
*root
);
2536 void vma_interval_tree_insert_after(struct vm_area_struct
*node
,
2537 struct vm_area_struct
*prev
,
2538 struct rb_root_cached
*root
);
2539 void vma_interval_tree_remove(struct vm_area_struct
*node
,
2540 struct rb_root_cached
*root
);
2541 struct vm_area_struct
*vma_interval_tree_iter_first(struct rb_root_cached
*root
,
2542 unsigned long start
, unsigned long last
);
2543 struct vm_area_struct
*vma_interval_tree_iter_next(struct vm_area_struct
*node
,
2544 unsigned long start
, unsigned long last
);
2546 #define vma_interval_tree_foreach(vma, root, start, last) \
2547 for (vma = vma_interval_tree_iter_first(root, start, last); \
2548 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2550 void anon_vma_interval_tree_insert(struct anon_vma_chain
*node
,
2551 struct rb_root_cached
*root
);
2552 void anon_vma_interval_tree_remove(struct anon_vma_chain
*node
,
2553 struct rb_root_cached
*root
);
2554 struct anon_vma_chain
*
2555 anon_vma_interval_tree_iter_first(struct rb_root_cached
*root
,
2556 unsigned long start
, unsigned long last
);
2557 struct anon_vma_chain
*anon_vma_interval_tree_iter_next(
2558 struct anon_vma_chain
*node
, unsigned long start
, unsigned long last
);
2559 #ifdef CONFIG_DEBUG_VM_RB
2560 void anon_vma_interval_tree_verify(struct anon_vma_chain
*node
);
2563 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2564 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2565 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2568 extern int __vm_enough_memory(struct mm_struct
*mm
, long pages
, int cap_sys_admin
);
2569 extern int __vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
2570 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
,
2571 struct vm_area_struct
*expand
);
2572 static inline int vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
2573 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
)
2575 return __vma_adjust(vma
, start
, end
, pgoff
, insert
, NULL
);
2577 extern struct vm_area_struct
*vma_merge(struct mm_struct
*,
2578 struct vm_area_struct
*prev
, unsigned long addr
, unsigned long end
,
2579 unsigned long vm_flags
, struct anon_vma
*, struct file
*, pgoff_t
,
2580 struct mempolicy
*, struct vm_userfaultfd_ctx
);
2581 extern struct anon_vma
*find_mergeable_anon_vma(struct vm_area_struct
*);
2582 extern int __split_vma(struct mm_struct
*, struct vm_area_struct
*,
2583 unsigned long addr
, int new_below
);
2584 extern int split_vma(struct mm_struct
*, struct vm_area_struct
*,
2585 unsigned long addr
, int new_below
);
2586 extern int insert_vm_struct(struct mm_struct
*, struct vm_area_struct
*);
2587 extern void __vma_link_rb(struct mm_struct
*, struct vm_area_struct
*,
2588 struct rb_node
**, struct rb_node
*);
2589 extern void unlink_file_vma(struct vm_area_struct
*);
2590 extern struct vm_area_struct
*copy_vma(struct vm_area_struct
**,
2591 unsigned long addr
, unsigned long len
, pgoff_t pgoff
,
2592 bool *need_rmap_locks
);
2593 extern void exit_mmap(struct mm_struct
*);
2595 static inline int check_data_rlimit(unsigned long rlim
,
2597 unsigned long start
,
2598 unsigned long end_data
,
2599 unsigned long start_data
)
2601 if (rlim
< RLIM_INFINITY
) {
2602 if (((new - start
) + (end_data
- start_data
)) > rlim
)
2609 extern int mm_take_all_locks(struct mm_struct
*mm
);
2610 extern void mm_drop_all_locks(struct mm_struct
*mm
);
2612 extern int set_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
);
2613 extern int replace_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
);
2614 extern struct file
*get_mm_exe_file(struct mm_struct
*mm
);
2615 extern struct file
*get_task_exe_file(struct task_struct
*task
);
2617 extern bool may_expand_vm(struct mm_struct
*, vm_flags_t
, unsigned long npages
);
2618 extern void vm_stat_account(struct mm_struct
*, vm_flags_t
, long npages
);
2620 extern bool vma_is_special_mapping(const struct vm_area_struct
*vma
,
2621 const struct vm_special_mapping
