Branch data Line data Source code
1 : : /* SPDX-License-Identifier: GPL-2.0 */
2 : : #ifndef _LINUX_MM_H
3 : : #define _LINUX_MM_H
4 : :
5 : : #include <linux/errno.h>
6 : :
7 : : #ifdef __KERNEL__
8 : :
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>
30 : :
31 : : struct mempolicy;
32 : : struct anon_vma;
33 : : struct anon_vma_chain;
34 : : struct file_ra_state;
35 : : struct user_struct;
36 : : struct writeback_control;
37 : : struct bdi_writeback;
38 : :
39 : : void init_mm_internals(void);
40 : :
41 : : #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
42 : : extern unsigned long max_mapnr;
43 : :
44 : : static inline void set_max_mapnr(unsigned long limit)
45 : : {
46 : 3 : max_mapnr = limit;
47 : : }
48 : : #else
49 : : static inline void set_max_mapnr(unsigned long limit) { }
50 : : #endif
51 : :
52 : : extern atomic_long_t _totalram_pages;
53 : : static inline unsigned long totalram_pages(void)
54 : : {
55 : : return (unsigned long)atomic_long_read(&_totalram_pages);
56 : : }
57 : :
58 : 0 : static inline void totalram_pages_inc(void)
59 : : {
60 : : atomic_long_inc(&_totalram_pages);
61 : 0 : }
62 : :
63 : : static inline void totalram_pages_dec(void)
64 : : {
65 : : atomic_long_dec(&_totalram_pages);
66 : : }
67 : :
68 : : static inline void totalram_pages_add(long count)
69 : : {
70 : : atomic_long_add(count, &_totalram_pages);
71 : : }
72 : :
73 : : static inline void totalram_pages_set(long val)
74 : : {
75 : : atomic_long_set(&_totalram_pages, val);
76 : : }
77 : :
78 : : extern void * high_memory;
79 : : extern int page_cluster;
80 : :
81 : : #ifdef CONFIG_SYSCTL
82 : : extern int sysctl_legacy_va_layout;
83 : : #else
84 : : #define sysctl_legacy_va_layout 0
85 : : #endif
86 : :
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;
91 : : #endif
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;
96 : : #endif
97 : :
98 : : #include <asm/page.h>
99 : : #include <asm/pgtable.h>
100 : : #include <asm/processor.h>
101 : :
102 : : /*
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.
108 : : */
109 : : #ifndef untagged_addr
110 : : #define untagged_addr(addr) (addr)
111 : : #endif
112 : :
113 : : #ifndef __pa_symbol
114 : : #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
115 : : #endif
116 : :
117 : : #ifndef page_to_virt
118 : : #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
119 : : #endif
120 : :
121 : : #ifndef lm_alias
122 : : #define lm_alias(x) __va(__pa_symbol(x))
123 : : #endif
124 : :
125 : : /*
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.
131 : : */
132 : : #ifndef mm_forbids_zeropage
133 : : #define mm_forbids_zeropage(X) (0)
134 : : #endif
135 : :
136 : : /*
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>
141 : : */
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.
148 : : */
149 : : #define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
150 : : static inline void __mm_zero_struct_page(struct page *page)
151 : : {
152 : : unsigned long *_pp = (void *)page;
153 : :
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);
158 : :
159 : : switch (sizeof(struct page)) {
160 : : case 80:
161 : : _pp[9] = 0; /* fallthrough */
162 : : case 72:
163 : : _pp[8] = 0; /* fallthrough */
164 : : case 64:
165 : : _pp[7] = 0; /* fallthrough */
166 : : case 56:
167 : : _pp[6] = 0;
168 : : _pp[5] = 0;
169 : : _pp[4] = 0;
170 : : _pp[3] = 0;
171 : : _pp[2] = 0;
172 : : _pp[1] = 0;
173 : : _pp[0] = 0;
174 : : }
175 : : }
176 : : #else
177 : : #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
178 : : #endif
179 : :
180 : : /*
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
183 : : * problem.
184 : : *
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.
191 : : *
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
194 : : * that.
195 : : */
196 : : #define MAPCOUNT_ELF_CORE_MARGIN (5)
197 : : #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
198 : :
199 : : extern int sysctl_max_map_count;
200 : :
201 : : extern unsigned long sysctl_user_reserve_kbytes;
202 : : extern unsigned long sysctl_admin_reserve_kbytes;
203 : :
204 : : extern int sysctl_overcommit_memory;
205 : : extern int sysctl_overcommit_ratio;
206 : : extern unsigned long sysctl_overcommit_kbytes;
207 : :
208 : : extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
209 : : size_t *, loff_t *);
210 : : extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
211 : : size_t *, loff_t *);
212 : :
213 : : #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
214 : :
215 : : /* to align the pointer to the (next) page boundary */
216 : : #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
217 : :
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)
220 : :
221 : : #define lru_to_page(head) (list_entry((head)->prev, struct page, lru))
222 : :
223 : : /*
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
229 : : * mmap() functions).
230 : : */
231 : :
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 *);
235 : :
236 : : #ifndef CONFIG_MMU
237 : : extern struct rb_root nommu_region_tree;
238 : : extern struct rw_semaphore nommu_region_sem;
239 : :
240 : : extern unsigned int kobjsize(const void *objp);
241 : : #endif
242 : :
243 : : /*
244 : : * vm_flags in vm_area_struct, see mm_types.h.
245 : : * When changing, update also include/trace/events/mmflags.h
246 : : */
247 : : #define VM_NONE 0x00000000
248 : :
249 : : #define VM_READ 0x00000001 /* currently active flags */
250 : : #define VM_WRITE 0x00000002
251 : : #define VM_EXEC 0x00000004
252 : : #define VM_SHARED 0x00000008
253 : :
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
259 : :
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 */
265 : :
266 : : #define VM_LOCKED 0x00002000
267 : : #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
268 : :
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 */
272 : :
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 */
283 : :
284 : : #ifdef CONFIG_MEM_SOFT_DIRTY
285 : : # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
286 : : #else
287 : : # define VM_SOFTDIRTY 0
288 : : #endif
289 : :
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 */
294 : :
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 */
307 : :
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
314 : : #ifdef CONFIG_PPC
315 : : # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
316 : : #else
317 : : # define VM_PKEY_BIT4 0
318 : : #endif
319 : : #endif /* CONFIG_ARCH_HAS_PKEYS */
320 : :
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) */
334 : : #endif
335 : :
336 : : #if defined(CONFIG_X86_INTEL_MPX)
337 : : /* MPX specific bounds table or bounds directory */
338 : : # define VM_MPX VM_HIGH_ARCH_4
339 : : #else
340 : : # define VM_MPX VM_NONE
341 : : #endif
342 : :
343 : : #ifndef VM_GROWSUP
344 : : # define VM_GROWSUP VM_NONE
345 : : #endif
346 : :
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)
349 : :
350 : : #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
351 : : #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
352 : : #endif
353 : :
354 : : #ifdef CONFIG_STACK_GROWSUP
355 : : #define VM_STACK VM_GROWSUP
356 : : #else
357 : : #define VM_STACK VM_GROWSDOWN
358 : : #endif
359 : :
360 : : #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
361 : :
362 : : /*
363 : : * Special vmas that are non-mergable, non-mlock()able.
364 : : * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
365 : : */
366 : : #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
367 : :
368 : : /* This mask defines which mm->def_flags a process can inherit its parent */
369 : : #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
370 : :
371 : : /* This mask is used to clear all the VMA flags used by mlock */
372 : : #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
373 : :
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
377 : : #endif
378 : : #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
379 : :
380 : : /*
381 : : * mapping from the currently active vm_flags protection bits (the
382 : : * low four bits) to a page protection mask..
383 : : */
384 : : extern pgprot_t protection_map[16];
385 : :
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 */
395 : :
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" }
406 : :
407 : : /*
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.
411 : : *
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.
414 : : *
415 : : * pgoff should be used in favour of virtual_address, if possible.
416 : : */
417 : : struct vm_fault {
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
424 : : * the 'address' */
425 : : pud_t *pud; /* Pointer to pud entry matching
426 : : * the 'address'
427 : : */
428 : : pte_t orig_pte; /* Value of PTE at the time of fault */
429 : :
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
435 : : * VM_FAULT_ERROR).
436 : : */
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.
441 : : */
442 : : spinlock_t *ptl; /* Page table lock.
443 : : * Protects pte page table if 'pte'
444 : : * is not NULL, otherwise pmd.
445 : : */
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
451 : : * atomic context.
452 : : */
453 : : };
454 : :
455 : : /* page entry size for vm->huge_fault() */
456 : : enum page_entry_size {
457 : : PE_SIZE_PTE = 0,
458 : : PE_SIZE_PMD,
459 : : PE_SIZE_PUD,
460 : : };
461 : :
462 : : /*
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.
466 : : */
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);
478 : :
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);
482 : :
483 : : /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
484 : : vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
485 : :
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
488 : : */
489 : : int (*access)(struct vm_area_struct *vma, unsigned long addr,
490 : : void *buf, int len, int write);
491 : :
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);
496 : :
497 : : #ifdef CONFIG_NUMA
498 : : /*
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
503 : : * mempolicy.
504 : : */
505 : : int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
506 : :
507 : : /*
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
515 : : * policy.
516 : : */
517 : : struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
518 : : unsigned long addr);
519 : : #endif
520 : : /*
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.
524 : : */
525 : : struct page *(*find_special_page)(struct vm_area_struct *vma,
526 : : unsigned long addr);
527 : : };
528 : :
529 : : static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
530 : : {
531 : : static const struct vm_operations_struct dummy_vm_ops = {};
532 : :
533 : 3 : memset(vma, 0, sizeof(*vma));
534 : 3 : vma->vm_mm = mm;
535 : 3 : vma->vm_ops = &dummy_vm_ops;
536 : 3 : INIT_LIST_HEAD(&vma->anon_vma_chain);
537 : : }
538 : :
539 : : static inline void vma_set_anonymous(struct vm_area_struct *vma)
540 : : {
541 : 3 : vma->vm_ops = NULL;
542 : : }
543 : :
544 : : static inline bool vma_is_anonymous(struct vm_area_struct *vma)
545 : : {
546 : 3 : return !vma->vm_ops;
547 : : }
548 : :
549 : : #ifdef CONFIG_SHMEM
550 : : /*
551 : : * The vma_is_shmem is not inline because it is used only by slow
552 : : * paths in userfault.
553 : : */
554 : : bool vma_is_shmem(struct vm_area_struct *vma);
555 : : #else
556 : : static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
557 : : #endif
558 : :
559 : : int vma_is_stack_for_current(struct vm_area_struct *vma);
560 : :
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) }
563 : :
564 : : struct mmu_gather;
565 : : struct inode;
566 : :
567 : : #if !defined(CONFIG_ARCH_HAS_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
568 : : static inline int pmd_devmap(pmd_t pmd)
569 : : {
570 : : return 0;
571 : : }
572 : : static inline int pud_devmap(pud_t pud)
573 : : {
574 : : return 0;
575 : : }
576 : : static inline int pgd_devmap(pgd_t pgd)
577 : : {
578 : : return 0;
579 : : }
580 : : #endif
581 : :
582 : : /*
583 : : * FIXME: take this include out, include page-flags.h in
584 : : * files which need it (119 of them)
585 : : */
586 : : #include <linux/page-flags.h>
587 : : #include <linux/huge_mm.h>
588 : :
589 : : /*
590 : : * Methods to modify the page usage count.
