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/range.h>
19 #include <linux/pfn.h>
20 #include <linux/percpu-refcount.h>
21 #include <linux/bit_spinlock.h>
22 #include <linux/shrinker.h>
23 #include <linux/resource.h>
24 #include <linux/page_ext.h>
25 #include <linux/err.h>
26 #include <linux/page_ref.h>
27 #include <linux/memremap.h>
28 #include <linux/overflow.h>
29 #include <linux/sizes.h>
33 struct anon_vma_chain
;
36 struct writeback_control
;
39 void init_mm_internals(void);
41 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
42 extern unsigned long max_mapnr
;
44 static inline void set_max_mapnr(unsigned long limit
)
49 static inline void set_max_mapnr(unsigned long limit
) { }
52 extern atomic_long_t _totalram_pages
;
53 static inline unsigned long totalram_pages(void)
55 return (unsigned long)atomic_long_read(&_totalram_pages
);
58 static inline void totalram_pages_inc(void)
60 atomic_long_inc(&_totalram_pages
);
63 static inline void totalram_pages_dec(void)
65 atomic_long_dec(&_totalram_pages
);
68 static inline void totalram_pages_add(long count
)
70 atomic_long_add(count
, &_totalram_pages
);
73 static inline void totalram_pages_set(long val
)
75 atomic_long_set(&_totalram_pages
, val
);
78 extern void * high_memory
;
79 extern int page_cluster
;
82 extern int sysctl_legacy_va_layout
;
84 #define sysctl_legacy_va_layout 0
87 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
88 extern const int mmap_rnd_bits_min
;
89 extern const int mmap_rnd_bits_max
;
90 extern int mmap_rnd_bits __read_mostly
;
92 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
93 extern const int mmap_rnd_compat_bits_min
;
94 extern const int mmap_rnd_compat_bits_max
;
95 extern int mmap_rnd_compat_bits __read_mostly
;
99 #include <asm/pgtable.h>
100 #include <asm/processor.h>
103 * Architectures that support memory tagging (assigning tags to memory regions,
104 * embedding these tags into addresses that point to these memory regions, and
105 * checking that the memory and the pointer tags match on memory accesses)
106 * redefine this macro to strip tags from pointers.
107 * It's defined as noop for arcitectures that don't support memory tagging.
109 #ifndef untagged_addr
110 #define untagged_addr(addr) (addr)
114 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
118 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
122 #define lm_alias(x) __va(__pa_symbol(x))
126 * To prevent common memory management code establishing
127 * a zero page mapping on a read fault.
128 * This macro should be defined within <asm/pgtable.h>.
129 * s390 does this to prevent multiplexing of hardware bits
130 * related to the physical page in case of virtualization.
132 #ifndef mm_forbids_zeropage
133 #define mm_forbids_zeropage(X) (0)
137 * On some architectures it is expensive to call memset() for small sizes.
138 * If an architecture decides to implement their own version of
139 * mm_zero_struct_page they should wrap the defines below in a #ifndef and
140 * define their own version of this macro in <asm/pgtable.h>
142 #if BITS_PER_LONG == 64
143 /* This function must be updated when the size of struct page grows above 80
144 * or reduces below 56. The idea that compiler optimizes out switch()
145 * statement, and only leaves move/store instructions. Also the compiler can
146 * combine write statments if they are both assignments and can be reordered,
147 * this can result in several of the writes here being dropped.
149 #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
150 static inline void __mm_zero_struct_page(struct page
*page
)
152 unsigned long *_pp
= (void *)page
;
154 /* Check that struct page is either 56, 64, 72, or 80 bytes */
155 BUILD_BUG_ON(sizeof(struct page
) & 7);
156 BUILD_BUG_ON(sizeof(struct page
) < 56);
157 BUILD_BUG_ON(sizeof(struct page
) > 80);
159 switch (sizeof(struct page
)) {
161 _pp
[9] = 0; /* fallthrough */
163 _pp
[8] = 0; /* fallthrough */
165 _pp
[7] = 0; /* fallthrough */
177 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
181 * Default maximum number of active map areas, this limits the number of vmas
182 * per mm struct. Users can overwrite this number by sysctl but there is a
185 * When a program's coredump is generated as ELF format, a section is created
186 * per a vma. In ELF, the number of sections is represented in unsigned short.
187 * This means the number of sections should be smaller than 65535 at coredump.
188 * Because the kernel adds some informative sections to a image of program at
189 * generating coredump, we need some margin. The number of extra sections is
190 * 1-3 now and depends on arch. We use "5" as safe margin, here.
192 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
193 * not a hard limit any more. Although some userspace tools can be surprised by
196 #define MAPCOUNT_ELF_CORE_MARGIN (5)
197 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
199 extern int sysctl_max_map_count
;
201 extern unsigned long sysctl_user_reserve_kbytes
;
202 extern unsigned long sysctl_admin_reserve_kbytes
;
204 extern int sysctl_overcommit_memory
;
205 extern int sysctl_overcommit_ratio
;
206 extern unsigned long sysctl_overcommit_kbytes
;
208 extern int overcommit_ratio_handler(struct ctl_table
*, int, void __user
*,
210 extern int overcommit_kbytes_handler(struct ctl_table
*, int, void __user
*,
213 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
215 /* to align the pointer to the (next) page boundary */
216 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
218 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
219 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
221 #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
224 * Linux kernel virtual memory manager primitives.
225 * The idea being to have a "virtual" mm in the same way
226 * we have a virtual fs - giving a cleaner interface to the
227 * mm details, and allowing different kinds of memory mappings
228 * (from shared memory to executable loading to arbitrary
232 struct vm_area_struct
*vm_area_alloc(struct mm_struct
*);
233 struct vm_area_struct
*vm_area_dup(struct vm_area_struct
*);
234 void vm_area_free(struct vm_area_struct
*);
237 extern struct rb_root nommu_region_tree
;
238 extern struct rw_semaphore nommu_region_sem
;
240 extern unsigned int kobjsize(const void *objp
);
244 * vm_flags in vm_area_struct, see mm_types.h.
245 * When changing, update also include/trace/events/mmflags.h
247 #define VM_NONE 0x00000000
249 #define VM_READ 0x00000001 /* currently active flags */
250 #define VM_WRITE 0x00000002
251 #define VM_EXEC 0x00000004
252 #define VM_SHARED 0x00000008
254 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
255 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
256 #define VM_MAYWRITE 0x00000020
257 #define VM_MAYEXEC 0x00000040
258 #define VM_MAYSHARE 0x00000080
260 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
261 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
262 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
263 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
264 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
266 #define VM_LOCKED 0x00002000
267 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
269 /* Used by sys_madvise() */
270 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
271 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
273 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
274 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
275 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
276 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
277 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
278 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
279 #define VM_SYNC 0x00800000 /* Synchronous page faults */
280 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
281 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
282 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
284 #ifdef CONFIG_MEM_SOFT_DIRTY
285 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
287 # define VM_SOFTDIRTY 0
290 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
291 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
292 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
293 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
295 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
296 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
297 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
298 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
299 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
300 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
301 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
302 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
303 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
304 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
305 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
306 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
308 #ifdef CONFIG_ARCH_HAS_PKEYS
309 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
310 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
311 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
312 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
313 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
315 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
317 # define VM_PKEY_BIT4 0
319 #endif /* CONFIG_ARCH_HAS_PKEYS */
321 #if defined(CONFIG_X86)
322 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
323 #elif defined(CONFIG_PPC)
324 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
325 #elif defined(CONFIG_PARISC)
326 # define VM_GROWSUP VM_ARCH_1
327 #elif defined(CONFIG_IA64)
328 # define VM_GROWSUP VM_ARCH_1
329 #elif defined(CONFIG_SPARC64)
330 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
331 # define VM_ARCH_CLEAR VM_SPARC_ADI
332 #elif !defined(CONFIG_MMU)
333 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
336 #if defined(CONFIG_X86_INTEL_MPX)
337 /* MPX specific bounds table or bounds directory */
338 # define VM_MPX VM_HIGH_ARCH_4
340 # define VM_MPX VM_NONE
344 # define VM_GROWSUP VM_NONE
347 /* Bits set in the VMA until the stack is in its final location */
348 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
350 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
351 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
354 #ifdef CONFIG_STACK_GROWSUP
355 #define VM_STACK VM_GROWSUP
357 #define VM_STACK VM_GROWSDOWN
360 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
363 * Special vmas that are non-mergable, non-mlock()able.
364 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
366 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
368 /* This mask defines which mm->def_flags a process can inherit its parent */
369 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
371 /* This mask is used to clear all the VMA flags used by mlock */
372 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
374 /* Arch-specific flags to clear when updating VM flags on protection change */
375 #ifndef VM_ARCH_CLEAR
376 # define VM_ARCH_CLEAR VM_NONE
378 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
381 * mapping from the currently active vm_flags protection bits (the
382 * low four bits) to a page protection mask..
384 extern pgprot_t protection_map
[16];
386 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
387 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
388 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
389 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
390 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
391 #define FAULT_FLAG_TRIED 0x20 /* Second try */
392 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
393 #define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */
394 #define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */
396 #define FAULT_FLAG_TRACE \
397 { FAULT_FLAG_WRITE, "WRITE" }, \
398 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
399 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
400 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
401 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
402 { FAULT_FLAG_TRIED, "TRIED" }, \
403 { FAULT_FLAG_USER, "USER" }, \
404 { FAULT_FLAG_REMOTE, "REMOTE" }, \
405 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }
408 * vm_fault is filled by the the pagefault handler and passed to the vma's
409 * ->fault function. The vma's ->fault is responsible for returning a bitmask
410 * of VM_FAULT_xxx flags that give details about how the fault was handled.
412 * MM layer fills up gfp_mask for page allocations but fault handler might
413 * alter it if its implementation requires a different allocation context.
415 * pgoff should be used in favour of virtual_address, if possible.
418 struct vm_area_struct
*vma
; /* Target VMA */
419 unsigned int flags
; /* FAULT_FLAG_xxx flags */
420 gfp_t gfp_mask
; /* gfp mask to be used for allocations */
421 pgoff_t pgoff
; /* Logical page offset based on vma */
422 unsigned long address
; /* Faulting virtual address */
423 pmd_t
*pmd
; /* Pointer to pmd entry matching
425 pud_t
*pud
; /* Pointer to pud entry matching
428 pte_t orig_pte
; /* Value of PTE at the time of fault */
430 struct page
*cow_page
; /* Page handler may use for COW fault */
431 struct mem_cgroup
*memcg
; /* Cgroup cow_page belongs to */
432 struct page
*page
; /* ->fault handlers should return a
433 * page here, unless VM_FAULT_NOPAGE
434 * is set (which is also implied by
437 /* These three entries are valid only while holding ptl lock */
438 pte_t
*pte
; /* Pointer to pte entry matching
439 * the 'address'. NULL if the page
440 * table hasn't been allocated.
442 spinlock_t
*ptl
; /* Page table lock.
443 * Protects pte page table if 'pte'
444 * is not NULL, otherwise pmd.
446 pgtable_t prealloc_pte
; /* Pre-allocated pte page table.
447 * vm_ops->map_pages() calls
448 * alloc_set_pte() from atomic context.
449 * do_fault_around() pre-allocates
450 * page table to avoid allocation from
455 /* page entry size for vm->huge_fault() */
456 enum page_entry_size
{
463 * These are the virtual MM functions - opening of an area, closing and
464 * unmapping it (needed to keep files on disk up-to-date etc), pointer
465 * to the functions called when a no-page or a wp-page exception occurs.
