Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
1/* SPDX-License-Identifier: GPL-2.0 */
2#ifndef _LINUX_MM_H
3#define _LINUX_MM_H
4
5#include <linux/args.h>
6#include <linux/errno.h>
7#include <linux/mmdebug.h>
8#include <linux/gfp.h>
9#include <linux/pgalloc_tag.h>
10#include <linux/bug.h>
11#include <linux/list.h>
12#include <linux/mmzone.h>
13#include <linux/rbtree.h>
14#include <linux/atomic.h>
15#include <linux/debug_locks.h>
16#include <linux/compiler.h>
17#include <linux/mm_types.h>
18#include <linux/mmap_lock.h>
19#include <linux/range.h>
20#include <linux/pfn.h>
21#include <linux/percpu-refcount.h>
22#include <linux/bit_spinlock.h>
23#include <linux/shrinker.h>
24#include <linux/resource.h>
25#include <linux/page_ext.h>
26#include <linux/err.h>
27#include <linux/page-flags.h>
28#include <linux/page_ref.h>
29#include <linux/overflow.h>
30#include <linux/sched.h>
31#include <linux/pgtable.h>
32#include <linux/kasan.h>
33#include <linux/memremap.h>
34#include <linux/slab.h>
35#include <linux/cacheinfo.h>
36#include <linux/rcuwait.h>
37#include <linux/bitmap.h>
38#include <linux/bitops.h>
39#include <linux/iommu-debug-pagealloc.h>
40
41struct mempolicy;
42struct anon_vma;
43struct anon_vma_chain;
44struct user_struct;
45struct pt_regs;
46struct folio_batch;
47
48void arch_mm_preinit(void);
49void mm_core_init_early(void);
50void mm_core_init(void);
51void init_mm_internals(void);
52
53extern atomic_long_t _totalram_pages;
54static inline unsigned long totalram_pages(void)
55{
56 return (unsigned long)atomic_long_read(&_totalram_pages);
57}
58
59static inline void totalram_pages_inc(void)
60{
61 atomic_long_inc(&_totalram_pages);
62}
63
64static inline void totalram_pages_dec(void)
65{
66 atomic_long_dec(&_totalram_pages);
67}
68
69static inline void totalram_pages_add(long count)
70{
71 atomic_long_add(count, &_totalram_pages);
72}
73
74extern void * high_memory;
75
76/*
77 * Convert between pages and MB
78 * 20 is the shift for 1MB (2^20 = 1MB)
79 * PAGE_SHIFT is the shift for page size (e.g., 12 for 4KB pages)
80 * So (20 - PAGE_SHIFT) converts between pages and MB
81 */
82#define PAGES_TO_MB(pages) ((pages) >> (20 - PAGE_SHIFT))
83#define MB_TO_PAGES(mb) ((mb) << (20 - PAGE_SHIFT))
84
85#ifdef CONFIG_SYSCTL
86extern int sysctl_legacy_va_layout;
87#else
88#define sysctl_legacy_va_layout 0
89#endif
90
91#ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
92extern const int mmap_rnd_bits_min;
93extern int mmap_rnd_bits_max __ro_after_init;
94extern int mmap_rnd_bits __read_mostly;
95#endif
96#ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
97extern const int mmap_rnd_compat_bits_min;
98extern const int mmap_rnd_compat_bits_max;
99extern int mmap_rnd_compat_bits __read_mostly;
100#endif
101
102#ifndef DIRECT_MAP_PHYSMEM_END
103# ifdef MAX_PHYSMEM_BITS
104# define DIRECT_MAP_PHYSMEM_END ((1ULL << MAX_PHYSMEM_BITS) - 1)
105# else
106# define DIRECT_MAP_PHYSMEM_END (((phys_addr_t)-1)&~(1ULL<<63))
107# endif
108#endif
109
110#define INVALID_PHYS_ADDR (~(phys_addr_t)0)
111
112#include <asm/page.h>
113#include <asm/processor.h>
114
115#ifndef __pa_symbol
116#define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
117#endif
118
119#ifndef page_to_virt
120#define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
121#endif
122
123#ifndef lm_alias
124#define lm_alias(x) __va(__pa_symbol(x))
125#endif
126
127/*
128 * To prevent common memory management code establishing
129 * a zero page mapping on a read fault.
130 * This macro should be defined within <asm/pgtable.h>.
131 * s390 does this to prevent multiplexing of hardware bits
132 * related to the physical page in case of virtualization.
133 */
134#ifndef mm_forbids_zeropage
135#define mm_forbids_zeropage(X) (0)
136#endif
137
138/*
139 * On some architectures it is expensive to call memset() for small sizes.
140 * If an architecture decides to implement their own version of
141 * mm_zero_struct_page they should wrap the defines below in a #ifndef and
142 * define their own version of this macro in <asm/pgtable.h>
143 */
144#if BITS_PER_LONG == 64
145/* This function must be updated when the size of struct page grows above 96
146 * or reduces below 56. The idea that compiler optimizes out switch()
147 * statement, and only leaves move/store instructions. Also the compiler can
148 * combine write statements if they are both assignments and can be reordered,
149 * this can result in several of the writes here being dropped.
150 */
151#define mm_zero_struct_page(pp) __mm_zero_struct_page(pp)
152static inline void __mm_zero_struct_page(struct page *page)
153{
154 unsigned long *_pp = (void *)page;
155
156 /* Check that struct page is either 56, 64, 72, 80, 88 or 96 bytes */
157 BUILD_BUG_ON(sizeof(struct page) & 7);
158 BUILD_BUG_ON(sizeof(struct page) < 56);
159 BUILD_BUG_ON(sizeof(struct page) > 96);
160
161 switch (sizeof(struct page)) {
162 case 96:
163 _pp[11] = 0;
164 fallthrough;
165 case 88:
166 _pp[10] = 0;
167 fallthrough;
168 case 80:
169 _pp[9] = 0;
170 fallthrough;
171 case 72:
172 _pp[8] = 0;
173 fallthrough;
174 case 64:
175 _pp[7] = 0;
176 fallthrough;
177 case 56:
178 _pp[6] = 0;
179 _pp[5] = 0;
180 _pp[4] = 0;
181 _pp[3] = 0;
182 _pp[2] = 0;
183 _pp[1] = 0;
184 _pp[0] = 0;
185 }
186}
187#else
188#define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
189#endif
190
191/*
192 * Default maximum number of active map areas, this limits the number of vmas
193 * per mm struct. Users can overwrite this number by sysctl but there is a
194 * problem.
195 *
196 * When a program's coredump is generated as ELF format, a section is created
197 * per a vma. In ELF, the number of sections is represented in unsigned short.
198 * This means the number of sections should be smaller than 65535 at coredump.
199 * Because the kernel adds some informative sections to a image of program at
200 * generating coredump, we need some margin. The number of extra sections is
201 * 1-3 now and depends on arch. We use "5" as safe margin, here.
202 *
203 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
204 * not a hard limit any more. Although some userspace tools can be surprised by
205 * that.
206 */
207#define MAPCOUNT_ELF_CORE_MARGIN (5)
208#define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
209
210extern unsigned long sysctl_user_reserve_kbytes;
211extern unsigned long sysctl_admin_reserve_kbytes;
212
213#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
214bool page_range_contiguous(const struct page *page, unsigned long nr_pages);
215#else
216static inline bool page_range_contiguous(const struct page *page,
217 unsigned long nr_pages)
218{
219 return true;
220}
221#endif
222
223/* to align the pointer to the (next) page boundary */
224#define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
225
226/* to align the pointer to the (prev) page boundary */
227#define PAGE_ALIGN_DOWN(addr) ALIGN_DOWN(addr, PAGE_SIZE)
228
229/* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
230#define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
231
232/**
233 * folio_page_idx - Return the number of a page in a folio.
234 * @folio: The folio.
235 * @page: The folio page.
236 *
237 * This function expects that the page is actually part of the folio.
238 * The returned number is relative to the start of the folio.
239 */
240static inline unsigned long folio_page_idx(const struct folio *folio,
241 const struct page *page)
242{
243 return page - &folio->page;
244}
245
246static inline struct folio *lru_to_folio(struct list_head *head)
247{
248 return list_entry((head)->prev, struct folio, lru);
249}
250
251void setup_initial_init_mm(void *start_code, void *end_code,
252 void *end_data, void *brk);
253
254/*
255 * Linux kernel virtual memory manager primitives.
256 * The idea being to have a "virtual" mm in the same way
257 * we have a virtual fs - giving a cleaner interface to the
258 * mm details, and allowing different kinds of memory mappings
259 * (from shared memory to executable loading to arbitrary
260 * mmap() functions).
261 */
262
263struct vm_area_struct *vm_area_alloc(struct mm_struct *);
264struct vm_area_struct *vm_area_dup(struct vm_area_struct *);
265void vm_area_free(struct vm_area_struct *);
266
267#ifndef CONFIG_MMU
268extern struct rb_root nommu_region_tree;
269extern struct rw_semaphore nommu_region_sem;
270
271extern unsigned int kobjsize(const void *objp);
272#endif
273
274/*
275 * vm_flags in vm_area_struct, see mm_types.h.
276 * When changing, update also include/trace/events/mmflags.h
277 */
278
279#define VM_NONE 0x00000000
280
281/**
282 * typedef vma_flag_t - specifies an individual VMA flag by bit number.
283 *
284 * This value is made type safe by sparse to avoid passing invalid flag values
285 * around.
286 */
287typedef int __bitwise vma_flag_t;
288
289#define DECLARE_VMA_BIT(name, bitnum) \
290 VMA_ ## name ## _BIT = ((__force vma_flag_t)bitnum)
291#define DECLARE_VMA_BIT_ALIAS(name, aliased) \
292 VMA_ ## name ## _BIT = (VMA_ ## aliased ## _BIT)
293enum {
294 DECLARE_VMA_BIT(READ, 0),
295 DECLARE_VMA_BIT(WRITE, 1),
296 DECLARE_VMA_BIT(EXEC, 2),
297 DECLARE_VMA_BIT(SHARED, 3),
298 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
299 DECLARE_VMA_BIT(MAYREAD, 4), /* limits for mprotect() etc. */
300 DECLARE_VMA_BIT(MAYWRITE, 5),
301 DECLARE_VMA_BIT(MAYEXEC, 6),
302 DECLARE_VMA_BIT(MAYSHARE, 7),
303 DECLARE_VMA_BIT(GROWSDOWN, 8), /* general info on the segment */
304#ifdef CONFIG_MMU
305 DECLARE_VMA_BIT(UFFD_MISSING, 9),/* missing pages tracking */
306#else
307 /* nommu: R/O MAP_PRIVATE mapping that might overlay a file mapping */
308 DECLARE_VMA_BIT(MAYOVERLAY, 9),
309#endif /* CONFIG_MMU */
310 /* Page-ranges managed without "struct page", just pure PFN */
311 DECLARE_VMA_BIT(PFNMAP, 10),
312 DECLARE_VMA_BIT(MAYBE_GUARD, 11),
313 DECLARE_VMA_BIT(UFFD_WP, 12), /* wrprotect pages tracking */
314 DECLARE_VMA_BIT(LOCKED, 13),
315 DECLARE_VMA_BIT(IO, 14), /* Memory mapped I/O or similar */
316 DECLARE_VMA_BIT(SEQ_READ, 15), /* App will access data sequentially */
317 DECLARE_VMA_BIT(RAND_READ, 16), /* App will not benefit from clustered reads */
318 DECLARE_VMA_BIT(DONTCOPY, 17), /* Do not copy this vma on fork */
319 DECLARE_VMA_BIT(DONTEXPAND, 18),/* Cannot expand with mremap() */
320 DECLARE_VMA_BIT(LOCKONFAULT, 19),/* Lock pages covered when faulted in */
321 DECLARE_VMA_BIT(ACCOUNT, 20), /* Is a VM accounted object */
322 DECLARE_VMA_BIT(NORESERVE, 21), /* should the VM suppress accounting */
323 DECLARE_VMA_BIT(HUGETLB, 22), /* Huge TLB Page VM */
324 DECLARE_VMA_BIT(SYNC, 23), /* Synchronous page faults */
325 DECLARE_VMA_BIT(ARCH_1, 24), /* Architecture-specific flag */
326 DECLARE_VMA_BIT(WIPEONFORK, 25),/* Wipe VMA contents in child. */
327 DECLARE_VMA_BIT(DONTDUMP, 26), /* Do not include in the core dump */
328 DECLARE_VMA_BIT(SOFTDIRTY, 27), /* NOT soft dirty clean area */
329 DECLARE_VMA_BIT(MIXEDMAP, 28), /* Can contain struct page and pure PFN pages */
330 DECLARE_VMA_BIT(HUGEPAGE, 29), /* MADV_HUGEPAGE marked this vma */
331 DECLARE_VMA_BIT(NOHUGEPAGE, 30),/* MADV_NOHUGEPAGE marked this vma */
332 DECLARE_VMA_BIT(MERGEABLE, 31), /* KSM may merge identical pages */
333 /* These bits are reused, we define specific uses below. */
334 DECLARE_VMA_BIT(HIGH_ARCH_0, 32),
335 DECLARE_VMA_BIT(HIGH_ARCH_1, 33),
336 DECLARE_VMA_BIT(HIGH_ARCH_2, 34),
337 DECLARE_VMA_BIT(HIGH_ARCH_3, 35),
338 DECLARE_VMA_BIT(HIGH_ARCH_4, 36),
339 DECLARE_VMA_BIT(HIGH_ARCH_5, 37),
340 DECLARE_VMA_BIT(HIGH_ARCH_6, 38),
341 /*
342 * This flag is used to connect VFIO to arch specific KVM code. It
343 * indicates that the memory under this VMA is safe for use with any
344 * non-cachable memory type inside KVM. Some VFIO devices, on some
345 * platforms, are thought to be unsafe and can cause machine crashes
346 * if KVM does not lock down the memory type.
347 */
348 DECLARE_VMA_BIT(ALLOW_ANY_UNCACHED, 39),
349#if defined(CONFIG_PPC32)
350 DECLARE_VMA_BIT_ALIAS(DROPPABLE, ARCH_1),
351#elif defined(CONFIG_64BIT)
352 DECLARE_VMA_BIT(DROPPABLE, 40),
353#endif
354 DECLARE_VMA_BIT(UFFD_MINOR, 41),
355 DECLARE_VMA_BIT(SEALED, 42),
356 /* Flags that reuse flags above. */
357 DECLARE_VMA_BIT_ALIAS(PKEY_BIT0, HIGH_ARCH_0),
358 DECLARE_VMA_BIT_ALIAS(PKEY_BIT1, HIGH_ARCH_1),
359 DECLARE_VMA_BIT_ALIAS(PKEY_BIT2, HIGH_ARCH_2),
360 DECLARE_VMA_BIT_ALIAS(PKEY_BIT3, HIGH_ARCH_3),
361 DECLARE_VMA_BIT_ALIAS(PKEY_BIT4, HIGH_ARCH_4),
362#if defined(CONFIG_X86_USER_SHADOW_STACK) || defined(CONFIG_RISCV_USER_CFI)
363 /*
364 * VM_SHADOW_STACK should not be set with VM_SHARED because of lack of
365 * support core mm.
366 *
367 * These VMAs will get a single end guard page. This helps userspace
368 * protect itself from attacks. A single page is enough for current
369 * shadow stack archs (x86). See the comments near alloc_shstk() in
370 * arch/x86/kernel/shstk.c for more details on the guard size.
371 */
372 DECLARE_VMA_BIT_ALIAS(SHADOW_STACK, HIGH_ARCH_5),
373#elif defined(CONFIG_ARM64_GCS)
374 /*
375 * arm64's Guarded Control Stack implements similar functionality and
376 * has similar constraints to shadow stacks.