591 : : *
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
597 : : *
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.
600 : : */
601 : :
602 : : /*
603 : : * Drop a ref, return true if the refcount fell to zero (the page has no users)
604 : : */
605 : : static inline int put_page_testzero(struct page *page)
606 : : {
607 : : VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
608 : : return page_ref_dec_and_test(page);
609 : : }
610 : :
611 : : /*
612 : : * Try to grab a ref unless the page has a refcount of zero, return false if
613 : : * that is the case.
614 : : * This can be called when MMU is off so it must not access
615 : : * any of the virtual mappings.
616 : : */
617 : : static inline int get_page_unless_zero(struct page *page)
618 : : {
619 : : return page_ref_add_unless(page, 1, 0);
620 : : }
621 : :
622 : : extern int page_is_ram(unsigned long pfn);
623 : :
624 : : enum {
625 : : REGION_INTERSECTS,
626 : : REGION_DISJOINT,
627 : : REGION_MIXED,
628 : : };
629 : :
630 : : int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
631 : : unsigned long desc);
632 : :
633 : : /* Support for virtually mapped pages */
634 : : struct page *vmalloc_to_page(const void *addr);
635 : : unsigned long vmalloc_to_pfn(const void *addr);
636 : :
637 : : /*
638 : : * Determine if an address is within the vmalloc range
639 : : *
640 : : * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
641 : : * is no special casing required.
642 : : */
643 : : static inline bool is_vmalloc_addr(const void *x)
644 : : {
645 : : #ifdef CONFIG_MMU
646 : 3 : unsigned long addr = (unsigned long)x;
647 : :
648 : 3 : return addr >= VMALLOC_START && addr < VMALLOC_END;
649 : : #else
650 : : return false;
651 : : #endif
652 : : }
653 : :
654 : : #ifndef is_ioremap_addr
655 : : #define is_ioremap_addr(x) is_vmalloc_addr(x)
656 : : #endif
657 : :
658 : : #ifdef CONFIG_MMU
659 : : extern int is_vmalloc_or_module_addr(const void *x);
660 : : #else
661 : : static inline int is_vmalloc_or_module_addr(const void *x)
662 : : {
663 : : return 0;
664 : : }
665 : : #endif
666 : :
667 : : extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
668 : : static inline void *kvmalloc(size_t size, gfp_t flags)
669 : : {
670 : 3 : return kvmalloc_node(size, flags, NUMA_NO_NODE);
671 : : }
672 : : static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
673 : : {
674 : : return kvmalloc_node(size, flags | __GFP_ZERO, node);
675 : : }
676 : : static inline void *kvzalloc(size_t size, gfp_t flags)
677 : : {
678 : 3 : return kvmalloc(size, flags | __GFP_ZERO);
679 : : }
680 : :
681 : 3 : static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
682 : : {
683 : : size_t bytes;
684 : :
685 : 3 : if (unlikely(check_mul_overflow(n, size, &bytes)))
686 : : return NULL;
687 : :
688 : 3 : return kvmalloc(bytes, flags);
689 : : }
690 : :
691 : : static inline void *kvcalloc(size_t n, size_t size, gfp_t flags)
692 : : {
693 : 3 : return kvmalloc_array(n, size, flags | __GFP_ZERO);
694 : : }
695 : :
696 : : extern void kvfree(const void *addr);
697 : : extern void kvfree_sensitive(const void *addr, size_t len);
698 : :
699 : : /*
700 : : * Mapcount of compound page as a whole, does not include mapped sub-pages.
701 : : *
702 : : * Must be called only for compound pages or any their tail sub-pages.
703 : : */
704 : : static inline int compound_mapcount(struct page *page)
705 : : {
706 : : VM_BUG_ON_PAGE(!PageCompound(page), page);
707 : : page = compound_head(page);
708 : 0 : return atomic_read(compound_mapcount_ptr(page)) + 1;
709 : : }
710 : :
711 : : /*
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).
715 : : */
716 : : static inline void page_mapcount_reset(struct page *page)
717 : : {
718 : : atomic_set(&(page)->_mapcount, -1);
719 : : }
720 : :
721 : : int __page_mapcount(struct page *page);
722 : :
723 : : /*
724 : : * Mapcount of 0-order page; when compound sub-page, includes
725 : : * compound_mapcount().
726 : : *
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.
730 : : */
731 : 3 : static inline int page_mapcount(struct page *page)
732 : : {
733 : 3 : if (unlikely(PageCompound(page)))
734 : 0 : return __page_mapcount(page);
735 : 3 : return atomic_read(&page->_mapcount) + 1;
736 : : }
737 : :
738 : : #ifdef CONFIG_TRANSPARENT_HUGEPAGE
739 : : int total_mapcount(struct page *page);
740 : : int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
741 : : #else
742 : : static inline int total_mapcount(struct page *page)
743 : : {
744 : 0 : return page_mapcount(page);
745 : : }
746 : : static inline int page_trans_huge_mapcount(struct page *page,
747 : : int *total_mapcount)
748 : : {
749 : 3 : int mapcount = page_mapcount(page);
750 : 3 : if (total_mapcount)
751 : 3 : *total_mapcount = mapcount;
752 : : return mapcount;
753 : : }
754 : : #endif
755 : :
756 : 0 : static inline struct page *virt_to_head_page(const void *x)
757 : : {
758 : 3 : struct page *page = virt_to_page(x);
759 : :
760 : 0 : return compound_head(page);
761 : : }
762 : :
763 : : void __put_page(struct page *page);
764 : :
765 : : void put_pages_list(struct list_head *pages);
766 : :
767 : : void split_page(struct page *page, unsigned int order);
768 : :
769 : : /*
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.
773 : : */
774 : : typedef void compound_page_dtor(struct page *);
775 : :
776 : : /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
777 : : enum compound_dtor_id {
778 : : NULL_COMPOUND_DTOR,
779 : : COMPOUND_PAGE_DTOR,
780 : : #ifdef CONFIG_HUGETLB_PAGE
781 : : HUGETLB_PAGE_DTOR,
782 : : #endif
783 : : #ifdef CONFIG_TRANSPARENT_HUGEPAGE
784 : : TRANSHUGE_PAGE_DTOR,
785 : : #endif
786 : : NR_COMPOUND_DTORS,
787 : : };
788 : : extern compound_page_dtor * const compound_page_dtors[];
789 : :
790 : : static inline void set_compound_page_dtor(struct page *page,
791 : : enum compound_dtor_id compound_dtor)
792 : : {
793 : : VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
794 : 3 : page[1].compound_dtor = compound_dtor;
795 : : }
796 : :
797 : : static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
798 : : {
799 : : VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
800 : 3 : return compound_page_dtors[page[1].compound_dtor];
801 : : }
802 : :
803 : : static inline unsigned int compound_order(struct page *page)
804 : : {
805 : 3 : if (!PageHead(page))
806 : : return 0;
807 : 3 : return page[1].compound_order;
808 : : }
809 : :
810 : : static inline void set_compound_order(struct page *page, unsigned int order)
811 : : {
812 : 3 : page[1].compound_order = order;
813 : : }
814 : :
815 : : /* Returns the number of pages in this potentially compound page. */
816 : : static inline unsigned long compound_nr(struct page *page)
817 : : {
818 : 3 : return 1UL << compound_order(page);
819 : : }
820 : :
821 : : /* Returns the number of bytes in this potentially compound page. */
822 : : static inline unsigned long page_size(struct page *page)
823 : : {
824 : 3 : return PAGE_SIZE << compound_order(page);
825 : : }
826 : :
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)
829 : : {
830 : : return PAGE_SHIFT + compound_order(page);
831 : : }
832 : :
833 : : void free_compound_page(struct page *page);
834 : :
835 : : #ifdef CONFIG_MMU
836 : : /*
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.
841 : : */
842 : : static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
843 : : {
844 : 3 : if (likely(vma->vm_flags & VM_WRITE))
845 : : pte = pte_mkwrite(pte);
846 : : return pte;
847 : : }
848 : :
849 : : vm_fault_t alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
850 : : struct page *page);
851 : : vm_fault_t finish_fault(struct vm_fault *vmf);
852 : : vm_fault_t finish_mkwrite_fault(struct vm_fault *vmf);
853 : : #endif
854 : :
855 : : /*
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.
860 : : *
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.
865 : : *
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).
871 : : *
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.
879 : : *
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.
883 : : *
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.
888 : : *
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.
892 : : *
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.
896 : : *
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.
900 : : *
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.
905 : : *
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
912 : : * back into memory.
913 : : */
914 : :
915 : : /*
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.
918 : : */
919 : :
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)
926 : :
927 : : /*
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.
931 : : */
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))
937 : :
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)
943 : : #else
944 : : #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
945 : : #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
946 : : NODES_PGOFF : ZONES_PGOFF)
947 : : #endif
948 : :
949 : : #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
950 : :
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
953 : : #endif
954 : :
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)
961 : :
962 : 3 : static inline enum zone_type page_zonenum(const struct page *page)
963 : : {
964 : 3 : return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
965 : : }
966 : :
967 : : #ifdef CONFIG_ZONE_DEVICE
968 : : static inline bool is_zone_device_page(const struct page *page)
969 : : {
970 : : return page_zonenum(page) == ZONE_DEVICE;
971 : : }
972 : : extern void memmap_init_zone_device(struct zone *, unsigned long,
973 : : unsigned long, struct dev_pagemap *);
974 : : #else
975 : : static inline bool is_zone_device_page(const struct page *page)
976 : : {
977 : : return false;
978 : : }
979 : : #endif
980 : :
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)
985 : : {
986 : : if (!static_branch_unlikely(&devmap_managed_key))
987 : : return false;
988 : : if (!is_zone_device_page(page))
989 : : return false;
990 : : switch (page->pgmap->type) {
991 : : case MEMORY_DEVICE_PRIVATE:
992 : : case MEMORY_DEVICE_FS_DAX:
993 : : __put_devmap_managed_page(page);
994 : : return true;
995 : : default:
996 : : break;
997 : : }
998 : : return false;
999 : : }
1000 : :
1001 : : #else /* CONFIG_DEV_PAGEMAP_OPS */
1002 : : static inline bool put_devmap_managed_page(struct page *page)
1003 : : {
1004 : : return false;
1005 : : }
1006 : : #endif /* CONFIG_DEV_PAGEMAP_OPS */
1007 : :
1008 : : static inline bool is_device_private_page(const struct page *page)
1009 : : {
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;
1014 : : }
1015 : :
1016 : : static inline bool is_pci_p2pdma_page(const struct page *page)
1017 : : {
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;
1022 : : }
1023 : :
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)
1027 : :
1028 : 3 : static inline void get_page(struct page *page)
1029 : : {
1030 : : page = compound_head(page);
1031 : : /*
1032 : : * Getting a normal page or the head of a compound page
1033 : : * requires to already have an elevated page->_refcount.