467 struct vm_operations_struct
{
468 void (*open
)(struct vm_area_struct
* area
);
469 void (*close
)(struct vm_area_struct
* area
);
470 int (*split
)(struct vm_area_struct
* area
, unsigned long addr
);
471 int (*mremap
)(struct vm_area_struct
* area
);
472 vm_fault_t (*fault
)(struct vm_fault
*vmf
);
473 vm_fault_t (*huge_fault
)(struct vm_fault
*vmf
,
474 enum page_entry_size pe_size
);
475 void (*map_pages
)(struct vm_fault
*vmf
,
476 pgoff_t start_pgoff
, pgoff_t end_pgoff
);
477 unsigned long (*pagesize
)(struct vm_area_struct
* area
);
479 /* notification that a previously read-only page is about to become
480 * writable, if an error is returned it will cause a SIGBUS */
481 vm_fault_t (*page_mkwrite
)(struct vm_fault
*vmf
);
483 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
484 vm_fault_t (*pfn_mkwrite
)(struct vm_fault
*vmf
);
486 /* called by access_process_vm when get_user_pages() fails, typically
487 * for use by special VMAs that can switch between memory and hardware
489 int (*access
)(struct vm_area_struct
*vma
, unsigned long addr
,
490 void *buf
, int len
, int write
);
492 /* Called by the /proc/PID/maps code to ask the vma whether it
493 * has a special name. Returning non-NULL will also cause this
494 * vma to be dumped unconditionally. */
495 const char *(*name
)(struct vm_area_struct
*vma
);
499 * set_policy() op must add a reference to any non-NULL @new mempolicy
500 * to hold the policy upon return. Caller should pass NULL @new to
501 * remove a policy and fall back to surrounding context--i.e. do not
502 * install a MPOL_DEFAULT policy, nor the task or system default
505 int (*set_policy
)(struct vm_area_struct
*vma
, struct mempolicy
*new);
508 * get_policy() op must add reference [mpol_get()] to any policy at
509 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
510 * in mm/mempolicy.c will do this automatically.
511 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
512 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
513 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
514 * must return NULL--i.e., do not "fallback" to task or system default
517 struct mempolicy
*(*get_policy
)(struct vm_area_struct
*vma
,
521 * Called by vm_normal_page() for special PTEs to find the
522 * page for @addr. This is useful if the default behavior
523 * (using pte_page()) would not find the correct page.
525 struct page
*(*find_special_page
)(struct vm_area_struct
*vma
,
529 static inline void vma_init(struct vm_area_struct
*vma
, struct mm_struct
*mm
)
531 static const struct vm_operations_struct dummy_vm_ops
= {};
533 memset(vma
, 0, sizeof(*vma
));
535 vma
->vm_ops
= &dummy_vm_ops
;
536 INIT_LIST_HEAD(&vma
->anon_vma_chain
);
539 static inline void vma_set_anonymous(struct vm_area_struct
*vma
)
544 static inline bool vma_is_anonymous(struct vm_area_struct
*vma
)
551 * The vma_is_shmem is not inline because it is used only by slow
552 * paths in userfault.
554 bool vma_is_shmem(struct vm_area_struct
*vma
);
556 static inline bool vma_is_shmem(struct vm_area_struct
*vma
) { return false; }
559 int vma_is_stack_for_current(struct vm_area_struct
*vma
);
561 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
562 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
567 #if !defined(CONFIG_ARCH_HAS_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
568 static inline int pmd_devmap(pmd_t pmd
)
572 static inline int pud_devmap(pud_t pud
)
576 static inline int pgd_devmap(pgd_t pgd
)
583 * FIXME: take this include out, include page-flags.h in
584 * files which need it (119 of them)
586 #include <linux/page-flags.h>
587 #include <linux/huge_mm.h>
590 * Methods to modify the page usage count.
592 * What counts for a page usage:
593 * - cache mapping (page->mapping)
594 * - private data (page->private)
595 * - page mapped in a task's page tables, each mapping
596 * is counted separately
598 * Also, many kernel routines increase the page count before a critical
599 * routine so they can be sure the page doesn't go away from under them.
603 * Drop a ref, return true if the refcount fell to zero (the page has no users)
605 static inline int put_page_testzero(struct page
*page
)
607 VM_BUG_ON_PAGE(page_ref_count(page
) == 0, page
);
608 return page_ref_dec_and_test(page
);
612 * Try to grab a ref unless the page has a refcount of zero, return false if
614 * This can be called when MMU is off so it must not access
615 * any of the virtual mappings.
617 static inline int get_page_unless_zero(struct page
*page
)
619 return page_ref_add_unless(page
, 1, 0);
622 extern int page_is_ram(unsigned long pfn
);
630 int region_intersects(resource_size_t offset
, size_t size
, unsigned long flags
,
633 /* Support for virtually mapped pages */
634 struct page
*vmalloc_to_page(const void *addr
);
635 unsigned long vmalloc_to_pfn(const void *addr
);
638 * Determine if an address is within the vmalloc range
640 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
641 * is no special casing required.
643 static inline bool is_vmalloc_addr(const void *x
)
646 unsigned long addr
= (unsigned long)x
;
648 return addr
>= VMALLOC_START
&& addr
< VMALLOC_END
;
654 #ifndef is_ioremap_addr
655 #define is_ioremap_addr(x) is_vmalloc_addr(x)
659 extern int is_vmalloc_or_module_addr(const void *x
);
661 static inline int is_vmalloc_or_module_addr(const void *x
)
667 extern void *kvmalloc_node(size_t size
, gfp_t flags
, int node
);
668 static inline void *kvmalloc(size_t size
, gfp_t flags
)
670 return kvmalloc_node(size
, flags
, NUMA_NO_NODE
);
672 static inline void *kvzalloc_node(size_t size
, gfp_t flags
, int node
)
674 return kvmalloc_node(size
, flags
| __GFP_ZERO
, node
);
676 static inline void *kvzalloc(size_t size
, gfp_t flags
)
678 return kvmalloc(size
, flags
| __GFP_ZERO
);
681 static inline void *kvmalloc_array(size_t n
, size_t size
, gfp_t flags
)
685 if (unlikely(check_mul_overflow(n
, size
, &bytes
)))
688 return kvmalloc(bytes
, flags
);
691 static inline void *kvcalloc(size_t n
, size_t size
, gfp_t flags
)
693 return kvmalloc_array(n
, size
, flags
| __GFP_ZERO
);
696 extern void kvfree(const void *addr
);
697 extern void kvfree_sensitive(const void *addr
, size_t len
);
700 * Mapcount of compound page as a whole, does not include mapped sub-pages.
702 * Must be called only for compound pages or any their tail sub-pages.
704 static inline int compound_mapcount(struct page
*page
)
706 VM_BUG_ON_PAGE(!PageCompound(page
), page
);
707 page
= compound_head(page
);
708 return atomic_read(compound_mapcount_ptr(page
)) + 1;
712 * The atomic page->_mapcount, starts from -1: so that transitions
713 * both from it and to it can be tracked, using atomic_inc_and_test
714 * and atomic_add_negative(-1).
716 static inline void page_mapcount_reset(struct page
*page
)
718 atomic_set(&(page
)->_mapcount
, -1);
721 int __page_mapcount(struct page
*page
);
724 * Mapcount of 0-order page; when compound sub-page, includes
725 * compound_mapcount().
727 * Result is undefined for pages which cannot be mapped into userspace.
728 * For example SLAB or special types of pages. See function page_has_type().
729 * They use this place in struct page differently.
731 static inline int page_mapcount(struct page
*page
)
733 if (unlikely(PageCompound(page
)))
734 return __page_mapcount(page
);
735 return atomic_read(&page
->_mapcount
) + 1;
738 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
739 int total_mapcount(struct page
*page
);
740 int page_trans_huge_mapcount(struct page
*page
, int *total_mapcount
);
742 static inline int total_mapcount(struct page
*page
)
744 return page_mapcount(page
);
746 static inline int page_trans_huge_mapcount(struct page
*page
,
749 int mapcount
= page_mapcount(page
);
751 *total_mapcount
= mapcount
;
756 static inline struct page
*virt_to_head_page(const void *x
)
758 struct page
*page
= virt_to_page(x
);
760 return compound_head(page
);
763 void __put_page(struct page
*page
);
765 void put_pages_list(struct list_head
*pages
);
767 void split_page(struct page
*page
, unsigned int order
);
770 * Compound pages have a destructor function. Provide a
771 * prototype for that function and accessor functions.
772 * These are _only_ valid on the head of a compound page.
774 typedef void compound_page_dtor(struct page
*);
776 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
777 enum compound_dtor_id
{
780 #ifdef CONFIG_HUGETLB_PAGE
783 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
788 extern compound_page_dtor
* const compound_page_dtors
[];
790 static inline void set_compound_page_dtor(struct page
*page
,
791 enum compound_dtor_id compound_dtor
)
793 VM_BUG_ON_PAGE(compound_dtor
>= NR_COMPOUND_DTORS
, page
);
794 page
[1].compound_dtor
= compound_dtor
;
797 static inline compound_page_dtor
*get_compound_page_dtor(struct page
*page
)
799 VM_BUG_ON_PAGE(page
[1].compound_dtor
>= NR_COMPOUND_DTORS
, page
);
800 return compound_page_dtors
[page
[1].compound_dtor
];
803 static inline unsigned int compound_order(struct page
*page
)
807 return page
[1].compound_order
;
810 static inline void set_compound_order(struct page
*page
, unsigned int order
)
812 page
[1].compound_order
= order
;
815 /* Returns the number of pages in this potentially compound page. */
816 static inline unsigned long compound_nr(struct page
*page
)
818 return 1UL << compound_order(page
);
821 /* Returns the number of bytes in this potentially compound page. */
822 static inline unsigned long page_size(struct page
*page
)
824 return PAGE_SIZE
<< compound_order(page
);
827 /* Returns the number of bits needed for the number of bytes in a page */
828 static inline unsigned int page_shift(struct page
*page
)
830 return PAGE_SHIFT
+ compound_order(page
);
833 void free_compound_page(struct page
*page
);
837 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
838 * servicing faults for write access. In the normal case, do always want
839 * pte_mkwrite. But get_user_pages can cause write faults for mappings
840 * that do not have writing enabled, when used by access_process_vm.
842 static inline pte_t
maybe_mkwrite(pte_t pte
, struct vm_area_struct
*vma
)
844 if (likely(vma
->vm_flags
& VM_WRITE
))
845 pte
= pte_mkwrite(pte
);
849 vm_fault_t
alloc_set_pte(struct vm_fault
*vmf
, struct mem_cgroup
*memcg
,
851 vm_fault_t
finish_fault(struct vm_fault
*vmf
);
852 vm_fault_t
finish_mkwrite_fault(struct vm_fault
*vmf
);
856 * Multiple processes may "see" the same page. E.g. for untouched
857 * mappings of /dev/null, all processes see the same page full of
858 * zeroes, and text pages of executables and shared libraries have
859 * only one copy in memory, at most, normally.
861 * For the non-reserved pages, page_count(page) denotes a reference count.
862 * page_count() == 0 means the page is free. page->lru is then used for
863 * freelist management in the buddy allocator.
864 * page_count() > 0 means the page has been allocated.
866 * Pages are allocated by the slab allocator in order to provide memory
867 * to kmalloc and kmem_cache_alloc. In this case, the management of the
868 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
869 * unless a particular usage is carefully commented. (the responsibility of
870 * freeing the kmalloc memory is the caller's, of course).
872 * A page may be used by anyone else who does a __get_free_page().
873 * In this case, page_count still tracks the references, and should only
874 * be used through the normal accessor functions. The top bits of page->flags
875 * and page->virtual store page management information, but all other fields
876 * are unused and could be used privately, carefully. The management of this
877 * page is the responsibility of the one who allocated it, and those who have
878 * subsequently been given references to it.
880 * The other pages (we may call them "pagecache pages") are completely
881 * managed by the Linux memory manager: I/O, buffers, swapping etc.
882 * The following discussion applies only to them.
884 * A pagecache page contains an opaque `private' member, which belongs to the
885 * page's address_space. Usually, this is the address of a circular list of
886 * the page's disk buffers. PG_private must be set to tell the VM to call
887 * into the filesystem to release these pages.
889 * A page may belong to an inode's memory mapping. In this case, page->mapping
890 * is the pointer to the inode, and page->index is the file offset of the page,
891 * in units of PAGE_SIZE.