377 */
378 DECLARE_VMA_BIT_ALIAS(SHADOW_STACK, HIGH_ARCH_6),
379#endif
380 DECLARE_VMA_BIT_ALIAS(SAO, ARCH_1), /* Strong Access Ordering (powerpc) */
381 DECLARE_VMA_BIT_ALIAS(GROWSUP, ARCH_1), /* parisc */
382 DECLARE_VMA_BIT_ALIAS(SPARC_ADI, ARCH_1), /* sparc64 */
383 DECLARE_VMA_BIT_ALIAS(ARM64_BTI, ARCH_1), /* arm64 */
384 DECLARE_VMA_BIT_ALIAS(ARCH_CLEAR, ARCH_1), /* sparc64, arm64 */
385 DECLARE_VMA_BIT_ALIAS(MAPPED_COPY, ARCH_1), /* !CONFIG_MMU */
386 DECLARE_VMA_BIT_ALIAS(MTE, HIGH_ARCH_4), /* arm64 */
387 DECLARE_VMA_BIT_ALIAS(MTE_ALLOWED, HIGH_ARCH_5),/* arm64 */
388#ifdef CONFIG_STACK_GROWSUP
389 DECLARE_VMA_BIT_ALIAS(STACK, GROWSUP),
390 DECLARE_VMA_BIT_ALIAS(STACK_EARLY, GROWSDOWN),
391#else
392 DECLARE_VMA_BIT_ALIAS(STACK, GROWSDOWN),
393#endif
394};
395#undef DECLARE_VMA_BIT
396#undef DECLARE_VMA_BIT_ALIAS
397
398#define INIT_VM_FLAG(name) BIT((__force int) VMA_ ## name ## _BIT)
399#define VM_READ INIT_VM_FLAG(READ)
400#define VM_WRITE INIT_VM_FLAG(WRITE)
401#define VM_EXEC INIT_VM_FLAG(EXEC)
402#define VM_SHARED INIT_VM_FLAG(SHARED)
403#define VM_MAYREAD INIT_VM_FLAG(MAYREAD)
404#define VM_MAYWRITE INIT_VM_FLAG(MAYWRITE)
405#define VM_MAYEXEC INIT_VM_FLAG(MAYEXEC)
406#define VM_MAYSHARE INIT_VM_FLAG(MAYSHARE)
407#define VM_GROWSDOWN INIT_VM_FLAG(GROWSDOWN)
408#ifdef CONFIG_MMU
409#define VM_UFFD_MISSING INIT_VM_FLAG(UFFD_MISSING)
410#else
411#define VM_UFFD_MISSING VM_NONE
412#define VM_MAYOVERLAY INIT_VM_FLAG(MAYOVERLAY)
413#endif
414#define VM_PFNMAP INIT_VM_FLAG(PFNMAP)
415#define VM_MAYBE_GUARD INIT_VM_FLAG(MAYBE_GUARD)
416#define VM_UFFD_WP INIT_VM_FLAG(UFFD_WP)
417#define VM_LOCKED INIT_VM_FLAG(LOCKED)
418#define VM_IO INIT_VM_FLAG(IO)
419#define VM_SEQ_READ INIT_VM_FLAG(SEQ_READ)
420#define VM_RAND_READ INIT_VM_FLAG(RAND_READ)
421#define VM_DONTCOPY INIT_VM_FLAG(DONTCOPY)
422#define VM_DONTEXPAND INIT_VM_FLAG(DONTEXPAND)
423#define VM_LOCKONFAULT INIT_VM_FLAG(LOCKONFAULT)
424#define VM_ACCOUNT INIT_VM_FLAG(ACCOUNT)
425#define VM_NORESERVE INIT_VM_FLAG(NORESERVE)
426#define VM_HUGETLB INIT_VM_FLAG(HUGETLB)
427#define VM_SYNC INIT_VM_FLAG(SYNC)
428#define VM_ARCH_1 INIT_VM_FLAG(ARCH_1)
429#define VM_WIPEONFORK INIT_VM_FLAG(WIPEONFORK)
430#define VM_DONTDUMP INIT_VM_FLAG(DONTDUMP)
431#ifdef CONFIG_MEM_SOFT_DIRTY
432#define VM_SOFTDIRTY INIT_VM_FLAG(SOFTDIRTY)
433#else
434#define VM_SOFTDIRTY VM_NONE
435#endif
436#define VM_MIXEDMAP INIT_VM_FLAG(MIXEDMAP)
437#define VM_HUGEPAGE INIT_VM_FLAG(HUGEPAGE)
438#define VM_NOHUGEPAGE INIT_VM_FLAG(NOHUGEPAGE)
439#define VM_MERGEABLE INIT_VM_FLAG(MERGEABLE)
440#define VM_STACK INIT_VM_FLAG(STACK)
441#ifdef CONFIG_STACK_GROWSUP
442#define VM_STACK_EARLY INIT_VM_FLAG(STACK_EARLY)
443#else
444#define VM_STACK_EARLY VM_NONE
445#endif
446#ifdef CONFIG_ARCH_HAS_PKEYS
447#define VM_PKEY_SHIFT ((__force int)VMA_HIGH_ARCH_0_BIT)
448/* Despite the naming, these are FLAGS not bits. */
449#define VM_PKEY_BIT0 INIT_VM_FLAG(PKEY_BIT0)
450#define VM_PKEY_BIT1 INIT_VM_FLAG(PKEY_BIT1)
451#define VM_PKEY_BIT2 INIT_VM_FLAG(PKEY_BIT2)
452#if CONFIG_ARCH_PKEY_BITS > 3
453#define VM_PKEY_BIT3 INIT_VM_FLAG(PKEY_BIT3)
454#else
455#define VM_PKEY_BIT3 VM_NONE
456#endif /* CONFIG_ARCH_PKEY_BITS > 3 */
457#if CONFIG_ARCH_PKEY_BITS > 4
458#define VM_PKEY_BIT4 INIT_VM_FLAG(PKEY_BIT4)
459#else
460#define VM_PKEY_BIT4 VM_NONE
461#endif /* CONFIG_ARCH_PKEY_BITS > 4 */
462#endif /* CONFIG_ARCH_HAS_PKEYS */
463#if defined(CONFIG_X86_USER_SHADOW_STACK) || defined(CONFIG_ARM64_GCS) || \
464 defined(CONFIG_RISCV_USER_CFI)
465#define VM_SHADOW_STACK INIT_VM_FLAG(SHADOW_STACK)
466#define VMA_STARTGAP_FLAGS mk_vma_flags(VMA_GROWSDOWN_BIT, VMA_SHADOW_STACK_BIT)
467#else
468#define VM_SHADOW_STACK VM_NONE
469#define VMA_STARTGAP_FLAGS mk_vma_flags(VMA_GROWSDOWN_BIT)
470#endif
471#if defined(CONFIG_PPC64)
472#define VM_SAO INIT_VM_FLAG(SAO)
473#elif defined(CONFIG_PARISC)
474#define VM_GROWSUP INIT_VM_FLAG(GROWSUP)
475#elif defined(CONFIG_SPARC64)
476#define VM_SPARC_ADI INIT_VM_FLAG(SPARC_ADI)
477#define VM_ARCH_CLEAR INIT_VM_FLAG(ARCH_CLEAR)
478#elif defined(CONFIG_ARM64)
479#define VM_ARM64_BTI INIT_VM_FLAG(ARM64_BTI)
480#define VM_ARCH_CLEAR INIT_VM_FLAG(ARCH_CLEAR)
481#elif !defined(CONFIG_MMU)
482#define VM_MAPPED_COPY INIT_VM_FLAG(MAPPED_COPY)
483#endif
484#ifndef VM_GROWSUP
485#define VM_GROWSUP VM_NONE
486#endif
487#ifdef CONFIG_ARM64_MTE
488#define VM_MTE INIT_VM_FLAG(MTE)
489#define VM_MTE_ALLOWED INIT_VM_FLAG(MTE_ALLOWED)
490#else
491#define VM_MTE VM_NONE
492#define VM_MTE_ALLOWED VM_NONE
493#endif
494#ifdef CONFIG_HAVE_ARCH_USERFAULTFD_MINOR
495#define VM_UFFD_MINOR INIT_VM_FLAG(UFFD_MINOR)
496#else
497#define VM_UFFD_MINOR VM_NONE
498#endif
499#ifdef CONFIG_64BIT
500#define VM_ALLOW_ANY_UNCACHED INIT_VM_FLAG(ALLOW_ANY_UNCACHED)
501#define VM_SEALED INIT_VM_FLAG(SEALED)
502#else
503#define VM_ALLOW_ANY_UNCACHED VM_NONE
504#define VM_SEALED VM_NONE
505#endif
506#if defined(CONFIG_64BIT) || defined(CONFIG_PPC32)
507#define VM_DROPPABLE INIT_VM_FLAG(DROPPABLE)
508#define VMA_DROPPABLE mk_vma_flags(VMA_DROPPABLE_BIT)
509#else
510#define VM_DROPPABLE VM_NONE
511#define VMA_DROPPABLE EMPTY_VMA_FLAGS
512#endif
513
514/* Bits set in the VMA until the stack is in its final location */
515#define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ | VM_STACK_EARLY)
516
517#define TASK_EXEC_BIT ((current->personality & READ_IMPLIES_EXEC) ? \
518 VMA_EXEC_BIT : VMA_READ_BIT)
519
520/* Common data flag combinations */
521#define VMA_DATA_FLAGS_TSK_EXEC mk_vma_flags(VMA_READ_BIT, VMA_WRITE_BIT, \
522 TASK_EXEC_BIT, VMA_MAYREAD_BIT, VMA_MAYWRITE_BIT, \
523 VMA_MAYEXEC_BIT)
524#define VMA_DATA_FLAGS_NON_EXEC mk_vma_flags(VMA_READ_BIT, VMA_WRITE_BIT, \
525 VMA_MAYREAD_BIT, VMA_MAYWRITE_BIT, VMA_MAYEXEC_BIT)
526#define VMA_DATA_FLAGS_EXEC mk_vma_flags(VMA_READ_BIT, VMA_WRITE_BIT, \
527 VMA_EXEC_BIT, VMA_MAYREAD_BIT, VMA_MAYWRITE_BIT, \
528 VMA_MAYEXEC_BIT)
529
530#ifndef VMA_DATA_DEFAULT_FLAGS /* arch can override this */
531#define VMA_DATA_DEFAULT_FLAGS VMA_DATA_FLAGS_EXEC
532#endif
533
534#ifndef VMA_STACK_DEFAULT_FLAGS /* arch can override this */
535#define VMA_STACK_DEFAULT_FLAGS VMA_DATA_DEFAULT_FLAGS
536#endif
537
538#define VMA_STACK_FLAGS append_vma_flags(VMA_STACK_DEFAULT_FLAGS, \
539 VMA_STACK_BIT, VMA_ACCOUNT_BIT)
540
541/* Temporary until VMA flags conversion complete. */
542#define VM_STACK_FLAGS vma_flags_to_legacy(VMA_STACK_FLAGS)
543
544#ifdef CONFIG_MSEAL_SYSTEM_MAPPINGS
545#define VM_SEALED_SYSMAP VM_SEALED
546#else
547#define VM_SEALED_SYSMAP VM_NONE
548#endif
549
550/* VMA basic access permission flags */
551#define VM_ACCESS_FLAGS (VM_READ | VM_WRITE | VM_EXEC)
552#define VMA_ACCESS_FLAGS mk_vma_flags(VMA_READ_BIT, VMA_WRITE_BIT, VMA_EXEC_BIT)
553
554/*
555 * Special vmas that are non-mergable, non-mlock()able.
556 */
557
558#define VMA_SPECIAL_FLAGS mk_vma_flags(VMA_IO_BIT, VMA_DONTEXPAND_BIT, \
559 VMA_PFNMAP_BIT, VMA_MIXEDMAP_BIT)
560#define VM_SPECIAL vma_flags_to_legacy(VMA_SPECIAL_FLAGS)
561
562/*
563 * Physically remapped pages are special. Tell the
564 * rest of the world about it:
565 * IO tells people not to look at these pages
566 * (accesses can have side effects).
567 * PFNMAP tells the core MM that the base pages are just
568 * raw PFN mappings, and do not have a "struct page" associated
569 * with them.
570 * DONTEXPAND
571 * Disable vma merging and expanding with mremap().
572 * DONTDUMP
573 * Omit vma from core dump, even when VM_IO turned off.
574 */
575#define VMA_REMAP_FLAGS mk_vma_flags(VMA_IO_BIT, VMA_PFNMAP_BIT, \
576 VMA_DONTEXPAND_BIT, VMA_DONTDUMP_BIT)
577
578/* This mask prevents VMA from being scanned with khugepaged */
579#define VM_NO_KHUGEPAGED (VM_SPECIAL | VM_HUGETLB)
580
581/* This mask defines which mm->def_flags a process can inherit its parent */
582#define VM_INIT_DEF_MASK VM_NOHUGEPAGE
583
584/* This mask represents all the VMA flag bits used by mlock */
585#define VM_LOCKED_MASK (VM_LOCKED | VM_LOCKONFAULT)
586
587#define VMA_LOCKED_MASK mk_vma_flags(VMA_LOCKED_BIT, VMA_LOCKONFAULT_BIT)
588
589/* These flags can be updated atomically via VMA/mmap read lock. */
590#define VM_ATOMIC_SET_ALLOWED VM_MAYBE_GUARD
591
592/* Arch-specific flags to clear when updating VM flags on protection change */
593#ifndef VM_ARCH_CLEAR
594#define VM_ARCH_CLEAR VM_NONE
595#endif
596#define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
597
598/*
599 * Flags which should be 'sticky' on merge - that is, flags which, when one VMA
600 * possesses it but the other does not, the merged VMA should nonetheless have
601 * applied to it:
602 *
603 * VMA_SOFTDIRTY_BIT - if a VMA is marked soft-dirty, that is has not had its
604 * references cleared via /proc/$pid/clear_refs, any
605 * merged VMA should be considered soft-dirty also as it
606 * operates at a VMA granularity.
607 *
608 * VMA_MAYBE_GUARD_BIT - If a VMA may have guard regions in place it implies
609 * that mapped page tables may contain metadata not
610 * described by the VMA and thus any merged VMA may also
611 * contain this metadata, and thus we must make this flag
612 * sticky.
613 */
614#ifdef CONFIG_MEM_SOFT_DIRTY
615#define VMA_STICKY_FLAGS mk_vma_flags(VMA_SOFTDIRTY_BIT, VMA_MAYBE_GUARD_BIT)
616#else
617#define VMA_STICKY_FLAGS mk_vma_flags(VMA_MAYBE_GUARD_BIT)
618#endif
619
620/*
621 * VMA flags we ignore for the purposes of merge, i.e. one VMA possessing one
622 * of these flags and the other not does not preclude a merge.
623 *
624 * VMA_STICKY_FLAGS - When merging VMAs, VMA flags must match, unless they
625 * are 'sticky'. If any sticky flags exist in either VMA,
626 * we simply set all of them on the merged VMA.
627 */
628#define VMA_IGNORE_MERGE_FLAGS VMA_STICKY_FLAGS
629
630/*
631 * Flags which should result in page tables being copied on fork. These are
632 * flags which indicate that the VMA maps page tables which cannot be
633 * reconsistuted upon page fault, so necessitate page table copying upon fork.
634 *
635 * Note that these flags should be compared with the DESTINATION VMA not the
636 * source, as VM_UFFD_WP may not be propagated to destination, while all other
637 * flags will be.
638 *
639 * VM_PFNMAP / VM_MIXEDMAP - These contain kernel-mapped data which cannot be
640 * reasonably reconstructed on page fault.
641 *
642 * VM_UFFD_WP - Encodes metadata about an installed uffd
643 * write protect handler, which cannot be
644 * reconstructed on page fault.
645 *
646 * We always copy pgtables when dst_vma has uffd-wp
647 * enabled even if it's file-backed
648 * (e.g. shmem). Because when uffd-wp is enabled,
649 * pgtable contains uffd-wp protection information,
650 * that's something we can't retrieve from page cache,
651 * and skip copying will lose those info.
652 *
653 * VM_MAYBE_GUARD - Could contain page guard region markers which
654 * by design are a property of the page tables
655 * only and thus cannot be reconstructed on page
656 * fault.
657 */
658#define VM_COPY_ON_FORK (VM_PFNMAP | VM_MIXEDMAP | VM_UFFD_WP | VM_MAYBE_GUARD)
659
660/*
661 * mapping from the currently active vm_flags protection bits (the
662 * low four bits) to a page protection mask..
663 */
664
665/*
666 * The default fault flags that should be used by most of the
667 * arch-specific page fault handlers.
668 */
669#define FAULT_FLAG_DEFAULT (FAULT_FLAG_ALLOW_RETRY | \
670 FAULT_FLAG_KILLABLE | \
671 FAULT_FLAG_INTERRUPTIBLE)
672
673/**
674 * fault_flag_allow_retry_first - check ALLOW_RETRY the first time
675 * @flags: Fault flags.
676 *
677 * This is mostly used for places where we want to try to avoid taking
678 * the mmap_lock for too long a time when waiting for another condition
679 * to change, in which case we can try to be polite to release the
680 * mmap_lock in the first round to avoid potential starvation of other
681 * processes that would also want the mmap_lock.
682 *
683 * Return: true if the page fault allows retry and this is the first
684 * attempt of the fault handling; false otherwise.
685 */
686static inline bool fault_flag_allow_retry_first(enum fault_flag flags)
687{
688 return (flags & FAULT_FLAG_ALLOW_RETRY) &&
689 (!(flags & FAULT_FLAG_TRIED));
690}
691
692#define FAULT_FLAG_TRACE \
693 { FAULT_FLAG_WRITE, "WRITE" }, \
694 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
695 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
696 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
697 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
698 { FAULT_FLAG_TRIED, "TRIED" }, \
699 { FAULT_FLAG_USER, "USER" }, \
700 { FAULT_FLAG_REMOTE, "REMOTE" }, \
701 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }, \
702 { FAULT_FLAG_INTERRUPTIBLE, "INTERRUPTIBLE" }, \
703 { FAULT_FLAG_VMA_LOCK, "VMA_LOCK" }
704
705/*
706 * vm_fault is filled by the pagefault handler and passed to the vma's
707 * ->fault function. The vma's ->fault is responsible for returning a bitmask
708 * of VM_FAULT_xxx flags that give details about how the fault was handled.