1034 : : */
1035 : : VM_BUG_ON_PAGE(page_ref_zero_or_close_to_overflow(page), page);
1036 : : page_ref_inc(page);
1037 : 3 : }
1038 : :
1039 : 3 : static inline __must_check bool try_get_page(struct page *page)
1040 : : {
1041 : : page = compound_head(page);
1042 : 3 : if (WARN_ON_ONCE(page_ref_count(page) <= 0))
1043 : : return false;
1044 : : page_ref_inc(page);
1045 : 3 : return true;
1046 : : }
1047 : :
1048 : 3 : static inline void put_page(struct page *page)
1049 : : {
1050 : : page = compound_head(page);
1051 : :
1052 : : /*
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.
1057 : : */
1058 : : if (put_devmap_managed_page(page))
1059 : 3 : return;
1060 : :
1061 : 3 : if (put_page_testzero(page))
1062 : 3 : __put_page(page);
1063 : : }
1064 : :
1065 : : /**
1066 : : * put_user_page() - release a gup-pinned page
1067 : : * @page: pointer to page to be released
1068 : : *
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
1074 : : * handling.
1075 : : *
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.
1079 : : */
1080 : : static inline void put_user_page(struct page *page)
1081 : : {
1082 : 0 : put_page(page);
1083 : : }
1084 : :
1085 : : void put_user_pages_dirty_lock(struct page **pages, unsigned long npages,
1086 : : bool make_dirty);
1087 : :
1088 : : void put_user_pages(struct page **pages, unsigned long npages);
1089 : :
1090 : : #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
1091 : : #define SECTION_IN_PAGE_FLAGS
1092 : : #endif
1093 : :
1094 : : /*
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
1100 : : * pages in a zone.
1101 : : */
1102 : : static inline int page_zone_id(struct page *page)
1103 : : {
1104 : 3 : return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
1105 : : }
1106 : :
1107 : : #ifdef NODE_NOT_IN_PAGE_FLAGS
1108 : : extern int page_to_nid(const struct page *page);
1109 : : #else
1110 : : static inline int page_to_nid(const struct page *page)
1111 : : {
1112 : : struct page *p = (struct page *)page;
1113 : :
1114 : : return (PF_POISONED_CHECK(p)->flags >> NODES_PGSHIFT) & NODES_MASK;
1115 : : }
1116 : : #endif
1117 : :
1118 : : #ifdef CONFIG_NUMA_BALANCING
1119 : : static inline int cpu_pid_to_cpupid(int cpu, int pid)
1120 : : {
1121 : : return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
1122 : : }
1123 : :
1124 : : static inline int cpupid_to_pid(int cpupid)
1125 : : {
1126 : : return cpupid & LAST__PID_MASK;
1127 : : }
1128 : :
1129 : : static inline int cpupid_to_cpu(int cpupid)
1130 : : {
1131 : : return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
1132 : : }
1133 : :
1134 : : static inline int cpupid_to_nid(int cpupid)
1135 : : {
1136 : : return cpu_to_node(cpupid_to_cpu(cpupid));
1137 : : }
1138 : :
1139 : : static inline bool cpupid_pid_unset(int cpupid)
1140 : : {
1141 : : return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
1142 : : }
1143 : :
1144 : : static inline bool cpupid_cpu_unset(int cpupid)
1145 : : {
1146 : : return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
1147 : : }
1148 : :
1149 : : static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
1150 : : {
1151 : : return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
1152 : : }
1153 : :
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)
1157 : : {
1158 : : return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
1159 : : }
1160 : :
1161 : : static inline int page_cpupid_last(struct page *page)
1162 : : {
1163 : : return page->_last_cpupid;
1164 : : }
1165 : : static inline void page_cpupid_reset_last(struct page *page)
1166 : : {
1167 : : page->_last_cpupid = -1 & LAST_CPUPID_MASK;
1168 : : }
1169 : : #else
1170 : : static inline int page_cpupid_last(struct page *page)
1171 : : {
1172 : : return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
1173 : : }
1174 : :
1175 : : extern int page_cpupid_xchg_last(struct page *page, int cpupid);
1176 : :
1177 : : static inline void page_cpupid_reset_last(struct page *page)
1178 : : {
1179 : : page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
1180 : : }
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)
1184 : : {
1185 : : return page_to_nid(page); /* XXX */
1186 : : }
1187 : :
1188 : : static inline int page_cpupid_last(struct page *page)
1189 : : {
1190 : : return page_to_nid(page); /* XXX */
1191 : : }
1192 : :
1193 : : static inline int cpupid_to_nid(int cpupid)
1194 : : {
1195 : : return -1;
1196 : : }
1197 : :
1198 : : static inline int cpupid_to_pid(int cpupid)
1199 : : {
1200 : : return -1;
1201 : : }
1202 : :
1203 : : static inline int cpupid_to_cpu(int cpupid)
1204 : : {
1205 : : return -1;
1206 : : }
1207 : :
1208 : : static inline int cpu_pid_to_cpupid(int nid, int pid)
1209 : : {
1210 : : return -1;
1211 : : }
1212 : :
1213 : : static inline bool cpupid_pid_unset(int cpupid)
1214 : : {
1215 : : return 1;
1216 : : }
1217 : :
1218 : : static inline void page_cpupid_reset_last(struct page *page)
1219 : : {
1220 : : }
1221 : :
1222 : : static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1223 : : {
1224 : : return false;
1225 : : }
1226 : : #endif /* CONFIG_NUMA_BALANCING */
1227 : :
1228 : : #ifdef CONFIG_KASAN_SW_TAGS
1229 : : static inline u8 page_kasan_tag(const struct page *page)
1230 : : {
1231 : : return (page->flags >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
1232 : : }
1233 : :
1234 : : static inline void page_kasan_tag_set(struct page *page, u8 tag)
1235 : : {
1236 : : page->flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
1237 : : page->flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
1238 : : }
1239 : :
1240 : : static inline void page_kasan_tag_reset(struct page *page)
1241 : : {
1242 : : page_kasan_tag_set(page, 0xff);
1243 : : }
1244 : : #else
1245 : : static inline u8 page_kasan_tag(const struct page *page)
1246 : : {
1247 : : return 0xff;
1248 : : }
1249 : :
1250 : : static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
1251 : : static inline void page_kasan_tag_reset(struct page *page) { }
1252 : : #endif
1253 : :
1254 : : static inline struct zone *page_zone(const struct page *page)
1255 : : {
1256 : 3 : return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1257 : : }
1258 : :
1259 : 3 : static inline pg_data_t *page_pgdat(const struct page *page)
1260 : : {
1261 : 3 : return NODE_DATA(page_to_nid(page));
1262 : : }
1263 : :
1264 : : #ifdef SECTION_IN_PAGE_FLAGS
1265 : : static inline void set_page_section(struct page *page, unsigned long section)
1266 : : {
1267 : : page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1268 : : page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1269 : : }
1270 : :
1271 : : static inline unsigned long page_to_section(const struct page *page)
1272 : : {
1273 : : return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1274 : : }
1275 : : #endif
1276 : :
1277 : : static inline void set_page_zone(struct page *page, enum zone_type zone)
1278 : : {
1279 : 3 : page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1280 : 3 : page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1281 : : }
1282 : :
1283 : : static inline void set_page_node(struct page *page, unsigned long node)
1284 : : {
1285 : : page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1286 : : page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1287 : : }
1288 : :
1289 : : static inline void set_page_links(struct page *page, enum zone_type zone,
1290 : : unsigned long node, unsigned long pfn)
1291 : : {
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));
1296 : : #endif
1297 : : }
1298 : :
1299 : : #ifdef CONFIG_MEMCG
1300 : : static inline struct mem_cgroup *page_memcg(struct page *page)
1301 : : {
1302 : 0 : return page->mem_cgroup;
1303 : : }
1304 : : static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1305 : : {
1306 : : WARN_ON_ONCE(!rcu_read_lock_held());
1307 : 3 : return READ_ONCE(page->mem_cgroup);
1308 : : }
1309 : : #else
1310 : : static inline struct mem_cgroup *page_memcg(struct page *page)
1311 : : {
1312 : : return NULL;
1313 : : }
1314 : : static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1315 : : {
1316 : : WARN_ON_ONCE(!rcu_read_lock_held());
1317 : : return NULL;
1318 : : }
1319 : : #endif
1320 : :
1321 : : /*
1322 : : * Some inline functions in vmstat.h depend on page_zone()
1323 : : */
1324 : : #include <linux/vmstat.h>
1325 : :
1326 : : static __always_inline void *lowmem_page_address(const struct page *page)
1327 : : {
1328 : 3 : return page_to_virt(page);
1329 : : }
1330 : :
1331 : : #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1332 : : #define HASHED_PAGE_VIRTUAL
1333 : : #endif
1334 : :
1335 : : #if defined(WANT_PAGE_VIRTUAL)
1336 : : static inline void *page_address(const struct page *page)
1337 : : {
1338 : : return page->virtual;
1339 : : }
1340 : : static inline void set_page_address(struct page *page, void *address)
1341 : : {
1342 : : page->virtual = address;
1343 : : }
1344 : : #define page_address_init() do { } while(0)
1345 : : #endif
1346 : :
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);
1351 : : #endif
1352 : :
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)
1357 : : #endif
1358 : :
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);
1362 : :
1363 : : extern struct address_space *__page_file_mapping(struct page *);
1364 : :
1365 : : static inline
1366 : 0 : struct address_space *page_file_mapping(struct page *page)
1367 : : {
1368 : 0 : if (unlikely(PageSwapCache(page)))
1369 : 0 : return __page_file_mapping(page);
1370 : :
1371 : 0 : return page->mapping;
1372 : : }
1373 : :
1374 : : extern pgoff_t __page_file_index(struct page *page);
1375 : :
1376 : : /*
1377 : : * Return the pagecache index of the passed page. Regular pagecache pages
1378 : : * use ->index whereas swapcache pages use swp_offset(->private)
1379 : : */
1380 : 3 : static inline pgoff_t page_index(struct page *page)
1381 : : {
1382 : 3 : if (unlikely(PageSwapCache(page)))
1383 : 0 : return __page_file_index(page);
1384 : 3 : return page->index;
1385 : : }
1386 : :
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);
1390 : :
1391 : : /*
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.
1395 : : */
1396 : : static inline bool page_is_pfmemalloc(struct page *page)
1397 : : {
1398 : : /*
1399 : : * Page index cannot be this large so this must be
1400 : : * a pfmemalloc page.
1401 : : */
1402 : 3 : return page->index == -1UL;
1403 : : }
1404 : :
1405 : : /*
1406 : : * Only to be called by the page allocator on a freshly allocated
1407 : : * page.