893 * If pagecache pages are not associated with an inode, they are said to be
894 * anonymous pages. These may become associated with the swapcache, and in that
895 * case PG_swapcache is set, and page->private is an offset into the swapcache.
897 * In either case (swapcache or inode backed), the pagecache itself holds one
898 * reference to the page. Setting PG_private should also increment the
899 * refcount. The each user mapping also has a reference to the page.
901 * The pagecache pages are stored in a per-mapping radix tree, which is
902 * rooted at mapping->i_pages, and indexed by offset.
903 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
904 * lists, we instead now tag pages as dirty/writeback in the radix tree.
906 * All pagecache pages may be subject to I/O:
907 * - inode pages may need to be read from disk,
908 * - inode pages which have been modified and are MAP_SHARED may need
909 * to be written back to the inode on disk,
910 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
911 * modified may need to be swapped out to swap space and (later) to be read
916 * The zone field is never updated after free_area_init_core()
917 * sets it, so none of the operations on it need to be atomic.
920 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
921 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
922 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
923 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
924 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
925 #define KASAN_TAG_PGOFF (LAST_CPUPID_PGOFF - KASAN_TAG_WIDTH)
928 * Define the bit shifts to access each section. For non-existent
929 * sections we define the shift as 0; that plus a 0 mask ensures
930 * the compiler will optimise away reference to them.
932 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
933 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
934 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
935 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
936 #define KASAN_TAG_PGSHIFT (KASAN_TAG_PGOFF * (KASAN_TAG_WIDTH != 0))
938 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
939 #ifdef NODE_NOT_IN_PAGE_FLAGS
940 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
941 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
942 SECTIONS_PGOFF : ZONES_PGOFF)
944 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
945 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
946 NODES_PGOFF : ZONES_PGOFF)
949 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
951 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
952 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
955 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
956 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
957 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
958 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
959 #define KASAN_TAG_MASK ((1UL << KASAN_TAG_WIDTH) - 1)
960 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
962 static inline enum zone_type
page_zonenum(const struct page
*page
)
964 return (page
->flags
>> ZONES_PGSHIFT
) & ZONES_MASK
;
967 #ifdef CONFIG_ZONE_DEVICE
968 static inline bool is_zone_device_page(const struct page
*page
)
970 return page_zonenum(page
) == ZONE_DEVICE
;
972 extern void memmap_init_zone_device(struct zone
*, unsigned long,
973 unsigned long, struct dev_pagemap
*);
975 static inline bool is_zone_device_page(const struct page
*page
)
981 #ifdef CONFIG_DEV_PAGEMAP_OPS
982 void __put_devmap_managed_page(struct page
*page
);
983 DECLARE_STATIC_KEY_FALSE(devmap_managed_key
);
984 static inline bool put_devmap_managed_page(struct page
*page
)
986 if (!static_branch_unlikely(&devmap_managed_key
))
988 if (!is_zone_device_page(page
))
990 switch (page
->pgmap
->type
) {
991 case MEMORY_DEVICE_PRIVATE
:
992 case MEMORY_DEVICE_FS_DAX
:
993 __put_devmap_managed_page(page
);
1001 #else /* CONFIG_DEV_PAGEMAP_OPS */
1002 static inline bool put_devmap_managed_page(struct page
*page
)
1006 #endif /* CONFIG_DEV_PAGEMAP_OPS */
1008 static inline bool is_device_private_page(const struct page
*page
)
1010 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS
) &&
1011 IS_ENABLED(CONFIG_DEVICE_PRIVATE
) &&
1012 is_zone_device_page(page
) &&
1013 page
->pgmap
->type
== MEMORY_DEVICE_PRIVATE
;
1016 static inline bool is_pci_p2pdma_page(const struct page
*page
)
1018 return IS_ENABLED(CONFIG_DEV_PAGEMAP_OPS
) &&
1019 IS_ENABLED(CONFIG_PCI_P2PDMA
) &&
1020 is_zone_device_page(page
) &&
1021 page
->pgmap
->type
== MEMORY_DEVICE_PCI_P2PDMA
;
1024 /* 127: arbitrary random number, small enough to assemble well */
1025 #define page_ref_zero_or_close_to_overflow(page) \
1026 ((unsigned int) page_ref_count(page) + 127u <= 127u)
1028 static inline void get_page(struct page
*page
)
1030 page
= compound_head(page
);
1032 * Getting a normal page or the head of a compound page
1033 * requires to already have an elevated page->_refcount.
1035 VM_BUG_ON_PAGE(page_ref_zero_or_close_to_overflow(page
), page
);
1039 static inline __must_check
bool try_get_page(struct page
*page
)
1041 page
= compound_head(page
);
1042 if (WARN_ON_ONCE(page_ref_count(page
) <= 0))
1048 static inline void put_page(struct page
*page
)
1050 page
= compound_head(page
);
1053 * For devmap managed pages we need to catch refcount transition from
1054 * 2 to 1, when refcount reach one it means the page is free and we
1055 * need to inform the device driver through callback. See
1056 * include/linux/memremap.h and HMM for details.
1058 if (put_devmap_managed_page(page
))
1061 if (put_page_testzero(page
))
1066 * put_user_page() - release a gup-pinned page
1067 * @page: pointer to page to be released
1069 * Pages that were pinned via get_user_pages*() must be released via
1070 * either put_user_page(), or one of the put_user_pages*() routines
1071 * below. This is so that eventually, pages that are pinned via
1072 * get_user_pages*() can be separately tracked and uniquely handled. In
1073 * particular, interactions with RDMA and filesystems need special
1076 * put_user_page() and put_page() are not interchangeable, despite this early
1077 * implementation that makes them look the same. put_user_page() calls must
1078 * be perfectly matched up with get_user_page() calls.
1080 static inline void put_user_page(struct page
*page
)
1085 void put_user_pages_dirty_lock(struct page
**pages
, unsigned long npages
,
1088 void put_user_pages(struct page
**pages
, unsigned long npages
);
1090 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1091 #define SECTION_IN_PAGE_FLAGS
1095 * The identification function is mainly used by the buddy allocator for
1096 * determining if two pages could be buddies. We are not really identifying
1097 * the zone since we could be using the section number id if we do not have
1098 * node id available in page flags.
1099 * We only guarantee that it will return the same value for two combinable
1102 static inline int page_zone_id(struct page
*page
)
1104 return (page
->flags
>> ZONEID_PGSHIFT
) & ZONEID_MASK
;
1107 #ifdef NODE_NOT_IN_PAGE_FLAGS
1108 extern int page_to_nid(const struct page
*page
);
1110 static inline int page_to_nid(const struct page
*page
)
1112 struct page
*p
= (struct page
*)page
;
1114 return (PF_POISONED_CHECK(p
)->flags
>> NODES_PGSHIFT
) & NODES_MASK
;
1118 #ifdef CONFIG_NUMA_BALANCING
1119 static inline int cpu_pid_to_cpupid(int cpu
, int pid
)
1121 return ((cpu
& LAST__CPU_MASK
) << LAST__PID_SHIFT
) | (pid
& LAST__PID_MASK
);
1124 static inline int cpupid_to_pid(int cpupid
)
1126 return cpupid
& LAST__PID_MASK
;
1129 static inline int cpupid_to_cpu(int cpupid
)
1131 return (cpupid
>> LAST__PID_SHIFT
) & LAST__CPU_MASK
;
1134 static inline int cpupid_to_nid(int cpupid
)
1136 return cpu_to_node(cpupid_to_cpu(cpupid
));
1139 static inline bool cpupid_pid_unset(int cpupid
)
1141 return cpupid_to_pid(cpupid
) == (-1 & LAST__PID_MASK
);
1144 static inline bool cpupid_cpu_unset(int cpupid
)
1146 return cpupid_to_cpu(cpupid
) == (-1 & LAST__CPU_MASK
);
1149 static inline bool __cpupid_match_pid(pid_t task_pid
, int cpupid
)
1151 return (task_pid
& LAST__PID_MASK
) == cpupid_to_pid(cpupid
);
1154 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1155 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1156 static inline int page_cpupid_xchg_last(struct page
*page
, int cpupid
)
1158 return xchg(&page
->_last_cpupid
, cpupid
& LAST_CPUPID_MASK
);
1161 static inline int page_cpupid_last(struct page
*page
)
1163 return page
->_last_cpupid
;
1165 static inline void page_cpupid_reset_last(struct page
*page
)
1167 page
->_last_cpupid
= -1 & LAST_CPUPID_MASK
;
1170 static inline int page_cpupid_last(struct page
*page
)
1172 return (page
->flags
>> LAST_CPUPID_PGSHIFT
) & LAST_CPUPID_MASK
;
1175 extern int page_cpupid_xchg_last(struct page
*page
, int cpupid
);
1177 static inline void page_cpupid_reset_last(struct page
*page
)
1179 page
->flags
|= LAST_CPUPID_MASK
<< LAST_CPUPID_PGSHIFT
;
1181 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1182 #else /* !CONFIG_NUMA_BALANCING */
1183 static inline int page_cpupid_xchg_last(struct page
*page
, int cpupid
)
1185 return page_to_nid(page
); /* XXX */
1188 static inline int page_cpupid_last(struct page
*page
)
1190 return page_to_nid(page
); /* XXX */
1193 static inline int cpupid_to_nid(int cpupid
)
1198 static inline int cpupid_to_pid(int cpupid
)
1203 static inline int cpupid_to_cpu(int cpupid
)
1208 static inline int cpu_pid_to_cpupid(int nid
, int pid
)
1213 static inline bool cpupid_pid_unset(int cpupid
)
1218 static inline void page_cpupid_reset_last(struct page
*page
)
1222 static inline bool cpupid_match_pid(struct task_struct
*task
, int cpupid
)
1226 #endif /* CONFIG_NUMA_BALANCING */
1228 #ifdef CONFIG_KASAN_SW_TAGS
1229 static inline u8
page_kasan_tag(const struct page
*page
)
1231 return (page
->flags
>> KASAN_TAG_PGSHIFT
) & KASAN_TAG_MASK
;
1234 static inline void page_kasan_tag_set(struct page
*page
, u8 tag
)
1236 page
->flags
&= ~(KASAN_TAG_MASK
<< KASAN_TAG_PGSHIFT
);
1237 page
->flags
|= (tag
& KASAN_TAG_MASK
) << KASAN_TAG_PGSHIFT
;
1240 static inline void page_kasan_tag_reset(struct page
*page
)
1242 page_kasan_tag_set(page
, 0xff);
1245 static inline u8
page_kasan_tag(const struct page
*page
)
1250 static inline void page_kasan_tag_set(struct page
*page
, u8 tag
) { }
1251 static inline void page_kasan_tag_reset(struct page
*page
) { }
1254 static inline struct zone
*page_zone(const struct page
*page
)
1256 return &NODE_DATA(page_to_nid(page
))->node_zones
[page_zonenum(page
)];
1259 static inline pg_data_t
*page_pgdat(const struct page
*page
)
1261 return NODE_DATA(page_to_nid(page
));
1264 #ifdef SECTION_IN_PAGE_FLAGS
1265 static inline void set_page_section(struct page
*page
, unsigned long section
)
1267 page
->flags
&= ~(SECTIONS_MASK
<< SECTIONS_PGSHIFT
);
1268 page
->flags
|= (section
& SECTIONS_MASK
) << SECTIONS_PGSHIFT
;
1271 static inline unsigned long page_to_section(const struct page
*page
)
1273 return (page
->flags
>> SECTIONS_PGSHIFT
) & SECTIONS_MASK
;
1277 static inline void set_page_zone(struct page
*page
, enum zone_type zone
)
1279 page
->flags
&= ~(ZONES_MASK
<< ZONES_PGSHIFT
);
1280 page
->flags
|= (zone
& ZONES_MASK
) << ZONES_PGSHIFT
;
1283 static inline void set_page_node(struct page
*page
, unsigned long node
)
1285 page
->flags
&= ~(NODES_MASK
<< NODES_PGSHIFT
);
1286 page
->flags
|= (node
& NODES_MASK
) << NODES_PGSHIFT
;
1289 static inline void set_page_links(struct page
*page
, enum zone_type zone
,
1290 unsigned long node
, unsigned long pfn
)
1292 set_page_zone(page
, zone
);
1293 set_page_node(page
, node
);
1294 #ifdef SECTION_IN_PAGE_FLAGS
1295 set_page_section(page
, pfn_to_section_nr(pfn
));
1300 static inline struct mem_cgroup
*page_memcg(struct page
*page
)
1302 return page
->mem_cgroup
;
1304 static inline struct mem_cgroup
*page_memcg_rcu(struct page
*page
)
1306 WARN_ON_ONCE(!rcu_read_lock_held());
1307 return READ_ONCE(page
->mem_cgroup
);
1310 static inline struct mem_cgroup
*page_memcg(struct page
*page
)
1314 static inline struct mem_cgroup
*page_memcg_rcu(struct page
*page
)
1316 WARN_ON_ONCE(!rcu_read_lock_held());
1322 * Some inline functions in vmstat.h depend on page_zone()
1324 #include <linux/vmstat.h>
1326 static __always_inline
void *lowmem_page_address(const struct page
*page
)
1328 return page_to_virt(page
);
1331 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1332 #define HASHED_PAGE_VIRTUAL
1335 #if defined(WANT_PAGE_VIRTUAL)
1336 static inline void *page_address(const struct page
*page
)
1338 return page
->virtual;
1340 static inline void set_page_address(struct page
*page
, void *address
)
1342 page
->virtual = address
;
1344 #define page_address_init() do { } while(0)
1347 #if defined(HASHED_PAGE_VIRTUAL)
1348 void *page_address(const struct page
*page
);
1349 void set_page_address(struct page
*page
, void *virtual);
1350 void page_address_init(void);
1353 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1354 #define page_address(page) lowmem_page_address(page)
1355 #define set_page_address(page, address) do { } while(0)
1356 #define page_address_init() do { } while(0)
1359 extern void *page_rmapping(struct page
*page
);
1360 extern struct anon_vma
*page_anon_vma(struct page
*page
);
1361 extern struct address_space
*page_mapping(struct page
*page
);
1363 extern struct address_space
*__page_file_mapping(struct page
*);
1366 struct address_space
*page_file_mapping(struct page
*page
)
1368 if (unlikely(PageSwapCache(page
)))
1369 return __page_file_mapping(page
);
1371 return page
->mapping
;
1374 extern pgoff_t
__page_file_index(struct page
*page
);
1377 * Return the pagecache index of the passed page. Regular pagecache pages
1378 * use ->index whereas swapcache pages use swp_offset(->private)
1380 static inline pgoff_t
page_index(struct page
*page
)
1382 if (unlikely(PageSwapCache(page
)))
1383 return __page_file_index(page
);
1387 bool page_mapped(struct page
*page
);
1388 struct address_space
*page_mapping(struct page
*page
);
1389 struct address_space
*page_mapping_file(struct page
*page
);
1392 * Return true only if the page has been allocated with
1393 * ALLOC_NO_WATERMARKS and the low watermark was not
1394 * met implying that the system is under some pressure.