709 *
710 * MM layer fills up gfp_mask for page allocations but fault handler might
711 * alter it if its implementation requires a different allocation context.
712 *
713 * pgoff should be used in favour of virtual_address, if possible.
714 */
715struct vm_fault {
716 const struct {
717 struct vm_area_struct *vma; /* Target VMA */
718 gfp_t gfp_mask; /* gfp mask to be used for allocations */
719 pgoff_t pgoff; /* Logical page offset based on vma */
720 unsigned long address; /* Faulting virtual address - masked */
721 unsigned long real_address; /* Faulting virtual address - unmasked */
722 };
723 enum fault_flag flags; /* FAULT_FLAG_xxx flags
724 * XXX: should really be 'const' */
725 pmd_t *pmd; /* Pointer to pmd entry matching
726 * the 'address' */
727 pud_t *pud; /* Pointer to pud entry matching
728 * the 'address'
729 */
730 union {
731 pte_t orig_pte; /* Value of PTE at the time of fault */
732 pmd_t orig_pmd; /* Value of PMD at the time of fault,
733 * used by PMD fault only.
734 */
735 };
736
737 struct page *cow_page; /* Page handler may use for COW fault */
738 struct page *page; /* ->fault handlers should return a
739 * page here, unless VM_FAULT_NOPAGE
740 * is set (which is also implied by
741 * VM_FAULT_ERROR).
742 */
743 /* These three entries are valid only while holding ptl lock */
744 pte_t *pte; /* Pointer to pte entry matching
745 * the 'address'. NULL if the page
746 * table hasn't been allocated.
747 */
748 spinlock_t *ptl; /* Page table lock.
749 * Protects pte page table if 'pte'
750 * is not NULL, otherwise pmd.
751 */
752 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
753 * vm_ops->map_pages() sets up a page
754 * table from atomic context.
755 * do_fault_around() pre-allocates
756 * page table to avoid allocation from
757 * atomic context.
758 */
759};
760
761struct vm_uffd_ops;
762
763/*
764 * These are the virtual MM functions - opening of an area, closing and
765 * unmapping it (needed to keep files on disk up-to-date etc), pointer
766 * to the functions called when a no-page or a wp-page exception occurs.
767 */
768struct vm_operations_struct {
769 /**
770 * @open: Called when a VMA is remapped, split or forked. Not called
771 * upon first mapping a VMA.
772 * Context: User context. May sleep. Caller holds mmap_lock.
773 */
774 void (*open)(struct vm_area_struct *vma);
775 /**
776 * @close: Called when the VMA is being removed from the MM.
777 * Context: User context. May sleep. Caller holds mmap_lock.
778 */
779 void (*close)(struct vm_area_struct *vma);
780 /**
781 * @mapped: Called when the VMA is first mapped in the MM. Not called if
782 * the new VMA is merged with an adjacent VMA.
783 *
784 * The @vm_private_data field is an output field allowing the user to
785 * modify vma->vm_private_data as necessary.
786 *
787 * ONLY valid if set from f_op->mmap_prepare. Will result in an error if
788 * set from f_op->mmap.
789 *
790 * Returns %0 on success, or an error otherwise. On error, the VMA will
791 * be unmapped.
792 *
793 * Context: User context. May sleep. Caller holds mmap_lock.
794 */
795 int (*mapped)(unsigned long start, unsigned long end, pgoff_t pgoff,
796 const struct file *file, void **vm_private_data);
797 /* Called any time before splitting to check if it's allowed */
798 int (*may_split)(struct vm_area_struct *vma, unsigned long addr);
799 int (*mremap)(struct vm_area_struct *vma);
800 /*
801 * Called by mprotect() to make driver-specific permission
802 * checks before mprotect() is finalised. The VMA must not
803 * be modified. Returns 0 if mprotect() can proceed.
804 */
805 int (*mprotect)(struct vm_area_struct *vma, unsigned long start,
806 unsigned long end, unsigned long newflags);
807 vm_fault_t (*fault)(struct vm_fault *vmf);
808 vm_fault_t (*huge_fault)(struct vm_fault *vmf, unsigned int order);
809 vm_fault_t (*map_pages)(struct vm_fault *vmf,
810 pgoff_t start_pgoff, pgoff_t end_pgoff);
811 unsigned long (*pagesize)(struct vm_area_struct *vma);
812
813 /* notification that a previously read-only page is about to become
814 * writable, if an error is returned it will cause a SIGBUS */
815 vm_fault_t (*page_mkwrite)(struct vm_fault *vmf);
816
817 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
818 vm_fault_t (*pfn_mkwrite)(struct vm_fault *vmf);
819
820 /* called by access_process_vm when get_user_pages() fails, typically
821 * for use by special VMAs. See also generic_access_phys() for a generic
822 * implementation useful for any iomem mapping.
823 */
824 int (*access)(struct vm_area_struct *vma, unsigned long addr,
825 void *buf, int len, int write);
826
827 /* Called by the /proc/PID/maps code to ask the vma whether it
828 * has a special name. Returning non-NULL will also cause this
829 * vma to be dumped unconditionally. */
830 const char *(*name)(struct vm_area_struct *vma);
831
832#ifdef CONFIG_NUMA
833 /*
834 * set_policy() op must add a reference to any non-NULL @new mempolicy
835 * to hold the policy upon return. Caller should pass NULL @new to
836 * remove a policy and fall back to surrounding context--i.e. do not
837 * install a MPOL_DEFAULT policy, nor the task or system default
838 * mempolicy.
839 */
840 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
841
842 /*
843 * get_policy() op must add reference [mpol_get()] to any policy at
844 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
845 * in mm/mempolicy.c will do this automatically.
846 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
847 * marked as MPOL_SHARED. vma policies are protected by the mmap_lock.
848 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
849 * must return NULL--i.e., do not "fallback" to task or system default
850 * policy.
851 */
852 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
853 unsigned long addr, pgoff_t *ilx);
854#endif
855#ifdef CONFIG_FIND_NORMAL_PAGE
856 /*
857 * Called by vm_normal_page() for special PTEs in @vma at @addr. This
858 * allows for returning a "normal" page from vm_normal_page() even
859 * though the PTE indicates that the "struct page" either does not exist
860 * or should not be touched: "special".
861 *
862 * Do not add new users: this really only works when a "normal" page
863 * was mapped, but then the PTE got changed to something weird (+
864 * marked special) that would not make pte_pfn() identify the originally
865 * inserted page.
866 */
867 struct page *(*find_normal_page)(struct vm_area_struct *vma,
868 unsigned long addr);
869#endif /* CONFIG_FIND_NORMAL_PAGE */
870#ifdef CONFIG_USERFAULTFD
871 const struct vm_uffd_ops *uffd_ops;
872#endif
873};
874
875#ifdef CONFIG_NUMA_BALANCING
876static inline void vma_numab_state_init(struct vm_area_struct *vma)
877{
878 vma->numab_state = NULL;
879}
880static inline void vma_numab_state_free(struct vm_area_struct *vma)
881{
882 kfree(vma->numab_state);
883}
884#else
885static inline void vma_numab_state_init(struct vm_area_struct *vma) {}
886static inline void vma_numab_state_free(struct vm_area_struct *vma) {}
887#endif /* CONFIG_NUMA_BALANCING */
888
889/*
890 * These must be here rather than mmap_lock.h as dependent on vm_fault type,
891 * declared in this header.
892 */
893#ifdef CONFIG_PER_VMA_LOCK
894static inline void release_fault_lock(struct vm_fault *vmf)
895{
896 if (vmf->flags & FAULT_FLAG_VMA_LOCK)
897 vma_end_read(vmf->vma);
898 else
899 mmap_read_unlock(vmf->vma->vm_mm);
900}
901
902static inline void assert_fault_locked(const struct vm_fault *vmf)
903{
904 if (vmf->flags & FAULT_FLAG_VMA_LOCK)
905 vma_assert_locked(vmf->vma);
906 else
907 mmap_assert_locked(vmf->vma->vm_mm);
908}
909#else
910static inline void release_fault_lock(struct vm_fault *vmf)
911{
912 mmap_read_unlock(vmf->vma->vm_mm);
913}
914
915static inline void assert_fault_locked(const struct vm_fault *vmf)
916{
917 mmap_assert_locked(vmf->vma->vm_mm);
918}
919#endif /* CONFIG_PER_VMA_LOCK */
920
921static inline bool mm_flags_test(int flag, const struct mm_struct *mm)
922{
923 return test_bit(flag, ACCESS_PRIVATE(&mm->flags, __mm_flags));
924}
925
926static inline bool mm_flags_test_and_set(int flag, struct mm_struct *mm)
927{
928 return test_and_set_bit(flag, ACCESS_PRIVATE(&mm->flags, __mm_flags));
929}
930
931static inline bool mm_flags_test_and_clear(int flag, struct mm_struct *mm)
932{
933 return test_and_clear_bit(flag, ACCESS_PRIVATE(&mm->flags, __mm_flags));
934}
935
936static inline void mm_flags_set(int flag, struct mm_struct *mm)
937{
938 set_bit(flag, ACCESS_PRIVATE(&mm->flags, __mm_flags));
939}
940
941static inline void mm_flags_clear(int flag, struct mm_struct *mm)
942{
943 clear_bit(flag, ACCESS_PRIVATE(&mm->flags, __mm_flags));
944}
945
946static inline void mm_flags_clear_all(struct mm_struct *mm)
947{
948 bitmap_zero(ACCESS_PRIVATE(&mm->flags, __mm_flags), NUM_MM_FLAG_BITS);
949}
950
951extern const struct vm_operations_struct vma_dummy_vm_ops;
952
953static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
954{
955 memset(vma, 0, sizeof(*vma));
956 vma->vm_mm = mm;
957 vma->vm_ops = &vma_dummy_vm_ops;
958 INIT_LIST_HEAD(&vma->anon_vma_chain);
959 vma_lock_init(vma, false);
960}
961
962/* Use when VMA is not part of the VMA tree and needs no locking */
963static inline void vm_flags_init(struct vm_area_struct *vma,
964 vm_flags_t flags)
965{
966 VM_WARN_ON_ONCE(!pgtable_supports_soft_dirty() && (flags & VM_SOFTDIRTY));
967 vma_flags_clear_all(&vma->flags);
968 vma_flags_overwrite_word(&vma->flags, flags);
969}
970
971/*
972 * Use when VMA is part of the VMA tree and modifications need coordination
973 * Note: vm_flags_reset and vm_flags_reset_once do not lock the vma and
974 * it should be locked explicitly beforehand.
975 */
976static inline void vm_flags_reset(struct vm_area_struct *vma,
977 vm_flags_t flags)
978{
979 VM_WARN_ON_ONCE(!pgtable_supports_soft_dirty() && (flags & VM_SOFTDIRTY));
980 vma_assert_write_locked(vma);
981 vm_flags_init(vma, flags);
982}
983
984static inline void vma_flags_reset_once(struct vm_area_struct *vma,
985 vma_flags_t *flags)
986{
987 const unsigned long word = flags->__vma_flags[0];
988
989 /* It is assumed only the first system word must be written once. */
990 vma_flags_overwrite_word_once(&vma->flags, word);
991 /* The remainder can be copied normally. */
992 if (NUM_VMA_FLAG_BITS > BITS_PER_LONG) {
993 unsigned long *dst = &vma->flags.__vma_flags[1];
994 const unsigned long *src = &flags->__vma_flags[1];
995
996 bitmap_copy(dst, src, NUM_VMA_FLAG_BITS - BITS_PER_LONG);
997 }
998}
999
1000static inline void vm_flags_set(struct vm_area_struct *vma,
1001 vm_flags_t flags)
1002{
1003 vma_start_write(vma);
1004 vma_flags_set_word(&vma->flags, flags);
1005}
1006
1007static inline void vm_flags_clear(struct vm_area_struct *vma,
1008 vm_flags_t flags)
1009{
1010 VM_WARN_ON_ONCE(!pgtable_supports_soft_dirty() && (flags & VM_SOFTDIRTY));
1011 vma_start_write(vma);
1012 vma_flags_clear_word(&vma->flags, flags);
1013}
1014
1015/*
1016 * Use only if VMA is not part of the VMA tree or has no other users and
1017 * therefore needs no locking.
1018 */
1019static inline void __vm_flags_mod(struct vm_area_struct *vma,
1020 vm_flags_t set, vm_flags_t clear)
1021{
1022 vm_flags_init(vma, (vma->vm_flags | set) & ~clear);
1023}
1024
1025/*
1026 * Use only when the order of set/clear operations is unimportant, otherwise
1027 * use vm_flags_{set|clear} explicitly.
1028 */
1029static inline void vm_flags_mod(struct vm_area_struct *vma,
1030 vm_flags_t set, vm_flags_t clear)
1031{
1032 vma_start_write(vma);
1033 __vm_flags_mod(vma, set, clear);
1034}
1035
1036static __always_inline bool __vma_atomic_valid_flag(struct vm_area_struct *vma,
1037 vma_flag_t bit)
1038{
1039 const vm_flags_t mask = BIT((__force int)bit);
1040
1041 /* Only specific flags are permitted */
1042 if (WARN_ON_ONCE(!(mask & VM_ATOMIC_SET_ALLOWED)))
1043 return false;
1044
1045 return true;
1046}
1047
1048/*
1049 * Set VMA flag atomically. Requires only VMA/mmap read lock. Only specific
1050 * valid flags are allowed to do this.
1051 */
1052static __always_inline void vma_set_atomic_flag(struct vm_area_struct *vma,
1053 vma_flag_t bit)
1054{
1055 unsigned long *bitmap = vma->flags.__vma_flags;
1056
1057 vma_assert_stabilised(vma);
1058 if (__vma_atomic_valid_flag(vma, bit))
1059 set_bit((__force int)bit, bitmap);
1060}
1061
1062/*
1063 * Test for VMA flag atomically. Requires no locks. Only specific valid flags
1064 * are allowed to do this.
1065 *
1066 * This is necessarily racey, so callers must ensure that serialisation is
1067 * achieved through some other means, or that races are permissible.
1068 */
1069static __always_inline bool vma_test_atomic_flag(struct vm_area_struct *vma,
1070 vma_flag_t bit)
1071{
1072 if (__vma_atomic_valid_flag(vma, bit))
1073 return test_bit((__force int)bit, &vma->vm_flags);
1074
1075 return false;
1076}
1077
1078/* Set an individual VMA flag in flags, non-atomically. */
1079static __always_inline void vma_flags_set_flag(vma_flags_t *flags,
1080 vma_flag_t bit)
1081{
1082 unsigned long *bitmap = flags->__vma_flags;
1083
1084 __set_bit((__force int)bit, bitmap);
1085}
1086
1087static __always_inline vma_flags_t __mk_vma_flags(vma_flags_t flags,
1088 size_t count, const vma_flag_t *bits)
1089{
1090 int i;
1091
1092 for (i = 0; i < count; i++)
1093 vma_flags_set_flag(&flags, bits[i]);
1094 return flags;
1095}
1096
1097/*
1098 * Helper macro which bitwise-or combines the specified input flags into a
1099 * vma_flags_t bitmap value. E.g.:
1100 *
1101 * vma_flags_t flags = mk_vma_flags(VMA_IO_BIT, VMA_PFNMAP_BIT,
1102 * VMA_DONTEXPAND_BIT, VMA_DONTDUMP_BIT);
1103 *
1104 * The compiler cleverly optimises away all of the work and this ends up being
1105 * equivalent to aggregating the values manually.
1106 */
1107#define mk_vma_flags(...) __mk_vma_flags(EMPTY_VMA_FLAGS, \
1108 COUNT_ARGS(__VA_ARGS__), (const vma_flag_t []){__VA_ARGS__})
1109
1110/*
1111 * Helper macro which acts like mk_vma_flags, only appending to a copy of the
1112 * specified flags rather than establishing new flags. E.g.:
1113 *
1114 * vma_flags_t flags = append_vma_flags(VMA_STACK_DEFAULT_FLAGS, VMA_STACK_BIT,
1115 * VMA_ACCOUNT_BIT);
1116 */
1117#define append_vma_flags(flags, ...) __mk_vma_flags(flags, \
1118 COUNT_ARGS(__VA_ARGS__), (const vma_flag_t []){__VA_ARGS__})
1119
1120/* Calculates the number of set bits in the specified VMA flags. */
1121static __always_inline int vma_flags_count(const vma_flags_t *flags)
1122{
1123 const unsigned long *bitmap = flags->__vma_flags;
1124
1125 return bitmap_weight(bitmap, NUM_VMA_FLAG_BITS);
1126}
1127
1128/*
1129 * Test whether a specific VMA flag is set, e.g.:
1130 *
1131 * if (vma_flags_test(flags, VMA_READ_BIT)) { ... }
1132 */
1133static __always_inline bool vma_flags_test(const vma_flags_t *flags,
1134 vma_flag_t bit)
1135{
1136 const unsigned long *bitmap = flags->__vma_flags;
1137
1138 return test_bit((__force int)bit, bitmap);
1139}
1140
1141/*
1142 * Obtain a set of VMA flags which contain the overlapping flags contained
1143 * within flags and to_and.