*sm
);
2622 extern struct vm_area_struct
*_install_special_mapping(struct mm_struct
*mm
,
2623 unsigned long addr
, unsigned long len
,
2624 unsigned long flags
,
2625 const struct vm_special_mapping
*spec
);
2626 /* This is an obsolete alternative to _install_special_mapping. */
2627 extern int install_special_mapping(struct mm_struct
*mm
,
2628 unsigned long addr
, unsigned long len
,
2629 unsigned long flags
, struct page
**pages
);
2631 unsigned long randomize_stack_top(unsigned long stack_top
);
2633 extern unsigned long get_unmapped_area(struct file
*, unsigned long, unsigned long, unsigned long, unsigned long);
2635 extern unsigned long mmap_region(struct file
*file
, unsigned long addr
,
2636 unsigned long len
, vm_flags_t vm_flags
, unsigned long pgoff
,
2637 struct list_head
*uf
);
2638 extern unsigned long do_mmap(struct file
*file
, unsigned long addr
,
2639 unsigned long len
, unsigned long prot
, unsigned long flags
,
2640 unsigned long pgoff
, unsigned long *populate
, struct list_head
*uf
);
2641 extern int __do_munmap(struct mm_struct
*, unsigned long, size_t,
2642 struct list_head
*uf
, bool downgrade
);
2643 extern int do_munmap(struct mm_struct
*, unsigned long, size_t,
2644 struct list_head
*uf
);
2645 extern int do_madvise(struct mm_struct
*mm
, unsigned long start
, size_t len_in
, int behavior
);
2648 extern int __mm_populate(unsigned long addr
, unsigned long len
,
2650 static inline void mm_populate(unsigned long addr
, unsigned long len
)
2653 (void) __mm_populate(addr
, len
, 1);
2656 static inline void mm_populate(unsigned long addr
, unsigned long len
) {}
2659 /* These take the mm semaphore themselves */
2660 extern int __must_check
vm_brk(unsigned long, unsigned long);
2661 extern int __must_check
vm_brk_flags(unsigned long, unsigned long, unsigned long);
2662 extern int vm_munmap(unsigned long, size_t);
2663 extern unsigned long __must_check
vm_mmap(struct file
*, unsigned long,
2664 unsigned long, unsigned long,
2665 unsigned long, unsigned long);
2667 struct vm_unmapped_area_info
{
2668 #define VM_UNMAPPED_AREA_TOPDOWN 1
2669 unsigned long flags
;
2670 unsigned long length
;
2671 unsigned long low_limit
;
2672 unsigned long high_limit
;
2673 unsigned long align_mask
;
2674 unsigned long align_offset
;
2677 extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info
*info
);
2680 extern void truncate_inode_pages(struct address_space
*, loff_t
);
2681 extern void truncate_inode_pages_range(struct address_space
*,
2682 loff_t lstart
, loff_t lend
);
2683 extern void truncate_inode_pages_final(struct address_space
*);
2685 /* generic vm_area_ops exported for stackable file systems */
2686 extern vm_fault_t
filemap_fault(struct vm_fault
*vmf
);
2687 extern vm_fault_t
filemap_map_pages(struct vm_fault
*vmf
,
2688 pgoff_t start_pgoff
, pgoff_t end_pgoff
);
2689 extern vm_fault_t
filemap_page_mkwrite(struct vm_fault
*vmf
);
2691 /* mm/page-writeback.c */
2692 int __must_check
write_one_page(struct page
*page
);
2693 void task_dirty_inc(struct task_struct
*tsk
);
2695 extern unsigned long stack_guard_gap
;
2696 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2697 extern int expand_stack(struct vm_area_struct
*vma
, unsigned long address
);
2699 /* CONFIG_STACK_GROWSUP still needs to grow downwards at some places */
2700 extern int expand_downwards(struct vm_area_struct
*vma
,
2701 unsigned long address
);
2703 extern int expand_upwards(struct vm_area_struct
*vma
, unsigned long address
);
2705 #define expand_upwards(vma, address) (0)
2708 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2709 extern struct vm_area_struct
* find_vma(struct mm_struct
* mm
, unsigned long addr
);
2710 extern struct vm_area_struct
* find_vma_prev(struct mm_struct
* mm
, unsigned long addr
,
2711 struct vm_area_struct
**pprev
);
2714 * find_vma_intersection() - Look up the first VMA which intersects the interval
2715 * @mm: The process address space.