1408 : : */
1409 : : static inline void set_page_pfmemalloc(struct page *page)
1410 : : {
1411 : 0 : page->index = -1UL;
1412 : : }
1413 : :
1414 : : static inline void clear_page_pfmemalloc(struct page *page)
1415 : : {
1416 : 3 : page->index = 0;
1417 : : }
1418 : :
1419 : : /*
1420 : : * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1421 : : */
1422 : : extern void pagefault_out_of_memory(void);
1423 : :
1424 : : #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1425 : :
1426 : : /*
1427 : : * Flags passed to show_mem() and show_free_areas() to suppress output in
1428 : : * various contexts.
1429 : : */
1430 : : #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1431 : :
1432 : : extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1433 : :
1434 : : #ifdef CONFIG_MMU
1435 : : extern bool can_do_mlock(void);
1436 : : #else
1437 : : static inline bool can_do_mlock(void) { return false; }
1438 : : #endif
1439 : : extern int user_shm_lock(size_t, struct user_struct *);
1440 : : extern void user_shm_unlock(size_t, struct user_struct *);
1441 : :
1442 : : /*
1443 : : * Parameter block passed down to zap_pte_range in exceptional cases.
1444 : : */
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 */
1449 : : };
1450 : :
1451 : : struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1452 : : pte_t pte);
1453 : : struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1454 : : pmd_t pmd);
1455 : :
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);
1462 : :
1463 : : struct mmu_notifier_range;
1464 : :
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);
1478 : :
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);
1486 : :
1487 : : #ifdef CONFIG_MMU
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,
1492 : : bool *unlocked);
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);
1497 : : #else
1498 : : static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
1499 : : unsigned long address, unsigned int flags)
1500 : : {
1501 : : /* should never happen if there's no MMU */
1502 : : BUG();
1503 : : return VM_FAULT_SIGBUS;
1504 : : }
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)
1508 : : {
1509 : : /* should never happen if there's no MMU */
1510 : : BUG();
1511 : : return -EFAULT;
1512 : : }
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) { }
1517 : : #endif
1518 : :
1519 : : static inline void unmap_shared_mapping_range(struct address_space *mapping,
1520 : : loff_t const holebegin, loff_t const holelen)
1521 : : {
1522 : : unmap_mapping_range(mapping, holebegin, holelen, 0);
1523 : : }
1524 : :
1525 : : extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
1526 : : void *buf, int len, unsigned int gup_flags);
1527 : : extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1528 : : void *buf, int len, unsigned int gup_flags);
1529 : : extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1530 : : unsigned long addr, void *buf, int len, unsigned int gup_flags);
1531 : :
1532 : : long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1533 : : unsigned long start, unsigned long nr_pages,
1534 : : unsigned int gup_flags, struct page **pages,
1535 : : struct vm_area_struct **vmas, int *locked);
1536 : : long get_user_pages(unsigned long start, unsigned long nr_pages,
1537 : : unsigned int gup_flags, struct page **pages,
1538 : : struct vm_area_struct **vmas);
1539 : : long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1540 : : unsigned int gup_flags, struct page **pages, int *locked);
1541 : : long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1542 : : struct page **pages, unsigned int gup_flags);
1543 : :
1544 : : int get_user_pages_fast(unsigned long start, int nr_pages,
1545 : : unsigned int gup_flags, struct page **pages);
1546 : :
1547 : : int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
1548 : : int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
1549 : : struct task_struct *task, bool bypass_rlim);
1550 : :
1551 : : /* Container for pinned pfns / pages */
1552 : : struct frame_vector {
1553 : : unsigned int nr_allocated; /* Number of frames we have space for */
1554 : : unsigned int nr_frames; /* Number of frames stored in ptrs array */
1555 : : bool got_ref; /* Did we pin pages by getting page ref? */
1556 : : bool is_pfns; /* Does array contain pages or pfns? */
1557 : : void *ptrs[0]; /* Array of pinned pfns / pages. Use
1558 : : * pfns_vector_pages() or pfns_vector_pfns()
1559 : : * for access */
1560 : : };
1561 : :
1562 : : struct frame_vector *frame_vector_create(unsigned int nr_frames);
1563 : : void frame_vector_destroy(struct frame_vector *vec);
1564 : : int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1565 : : unsigned int gup_flags, struct frame_vector *vec);
1566 : : void put_vaddr_frames(struct frame_vector *vec);
1567 : : int frame_vector_to_pages(struct frame_vector *vec);
1568 : : void frame_vector_to_pfns(struct frame_vector *vec);
1569 : :
1570 : : static inline unsigned int frame_vector_count(struct frame_vector *vec)
1571 : : {
1572 : 0 : return vec->nr_frames;
1573 : : }
1574 : :
1575 : : static inline struct page **frame_vector_pages(struct frame_vector *vec)
1576 : : {
1577 : 0 : if (vec->is_pfns) {
1578 : 0 : int err = frame_vector_to_pages(vec);
1579 : :
1580 : 0 : if (err)
1581 : : return ERR_PTR(err);
1582 : : }
1583 : 0 : return (struct page **)(vec->ptrs);
1584 : : }
1585 : :
1586 : : static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1587 : : {
1588 : 0 : if (!vec->is_pfns)
1589 : 0 : frame_vector_to_pfns(vec);
1590 : 0 : return (unsigned long *)(vec->ptrs);
1591 : : }
1592 : :
1593 : : struct kvec;
1594 : : int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1595 : : struct page **pages);
1596 : : int get_kernel_page(unsigned long start, int write, struct page **pages);
1597 : : struct page *get_dump_page(unsigned long addr);
1598 : :
1599 : : extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1600 : : extern void do_invalidatepage(struct page *page, unsigned int offset,
1601 : : unsigned int length);
1602 : :
1603 : : void __set_page_dirty(struct page *, struct address_space *, int warn);
1604 : : int __set_page_dirty_nobuffers(struct page *page);
1605 : : int __set_page_dirty_no_writeback(struct page *page);
1606 : : int redirty_page_for_writepage(struct writeback_control *wbc,
1607 : : struct page *page);
1608 : : void account_page_dirtied(struct page *page, struct address_space *mapping);
1609 : : void account_page_cleaned(struct page *page, struct address_space *mapping,
1610 : : struct bdi_writeback *wb);
1611 : : int set_page_dirty(struct page *page);
1612 : : int set_page_dirty_lock(struct page *page);
1613 : : void __cancel_dirty_page(struct page *page);
1614 : 3 : static inline void cancel_dirty_page(struct page *page)
1615 : : {
1616 : : /* Avoid atomic ops, locking, etc. when not actually needed. */
1617 : 3 : if (PageDirty(page))
1618 : 3 : __cancel_dirty_page(page);
1619 : 3 : }
1620 : : int clear_page_dirty_for_io(struct page *page);
1621 : :
1622 : : int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1623 : :
1624 : : extern unsigned long move_page_tables(struct vm_area_struct *vma,
1625 : : unsigned long old_addr, struct vm_area_struct *new_vma,
1626 : : unsigned long new_addr, unsigned long len,
1627 : : bool need_rmap_locks);
1628 : : extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1629 : : unsigned long end, pgprot_t newprot,
1630 : : int dirty_accountable, int prot_numa);
1631 : : extern int mprotect_fixup(struct vm_area_struct *vma,
1632 : : struct vm_area_struct **pprev, unsigned long start,
1633 : : unsigned long end, unsigned long newflags);
1634 : :
1635 : : /*
1636 : : * doesn't attempt to fault and will return short.
1637 : : */
1638 : : int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1639 : : struct page **pages);
1640 : : /*
1641 : : * per-process(per-mm_struct) statistics.
1642 : : */
1643 : : static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1644 : : {
1645 : : long val = atomic_long_read(&mm->rss_stat.count[member]);
1646 : :
1647 : : #ifdef SPLIT_RSS_COUNTING
1648 : : /*
1649 : : * counter is updated in asynchronous manner and may go to minus.
1650 : : * But it's never be expected number for users.
1651 : : */
1652 : 3 : if (val < 0)
1653 : : val = 0;
1654 : : #endif
1655 : 3 : return (unsigned long)val;
1656 : : }
1657 : :
1658 : : static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1659 : : {
1660 : 3 : atomic_long_add(value, &mm->rss_stat.count[member]);
1661 : : }
1662 : :
1663 : : static inline void inc_mm_counter(struct mm_struct *mm, int member)
1664 : : {
1665 : 0 : atomic_long_inc(&mm->rss_stat.count[member]);
1666 : : }
1667 : :
1668 : : static inline void dec_mm_counter(struct mm_struct *mm, int member)
1669 : : {
1670 : 0 : atomic_long_dec(&mm->rss_stat.count[member]);
1671 : : }
1672 : :
1673 : : /* Optimized variant when page is already known not to be PageAnon */
1674 : 3 : static inline int mm_counter_file(struct page *page)
1675 : : {
1676 : 3 : if (PageSwapBacked(page))
1677 : : return MM_SHMEMPAGES;
1678 : 3 : return MM_FILEPAGES;
1679 : : }
1680 : :
1681 : 3 : static inline int mm_counter(struct page *page)
1682 : : {
1683 : 3 : if (PageAnon(page))
1684 : : return MM_ANONPAGES;
1685 : 3 : return mm_counter_file(page);
1686 : : }
1687 : :
1688 : : static inline unsigned long get_mm_rss(struct mm_struct *mm)
1689 : : {
1690 : 3 : return get_mm_counter(mm, MM_FILEPAGES) +
1691 : : get_mm_counter(mm, MM_ANONPAGES) +
1692 : : get_mm_counter(mm, MM_SHMEMPAGES);
1693 : : }
1694 : :
1695 : : static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1696 : : {
1697 : 3 : return max(mm->hiwater_rss, get_mm_rss(mm));
1698 : : }
1699 : :
1700 : : static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1701 : : {
1702 : 0 : return max(mm->hiwater_vm, mm->total_vm);
1703 : : }
1704 : :
1705 : : static inline void update_hiwater_rss(struct mm_struct *mm)
1706 : : {
1707 : : unsigned long _rss = get_mm_rss(mm);
1708 : :
1709 : 3 : if ((mm)->hiwater_rss < _rss)
1710 : 3 : (mm)->hiwater_rss = _rss;
1711 : : }
1712 : :
1713 : : static inline void update_hiwater_vm(struct mm_struct *mm)
1714 : : {
1715 : 3 : if (mm->hiwater_vm < mm->total_vm)
1716 : 3 : mm->hiwater_vm = mm->total_vm;
1717 : : }
1718 : :
1719 : : static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1720 : : {
1721 : 0 : mm->hiwater_rss = get_mm_rss(mm);
1722 : : }
1723 : :
1724 : : static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1725 : : struct mm_struct *mm)
1726 : : {
1727 : : unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1728 : :
1729 : 3 : if (*maxrss < hiwater_rss)
1730 : 3 : *maxrss = hiwater_rss;
1731 : : }
1732 : :
1733 : : #if defined(SPLIT_RSS_COUNTING)
1734 : : void sync_mm_rss(struct mm_struct *mm);
1735 : : #else
1736 : : static inline void sync_mm_rss(struct mm_struct *mm)
1737 : : {
1738 : : }
1739 : : #endif
1740 : :
1741 : : #ifndef CONFIG_ARCH_HAS_PTE_DEVMAP
1742 : : static inline int pte_devmap(pte_t pte)
1743 : : {
1744 : : return 0;
1745 : : }
1746 : : #endif
1747 : :
1748 : : int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1749 : :
1750 : : extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1751 : : spinlock_t **ptl);
1752 : : static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1753 : : spinlock_t **ptl)
1754 : : {
1755 : : pte_t *ptep;
1756 : 3 : __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1757 : : return ptep;
1758 : : }
1759 : :
1760 : : #ifdef __PAGETABLE_P4D_FOLDED
1761 : : static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1762 : : unsigned long address)
1763 : : {
1764 : : return 0;
1765 : : }
1766 : : #else
1767 : : int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1768 : : #endif
1769 : :
1770 : : #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
1771 : : static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1772 : : unsigned long address)
1773 : : {
1774 : : return 0;
1775 : : }
1776 : : static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
1777 : : static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
1778 : :
1779 : : #else
1780 : : int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
1781 : :
1782 : : static inline void mm_inc_nr_puds(struct mm_struct *mm)
1783 : : {
1784 : : if (mm_pud_folded(mm))
1785 : : return;
1786 : : atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1787 : : }
1788 : :
1789 : : static inline void mm_dec_nr_puds(struct mm_struct *mm)
1790 : : {
1791 : : if (mm_pud_folded(mm))
1792 : : return;
1793 : : atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
1794 : : }
1795 : : #endif
1796 : :
1797 : : #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1798 : : static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1799 : : unsigned long address)
1800 : : {
1801 : : return 0;
1802 : : }
1803 : :
1804 : : static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1805 : : static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1806 : :
1807 : : #else
1808 : : int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1809 : :
1810 : : static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1811 : : {
1812 : : if (mm_pmd_folded(mm))
1813 : : return;
1814 : : atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1815 : : }
1816 : :
1817 : : static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1818 : : {
1819 : : if (mm_pmd_folded(mm))
1820 : : return;
1821 : : atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
1822 : : }
1823 : : #endif
1824 : :
1825 : : #ifdef CONFIG_MMU
1826 : : static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
1827 : : {
1828 : : atomic_long_set(&mm->pgtables_bytes, 0);
1829 : : }
1830 : :
1831 : : static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1832 : : {
1833 : : return atomic_long_read(&mm->pgtables_bytes);
1834 : : }
1835 : :
1836 : : static inline void mm_inc_nr_ptes(struct mm_struct *mm)
1837 : : {
1838 : 3 : atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1839 : : }
1840 : :
1841 : : static inline void mm_dec_nr_ptes(struct mm_struct *mm)
1842 : : {
1843 : 3 : atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
1844 : : }
1845 : : #else
1846 : :
1847 : : static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
1848 : : static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
1849 : : {
1850 : : return 0;
1851 : : }
1852 : :
1853 : : static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
1854 : : static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
1855 : : #endif
1856 : :
1857 : : int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
1858 : : int __pte_alloc_kernel(pmd_t *pmd);
1859 : :
1860 : : /*
1861 : : * The following ifdef needed to get the 4level-fixup.h header to work.