1144 */
1145static __always_inline vma_flags_t vma_flags_and_mask(const vma_flags_t *flags,
1146 vma_flags_t to_and)
1147{
1148 vma_flags_t dst;
1149 unsigned long *bitmap_dst = dst.__vma_flags;
1150 const unsigned long *bitmap = flags->__vma_flags;
1151 const unsigned long *bitmap_to_and = to_and.__vma_flags;
1152
1153 bitmap_and(bitmap_dst, bitmap, bitmap_to_and, NUM_VMA_FLAG_BITS);
1154 return dst;
1155}
1156
1157/*
1158 * Obtain a set of VMA flags which contains the specified overlapping flags,
1159 * e.g.:
1160 *
1161 * vma_flags_t read_flags = vma_flags_and(&flags, VMA_READ_BIT,
1162 * VMA_MAY_READ_BIT);
1163 */
1164#define vma_flags_and(flags, ...) \
1165 vma_flags_and_mask(flags, mk_vma_flags(__VA_ARGS__))
1166
1167/* Test each of to_test flags in flags, non-atomically. */
1168static __always_inline bool vma_flags_test_any_mask(const vma_flags_t *flags,
1169 vma_flags_t to_test)
1170{
1171 const unsigned long *bitmap = flags->__vma_flags;
1172 const unsigned long *bitmap_to_test = to_test.__vma_flags;
1173
1174 return bitmap_intersects(bitmap_to_test, bitmap, NUM_VMA_FLAG_BITS);
1175}
1176
1177/*
1178 * Test whether any specified VMA flag is set, e.g.:
1179 *
1180 * if (vma_flags_test_any(flags, VMA_READ_BIT, VMA_MAYREAD_BIT)) { ... }
1181 */
1182#define vma_flags_test_any(flags, ...) \
1183 vma_flags_test_any_mask(flags, mk_vma_flags(__VA_ARGS__))
1184
1185/* Test that ALL of the to_test flags are set, non-atomically. */
1186static __always_inline bool vma_flags_test_all_mask(const vma_flags_t *flags,
1187 vma_flags_t to_test)
1188{
1189 const unsigned long *bitmap = flags->__vma_flags;
1190 const unsigned long *bitmap_to_test = to_test.__vma_flags;
1191
1192 return bitmap_subset(bitmap_to_test, bitmap, NUM_VMA_FLAG_BITS);
1193}
1194
1195/*
1196 * Test whether ALL specified VMA flags are set, e.g.:
1197 *
1198 * if (vma_flags_test_all(flags, VMA_READ_BIT, VMA_MAYREAD_BIT)) { ... }
1199 */
1200#define vma_flags_test_all(flags, ...) \
1201 vma_flags_test_all_mask(flags, mk_vma_flags(__VA_ARGS__))
1202
1203/*
1204 * Helper to test that a flag mask of type vma_flags_t has a SINGLE flag set
1205 * (returning false if flagmask has no flags set).
1206 *
1207 * This is defined to make the semantics clearer when testing an optionally
1208 * defined VMA flags mask, e.g.:
1209 *
1210 * if (vma_flags_test_single_mask(&flags, VMA_DROPPABLE)) { ... }
1211 *
1212 * When VMA_DROPPABLE is defined if available, or set to EMPTY_VMA_FLAGS
1213 * otherwise.
1214 */
1215static __always_inline bool vma_flags_test_single_mask(const vma_flags_t *flags,
1216 vma_flags_t flagmask)
1217{
1218 VM_WARN_ON_ONCE(vma_flags_count(&flagmask) > 1);
1219
1220 return vma_flags_test_any_mask(flags, flagmask);
1221}
1222
1223/* Set each of the to_set flags in flags, non-atomically. */
1224static __always_inline void vma_flags_set_mask(vma_flags_t *flags,
1225 vma_flags_t to_set)
1226{
1227 unsigned long *bitmap = flags->__vma_flags;
1228 const unsigned long *bitmap_to_set = to_set.__vma_flags;
1229
1230 bitmap_or(bitmap, bitmap, bitmap_to_set, NUM_VMA_FLAG_BITS);
1231}
1232
1233/*
1234 * Set all specified VMA flags, e.g.:
1235 *
1236 * vma_flags_set(&flags, VMA_READ_BIT, VMA_WRITE_BIT, VMA_EXEC_BIT);
1237 */
1238#define vma_flags_set(flags, ...) \
1239 vma_flags_set_mask(flags, mk_vma_flags(__VA_ARGS__))
1240
1241/* Clear all of the to-clear flags in flags, non-atomically. */
1242static __always_inline void vma_flags_clear_mask(vma_flags_t *flags,
1243 vma_flags_t to_clear)
1244{
1245 unsigned long *bitmap = flags->__vma_flags;
1246 const unsigned long *bitmap_to_clear = to_clear.__vma_flags;
1247
1248 bitmap_andnot(bitmap, bitmap, bitmap_to_clear, NUM_VMA_FLAG_BITS);
1249}
1250
1251/*
1252 * Clear all specified individual flags, e.g.:
1253 *
1254 * vma_flags_clear(&flags, VMA_READ_BIT, VMA_WRITE_BIT, VMA_EXEC_BIT);
1255 */
1256#define vma_flags_clear(flags, ...) \
1257 vma_flags_clear_mask(flags, mk_vma_flags(__VA_ARGS__))
1258
1259/*
1260 * Obtain a VMA flags value containing those flags that are present in flags or
1261 * flags_other but not in both.
1262 */
1263static __always_inline vma_flags_t vma_flags_diff_pair(const vma_flags_t *flags,
1264 const vma_flags_t *flags_other)
1265{
1266 vma_flags_t dst;
1267 const unsigned long *bitmap_other = flags_other->__vma_flags;
1268 const unsigned long *bitmap = flags->__vma_flags;
1269 unsigned long *bitmap_dst = dst.__vma_flags;
1270
1271 bitmap_xor(bitmap_dst, bitmap, bitmap_other, NUM_VMA_FLAG_BITS);
1272 return dst;
1273}
1274
1275/* Determine if flags and flags_other have precisely the same flags set. */
1276static __always_inline bool vma_flags_same_pair(const vma_flags_t *flags,
1277 const vma_flags_t *flags_other)
1278{
1279 const unsigned long *bitmap = flags->__vma_flags;
1280 const unsigned long *bitmap_other = flags_other->__vma_flags;
1281
1282 return bitmap_equal(bitmap, bitmap_other, NUM_VMA_FLAG_BITS);
1283}
1284
1285/* Determine if flags and flags_other have precisely the same flags set. */
1286static __always_inline bool vma_flags_same_mask(const vma_flags_t *flags,
1287 vma_flags_t flags_other)
1288{
1289 const unsigned long *bitmap = flags->__vma_flags;
1290 const unsigned long *bitmap_other = flags_other.__vma_flags;
1291
1292 return bitmap_equal(bitmap, bitmap_other, NUM_VMA_FLAG_BITS);
1293}
1294
1295/*
1296 * Helper macro to determine if only the specific flags are set, e.g.:
1297 *
1298 * if (vma_flags_same(&flags, VMA_WRITE_BIT) { ... }
1299 */
1300#define vma_flags_same(flags, ...) \
1301 vma_flags_same_mask(flags, mk_vma_flags(__VA_ARGS__))
1302
1303/*
1304 * Test whether a specific flag in the VMA is set, e.g.:
1305 *
1306 * if (vma_test(vma, VMA_READ_BIT)) { ... }
1307 */
1308static __always_inline bool vma_test(const struct vm_area_struct *vma,
1309 vma_flag_t bit)
1310{
1311 return vma_flags_test(&vma->flags, bit);
1312}
1313
1314/* Helper to test any VMA flags in a VMA . */
1315static __always_inline bool vma_test_any_mask(const struct vm_area_struct *vma,
1316 vma_flags_t flags)
1317{
1318 return vma_flags_test_any_mask(&vma->flags, flags);
1319}
1320
1321/*
1322 * Helper macro for testing whether any VMA flags are set in a VMA,
1323 * e.g.:
1324 *
1325 * if (vma_test_any(vma, VMA_IO_BIT, VMA_PFNMAP_BIT,
1326 * VMA_DONTEXPAND_BIT, VMA_DONTDUMP_BIT)) { ... }
1327 */
1328#define vma_test_any(vma, ...) \
1329 vma_test_any_mask(vma, mk_vma_flags(__VA_ARGS__))
1330
1331/*
1332 * Helper to test that ALL specified flags are set in a VMA.
1333 *
1334 * Note: appropriate locks must be held, this function does not acquire them for
1335 * you.
1336 */
1337static __always_inline bool vma_test_all_mask(const struct vm_area_struct *vma,
1338 vma_flags_t flags)
1339{
1340 return vma_flags_test_all_mask(&vma->flags, flags);
1341}
1342
1343/*
1344 * Helper macro for checking that ALL specified flags are set in a VMA, e.g.:
1345 *
1346 * if (vma_test_all(vma, VMA_READ_BIT, VMA_MAYREAD_BIT) { ... }
1347 */
1348#define vma_test_all(vma, ...) \
1349 vma_test_all_mask(vma, mk_vma_flags(__VA_ARGS__))
1350
1351/*
1352 * Helper to test that a flag mask of type vma_flags_t has a SINGLE flag set
1353 * (returning false if flagmask has no flags set).
1354 *
1355 * This is useful when a flag needs to be either defined or not depending upon
1356 * kernel configuration, e.g.:
1357 *
1358 * if (vma_test_single_mask(vma, VMA_DROPPABLE)) { ... }
1359 *
1360 * When VMA_DROPPABLE is defined if available, or set to EMPTY_VMA_FLAGS
1361 * otherwise.
1362 */
1363static __always_inline bool
1364vma_test_single_mask(const struct vm_area_struct *vma, vma_flags_t flagmask)
1365{
1366 return vma_flags_test_single_mask(&vma->flags, flagmask);
1367}
1368
1369/*
1370 * Helper to set all VMA flags in a VMA.
1371 *
1372 * Note: appropriate locks must be held, this function does not acquire them for
1373 * you.
1374 */
1375static __always_inline void vma_set_flags_mask(struct vm_area_struct *vma,
1376 vma_flags_t flags)
1377{
1378 vma_flags_set_mask(&vma->flags, flags);
1379}
1380
1381/*
1382 * Helper macro for specifying VMA flags in a VMA, e.g.:
1383 *
1384 * vma_set_flags(vma, VMA_IO_BIT, VMA_PFNMAP_BIT, VMA_DONTEXPAND_BIT,
1385 * VMA_DONTDUMP_BIT);
1386 *
1387 * Note: appropriate locks must be held, this function does not acquire them for
1388 * you.
1389 */
1390#define vma_set_flags(vma, ...) \
1391 vma_set_flags_mask(vma, mk_vma_flags(__VA_ARGS__))
1392
1393/* Helper to clear all VMA flags in a VMA. */
1394static __always_inline void vma_clear_flags_mask(struct vm_area_struct *vma,
1395 vma_flags_t flags)
1396{
1397 vma_flags_clear_mask(&vma->flags, flags);
1398}
1399
1400/*
1401 * Helper macro for clearing VMA flags, e.g.:
1402 *
1403 * vma_clear_flags(vma, VMA_IO_BIT, VMA_PFNMAP_BIT, VMA_DONTEXPAND_BIT,
1404 * VMA_DONTDUMP_BIT);
1405 */
1406#define vma_clear_flags(vma, ...) \
1407 vma_clear_flags_mask(vma, mk_vma_flags(__VA_ARGS__))
1408
1409/*
1410 * Test whether a specific VMA flag is set in a VMA descriptor, e.g.:
1411 *
1412 * if (vma_desc_test(desc, VMA_READ_BIT)) { ... }
1413 */
1414static __always_inline bool vma_desc_test(const struct vm_area_desc *desc,
1415 vma_flag_t bit)
1416{
1417 return vma_flags_test(&desc->vma_flags, bit);
1418}
1419
1420/* Helper to test any VMA flags in a VMA descriptor. */
1421static __always_inline bool vma_desc_test_any_mask(const struct vm_area_desc *desc,
1422 vma_flags_t flags)
1423{
1424 return vma_flags_test_any_mask(&desc->vma_flags, flags);
1425}
1426
1427/*
1428 * Helper macro for testing whether any VMA flags are set in a VMA descriptor,
1429 * e.g.:
1430 *
1431 * if (vma_desc_test_any(desc, VMA_IO_BIT, VMA_PFNMAP_BIT,
1432 * VMA_DONTEXPAND_BIT, VMA_DONTDUMP_BIT)) { ... }
1433 */
1434#define vma_desc_test_any(desc, ...) \
1435 vma_desc_test_any_mask(desc, mk_vma_flags(__VA_ARGS__))
1436
1437/* Helper to test all VMA flags in a VMA descriptor. */
1438static __always_inline bool vma_desc_test_all_mask(const struct vm_area_desc *desc,
1439 vma_flags_t flags)
1440{
1441 return vma_flags_test_all_mask(&desc->vma_flags, flags);
1442}
1443
1444/*
1445 * Helper macro for testing whether ALL VMA flags are set in a VMA descriptor,
1446 * e.g.:
1447 *
1448 * if (vma_desc_test_all(desc, VMA_READ_BIT, VMA_MAYREAD_BIT)) { ... }
1449 */
1450#define vma_desc_test_all(desc, ...) \
1451 vma_desc_test_all_mask(desc, mk_vma_flags(__VA_ARGS__))
1452
1453/* Helper to set all VMA flags in a VMA descriptor. */
1454static __always_inline void vma_desc_set_flags_mask(struct vm_area_desc *desc,
1455 vma_flags_t flags)
1456{
1457 vma_flags_set_mask(&desc->vma_flags, flags);
1458}
1459
1460/*
1461 * Helper macro for specifying VMA flags for an input pointer to a struct
1462 * vm_area_desc object describing a proposed VMA, e.g.:
1463 *
1464 * vma_desc_set_flags(desc, VMA_IO_BIT, VMA_PFNMAP_BIT, VMA_DONTEXPAND_BIT,
1465 * VMA_DONTDUMP_BIT);
1466 */
1467#define vma_desc_set_flags(desc, ...) \
1468 vma_desc_set_flags_mask(desc, mk_vma_flags(__VA_ARGS__))
1469
1470/* Helper to clear all VMA flags in a VMA descriptor. */
1471static __always_inline void vma_desc_clear_flags_mask(struct vm_area_desc *desc,
1472 vma_flags_t flags)
1473{
1474 vma_flags_clear_mask(&desc->vma_flags, flags);
1475}
1476
1477/*
1478 * Helper macro for clearing VMA flags for an input pointer to a struct
1479 * vm_area_desc object describing a proposed VMA, e.g.:
1480 *
1481 * vma_desc_clear_flags(desc, VMA_IO_BIT, VMA_PFNMAP_BIT, VMA_DONTEXPAND_BIT,
1482 * VMA_DONTDUMP_BIT);
1483 */
1484#define vma_desc_clear_flags(desc, ...) \
1485 vma_desc_clear_flags_mask(desc, mk_vma_flags(__VA_ARGS__))
1486
1487static inline void vma_set_anonymous(struct vm_area_struct *vma)
1488{
1489 vma->vm_ops = NULL;
1490}
1491
1492static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1493{
1494 return !vma->vm_ops;
1495}
1496
1497/*
1498 * Indicate if the VMA is a heap for the given task; for
1499 * /proc/PID/maps that is the heap of the main task.
1500 */
1501static inline bool vma_is_initial_heap(const struct vm_area_struct *vma)
1502{
1503 return vma->vm_start < vma->vm_mm->brk &&
1504 vma->vm_end > vma->vm_mm->start_brk;
1505}
1506
1507/*
1508 * Indicate if the VMA is a stack for the given task; for
1509 * /proc/PID/maps that is the stack of the main task.
1510 */
1511static inline bool vma_is_initial_stack(const struct vm_area_struct *vma)
1512{
1513 /*
1514 * We make no effort to guess what a given thread considers to be
1515 * its "stack". It's not even well-defined for programs written
1516 * languages like Go.
