VirtualBox

Ticket #16966: mm.h

File mm.h, 74.4 KB (added by Dmitri Chubarov, 7 years ago)

include/linux/mm.h file provided with OpenSUSE 42.3

Line 
1#ifndef _LINUX_MM_H
2#define _LINUX_MM_H
3
4#include <linux/errno.h>
5
6#ifdef __KERNEL__
7
8#include <linux/mmdebug.h>
9#include <linux/gfp.h>
10#include <linux/bug.h>
11#include <linux/list.h>
12#include <linux/mmzone.h>
13#include <linux/rbtree.h>
14#include <linux/atomic.h>
15#include <linux/debug_locks.h>
16#include <linux/mm_types.h>
17#include <linux/range.h>
18#include <linux/pfn.h>
19#include <linux/percpu-refcount.h>
20#include <linux/bit_spinlock.h>
21#include <linux/shrinker.h>
22#include <linux/resource.h>
23#include <linux/page_ext.h>
24#include <linux/err.h>
25
26struct mempolicy;
27struct anon_vma;
28struct anon_vma_chain;
29struct file_ra_state;
30struct user_struct;
31struct writeback_control;
32struct bdi_writeback;
33
34#ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
35extern unsigned long max_mapnr;
36
37static inline void set_max_mapnr(unsigned long limit)
38{
39 max_mapnr = limit;
40}
41#else
42static inline void set_max_mapnr(unsigned long limit) { }
43#endif
44
45extern unsigned long totalram_pages;
46extern void * high_memory;
47extern int page_cluster;
48
49#ifdef CONFIG_SYSCTL
50extern int sysctl_legacy_va_layout;
51#else
52#define sysctl_legacy_va_layout 0
53#endif
54
55#include <asm/page.h>
56#include <asm/pgtable.h>
57#include <asm/processor.h>
58
59#ifndef __pa_symbol
60#define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
61#endif
62
63#ifndef lm_alias
64#define lm_alias(x) __va(__pa_symbol(x))
65#endif
66
67/*
68 * To prevent common memory management code establishing
69 * a zero page mapping on a read fault.
70 * This macro should be defined within <asm/pgtable.h>.
71 * s390 does this to prevent multiplexing of hardware bits
72 * related to the physical page in case of virtualization.
73 */
74#ifndef mm_forbids_zeropage
75#define mm_forbids_zeropage(X) (0)
76#endif
77
78extern unsigned long sysctl_user_reserve_kbytes;
79extern unsigned long sysctl_admin_reserve_kbytes;
80
81extern int sysctl_overcommit_memory;
82extern int sysctl_overcommit_ratio;
83extern unsigned long sysctl_overcommit_kbytes;
84
85extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
86 size_t *, loff_t *);
87extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
88 size_t *, loff_t *);
89
90#define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
91
92/* to align the pointer to the (next) page boundary */
93#define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
94
95/* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
96#define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)addr, PAGE_SIZE)
97
98/*
99 * Linux kernel virtual memory manager primitives.
100 * The idea being to have a "virtual" mm in the same way
101 * we have a virtual fs - giving a cleaner interface to the
102 * mm details, and allowing different kinds of memory mappings
103 * (from shared memory to executable loading to arbitrary
104 * mmap() functions).
105 */
106
107extern struct kmem_cache *vm_area_cachep;
108
109#ifndef CONFIG_MMU
110extern struct rb_root nommu_region_tree;
111extern struct rw_semaphore nommu_region_sem;
112
113extern unsigned int kobjsize(const void *objp);
114#endif
115
116/*
117 * vm_flags in vm_area_struct, see mm_types.h.
118 * When changing, update also include/trace/events/mmflags.h
119 */
120#define VM_NONE 0x00000000
121
122#define VM_READ 0x00000001 /* currently active flags */
123#define VM_WRITE 0x00000002
124#define VM_EXEC 0x00000004
125#define VM_SHARED 0x00000008
126
127/* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
128#define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
129#define VM_MAYWRITE 0x00000020
130#define VM_MAYEXEC 0x00000040
131#define VM_MAYSHARE 0x00000080
132
133#define VM_GROWSDOWN 0x00000100 /* general info on the segment */
134#define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
135#define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
136#define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
137#define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
138
139#define VM_LOCKED 0x00002000
140#define VM_IO 0x00004000 /* Memory mapped I/O or similar */
141
142 /* Used by sys_madvise() */
143#define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
144#define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
145
146#define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
147#define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
148#define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
149#define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
150#define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
151#define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
152#define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
153#define VM_ARCH_2 0x02000000
154#define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
155
156#ifdef CONFIG_MEM_SOFT_DIRTY
157# define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
158#else
159# define VM_SOFTDIRTY 0
160#endif
161
162#define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
163#define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
164#define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
165#define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
166
167#ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
168#define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
169#define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
170#define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
171#define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
172#define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
173#define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
174#define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
175#define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
176#endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
177
178#if defined(CONFIG_X86)
179# define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
180#if defined (CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS)
181# define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
182# define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
183# define VM_PKEY_BIT1 VM_HIGH_ARCH_1
184# define VM_PKEY_BIT2 VM_HIGH_ARCH_2
185# define VM_PKEY_BIT3 VM_HIGH_ARCH_3
186#endif
187#elif defined(CONFIG_PPC)
188# define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
189#elif defined(CONFIG_PARISC)
190# define VM_GROWSUP VM_ARCH_1
191#elif defined(CONFIG_METAG)
192# define VM_GROWSUP VM_ARCH_1
193#elif defined(CONFIG_IA64)
194# define VM_GROWSUP VM_ARCH_1
195#elif !defined(CONFIG_MMU)
196# define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
197#endif
198
199#if defined(CONFIG_X86)
200/* MPX specific bounds table or bounds directory */
201# define VM_MPX VM_ARCH_2
202#endif
203
204#ifndef VM_GROWSUP
205# define VM_GROWSUP VM_NONE
206#endif
207
208/* Bits set in the VMA until the stack is in its final location */
209#define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
210
211#ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
212#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
213#endif
214
215#ifdef CONFIG_STACK_GROWSUP
216#define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
217#else
218#define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
219#endif
220
221/*
222 * Special vmas that are non-mergable, non-mlock()able.
223 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
224 */
225#define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
226
227/* This mask defines which mm->def_flags a process can inherit its parent */
228#define VM_INIT_DEF_MASK VM_NOHUGEPAGE
229
230/* This mask is used to clear all the VMA flags used by mlock */
231#define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
232
233/*
234 * mapping from the currently active vm_flags protection bits (the
235 * low four bits) to a page protection mask..
236 */
237extern pgprot_t protection_map[16];
238
239#define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
240#define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
241#define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
242#define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
243#define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
244#define FAULT_FLAG_TRIED 0x20 /* Second try */
245#define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
246#define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */
247#define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */
248
249/*
250 * vm_fault is filled by the the pagefault handler and passed to the vma's
251 * ->fault function. The vma's ->fault is responsible for returning a bitmask
252 * of VM_FAULT_xxx flags that give details about how the fault was handled.
253 *
254 * MM layer fills up gfp_mask for page allocations but fault handler might
255 * alter it if its implementation requires a different allocation context.
256 *
257 * pgoff should be used in favour of virtual_address, if possible.
258 */
259struct vm_fault {
260 unsigned int flags; /* FAULT_FLAG_xxx flags */
261 gfp_t gfp_mask; /* gfp mask to be used for allocations */
262 pgoff_t pgoff; /* Logical page offset based on vma */
263 void __user *virtual_address; /* Faulting virtual address */
264
265 struct page *cow_page; /* Handler may choose to COW */
266 struct page *page; /* ->fault handlers should return a
267 * page here, unless VM_FAULT_NOPAGE
268 * is set (which is also implied by
269 * VM_FAULT_ERROR).
270 */
271 void *entry; /* ->fault handler can alternatively
272 * return locked DAX entry. In that
273 * case handler should return
274 * VM_FAULT_DAX_LOCKED and fill in
275 * entry here.
276 */
277 /* for ->map_pages() only */
278 pgoff_t max_pgoff; /* map pages for offset from pgoff till
279 * max_pgoff inclusive */
280 pte_t *pte; /* pte entry associated with ->pgoff */
281 void *suse_kabi_padding;
282};
283
284/*
285 * These are the virtual MM functions - opening of an area, closing and
286 * unmapping it (needed to keep files on disk up-to-date etc), pointer
287 * to the functions called when a no-page or a wp-page exception occurs.