1862 : : * Remove it when 4level-fixup.h has been removed.
1863 : : */
1864 : : #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1865 : :
1866 : : #ifndef __ARCH_HAS_5LEVEL_HACK
1867 : : static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1868 : : unsigned long address)
1869 : : {
1870 : : return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
1871 : : NULL : p4d_offset(pgd, address);
1872 : : }
1873 : :
1874 : : static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1875 : : unsigned long address)
1876 : : {
1877 : : return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
1878 : : NULL : pud_offset(p4d, address);
1879 : : }
1880 : : #endif /* !__ARCH_HAS_5LEVEL_HACK */
1881 : :
1882 : : static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1883 : : {
1884 : : return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1885 : : NULL: pmd_offset(pud, address);
1886 : : }
1887 : : #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1888 : :
1889 : : #if USE_SPLIT_PTE_PTLOCKS
1890 : : #if ALLOC_SPLIT_PTLOCKS
1891 : : void __init ptlock_cache_init(void);
1892 : : extern bool ptlock_alloc(struct page *page);
1893 : : extern void ptlock_free(struct page *page);
1894 : :
1895 : : static inline spinlock_t *ptlock_ptr(struct page *page)
1896 : : {
1897 : : return page->ptl;
1898 : : }
1899 : : #else /* ALLOC_SPLIT_PTLOCKS */
1900 : : static inline void ptlock_cache_init(void)
1901 : : {
1902 : : }
1903 : :
1904 : : static inline bool ptlock_alloc(struct page *page)
1905 : : {
1906 : : return true;
1907 : : }
1908 : :
1909 : : static inline void ptlock_free(struct page *page)
1910 : : {
1911 : : }
1912 : :
1913 : : static inline spinlock_t *ptlock_ptr(struct page *page)
1914 : : {
1915 : 3 : return &page->ptl;
1916 : : }
1917 : : #endif /* ALLOC_SPLIT_PTLOCKS */
1918 : :
1919 : : static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1920 : : {
1921 : 3 : return ptlock_ptr(pmd_page(*pmd));
1922 : : }
1923 : :
1924 : : static inline bool ptlock_init(struct page *page)
1925 : : {
1926 : : /*
1927 : : * prep_new_page() initialize page->private (and therefore page->ptl)
1928 : : * with 0. Make sure nobody took it in use in between.
1929 : : *
1930 : : * It can happen if arch try to use slab for page table allocation:
1931 : : * slab code uses page->slab_cache, which share storage with page->ptl.
1932 : : */
1933 : : VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1934 : : if (!ptlock_alloc(page))
1935 : : return false;
1936 : 3 : spin_lock_init(ptlock_ptr(page));
1937 : : return true;
1938 : : }
1939 : :
1940 : : #else /* !USE_SPLIT_PTE_PTLOCKS */
1941 : : /*
1942 : : * We use mm->page_table_lock to guard all pagetable pages of the mm.
1943 : : */
1944 : : static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1945 : : {
1946 : : return &mm->page_table_lock;
1947 : : }
1948 : : static inline void ptlock_cache_init(void) {}
1949 : : static inline bool ptlock_init(struct page *page) { return true; }
1950 : : static inline void ptlock_free(struct page *page) {}
1951 : : #endif /* USE_SPLIT_PTE_PTLOCKS */
1952 : :
1953 : : static inline void pgtable_init(void)
1954 : : {
1955 : : ptlock_cache_init();
1956 : 3 : pgtable_cache_init();
1957 : : }
1958 : :
1959 : : static inline bool pgtable_pte_page_ctor(struct page *page)
1960 : : {
1961 : : if (!ptlock_init(page))
1962 : : return false;
1963 : : __SetPageTable(page);
1964 : 3 : inc_zone_page_state(page, NR_PAGETABLE);
1965 : : return true;
1966 : : }
1967 : :
1968 : : static inline void pgtable_pte_page_dtor(struct page *page)
1969 : : {
1970 : : ptlock_free(page);
1971 : : __ClearPageTable(page);
1972 : 3 : dec_zone_page_state(page, NR_PAGETABLE);
1973 : : }
1974 : :
1975 : : #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1976 : : ({ \
1977 : : spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1978 : : pte_t *__pte = pte_offset_map(pmd, address); \
1979 : : *(ptlp) = __ptl; \
1980 : : spin_lock(__ptl); \
1981 : : __pte; \
1982 : : })
1983 : :
1984 : : #define pte_unmap_unlock(pte, ptl) do { \
1985 : : spin_unlock(ptl); \
1986 : : pte_unmap(pte); \
1987 : : } while (0)
1988 : :
1989 : : #define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
1990 : :
1991 : : #define pte_alloc_map(mm, pmd, address) \
1992 : : (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
1993 : :
1994 : : #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1995 : : (pte_alloc(mm, pmd) ? \
1996 : : NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
1997 : :
1998 : : #define pte_alloc_kernel(pmd, address) \
1999 : : ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
2000 : : NULL: pte_offset_kernel(pmd, address))
2001 : :
2002 : : #if USE_SPLIT_PMD_PTLOCKS
2003 : :
2004 : : static struct page *pmd_to_page(pmd_t *pmd)
2005 : : {
2006 : : unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
2007 : : return virt_to_page((void *)((unsigned long) pmd & mask));
2008 : : }
2009 : :
2010 : : static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2011 : : {
2012 : : return ptlock_ptr(pmd_to_page(pmd));
2013 : : }
2014 : :
2015 : : static inline bool pgtable_pmd_page_ctor(struct page *page)
2016 : : {
2017 : : #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2018 : : page->pmd_huge_pte = NULL;
2019 : : #endif
2020 : : return ptlock_init(page);
2021 : : }
2022 : :
2023 : : static inline void pgtable_pmd_page_dtor(struct page *page)
2024 : : {
2025 : : #ifdef CONFIG_TRANSPARENT_HUGEPAGE
2026 : : VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
2027 : : #endif
2028 : : ptlock_free(page);
2029 : : }
2030 : :
2031 : : #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2032 : :
2033 : : #else
2034 : :
2035 : : static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
2036 : : {
2037 : 3 : return &mm->page_table_lock;
2038 : : }
2039 : :
2040 : : static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
2041 : : static inline void pgtable_pmd_page_dtor(struct page *page) {}
2042 : :
2043 : : #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2044 : :
2045 : : #endif
2046 : :
2047 : : static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
2048 : : {
2049 : : spinlock_t *ptl = pmd_lockptr(mm, pmd);
2050 : : spin_lock(ptl);
2051 : : return ptl;
2052 : : }
2053 : :
2054 : : /*
2055 : : * No scalability reason to split PUD locks yet, but follow the same pattern
2056 : : * as the PMD locks to make it easier if we decide to. The VM should not be
2057 : : * considered ready to switch to split PUD locks yet; there may be places
2058 : : * which need to be converted from page_table_lock.
2059 : : */
2060 : : static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
2061 : : {
2062 : : return &mm->page_table_lock;
2063 : : }
2064 : :
2065 : : static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
2066 : : {
2067 : : spinlock_t *ptl = pud_lockptr(mm, pud);
2068 : :
2069 : : spin_lock(ptl);
2070 : : return ptl;
2071 : : }
2072 : :
2073 : : extern void __init pagecache_init(void);
2074 : : extern void free_area_init(unsigned long * zones_size);
2075 : : extern void __init free_area_init_node(int nid, unsigned long * zones_size,
2076 : : unsigned long zone_start_pfn, unsigned long *zholes_size);
2077 : : extern void free_initmem(void);
2078 : :
2079 : : /*
2080 : : * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2081 : : * into the buddy system. The freed pages will be poisoned with pattern
2082 : : * "poison" if it's within range [0, UCHAR_MAX].
2083 : : * Return pages freed into the buddy system.
2084 : : */
2085 : : extern unsigned long free_reserved_area(void *start, void *end,
2086 : : int poison, const char *s);
2087 : :
2088 : : #ifdef CONFIG_HIGHMEM
2089 : : /*
2090 : : * Free a highmem page into the buddy system, adjusting totalhigh_pages
2091 : : * and totalram_pages.