2716 * @start_addr: The inclusive start user address.
2717 * @end_addr: The exclusive end user address.
2719 * Returns: The first VMA within the provided range, %NULL otherwise. Assumes
2720 * start_addr < end_addr.
2723 struct vm_area_struct
*find_vma_intersection(struct mm_struct
*mm
,
2724 unsigned long start_addr
,
2725 unsigned long end_addr
)
2727 struct vm_area_struct
*vma
= find_vma(mm
, start_addr
);
2729 if (vma
&& end_addr
<= vma
->vm_start
)
2735 * vma_lookup() - Find a VMA at a specific address
2736 * @mm: The process address space.
2737 * @addr: The user address.
2739 * Return: The vm_area_struct at the given address, %NULL otherwise.
2742 struct vm_area_struct
*vma_lookup(struct mm_struct
*mm
, unsigned long addr
)
2744 struct vm_area_struct
*vma
= find_vma(mm
, addr
);
2746 if (vma
&& addr
< vma
->vm_start
)
2752 static inline unsigned long vm_start_gap(struct vm_area_struct
*vma
)
2754 unsigned long vm_start
= vma
->vm_start
;
2756 if (vma
->vm_flags
& VM_GROWSDOWN
) {
2757 vm_start
-= stack_guard_gap
;
2758 if (vm_start
> vma
->vm_start
)
2764 static inline unsigned long vm_end_gap(struct vm_area_struct
*vma
)
2766 unsigned long vm_end
= vma
->vm_end
;
2768 if (vma
->vm_flags
& VM_GROWSUP
) {
2769 vm_end
+= stack_guard_gap
;
2770 if (vm_end
< vma
->vm_end
)
2771 vm_end
= -PAGE_SIZE
;
2776 static inline unsigned long vma_pages(struct vm_area_struct
*vma
)
2778 return (vma
->vm_end
- vma
->vm_start
) >> PAGE_SHIFT
;
2781 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2782 static inline struct vm_area_struct
*find_exact_vma(struct mm_struct
*mm
,
2783 unsigned long vm_start
, unsigned long vm_end
)
2785 struct vm_area_struct
*vma
= find_vma(mm
, vm_start
);
2787 if (vma
&& (vma
->vm_start
!= vm_start
|| vma
->vm_end
!= vm_end
))
2793 static inline bool range_in_vma(struct vm_area_struct
*vma
,
2794 unsigned long start
, unsigned long end
)
2796 return (vma
&& vma
->vm_start
<= start
&& end
<= vma
->vm_end
);
2800 pgprot_t
vm_get_page_prot(unsigned long vm_flags
);
2801 void vma_set_page_prot(struct vm_area_struct
*vma
);
2803 static inline pgprot_t
vm_get_page_prot(unsigned long vm_flags
)
2807 static inline void vma_set_page_prot(struct vm_area_struct
*vma
)
2809 vma
->vm_page_prot
= vm_get_page_prot(vma
->vm_flags
);
2813 void vma_set_file(struct vm_area_struct
*vma
, struct file
*file
);
2815 #ifdef CONFIG_NUMA_BALANCING
2816 unsigned long change_prot_numa(struct vm_area_struct
*vma
,
2817 unsigned long start
, unsigned long end
);
2820 struct vm_area_struct
*find_extend_vma(struct mm_struct
*, unsigned long addr
);
2821 int remap_pfn_range(struct vm_area_struct
*, unsigned long addr
,
2822 unsigned long pfn
, unsigned long size
, pgprot_t
);
2823 int remap_pfn_range_notrack(struct vm_area_struct
*vma
, unsigned long addr
,
2824 unsigned long pfn
, unsigned long size
, pgprot_t prot
);
2825 int vm_insert_page(struct vm_area_struct
*, unsigned long addr
, struct page
*);
2826 int vm_insert_pages(struct vm_area_struct
*vma
, unsigned long addr
,
2827 struct page
**pages
, unsigned long *num
);
2828 int vm_map_pages(struct vm_area_struct
*vma
, struct page
**pages
,
2830 int vm_map_pages_zero(struct vm_area_struct
*vma
, struct page
**pages
,
2832 vm_fault_t
vmf_insert_pfn(struct vm_area_struct
*vma
, unsigned long addr
,
2834 vm_fault_t
vmf_insert_pfn_prot(struct vm_area_struct
*vma
, unsigned long addr
,
2835 unsigned long pfn