288 */
289struct vm_operations_struct {
290 void (*open)(struct vm_area_struct * area);
291 void (*close)(struct vm_area_struct * area);
292 int (*mremap)(struct vm_area_struct * area);
293 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
294 int (*pmd_fault)(struct vm_area_struct *, unsigned long address,
295 pmd_t *, unsigned int flags);
296 void (*map_pages)(struct vm_area_struct *vma, struct vm_fault *vmf);
297
298 /* notification that a previously read-only page is about to become
299 * writable, if an error is returned it will cause a SIGBUS */
300 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
301
302 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
303 int (*pfn_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
304
305 /* called by access_process_vm when get_user_pages() fails, typically
306 * for use by special VMAs that can switch between memory and hardware
307 */
308 int (*access)(struct vm_area_struct *vma, unsigned long addr,
309 void *buf, int len, int write);
310
311 /* Called by the /proc/PID/maps code to ask the vma whether it
312 * has a special name. Returning non-NULL will also cause this
313 * vma to be dumped unconditionally. */
314 const char *(*name)(struct vm_area_struct *vma);
315
316#ifdef CONFIG_NUMA
317 /*
318 * set_policy() op must add a reference to any non-NULL @new mempolicy
319 * to hold the policy upon return. Caller should pass NULL @new to
320 * remove a policy and fall back to surrounding context--i.e. do not
321 * install a MPOL_DEFAULT policy, nor the task or system default
322 * mempolicy.
323 */
324 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
325
326 /*
327 * get_policy() op must add reference [mpol_get()] to any policy at
328 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
329 * in mm/mempolicy.c will do this automatically.
330 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
331 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
332 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
333 * must return NULL--i.e., do not "fallback" to task or system default
334 * policy.
335 */
336 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
337 unsigned long addr);
338#endif
339 /*
340 * Called by vm_normal_page() for special PTEs to find the
341 * page for @addr. This is useful if the default behavior
342 * (using pte_page()) would not find the correct page.
343 */
344 struct page *(*find_special_page)(struct vm_area_struct *vma,
345 unsigned long addr);
346};
347
348struct mmu_gather;
349struct inode;
350
351#define page_private(page) ((page)->private)
352#define set_page_private(page, v) ((page)->private = (v))
353
354#if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
355static inline int pmd_devmap(pmd_t pmd)
356{
357 return 0;
358}
359#endif
360
361/*
362 * FIXME: take this include out, include page-flags.h in
363 * files which need it (119 of them)
364 */
365#include <linux/page-flags.h>
366#include <linux/huge_mm.h>
367
368/*
369 * Methods to modify the page usage count.
370 *
371 * What counts for a page usage:
372 * - cache mapping (page->mapping)
373 * - private data (page->private)
374 * - page mapped in a task's page tables, each mapping
375 * is counted separately
376 *
377 * Also, many kernel routines increase the page count before a critical
378 * routine so they can be sure the page doesn't go away from under them.
379 */
380
381/*
382 * Drop a ref, return true if the refcount fell to zero (the page has no users)
383 */
384static inline int put_page_testzero(struct page *page)
385{
386 VM_BUG_ON_PAGE(atomic_read(&page->_count) == 0, page);
387 return atomic_dec_and_test(&page->_count);
388}
389
390/*
391 * Try to grab a ref unless the page has a refcount of zero, return false if
392 * that is the case.
393 * This can be called when MMU is off so it must not access
394 * any of the virtual mappings.
395 */
396static inline int get_page_unless_zero(struct page *page)
397{
398 return atomic_inc_not_zero(&page->_count);
399}
400
401extern int page_is_ram(unsigned long pfn);
402
403enum {
404 REGION_INTERSECTS,
405 REGION_DISJOINT,
406 REGION_MIXED,
407};
408
409int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
410 unsigned long desc);
411
412/* Support for virtually mapped pages */
413struct page *vmalloc_to_page(const void *addr);
414unsigned long vmalloc_to_pfn(const void *addr);
415
416/*
417 * Determine if an address is within the vmalloc range
418 *
419 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
420 * is no special casing required.
421 */
422static inline int is_vmalloc_addr(const void *x)
423{
424#ifdef CONFIG_MMU
425 unsigned long addr = (unsigned long)x;
426
427 return addr >= VMALLOC_START && addr < VMALLOC_END;
428#else
429 return 0;
430#endif
431}
432#ifdef CONFIG_MMU
433extern int is_vmalloc_or_module_addr(const void *x);
434#else
435static inline int is_vmalloc_or_module_addr(const void *x)
436{
437 return 0;
438}
439#endif
440
441extern void kvfree(const void *addr);
442
443static inline void compound_lock(struct page *page)
444{
445#ifdef CONFIG_TRANSPARENT_HUGEPAGE
446 VM_BUG_ON_PAGE(PageSlab(page), page);
447 bit_spin_lock(PG_compound_lock, &page->flags);
448#endif
449}
450
451static inline void compound_unlock(struct page *page)
452{
453#ifdef CONFIG_TRANSPARENT_HUGEPAGE
454 VM_BUG_ON_PAGE(PageSlab(page), page);
455 bit_spin_unlock(PG_compound_lock, &page->flags);
456#endif
457}
458
459static inline unsigned long compound_lock_irqsave(struct page *page)
460{
461 unsigned long uninitialized_var(flags);
462#ifdef CONFIG_TRANSPARENT_HUGEPAGE
463 local_irq_save(flags);
464 compound_lock(page);
465#endif
466 return flags;
467}
468
469static inline void compound_unlock_irqrestore(struct page *page,
470 unsigned long flags)
471{
472#ifdef CONFIG_TRANSPARENT_HUGEPAGE
473 compound_unlock(page);
474 local_irq_restore(flags);
475#endif
476}
477
478/*
479 * The atomic page->_mapcount, starts from -1: so that transitions
480 * both from it and to it can be tracked, using atomic_inc_and_test
481 * and atomic_add_negative(-1).
482 */
483static inline void page_mapcount_reset(struct page *page)
484{
485 atomic_set(&(page)->_mapcount, -1);
486}
487
488static inline int page_mapcount(struct page *page)
489{
490 VM_BUG_ON_PAGE(PageSlab(page), page);
491 return atomic_read(&page->_mapcount) + 1;
492}
493
494static inline int page_count(struct page *page)
495{
496 return atomic_read(&compound_head(page)->_count);
497}
498
499static inline bool __compound_tail_refcounted(struct page *page)
500{
501 return PageAnon(page) && !PageSlab(page) && !PageHeadHuge(page);
502}
503
504/*
505 * This takes a head page as parameter and tells if the
506 * tail page reference counting can be skipped.
507 *
508 * For this to be safe, PageSlab and PageHeadHuge must remain true on
509 * any given page where they return true here, until all tail pins
510 * have been released.
511 */
512static inline bool compound_tail_refcounted(struct page *page)
513{
514 VM_BUG_ON_PAGE(!PageHead(page), page);
515 return __compound_tail_refcounted(page);
516}
517
518static inline void get_huge_page_tail(struct page *page)
519{
520 /*
521 * __split_huge_page_refcount() cannot run from under us.
522 */
523 VM_BUG_ON_PAGE(!PageTail(page), page);
524 VM_BUG_ON_PAGE(page_mapcount(page) < 0, page);
525 VM_BUG_ON_PAGE(atomic_read(&page->_count) != 0, page);
526 if (compound_tail_refcounted(compound_head(page)))
527 atomic_inc(&page->_mapcount);
528}
529
530extern bool __get_page_tail(struct page *page);
531
532static inline struct page *virt_to_head_page(const void *x)
533{
534 struct page *page = virt_to_page(x);
535
536 return compound_head(page);
537}
538
539/*
540 * Setup the page count before being freed into the page allocator for
541 * the first time (boot or memory hotplug)
542 */
543static inline void init_page_count(struct page *page)
544{
545 atomic_set(&page->_count, 1);
546}
547
548void put_page(struct page *page);
549void put_pages_list(struct list_head *pages);
550
551void split_page(struct page *page, unsigned int order);
552
553/*
554 * Compound pages have a destructor function. Provide a
555 * prototype for that function and accessor functions.
556 * These are _only_ valid on the head of a compound page.
557 */
558typedef void compound_page_dtor(struct page *);
559
560/* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
561enum compound_dtor_id {
562 NULL_COMPOUND_DTOR,
563 COMPOUND_PAGE_DTOR,
564#ifdef CONFIG_HUGETLB_PAGE
565 HUGETLB_PAGE_DTOR,
566#endif
567 NR_COMPOUND_DTORS,
568};
569extern compound_page_dtor * const compound_page_dtors[];
570
571static inline void set_compound_page_dtor(struct page *page,
572 enum compound_dtor_id compound_dtor)
573{
574 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
575 page[1].compound_dtor = compound_dtor;
576}
577
578static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
579{
580 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
581 return compound_page_dtors[page[1].compound_dtor];
582}
583
584static inline unsigned int compound_order(struct page *page)
585{
586 if (!PageHead(page))
587 return 0;
588 return page[1].compound_order;
589}
590
591static inline void set_compound_order(struct page *page, unsigned int order)
592{
593 page[1].compound_order = order;
594}
595
596#ifdef CONFIG_MMU
597/*
598 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
599 * servicing faults for write access. In the normal case, do always want
600 * pte_mkwrite. But get_user_pages can cause write faults for mappings
601 * that do not have writing enabled, when used by access_process_vm.
602 */
603static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
604{
605 if (likely(vma->vm_flags & VM_WRITE))
606 pte = pte_mkwrite(pte);
607 return pte;
608}
609
610void do_set_pte(struct vm_area_struct *vma, unsigned long address,
611 struct page *page, pte_t *pte, bool write, bool anon);
612#endif
613
614/*
615 * Multiple processes may "see" the same page. E.g. for untouched
616 * mappings of /dev/null, all processes see the same page full of
617 * zeroes, and text pages of executables and shared libraries have
618 * only one copy in memory, at most, normally.