2092 : : */
2093 : : extern void free_highmem_page(struct page *page);
2094 : : #endif
2095 : :
2096 : : extern void adjust_managed_page_count(struct page *page, long count);
2097 : : extern void mem_init_print_info(const char *str);
2098 : :
2099 : : extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
2100 : :
2101 : : /* Free the reserved page into the buddy system, so it gets managed. */
2102 : 3 : static inline void __free_reserved_page(struct page *page)
2103 : : {
2104 : : ClearPageReserved(page);
2105 : : init_page_count(page);
2106 : 3 : __free_page(page);
2107 : 3 : }
2108 : :
2109 : : static inline void free_reserved_page(struct page *page)
2110 : : {
2111 : 3 : __free_reserved_page(page);
2112 : 3 : adjust_managed_page_count(page, 1);
2113 : : }
2114 : :
2115 : : static inline void mark_page_reserved(struct page *page)
2116 : : {
2117 : : SetPageReserved(page);
2118 : : adjust_managed_page_count(page, -1);
2119 : : }
2120 : :
2121 : : /*
2122 : : * Default method to free all the __init memory into the buddy system.
2123 : : * The freed pages will be poisoned with pattern "poison" if it's within
2124 : : * range [0, UCHAR_MAX].
2125 : : * Return pages freed into the buddy system.
2126 : : */
2127 : : static inline unsigned long free_initmem_default(int poison)
2128 : : {
2129 : : extern char __init_begin[], __init_end[];
2130 : :
2131 : 3 : return free_reserved_area(&__init_begin, &__init_end,
2132 : : poison, "unused kernel");
2133 : : }
2134 : :
2135 : : static inline unsigned long get_num_physpages(void)
2136 : : {
2137 : : int nid;
2138 : : unsigned long phys_pages = 0;
2139 : :
2140 : 3 : for_each_online_node(nid)
2141 : 3 : phys_pages += node_present_pages(nid);
2142 : :
2143 : 3 : return phys_pages;
2144 : : }
2145 : :
2146 : : #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
2147 : : /*
2148 : : * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
2149 : : * zones, allocate the backing mem_map and account for memory holes in a more
2150 : : * architecture independent manner. This is a substitute for creating the
2151 : : * zone_sizes[] and zholes_size[] arrays and passing them to
2152 : : * free_area_init_node()
2153 : : *
2154 : : * An architecture is expected to register range of page frames backed by
2155 : : * physical memory with memblock_add[_node]() before calling
2156 : : * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
2157 : : * usage, an architecture is expected to do something like
2158 : : *
2159 : : * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2160 : : * max_highmem_pfn};
2161 : : * for_each_valid_physical_page_range()
2162 : : * memblock_add_node(base, size, nid)
2163 : : * free_area_init_nodes(max_zone_pfns);
2164 : : *
2165 : : * free_bootmem_with_active_regions() calls free_bootmem_node() for each
2166 : : * registered physical page range. Similarly
2167 : : * sparse_memory_present_with_active_regions() calls memory_present() for
2168 : : * each range when SPARSEMEM is enabled.
2169 : : *
2170 : : * See mm/page_alloc.c for more information on each function exposed by
2171 : : * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
2172 : : */
2173 : : extern void free_area_init_nodes(unsigned long *max_zone_pfn);
2174 : : unsigned long node_map_pfn_alignment(void);
2175 : : unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2176 : : unsigned long end_pfn);
2177 : : extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2178 : : unsigned long end_pfn);
2179 : : extern void get_pfn_range_for_nid(unsigned int nid,
2180 : : unsigned long *start_pfn, unsigned long *end_pfn);
2181 : : extern unsigned long find_min_pfn_with_active_regions(void);
2182 : : extern void free_bootmem_with_active_regions(int nid,
2183 : : unsigned long max_low_pfn);
2184 : : extern void sparse_memory_present_with_active_regions(int nid);
2185 : :
2186 : : #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
2187 : :
2188 : : #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
2189 : : !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
2190 : : static inline int __early_pfn_to_nid(unsigned long pfn,
2191 : : struct mminit_pfnnid_cache *state)
2192 : : {
2193 : : return 0;
2194 : : }
2195 : : #else
2196 : : /* please see mm/page_alloc.c */
2197 : : extern int __meminit early_pfn_to_nid(unsigned long pfn);
2198 : : /* there is a per-arch backend function. */
2199 : : extern int __meminit __early_pfn_to_nid(unsigned long pfn,
2200 : : struct mminit_pfnnid_cache *state);
2201 : : #endif
2202 : :
2203 : : #if !defined(CONFIG_FLAT_NODE_MEM_MAP)
2204 : : void zero_resv_unavail(void);
2205 : : #else
2206 : : static inline void zero_resv_unavail(void) {}
2207 : : #endif
2208 : :
2209 : : extern void set_dma_reserve(unsigned long new_dma_reserve);
2210 : : extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long,
2211 : : enum memmap_context, struct vmem_altmap *);
2212 : : extern void setup_per_zone_wmarks(void);
2213 : : extern int __meminit init_per_zone_wmark_min(void);
2214 : : extern void mem_init(void);
2215 : : extern void __init mmap_init(void);
2216 : : extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2217 : : extern long si_mem_available(void);
2218 : : extern void si_meminfo(struct sysinfo * val);
2219 : : extern void si_meminfo_node(struct sysinfo *val, int nid);
2220 : : #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2221 : : extern unsigned long arch_reserved_kernel_pages(void);
2222 : : #endif
2223 : :
2224 : : extern __printf(3, 4)
2225 : : void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2226 : :
2227 : : extern void setup_per_cpu_pageset(void);
2228 : :
2229 : : extern void zone_pcp_update(struct zone *zone);
2230 : : extern void zone_pcp_reset(struct zone *zone);
2231 : :
2232 : : /* page_alloc.c */
2233 : : extern int min_free_kbytes;
2234 : : extern int watermark_boost_factor;
2235 : : extern int watermark_scale_factor;
2236 : :
2237 : : /* nommu.c */
2238 : : extern atomic_long_t mmap_pages_allocated;
2239 : : extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2240 : :
2241 : : /* interval_tree.c */
2242 : : void vma_interval_tree_insert(struct vm_area_struct *node,
2243 : : struct rb_root_cached *root);
2244 : : void vma_interval_tree_insert_after(struct vm_area_struct *node,
2245 : : struct vm_area_struct *prev,
2246 : : struct rb_root_cached *root);
2247 : : void vma_interval_tree_remove(struct vm_area_struct *node,
2248 : : struct rb_root_cached *root);
2249 : : struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2250 : : unsigned long start, unsigned long last);
2251 : : struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2252 : : unsigned long start, unsigned long last);
2253 : :
2254 : : #define vma_interval_tree_foreach(vma, root, start, last) \
2255 : : for (vma = vma_interval_tree_iter_first(root, start, last); \
2256 : : vma; vma = vma_interval_tree_iter_next(vma, start, last))
2257 : :
2258 : : void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2259 : : struct rb_root_cached *root);
2260 : : void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2261 : : struct rb_root_cached *root);
2262 : : struct anon_vma_chain *
2263 : : anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2264 : : unsigned long start, unsigned long last);
2265 : : struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2266 : : struct anon_vma_chain *node, unsigned long start, unsigned long last);
2267 : : #ifdef CONFIG_DEBUG_VM_RB
2268 : : void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2269 : : #endif
2270 : :
2271 : : #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2272 : : for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2273 : : avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2274 : :
2275 : : /* mmap.c */
2276 : : extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2277 : : extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2278 : : unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2279 : : struct vm_area_struct *expand);
2280 : : static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2281 : : unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2282 : : {
2283 : 3 : return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2284 : : }
2285 : : extern struct vm_area_struct *vma_merge(struct mm_struct *,
2286 : : struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2287 : : unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2288 : : struct mempolicy *, struct vm_userfaultfd_ctx);
2289 : : extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2290 : : extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2291 : : unsigned long addr, int new_below);
2292 : : extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2293 : : unsigned long addr, int new_below);
2294 : : extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2295 : : extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2296 : : struct rb_node **, struct rb_node *);
2297 : : extern void unlink_file_vma(struct vm_area_struct *);
2298 : : extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2299 : : unsigned long addr, unsigned long len, pgoff_t pgoff,
2300 : : bool *need_rmap_locks);
2301 : : extern void exit_mmap(struct mm_struct *);
2302 : :
2303 : : static inline int check_data_rlimit(unsigned long rlim,
2304 : : unsigned long new,
2305 : : unsigned long start,
2306 : : unsigned long end_data,
2307 : : unsigned long start_data)
2308 : : {
2309 : 3 : if (rlim < RLIM_INFINITY) {
2310 : 0 : if (((new - start) + (end_data - start_data)) > rlim)
2311 : : return -ENOSPC;
2312 : : }
2313 : :
2314 : : return 0;
2315 : : }
2316 : :
2317 : : extern int mm_take_all_locks(struct mm_struct *mm);
2318 : : extern void mm_drop_all_locks(struct mm_struct *mm);
2319 : :
2320 : : extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2321 : : extern struct file *get_mm_exe_file(struct mm_struct *mm);
2322 : : extern struct file *get_task_exe_file(struct task_struct *task);
2323 : :
2324 : : extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2325 : : extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2326 : :
2327 : : extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2328 : : const struct vm_special_mapping *sm);
2329 : : extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2330 : : unsigned long addr, unsigned long len,
2331 : : unsigned long flags,
2332 : : const struct vm_special_mapping *spec);
2333 : : /* This is an obsolete alternative to _install_special_mapping. */
2334 : : extern int install_special_mapping(struct mm_struct *mm,
2335 : : unsigned long addr, unsigned long len,
2336 : : unsigned long flags, struct page **pages);
2337 : :
2338 : : unsigned long randomize_stack_top(unsigned long stack_top);
2339 : :
2340 : : extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2341 : :
2342 : : extern unsigned long mmap_region(struct file *file, unsigned long addr,
2343 : : unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2344 : : struct list_head *uf);
2345 : : extern unsigned long do_mmap(struct file *file, unsigned long addr,
2346 : : unsigned long len, unsigned long prot, unsigned long flags,
2347 : : vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate,
2348 : : struct list_head *uf);
2349 : : extern int __do_munmap(struct mm_struct *, unsigned long, size_t,
2350 : : struct list_head *uf, bool downgrade);
2351 : : extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2352 : : struct list_head *uf);
2353 : :
2354 : : static inline unsigned long
2355 : : do_mmap_pgoff(struct file *file, unsigned long addr,
2356 : : unsigned long len, unsigned long prot, unsigned long flags,
2357 : : unsigned long pgoff, unsigned long *populate,
2358 : : struct list_head *uf)
2359 : : {
2360 : 3 : return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate, uf);
2361 : : }
2362 : :
2363 : : #ifdef CONFIG_MMU
2364 : : extern int __mm_populate(unsigned long addr, unsigned long len,
2365 : : int ignore_errors);
2366 : : static inline void mm_populate(unsigned long addr, unsigned long len)
2367 : : {
2368 : : /* Ignore errors */
2369 : 2 : (void) __mm_populate(addr, len, 1);
2370 : : }
2371 : : #else
2372 : : static inline void mm_populate(unsigned long addr, unsigned long len) {}
2373 : : #endif
2374 : :
2375 : : /* These take the mm semaphore themselves */
2376 : : extern int __must_check vm_brk(unsigned long, unsigned long);
2377 : : extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2378 : : extern int vm_munmap(unsigned long, size_t);
2379 : : extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2380 : : unsigned long, unsigned long,
2381 : : unsigned long, unsigned long);
2382 : :
2383 : : struct vm_unmapped_area_info {
2384 : : #define VM_UNMAPPED_AREA_TOPDOWN 1
2385 : : unsigned long flags;
2386 : : unsigned long length;
2387 : : unsigned long low_limit;
2388 : : unsigned long high_limit;
2389 : : unsigned long align_mask;
2390 : : unsigned long align_offset;
2391 : : };
2392 : :
2393 : : extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2394 : : extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2395 : :
2396 : : /*
2397 : : * Search for an unmapped address range.