, pgprot_t pgprot
);
2836 vm_fault_t
vmf_insert_mixed(struct vm_area_struct
*vma
, unsigned long addr
,
2838 vm_fault_t
vmf_insert_mixed_prot(struct vm_area_struct
*vma
, unsigned long addr
,
2839 pfn_t pfn
, pgprot_t pgprot
);
2840 vm_fault_t
vmf_insert_mixed_mkwrite(struct vm_area_struct
*vma
,
2841 unsigned long addr
, pfn_t pfn
);
2842 int vm_iomap_memory(struct vm_area_struct
*vma
, phys_addr_t start
, unsigned long len
);
2844 static inline vm_fault_t
vmf_insert_page(struct vm_area_struct
*vma
,
2845 unsigned long addr
, struct page
*page
)
2847 int err
= vm_insert_page(vma
, addr
, page
);
2850 return VM_FAULT_OOM
;
2851 if (err
< 0 && err
!= -EBUSY
)
2852 return VM_FAULT_SIGBUS
;
2854 return VM_FAULT_NOPAGE
;
2857 #ifndef io_remap_pfn_range
2858 static inline int io_remap_pfn_range(struct vm_area_struct
*vma
,
2859 unsigned long addr
, unsigned long pfn
,
2860 unsigned long size
, pgprot_t prot
)
2862 return remap_pfn_range(vma
, addr
, pfn
, size
, pgprot_decrypted(prot
));
2866 static inline vm_fault_t
vmf_error(int err
)
2869 return VM_FAULT_OOM
;
2870 return VM_FAULT_SIGBUS
;
2873 struct page
*follow_page(struct vm_area_struct
*vma
, unsigned long address
,
2874 unsigned int foll_flags
);
2876 #define FOLL_WRITE 0x01 /* check pte is writable */
2877 #define FOLL_TOUCH 0x02 /* mark page accessed */
2878 #define FOLL_GET 0x04 /* do get_page on page */
2879 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2880 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2881 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2882 * and return without waiting upon it */
2883 #define FOLL_POPULATE 0x40 /* fault in page */
2884 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2885 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2886 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2887 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2888 #define FOLL_MLOCK 0x1000 /* lock present pages */
2889 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2890 #define FOLL_COW 0x4000 /* internal GUP flag */
2891 #define FOLL_ANON 0x8000 /* don't do file mappings */
2892 #define FOLL_LONGTERM 0x10000 /* mapping lifetime is indefinite: see below */
2893 #define FOLL_SPLIT_PMD 0x20000 /* split huge pmd before returning */
2894 #define FOLL_PIN 0x40000 /* pages must be released via unpin_user_page */
2895 #define FOLL_FAST_ONLY 0x80000 /* gup_fast: prevent fall-back to slow gup */
2898 * FOLL_PIN and FOLL_LONGTERM may be used in various combinations with each
2899 * other. Here is what they mean, and how to use them:
2901 * FOLL_LONGTERM indicates that the page will be held for an indefinite time
2902 * period _often_ under userspace control. This is in contrast to
2903 * iov_iter_get_pages(), whose usages are transient.
2905 * FIXME: For pages which are part of a filesystem, mappings are subject to the
2906 * lifetime enforced by the filesystem and we need guarantees that longterm
2907 * users like RDMA and V4L2 only establish mappings which coordinate usage with
2908 * the filesystem. Ideas for this coordination include revoking the longterm
2909 * pin, delaying writeback, bounce buffer page writeback, etc. As FS DAX was
2910 * added after the problem with filesystems was found FS DAX VMAs are
2911 * specifically failed. Filesystem pages are still subject to bugs and use of
2912 * FOLL_LONGTERM should be avoided on those pages.