1396 static inline bool page_is_pfmemalloc(struct page
*page
)
1399 * Page index cannot be this large so this must be
1400 * a pfmemalloc page.
1402 return page
->index
== -1UL;
1406 * Only to be called by the page allocator on a freshly allocated
1409 static inline void set_page_pfmemalloc(struct page
*page
)
1414 static inline void clear_page_pfmemalloc(struct page
*page
)
1420 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1422 extern void pagefault_out_of_memory(void);
1424 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1427 * Flags passed to show_mem() and show_free_areas() to suppress output in
1430 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1432 extern void show_free_areas(unsigned int flags
, nodemask_t
*nodemask
);
1435 extern bool can_do_mlock(void);
1437 static inline bool can_do_mlock(void) { return false; }
1439 extern int user_shm_lock(size_t, struct user_struct
*);
1440 extern void user_shm_unlock(size_t, struct user_struct
*);
1443 * Parameter block passed down to zap_pte_range in exceptional cases.
1445 struct zap_details
{
1446 struct address_space
*check_mapping
; /* Check page->mapping if set */
1447 pgoff_t first_index
; /* Lowest page->index to unmap */
1448 pgoff_t last_index
; /* Highest page->index to unmap */
1451 struct page
*vm_normal_page(struct vm_area_struct
*vma
, unsigned long addr
,
1453 struct page
*vm_normal_page_pmd(struct vm_area_struct
*vma
, unsigned long addr
,
1456 void zap_vma_ptes(struct vm_area_struct
*vma
, unsigned long address
,
1457 unsigned long size
);
1458 void zap_page_range(struct vm_area_struct
*vma
, unsigned long address
,
1459 unsigned long size
);
1460 void unmap_vmas(struct mmu_gather
*tlb
, struct vm_area_struct
*start_vma
,
1461 unsigned long start
, unsigned long end
);
1463 struct mmu_notifier_range
;
1465 void free_pgd_range(struct mmu_gather
*tlb
, unsigned long addr
,
1466 unsigned long end
, unsigned long floor
, unsigned long ceiling
);
1467 int copy_page_range(struct mm_struct
*dst
, struct mm_struct
*src
,
1468 struct vm_area_struct
*vma
);
1469 int follow_pte_pmd(struct mm_struct
*mm
, unsigned long address
,
1470 struct mmu_notifier_range
*range
,
1471 pte_t
**ptepp
, pmd_t
**pmdpp
, spinlock_t
**ptlp
);
1472 int follow_pfn(struct vm_area_struct
*vma
, unsigned long address
,
1473 unsigned long *pfn
);
1474 int follow_phys(struct vm_area_struct
*vma
, unsigned long address
,
1475 unsigned int flags
, unsigned long *prot
, resource_size_t
*phys
);
1476 int generic_access_phys(struct vm_area_struct
*vma
, unsigned long addr
,
1477 void *buf
, int len
, int write
);
1479 extern void truncate_pagecache(struct inode
*inode
, loff_t
new);
1480 extern void truncate_setsize(struct inode
*inode
, loff_t newsize
);
1481 void pagecache_isize_extended(struct inode
*inode
, loff_t from
, loff_t to
);
1482 void truncate_pagecache_range(struct inode
*inode
, loff_t offset
, loff_t end
);
1483 int truncate_inode_page(struct address_space
*mapping
, struct page
*page
);
1484 int generic_error_remove_page(struct address_space
*mapping
, struct page
*page
);
1485 int invalidate_inode_page(struct page
*page
);
1488 extern vm_fault_t
handle_mm_fault(struct vm_area_struct
*vma
,
1489 unsigned long address
, unsigned int flags
);
1490 extern int fixup_user_fault(struct task_struct
*tsk
, struct mm_struct
*mm
,
1491 unsigned long address
, unsigned int fault_flags
,
1493 void unmap_mapping_pages(struct address_space
*mapping
,
1494 pgoff_t start
, pgoff_t nr
, bool even_cows
);
1495 void unmap_mapping_range(struct address_space
*mapping
,
1496 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
);
1498 static inline vm_fault_t
handle_mm_fault(struct vm_area_struct
*vma
,
1499 unsigned long address
, unsigned int flags
)
1501 /* should never happen if there's no MMU */
1503 return VM_FAULT_SIGBUS
;
1505 static inline int fixup_user_fault(struct task_struct
*tsk
,
1506 struct mm_struct
*mm
, unsigned long address
,
1507 unsigned int fault_flags
, bool *unlocked
)
1509 /* should never happen if there's no MMU */
1513 static inline void unmap_mapping_pages(struct address_space
*mapping
,
1514 pgoff_t start
, pgoff_t nr
, bool even_cows
) { }
1515 static inline void unmap_mapping_range(struct address_space
*mapping
,
1516 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
) { }
1519 static inline void unmap_shared_mapping_range(struct address_space
*mapping
,
1520 loff_t
const holebegin
, loff_t
const holelen
)
1522 unmap_mapping_range(mapping
, holebegin
, holelen
, 0);
1525 extern void vma_do_file_update_time(struct vm_area_struct
*, const char[], int);
1526 extern struct file
*vma_do_pr_or_file(struct vm_area_struct
*, const char[],
1528 extern void vma_do_get_file(struct vm_area_struct
*, const char[], int);
1529 extern void vma_do_fput(struct vm_area_struct
*, const char[], int);
1531 #define vma_file_update_time(vma) vma_do_file_update_time(vma, __func__, \
1533 #define vma_pr_or_file(vma) vma_do_pr_or_file(vma, __func__, \
1535 #define vma_get_file(vma) vma_do_get_file(vma, __func__, __LINE__)
1536 #define vma_fput(vma) vma_do_fput(vma, __func__, __LINE__)
1539 extern struct file
*vmr_do_pr_or_file(struct vm_region
*, const char[], int);
1540 extern void vmr_do_fput(struct vm_region
*, const char[], int);
1542 #define vmr_pr_or_file(region) vmr_do_pr_or_file(region, __func__, \
1544 #define vmr_fput(region) vmr_do_fput(region, __func__, __LINE__)
1545 #endif /* !CONFIG_MMU */
1547 extern int access_process_vm(struct task_struct
*tsk
, unsigned long addr
,
1548 void *buf
, int len
, unsigned int gup_flags
);
1549 extern int access_remote_vm(struct mm_struct
*mm
, unsigned long addr
,
1550 void *buf
, int len
, unsigned int gup_flags
);
1551 extern int __access_remote_vm(struct task_struct
*tsk
, struct mm_struct
*mm
,
1552 unsigned long addr
, void *buf
, int len
, unsigned int gup_flags
);
1554 long get_user_pages_remote(struct task_struct
*tsk
, struct mm_struct
*mm
,
1555 unsigned long start
, unsigned long nr_pages
,
1556 unsigned int gup_flags
, struct page
**pages
,
1557 struct vm_area_struct
**vmas
, int *locked
);
1558 long get_user_pages(unsigned long start
, unsigned long nr_pages
,
1559 unsigned int gup_flags
, struct page
**pages
,
1560 struct vm_area_struct
**vmas
);
1561 long get_user_pages_locked(unsigned long start
, unsigned long nr_pages
,
1562 unsigned int gup_flags
, struct page
**pages
, int *locked
);
1563 long get_user_pages_unlocked(unsigned long start
, unsigned long nr_pages
,
1564 struct page
**pages
, unsigned int gup_flags
);
1566 int get_user_pages_fast(unsigned long start
, int nr_pages
,
1567 unsigned int gup_flags
, struct page
**pages
);
1569 int account_locked_vm(struct mm_struct
*mm
, unsigned long pages
, bool inc
);
1570 int __account_locked_vm(struct mm_struct
*mm
, unsigned long pages
, bool inc
,
1571 struct task_struct
*task
, bool bypass_rlim
);
1573 /* Container for pinned pfns / pages */
1574 struct frame_vector
{
1575 unsigned int nr_allocated
; /* Number of frames we have space for */
1576 unsigned int nr_frames
; /* Number of frames stored in ptrs array */
1577 bool got_ref
; /* Did we pin pages by getting page ref? */
1578 bool is_pfns
; /* Does array contain pages or pfns? */
1579 void *ptrs
[0]; /* Array of pinned pfns / pages. Use
1580 * pfns_vector_pages() or pfns_vector_pfns()
1584 struct frame_vector
*frame_vector_create(unsigned int nr_frames
);
1585 void frame_vector_destroy(struct frame_vector
*vec
);
1586 int get_vaddr_frames(unsigned long start
, unsigned int nr_pfns
,
1587 unsigned int gup_flags
, struct frame_vector
*vec
);
1588 void put_vaddr_frames(struct frame_vector
*vec
);
1589 int frame_vector_to_pages(struct frame_vector
*vec
);
1590 void frame_vector_to_pfns(struct frame_vector
*vec
);
1592 static inline unsigned int frame_vector_count(struct frame_vector
*vec
)
1594 return vec
->nr_frames
;
1597 static inline struct page
**frame_vector_pages(struct frame_vector
*vec
)
1600 int err
= frame_vector_to_pages(vec
);
1603 return ERR_PTR(err
);
1605 return (struct page
**)(vec
->ptrs
);
1608 static inline unsigned long *frame_vector_pfns(struct frame_vector
*vec
)
1611 frame_vector_to_pfns(vec
);
1612 return (unsigned long *)(vec
->ptrs
);
1616 int get_kernel_pages(const struct kvec
*iov
, int nr_pages
, int write
,
1617 struct page
**pages
);
1618 int get_kernel_page(unsigned long start
, int write
, struct page
**pages
);
1619 struct page
*get_dump_page(unsigned long addr
);
1621 extern int try_to_release_page(struct page
* page
, gfp_t gfp_mask
);
1622 extern void do_invalidatepage(struct page
*page
, unsigned int offset
,
1623 unsigned int length
);
1625 void __set_page_dirty(struct page
*, struct address_space
*, int warn
);
1626 int __set_page_dirty_nobuffers(struct page
*page
);
1627 int __set_page_dirty_no_writeback(struct page
*page
);
1628 int redirty_page_for_writepage(struct writeback_control
*wbc
,
1630 void account_page_dirtied(struct page
*page
, struct address_space
*mapping
);
1631 void account_page_cleaned(struct page
*page
, struct address_space
*mapping
,
1632 struct bdi_writeback
*wb
);
1633 int set_page_dirty(struct page
*page
);
1634 int set_page_dirty_lock(struct page
*page
);
1635 void __cancel_dirty_page(struct page
*page
);
1636 static inline void cancel_dirty_page(struct page
*page
)
1638 /* Avoid atomic ops, locking, etc. when not actually needed. */
1639 if (PageDirty(page
))
1640 __cancel_dirty_page(page
);
1642 int clear_page_dirty_for_io(struct page
*page
);
1644 int get_cmdline(struct task_struct
*task
, char *buffer
, int buflen
);
1646 extern unsigned long move_page_tables(struct vm_area_struct
*vma
,
1647 unsigned long old_addr
, struct vm_area_struct
*new_vma
,
1648 unsigned long new_addr
, unsigned long len
,
1649 bool need_rmap_locks
);
1650 extern unsigned long change_protection(struct vm_area_struct
*vma
, unsigned long start
,
1651 unsigned long end
, pgprot_t newprot
,
1652 int dirty_accountable
, int prot_numa
);
1653 extern int mprotect_fixup(struct vm_area_struct
*vma
,
1654 struct vm_area_struct
**pprev
, unsigned long start
,
1655 unsigned long end
, unsigned long newflags
);
1658 * doesn't attempt to fault and will return short.