2398 : : *
2399 : : * We are looking for a range that:
2400 : : * - does not intersect with any VMA;
2401 : : * - is contained within the [low_limit, high_limit) interval;
2402 : : * - is at least the desired size.
2403 : : * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2404 : : */
2405 : : static inline unsigned long
2406 : 3 : vm_unmapped_area(struct vm_unmapped_area_info *info)
2407 : : {
2408 : 3 : if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2409 : 3 : return unmapped_area_topdown(info);
2410 : : else
2411 : 0 : return unmapped_area(info);
2412 : : }
2413 : :
2414 : : /* truncate.c */
2415 : : extern void truncate_inode_pages(struct address_space *, loff_t);
2416 : : extern void truncate_inode_pages_range(struct address_space *,
2417 : : loff_t lstart, loff_t lend);
2418 : : extern void truncate_inode_pages_final(struct address_space *);
2419 : :
2420 : : /* generic vm_area_ops exported for stackable file systems */
2421 : : extern vm_fault_t filemap_fault(struct vm_fault *vmf);
2422 : : extern void filemap_map_pages(struct vm_fault *vmf,
2423 : : pgoff_t start_pgoff, pgoff_t end_pgoff);
2424 : : extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
2425 : :
2426 : : /* mm/page-writeback.c */
2427 : : int __must_check write_one_page(struct page *page);
2428 : : void task_dirty_inc(struct task_struct *tsk);
2429 : :
2430 : : /* readahead.c */
2431 : : #define VM_READAHEAD_PAGES (SZ_128K / PAGE_SIZE)
2432 : :
2433 : : int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2434 : : pgoff_t offset, unsigned long nr_to_read);
2435 : :
2436 : : void page_cache_sync_readahead(struct address_space *mapping,
2437 : : struct file_ra_state *ra,
2438 : : struct file *filp,
2439 : : pgoff_t offset,
2440 : : unsigned long size);
2441 : :
2442 : : void page_cache_async_readahead(struct address_space *mapping,
2443 : : struct file_ra_state *ra,
2444 : : struct file *filp,
2445 : : struct page *pg,
2446 : : pgoff_t offset,
2447 : : unsigned long size);
2448 : :
2449 : : extern unsigned long stack_guard_gap;
2450 : : /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2451 : : extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2452 : :
2453 : : /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2454 : : extern int expand_downwards(struct vm_area_struct *vma,
2455 : : unsigned long address);
2456 : : #if VM_GROWSUP
2457 : : extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2458 : : #else
2459 : : #define expand_upwards(vma, address) (0)
2460 : : #endif
2461 : :
2462 : : /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2463 : : extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2464 : : extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2465 : : struct vm_area_struct **pprev);
2466 : :
2467 : : /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2468 : : NULL if none. Assume start_addr < end_addr. */
2469 : : static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2470 : : {
2471 : 0 : struct vm_area_struct * vma = find_vma(mm,start_addr);
2472 : :
2473 : 0 : if (vma && end_addr <= vma->vm_start)
2474 : : vma = NULL;
2475 : : return vma;
2476 : : }
2477 : :
2478 : : static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2479 : : {
2480 : 3 : unsigned long vm_start = vma->vm_start;
2481 : :
2482 : 3 : if (vma->vm_flags & VM_GROWSDOWN) {
2483 : 3 : vm_start -= stack_guard_gap;
2484 : 3 : if (vm_start > vma->vm_start)
2485 : : vm_start = 0;
2486 : : }
2487 : : return vm_start;
2488 : : }
2489 : :
2490 : : static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2491 : : {
2492 : 3 : unsigned long vm_end = vma->vm_end;
2493 : :
2494 : : if (vma->vm_flags & VM_GROWSUP) {
2495 : : vm_end += stack_guard_gap;
2496 : : if (vm_end < vma->vm_end)
2497 : : vm_end = -PAGE_SIZE;
2498 : : }
2499 : : return vm_end;
2500 : : }
2501 : :
2502 : : static inline unsigned long vma_pages(struct vm_area_struct *vma)
2503 : : {
2504 : 3 : return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2505 : : }
2506 : :
2507 : : /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2508 : 0 : static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2509 : : unsigned long vm_start, unsigned long vm_end)
2510 : : {
2511 : 0 : struct vm_area_struct *vma = find_vma(mm, vm_start);
2512 : :
2513 : 0 : if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2514 : : vma = NULL;
2515 : :
2516 : 0 : return vma;
2517 : : }
2518 : :
2519 : : static inline bool range_in_vma(struct vm_area_struct *vma,
2520 : : unsigned long start, unsigned long end)
2521 : : {
2522 : : return (vma && vma->vm_start <= start && end <= vma->vm_end);
2523 : : }
2524 : :
2525 : : #ifdef CONFIG_MMU
2526 : : pgprot_t vm_get_page_prot(unsigned long vm_flags);
2527 : : void vma_set_page_prot(struct vm_area_struct *vma);
2528 : : #else
2529 : : static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2530 : : {
2531 : : return __pgprot(0);
2532 : : }
2533 : : static inline void vma_set_page_prot(struct vm_area_struct *vma)
2534 : : {
2535 : : vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2536 : : }
2537 : : #endif
2538 : :
2539 : : #ifdef CONFIG_NUMA_BALANCING
2540 : : unsigned long change_prot_numa(struct vm_area_struct *vma,
2541 : : unsigned long start, unsigned long end);
2542 : : #endif
2543 : :
2544 : : struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2545 : : int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2546 : : unsigned long pfn, unsigned long size, pgprot_t);
2547 : : int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2548 : : int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
2549 : : unsigned long num);
2550 : : int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
2551 : : unsigned long num);
2552 : : vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2553 : : unsigned long pfn);
2554 : : vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2555 : : unsigned long pfn, pgprot_t pgprot);
2556 : : vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2557 : : pfn_t pfn);
2558 : : vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
2559 : : unsigned long addr, pfn_t pfn);
2560 : : int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2561 : :
2562 : : static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
2563 : : unsigned long addr, struct page *page)
2564 : : {
2565 : : int err = vm_insert_page(vma, addr, page);
2566 : :
2567 : : if (err == -ENOMEM)
2568 : : return VM_FAULT_OOM;
2569 : : if (err < 0 && err != -EBUSY)
2570 : : return VM_FAULT_SIGBUS;
2571 : :
2572 : : return VM_FAULT_NOPAGE;
2573 : : }
2574 : :
2575 : : static inline vm_fault_t vmf_error(int err)
2576 : : {
2577 : 0 : if (err == -ENOMEM)
2578 : : return VM_FAULT_OOM;
2579 : : return VM_FAULT_SIGBUS;
2580 : : }
2581 : :
2582 : : struct page *follow_page(struct vm_area_struct *vma, unsigned long address,
2583 : : unsigned int foll_flags);
2584 : :
2585 : : #define FOLL_WRITE 0x01 /* check pte is writable */
2586 : : #define FOLL_TOUCH 0x02 /* mark page accessed */
2587 : : #define FOLL_GET 0x04 /* do get_page on page */
2588 : : #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2589 : : #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2590 : : #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2591 : : * and return without waiting upon it */
2592 : : #define FOLL_POPULATE 0x40 /* fault in page */
2593 : : #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2594 : : #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2595 : : #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2596 : : #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2597 : : #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2598 : : #define FOLL_MLOCK 0x1000 /* lock present pages */
2599 : : #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2600 : : #define FOLL_COW 0x4000 /* internal GUP flag */
2601 : : #define FOLL_ANON 0x8000 /* don't do file mappings */
2602 : : #define FOLL_LONGTERM 0x10000 /* mapping lifetime is indefinite: see below */
2603 : : #define FOLL_SPLIT_PMD 0x20000 /* split huge pmd before returning */
2604 : :
2605 : : /*
2606 : : * NOTE on FOLL_LONGTERM:
2607 : : *
2608 : : * FOLL_LONGTERM indicates that the page will be held for an indefinite time
2609 : : * period _often_ under userspace control. This is contrasted with
2610 : : * iov_iter_get_pages() where usages which are transient.
2611 : : *
2612 : : * FIXME: For pages which are part of a filesystem, mappings are subject to the
2613 : : * lifetime enforced by the filesystem and we need guarantees that longterm
2614 : : * users like RDMA and V4L2 only establish mappings which coordinate usage with
2615 : : * the filesystem. Ideas for this coordination include revoking the longterm
2616 : : * pin, delaying writeback, bounce buffer page writeback, etc. As FS DAX was
2617 : : * added after the problem with filesystems was found FS DAX VMAs are
2618 : : * specifically failed. Filesystem pages are still subject to bugs and use of
2619 : : * FOLL_LONGTERM should be avoided on those pages.
2620 : : *
2621 : : * FIXME: Also NOTE that FOLL_LONGTERM is not supported in every GUP call.
2622 : : * Currently only get_user_pages() and get_user_pages_fast() support this flag
2623 : : * and calls to get_user_pages_[un]locked are specifically not allowed. This
2624 : : * is due to an incompatibility with the FS DAX check and
2625 : : * FAULT_FLAG_ALLOW_RETRY
2626 : : *
2627 : : * In the CMA case: longterm pins in a CMA region would unnecessarily fragment
2628 : : * that region. And so CMA attempts to migrate the page before pinning when
2629 : : * FOLL_LONGTERM is specified.