619 *
620 * For the non-reserved pages, page_count(page) denotes a reference count.
621 * page_count() == 0 means the page is free. page->lru is then used for
622 * freelist management in the buddy allocator.
623 * page_count() > 0 means the page has been allocated.
624 *
625 * Pages are allocated by the slab allocator in order to provide memory
626 * to kmalloc and kmem_cache_alloc. In this case, the management of the
627 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
628 * unless a particular usage is carefully commented. (the responsibility of
629 * freeing the kmalloc memory is the caller's, of course).
630 *
631 * A page may be used by anyone else who does a __get_free_page().
632 * In this case, page_count still tracks the references, and should only
633 * be used through the normal accessor functions. The top bits of page->flags
634 * and page->virtual store page management information, but all other fields
635 * are unused and could be used privately, carefully. The management of this
636 * page is the responsibility of the one who allocated it, and those who have
637 * subsequently been given references to it.
638 *
639 * The other pages (we may call them "pagecache pages") are completely
640 * managed by the Linux memory manager: I/O, buffers, swapping etc.
641 * The following discussion applies only to them.
642 *
643 * A pagecache page contains an opaque `private' member, which belongs to the
644 * page's address_space. Usually, this is the address of a circular list of
645 * the page's disk buffers. PG_private must be set to tell the VM to call
646 * into the filesystem to release these pages.
647 *
648 * A page may belong to an inode's memory mapping. In this case, page->mapping
649 * is the pointer to the inode, and page->index is the file offset of the page,
650 * in units of PAGE_CACHE_SIZE.
651 *
652 * If pagecache pages are not associated with an inode, they are said to be
653 * anonymous pages. These may become associated with the swapcache, and in that
654 * case PG_swapcache is set, and page->private is an offset into the swapcache.
655 *
656 * In either case (swapcache or inode backed), the pagecache itself holds one
657 * reference to the page. Setting PG_private should also increment the
658 * refcount. The each user mapping also has a reference to the page.
659 *
660 * The pagecache pages are stored in a per-mapping radix tree, which is
661 * rooted at mapping->page_tree, and indexed by offset.
662 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
663 * lists, we instead now tag pages as dirty/writeback in the radix tree.
664 *
665 * All pagecache pages may be subject to I/O:
666 * - inode pages may need to be read from disk,
667 * - inode pages which have been modified and are MAP_SHARED may need
668 * to be written back to the inode on disk,
669 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
670 * modified may need to be swapped out to swap space and (later) to be read
671 * back into memory.
672 */
673
674/*
675 * The zone field is never updated after free_area_init_core()
676 * sets it, so none of the operations on it need to be atomic.
677 */
678
679/* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
680#define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
681#define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
682#define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
683#define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
684
685/*
686 * Define the bit shifts to access each section. For non-existent
687 * sections we define the shift as 0; that plus a 0 mask ensures
688 * the compiler will optimise away reference to them.
689 */
690#define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
691#define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
692#define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
693#define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
694
695/* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
696#ifdef NODE_NOT_IN_PAGE_FLAGS
697#define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
698#define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
699 SECTIONS_PGOFF : ZONES_PGOFF)
700#else
701#define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
702#define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
703 NODES_PGOFF : ZONES_PGOFF)
704#endif
705
706#define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
707
708#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
709#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
710#endif
711
712#define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
713#define NODES_MASK ((1UL << NODES_WIDTH) - 1)
714#define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
715#define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
716#define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
717
718static inline enum zone_type page_zonenum(const struct page *page)
719{
720 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
721}
722
723#ifdef CONFIG_ZONE_DEVICE
724void get_zone_device_page(struct page *page);
725void put_zone_device_page(struct page *page);
726static inline bool is_zone_device_page(const struct page *page)
727{
728 return page_zonenum(page) == ZONE_DEVICE;
729}
730#else
731static inline void get_zone_device_page(struct page *page)
732{
733}
734static inline void put_zone_device_page(struct page *page)
735{
736}
737static inline bool is_zone_device_page(const struct page *page)
738{
739 return false;
740}
741#endif
742
743static inline void get_page(struct page *page)
744{
745 if (unlikely(PageTail(page)))
746 if (likely(__get_page_tail(page))) {
747 if (unlikely(is_zone_device_page(page)))
748 get_zone_device_page(page);
749 return;
750 }
751 /*
752 * Getting a normal page or the head of a compound page
753 * requires to already have an elevated page->_count.
754 */
755 VM_BUG_ON_PAGE(atomic_read(&page->_count) <= 0, page);
756 atomic_inc(&page->_count);
757
758 if (unlikely(is_zone_device_page(page)))
759 get_zone_device_page(page);
760}
761
762#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
763#define SECTION_IN_PAGE_FLAGS
764#endif
765
766/*
767 * The identification function is mainly used by the buddy allocator for
768 * determining if two pages could be buddies. We are not really identifying
769 * the zone since we could be using the section number id if we do not have
770 * node id available in page flags.
771 * We only guarantee that it will return the same value for two combinable
772 * pages in a zone.
773 */
774static inline int page_zone_id(struct page *page)
775{
776 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
777}
778
779static inline int zone_to_nid(struct zone *zone)
780{
781#ifdef CONFIG_NUMA
782 return zone->node;
783#else
784 return 0;
785#endif
786}
787
788#ifdef NODE_NOT_IN_PAGE_FLAGS
789extern int page_to_nid(const struct page *page);
790#else
791static inline int page_to_nid(const struct page *page)
792{
793 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
794}
795#endif
796
797#ifdef CONFIG_NUMA_BALANCING
798static inline int cpu_pid_to_cpupid(int cpu, int pid)
799{
800 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
801}
802
803static inline int cpupid_to_pid(int cpupid)
804{
805 return cpupid & LAST__PID_MASK;
806}
807
808static inline int cpupid_to_cpu(int cpupid)
809{
810 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
811}
812
813static inline int cpupid_to_nid(int cpupid)
814{
815 return cpu_to_node(cpupid_to_cpu(cpupid));
816}
817
818static inline bool cpupid_pid_unset(int cpupid)
819{
820 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
821}
822
823static inline bool cpupid_cpu_unset(int cpupid)
824{
825 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
826}
827
828static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
829{
830 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
831}
832
833#define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
834#ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
835static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
836{
837 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
838}
839
840static inline int page_cpupid_last(struct page *page)
841{
842 return page->_last_cpupid;
843}
844static inline void page_cpupid_reset_last(struct page *page)
845{
846 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
847}
848#else
849static inline int page_cpupid_last(struct page *page)
850{
851 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
852}
853
854extern int page_cpupid_xchg_last(struct page *page, int cpupid);
855
856static inline void page_cpupid_reset_last(struct page *page)
857{
858 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
859}
860#endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
861#else /* !CONFIG_NUMA_BALANCING */
862static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
863{
864 return page_to_nid(page); /* XXX */
865}
866
867static inline int page_cpupid_last(struct page *page)
868{
869 return page_to_nid(page); /* XXX */
870}
871
872static inline int cpupid_to_nid(int cpupid)
873{
874 return -1;
875}
876
877static inline int cpupid_to_pid(int cpupid)
878{
879 return -1;
880}
881
882static inline int cpupid_to_cpu(int cpupid)
883{
884 return -1;
885}
886
887static inline int cpu_pid_to_cpupid(int nid, int pid)
888{
889 return -1;
890}
891
892static inline bool cpupid_pid_unset(int cpupid)
893{
894 return 1;
895}
896
897static inline void page_cpupid_reset_last(struct page *page)
898{
899}
900
901static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
902{
903 return false;
904}
905#endif /* CONFIG_NUMA_BALANCING */
906
907static inline struct zone *page_zone(const struct page *page)
908{
909 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
910}
911
912#ifdef SECTION_IN_PAGE_FLAGS
913static inline void set_page_section(struct page *page, unsigned long section)
914{
915 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
916 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
917}
918
919static inline unsigned long page_to_section(const struct page *page)
920{
921 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
922}
923#endif
924
925static inline void set_page_zone(struct page *page, enum zone_type zone)
926{
927 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
928 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
929}
930
931static inline void set_page_node(struct page *page, unsigned long node)
932{
933 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
934 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
935}
936
937static inline void set_page_links(struct page *page, enum zone_type zone,
938 unsigned long node, unsigned long pfn)
939{
940 set_page_zone(page, zone);
941 set_page_node(page, node);
942#ifdef SECTION_IN_PAGE_FLAGS
943 set_page_section(page, pfn_to_section_nr(pfn));
944#endif
945}
946
947#ifdef CONFIG_MEMCG
948static inline struct mem_cgroup *page_memcg(struct page *page)
949{
950 return page->mem_cgroup;
951}
952
953static inline void set_page_memcg(struct page *page, struct mem_cgroup *memcg)
954{
955 page->mem_cgroup = memcg;
956}
957#else
958static inline struct mem_cgroup *page_memcg(struct page *page)
959{
960 return NULL;
961}
962
963static inline void set_page_memcg(struct page *page, struct mem_cgroup *memcg)
964{
965}
966#endif
967
968/*
969 * Some inline functions in vmstat.h depend on page_zone()
970 */
971#include <linux/vmstat.h>
972
973static __always_inline void *lowmem_page_address(const struct page *page)
974{
975 return __va(PFN_PHYS(page_to_pfn(page)));
976}
977
978#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
979#define HASHED_PAGE_VIRTUAL
980#endif
981
982#if defined(WANT_PAGE_VIRTUAL)
983static inline void *page_address(const struct page *page)
984{
985 return page->virtual;
986}
987static inline void set_page_address(struct page *page, void *address)
988{
989 page->virtual = address;
990}
991#define page_address_init() do { } while(0)
992#endif
993
994#if defined(HASHED_PAGE_VIRTUAL)
995void *page_address(const struct page *page);
996void set_page_address(struct page *page, void *virtual);
997void page_address_init(void);
998#endif
999
1000#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1001#define page_address(page) lowmem_page_address(page)
1002#define set_page_address(page, address) do { } while(0)
1003#define page_address_init() do { } while(0)
1004#endif
1005
1006extern void *page_rmapping(struct page *page);
1007extern struct anon_vma *page_anon_vma(struct page *page);
1008extern struct address_space *page_mapping(struct page *page);
1009
1010extern struct address_space *__page_file_mapping(struct page *);
1011
1012static inline
1013struct address_space *page_file_mapping(struct page *page)
1014{
1015 if (unlikely(PageSwapCache(page)))
1016 return __page_file_mapping(page);
1017
1018 return page->mapping;
1019}
1020
1021/*
1022 * Return the pagecache index of the passed page. Regular pagecache pages
1023 * use ->index whereas swapcache pages use ->private
1024 */
1025static inline pgoff_t page_index(struct page *page)
1026{
1027 if (unlikely(PageSwapCache(page)))
1028 return page_private(page);
1029 return page->index;
1030}
1031
1032extern pgoff_t __page_file_index(struct page *page);
1033
1034/*
1035 * Return the file index of the page. Regular pagecache pages use ->index
1036 * whereas swapcache pages use swp_offset(->private)
1037 */
1038static inline pgoff_t page_file_index(struct page *page)
1039{
1040 if (unlikely(PageSwapCache(page)))
1041 return __page_file_index(page);
1042
1043 return page->index;
1044}
1045
1046/*
1047 * Return true if this page is mapped into pagetables.