1660 int __get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
1661 struct page
**pages
);
1663 * per-process(per-mm_struct) statistics.
1665 static inline unsigned long get_mm_counter(struct mm_struct
*mm
, int member
)
1667 long val
= atomic_long_read(&mm
->rss_stat
.count
[member
]);
1669 #ifdef SPLIT_RSS_COUNTING
1671 * counter is updated in asynchronous manner and may go to minus.
1672 * But it's never be expected number for users.
1677 return (unsigned long)val
;
1680 static inline void add_mm_counter(struct mm_struct
*mm
, int member
, long value
)
1682 atomic_long_add(value
, &mm
->rss_stat
.count
[member
]);
1685 static inline void inc_mm_counter(struct mm_struct
*mm
, int member
)
1687 atomic_long_inc(&mm
->rss_stat
.count
[member
]);
1690 static inline void dec_mm_counter(struct mm_struct
*mm
, int member
)
1692 atomic_long_dec(&mm
->rss_stat
.count
[member
]);
1695 /* Optimized variant when page is already known not to be PageAnon */
1696 static inline int mm_counter_file(struct page
*page
)
1698 if (PageSwapBacked(page
))
1699 return MM_SHMEMPAGES
;
1700 return MM_FILEPAGES
;
1703 static inline int mm_counter(struct page
*page
)
1706 return MM_ANONPAGES
;
1707 return mm_counter_file(page
);
1710 static inline unsigned long get_mm_rss(struct mm_struct
*mm
)
1712 return get_mm_counter(mm
, MM_FILEPAGES
) +
1713 get_mm_counter(mm
, MM_ANONPAGES
) +
1714 get_mm_counter(mm
, MM_SHMEMPAGES
);
1717 static inline unsigned long get_mm_hiwater_rss(struct mm_struct
*mm
)
1719 return max(mm
->hiwater_rss
, get_mm_rss(mm
));
1722 static inline unsigned long get_mm_hiwater_vm(struct mm_struct
*mm
)
1724 return max(mm
->hiwater_vm
, mm
->total_vm
);
1727 static inline void update_hiwater_rss(struct mm_struct
*mm
)
1729 unsigned long _rss
= get_mm_rss(mm
);
1731 if ((mm
)->hiwater_rss
< _rss
)
1732 (mm
)->hiwater_rss
= _rss
;
1735 static inline void update_hiwater_vm(struct mm_struct
*mm
)
1737 if (mm
->hiwater_vm
< mm
->total_vm
)
1738 mm
->hiwater_vm
= mm
->total_vm
;
1741 static inline void reset_mm_hiwater_rss(struct mm_struct
*mm
)
1743 mm
->hiwater_rss
= get_mm_rss(mm
);
1746 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss
,
1747 struct mm_struct
*mm
)
1749 unsigned long hiwater_rss
= get_mm_hiwater_rss(mm
);
1751 if (*maxrss
< hiwater_rss
)
1752 *maxrss
= hiwater_rss
;
1755 #if defined(SPLIT_RSS_COUNTING)
1756 void sync_mm_rss(struct mm_struct
*mm
);
1758 static inline void sync_mm_rss(struct mm_struct
*mm
)
1763 #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
1764 static inline int pte_devmap(pte_t pte
)
1770 int vma_wants_writenotify(struct vm_area_struct
*vma
, pgprot_t vm_page_prot
);
1772 extern pte_t
*__get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1774 static inline pte_t
*get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1778 __cond_lock(*ptl
, ptep
= __get_locked_pte(mm
, addr
, ptl
));
1782 #ifdef __PAGETABLE_P4D_FOLDED
1783 static inline int __p4d_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
1784 unsigned long address
)
1789 int __p4d_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
);
1792 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
1793 static inline int __pud_alloc(struct mm_struct
*mm
, p4d_t
*p4d
,
1794 unsigned long address
)
1798 static inline void mm_inc_nr_puds(struct mm_struct
*mm
) {}
1799 static inline void mm_dec_nr_puds(struct mm_struct
*mm
) {}
1802 int __pud_alloc(struct mm_struct
*mm
, p4d_t
*p4d
, unsigned long address
);
1804 static inline void mm_inc_nr_puds(struct mm_struct
*mm
)
1806 if (mm_pud_folded(mm
))
1808 atomic_long_add(PTRS_PER_PUD
* sizeof(pud_t
), &mm
->pgtables_bytes
);
1811 static inline void mm_dec_nr_puds(struct mm_struct
*mm
)
1813 if (mm_pud_folded(mm
))
1815 atomic_long_sub(PTRS_PER_PUD
* sizeof(pud_t
), &mm
->pgtables_bytes
);
1819 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1820 static inline int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
,
1821 unsigned long address
)
1826 static inline void mm_inc_nr_pmds(struct mm_struct
*mm
) {}
1827 static inline void mm_dec_nr_pmds(struct mm_struct
*mm
) {}
1830 int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
);
1832 static inline void mm_inc_nr_pmds(struct mm_struct
*mm
)
1834 if (mm_pmd_folded(mm
))
1836 atomic_long_add(PTRS_PER_PMD
* sizeof(pmd_t
), &mm
->pgtables_bytes
);
1839 static inline void mm_dec_nr_pmds(struct mm_struct
*mm
)
1841 if (mm_pmd_folded(mm
))
1843 atomic_long_sub(PTRS_PER_PMD
* sizeof(pmd_t
), &mm
->pgtables_bytes
);
1848 static inline void mm_pgtables_bytes_init(struct mm_struct
*mm
)
1850 atomic_long_set(&mm
->pgtables_bytes
, 0);
1853 static inline unsigned long mm_pgtables_bytes(const struct mm_struct
*mm
)
1855 return atomic_long_read(&mm
->pgtables_bytes
);
1858 static inline void mm_inc_nr_ptes(struct mm_struct
*mm
)
1860 atomic_long_add(PTRS_PER_PTE
* sizeof(pte_t
), &mm
->pgtables_bytes
);
1863 static inline void mm_dec_nr_ptes(struct mm_struct
*mm
)
1865 atomic_long_sub(PTRS_PER_PTE
* sizeof(pte_t
), &mm
->pgtables_bytes
);
1869 static inline void mm_pgtables_bytes_init(struct mm_struct
*mm
) {}
1870 static inline unsigned long mm_pgtables_bytes(const struct mm_struct
*mm
)
1875 static inline void mm_inc_nr_ptes(struct mm_struct
*mm
) {}
1876 static inline void mm_dec_nr_ptes(struct mm_struct
*mm
) {}
1879 int __pte_alloc(struct mm_struct
*mm
, pmd_t
*pmd
);
1880 int __pte_alloc_kernel(pmd_t
*pmd
);
1883 * The following ifdef needed to get the 4level-fixup.h header to work.
1884 * Remove it when 4level-fixup.h has been removed.
1886 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1888 #ifndef __ARCH_HAS_5LEVEL_HACK
1889 static inline p4d_t
*p4d_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
1890 unsigned long address
)
1892 return (unlikely(pgd_none(*pgd
)) && __p4d_alloc(mm
, pgd
, address
)) ?
1893 NULL
: p4d_offset(pgd
, address
);
1896 static inline pud_t
*pud_alloc(struct mm_struct
*mm
, p4d_t
*p4d
,
1897 unsigned long address
)
1899 return (unlikely(p4d_none(*p4d
)) && __pud_alloc(mm
, p4d
, address
)) ?
1900 NULL
: pud_offset(p4d
, address
);
1902 #endif /* !__ARCH_HAS_5LEVEL_HACK */
1904 static inline pmd_t
*pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
)
1906 return (unlikely(pud_none(*pud
)) && __pmd_alloc(mm
, pud
, address
))?
1907 NULL
: pmd_offset(pud
, address
);
1909 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1911 #if USE_SPLIT_PTE_PTLOCKS
1912 #if ALLOC_SPLIT_PTLOCKS
1913 void __init
ptlock_cache_init(void);
1914 extern bool ptlock_alloc(struct page
*page
);
1915 extern void ptlock_free(struct page
*page
);
1917 static inline spinlock_t
*ptlock_ptr(struct page
*page
)
1921 #else /* ALLOC_SPLIT_PTLOCKS */
1922 static inline void ptlock_cache_init(void)
1926 static inline bool ptlock_alloc(struct page
*page
)
1931 static inline void ptlock_free(struct page
*page
)
1935 static inline spinlock_t
*ptlock_ptr(struct page
*page
)
1939 #endif /* ALLOC_SPLIT_PTLOCKS */
1941 static inline spinlock_t
*pte_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1943 return ptlock_ptr(pmd_page(*pmd
));
1946 static inline bool ptlock_init(struct page
*page
)
1949 * prep_new_page() initialize page->private (and therefore page->ptl)
1950 * with 0. Make sure nobody took it in use in between.
1952 * It can happen if arch try to use slab for page table allocation:
1953 * slab code uses page->slab_cache, which share storage with page->ptl.