2914 * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call.
2915 * Currently only get_user_pages() and get_user_pages_fast() support this flag
2916 * and calls to get_user_pages_[un]locked are specifically not allowed. This
2917 * is due to an incompatibility with the FS DAX check and
2918 * FAULT_FLAG_ALLOW_RETRY.
2920 * In the CMA case: long term pins in a CMA region would unnecessarily fragment
2921 * that region. And so, CMA attempts to migrate the page before pinning, when
2922 * FOLL_LONGTERM is specified.
2924 * FOLL_PIN indicates that a special kind of tracking (not just page->_refcount,
2925 * but an additional pin counting system) will be invoked. This is intended for
2926 * anything that gets a page reference and then touches page data (for example,
2927 * Direct IO). This lets the filesystem know that some non-file-system entity is
2928 * potentially changing the pages' data. In contrast to FOLL_GET (whose pages
2929 * are released via put_page()), FOLL_PIN pages must be released, ultimately, by
2930 * a call to unpin_user_page().
2932 * FOLL_PIN is similar to FOLL_GET: both of these pin pages. They use different
2933 * and separate refcounting mechanisms, however, and that means that each has
2934 * its own acquire and release mechanisms:
2936 * FOLL_GET: get_user_pages*() to acquire, and put_page() to release.
2938 * FOLL_PIN: pin_user_pages*() to acquire, and unpin_user_pages to release.
2940 * FOLL_PIN and FOLL_GET are mutually exclusive for a given function call.
2941 * (The underlying pages may experience both FOLL_GET-based and FOLL_PIN-based
2942 * calls applied to them, and that's perfectly OK. This is a constraint on the
2943 * callers, not on the pages.)
2945 * FOLL_PIN should be set internally by the pin_user_pages*() APIs, never
2946 * directly by the caller. That's in order to help avoid mismatches when
2947 * releasing pages: get_user_pages*() pages must be released via put_page(),
2948 * while pin_user_pages*() pages must be released via unpin_user_page().
2950 * Please see Documentation/core-api/pin_user_pages.rst for more information.
2953 static inline int vm_fault_to_errno(vm_fault_t vm_fault
, int foll_flags
)
2955 if (vm_fault
& VM_FAULT_OOM
)
2957 if (vm_fault
& (VM_FAULT_HWPOISON
| VM_FAULT_HWPOISON_LARGE
))
2958 return (foll_flags
& FOLL_HWPOISON
) ? -EHWPOISON
: -EFAULT
;
2959 if (vm_fault
& (VM_FAULT_SIGBUS
| VM_FAULT_SIGSEGV
))
2964 typedef int (*pte_fn_t
)(pte_t
*pte
, unsigned long addr
, void *data
);
2965 extern int apply_to_page_range(struct mm_struct
*mm
, unsigned long address
,
2966 unsigned long size
, pte_fn_t fn
, void *data
);
2967 extern int apply_to_existing_page_range(struct mm_struct
*mm
,
2968 unsigned long address
, unsigned long size
,
2969 pte_fn_t fn
, void *data
);
2971 extern void init_mem_debugging_and_hardening(void);
2972 #ifdef CONFIG_PAGE_POISONING
2973 extern void __kernel_poison_pages(struct page
*page
, int numpages
);
2974 extern void __kernel_unpoison_pages(struct page
*page
, int numpages
);
2975 extern bool _page_poisoning_enabled_early
;
2976 DECLARE_STATIC_KEY_FALSE(_page_poisoning_enabled
);
2977 static inline bool page_poisoning_enabled(void)
2979 return _page_poisoning_enabled_early
;
2982 * For use in fast paths after init_mem_debugging() has run, or when a
2983 * false negative result is not harmful when called too early.