1517 */
1518 return vma->vm_start <= vma->vm_mm->start_stack &&
1519 vma->vm_end >= vma->vm_mm->start_stack;
1520}
1521
1522static inline bool vma_is_temporary_stack(const struct vm_area_struct *vma)
1523{
1524 int maybe_stack = vma->vm_flags & (VM_GROWSDOWN | VM_GROWSUP);
1525
1526 if (!maybe_stack)
1527 return false;
1528
1529 if ((vma->vm_flags & VM_STACK_INCOMPLETE_SETUP) ==
1530 VM_STACK_INCOMPLETE_SETUP)
1531 return true;
1532
1533 return false;
1534}
1535
1536static inline bool vma_is_foreign(const struct vm_area_struct *vma)
1537{
1538 if (!current->mm)
1539 return true;
1540
1541 if (current->mm != vma->vm_mm)
1542 return true;
1543
1544 return false;
1545}
1546
1547static inline bool vma_is_accessible(const struct vm_area_struct *vma)
1548{
1549 return vma->vm_flags & VM_ACCESS_FLAGS;
1550}
1551
1552static inline bool is_shared_maywrite(const vma_flags_t *flags)
1553{
1554 return vma_flags_test_all(flags, VMA_SHARED_BIT, VMA_MAYWRITE_BIT);
1555}
1556
1557static inline bool vma_is_shared_maywrite(const struct vm_area_struct *vma)
1558{
1559 return is_shared_maywrite(&vma->flags);
1560}
1561
1562/**
1563 * vma_kernel_pagesize - Default page size granularity for this VMA.
1564 * @vma: The user mapping.
1565 *
1566 * The kernel page size specifies in which granularity VMA modifications
1567 * can be performed. Folios in this VMA will be aligned to, and at least
1568 * the size of the number of bytes returned by this function.
1569 *
1570 * The default kernel page size is not affected by Transparent Huge Pages
1571 * being in effect.
1572 *
1573 * Return: The default page size granularity for this VMA.
1574 */
1575static inline unsigned long vma_kernel_pagesize(struct vm_area_struct *vma)
1576{
1577 if (unlikely(vma->vm_ops && vma->vm_ops->pagesize))
1578 return vma->vm_ops->pagesize(vma);
1579 return PAGE_SIZE;
1580}
1581
1582unsigned long vma_mmu_pagesize(struct vm_area_struct *vma);
1583
1584static inline
1585struct vm_area_struct *vma_find(struct vma_iterator *vmi, unsigned long max)
1586{
1587 return mas_find(&vmi->mas, max - 1);
1588}
1589
1590static inline struct vm_area_struct *vma_next(struct vma_iterator *vmi)
1591{
1592 /*
1593 * Uses mas_find() to get the first VMA when the iterator starts.
1594 * Calling mas_next() could skip the first entry.
1595 */
1596 return mas_find(&vmi->mas, ULONG_MAX);
1597}
1598
1599static inline
1600struct vm_area_struct *vma_iter_next_range(struct vma_iterator *vmi)
1601{
1602 return mas_next_range(&vmi->mas, ULONG_MAX);
1603}
1604
1605
1606static inline struct vm_area_struct *vma_prev(struct vma_iterator *vmi)
1607{
1608 return mas_prev(&vmi->mas, 0);
1609}
1610
1611static inline int vma_iter_clear_gfp(struct vma_iterator *vmi,
1612 unsigned long start, unsigned long end, gfp_t gfp)
1613{
1614 __mas_set_range(&vmi->mas, start, end - 1);
1615 mas_store_gfp(&vmi->mas, NULL, gfp);
1616 if (unlikely(mas_is_err(&vmi->mas)))
1617 return -ENOMEM;
1618
1619 return 0;
1620}
1621
1622/* Free any unused preallocations */
1623static inline void vma_iter_free(struct vma_iterator *vmi)
1624{
1625 mas_destroy(&vmi->mas);
1626}
1627
1628static inline int vma_iter_bulk_store(struct vma_iterator *vmi,
1629 struct vm_area_struct *vma)
1630{
1631 vmi->mas.index = vma->vm_start;
1632 vmi->mas.last = vma->vm_end - 1;
1633 mas_store(&vmi->mas, vma);
1634 if (unlikely(mas_is_err(&vmi->mas)))
1635 return -ENOMEM;
1636
1637 vma_mark_attached(vma);
1638 return 0;
1639}
1640
1641static inline void vma_iter_invalidate(struct vma_iterator *vmi)
1642{
1643 mas_pause(&vmi->mas);
1644}
1645
1646static inline void vma_iter_set(struct vma_iterator *vmi, unsigned long addr)
1647{
1648 mas_set(&vmi->mas, addr);
1649}
1650
1651#define for_each_vma(__vmi, __vma) \
1652 while (((__vma) = vma_next(&(__vmi))) != NULL)
1653
1654/* The MM code likes to work with exclusive end addresses */
1655#define for_each_vma_range(__vmi, __vma, __end) \
1656 while (((__vma) = vma_find(&(__vmi), (__end))) != NULL)
1657
1658#ifdef CONFIG_SHMEM
1659/*
1660 * The vma_is_shmem is not inline because it is used only by slow
1661 * paths in userfault.
1662 */
1663bool vma_is_shmem(const struct vm_area_struct *vma);
1664bool vma_is_anon_shmem(const struct vm_area_struct *vma);
1665#else
1666static inline bool vma_is_shmem(const struct vm_area_struct *vma) { return false; }
1667static inline bool vma_is_anon_shmem(const struct vm_area_struct *vma) { return false; }
1668#endif
1669
1670int vma_is_stack_for_current(const struct vm_area_struct *vma);
1671
1672/* flush_tlb_range() takes a vma, not a mm, and can care about flags */
1673#define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
1674
1675struct mmu_gather;
1676struct inode;
1677
1678extern void prep_compound_page(struct page *page, unsigned int order);
1679
1680static inline unsigned int folio_large_order(const struct folio *folio)
1681{
1682 return folio->_flags_1 & 0xff;
1683}
1684
1685#ifdef NR_PAGES_IN_LARGE_FOLIO
1686static inline unsigned long folio_large_nr_pages(const struct folio *folio)
1687{
1688 return folio->_nr_pages;
1689}
1690#else
1691static inline unsigned long folio_large_nr_pages(const struct folio *folio)
1692{
1693 return 1L << folio_large_order(folio);
1694}
1695#endif
1696
1697/*
1698 * compound_order() can be called without holding a reference, which means
1699 * that niceties like page_folio() don't work. These callers should be
1700 * prepared to handle wild return values. For example, PG_head may be
1701 * set before the order is initialised, or this may be a tail page.
1702 * See compaction.c for some good examples.
1703 */
1704static inline unsigned int compound_order(const struct page *page)
1705{
1706 const struct folio *folio = (struct folio *)page;
1707
1708 if (!test_bit(PG_head, &folio->flags.f))
1709 return 0;
1710 return folio_large_order(folio);
1711}
1712
1713/**
1714 * folio_order - The allocation order of a folio.
1715 * @folio: The folio.
1716 *
1717 * A folio is composed of 2^order pages. See get_order() for the definition
1718 * of order.
1719 *
1720 * Return: The order of the folio.
1721 */
1722static inline unsigned int folio_order(const struct folio *folio)
1723{
1724 if (!folio_test_large(folio))
1725 return 0;
1726 return folio_large_order(folio);
1727}
1728
1729/**
1730 * folio_reset_order - Reset the folio order and derived _nr_pages
1731 * @folio: The folio.
1732 *
1733 * Reset the order and derived _nr_pages to 0. Must only be used in the
1734 * process of splitting large folios.
1735 */
1736static inline void folio_reset_order(struct folio *folio)
1737{
1738 if (WARN_ON_ONCE(!folio_test_large(folio)))
1739 return;
1740 folio->_flags_1 &= ~0xffUL;
1741#ifdef NR_PAGES_IN_LARGE_FOLIO
1742 folio->_nr_pages = 0;
1743#endif
1744}
1745
1746#include <linux/huge_mm.h>
1747
1748/*
1749 * Methods to modify the page usage count.
1750 *
1751 * What counts for a page usage:
1752 * - cache mapping (page->mapping)
1753 * - private data (page->private)
1754 * - page mapped in a task's page tables, each mapping
1755 * is counted separately
1756 *
1757 * Also, many kernel routines increase the page count before a critical
1758 * routine so they can be sure the page doesn't go away from under them.
1759 */
1760
1761/*
1762 * Drop a ref, return true if the refcount fell to zero (the page has no users)
1763 */
1764static inline int put_page_testzero(struct page *page)
1765{
1766 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
1767 return page_ref_dec_and_test(page);
1768}
1769
1770static inline int folio_put_testzero(struct folio *folio)
1771{
1772 return put_page_testzero(&folio->page);
1773}
1774
1775/*
1776 * Try to grab a ref unless the page has a refcount of zero, return false if
1777 * that is the case.
1778 * This can be called when MMU is off so it must not access
1779 * any of the virtual mappings.
1780 */
1781static inline bool get_page_unless_zero(struct page *page)
1782{
1783 return page_ref_add_unless_zero(page, 1);
1784}
1785
1786static inline struct folio *folio_get_nontail_page(struct page *page)
1787{
1788 if (unlikely(!get_page_unless_zero(page)))
1789 return NULL;
1790 return (struct folio *)page;
1791}
1792
1793extern int page_is_ram(unsigned long pfn);
1794
1795enum {
1796 REGION_INTERSECTS,
1797 REGION_DISJOINT,
1798 REGION_MIXED,
1799};
1800
1801int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
1802 unsigned long desc);
1803
1804/* Support for virtually mapped pages */
1805struct page *vmalloc_to_page(const void *addr);
1806unsigned long vmalloc_to_pfn(const void *addr);
1807
1808/*
1809 * Determine if an address is within the vmalloc range
1810 *
1811 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
1812 * is no special casing required.
1813 */
1814#ifdef CONFIG_MMU
1815extern bool is_vmalloc_addr(const void *x);
1816extern int is_vmalloc_or_module_addr(const void *x);
1817#else
1818static inline bool is_vmalloc_addr(const void *x)
1819{
1820 return false;
1821}
1822static inline int is_vmalloc_or_module_addr(const void *x)
1823{
1824 return 0;
1825}
1826#endif
1827
1828/*
1829 * How many times the entire folio is mapped as a single unit (eg by a
1830 * PMD or PUD entry). This is probably not what you want, except for
1831 * debugging purposes or implementation of other core folio_*() primitives.
1832 */
1833static inline int folio_entire_mapcount(const struct folio *folio)
1834{
1835 VM_BUG_ON_FOLIO(!folio_test_large(folio), folio);
1836 if (!IS_ENABLED(CONFIG_64BIT) && unlikely(folio_large_order(folio) == 1))
1837 return 0;
1838 return atomic_read(&folio->_entire_mapcount) + 1;
1839}
1840
1841static inline int folio_large_mapcount(const struct folio *folio)
1842{
1843 VM_WARN_ON_FOLIO(!folio_test_large(folio), folio);
1844 return atomic_read(&folio->_large_mapcount) + 1;
1845}
1846
1847/**
1848 * folio_mapcount() - Number of mappings of this folio.
1849 * @folio: The folio.
1850 *
1851 * The folio mapcount corresponds to the number of present user page table
1852 * entries that reference any part of a folio. Each such present user page
1853 * table entry must be paired with exactly on folio reference.
1854 *
1855 * For ordindary folios, each user page table entry (PTE/PMD/PUD/...) counts
1856 * exactly once.
1857 *
1858 * For hugetlb folios, each abstracted "hugetlb" user page table entry that
1859 * references the entire folio counts exactly once, even when such special
1860 * page table entries are comprised of multiple ordinary page table entries.
1861 *
1862 * Will report 0 for pages which cannot be mapped into userspace, such as
1863 * slab, page tables and similar.
1864 *
1865 * Return: The number of times this folio is mapped.
1866 */
1867static inline int folio_mapcount(const struct folio *folio)
1868{
1869 int mapcount;
1870
1871 if (likely(!folio_test_large(folio))) {
1872 mapcount = atomic_read(&folio->_mapcount) + 1;
1873 if (page_mapcount_is_type(mapcount))
1874 mapcount = 0;
1875 return mapcount;
1876 }
1877 return folio_large_mapcount(folio);
1878}
1879
1880/**
1881 * folio_mapped - Is this folio mapped into userspace?
1882 * @folio: The folio.
1883 *
1884 * Return: True if any page in this folio is referenced by user page tables.
1885 */
1886static inline bool folio_mapped(const struct folio *folio)
1887{
1888 return folio_mapcount(folio) >= 1;
1889}
1890
1891/*
1892 * Return true if this page is mapped into pagetables.
1893 * For compound page it returns true if any sub-page of compound page is mapped,
1894 * even if this particular sub-page is not itself mapped by any PTE or PMD.
1895 */
1896static inline bool page_mapped(const struct page *page)
1897{
1898 return folio_mapped(page_folio(page));
1899}
1900
1901static inline struct page *virt_to_head_page(const void *x)
1902{
1903 struct page *page = virt_to_page(x);
1904
1905 return compound_head(page);
1906}
1907
1908static inline struct folio *virt_to_folio(const void *x)
1909{
1910 struct page *page = virt_to_page(x);
1911
1912 return page_folio(page);
1913}
1914
1915void __folio_put(struct folio *folio);
1916
1917void split_page(struct page *page, unsigned int order);
1918void folio_copy(struct folio *dst, struct folio *src);
1919int folio_mc_copy(struct folio *dst, struct folio *src);
1920
1921unsigned long nr_free_buffer_pages(void);
1922
1923/* Returns the number of bytes in this potentially compound page. */
1924static inline unsigned long page_size(const struct page *page)
1925{
1926 return PAGE_SIZE << compound_order(page);
1927}
1928
1929/* Returns the number of bits needed for the number of bytes in a page */
1930static inline unsigned int page_shift(struct page *page)
1931{
1932 return PAGE_SHIFT + compound_order(page);
1933}
1934
1935/**
1936 * thp_order - Order of a transparent huge page.
1937 * @page: Head page of a transparent huge page.
1938 */
1939static inline unsigned int thp_order(struct page *page)
1940{
1941 VM_BUG_ON_PGFLAGS(PageTail(page), page);
1942 return compound_order(page);
1943}
1944
1945/**
1946 * thp_size - Size of a transparent huge page.
1947 * @page: Head page of a transparent huge page.
1948 *
1949 * Return: Number of bytes in this page.
1950 */
1951static inline unsigned long thp_size(struct page *page)
1952{
1953 return PAGE_SIZE << thp_order(page);
1954}
1955
1956#ifdef CONFIG_MMU
1957/*
1958 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
1959 * servicing faults for write access. In the normal case, do always want
1960 * pte_mkwrite. But get_user_pages can cause write faults for mappings
1961 * that do not have writing enabled, when used by access_process_vm.
1962 */
1963static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
1964{
1965 if (likely(vma->vm_flags & VM_WRITE))
1966 pte = pte_mkwrite(pte, vma);
1967 return pte;
1968}
1969
1970vm_fault_t do_set_pmd(struct vm_fault *vmf, struct folio *folio, struct page *page);
1971void set_pte_range(struct vm_fault *vmf, struct folio *folio,
1972 struct page *page, unsigned int nr, unsigned long addr);
1973
1974vm_fault_t finish_fault(struct vm_fault *vmf);
1975#endif
1976
1977/*
1978 * Multiple processes may "see" the same page. E.g. for untouched
1979 * mappings of /dev/null, all processes see the same page full of
1980 * zeroes, and text pages of executables and shared libraries have
1981 * only one copy in memory, at most, normally.
1982 *
1983 * For the non-reserved pages, page_count(page) denotes a reference count.
1984 * page_count() == 0 means the page is free. page->lru is then used for
1985 * freelist management in the buddy allocator.
1986 * page_count() > 0 means the page has been allocated.
1987 *
1988 * Pages are allocated by the slab allocator in order to provide memory
1989 * to kmalloc and kmem_cache_alloc. In this case, the management of the
1990 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
1991 * unless a particular usage is carefully commented. (the responsibility of
1992 * freeing the kmalloc memory is the caller's, of course).
1993 *
1994 * A page may be used by anyone else who does a __get_free_page().
1995 * In this case, page_count still tracks the references, and should only
1996 * be used through the normal accessor functions. The top bits of page->flags
1997 * and page->virtual store page management information, but all other fields
1998 * are unused and could be used privately, carefully. The management of this
1999 * page is the responsibility of the one who allocated it, and those who have
2000 * subsequently been given references to it.
2001 *
2002 * The other pages (we may call them "pagecache pages") are completely
2003 * managed by the Linux memory manager: I/O, buffers, swapping etc.
2004 * The following discussion applies only to them.
2005 *
2006 * A pagecache page contains an opaque `private' member, which belongs to the
2007 * page's address_space. Usually, this is the address of a circular list of
2008 * the page's disk buffers. PG_private must be set to tell the VM to call
2009 * into the filesystem to release these pages.