1048 */
1049static inline int page_mapped(struct page *page)
1050{
1051 return atomic_read(&(page)->_mapcount) >= 0;
1052}
1053
1054/*
1055 * Return true only if the page has been allocated with
1056 * ALLOC_NO_WATERMARKS and the low watermark was not
1057 * met implying that the system is under some pressure.
1058 */
1059static inline bool page_is_pfmemalloc(struct page *page)
1060{
1061 /*
1062 * Page index cannot be this large so this must be
1063 * a pfmemalloc page.
1064 */
1065 return page->index == -1UL;
1066}
1067
1068/*
1069 * Only to be called by the page allocator on a freshly allocated
1070 * page.
1071 */
1072static inline void set_page_pfmemalloc(struct page *page)
1073{
1074 page->index = -1UL;
1075}
1076
1077static inline void clear_page_pfmemalloc(struct page *page)
1078{
1079 page->index = 0;
1080}
1081
1082/*
1083 * Different kinds of faults, as returned by handle_mm_fault().
1084 * Used to decide whether a process gets delivered SIGBUS or
1085 * just gets major/minor fault counters bumped up.
1086 */
1087
1088#define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
1089
1090#define VM_FAULT_OOM 0x0001
1091#define VM_FAULT_SIGBUS 0x0002
1092#define VM_FAULT_MAJOR 0x0004
1093#define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1094#define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1095#define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1096#define VM_FAULT_SIGSEGV 0x0040
1097
1098#define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1099#define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1100#define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1101#define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1102#define VM_FAULT_DAX_LOCKED 0x1000 /* ->fault has locked DAX entry */
1103
1104#define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1105
1106#define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1107 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1108 VM_FAULT_FALLBACK)
1109
1110/* Encode hstate index for a hwpoisoned large page */
1111#define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1112#define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1113
1114/*
1115 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1116 */
1117extern void pagefault_out_of_memory(void);
1118
1119#define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1120
1121/*
1122 * Flags passed to show_mem() and show_free_areas() to suppress output in
1123 * various contexts.
1124 */
1125#define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1126
1127extern void show_free_areas(unsigned int flags);
1128extern bool skip_free_areas_node(unsigned int flags, int nid);
1129
1130int shmem_zero_setup(struct vm_area_struct *);
1131#ifdef CONFIG_SHMEM
1132bool shmem_mapping(struct address_space *mapping);
1133#else
1134static inline bool shmem_mapping(struct address_space *mapping)
1135{
1136 return false;
1137}
1138#endif
1139
1140extern int can_do_mlock(void);
1141extern int user_shm_lock(size_t, struct user_struct *);
1142extern void user_shm_unlock(size_t, struct user_struct *);
1143
1144/*
1145 * Parameter block passed down to zap_pte_range in exceptional cases.
1146 */
1147struct zap_details {
1148 struct address_space *check_mapping; /* Check page->mapping if set */
1149 pgoff_t first_index; /* Lowest page->index to unmap */
1150 pgoff_t last_index; /* Highest page->index to unmap */
1151};
1152
1153struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1154 pte_t pte);
1155struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1156 pmd_t pmd);
1157
1158int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1159 unsigned long size);
1160void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1161 unsigned long size, struct zap_details *);
1162void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1163 unsigned long start, unsigned long end);
1164
1165/**
1166 * mm_walk - callbacks for walk_page_range
1167 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1168 * this handler is required to be able to handle
1169 * pmd_trans_huge() pmds. They may simply choose to
1170 * split_huge_page() instead of handling it explicitly.
1171 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1172 * @pte_hole: if set, called for each hole at all levels
1173 * @hugetlb_entry: if set, called for each hugetlb entry
1174 * @test_walk: caller specific callback function to determine whether
1175 * we walk over the current vma or not. A positive returned
1176 * value means "do page table walk over the current vma,"
1177 * and a negative one means "abort current page table walk
1178 * right now." 0 means "skip the current vma."