1955 VM_BUG_ON_PAGE(*(unsigned long *)&page
->ptl
, page
);
1956 if (!ptlock_alloc(page
))
1958 spin_lock_init(ptlock_ptr(page
));
1962 #else /* !USE_SPLIT_PTE_PTLOCKS */
1964 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1966 static inline spinlock_t
*pte_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1968 return &mm
->page_table_lock
;
1970 static inline void ptlock_cache_init(void) {}
1971 static inline bool ptlock_init(struct page
*page
) { return true; }
1972 static inline void ptlock_free(struct page
*page
) {}
1973 #endif /* USE_SPLIT_PTE_PTLOCKS */
1975 static inline void pgtable_init(void)
1977 ptlock_cache_init();
1978 pgtable_cache_init();
1981 static inline bool pgtable_pte_page_ctor(struct page
*page
)
1983 if (!ptlock_init(page
))
1985 __SetPageTable(page
);
1986 inc_zone_page_state(page
, NR_PAGETABLE
);
1990 static inline void pgtable_pte_page_dtor(struct page
*page
)
1993 __ClearPageTable(page
);
1994 dec_zone_page_state(page
, NR_PAGETABLE
);
1997 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1999 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
2000 pte_t *__pte = pte_offset_map(pmd, address); \
2006 #define pte_unmap_unlock(pte, ptl) do { \
2011 #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
2013 #define pte_alloc_map(mm, pmd, address) \
2014 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
2016 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
2017 (pte_alloc(mm, pmd) ? \
2018 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
2020 #define pte_alloc_kernel(pmd, address) \
2021 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2022 NULL: pte_offset_kernel(pmd, address))
2024 #if USE_SPLIT_PMD_PTLOCKS
2026 static struct page
*pmd_to_page(pmd_t
*pmd
)
2028 unsigned long mask
= ~(PTRS_PER_PMD
* sizeof(pmd_t
) - 1);
2029 return virt_to_page((void *)((unsigned long) pmd
& mask
));
2032 static inline spinlock_t
*pmd_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
2034 return ptlock_ptr(pmd_to_page(pmd
));
2037 static inline bool pgtable_pmd_page_ctor(struct page
*page
)
2039 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2040 page
->pmd_huge_pte
= NULL
;
2042 return ptlock_init(page
);
2045 static inline void pgtable_pmd_page_dtor(struct page
*page
)
2047 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2048 VM_BUG_ON_PAGE(page
->pmd_huge_pte
, page
);
2053 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2057 static inline spinlock_t
*pmd_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
2059 return &mm
->page_table_lock
;
2062 static inline bool pgtable_pmd_page_ctor(struct page
*page
) { return true; }
2063 static inline void pgtable_pmd_page_dtor(struct page
*page
) {}
2065 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2069 static inline spinlock_t
*pmd_lock(struct mm_struct
*mm
, pmd_t
*pmd
)
2071 spinlock_t
*ptl
= pmd_lockptr(mm
, pmd
);
2077 * No scalability reason to split PUD locks yet, but follow the same pattern
2078 * as the PMD locks to make it easier if we decide to. The VM should not be
2079 * considered ready to switch to split PUD locks yet; there may be places
2080 * which need to be converted from page_table_lock.
2082 static inline spinlock_t
*pud_lockptr(struct mm_struct
*mm
, pud_t
*pud
)
2084 return &mm
->page_table_lock
;
2087 static inline spinlock_t
*pud_lock(struct mm_struct
*mm
, pud_t
*pud
)
2089 spinlock_t
*ptl
= pud_lockptr(mm
, pud
);
2095 extern void __init
pagecache_init(void);
2096 extern void free_area_init(unsigned long * zones_size
);
2097 extern void __init
free_area_init_node(int nid
, unsigned long * zones_size
,
2098 unsigned long zone_start_pfn
, unsigned long *zholes_size
);
2099 extern void free_initmem(void);
2102 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2103 * into the buddy system. The freed pages will be poisoned with pattern
2104 * "poison" if it's within range [0, UCHAR_MAX].
2105 * Return pages freed into the buddy system.
2107 extern unsigned long free_reserved_area(void *start
, void *end
,
2108 int poison
, const char *s
);
2110 #ifdef CONFIG_HIGHMEM
2112 * Free a highmem page into the buddy system, adjusting totalhigh_pages
2113 * and totalram_pages.
2115 extern void free_highmem_page(struct page
*page
);
2118 extern void adjust_managed_page_count(struct page
*page
, long count
);
2119 extern void mem_init_print_info(const char *str
);
2121 extern void reserve_bootmem_region(phys_addr_t start
, phys_addr_t end
);
2123 /* Free the reserved page into the buddy system, so it gets managed. */
2124 static inline void __free_reserved_page(struct page
*page
)
2126 ClearPageReserved(page
);
2127 init_page_count(page
);
2131 static inline void free_reserved_page(struct page
*page
)
2133 __free_reserved_page(page
);
2134 adjust_managed_page_count(page
, 1);
2137 static inline void mark_page_reserved(struct page
*page
)
2139 SetPageReserved(page
);
2140 adjust_managed_page_count(page
, -1);
2144 * Default method to free all the __init memory into the buddy system.
2145 * The freed pages will be poisoned with pattern "poison" if it's within
2146 * range [0, UCHAR_MAX].
2147 * Return pages freed into the buddy system.
2149 static inline unsigned long free_initmem_default(int poison
)
2151 extern char __init_begin
[], __init_end
[];
2153 return free_reserved_area(&__init_begin
, &__init_end
,
2154 poison
, "unused kernel");
2157 static inline unsigned long get_num_physpages(void)
2160 unsigned long phys_pages
= 0;
2162 for_each_online_node(nid
)
2163 phys_pages
+= node_present_pages(nid
);
2168 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
2170 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
2171 * zones, allocate the backing mem_map and account for memory holes in a more
2172 * architecture independent manner. This is a substitute for creating the
2173 * zone_sizes[] and zholes_size[] arrays and passing them to
2174 * free_area_init_node()
2176 * An architecture is expected to register range of page frames backed by
2177 * physical memory with memblock_add[_node]() before calling
2178 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
2179 * usage, an architecture is expected to do something like
2181 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2183 * for_each_valid_physical_page_range()
2184 * memblock_add_node(base, size, nid)
2185 * free_area_init_nodes(max_zone_pfns);
2187 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
2188 * registered physical page range. Similarly
2189 * sparse_memory_present_with_active_regions() calls memory_present() for
2190 * each range when SPARSEMEM is enabled.
2192 * See mm/page_alloc.c for more information on each function exposed by
2193 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
2195 extern void free_area_init_nodes(unsigned long *max_zone_pfn
);
2196 unsigned long node_map_pfn_alignment(void);
2197 unsigned long __absent_pages_in_range(int nid
, unsigned long start_pfn
,
2198 unsigned long end_pfn
);
2199 extern unsigned long absent_pages_in_range(unsigned long start_pfn
,
2200 unsigned long end_pfn
);
2201 extern void get_pfn_range_for_nid(unsigned int nid
,
2202 unsigned long *start_pfn
, unsigned long *end_pfn
);
2203 extern unsigned long find_min_pfn_with_active_regions(void);
2204 extern void free_bootmem_with_active_regions(int nid
,
2205 unsigned long max_low_pfn
);
2206 extern void sparse_memory_present_with_active_regions(int nid
);
2208 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
2210 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
2211 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
2212 static inline int __early_pfn_to_nid(unsigned long pfn
,
2213 struct mminit_pfnnid_cache
*state
)
2218 /* please see mm/page_alloc.c */
2219 extern int __meminit
early_pfn_to_nid(unsigned long pfn
);
2220 /* there is a per-arch backend function. */
2221 extern int __meminit
__early_pfn_to_nid(unsigned long pfn
,
2222 struct mminit_pfnnid_cache
*state
);
2225 #if !defined(CONFIG_FLAT_NODE_MEM_MAP)
2226 void zero_resv_unavail(void);
2228 static inline void zero_resv_unavail(void) {}
2231 extern void set_dma_reserve(unsigned long new_dma_reserve
);
2232 extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long,
2233 enum meminit_context
, struct vmem_altmap
*);
2234 extern void setup_per_zone_wmarks(void);
2235 extern int __meminit
init_per_zone_wmark_min(void);
2236 extern void mem_init(void);
2237 extern void __init
mmap_init(void);
2238 extern void show_mem(unsigned int flags
, nodemask_t
*nodemask
);
2239 extern long si_mem_available(void);
2240 extern void si_meminfo(struct sysinfo
* val
);
2241 extern void si_meminfo_node(struct sysinfo
*val
, int nid
);
2242 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2243 extern unsigned long arch_reserved_kernel_pages(void);
2246 extern __printf(3, 4)
2247 void warn_alloc(gfp_t gfp_mask
, nodemask_t
*nodemask
, const char *fmt
, ...);
2249 extern void setup_per_cpu_pageset(void);
2251 extern void zone_pcp_update(struct zone
*zone
);
2252 extern void zone_pcp_reset(struct zone
*zone
);
2255 extern int min_free_kbytes
;
2256 extern int watermark_boost_factor
;
2257 extern int watermark_scale_factor
;
2260 extern atomic_long_t mmap_pages_allocated
;
2261 extern int nommu_shrink_inode_mappings(struct inode
*, size_t, size_t);
2263 /* interval_tree.c */
2264 void vma_interval_tree_insert(struct vm_area_struct
*node
,
2265 struct rb_root_cached
*root
);
2266 void vma_interval_tree_insert_after(struct vm_area_struct
*node
,
2267 struct vm_area_struct
*prev
,
2268 struct rb_root_cached
*root
);
2269 void vma_interval_tree_remove(struct vm_area_struct
*node
,
2270 struct rb_root_cached
*root
);
2271 struct vm_area_struct
*vma_interval_tree_iter_first(struct rb_root_cached
*root
,
2272 unsigned long start
, unsigned long last
);
2273 struct vm_area_struct
*vma_interval_tree_iter_next(struct vm_area_struct
*node
,
2274 unsigned long start
, unsigned long last
);
2276 #define vma_interval_tree_foreach(vma, root, start, last) \
2277 for (vma = vma_interval_tree_iter_first(root, start, last); \
2278 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2280 void anon_vma_interval_tree_insert(struct anon_vma_chain
*node
,
2281 struct rb_root_cached
*root
);
2282 void anon_vma_interval_tree_remove(struct anon_vma_chain
*node
,
2283 struct rb_root_cached
*root
);
2284 struct anon_vma_chain
*
2285 anon_vma_interval_tree_iter_first(struct rb_root_cached
*root
,
2286 unsigned long start
, unsigned long last
);
2287 struct anon_vma_chain
*anon_vma_interval_tree_iter_next(
2288 struct anon_vma_chain
*node
, unsigned long start
, unsigned long last
);
2289 #ifdef CONFIG_DEBUG_VM_RB
2290 void anon_vma_interval_tree_verify(struct anon_vma_chain
*node
);
2293 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2294 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2295 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2298 extern int __vm_enough_memory(struct mm_struct
*mm
, long pages
, int cap_sys_admin
);
2299 extern int __vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
2300 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
,
2301 struct vm_area_struct
*expand
);
2302 static inline int vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
2303 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
)
2305 return __vma_adjust(vma
, start
, end
, pgoff
, insert
, NULL
);
2307 extern struct vm_area_struct
*vma_merge(struct mm_struct
*,
2308 struct vm_area_struct
*prev
, unsigned long addr
, unsigned long end
,
2309 unsigned long vm_flags