2010 *
2011 * A folio may belong to an inode's memory mapping. In this case,
2012 * folio->mapping points to the inode, and folio->index is the file
2013 * offset of the folio, in units of PAGE_SIZE.
2014 *
2015 * If pagecache pages are not associated with an inode, they are said to be
2016 * anonymous pages. These may become associated with the swapcache, and in that
2017 * case PG_swapcache is set, and page->private is an offset into the swapcache.
2018 *
2019 * In either case (swapcache or inode backed), the pagecache itself holds one
2020 * reference to the page. Setting PG_private should also increment the
2021 * refcount. The each user mapping also has a reference to the page.
2022 *
2023 * The pagecache pages are stored in a per-mapping radix tree, which is
2024 * rooted at mapping->i_pages, and indexed by offset.
2025 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
2026 * lists, we instead now tag pages as dirty/writeback in the radix tree.
2027 *
2028 * All pagecache pages may be subject to I/O:
2029 * - inode pages may need to be read from disk,
2030 * - inode pages which have been modified and are MAP_SHARED may need
2031 * to be written back to the inode on disk,
2032 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
2033 * modified may need to be swapped out to swap space and (later) to be read
2034 * back into memory.
2035 */
2036
2037/* 127: arbitrary random number, small enough to assemble well */
2038#define folio_ref_zero_or_close_to_overflow(folio) \
2039 ((unsigned int) folio_ref_count(folio) + 127u <= 127u)
2040
2041/**
2042 * folio_get - Increment the reference count on a folio.
2043 * @folio: The folio.
2044 *
2045 * Context: May be called in any context, as long as you know that
2046 * you have a refcount on the folio. If you do not already have one,
2047 * folio_try_get() may be the right interface for you to use.
2048 */
2049static inline void folio_get(struct folio *folio)
2050{
2051 VM_BUG_ON_FOLIO(folio_ref_zero_or_close_to_overflow(folio), folio);
2052 folio_ref_inc(folio);
2053}
2054
2055static inline void get_page(struct page *page)
2056{
2057 struct folio *folio = page_folio(page);
2058 if (WARN_ON_ONCE(folio_test_slab(folio)))
2059 return;
2060 if (WARN_ON_ONCE(folio_test_large_kmalloc(folio)))
2061 return;
2062 folio_get(folio);
2063}
2064
2065static inline __must_check bool try_get_page(struct page *page)
2066{
2067 page = compound_head(page);
2068 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
2069 return false;
2070 page_ref_inc(page);
2071 return true;
2072}
2073
2074/**
2075 * folio_put - Decrement the reference count on a folio.
2076 * @folio: The folio.
2077 *
2078 * If the folio's reference count reaches zero, the memory will be
2079 * released back to the page allocator and may be used by another
2080 * allocation immediately. Do not access the memory or the struct folio
2081 * after calling folio_put() unless you can be sure that it wasn't the
2082 * last reference.
2083 *
2084 * Context: May be called in process or interrupt context, but not in NMI
2085 * context. May be called while holding a spinlock.
2086 */
2087static inline void folio_put(struct folio *folio)
2088{
2089 if (folio_put_testzero(folio))
2090 __folio_put(folio);
2091}
2092
2093/**
2094 * folio_put_refs - Reduce the reference count on a folio.
2095 * @folio: The folio.
2096 * @refs: The amount to subtract from the folio's reference count.
2097 *
2098 * If the folio's reference count reaches zero, the memory will be
2099 * released back to the page allocator and may be used by another
2100 * allocation immediately. Do not access the memory or the struct folio
2101 * after calling folio_put_refs() unless you can be sure that these weren't
2102 * the last references.
2103 *
2104 * Context: May be called in process or interrupt context, but not in NMI
2105 * context. May be called while holding a spinlock.
2106 */
2107static inline void folio_put_refs(struct folio *folio, int refs)
2108{
2109 if (folio_ref_sub_and_test(folio, refs))
2110 __folio_put(folio);
2111}
2112
2113void folios_put_refs(struct folio_batch *folios, unsigned int *refs);
2114
2115/*
2116 * union release_pages_arg - an array of pages or folios
2117 *
2118 * release_pages() releases a simple array of multiple pages, and
2119 * accepts various different forms of said page array: either
2120 * a regular old boring array of pages, an array of folios, or
2121 * an array of encoded page pointers.
2122 *
2123 * The transparent union syntax for this kind of "any of these
2124 * argument types" is all kinds of ugly, so look away.
2125 */
2126typedef union {
2127 struct page **pages;
2128 struct folio **folios;
2129 struct encoded_page **encoded_pages;
2130} release_pages_arg __attribute__ ((__transparent_union__));
2131
2132void release_pages(release_pages_arg, int nr);
2133
2134/**
2135 * folios_put - Decrement the reference count on an array of folios.
2136 * @folios: The folios.
2137 *
2138 * Like folio_put(), but for a batch of folios. This is more efficient
2139 * than writing the loop yourself as it will optimise the locks which need
2140 * to be taken if the folios are freed. The folios batch is returned
2141 * empty and ready to be reused for another batch; there is no need to
2142 * reinitialise it.
2143 *
2144 * Context: May be called in process or interrupt context, but not in NMI
2145 * context. May be called while holding a spinlock.
2146 */
2147static inline void folios_put(struct folio_batch *folios)
2148{
2149 folios_put_refs(folios, NULL);
2150}
2151
2152static inline void put_page(struct page *page)
2153{
2154 struct folio *folio = page_folio(page);
2155
2156 if (folio_test_slab(folio) || folio_test_large_kmalloc(folio))
2157 return;
2158
2159 folio_put(folio);
2160}
2161
2162/*
2163 * GUP_PIN_COUNTING_BIAS, and the associated functions that use it, overload
2164 * the page's refcount so that two separate items are tracked: the original page
2165 * reference count, and also a new count of how many pin_user_pages() calls were
2166 * made against the page. ("gup-pinned" is another term for the latter).
2167 *
2168 * With this scheme, pin_user_pages() becomes special: such pages are marked as
2169 * distinct from normal pages. As such, the unpin_user_page() call (and its
2170 * variants) must be used in order to release gup-pinned pages.
2171 *
2172 * Choice of value:
2173 *
2174 * By making GUP_PIN_COUNTING_BIAS a power of two, debugging of page reference
2175 * counts with respect to pin_user_pages() and unpin_user_page() becomes
2176 * simpler, due to the fact that adding an even power of two to the page
2177 * refcount has the effect of using only the upper N bits, for the code that
2178 * counts up using the bias value. This means that the lower bits are left for
2179 * the exclusive use of the original code that increments and decrements by one
2180 * (or at least, by much smaller values than the bias value).
2181 *
2182 * Of course, once the lower bits overflow into the upper bits (and this is
2183 * OK, because subtraction recovers the original values), then visual inspection
2184 * no longer suffices to directly view the separate counts. However, for normal
2185 * applications that don't have huge page reference counts, this won't be an
2186 * issue.
2187 *
2188 * Locking: the lockless algorithm described in folio_try_get_rcu()
2189 * provides safe operation for get_user_pages(), folio_mkclean() and
2190 * other calls that race to set up page table entries.
2191 */
2192#define GUP_PIN_COUNTING_BIAS (1U << 10)
2193
2194void unpin_user_page(struct page *page);
2195void unpin_folio(struct folio *folio);
2196void unpin_user_pages_dirty_lock(struct page **pages, unsigned long npages,
2197 bool make_dirty);
2198void unpin_user_page_range_dirty_lock(struct page *page, unsigned long npages,
2199 bool make_dirty);
2200void unpin_user_pages(struct page **pages, unsigned long npages);
2201void unpin_user_folio(struct folio *folio, unsigned long npages);
2202void unpin_folios(struct folio **folios, unsigned long nfolios);
2203
2204static inline bool is_cow_mapping(vm_flags_t flags)
2205{
2206 return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
2207}
2208
2209static inline bool vma_desc_is_cow_mapping(struct vm_area_desc *desc)
2210{
2211 const vma_flags_t *flags = &desc->vma_flags;
2212
2213 return vma_flags_test(flags, VMA_MAYWRITE_BIT) &&
2214 !vma_flags_test(flags, VMA_SHARED_BIT);
2215}
2216
2217#ifndef CONFIG_MMU
2218static inline bool is_nommu_shared_mapping(vm_flags_t flags)
2219{
2220 /*
2221 * NOMMU shared mappings are ordinary MAP_SHARED mappings and selected
2222 * R/O MAP_PRIVATE file mappings that are an effective R/O overlay of
2223 * a file mapping. R/O MAP_PRIVATE mappings might still modify
2224 * underlying memory if ptrace is active, so this is only possible if
2225 * ptrace does not apply. Note that there is no mprotect() to upgrade
2226 * write permissions later.
2227 */
2228 return flags & (VM_MAYSHARE | VM_MAYOVERLAY);
2229}
2230
2231static inline bool is_nommu_shared_vma_flags(const vma_flags_t *flags)
2232{
2233 return vma_flags_test_any(flags, VMA_MAYSHARE_BIT, VMA_MAYOVERLAY_BIT);
2234}
2235#endif
2236
2237#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
2238#define SECTION_IN_PAGE_FLAGS
2239#endif
2240
2241/*
2242 * The identification function is mainly used by the buddy allocator for
2243 * determining if two pages could be buddies. We are not really identifying
2244 * the zone since we could be using the section number id if we do not have
2245 * node id available in page flags.
2246 * We only guarantee that it will return the same value for two combinable
2247 * pages in a zone.
2248 */
2249static inline int page_zone_id(struct page *page)
2250{
2251 return (page->flags.f >> ZONEID_PGSHIFT) & ZONEID_MASK;
2252}
2253
2254#ifdef NODE_NOT_IN_PAGE_FLAGS
2255int memdesc_nid(memdesc_flags_t mdf);
2256#else
2257static inline int memdesc_nid(memdesc_flags_t mdf)
2258{
2259 return (mdf.f >> NODES_PGSHIFT) & NODES_MASK;
2260}
2261#endif
2262
2263static inline int page_to_nid(const struct page *page)
2264{
2265 return memdesc_nid(PF_POISONED_CHECK(page)->flags);
2266}
2267
2268static inline int folio_nid(const struct folio *folio)
2269{
2270 return memdesc_nid(folio->flags);
2271}
2272
2273#ifdef CONFIG_NUMA_BALANCING
2274/* page access time bits needs to hold at least 4 seconds */
2275#define PAGE_ACCESS_TIME_MIN_BITS 12
2276#if LAST_CPUPID_SHIFT < PAGE_ACCESS_TIME_MIN_BITS
2277#define PAGE_ACCESS_TIME_BUCKETS \
2278 (PAGE_ACCESS_TIME_MIN_BITS - LAST_CPUPID_SHIFT)
2279#else
2280#define PAGE_ACCESS_TIME_BUCKETS 0
2281#endif
2282
2283#define PAGE_ACCESS_TIME_MASK \
2284 (LAST_CPUPID_MASK << PAGE_ACCESS_TIME_BUCKETS)
2285
2286static inline int cpu_pid_to_cpupid(int cpu, int pid)
2287{
2288 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
2289}
2290
2291static inline int cpupid_to_pid(int cpupid)
2292{
2293 return cpupid & LAST__PID_MASK;
2294}
2295
2296static inline int cpupid_to_cpu(int cpupid)
2297{
2298 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
2299}
2300
2301static inline int cpupid_to_nid(int cpupid)
2302{
2303 return cpu_to_node(cpupid_to_cpu(cpupid));
2304}
2305
2306static inline bool cpupid_pid_unset(int cpupid)
2307{
2308 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
2309}
2310
2311static inline bool cpupid_cpu_unset(int cpupid)
2312{
2313 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
2314}
2315
2316static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
2317{
2318 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
2319}
2320
2321#define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
2322#ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
2323static inline int folio_xchg_last_cpupid(struct folio *folio, int cpupid)
2324{
2325 return xchg(&folio->_last_cpupid, cpupid & LAST_CPUPID_MASK);
2326}
2327
2328static inline int folio_last_cpupid(struct folio *folio)
2329{
2330 return folio->_last_cpupid;
2331}
2332static inline void page_cpupid_reset_last(struct page *page)
2333{
2334 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
2335}
2336#else
2337static inline int folio_last_cpupid(struct folio *folio)
2338{
2339 return (folio->flags.f >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
2340}
2341
2342int folio_xchg_last_cpupid(struct folio *folio, int cpupid);
2343
2344static inline void page_cpupid_reset_last(struct page *page)
2345{
2346 page->flags.f |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
2347}
2348#endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
2349
2350static inline int folio_xchg_access_time(struct folio *folio, int time)
2351{
2352 int last_time;
2353
2354 last_time = folio_xchg_last_cpupid(folio,
2355 time >> PAGE_ACCESS_TIME_BUCKETS);
2356 return last_time << PAGE_ACCESS_TIME_BUCKETS;
2357}
2358
2359static inline void vma_set_access_pid_bit(struct vm_area_struct *vma)
2360{
2361 unsigned int pid_bit;
2362
2363 pid_bit = hash_32(current->pid, ilog2(BITS_PER_LONG));
2364 if (vma->numab_state && !test_bit(pid_bit, &vma->numab_state->pids_active[1])) {
2365 __set_bit(pid_bit, &vma->numab_state->pids_active[1]);
2366 }
2367}
2368
2369bool folio_use_access_time(struct folio *folio);
2370#else /* !CONFIG_NUMA_BALANCING */
2371static inline int folio_xchg_last_cpupid(struct folio *folio, int cpupid)
2372{
2373 return folio_nid(folio); /* XXX */
2374}
2375
2376static inline int folio_xchg_access_time(struct folio *folio, int time)
2377{
2378 return 0;
2379}
2380
2381static inline int folio_last_cpupid(struct folio *folio)
2382{
2383 return folio_nid(folio); /* XXX */
2384}
2385
2386static inline int cpupid_to_nid(int cpupid)
2387{
2388 return -1;
2389}
2390
2391static inline int cpupid_to_pid(int cpupid)
2392{
2393 return -1;
2394}
2395
2396static inline int cpupid_to_cpu(int cpupid)
2397{
2398 return -1;
2399}
2400
2401static inline int cpu_pid_to_cpupid(int nid, int pid)
2402{
2403 return -1;
2404}
2405
2406static inline bool cpupid_pid_unset(int cpupid)
2407{
2408 return true;
2409}
2410
2411static inline void page_cpupid_reset_last(struct page *page)
2412{
2413}
2414
2415static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
2416{
2417 return false;
2418}
2419
2420static inline void vma_set_access_pid_bit(struct vm_area_struct *vma)
2421{
2422}
2423static inline bool folio_use_access_time(struct folio *folio)
2424{
2425 return false;
2426}
2427#endif /* CONFIG_NUMA_BALANCING */
2428
2429#if defined(CONFIG_KASAN_SW_TAGS) || defined(CONFIG_KASAN_HW_TAGS)
2430
2431/*
2432 * KASAN per-page tags are stored xor'ed with 0xff. This allows to avoid
2433 * setting tags for all pages to native kernel tag value 0xff, as the default
2434 * value 0x00 maps to 0xff.
2435 */
2436
2437static inline u8 page_kasan_tag(const struct page *page)
2438{
2439 u8 tag = KASAN_TAG_KERNEL;
2440
2441 if (kasan_enabled()) {
2442 tag = (page->flags.f >> KASAN_TAG_PGSHIFT) & KASAN_TAG_MASK;
2443 tag ^= 0xff;
2444 }
2445
2446 return tag;
2447}
2448
2449static inline void page_kasan_tag_set(struct page *page, u8 tag)
2450{
2451 unsigned long old_flags, flags;
2452
2453 if (!kasan_enabled())
2454 return;
2455
2456 tag ^= 0xff;
2457 old_flags = READ_ONCE(page->flags.f);
2458 do {
2459 flags = old_flags;
2460 flags &= ~(KASAN_TAG_MASK << KASAN_TAG_PGSHIFT);
2461 flags |= (tag & KASAN_TAG_MASK) << KASAN_TAG_PGSHIFT;
2462 } while (unlikely(!try_cmpxchg(&page->flags.f, &old_flags, flags)));
2463}
2464
2465static inline void page_kasan_tag_reset(struct page *page)
2466{
2467 if (kasan_enabled())
2468 page_kasan_tag_set(page, KASAN_TAG_KERNEL);
2469}
2470
2471#else /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
2472
2473static inline u8 page_kasan_tag(const struct page *page)
2474{
2475 return 0xff;
2476}
2477
2478static inline void page_kasan_tag_set(struct page *page, u8 tag) { }
2479static inline void page_kasan_tag_reset(struct page *page) { }
2480
2481#endif /* CONFIG_KASAN_SW_TAGS || CONFIG_KASAN_HW_TAGS */
2482
2483static inline struct zone *page_zone(const struct page *page)
2484{
2485 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
2486}
2487
2488static inline pg_data_t *page_pgdat(const struct page *page)
2489{
2490 return NODE_DATA(page_to_nid(page));
2491}
2492
2493static inline pg_data_t *folio_pgdat(const struct folio *folio)
2494{
2495 return NODE_DATA(folio_nid(folio));
2496}
2497
2498static inline struct zone *folio_zone(const struct folio *folio)
2499{
2500 return &folio_pgdat(folio)->node_zones[folio_zonenum(folio)];
2501}
2502
2503#ifdef SECTION_IN_PAGE_FLAGS
2504static inline void set_page_section(struct page *page, unsigned long section)
2505{
2506 page->flags.f &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
2507 page->flags.f |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
2508}
2509
2510static inline unsigned long memdesc_section(memdesc_flags_t mdf)
2511{
2512 return (mdf.f >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
2513}
2514#else /* !SECTION_IN_PAGE_FLAGS */
2515static inline unsigned long memdesc_section(memdesc_flags_t mdf)
2516{
2517 return 0;
2518}
2519#endif /* SECTION_IN_PAGE_FLAGS */
2520
2521/**
2522 * folio_pfn - Return the Page Frame Number of a folio.