1179 * @mm: mm_struct representing the target process of page table walk
1180 * @vma: vma currently walked (NULL if walking outside vmas)
1181 * @private: private data for callbacks' usage
1182 *
1183 * (see the comment on walk_page_range() for more details)
1184 */
1185struct mm_walk {
1186 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1187 unsigned long next, struct mm_walk *walk);
1188 int (*pte_entry)(pte_t *pte, unsigned long addr,
1189 unsigned long next, struct mm_walk *walk);
1190 int (*pte_hole)(unsigned long addr, unsigned long next,
1191 struct mm_walk *walk);
1192 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1193 unsigned long addr, unsigned long next,
1194 struct mm_walk *walk);
1195 int (*test_walk)(unsigned long addr, unsigned long next,
1196 struct mm_walk *walk);
1197 struct mm_struct *mm;
1198 struct vm_area_struct *vma;
1199 void *private;
1200};
1201
1202int walk_page_range(unsigned long addr, unsigned long end,
1203 struct mm_walk *walk);
1204int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1205void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1206 unsigned long end, unsigned long floor, unsigned long ceiling);
1207int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1208 struct vm_area_struct *vma);
1209void unmap_mapping_range(struct address_space *mapping,
1210 loff_t const holebegin, loff_t const holelen, int even_cows);
1211int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1212 unsigned long *pfn);
1213int follow_phys(struct vm_area_struct *vma, unsigned long address,
1214 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1215int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1216 void *buf, int len, int write);
1217
1218static inline void unmap_shared_mapping_range(struct address_space *mapping,
1219 loff_t const holebegin, loff_t const holelen)
1220{
1221 unmap_mapping_range(mapping, holebegin, holelen, 0);
1222}
1223
1224extern void truncate_pagecache(struct inode *inode, loff_t new);
1225extern void truncate_setsize(struct inode *inode, loff_t newsize);
1226void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1227void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1228int truncate_inode_page(struct address_space *mapping, struct page *page);
1229int generic_error_remove_page(struct address_space *mapping, struct page *page);
1230int invalidate_inode_page(struct page *page);
1231
1232#ifdef CONFIG_MMU
1233extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
1234 unsigned long address, unsigned int flags);
1235extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1236 unsigned long address, unsigned int fault_flags);
1237#else
1238static inline int handle_mm_fault(struct mm_struct *mm,
1239 struct vm_area_struct *vma, unsigned long address,
1240 unsigned int flags)
1241{
1242 /* should never happen if there's no MMU */
1243 BUG();
1244 return VM_FAULT_SIGBUS;
1245}
1246static inline int fixup_user_fault(struct task_struct *tsk,
1247 struct mm_struct *mm, unsigned long address,
1248 unsigned int fault_flags)
1249{
1250 /* should never happen if there's no MMU */
1251 BUG();
1252 return -EFAULT;
1253}
1254#endif
1255
1256extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1257extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1258 void *buf, int len, int write);
1259
1260long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1261 unsigned long start, unsigned long nr_pages,
1262 unsigned int gup_flags, struct page **pages,
1263 struct vm_area_struct **vmas, int *locked);
1264long get_user_pages(unsigned long start, unsigned long nr_pages,
1265 unsigned int gup_flags, struct page **pages,
1266 struct vm_area_struct **vmas);
1267long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1268 unsigned int gup_flags, struct page **pages, int *locked);
1269long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1270 struct page **pages, unsigned int gup_flags);
1271int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1272 struct page **pages);
1273
1274/* Container for pinned pfns / pages */
1275struct frame_vector {
1276 unsigned int nr_allocated; /* Number of frames we have space for */
1277 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1278 bool got_ref; /* Did we pin pages by getting page ref? */
1279 bool is_pfns; /* Does array contain pages or pfns? */
1280 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1281 * pfns_vector_pages() or pfns_vector_pfns()
1282 * for access */
1283};
1284
1285struct frame_vector *frame_vector_create(unsigned int nr_frames);
1286void frame_vector_destroy(struct frame_vector *vec);
1287int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1288 bool write, bool force, struct frame_vector *vec);
1289void put_vaddr_frames(struct frame_vector *vec);
1290int frame_vector_to_pages(struct frame_vector *vec);
1291void frame_vector_to_pfns(struct frame_vector *vec);
1292
1293static inline unsigned int frame_vector_count(struct frame_vector *vec)
1294{
1295 return vec->nr_frames;
1296}
1297
1298static inline struct page **frame_vector_pages(struct frame_vector *vec)
1299{
1300 if (vec->is_pfns) {
1301 int err = frame_vector_to_pages(vec);
1302
1303 if (err)
1304 return ERR_PTR(err);
1305 }
1306 return (struct page **)(vec->ptrs);
1307}
1308
1309static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1310{
1311 if (!vec->is_pfns)
1312 frame_vector_to_pfns(vec);
1313 return (unsigned long *)(vec->ptrs);
1314}
1315
1316struct kvec;
1317int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1318 struct page **pages);
1319int get_kernel_page(unsigned long start, int write, struct page **pages);
1320struct page *get_dump_page(unsigned long addr);
1321
1322extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1323extern void do_invalidatepage(struct page *page, unsigned int offset,
1324 unsigned int length);
1325
1326int __set_page_dirty_nobuffers(struct page *page);
1327int __set_page_dirty_no_writeback(struct page *page);
1328int redirty_page_for_writepage(struct writeback_control *wbc,
1329 struct page *page);
1330void account_page_dirtied(struct page *page, struct address_space *mapping,
1331 struct mem_cgroup *memcg);
1332void account_page_cleaned(struct page *page, struct address_space *mapping,
1333 struct mem_cgroup *memcg, struct bdi_writeback *wb);
1334int set_page_dirty(struct page *page);
1335int set_page_dirty_lock(struct page *page);
1336void cancel_dirty_page(struct page *page);
1337int clear_page_dirty_for_io(struct page *page);
1338
1339int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1340
1341static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1342{
1343 return !vma->vm_ops;
1344}
1345
1346int vma_is_stack_for_task(struct vm_area_struct *vma, struct task_struct *t);
1347
1348extern unsigned long move_page_tables(struct vm_area_struct *vma,
1349 unsigned long old_addr, struct vm_area_struct *new_vma,
1350 unsigned long new_addr, unsigned long len,
1351 bool need_rmap_locks);
1352extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1353 unsigned long end, pgprot_t newprot,
1354 int dirty_accountable, int prot_numa);
1355extern int mprotect_fixup(struct vm_area_struct *vma,
1356 struct vm_area_struct **pprev, unsigned long start,
1357 unsigned long end, unsigned long newflags);
1358
1359/*
1360 * doesn't attempt to fault and will return short.
1361 */
1362int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1363 struct page **pages);
1364/*
1365 * per-process(per-mm_struct) statistics.
1366 */
1367static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1368{
1369 long val = atomic_long_read(&mm->rss_stat.count[member]);
1370
1371#ifdef SPLIT_RSS_COUNTING
1372 /*
1373 * counter is updated in asynchronous manner and may go to minus.
1374 * But it's never be expected number for users.
1375 */
1376 if (val < 0)
1377 val = 0;
1378#endif
1379 return (unsigned long)val;
1380}
1381
1382static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1383{
1384 atomic_long_add(value, &mm->rss_stat.count[member]);
1385}
1386
1387static inline void inc_mm_counter(struct mm_struct *mm, int member)
1388{
1389 atomic_long_inc(&mm->rss_stat.count[member]);
1390}
1391
1392static inline void dec_mm_counter(struct mm_struct *mm, int member)
1393{
1394 atomic_long_dec(&mm->rss_stat.count[member]);
1395}
1396
1397/* Optimized variant when page is already known not to be PageAnon */
1398static inline int mm_counter_file(struct page *page)
1399{
1400 if (PageSwapBacked(page))
1401 return MM_SHMEMPAGES;
1402 return MM_FILEPAGES;
1403}
1404
1405static inline int mm_counter(struct page *page)
1406{
1407 if (PageAnon(page))
1408 return MM_ANONPAGES;
1409 return mm_counter_file(page);
1410}
1411
1412static inline unsigned long get_mm_rss(struct mm_struct *mm)
1413{
1414 return get_mm_counter(mm, MM_FILEPAGES) +
1415 get_mm_counter(mm, MM_ANONPAGES) +
1416 get_mm_counter(mm, MM_SHMEMPAGES);
1417}
1418
1419static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1420{
1421 return max(mm->hiwater_rss, get_mm_rss(mm));
1422}
1423
1424static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1425{
1426 return max(mm->hiwater_vm, mm->total_vm);
1427}
1428
1429static inline void update_hiwater_rss(struct mm_struct *mm)
1430{
1431 unsigned long _rss = get_mm_rss(mm);
1432
1433 if ((mm)->hiwater_rss < _rss)
1434 (mm)->hiwater_rss = _rss;
1435}
1436
1437static inline void update_hiwater_vm(struct mm_struct *mm)
1438{
1439 if (mm->hiwater_vm < mm->total_vm)
1440 mm->hiwater_vm = mm->total_vm;
1441}
1442
1443static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1444{
1445 mm->hiwater_rss = get_mm_rss(mm);
1446}
1447
1448static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1449 struct mm_struct *mm)
1450{
1451 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1452
1453 if (*maxrss < hiwater_rss)
1454 *maxrss = hiwater_rss;
1455}
1456
1457#if defined(SPLIT_RSS_COUNTING)
1458void sync_mm_rss(struct mm_struct *mm);
1459#else
1460static inline void sync_mm_rss(struct mm_struct *mm)
1461{
1462}
1463#endif
1464
1465#ifndef __HAVE_ARCH_PTE_DEVMAP
1466static inline int pte_devmap(pte_t pte)
1467{
1468 return 0;
1469}
1470#endif
1471
1472int vma_wants_writenotify(struct vm_area_struct *vma);
1473
1474extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1475 spinlock_t **ptl);
1476static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1477 spinlock_t **ptl)
1478{
1479 pte_t *ptep;
1480 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1481 return ptep;
1482}
1483
1484#ifdef __PAGETABLE_PUD_FOLDED
1485static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1486 unsigned long address)
1487{
1488 return 0;
1489}
1490#else
1491int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1492#endif
1493
1494#if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1495static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1496 unsigned long address)
1497{
1498 return 0;
1499}
1500
1501static inline void mm_nr_pmds_init(struct mm_struct *mm) {}
1502
1503static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1504{
1505 return 0;
1506}
1507
1508static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1509static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1510
1511#else
1512int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1513
1514static inline void mm_nr_pmds_init(struct mm_struct *mm)
1515{
1516 atomic_long_set(&mm->nr_pmds, 0);
1517}
1518
1519static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1520{
1521 return atomic_long_read(&mm->nr_pmds);
1522}
1523
1524static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1525{
1526 atomic_long_inc(&mm->nr_pmds);
1527}
1528
1529static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1530{
1531 atomic_long_dec(&mm->nr_pmds);
1532}
1533#endif
1534
1535int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1536 pmd_t *pmd, unsigned long address);
1537int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1538
1539/*
1540 * The following ifdef needed to get the 4level-fixup.h header to work.
1541 * Remove it when 4level-fixup.h has been removed.