2630 : : */
2631 : :
2632 : : static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
2633 : : {
2634 : 0 : if (vm_fault & VM_FAULT_OOM)
2635 : : return -ENOMEM;
2636 : 0 : if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2637 : 0 : return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2638 : 0 : if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2639 : : return -EFAULT;
2640 : : return 0;
2641 : : }
2642 : :
2643 : : typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
2644 : : extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2645 : : unsigned long size, pte_fn_t fn, void *data);
2646 : :
2647 : :
2648 : : #ifdef CONFIG_PAGE_POISONING
2649 : : extern bool page_poisoning_enabled(void);
2650 : : extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2651 : : #else
2652 : : static inline bool page_poisoning_enabled(void) { return false; }
2653 : : static inline void kernel_poison_pages(struct page *page, int numpages,
2654 : : int enable) { }
2655 : : #endif
2656 : :
2657 : : #ifdef CONFIG_INIT_ON_ALLOC_DEFAULT_ON
2658 : : DECLARE_STATIC_KEY_TRUE(init_on_alloc);
2659 : : #else
2660 : : DECLARE_STATIC_KEY_FALSE(init_on_alloc);
2661 : : #endif
2662 : : static inline bool want_init_on_alloc(gfp_t flags)
2663 : : {
2664 : 3 : if (static_branch_unlikely(&init_on_alloc) &&
2665 : : !page_poisoning_enabled())
2666 : : return true;
2667 : 3 : return flags & __GFP_ZERO;
2668 : : }
2669 : :
2670 : : #ifdef CONFIG_INIT_ON_FREE_DEFAULT_ON
2671 : : DECLARE_STATIC_KEY_TRUE(init_on_free);
2672 : : #else
2673 : : DECLARE_STATIC_KEY_FALSE(init_on_free);
2674 : : #endif
2675 : 3 : static inline bool want_init_on_free(void)
2676 : : {
2677 : 3 : return static_branch_unlikely(&init_on_free) &&
2678 : : !page_poisoning_enabled();
2679 : : }
2680 : :
2681 : : #ifdef CONFIG_DEBUG_PAGEALLOC
2682 : : extern void init_debug_pagealloc(void);
2683 : : #else
2684 : : static inline void init_debug_pagealloc(void) {}
2685 : : #endif
2686 : : extern bool _debug_pagealloc_enabled_early;
2687 : : DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
2688 : :
2689 : : static inline bool debug_pagealloc_enabled(void)
2690 : : {
2691 : : return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
2692 : : _debug_pagealloc_enabled_early;
2693 : : }
2694 : :
2695 : : /*
2696 : : * For use in fast paths after init_debug_pagealloc() has run, or when a
2697 : : * false negative result is not harmful when called too early.
2698 : : */
2699 : : static inline bool debug_pagealloc_enabled_static(void)
2700 : : {
2701 : : if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
2702 : : return false;
2703 : :
2704 : : return static_branch_unlikely(&_debug_pagealloc_enabled);
2705 : : }
2706 : :
2707 : : #if defined(CONFIG_DEBUG_PAGEALLOC) || defined(CONFIG_ARCH_HAS_SET_DIRECT_MAP)
2708 : : extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2709 : :
2710 : : /*
2711 : : * When called in DEBUG_PAGEALLOC context, the call should most likely be
2712 : : * guarded by debug_pagealloc_enabled() or debug_pagealloc_enabled_static()
2713 : : */
2714 : : static inline void
2715 : : kernel_map_pages(struct page *page, int numpages, int enable)
2716 : : {
2717 : : __kernel_map_pages(page, numpages, enable);
2718 : : }
2719 : : #ifdef CONFIG_HIBERNATION
2720 : : extern bool kernel_page_present(struct page *page);
2721 : : #endif /* CONFIG_HIBERNATION */
2722 : : #else /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */
2723 : : static inline void
2724 : : kernel_map_pages(struct page *page, int numpages, int enable) {}
2725 : : #ifdef CONFIG_HIBERNATION
2726 : : static inline bool kernel_page_present(struct page *page) { return true; }
2727 : : #endif /* CONFIG_HIBERNATION */
2728 : : #endif /* CONFIG_DEBUG_PAGEALLOC || CONFIG_ARCH_HAS_SET_DIRECT_MAP */
2729 : :
2730 : : #ifdef __HAVE_ARCH_GATE_AREA
2731 : : extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2732 : : extern int in_gate_area_no_mm(unsigned long addr);
2733 : : extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2734 : : #else
2735 : : static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2736 : : {
2737 : : return NULL;
2738 : : }
2739 : : static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2740 : : static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2741 : : {
2742 : : return 0;
2743 : : }
2744 : : #endif /* __HAVE_ARCH_GATE_AREA */
2745 : :
2746 : : extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2747 : :
2748 : : #ifdef CONFIG_SYSCTL
2749 : : extern int sysctl_drop_caches;
2750 : : int drop_caches_sysctl_handler(struct ctl_table *, int,
2751 : : void __user *, size_t *, loff_t *);
2752 : : #endif
2753 : :
2754 : : void drop_slab(void);
2755 : : void drop_slab_node(int nid);
2756 : :
2757 : : #ifndef CONFIG_MMU
2758 : : #define randomize_va_space 0
2759 : : #else
2760 : : extern int randomize_va_space;
2761 : : #endif
2762 : :
2763 : : const char * arch_vma_name(struct vm_area_struct *vma);
2764 : : #ifdef CONFIG_MMU
2765 : : void print_vma_addr(char *prefix, unsigned long rip);
2766 : : #else
2767 : : static inline void print_vma_addr(char *prefix, unsigned long rip)
2768 : : {
2769 : : }
2770 : : #endif
2771 : :
2772 : : void *sparse_buffer_alloc(unsigned long size);
2773 : : struct page * __populate_section_memmap(unsigned long pfn,
2774 : : unsigned long nr_pages, int nid, struct vmem_altmap *altmap);
2775 : : pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2776 : : p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
2777 : : pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
2778 : : pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2779 : : pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2780 : : void *vmemmap_alloc_block(unsigned long size, int node);
2781 : : struct vmem_altmap;
2782 : : void *vmemmap_alloc_block_buf(unsigned long size, int node);
2783 : : void *altmap_alloc_block_buf(unsigned long size, struct vmem_altmap *altmap);
2784 : : void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2785 : : int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2786 : : int node);
2787 : : int vmemmap_populate(unsigned long start, unsigned long end, int node,
2788 : : struct vmem_altmap *altmap);
2789 : : void vmemmap_populate_print_last(void);
2790 : : #ifdef CONFIG_MEMORY_HOTPLUG
2791 : : void vmemmap_free(unsigned long start, unsigned long end,
2792 : : struct vmem_altmap *altmap);
2793 : : #endif
2794 : : void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2795 : : unsigned long nr_pages);
2796 : :
2797 : : enum mf_flags {
2798 : : MF_COUNT_INCREASED = 1 << 0,
2799 : : MF_ACTION_REQUIRED = 1 << 1,
2800 : : MF_MUST_KILL = 1 << 2,
2801 : : MF_SOFT_OFFLINE = 1 << 3,
2802 : : };
2803 : : extern int memory_failure(unsigned long pfn, int flags);
2804 : : extern void memory_failure_queue(unsigned long pfn, int flags);
2805 : : extern int unpoison_memory(unsigned long pfn);
2806 : : extern int get_hwpoison_page(struct page *page);
2807 : : #define put_hwpoison_page(page) put_page(page)
2808 : : extern int sysctl_memory_failure_early_kill;
2809 : : extern int sysctl_memory_failure_recovery;
2810 : : extern void shake_page(struct page *p, int access);
2811 : : extern atomic_long_t num_poisoned_pages __read_mostly;
2812 : : extern int soft_offline_page(struct page *page, int flags);
2813 : :
2814 : :
2815 : : /*
2816 : : * Error handlers for various types of pages.
2817 : : */
2818 : : enum mf_result {
2819 : : MF_IGNORED, /* Error: cannot be handled */
2820 : : MF_FAILED, /* Error: handling failed */
2821 : : MF_DELAYED, /* Will be handled later */
2822 : : MF_RECOVERED, /* Successfully recovered */
2823 : : };
2824 : :
2825 : : enum mf_action_page_type {
2826 : : MF_MSG_KERNEL,
2827 : : MF_MSG_KERNEL_HIGH_ORDER,
2828 : : MF_MSG_SLAB,
2829 : : MF_MSG_DIFFERENT_COMPOUND,
2830 : : MF_MSG_POISONED_HUGE,
2831 : : MF_MSG_HUGE,
2832 : : MF_MSG_FREE_HUGE,
2833 : : MF_MSG_NON_PMD_HUGE,
2834 : : MF_MSG_UNMAP_FAILED,
2835 : : MF_MSG_DIRTY_SWAPCACHE,
2836 : : MF_MSG_CLEAN_SWAPCACHE,
2837 : : MF_MSG_DIRTY_MLOCKED_LRU,
2838 : : MF_MSG_CLEAN_MLOCKED_LRU,
2839 : : MF_MSG_DIRTY_UNEVICTABLE_LRU,
2840 : : MF_MSG_CLEAN_UNEVICTABLE_LRU,
2841 : : MF_MSG_DIRTY_LRU,
2842 : : MF_MSG_CLEAN_LRU,
2843 : : MF_MSG_TRUNCATED_LRU,
2844 : : MF_MSG_BUDDY,
2845 : : MF_MSG_BUDDY_2ND,
2846 : : MF_MSG_DAX,
2847 : : MF_MSG_UNKNOWN,
2848 : : };
2849 : :
2850 : : #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2851 : : extern void clear_huge_page(struct page *page,
2852 : : unsigned long addr_hint,
2853 : : unsigned int pages_per_huge_page);
2854 : : extern void copy_user_huge_page(struct page *dst, struct page *src,
2855 : : unsigned long addr_hint,
2856 : : struct vm_area_struct *vma,
2857 : : unsigned int pages_per_huge_page);
2858 : : extern long copy_huge_page_from_user(struct page *dst_page,
2859 : : const void __user *usr_src,
2860 : : unsigned int pages_per_huge_page,
2861 : : bool allow_pagefault);
2862 : : #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2863 : :
2864 : : #ifdef CONFIG_DEBUG_PAGEALLOC
2865 : : extern unsigned int _debug_guardpage_minorder;
2866 : : DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
2867 : :
2868 : : static inline unsigned int debug_guardpage_minorder(void)
2869 : : {
2870 : : return _debug_guardpage_minorder;
2871 : : }
2872 : :
2873 : : static inline bool debug_guardpage_enabled(void)
2874 : : {
2875 : : return static_branch_unlikely(&_debug_guardpage_enabled);
2876 : : }
2877 : :
2878 : : static inline bool page_is_guard(struct page *page)
2879 : : {
2880 : : if (!debug_guardpage_enabled())
2881 : : return false;
2882 : :
2883 : : return PageGuard(page);
2884 : : }
2885 : : #else
2886 : : static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2887 : : static inline bool debug_guardpage_enabled(void) { return false; }
2888 : : static inline bool page_is_guard(struct page *page) { return false; }
2889 : : #endif /* CONFIG_DEBUG_PAGEALLOC */
2890 : :
2891 : : #if MAX_NUMNODES > 1
2892 : : void __init setup_nr_node_ids(void);
2893 : : #else
2894 : : static inline void setup_nr_node_ids(void) {}
2895 : : #endif
2896 : :
2897 : : extern int memcmp_pages(struct page *page1, struct page *page2);
2898 : :
2899 : : static inline int pages_identical(struct page *page1, struct page *page2)
2900 : : {
2901 : : return !memcmp_pages(page1, page2);
2902 : : }
2903 : :
2904 : : #endif /* __KERNEL__ */
2905 : : #endif /* _LINUX_MM_H */
|