2523 * @folio: The folio.
2524 *
2525 * A folio may contain multiple pages. The pages have consecutive
2526 * Page Frame Numbers.
2527 *
2528 * Return: The Page Frame Number of the first page in the folio.
2529 */
2530static inline unsigned long folio_pfn(const struct folio *folio)
2531{
2532 return page_to_pfn(&folio->page);
2533}
2534
2535static inline struct folio *pfn_folio(unsigned long pfn)
2536{
2537 return page_folio(pfn_to_page(pfn));
2538}
2539
2540#ifdef CONFIG_MMU
2541static inline pte_t mk_pte(const struct page *page, pgprot_t pgprot)
2542{
2543 return pfn_pte(page_to_pfn(page), pgprot);
2544}
2545
2546/**
2547 * folio_mk_pte - Create a PTE for this folio
2548 * @folio: The folio to create a PTE for
2549 * @pgprot: The page protection bits to use
2550 *
2551 * Create a page table entry for the first page of this folio.
2552 * This is suitable for passing to set_ptes().
2553 *
2554 * Return: A page table entry suitable for mapping this folio.
2555 */
2556static inline pte_t folio_mk_pte(const struct folio *folio, pgprot_t pgprot)
2557{
2558 return pfn_pte(folio_pfn(folio), pgprot);
2559}
2560
2561#ifdef CONFIG_TRANSPARENT_HUGEPAGE
2562/**
2563 * folio_mk_pmd - Create a PMD for this folio
2564 * @folio: The folio to create a PMD for
2565 * @pgprot: The page protection bits to use
2566 *
2567 * Create a page table entry for the first page of this folio.
2568 * This is suitable for passing to set_pmd_at().
2569 *
2570 * Return: A page table entry suitable for mapping this folio.
2571 */
2572static inline pmd_t folio_mk_pmd(const struct folio *folio, pgprot_t pgprot)
2573{
2574 return pmd_mkhuge(pfn_pmd(folio_pfn(folio), pgprot));
2575}
2576
2577#ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
2578/**
2579 * folio_mk_pud - Create a PUD for this folio
2580 * @folio: The folio to create a PUD for
2581 * @pgprot: The page protection bits to use
2582 *
2583 * Create a page table entry for the first page of this folio.
2584 * This is suitable for passing to set_pud_at().
2585 *
2586 * Return: A page table entry suitable for mapping this folio.
2587 */
2588static inline pud_t folio_mk_pud(const struct folio *folio, pgprot_t pgprot)
2589{
2590 return pud_mkhuge(pfn_pud(folio_pfn(folio), pgprot));
2591}
2592#endif /* CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD */
2593#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
2594#endif /* CONFIG_MMU */
2595
2596static inline bool folio_has_pincount(const struct folio *folio)
2597{
2598 if (IS_ENABLED(CONFIG_64BIT))
2599 return folio_test_large(folio);
2600 return folio_order(folio) > 1;
2601}
2602
2603/**
2604 * folio_maybe_dma_pinned - Report if a folio may be pinned for DMA.
2605 * @folio: The folio.
2606 *
2607 * This function checks if a folio has been pinned via a call to
2608 * a function in the pin_user_pages() family.
2609 *
2610 * For small folios, the return value is partially fuzzy: false is not fuzzy,
2611 * because it means "definitely not pinned for DMA", but true means "probably
2612 * pinned for DMA, but possibly a false positive due to having at least
2613 * GUP_PIN_COUNTING_BIAS worth of normal folio references".
2614 *
2615 * False positives are OK, because: a) it's unlikely for a folio to
2616 * get that many refcounts, and b) all the callers of this routine are
2617 * expected to be able to deal gracefully with a false positive.
2618 *
2619 * For most large folios, the result will be exactly correct. That's because
2620 * we have more tracking data available: the _pincount field is used
2621 * instead of the GUP_PIN_COUNTING_BIAS scheme.
2622 *
2623 * For more information, please see Documentation/core-api/pin_user_pages.rst.
2624 *
2625 * Return: True, if it is likely that the folio has been "dma-pinned".
2626 * False, if the folio is definitely not dma-pinned.
2627 */
2628static inline bool folio_maybe_dma_pinned(struct folio *folio)
2629{
2630 if (folio_has_pincount(folio))
2631 return atomic_read(&folio->_pincount) > 0;
2632
2633 /*
2634 * folio_ref_count() is signed. If that refcount overflows, then
2635 * folio_ref_count() returns a negative value, and callers will avoid
2636 * further incrementing the refcount.
2637 *
2638 * Here, for that overflow case, use the sign bit to count a little
2639 * bit higher via unsigned math, and thus still get an accurate result.
2640 */
2641 return ((unsigned int)folio_ref_count(folio)) >=
2642 GUP_PIN_COUNTING_BIAS;
2643}
2644
2645/*
2646 * This should most likely only be called during fork() to see whether we
2647 * should break the cow immediately for an anon page on the src mm.
2648 *
2649 * The caller has to hold the PT lock and the vma->vm_mm->->write_protect_seq.
2650 */
2651static inline bool folio_needs_cow_for_dma(struct vm_area_struct *vma,
2652 struct folio *folio)
2653{
2654 VM_BUG_ON(!(raw_read_seqcount(&vma->vm_mm->write_protect_seq) & 1));
2655
2656 if (!mm_flags_test(MMF_HAS_PINNED, vma->vm_mm))
2657 return false;
2658
2659 return folio_maybe_dma_pinned(folio);
2660}
2661
2662/**
2663 * is_zero_page - Query if a page is a zero page
2664 * @page: The page to query
2665 *
2666 * This returns true if @page is one of the permanent zero pages.
2667 */
2668static inline bool is_zero_page(const struct page *page)
2669{
2670 return is_zero_pfn(page_to_pfn(page));
2671}
2672
2673/**
2674 * is_zero_folio - Query if a folio is a zero page
2675 * @folio: The folio to query
2676 *
2677 * This returns true if @folio is one of the permanent zero pages.
2678 */
2679static inline bool is_zero_folio(const struct folio *folio)
2680{
2681 return is_zero_page(&folio->page);
2682}
2683
2684/* MIGRATE_CMA and ZONE_MOVABLE do not allow pin folios */
2685#ifdef CONFIG_MIGRATION
2686static inline bool folio_is_longterm_pinnable(struct folio *folio)
2687{
2688#ifdef CONFIG_CMA
2689 int mt = folio_migratetype(folio);
2690
2691 if (mt == MIGRATE_CMA || mt == MIGRATE_ISOLATE)
2692 return false;
2693#endif
2694 /* The zero page can be "pinned" but gets special handling. */
2695 if (is_zero_folio(folio))
2696 return true;
2697
2698 /* Coherent device memory must always allow eviction. */
2699 if (folio_is_device_coherent(folio))
2700 return false;
2701
2702 /*
2703 * Filesystems can only tolerate transient delays to truncate and
2704 * hole-punch operations
2705 */
2706 if (folio_is_fsdax(folio))
2707 return false;
2708
2709 /* Otherwise, non-movable zone folios can be pinned. */
2710 return !folio_is_zone_movable(folio);
2711
2712}
2713#else
2714static inline bool folio_is_longterm_pinnable(struct folio *folio)
2715{
2716 return true;
2717}
2718#endif
2719
2720static inline void set_page_zone(struct page *page, enum zone_type zone)
2721{
2722 page->flags.f &= ~(ZONES_MASK << ZONES_PGSHIFT);
2723 page->flags.f |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
2724}
2725
2726static inline void set_page_node(struct page *page, unsigned long node)
2727{
2728 page->flags.f &= ~(NODES_MASK << NODES_PGSHIFT);
2729 page->flags.f |= (node & NODES_MASK) << NODES_PGSHIFT;
2730}
2731
2732static inline void set_page_links(struct page *page, enum zone_type zone,
2733 unsigned long node, unsigned long pfn)
2734{
2735 set_page_zone(page, zone);
2736 set_page_node(page, node);
2737#ifdef SECTION_IN_PAGE_FLAGS
2738 set_page_section(page, pfn_to_section_nr(pfn));
2739#endif
2740}
2741
2742/**
2743 * folio_nr_pages - The number of pages in the folio.
2744 * @folio: The folio.
2745 *
2746 * Return: A positive power of two.
2747 */
2748static inline unsigned long folio_nr_pages(const struct folio *folio)
2749{
2750 if (!folio_test_large(folio))
2751 return 1;
2752 return folio_large_nr_pages(folio);
2753}
2754
2755/*
2756 * compound_nr() returns the number of pages in this potentially compound
2757 * page. compound_nr() can be called on a tail page, and is defined to
2758 * return 1 in that case.
2759 */
2760static inline unsigned long compound_nr(const struct page *page)
2761{
2762 const struct folio *folio = (struct folio *)page;
2763
2764 if (!test_bit(PG_head, &folio->flags.f))
2765 return 1;
2766 return folio_large_nr_pages(folio);
2767}
2768
2769/**
2770 * folio_next - Move to the next physical folio.
2771 * @folio: The folio we're currently operating on.
2772 *
2773 * If you have physically contiguous memory which may span more than
2774 * one folio (eg a &struct bio_vec), use this function to move from one
2775 * folio to the next. Do not use it if the memory is only virtually
2776 * contiguous as the folios are almost certainly not adjacent to each
2777 * other. This is the folio equivalent to writing ``page++``.
2778 *
2779 * Context: We assume that the folios are refcounted and/or locked at a
2780 * higher level and do not adjust the reference counts.
2781 * Return: The next struct folio.
2782 */
2783static inline struct folio *folio_next(struct folio *folio)
2784{
2785 return (struct folio *)folio_page(folio, folio_nr_pages(folio));
2786}
2787
2788/**
2789 * folio_shift - The size of the memory described by this folio.
2790 * @folio: The folio.
2791 *
2792 * A folio represents a number of bytes which is a power-of-two in size.
2793 * This function tells you which power-of-two the folio is. See also
2794 * folio_size() and folio_order().
2795 *
2796 * Context: The caller should have a reference on the folio to prevent
2797 * it from being split. It is not necessary for the folio to be locked.
2798 * Return: The base-2 logarithm of the size of this folio.
2799 */
2800static inline unsigned int folio_shift(const struct folio *folio)
2801{
2802 return PAGE_SHIFT + folio_order(folio);
2803}
2804
2805/**
2806 * folio_size - The number of bytes in a folio.
2807 * @folio: The folio.
2808 *
2809 * Context: The caller should have a reference on the folio to prevent
2810 * it from being split. It is not necessary for the folio to be locked.
2811 * Return: The number of bytes in this folio.
2812 */
2813static inline size_t folio_size(const struct folio *folio)
2814{
2815 return PAGE_SIZE << folio_order(folio);
2816}
2817
2818/**
2819 * folio_maybe_mapped_shared - Whether the folio is mapped into the page
2820 * tables of more than one MM
2821 * @folio: The folio.
2822 *
2823 * This function checks if the folio maybe currently mapped into more than one
2824 * MM ("maybe mapped shared"), or if the folio is certainly mapped into a single
2825 * MM ("mapped exclusively").
2826 *
2827 * For KSM folios, this function also returns "mapped shared" when a folio is
2828 * mapped multiple times into the same MM, because the individual page mappings
2829 * are independent.
2830 *
2831 * For small anonymous folios and anonymous hugetlb folios, the return
2832 * value will be exactly correct: non-KSM folios can only be mapped at most once
2833 * into an MM, and they cannot be partially mapped. KSM folios are
2834 * considered shared even if mapped multiple times into the same MM.
2835 *
2836 * For other folios, the result can be fuzzy:
2837 * #. For partially-mappable large folios (THP), the return value can wrongly
2838 * indicate "mapped shared" (false positive) if a folio was mapped by
2839 * more than two MMs at one point in time.
2840 * #. For pagecache folios (including hugetlb), the return value can wrongly
2841 * indicate "mapped shared" (false positive) when two VMAs in the same MM
2842 * cover the same file range.
2843 *
2844 * Further, this function only considers current page table mappings that
2845 * are tracked using the folio mapcount(s).
2846 *
2847 * This function does not consider:
2848 * #. If the folio might get mapped in the (near) future (e.g., swapcache,
2849 * pagecache, temporary unmapping for migration).
2850 * #. If the folio is mapped differently (VM_PFNMAP).
2851 * #. If hugetlb page table sharing applies. Callers might want to check
2852 * hugetlb_pmd_shared().
2853 *
2854 * Return: Whether the folio is estimated to be mapped into more than one MM.
2855 */
2856static inline bool folio_maybe_mapped_shared(struct folio *folio)
2857{
2858 int mapcount = folio_mapcount(folio);
2859
2860 /* Only partially-mappable folios require more care. */
2861 if (!folio_test_large(folio) || unlikely(folio_test_hugetlb(folio)))
2862 return mapcount > 1;
2863
2864 /*
2865 * vm_insert_page() without CONFIG_TRANSPARENT_HUGEPAGE ...
2866 * simply assume "mapped shared", nobody should really care
2867 * about this for arbitrary kernel allocations.
2868 */
2869 if (!IS_ENABLED(CONFIG_MM_ID))
2870 return true;
2871
2872 /*
2873 * A single mapping implies "mapped exclusively", even if the
2874 * folio flag says something different: it's easier to handle this
2875 * case here instead of on the RMAP hot path.
2876 */
2877 if (mapcount <= 1)
2878 return false;
2879 return test_bit(FOLIO_MM_IDS_SHARED_BITNUM, &folio->_mm_ids);
2880}
2881
2882/**
2883 * folio_expected_ref_count - calculate the expected folio refcount
2884 * @folio: the folio
2885 *
2886 * Calculate the expected folio refcount, taking references from the pagecache,
2887 * swapcache, PG_private and page table mappings into account. Useful in
2888 * combination with folio_ref_count() to detect unexpected references (e.g.,
2889 * GUP or other temporary references).
2890 *
2891 * Does currently not consider references from the LRU cache. If the folio
2892 * was isolated from the LRU (which is the case during migration or split),
2893 * the LRU cache does not apply.
2894 *
2895 * Calling this function on an unmapped folio -- !folio_mapped() -- that is
2896 * locked will return a stable result.
2897 *
2898 * Calling this function on a mapped folio will not result in a stable result,
2899 * because nothing stops additional page table mappings from coming (e.g.,
2900 * fork()) or going (e.g., munmap()).
2901 *
2902 * Calling this function without the folio lock will also not result in a
2903 * stable result: for example, the folio might get dropped from the swapcache
2904 * concurrently.
2905 *
2906 * However, even when called without the folio lock or on a mapped folio,
2907 * this function can be used to detect unexpected references early (for example,
2908 * if it makes sense to even lock the folio and unmap it).
2909 *
2910 * The caller must add any reference (e.g., from folio_try_get()) it might be
2911 * holding itself to the result.
2912 *
2913 * Returns: the expected folio refcount.