1542 */
1543#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1544static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1545{
1546 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1547 NULL: pud_offset(pgd, address);
1548}
1549
1550static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1551{
1552 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1553 NULL: pmd_offset(pud, address);
1554}
1555#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1556
1557#if USE_SPLIT_PTE_PTLOCKS
1558#if ALLOC_SPLIT_PTLOCKS
1559void __init ptlock_cache_init(void);
1560extern bool ptlock_alloc(struct page *page);
1561extern void ptlock_free(struct page *page);
1562
1563static inline spinlock_t *ptlock_ptr(struct page *page)
1564{
1565 return page->ptl;
1566}
1567#else /* ALLOC_SPLIT_PTLOCKS */
1568static inline void ptlock_cache_init(void)
1569{
1570}
1571
1572static inline bool ptlock_alloc(struct page *page)
1573{
1574 return true;
1575}
1576
1577static inline void ptlock_free(struct page *page)
1578{
1579}
1580
1581static inline spinlock_t *ptlock_ptr(struct page *page)
1582{
1583 return &page->ptl;
1584}
1585#endif /* ALLOC_SPLIT_PTLOCKS */
1586
1587static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1588{
1589 return ptlock_ptr(pmd_page(*pmd));
1590}
1591
1592static inline bool ptlock_init(struct page *page)
1593{
1594 /*
1595 * prep_new_page() initialize page->private (and therefore page->ptl)
1596 * with 0. Make sure nobody took it in use in between.
1597 *
1598 * It can happen if arch try to use slab for page table allocation:
1599 * slab code uses page->slab_cache, which share storage with page->ptl.
1600 */
1601 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1602 if (!ptlock_alloc(page))
1603 return false;
1604 spin_lock_init(ptlock_ptr(page));
1605 return true;
1606}
1607
1608/* Reset page->mapping so free_pages_check won't complain. */
1609static inline void pte_lock_deinit(struct page *page)
1610{
1611 page->mapping = NULL;
1612 ptlock_free(page);
1613}
1614
1615#else /* !USE_SPLIT_PTE_PTLOCKS */
1616/*
1617 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1618 */
1619static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1620{
1621 return &mm->page_table_lock;
1622}
1623static inline void ptlock_cache_init(void) {}
1624static inline bool ptlock_init(struct page *page) { return true; }
1625static inline void pte_lock_deinit(struct page *page) {}
1626#endif /* USE_SPLIT_PTE_PTLOCKS */
1627
1628static inline void pgtable_init(void)
1629{
1630 ptlock_cache_init();
1631 pgtable_cache_init();
1632}
1633
1634static inline bool pgtable_page_ctor(struct page *page)
1635{
1636 if (!ptlock_init(page))
1637 return false;
1638 inc_zone_page_state(page, NR_PAGETABLE);
1639 return true;
1640}
1641
1642static inline void pgtable_page_dtor(struct page *page)
1643{
1644 pte_lock_deinit(page);
1645 dec_zone_page_state(page, NR_PAGETABLE);
1646}
1647
1648#define pte_offset_map_lock(mm, pmd, address, ptlp) \
1649({ \
1650 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1651 pte_t *__pte = pte_offset_map(pmd, address); \
1652 *(ptlp) = __ptl; \
1653 spin_lock(__ptl); \
1654 __pte; \
1655})
1656
1657#define pte_unmap_unlock(pte, ptl) do { \
1658 spin_unlock(ptl); \
1659 pte_unmap(pte); \
1660} while (0)
1661
1662#define pte_alloc_map(mm, vma, pmd, address) \
1663 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1664 pmd, address))? \
1665 NULL: pte_offset_map(pmd, address))
1666
1667#define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1668 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1669 pmd, address))? \
1670 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1671
1672#define pte_alloc_kernel(pmd, address) \
1673 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1674 NULL: pte_offset_kernel(pmd, address))
1675
1676#if USE_SPLIT_PMD_PTLOCKS
1677
1678static struct page *pmd_to_page(pmd_t *pmd)
1679{
1680 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1681 return virt_to_page((void *)((unsigned long) pmd & mask));
1682}
1683
1684static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1685{
1686 return ptlock_ptr(pmd_to_page(pmd));
1687}
1688
1689static inline bool pgtable_pmd_page_ctor(struct page *page)
1690{
1691#ifdef CONFIG_TRANSPARENT_HUGEPAGE
1692 page->pmd_huge_pte = NULL;
1693#endif
1694 return ptlock_init(page);
1695}
1696
1697static inline void pgtable_pmd_page_dtor(struct page *page)
1698{
1699#ifdef CONFIG_TRANSPARENT_HUGEPAGE
1700 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1701#endif
1702 ptlock_free(page);
1703}
1704
1705#define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1706
1707#else
1708
1709static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1710{
1711 return &mm->page_table_lock;
1712}
1713
1714static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1715static inline void pgtable_pmd_page_dtor(struct page *page) {}
1716
1717#define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1718
1719#endif
1720
1721static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1722{
1723 spinlock_t *ptl = pmd_lockptr(mm, pmd);
1724 spin_lock(ptl);
1725 return ptl;
1726}
1727
1728extern void free_area_init(unsigned long * zones_size);
1729extern void free_area_init_node(int nid, unsigned long * zones_size,
1730 unsigned long zone_start_pfn, unsigned long *zholes_size);
1731extern void free_initmem(void);
1732
1733/*
1734 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1735 * into the buddy system. The freed pages will be poisoned with pattern
1736 * "poison" if it's within range [0, UCHAR_MAX].
1737 * Return pages freed into the buddy system.
1738 */
1739extern unsigned long free_reserved_area(void *start, void *end,
1740 int poison, char *s);
1741
1742#ifdef CONFIG_HIGHMEM
1743/*
1744 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1745 * and totalram_pages.
1746 */
1747extern void free_highmem_page(struct page *page);
1748#endif
1749
1750extern void adjust_managed_page_count(struct page *page, long count);
1751extern void mem_init_print_info(const char *str);
1752
1753extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
1754
1755/* Free the reserved page into the buddy system, so it gets managed. */
1756static inline void __free_reserved_page(struct page *page)
1757{
1758 ClearPageReserved(page);
1759 init_page_count(page);
1760 __free_page(page);
1761}
1762
1763static inline void free_reserved_page(struct page *page)
1764{
1765 __free_reserved_page(page);
1766 adjust_managed_page_count(page, 1);
1767}
1768
1769static inline void mark_page_reserved(struct page *page)
1770{
1771 SetPageReserved(page);
1772 adjust_managed_page_count(page, -1);
1773}
1774
1775/*
1776 * Default method to free all the __init memory into the buddy system.
1777 * The freed pages will be poisoned with pattern "poison" if it's within
1778 * range [0, UCHAR_MAX].
1779 * Return pages freed into the buddy system.
1780 */
1781static inline unsigned long free_initmem_default(int poison)
1782{
1783 extern char __init_begin[], __init_end[];
1784
1785 return free_reserved_area(&__init_begin, &__init_end,
1786 poison, "unused kernel");
1787}
1788
1789static inline unsigned long get_num_physpages(void)
1790{
1791 int nid;
1792 unsigned long phys_pages = 0;
1793
1794 for_each_online_node(nid)
1795 phys_pages += node_present_pages(nid);
1796
1797 return phys_pages;
1798}
1799
1800#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1801/*
1802 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1803 * zones, allocate the backing mem_map and account for memory holes in a more
1804 * architecture independent manner. This is a substitute for creating the
1805 * zone_sizes[] and zholes_size[] arrays and passing them to
1806 * free_area_init_node()
1807 *
1808 * An architecture is expected to register range of page frames backed by
1809 * physical memory with memblock_add[_node]() before calling
1810 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1811 * usage, an architecture is expected to do something like
1812 *
1813 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1814 * max_highmem_pfn};
1815 * for_each_valid_physical_page_range()
1816 * memblock_add_node(base, size, nid)
1817 * free_area_init_nodes(max_zone_pfns);
1818 *
1819 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1820 * registered physical page range. Similarly
1821 * sparse_memory_present_with_active_regions() calls memory_present() for
1822 * each range when SPARSEMEM is enabled.