2985 static inline bool page_poisoning_enabled_static(void)
2987 return static_branch_unlikely(&_page_poisoning_enabled
);
2989 static inline void kernel_poison_pages(struct page
*page
, int numpages
)
2991 if (page_poisoning_enabled_static())
2992 __kernel_poison_pages(page
, numpages
);
2994 static inline void kernel_unpoison_pages(struct page
*page
, int numpages
)
2996 if (page_poisoning_enabled_static())
2997 __kernel_unpoison_pages(page
, numpages
);
3000 static inline bool page_poisoning_enabled(void) { return false; }
3001 static inline bool page_poisoning_enabled_static(void) { return false; }
3002 static inline void __kernel_poison_pages(struct page
*page
, int nunmpages
) { }
3003 static inline void kernel_poison_pages(struct page
*page
, int numpages
) { }
3004 static inline void kernel_unpoison_pages(struct page
*page
, int numpages
) { }
3007 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_ALLOC_DEFAULT_ON
, init_on_alloc
);
3008 static inline bool want_init_on_alloc(gfp_t flags
)
3010 if (static_branch_maybe(CONFIG_INIT_ON_ALLOC_DEFAULT_ON
,
3013 return flags
& __GFP_ZERO
;
3016 DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_FREE_DEFAULT_ON
, init_on_free
);
3017 static inline bool want_init_on_free(void)
3019 return static_branch_maybe(CONFIG_INIT_ON_FREE_DEFAULT_ON
,
3023 extern bool _debug_pagealloc_enabled_early
;
3024 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled
);
3026 static inline bool debug_pagealloc_enabled(void)
3028 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC
) &&
3029 _debug_pagealloc_enabled_early
;
3033 * For use in fast paths after init_debug_pagealloc() has run, or when a
3034 * false negative result is not harmful when called too early.
3036 static inline bool debug_pagealloc_enabled_static(void)
3038 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC
))
3041 return static_branch_unlikely(&_debug_pagealloc_enabled
);
3044 #ifdef CONFIG_DEBUG_PAGEALLOC
3046 * To support DEBUG_PAGEALLOC architecture must ensure that
3047 * __kernel_map_pages() never fails
3049 extern void __kernel_map_pages(struct page
*page
, int numpages
, int enable
);
3051 static inline void debug_pagealloc_map_pages(struct page
*page
, int numpages
)
3053 if (debug_pagealloc_enabled_static())
3054 __kernel_map_pages(page
, numpages
, 1);
3057 static inline void debug_pagealloc_unmap_pages(struct page
*page
, int numpages
)
3059 if (debug_pagealloc_enabled_static())
3060 __kernel_map_pages(page
, numpages
, 0);
3062 #else /* CONFIG_DEBUG_PAGEALLOC */
3063 static inline void debug_pagealloc_map_pages(struct page
*page
, int numpages
) {}
3064 static inline void debug_pagealloc_unmap_pages(struct page
*page
, int numpages
) {}
3065 #endif /* CONFIG_DEBUG_PAGEALLOC */
3067 #ifdef __HAVE_ARCH_GATE_AREA
3068 extern struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
);
3069 extern int in_gate_area_no_mm(unsigned long addr
);
3070 extern int in_gate_area(struct mm_struct
*mm
, unsigned long addr
);
3072 static inline struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
)
3076 static inline int in_gate_area_no_mm(unsigned long addr
) { return 0; }
3077 static inline int in_gate_area(struct mm_struct
*mm
, unsigned long addr
)
3081 #endif /* __HAVE_ARCH_GATE_AREA */
3083 extern bool process_shares_mm(struct task_struct
*p
, struct mm_struct
*mm
);
3085 #ifdef CONFIG_SYSCTL
3086 extern int sysctl_drop_caches
;
3087 int drop_caches_sysctl_handler(struct ctl_table
*, int, void *, size_t *,
3091 void drop_slab(void);
3092 void drop_slab_node(int nid
);
3095 #define randomize_va_space 0
3097 extern int randomize_va_space
;
3100 const char * arch_vma_name(struct vm_area_struct