, struct anon_vma
*, struct file
*, pgoff_t
,
2310 struct mempolicy
*, struct vm_userfaultfd_ctx
);
2311 extern struct anon_vma
*find_mergeable_anon_vma(struct vm_area_struct
*);
2312 extern int __split_vma(struct mm_struct
*, struct vm_area_struct
*,
2313 unsigned long addr
, int new_below
);
2314 extern int split_vma(struct mm_struct
*, struct vm_area_struct
*,
2315 unsigned long addr
, int new_below
);
2316 extern int insert_vm_struct(struct mm_struct
*, struct vm_area_struct
*);
2317 extern void __vma_link_rb(struct mm_struct
*, struct vm_area_struct
*,
2318 struct rb_node
**, struct rb_node
*);
2319 extern void unlink_file_vma(struct vm_area_struct
*);
2320 extern struct vm_area_struct
*copy_vma(struct vm_area_struct
**,
2321 unsigned long addr
, unsigned long len
, pgoff_t pgoff
,
2322 bool *need_rmap_locks
);
2323 extern void exit_mmap(struct mm_struct
*);
2325 static inline int check_data_rlimit(unsigned long rlim
,
2327 unsigned long start
,
2328 unsigned long end_data
,
2329 unsigned long start_data
)
2331 if (rlim
< RLIM_INFINITY
) {
2332 if (((new - start
) + (end_data
- start_data
)) > rlim
)
2339 extern int mm_take_all_locks(struct mm_struct
*mm
);
2340 extern void mm_drop_all_locks(struct mm_struct
*mm
);
2342 extern void set_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
);
2343 extern struct file
*get_mm_exe_file(struct mm_struct
*mm
);
2344 extern struct file
*get_task_exe_file(struct task_struct
*task
);
2346 extern bool may_expand_vm(struct mm_struct
*, vm_flags_t
, unsigned long npages
);
2347 extern void vm_stat_account(struct mm_struct
*, vm_flags_t
, long npages
);
2349 extern bool vma_is_special_mapping(const struct vm_area_struct
*vma
,
2350 const struct vm_special_mapping
*sm
);
2351 extern struct vm_area_struct
*_install_special_mapping(struct mm_struct
*mm
,
2352 unsigned long addr
, unsigned long len
,
2353 unsigned long flags
,
2354 const struct vm_special_mapping
*spec
);
2355 /* This is an obsolete alternative to _install_special_mapping. */
2356 extern int install_special_mapping(struct mm_struct
*mm
,
2357 unsigned long addr
, unsigned long len
,
2358 unsigned long flags
, struct page
**pages
);
2360 unsigned long randomize_stack_top(unsigned long stack_top
);
2362 extern unsigned long get_unmapped_area(struct file
*, unsigned long, unsigned long, unsigned long, unsigned long);
2364 extern unsigned long mmap_region(struct file
*file
, unsigned long addr
,
2365 unsigned long len
, vm_flags_t vm_flags
, unsigned long pgoff
,
2366 struct list_head
*uf
);
2367 extern unsigned long do_mmap(struct file
*file
, unsigned long addr
,
2368 unsigned long len
, unsigned long prot
, unsigned long flags
,
2369 vm_flags_t vm_flags
, unsigned long pgoff
, unsigned long *populate
,
2370 struct list_head
*uf
);
2371 extern int __do_munmap(struct mm_struct
*, unsigned long, size_t,
2372 struct list_head
*uf
, bool downgrade
);
2373 extern int do_munmap(struct mm_struct
*, unsigned long, size_t,
2374 struct list_head
*uf
);
2376 static inline unsigned long
2377 do_mmap_pgoff(struct file
*file
, unsigned long addr
,
2378 unsigned long len
, unsigned long prot
, unsigned long flags
,
2379 unsigned long pgoff
, unsigned long *populate
,
2380 struct list_head
*uf
)
2382 return do_mmap(file
, addr
, len
, prot
, flags
, 0, pgoff
, populate
, uf
);
2386 extern int __mm_populate(unsigned long addr
, unsigned long len
,
2388 static inline void mm_populate(unsigned long addr
, unsigned long len
)
2391 (void) __mm_populate(addr
, len
, 1);
2394 static inline void mm_populate(unsigned long addr
, unsigned long len
) {}
2397 /* These take the mm semaphore themselves */
2398 extern int __must_check
vm_brk(unsigned long, unsigned long);
2399 extern int __must_check
vm_brk_flags(unsigned long, unsigned long, unsigned long);
2400 extern int vm_munmap(unsigned long, size_t);
2401 extern unsigned long __must_check
vm_mmap(struct file
*, unsigned long,
2402 unsigned long, unsigned long,
2403 unsigned long, unsigned long);
2405 struct vm_unmapped_area_info
{
2406 #define VM_UNMAPPED_AREA_TOPDOWN 1
2407 unsigned long flags
;
2408 unsigned long length
;
2409 unsigned long low_limit
;
2410 unsigned long high_limit
;
2411 unsigned long align_mask
;
2412 unsigned long align_offset
;
2415 extern unsigned long unmapped_area(struct vm_unmapped_area_info
*info
);
2416 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info
*info
);
2419 * Search for an unmapped address range.
2421 * We are looking for a range that:
2422 * - does not intersect with any VMA;
2423 * - is contained within the [low_limit, high_limit) interval;
2424 * - is at least the desired size.
2425 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2427 static inline unsigned long
2428 vm_unmapped_area(struct vm_unmapped_area_info
*info
)
2430 if (info
->flags
& VM_UNMAPPED_AREA_TOPDOWN
)
2431 return unmapped_area_topdown(info
);
2433 return unmapped_area(info
);
2437 extern void truncate_inode_pages(struct address_space
*, loff_t
);
2438 extern void truncate_inode_pages_range(struct address_space
*,
2439 loff_t lstart
, loff_t lend
);
2440 extern void truncate_inode_pages_final(struct address_space
*);
2442 /* generic vm_area_ops exported for stackable file systems */
2443 extern vm_fault_t
filemap_fault(struct vm_fault
*vmf
);
2444 extern void filemap_map_pages(struct vm_fault
*vmf
,
2445 pgoff_t start_pgoff
, pgoff_t end_pgoff
);
2446 extern vm_fault_t
filemap_page_mkwrite(struct vm_fault
*vmf
);
2448 /* mm/page-writeback.c */
2449 int __must_check
write_one_page(struct page
*page
);
2450 void task_dirty_inc(struct task_struct
*tsk
);
2453 #define VM_READAHEAD_PAGES (SZ_128K / PAGE_SIZE)
2455 int force_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
2456 pgoff_t offset
, unsigned long nr_to_read
);
2458 void page_cache_sync_readahead(struct address_space
*mapping
,
2459 struct file_ra_state
*ra
,
2462 unsigned long size
);
2464 void page_cache_async_readahead(struct address_space
*mapping
,
2465 struct file_ra_state
*ra
,
2469 unsigned long size
);
2471 extern unsigned long stack_guard_gap
;
2472 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2473 extern int expand_stack(struct vm_area_struct
*vma
, unsigned long address
);
2475 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2476 extern int expand_downwards(struct vm_area_struct
*vma
,
2477 unsigned long address
);
2479 extern int expand_upwards(struct vm_area_struct
*vma
, unsigned long address
);
2481 #define expand_upwards(vma, address) (0)
2484 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2485 extern struct vm_area_struct
* find_vma(struct mm_struct
* mm
, unsigned long addr
);
2486 extern struct vm_area_struct
* find_vma_prev(struct mm_struct
* mm
, unsigned long addr
,
2487 struct vm_area_struct
**pprev
);
2489 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2490 NULL if none. Assume start_addr < end_addr. */
2491 static inline struct vm_area_struct
* find_vma_intersection(struct mm_struct
* mm
, unsigned long start_addr
, unsigned long end_addr
)
2493 struct vm_area_struct
* vma
= find_vma(mm
,start_addr
);
2495 if (vma
&& end_addr
<= vma
->vm_start
)
2500 static inline unsigned long vm_start_gap(struct vm_area_struct
*vma
)
2502 unsigned long vm_start
= vma
->vm_start
;
2504 if (vma
->vm_flags
& VM_GROWSDOWN
) {
2505 vm_start
-= stack_guard_gap
;
2506 if (vm_start
> vma
->vm_start
)
2512 static inline unsigned long vm_end_gap(struct vm_area_struct
*vma
)
2514 unsigned long vm_end
= vma
->vm_end
;
2516 if (vma
->vm_flags
& VM_GROWSUP
) {
2517 vm_end
+= stack_guard_gap
;
2518 if (vm_end
< vma
->vm_end
)
2519 vm_end
= -PAGE_SIZE
;
2524 static inline unsigned long vma_pages(struct vm_area_struct
*vma
)
2526 return (vma
->vm_end
- vma
->vm_start
) >> PAGE_SHIFT
;
2529 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2530 static inline struct vm_area_struct
*find_exact_vma(struct mm_struct
*mm
,
2531 unsigned long vm_start
, unsigned long vm_end
)
2533 struct vm_area_struct
*vma
= find_vma(mm
, vm_start
);
2535 if (vma
&& (vma
->vm_start
!= vm_start
|| vma
->vm_end
!= vm_end
))
2541 static inline bool range_in_vma(struct vm_area_struct
*vma
,
2542 unsigned long start
, unsigned long end
)
2544 return (vma
&& vma
->vm_start
<= start
&& end
<= vma
->vm_end
);
2548 pgprot_t
vm_get_page_prot(unsigned long vm_flags
);
2549 void vma_set_page_prot(struct vm_area_struct
*vma
);
2551 static inline pgprot_t
vm_get_page_prot(unsigned long vm_flags
)
2555 static inline void vma_set_page_prot(struct vm_area_struct
*vma
)
2557 vma
->vm_page_prot
= vm_get_page_prot(vma
->vm_flags
);
2561 #ifdef CONFIG_NUMA_BALANCING
2562 unsigned long change_prot_numa(struct vm_area_struct
*vma
,
2563 unsigned long start
, unsigned long end
);
2566 struct vm_area_struct
*find_extend_vma(struct mm_struct
*, unsigned long addr
);
2567 int remap_pfn_range(struct vm_area_struct
*, unsigned long addr
,
2568 unsigned long pfn
, unsigned long size
, pgprot_t
);
2569 int vm_insert_page(struct vm_area_struct
*, unsigned long addr
, struct page
*);
2570 int vm_map_pages(struct vm_area_struct
*vma
, struct page
**pages
,
2572 int vm_map_pages_zero(struct vm_area_struct
*vma
, struct page
**pages
,
2574 vm_fault_t
vmf_insert_pfn(struct vm_area_struct
*vma
, unsigned long addr
,
2576 vm_fault_t
vmf_insert_pfn_prot(struct vm_area_struct
*vma
, unsigned long addr
,
2577 unsigned long pfn
, pgprot_t pgprot
);
2578 vm_fault_t
vmf_insert_mixed(struct vm_area_struct
*vma
, unsigned long addr
,
2580 vm_fault_t
vmf_insert_mixed_mkwrite(struct vm_area_struct
*vma
,
2581 unsigned long addr
, pfn_t pfn
);
2582 int vm_iomap_memory(struct vm_area_struct
*vma
, phys_addr_t start
, unsigned long len
);
2584 static inline vm_fault_t
vmf_insert_page(struct vm_area_struct
*vma
,
2585 unsigned long addr
, struct page
*page
)
2587 int err
= vm_insert_page(vma
, addr
, page
);
2590 return VM_FAULT_OOM
;
2591 if (err
< 0 && err
!= -EBUSY
)
2592 return VM_FAULT_SIGBUS
;
2594 return VM_FAULT_NOPAGE
;
2597 static inline vm_fault_t
vmf_error(int err
)
2600 return VM_FAULT_OOM
;
2601 return VM_FAULT_SIGBUS
;
2604 struct page
*follow_page(struct vm_area_struct
*vma
, unsigned long address
,
2605 unsigned int foll_flags
);
2607 #define FOLL_WRITE 0x01 /* check pte is writable */
2608 #define FOLL_TOUCH 0x02 /* mark page accessed */
2609 #define FOLL_GET 0x04 /* do get_page on page */
2610 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2611 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2612 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2613 * and return without waiting upon it */
2614 #define FOLL_POPULATE 0x40 /* fault in page */
2615 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2616 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2617 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2618 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2619 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2620 #define FOLL_MLOCK 0x1000 /* lock present pages */
2621 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2622 #define FOLL_COW 0x4000 /* internal GUP flag */
2623 #define FOLL_ANON 0x8000 /* don't do file mappings */
2624 #define FOLL_LONGTERM 0x10000 /* mapping lifetime is indefinite: see below */
2625 #define FOLL_SPLIT_PMD 0x20000 /* split huge pmd before returning */
2628 * NOTE on FOLL_LONGTERM:
2630 * FOLL_LONGTERM indicates that the page will be held for an indefinite time
2631 * period _often_ under userspace control. This is contrasted with
2632 * iov_iter_get_pages() where usages which are transient.