1823 *
1824 * See mm/page_alloc.c for more information on each function exposed by
1825 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1826 */
1827extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1828unsigned long node_map_pfn_alignment(void);
1829unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1830 unsigned long end_pfn);
1831extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1832 unsigned long end_pfn);
1833extern void get_pfn_range_for_nid(unsigned int nid,
1834 unsigned long *start_pfn, unsigned long *end_pfn);
1835extern unsigned long find_min_pfn_with_active_regions(void);
1836extern void free_bootmem_with_active_regions(int nid,
1837 unsigned long max_low_pfn);
1838extern void sparse_memory_present_with_active_regions(int nid);
1839
1840#endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1841
1842#if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1843 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1844static inline int __early_pfn_to_nid(unsigned long pfn,
1845 struct mminit_pfnnid_cache *state)
1846{
1847 return 0;
1848}
1849#else
1850/* please see mm/page_alloc.c */
1851extern int __meminit early_pfn_to_nid(unsigned long pfn);
1852/* there is a per-arch backend function. */
1853extern int __meminit __early_pfn_to_nid(unsigned long pfn,
1854 struct mminit_pfnnid_cache *state);
1855#endif
1856
1857extern void set_dma_reserve(unsigned long new_dma_reserve);
1858extern void memmap_init_zone(unsigned long, int, unsigned long,
1859 unsigned long, enum memmap_context);
1860extern void setup_per_zone_wmarks(void);
1861extern int __meminit init_per_zone_wmark_min(void);
1862extern void mem_init(void);
1863extern void __init mmap_init(void);
1864extern void show_mem(unsigned int flags);
1865extern long si_mem_available(void);
1866extern void si_meminfo(struct sysinfo * val);
1867extern void si_meminfo_node(struct sysinfo *val, int nid);
1868#ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
1869extern unsigned long arch_reserved_kernel_pages(void);
1870#endif
1871
1872extern __printf(3, 4)
1873void warn_alloc_failed(gfp_t gfp_mask, unsigned int order,
1874 const char *fmt, ...);
1875
1876extern void setup_per_cpu_pageset(void);
1877
1878extern void zone_pcp_update(struct zone *zone);
1879extern void zone_pcp_reset(struct zone *zone);
1880
1881/* page_alloc.c */
1882extern int min_free_kbytes;
1883extern int watermark_scale_factor;
1884
1885/* nommu.c */
1886extern atomic_long_t mmap_pages_allocated;
1887extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1888
1889/* interval_tree.c */
1890void vma_interval_tree_insert(struct vm_area_struct *node,
1891 struct rb_root *root);
1892void vma_interval_tree_insert_after(struct vm_area_struct *node,
1893 struct vm_area_struct *prev,
1894 struct rb_root *root);
1895void vma_interval_tree_remove(struct vm_area_struct *node,
1896 struct rb_root *root);
1897struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1898 unsigned long start, unsigned long last);
1899struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1900 unsigned long start, unsigned long last);
1901
1902#define vma_interval_tree_foreach(vma, root, start, last) \
1903 for (vma = vma_interval_tree_iter_first(root, start, last); \
1904 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1905
1906void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1907 struct rb_root *root);
1908void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1909 struct rb_root *root);
1910struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1911 struct rb_root *root, unsigned long start, unsigned long last);
1912struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1913 struct anon_vma_chain *node, unsigned long start, unsigned long last);
1914#ifdef CONFIG_DEBUG_VM_RB
1915void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1916#endif
1917
1918#define anon_vma_interval_tree_foreach(avc, root, start, last) \
1919 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1920 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1921
1922/* mmap.c */
1923extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1924extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1925 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1926extern struct vm_area_struct *vma_merge(struct mm_struct *,
1927 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1928 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1929 struct mempolicy *, struct vm_userfaultfd_ctx);
1930extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1931extern int split_vma(struct mm_struct *,
1932 struct vm_area_struct *, unsigned long addr, int new_below);
1933extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1934extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1935 struct rb_node **, struct rb_node *);
1936extern void unlink_file_vma(struct vm_area_struct *);
1937extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1938 unsigned long addr, unsigned long len, pgoff_t pgoff,
1939 bool *need_rmap_locks);
1940extern void exit_mmap(struct mm_struct *);
1941
1942static inline int check_data_rlimit(unsigned long rlim,
1943 unsigned long new,
1944 unsigned long start,
1945 unsigned long end_data,
1946 unsigned long start_data)
1947{
1948 if (rlim < RLIM_INFINITY) {
1949 if (((new - start) + (end_data - start_data)) > rlim)
1950 return -ENOSPC;
1951 }
1952
1953 return 0;
1954}
1955
1956extern int mm_take_all_locks(struct mm_struct *mm);
1957extern void mm_drop_all_locks(struct mm_struct *mm);
1958
1959extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1960extern struct file *get_mm_exe_file(struct mm_struct *mm);
1961extern struct file *get_task_exe_file(struct task_struct *task);
1962
1963extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1964extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
1965 unsigned long addr, unsigned long len,
1966 unsigned long flags,
1967 const struct vm_special_mapping *spec);
1968/* This is an obsolete alternative to _install_special_mapping. */
1969extern int install_special_mapping(struct mm_struct *mm,
1970 unsigned long addr, unsigned long len,
1971 unsigned long flags, struct page **pages);
1972
1973extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1974
1975extern unsigned long mmap_region(struct file *file, unsigned long addr,
1976 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
1977extern unsigned long do_mmap(struct file *file, unsigned long addr,
1978 unsigned long len, unsigned long prot, unsigned long flags,
1979 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate);
1980extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1981
1982static inline unsigned long
1983do_mmap_pgoff(struct file *file, unsigned long addr,
1984 unsigned long len, unsigned long prot, unsigned long flags,
1985 unsigned long pgoff, unsigned long *populate)
1986{
1987 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate);
1988}
1989
1990#ifdef CONFIG_MMU
1991extern int __mm_populate(unsigned long addr, unsigned long len,
1992 int ignore_errors);
1993static inline void mm_populate(unsigned long addr, unsigned long len)
1994{
1995 /* Ignore errors */
1996 (void) __mm_populate(addr, len, 1);
1997}
1998#else
1999static inline void mm_populate(unsigned long addr, unsigned long len) {}
2000#endif
2001
2002/* These take the mm semaphore themselves */
2003extern unsigned long vm_brk(unsigned long, unsigned long);
2004extern int vm_munmap(unsigned long, size_t);
2005extern unsigned long vm_mmap(struct file *, unsigned long,
2006 unsigned long, unsigned long,
2007 unsigned long, unsigned long);
2008
2009struct vm_unmapped_area_info {
2010#define VM_UNMAPPED_AREA_TOPDOWN 1
2011 unsigned long flags;
2012 unsigned long length;
2013 unsigned long low_limit;
2014 unsigned long high_limit;
2015 unsigned long align_mask;
2016 unsigned long align_offset;
2017};
2018
2019extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2020extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2021
2022/*
2023 * Search for an unmapped address range.
2024 *
2025 * We are looking for a range that:
2026 * - does not intersect with any VMA;
2027 * - is contained within the [low_limit, high_limit) interval;
2028 * - is at least the desired size.
2029 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2030 */
2031static inline unsigned long
2032vm_unmapped_area(struct vm_unmapped_area_info *info)
2033{
2034 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2035 return unmapped_area_topdown(info);
2036 else
2037 return unmapped_area(info);
2038}
2039
2040/* truncate.c */
2041extern void truncate_inode_pages(struct address_space *, loff_t);
2042extern void truncate_inode_pages_range(struct address_space *,
2043 loff_t lstart, loff_t lend);
2044extern void truncate_inode_pages_final(struct address_space *);
2045
2046/* generic vm_area_ops exported for stackable file systems */
2047extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
2048extern void filemap_map_pages(struct vm_area_struct *vma, struct vm_fault *vmf);
2049extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
2050
2051/* mm/page-writeback.c */
2052int write_one_page(struct page *page, int wait);
2053void task_dirty_inc(struct task_struct *tsk);
2054
2055/* readahead.c */
2056#define VM_MAX_READAHEAD 512 /* kbytes */
2057#define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
2058
2059int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2060 pgoff_t offset, unsigned long nr_to_read);
2061
2062void page_cache_sync_readahead(struct address_space *mapping,
2063 struct file_ra_state *ra,
2064 struct file *filp,
2065 pgoff_t offset,
2066 unsigned long size);
2067
2068void page_cache_async_readahead(struct address_space *mapping,
2069 struct file_ra_state *ra,
2070 struct file *filp,
2071 struct page *pg,
2072 pgoff_t offset,
2073 unsigned long size);
2074
2075extern unsigned long stack_guard_gap;
2076/* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2077extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2078
2079/* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2080extern int expand_downwards(struct vm_area_struct *vma,
2081 unsigned long address);
2082#if VM_GROWSUP
2083extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2084#else
2085 #define expand_upwards(vma, address) (0)
2086#endif
2087
2088/* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2089extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2090extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2091 struct vm_area_struct **pprev);
2092
2093/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2094 NULL if none. Assume start_addr < end_addr. */