*vma
);
3102 void print_vma_addr(char *prefix
, unsigned long rip
);
3104 static inline void print_vma_addr(char *prefix
, unsigned long rip
)
3109 int vmemmap_remap_free(unsigned long start
, unsigned long end
,
3110 unsigned long reuse
);
3111 int vmemmap_remap_alloc(unsigned long start
, unsigned long end
,
3112 unsigned long reuse
, gfp_t gfp_mask
);
3114 void *sparse_buffer_alloc(unsigned long size
);
3115 struct page
* __populate_section_memmap(unsigned long pfn
,
3116 unsigned long nr_pages
, int nid
, struct vmem_altmap
*altmap
);
3117 pgd_t
*vmemmap_pgd_populate(unsigned long addr
, int node
);
3118 p4d_t
*vmemmap_p4d_populate(pgd_t
*pgd
, unsigned long addr
, int node
);
3119 pud_t
*vmemmap_pud_populate(p4d_t
*p4d
, unsigned long addr
, int node
);
3120 pmd_t
*vmemmap_pmd_populate(pud_t
*pud
, unsigned long addr
, int node
);
3121 pte_t
*vmemmap_pte_populate(pmd_t
*pmd
, unsigned long addr
, int node
,
3122 struct vmem_altmap
*altmap
);
3123 void *vmemmap_alloc_block(unsigned long size
, int node
);
3125 void *vmemmap_alloc_block_buf(unsigned long size
, int node
,
3126 struct vmem_altmap
*altmap
);
3127 void vmemmap_verify(pte_t
*, int, unsigned long, unsigned long);
3128 int vmemmap_populate_basepages(unsigned long start
, unsigned long end
,
3129 int node
, struct vmem_altmap
*altmap
);
3130 int vmemmap_populate(unsigned long start
, unsigned long end
, int node
,
3131 struct vmem_altmap
*altmap
);
3132 void vmemmap_populate_print_last(void);
3133 #ifdef CONFIG_MEMORY_HOTPLUG
3134 void vmemmap_free(unsigned long start
, unsigned long end
,
3135 struct vmem_altmap
*altmap
);
3137 void register_page_bootmem_memmap(unsigned long section_nr
, struct page
*map
,
3138 unsigned long nr_pages
);
3141 MF_COUNT_INCREASED
= 1 << 0,
3142 MF_ACTION_REQUIRED
= 1 << 1,
3143 MF_MUST_KILL
= 1 << 2,
3144 MF_SOFT_OFFLINE
= 1 << 3,
3146 extern int memory_failure(unsigned long pfn
, int flags
);
3147 extern void memory_failure_queue(unsigned long pfn
, int flags
);
3148 extern void memory_failure_queue_kick(int cpu
);
3149 extern int unpoison_memory(unsigned long pfn
);
3150 extern int sysctl_memory_failure_early_kill
;
3151 extern int sysctl_memory_failure_recovery
;
3152 extern void shake_page(struct page
*p
);
3153 extern atomic_long_t num_poisoned_pages __read_mostly
;
3154 extern int soft_offline_page(unsigned long pfn
, int flags
);
3158 * Error handlers for various types of pages.
3161 MF_IGNORED
, /* Error: cannot be handled */
3162 MF_FAILED
, /* Error: handling failed */
3163 MF_DELAYED
, /* Will be handled later */
3164 MF_RECOVERED
, /* Successfully recovered */
3167 enum mf_action_page_type
{
3169 MF_MSG_KERNEL_HIGH_ORDER
,
3171 MF_MSG_DIFFERENT_COMPOUND
,
3172 MF_MSG_POISONED_HUGE
,
3175 MF_MSG_NON_PMD_HUGE
,
3176 MF_MSG_UNMAP_FAILED
,
3177 MF_MSG_DIRTY_SWAPCACHE
,
3178 MF_MSG_CLEAN_SWAPCACHE
,
3179 MF_MSG_DIRTY_MLOCKED_LRU
,
3180 MF_MSG_CLEAN_MLOCKED_LRU
,
3181 MF_MSG_DIRTY_UNEVICTABLE_LRU
,
3182 MF_MSG_CLEAN_UNEVICTABLE_LRU
,
3185 MF_MSG_TRUNCATED_LRU
,
3193 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
3194 extern void clear_huge_page(struct page
*page
,
3195 unsigned long addr_hint
,
3196 unsigned int pages_per_huge_page
);
3197 extern void copy_user_huge_page(struct page
*dst
, struct page
*src
,
3198 unsigned long addr_hint
,
3199 struct vm_area_struct
*vma
,
3200 unsigned int pages_per_huge_page
);
3201 extern long copy_huge_page_from_user(struct page
*dst_page
,
3202 const void __user
*usr_src
,
3203 unsigned int pages_per_huge_page
,
3204 bool allow_pagefault
);
3207 * vma_is_special_huge - Are transhuge page-table entries considered special?