2634 * FIXME: For pages which are part of a filesystem, mappings are subject to the
2635 * lifetime enforced by the filesystem and we need guarantees that longterm
2636 * users like RDMA and V4L2 only establish mappings which coordinate usage with
2637 * the filesystem. Ideas for this coordination include revoking the longterm
2638 * pin, delaying writeback, bounce buffer page writeback, etc. As FS DAX was
2639 * added after the problem with filesystems was found FS DAX VMAs are
2640 * specifically failed. Filesystem pages are still subject to bugs and use of
2641 * FOLL_LONGTERM should be avoided on those pages.
2643 * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call.
2644 * Currently only get_user_pages() and get_user_pages_fast() support this flag
2645 * and calls to get_user_pages_[un]locked are specifically not allowed. This
2646 * is due to an incompatibility with the FS DAX check and
2647 * FAULT_FLAG_ALLOW_RETRY
2649 * In the CMA case: longterm pins in a CMA region would unnecessarily fragment
2650 * that region. And so CMA attempts to migrate the page before pinning when
2651 * FOLL_LONGTERM is specified.
2654 static inline int vm_fault_to_errno(vm_fault_t vm_fault
, int foll_flags
)
2656 if (vm_fault
& VM_FAULT_OOM
)
2658 if (vm_fault
& (VM_FAULT_HWPOISON
| VM_FAULT_HWPOISON_LARGE
))
2659 return (foll_flags
& FOLL_HWPOISON
) ? -EHWPOISON
: -EFAULT
;
2660 if (vm_fault
& (VM_FAULT_SIGBUS
| VM_FAULT_SIGSEGV
))
2665 typedef int (*pte_fn_t
)(pte_t
*pte
, unsigned long addr
, void *data
);
2666 extern int apply_to_page_range(struct mm_struct
*mm
, unsigned long address
,
2667 unsigned long size
, pte_fn_t fn
, void *data
);
2670 #ifdef CONFIG_PAGE_POISONING
2671 extern bool page_poisoning_enabled(void);
2672 extern void kernel_poison_pages(struct page
*page
, int numpages
, int enable
);
2674 static inline bool page_poisoning_enabled(void) { return false; }
2675 static inline void kernel_poison_pages(struct page
*page
, int numpages
,
2679 #ifdef CONFIG_INIT_ON_ALLOC_DEFAULT_ON
2680 DECLARE_STATIC_KEY_TRUE(init_on_alloc
);
2682 DECLARE_STATIC_KEY_FALSE(init_on_alloc
);
2684 static inline bool want_init_on_alloc(gfp_t flags
)
2686 if (static_branch_unlikely(&init_on_alloc
) &&
2687 !page_poisoning_enabled())
2689 return flags
& __GFP_ZERO
;
2692 #ifdef CONFIG_INIT_ON_FREE_DEFAULT_ON
2693 DECLARE_STATIC_KEY_TRUE(init_on_free
);
2695 DECLARE_STATIC_KEY_FALSE(init_on_free
);
2697 static inline bool want_init_on_free(void)
2699 return static_branch_unlikely(&init_on_free
) &&
2700 !page_poisoning_enabled();
2703 #ifdef CONFIG_DEBUG_PAGEALLOC
2704 extern void init_debug_pagealloc(void);
2706 static inline void init_debug_pagealloc(void) {}
2708 extern bool _debug_pagealloc_enabled_early
;
2709 DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled
);
2711 static inline bool debug_pagealloc_enabled(void)
2713 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC
) &&
2714 _debug_pagealloc_enabled_early
;
2718 * For use in fast paths after init_debug_pagealloc() has run, or when a
2719 * false negative result is not harmful when called too early.
2721 static inline bool debug_pagealloc_enabled_static(void)
2723 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC
))
2726 return static_branch_unlikely(&_debug_pagealloc_enabled
);
2729 #if defined(CONFIG_DEBUG_PAGEALLOC) || defined(CONFIG_ARCH_HAS_SET_DIRECT_MAP)
2730 extern void __kernel_map_pages(struct page
*page
, int numpages
, int enable
);
2733 * When called in DEBUG_PAGEALLOC context, the call should most likely be
2734 * guarded by debug_pagealloc_enabled() or debug_pagealloc_enabled_static()
2737 kernel_map_pages(struct page
*page
, int numpages
, int enable
)
2739 __kernel_map_pages(page
, numpages
, enable
);
2741 #ifdef CONFIG_HIBERNATION
2742 extern bool kernel_page_present(struct page
*page
);
2743 #endif /* CONFIG_HIBERNATION */
2744 #else /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */
2746 kernel_map_pages(struct page
*page
, int numpages
, int enable
) {}
2747 #ifdef CONFIG_HIBERNATION
2748 static inline bool kernel_page_present(struct page
*page
) { return true; }
2749 #endif /* CONFIG_HIBERNATION */
2750 #endif /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */
2752 #ifdef __HAVE_ARCH_GATE_AREA
2753 extern struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
);
2754 extern int in_gate_area_no_mm(unsigned long addr
);
2755 extern int in_gate_area(struct mm_struct
*mm
, unsigned long addr
);
2757 static inline struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
)
2761 static inline int in_gate_area_no_mm(unsigned long addr
) { return 0; }
2762 static inline int in_gate_area(struct mm_struct
*mm
, unsigned long addr
)
2766 #endif /* __HAVE_ARCH_GATE_AREA */
2768 extern bool process_shares_mm(struct task_struct
*p
, struct mm_struct
*mm
);
2770 #ifdef CONFIG_SYSCTL
2771 extern int sysctl_drop_caches
;
2772 int drop_caches_sysctl_handler(struct ctl_table
*, int,
2773 void __user
*, size_t *, loff_t
*);
2776 void drop_slab(void);
2777 void drop_slab_node(int nid
);
2780 #define randomize_va_space 0
2782 extern int randomize_va_space
;
2785 const char * arch_vma_name(struct vm_area_struct
*vma
);
2787 void print_vma_addr(char *prefix
, unsigned long rip
);
2789 static inline void print_vma_addr(char *prefix
, unsigned long rip
)
2794 void *sparse_buffer_alloc(unsigned long size
);
2795 struct page
* __populate_section_memmap(unsigned long pfn
,
2796 unsigned long nr_pages
, int nid
, struct vmem_altmap
*altmap
);
2797 pgd_t
*vmemmap_pgd_populate(unsigned long addr
, int node
);
2798 p4d_t
*vmemmap_p4d_populate(pgd_t
*pgd
, unsigned long addr
, int node
);
2799 pud_t
*vmemmap_pud_populate(p4d_t
*p4d
, unsigned long addr
, int node
);
2800 pmd_t
*vmemmap_pmd_populate(pud_t
*pud
, unsigned long addr
, int node
);
2801 pte_t
*vmemmap_pte_populate(pmd_t
*pmd
, unsigned long addr
, int node
);
2802 void *vmemmap_alloc_block(unsigned long size
, int node
);
2804 void *vmemmap_alloc_block_buf(unsigned long size
, int node
);
2805 void *altmap_alloc_block_buf(unsigned long size
, struct vmem_altmap
*altmap
);
2806 void vmemmap_verify(pte_t
*, int, unsigned long, unsigned long);
2807 int vmemmap_populate_basepages(unsigned long start
, unsigned long end
,
2809 int vmemmap_populate(unsigned long start
, unsigned long end
, int node
,
2810 struct vmem_altmap
*altmap
);
2811 void vmemmap_populate_print_last(void);
2812 #ifdef CONFIG_MEMORY_HOTPLUG
2813 void vmemmap_free(unsigned long start
, unsigned long end
,
2814 struct vmem_altmap
*altmap
);
2816 void register_page_bootmem_memmap(unsigned long section_nr
, struct page
*map
,
2817 unsigned long nr_pages
);
2820 MF_COUNT_INCREASED
= 1 << 0,
2821 MF_ACTION_REQUIRED
= 1 << 1,
2822 MF_MUST_KILL
= 1 << 2,
2823 MF_SOFT_OFFLINE
= 1 << 3,
2825 extern int memory_failure(unsigned long pfn
, int flags
);
2826 extern void memory_failure_queue(unsigned long pfn
, int flags
);
2827 extern int unpoison_memory(unsigned long pfn
);
2828 extern int get_hwpoison_page(struct page
*page
);
2829 #define put_hwpoison_page(page) put_page(page)
2830 extern int sysctl_memory_failure_early_kill
;
2831 extern int sysctl_memory_failure_recovery
;
2832 extern void shake_page(struct page
*p
, int access
);
2833 extern atomic_long_t num_poisoned_pages __read_mostly
;
2834 extern int soft_offline_page(struct page
*page
, int flags
);
2838 * Error handlers for various types of pages.
2841 MF_IGNORED
, /* Error: cannot be handled */
2842 MF_FAILED
, /* Error: handling failed */
2843 MF_DELAYED
, /* Will be handled later */
2844 MF_RECOVERED
, /* Successfully recovered */
2847 enum mf_action_page_type
{
2849 MF_MSG_KERNEL_HIGH_ORDER
,
2851 MF_MSG_DIFFERENT_COMPOUND
,
2852 MF_MSG_POISONED_HUGE
,
2855 MF_MSG_NON_PMD_HUGE
,
2856 MF_MSG_UNMAP_FAILED
,
2857 MF_MSG_DIRTY_SWAPCACHE
,
2858 MF_MSG_CLEAN_SWAPCACHE
,
2859 MF_MSG_DIRTY_MLOCKED_LRU
,
2860 MF_MSG_CLEAN_MLOCKED_LRU
,
2861 MF_MSG_DIRTY_UNEVICTABLE_LRU
,
2862 MF_MSG_CLEAN_UNEVICTABLE_LRU
,
2865 MF_MSG_TRUNCATED_LRU
,
2872 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2873 extern void clear_huge_page(struct page
*page
,
2874 unsigned long addr_hint
,
2875 unsigned int pages_per_huge_page
);
2876 extern void copy_user_huge_page(struct page
*dst
, struct page
*src
,
2877 unsigned long addr_hint
,
2878 struct vm_area_struct
*vma
,
2879 unsigned int pages_per_huge_page
);
2880 extern long copy_huge_page_from_user(struct page
*dst_page
,
2881 const void __user
*usr_src
,
2882 unsigned int pages_per_huge_page
,
2883 bool allow_pagefault
);
2884 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2886 #ifdef CONFIG_DEBUG_PAGEALLOC
2887 extern unsigned int _debug_guardpage_minorder
;
2888 DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled
);
2890 static inline unsigned int debug_guardpage_minorder(void)
2892 return _debug_guardpage_minorder
;
2895 static inline bool debug_guardpage_enabled(void)
2897 return static_branch_unlikely(&_debug_guardpage_enabled
);
2900 static inline bool page_is_guard(struct page
*page
)
2902 if (!debug_guardpage_enabled())
2905 return PageGuard(page
);
2908 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2909 static inline bool debug_guardpage_enabled(void) { return false; }
2910 static inline bool page_is_guard(struct page
*page
) { return false; }
2911 #endif /* CONFIG_DEBUG_PAGEALLOC */
2913 #if MAX_NUMNODES > 1
2914 void __init
setup_nr_node_ids(void);
2916 static inline void setup_nr_node_ids(void) {}
2919 extern int memcmp_pages(struct page
*page1
, struct page
*page2
);
2921 static inline int pages_identical(struct page
*page1
, struct page
*page2
)
2923 return !memcmp_pages(page1
, page2
);
2926 #endif /* __KERNEL__ */
2927 #endif /* _LINUX_MM_H */