2095static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2096{
2097 struct vm_area_struct * vma = find_vma(mm,start_addr);
2098
2099 if (vma && end_addr <= vma->vm_start)
2100 vma = NULL;
2101 return vma;
2102}
2103
2104static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2105{
2106 unsigned long vm_start = vma->vm_start;
2107
2108 if (vma->vm_flags & VM_GROWSDOWN) {
2109 vm_start -= stack_guard_gap;
2110 if (vm_start > vma->vm_start)
2111 vm_start = 0;
2112 }
2113 return vm_start;
2114}
2115
2116static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2117{
2118 unsigned long vm_end = vma->vm_end;
2119
2120 if (vma->vm_flags & VM_GROWSUP) {
2121 vm_end += stack_guard_gap;
2122 if (vm_end < vma->vm_end)
2123 vm_end = -PAGE_SIZE;
2124 }
2125 return vm_end;
2126}
2127
2128static inline unsigned long vma_pages(struct vm_area_struct *vma)
2129{
2130 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2131}
2132
2133/* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2134static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2135 unsigned long vm_start, unsigned long vm_end)
2136{
2137 struct vm_area_struct *vma = find_vma(mm, vm_start);
2138
2139 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2140 vma = NULL;
2141
2142 return vma;
2143}
2144
2145#ifdef CONFIG_MMU
2146pgprot_t vm_get_page_prot(unsigned long vm_flags);
2147void vma_set_page_prot(struct vm_area_struct *vma);
2148#else
2149static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2150{
2151 return __pgprot(0);
2152}
2153static inline void vma_set_page_prot(struct vm_area_struct *vma)
2154{
2155 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2156}
2157#endif
2158
2159#ifdef CONFIG_NUMA_BALANCING
2160unsigned long change_prot_numa(struct vm_area_struct *vma,
2161 unsigned long start, unsigned long end);
2162#endif
2163
2164struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2165int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2166 unsigned long pfn, unsigned long size, pgprot_t);
2167int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2168int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2169 unsigned long pfn);
2170int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2171 pfn_t pfn);
2172int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2173
2174
2175struct page *follow_page_mask(struct vm_area_struct *vma,
2176 unsigned long address, unsigned int foll_flags,
2177 unsigned int *page_mask);
2178
2179static inline struct page *follow_page(struct vm_area_struct *vma,
2180 unsigned long address, unsigned int foll_flags)
2181{
2182 unsigned int unused_page_mask;
2183 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2184}
2185
2186#define FOLL_WRITE 0x01 /* check pte is writable */
2187#define FOLL_TOUCH 0x02 /* mark page accessed */
2188#define FOLL_GET 0x04 /* do get_page on page */
2189#define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2190#define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2191#define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2192 * and return without waiting upon it */
2193#define FOLL_POPULATE 0x40 /* fault in page */
2194#define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2195#define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2196#define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2197#define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2198#define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2199#define FOLL_MLOCK 0x1000 /* lock present pages */
2200#define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2201#define FOLL_COW 0x4000 /* internal GUP flag */
2202
2203typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2204 void *data);
2205extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2206 unsigned long size, pte_fn_t fn, void *data);
2207
2208#ifdef CONFIG_PROC_FS
2209void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
2210#else
2211static inline void vm_stat_account(struct mm_struct *mm,
2212 unsigned long flags, struct file *file, long pages)
2213{
2214 mm->total_vm += pages;
2215}
2216#endif /* CONFIG_PROC_FS */
2217
2218#ifdef CONFIG_DEBUG_PAGEALLOC
2219extern bool _debug_pagealloc_enabled;
2220extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2221
2222static inline bool debug_pagealloc_enabled(void)
2223{
2224 return _debug_pagealloc_enabled;
2225}
2226
2227static inline void
2228kernel_map_pages(struct page *page, int numpages, int enable)
2229{
2230 if (!debug_pagealloc_enabled())
2231 return;
2232
2233 __kernel_map_pages(page, numpages, enable);
2234}
2235#ifdef CONFIG_HIBERNATION
2236extern bool kernel_page_present(struct page *page);
2237#endif /* CONFIG_HIBERNATION */
2238#else /* CONFIG_DEBUG_PAGEALLOC */
2239static inline void
2240kernel_map_pages(struct page *page, int numpages, int enable) {}
2241#ifdef CONFIG_HIBERNATION
2242static inline bool kernel_page_present(struct page *page) { return true; }
2243#endif /* CONFIG_HIBERNATION */
2244static inline bool debug_pagealloc_enabled(void)
2245{
2246 return false;
2247}
2248#endif /* CONFIG_DEBUG_PAGEALLOC */
2249
2250#ifdef __HAVE_ARCH_GATE_AREA
2251extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2252extern int in_gate_area_no_mm(unsigned long addr);
2253extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2254#else
2255static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2256{
2257 return NULL;
2258}
2259static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2260static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2261{
2262 return 0;
2263}
2264#endif /* __HAVE_ARCH_GATE_AREA */
2265
2266#ifdef CONFIG_SYSCTL
2267extern int sysctl_drop_caches;
2268int drop_caches_sysctl_handler(struct ctl_table *, int,
2269 void __user *, size_t *, loff_t *);
2270#endif
2271
2272void drop_slab(void);
2273void drop_slab_node(int nid);
2274
2275#ifndef CONFIG_MMU
2276#define randomize_va_space 0
2277#else
2278extern int randomize_va_space;
2279#endif
2280
2281const char * arch_vma_name(struct vm_area_struct *vma);
2282void print_vma_addr(char *prefix, unsigned long rip);
2283
2284void sparse_mem_maps_populate_node(struct page **map_map,
2285 unsigned long pnum_begin,
2286 unsigned long pnum_end,
2287 unsigned long map_count,
2288 int nodeid);
2289
2290struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
2291pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2292pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
2293pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2294pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2295void *vmemmap_alloc_block(unsigned long size, int node);
2296struct vmem_altmap;
2297void *__vmemmap_alloc_block_buf(unsigned long size, int node,
2298 struct vmem_altmap *altmap);
2299static inline void *vmemmap_alloc_block_buf(unsigned long size, int node)
2300{
2301 return __vmemmap_alloc_block_buf(size, node, NULL);
2302}
2303
2304void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2305int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2306 int node);
2307int vmemmap_populate(unsigned long start, unsigned long end, int node);
2308void vmemmap_populate_print_last(void);
2309#ifdef CONFIG_MEMORY_HOTPLUG
2310void vmemmap_free(unsigned long start, unsigned long end);
2311#endif
2312void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2313 unsigned long size);
2314
2315enum mf_flags {
2316 MF_COUNT_INCREASED = 1 << 0,
2317 MF_ACTION_REQUIRED = 1 << 1,
2318 MF_MUST_KILL = 1 << 2,
2319 MF_SOFT_OFFLINE = 1 << 3,
2320};
2321extern int memory_failure(unsigned long pfn, int trapno, int flags);
2322extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
2323extern int unpoison_memory(unsigned long pfn);
2324extern int get_hwpoison_page(struct page *page);
2325extern void put_hwpoison_page(struct page *page);
2326extern int sysctl_memory_failure_early_kill;
2327extern int sysctl_memory_failure_recovery;
2328extern void shake_page(struct page *p, int access);
2329extern atomic_long_t num_poisoned_pages;
2330extern int soft_offline_page(struct page *page, int flags);
2331
2332
2333/*
2334 * Error handlers for various types of pages.
2335 */
2336enum mf_result {
2337 MF_IGNORED, /* Error: cannot be handled */
2338 MF_FAILED, /* Error: handling failed */
2339 MF_DELAYED, /* Will be handled later */
2340 MF_RECOVERED, /* Successfully recovered */
2341};
2342
2343enum mf_action_page_type {
2344 MF_MSG_KERNEL,
2345 MF_MSG_KERNEL_HIGH_ORDER,
2346 MF_MSG_SLAB,
2347 MF_MSG_DIFFERENT_COMPOUND,
2348 MF_MSG_POISONED_HUGE,
2349 MF_MSG_HUGE,
2350 MF_MSG_FREE_HUGE,
2351 MF_MSG_UNMAP_FAILED,
2352 MF_MSG_DIRTY_SWAPCACHE,
2353 MF_MSG_CLEAN_SWAPCACHE,
2354 MF_MSG_DIRTY_MLOCKED_LRU,
2355 MF_MSG_CLEAN_MLOCKED_LRU,
2356 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2357 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2358 MF_MSG_DIRTY_LRU,
2359 MF_MSG_CLEAN_LRU,
2360 MF_MSG_TRUNCATED_LRU,
2361 MF_MSG_BUDDY,
2362 MF_MSG_BUDDY_2ND,
2363 MF_MSG_UNKNOWN,
2364};
2365
2366#if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2367extern void clear_huge_page(struct page *page,
2368 unsigned long addr,
2369 unsigned int pages_per_huge_page);
2370extern void copy_user_huge_page(struct page *dst, struct page *src,
2371 unsigned long addr, struct vm_area_struct *vma,
2372 unsigned int pages_per_huge_page);
2373#endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2374
2375extern struct page_ext_operations debug_guardpage_ops;
2376extern struct page_ext_operations page_poisoning_ops;
2377
2378#ifdef CONFIG_DEBUG_PAGEALLOC
2379extern unsigned int _debug_guardpage_minorder;
2380extern bool _debug_guardpage_enabled;
2381
2382static inline unsigned int debug_guardpage_minorder(void)
2383{
2384 return _debug_guardpage_minorder;
2385}
2386
2387static inline bool debug_guardpage_enabled(void)
2388{
2389 return _debug_guardpage_enabled;
2390}
2391
2392static inline bool page_is_guard(struct page *page)
2393{
2394 struct page_ext *page_ext;
2395
2396 if (!debug_guardpage_enabled())
2397 return false;
2398
2399 page_ext = lookup_page_ext(page);
2400 return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2401}
2402#else
2403static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2404static inline bool debug_guardpage_enabled(void) { return false; }
2405static inline bool page_is_guard(struct page *page) { return false; }
2406#endif /* CONFIG_DEBUG_PAGEALLOC */
2407
2408#if MAX_NUMNODES > 1
2409void __init setup_nr_node_ids(void);
2410#else
2411static inline void setup_nr_node_ids(void) {}
2412#endif
2413
2414#endif /* __KERNEL__ */
2415#endif /* _LINUX_MM_H */

© 2024 Oracle Support Privacy / Do Not Sell My Info Terms of Use Trademark Policy