3208 * @vma: Pointer to the struct vm_area_struct to consider
3210 * Whether transhuge page-table entries are considered "special" following
3211 * the definition in vm_normal_page().
3213 * Return: true if transhuge page-table entries should be considered special,
3216 static inline bool vma_is_special_huge(const struct vm_area_struct
*vma
)
3218 return vma_is_dax(vma
) || (vma
->vm_file
&&
3219 (vma
->vm_flags
& (VM_PFNMAP
| VM_MIXEDMAP
)));
3222 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
3224 #ifdef CONFIG_DEBUG_PAGEALLOC
3225 extern unsigned int _debug_guardpage_minorder
;
3226 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled
);
3228 static inline unsigned int debug_guardpage_minorder(void)
3230 return _debug_guardpage_minorder
;
3233 static inline bool debug_guardpage_enabled(void)
3235 return static_branch_unlikely(&_debug_guardpage_enabled
);
3238 static inline bool page_is_guard(struct page
*page
)
3240 if (!debug_guardpage_enabled())
3243 return PageGuard(page
);
3246 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
3247 static inline bool debug_guardpage_enabled(void) { return false; }
3248 static inline bool page_is_guard(struct page
*page
) { return false; }
3249 #endif /* CONFIG_DEBUG_PAGEALLOC */
3251 #if MAX_NUMNODES > 1
3252 void __init
setup_nr_node_ids(void);
3254 static inline void setup_nr_node_ids(void) {}
3257 extern int memcmp_pages(struct page
*page1
, struct page
*page2
);
3259 static inline int pages_identical(struct page
*page1
, struct page
*page2
)
3261 return !memcmp_pages(page1
, page2
);
3264 #ifdef CONFIG_MAPPING_DIRTY_HELPERS
3265 unsigned long clean_record_shared_mapping_range(struct address_space
*mapping
,
3266 pgoff_t first_index
, pgoff_t nr
,
3267 pgoff_t bitmap_pgoff
,
3268 unsigned long *bitmap
,
3272 unsigned long wp_shared_mapping_range(struct address_space
*mapping
,
3273 pgoff_t first_index
, pgoff_t nr
);
3276 extern int sysctl_nr_trim_pages
;
3278 #ifdef CONFIG_PRINTK
3279 void mem_dump_obj(void *object
);
3281 static inline void mem_dump_obj(void *object
) {}
3285 * seal_check_future_write - Check for F_SEAL_FUTURE_WRITE flag and handle it
3286 * @seals: the seals to check
3287 * @vma: the vma to operate on
3289 * Check whether F_SEAL_FUTURE_WRITE is set; if so, do proper check/handling on
3290 * the vma flags. Return 0 if check pass, or <0 for errors.
3292 static inline int seal_check_future_write(int seals
, struct vm_area_struct
*vma
)
3294 if (seals
& F_SEAL_FUTURE_WRITE
) {
3296 * New PROT_WRITE and MAP_SHARED mmaps are not allowed when
3297 * "future write" seal active.
3299 if ((vma
->vm_flags
& VM_SHARED
) && (vma
->vm_flags
& VM_WRITE
))
3303 * Since an F_SEAL_FUTURE_WRITE sealed memfd can be mapped as
3304 * MAP_SHARED and read-only, take care to not allow mprotect to
3305 * revert protections on such mappings. Do this only for shared
3306 * mappings. For private mappings, don't need to mask
3307 * VM_MAYWRITE as we still want them to be COW-writable.
3309 if (vma
->vm_flags
& VM_SHARED
)
3310 vma
->vm_flags
&= ~(VM_MAYWRITE
);
3316 #endif /* __KERNEL__ */
3317 #endif /* _LINUX_MM_H */