2914 */
2915static inline int folio_expected_ref_count(const struct folio *folio)
2916{
2917 const int order = folio_order(folio);
2918 int ref_count = 0;
2919
2920 if (WARN_ON_ONCE(page_has_type(&folio->page) && !folio_test_hugetlb(folio)))
2921 return 0;
2922
2923 /* One reference per page from the swapcache. */
2924 ref_count += folio_test_swapcache(folio) << order;
2925
2926 if (!folio_test_anon(folio)) {
2927 /* One reference per page from the pagecache. */
2928 ref_count += !!folio->mapping << order;
2929 /* One reference from PG_private. */
2930 ref_count += folio_test_private(folio);
2931 }
2932
2933 /* One reference per page table mapping. */
2934 return ref_count + folio_mapcount(folio);
2935}
2936
2937#ifndef HAVE_ARCH_MAKE_FOLIO_ACCESSIBLE
2938static inline int arch_make_folio_accessible(struct folio *folio)
2939{
2940 return 0;
2941}
2942#endif
2943
2944/*
2945 * Some inline functions in vmstat.h depend on page_zone()
2946 */
2947#include <linux/vmstat.h>
2948
2949#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
2950#define HASHED_PAGE_VIRTUAL
2951#endif
2952
2953#if defined(WANT_PAGE_VIRTUAL)
2954static inline void *page_address(const struct page *page)
2955{
2956 return page->virtual;
2957}
2958static inline void set_page_address(struct page *page, void *address)
2959{
2960 page->virtual = address;
2961}
2962#define page_address_init() do { } while(0)
2963#endif
2964
2965#if defined(HASHED_PAGE_VIRTUAL)
2966void *page_address(const struct page *page);
2967void set_page_address(struct page *page, void *virtual);
2968void page_address_init(void);
2969#endif
2970
2971static __always_inline void *lowmem_page_address(const struct page *page)
2972{
2973 return page_to_virt(page);
2974}
2975
2976#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
2977#define page_address(page) lowmem_page_address(page)
2978#define set_page_address(page, address) do { } while(0)
2979#define page_address_init() do { } while(0)
2980#endif
2981
2982static inline void *folio_address(const struct folio *folio)
2983{
2984 return page_address(&folio->page);
2985}
2986
2987/*
2988 * Return true only if the page has been allocated with
2989 * ALLOC_NO_WATERMARKS and the low watermark was not
2990 * met implying that the system is under some pressure.
2991 */
2992static inline bool page_is_pfmemalloc(const struct page *page)
2993{
2994 /*
2995 * lru.next has bit 1 set if the page is allocated from the
2996 * pfmemalloc reserves. Callers may simply overwrite it if
2997 * they do not need to preserve that information.
2998 */
2999 return (uintptr_t)page->lru.next & BIT(1);
3000}
3001
3002/*
3003 * Return true only if the folio has been allocated with
3004 * ALLOC_NO_WATERMARKS and the low watermark was not
3005 * met implying that the system is under some pressure.
3006 */
3007static inline bool folio_is_pfmemalloc(const struct folio *folio)
3008{
3009 /*
3010 * lru.next has bit 1 set if the page is allocated from the
3011 * pfmemalloc reserves. Callers may simply overwrite it if
3012 * they do not need to preserve that information.
3013 */
3014 return (uintptr_t)folio->lru.next & BIT(1);
3015}
3016
3017/*
3018 * Only to be called by the page allocator on a freshly allocated
3019 * page.
3020 */
3021static inline void set_page_pfmemalloc(struct page *page)
3022{
3023 page->lru.next = (void *)BIT(1);
3024}
3025
3026static inline void clear_page_pfmemalloc(struct page *page)
3027{
3028 page->lru.next = NULL;
3029}
3030
3031/*
3032 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
3033 */
3034extern void pagefault_out_of_memory(void);
3035
3036#define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
3037#define offset_in_folio(folio, p) ((unsigned long)(p) & (folio_size(folio) - 1))
3038
3039/*
3040 * Parameter block passed down to zap_pte_range in exceptional cases.
3041 */
3042struct zap_details {
3043 struct folio *single_folio; /* Locked folio to be unmapped */
3044 bool skip_cows; /* Do not zap COWed private pages */
3045 bool reclaim_pt; /* Need reclaim page tables? */
3046 bool reaping; /* Reaping, do not block. */
3047 zap_flags_t zap_flags; /* Extra flags for zapping */
3048};
3049
3050/*
3051 * Whether to drop the pte markers, for example, the uffd-wp information for
3052 * file-backed memory. This should only be specified when we will completely
3053 * drop the page in the mm, either by truncation or unmapping of the vma. By
3054 * default, the flag is not set.
3055 */
3056#define ZAP_FLAG_DROP_MARKER ((__force zap_flags_t) BIT(0))
3057/* Set in unmap_vmas() to indicate a final unmap call. Only used by hugetlb */
3058#define ZAP_FLAG_UNMAP ((__force zap_flags_t) BIT(1))
3059
3060#ifdef CONFIG_MMU
3061extern bool can_do_mlock(void);
3062#else
3063static inline bool can_do_mlock(void) { return false; }
3064#endif
3065extern int user_shm_lock(size_t, struct ucounts *);
3066extern void user_shm_unlock(size_t, struct ucounts *);
3067
3068struct folio *vm_normal_folio(struct vm_area_struct *vma, unsigned long addr,
3069 pte_t pte);
3070struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
3071 pte_t pte);
3072struct folio *vm_normal_folio_pmd(struct vm_area_struct *vma,
3073 unsigned long addr, pmd_t pmd);
3074struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
3075 pmd_t pmd);
3076struct page *vm_normal_page_pud(struct vm_area_struct *vma, unsigned long addr,
3077 pud_t pud);
3078
3079void zap_special_vma_range(struct vm_area_struct *vma, unsigned long address,
3080 unsigned long size);
3081void zap_vma_range(struct vm_area_struct *vma, unsigned long address,
3082 unsigned long size);
3083/**
3084 * zap_vma - zap all page table entries in a vma
3085 * @vma: The vma to zap.
3086 */
3087static inline void zap_vma(struct vm_area_struct *vma)
3088{
3089 zap_vma_range(vma, vma->vm_start, vma->vm_end - vma->vm_start);
3090}
3091struct mmu_notifier_range;
3092
3093void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
3094 unsigned long end, unsigned long floor, unsigned long ceiling);
3095int
3096copy_page_range(struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma);
3097int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
3098 void *buf, int len, int write);
3099
3100struct follow_pfnmap_args {
3101 /**
3102 * Inputs:
3103 * @vma: Pointer to @vm_area_struct struct
3104 * @address: the virtual address to walk
3105 */
3106 struct vm_area_struct *vma;
3107 unsigned long address;
3108 /**
3109 * Internals:
3110 *
3111 * The caller shouldn't touch any of these.
3112 */
3113 spinlock_t *lock;
3114 pte_t *ptep;
3115 /**
3116 * Outputs:
3117 *
3118 * @pfn: the PFN of the address
3119 * @addr_mask: address mask covering pfn
3120 * @pgprot: the pgprot_t of the mapping
3121 * @writable: whether the mapping is writable
3122 * @special: whether the mapping is a special mapping (real PFN maps)
3123 */
3124 unsigned long pfn;
3125 unsigned long addr_mask;
3126 pgprot_t pgprot;
3127 bool writable;
3128 bool special;
3129};
3130int follow_pfnmap_start(struct follow_pfnmap_args *args);
3131void follow_pfnmap_end(struct follow_pfnmap_args *args);
3132
3133extern void truncate_pagecache(struct inode *inode, loff_t new);
3134extern void truncate_setsize(struct inode *inode, loff_t newsize);
3135void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
3136void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
3137int generic_error_remove_folio(struct address_space *mapping,
3138 struct folio *folio);
3139
3140struct vm_area_struct *lock_mm_and_find_vma(struct mm_struct *mm,
3141 unsigned long address, struct pt_regs *regs);
3142
3143#ifdef CONFIG_MMU
3144extern vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
3145 unsigned long address, unsigned int flags,
3146 struct pt_regs *regs);
3147extern int fixup_user_fault(struct mm_struct *mm,
3148 unsigned long address, unsigned int fault_flags,
3149 bool *unlocked);
3150void unmap_mapping_pages(struct address_space *mapping,
3151 pgoff_t start, pgoff_t nr, bool even_cows);
3152void unmap_mapping_range(struct address_space *mapping,
3153 loff_t const holebegin, loff_t const holelen, int even_cows);
3154#else
3155static inline vm_fault_t handle_mm_fault(struct vm_area_struct *vma,
3156 unsigned long address, unsigned int flags,
3157 struct pt_regs *regs)
3158{
3159 /* should never happen if there's no MMU */
3160 BUG();
3161 return VM_FAULT_SIGBUS;
3162}
3163static inline int fixup_user_fault(struct mm_struct *mm, unsigned long address,
3164 unsigned int fault_flags, bool *unlocked)
3165{
3166 /* should never happen if there's no MMU */
3167 BUG();
3168 return -EFAULT;
3169}
3170static inline void unmap_mapping_pages(struct address_space *mapping,
3171 pgoff_t start, pgoff_t nr, bool even_cows) { }
3172static inline void unmap_mapping_range(struct address_space *mapping,
3173 loff_t const holebegin, loff_t const holelen, int even_cows) { }
3174#endif
3175
3176static inline void unmap_shared_mapping_range(struct address_space *mapping,
3177 loff_t const holebegin, loff_t const holelen)
3178{
3179 unmap_mapping_range(mapping, holebegin, holelen, 0);
3180}
3181
3182static inline struct vm_area_struct *vma_lookup(struct mm_struct *mm,
3183 unsigned long addr);
3184
3185extern int access_process_vm(struct task_struct *tsk, unsigned long addr,
3186 void *buf, int len, unsigned int gup_flags);
3187extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
3188 void *buf, int len, unsigned int gup_flags);
3189
3190#ifdef CONFIG_BPF_SYSCALL
3191extern int copy_remote_vm_str(struct task_struct *tsk, unsigned long addr,
3192 void *buf, int len, unsigned int gup_flags);
3193#endif
3194
3195long get_user_pages_remote(struct mm_struct *mm,
3196 unsigned long start, unsigned long nr_pages,
3197 unsigned int gup_flags, struct page **pages,
3198 int *locked);
3199long pin_user_pages_remote(struct mm_struct *mm,
3200 unsigned long start, unsigned long nr_pages,
3201 unsigned int gup_flags, struct page **pages,
3202 int *locked);
3203
3204/*
3205 * Retrieves a single page alongside its VMA. Does not support FOLL_NOWAIT.
3206 */
3207static inline struct page *get_user_page_vma_remote(struct mm_struct *mm,
3208 unsigned long addr,
3209 int gup_flags,
3210 struct vm_area_struct **vmap)
3211{
3212 struct page *page;
3213 struct vm_area_struct *vma;
3214 int got;
3215
3216 if (WARN_ON_ONCE(unlikely(gup_flags & FOLL_NOWAIT)))
3217 return ERR_PTR(-EINVAL);
3218
3219 got = get_user_pages_remote(mm, addr, 1, gup_flags, &page, NULL);
3220
3221 if (got < 0)
3222 return ERR_PTR(got);
3223
3224 vma = vma_lookup(mm, addr);
3225 if (WARN_ON_ONCE(!vma)) {
3226 put_page(page);
3227 return ERR_PTR(-EINVAL);
3228 }
3229
3230 *vmap = vma;
3231 return page;
3232}
3233
3234long get_user_pages(unsigned long start, unsigned long nr_pages,
3235 unsigned int gup_flags, struct page **pages);
3236long pin_user_pages(unsigned long start, unsigned long nr_pages,
3237 unsigned int gup_flags, struct page **pages);
3238long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
3239 struct page **pages, unsigned int gup_flags);
3240long pin_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
3241 struct page **pages, unsigned int gup_flags);
3242long memfd_pin_folios(struct file *memfd, loff_t start, loff_t end,
3243 struct folio **folios, unsigned int max_folios,
3244 pgoff_t *offset);
3245int folio_add_pins(struct folio *folio, unsigned int pins);
3246
3247int get_user_pages_fast(unsigned long start, int nr_pages,
3248 unsigned int gup_flags, struct page **pages);
3249int pin_user_pages_fast(unsigned long start, int nr_pages,
3250 unsigned int gup_flags, struct page **pages);
3251void folio_add_pin(struct folio *folio);
3252
3253int account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc);
3254int __account_locked_vm(struct mm_struct *mm, unsigned long pages, bool inc,
3255 const struct task_struct *task, bool bypass_rlim);
3256
3257struct kvec;
3258struct page *get_dump_page(unsigned long addr, int *locked);
3259
3260bool folio_mark_dirty(struct folio *folio);
3261bool folio_mark_dirty_lock(struct folio *folio);
3262bool set_page_dirty(struct page *page);
3263int set_page_dirty_lock(struct page *page);
3264
3265int get_cmdline(struct task_struct *task, char *buffer, int buflen);
3266
3267/*
3268 * Flags used by change_protection(). For now we make it a bitmap so
3269 * that we can pass in multiple flags just like parameters. However
3270 * for now all the callers are only use one of the flags at the same
3271 * time.
3272 */
3273/*
3274 * Whether we should manually check if we can map individual PTEs writable,
3275 * because something (e.g., COW, uffd-wp) blocks that from happening for all
3276 * PTEs automatically in a writable mapping.
3277 */
3278#define MM_CP_TRY_CHANGE_WRITABLE (1UL << 0)
3279/* Whether this protection change is for NUMA hints */
3280#define MM_CP_PROT_NUMA (1UL << 1)
3281/* Whether this change is for write protecting */
3282#define MM_CP_UFFD_WP (1UL << 2) /* do wp */
3283#define MM_CP_UFFD_WP_RESOLVE (1UL << 3) /* Resolve wp */
3284#define MM_CP_UFFD_WP_ALL (MM_CP_UFFD_WP | \
3285 MM_CP_UFFD_WP_RESOLVE)
3286
3287bool can_change_pte_writable(struct vm_area_struct *vma, unsigned long addr,
3288 pte_t pte);
3289extern long change_protection(struct mmu_gather *tlb,
3290 struct vm_area_struct *vma, unsigned long start,
3291 unsigned long end, unsigned long cp_flags);
3292extern int mprotect_fixup(struct vma_iterator *vmi, struct mmu_gather *tlb,
3293 struct vm_area_struct *vma, struct vm_area_struct **pprev,
3294 unsigned long start, unsigned long end, vm_flags_t newflags);
3295
3296/*
3297 * doesn't attempt to fault and will return short.
3298 */
3299int get_user_pages_fast_only(unsigned long start, int nr_pages,
3300 unsigned int gup_flags, struct page **pages);
3301
3302static inline bool get_user_page_fast_only(unsigned long addr,
3303 unsigned int gup_flags, struct page **pagep)
3304{
3305 return get_user_pages_fast_only(addr, 1, gup_flags, pagep) == 1;
3306}
3307/*
3308 * per-process(per-mm_struct) statistics.
3309 */
3310static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
3311{
3312 return percpu_counter_read_positive(&mm->rss_stat[member]);
3313}
3314
3315static inline unsigned long get_mm_counter_sum(struct mm_struct *mm, int member)
3316{
3317 return percpu_counter_sum_positive(&mm->rss_stat[member]);
3318}
3319
3320void mm_trace_rss_stat(struct mm_struct *mm, int member);
3321
3322static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
3323{
3324 percpu_counter_add(&mm->rss_stat[member], value);
3325
3326 mm_trace_rss_stat(mm, member);
3327}
3328
3329static inline void inc_mm_counter(struct mm_struct *mm, int member)
3330{
3331 percpu_counter_inc(&mm->rss_stat[member]);
3332
3333 mm_trace_rss_stat(mm, member);
3334}
3335
3336static inline void dec_mm_counter(struct mm_struct *mm, int member)
3337{
3338 percpu_counter_dec(&mm->rss_stat[member]);
3339
3340 mm_trace_rss_stat(mm, member);
3341}
3342
3343/* Optimized variant when folio is already known not to be anon */
3344static inline int mm_counter_file(struct folio *folio)
3345{
3346 if (folio_test_swapbacked(folio))
3347 return MM_SHMEMPAGES;
3348 return MM_FILEPAGES;
3349}
3350
3351static inline int mm_counter(struct folio *folio)
3352{
3353 if (folio_test_anon(folio))
3354 return MM_ANONPAGES;
3355 return mm_counter_file(folio);
3356}
3357
3358static inline unsigned long get_mm_rss(struct mm_struct *mm)
3359{
3360 return get_mm_counter(mm, MM_FILEPAGES) +
3361 get_mm_counter(mm, MM_ANONPAGES) +
3362 get_mm_counter(mm, MM_SHMEMPAGES);
3363}
3364
3365static inline unsigned long get_mm_rss_sum(struct mm_struct *mm)
3366{
3367 return get_mm_counter_sum(mm, MM_FILEPAGES) +
3368 get_mm_counter_sum(mm, MM_ANONPAGES) +
3369 get_mm_counter_sum(mm, MM_SHMEMPAGES);
3370}
3371
3372static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
3373{
3374 return max(mm->hiwater_rss, get_mm_rss(mm));
3375}
3376
3377static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
3378{
3379 return max(mm->hiwater_vm, mm->total_vm);
3380}
3381
3382static inline void update_hiwater_rss(struct mm_struct *mm)
3383{
3384 unsigned long _rss = get_mm_rss(mm);
3385
3386 if (data_race(mm->hiwater_rss) < _rss)
3387 data_race(mm->hiwater_rss = _rss);
3388}
3389
3390static inline void update_hiwater_vm(struct mm_struct *mm)
3391{
3392 if (mm->hiwater_vm < mm->total_vm)
3393 mm->hiwater_vm = mm->total_vm;
3394}
3395
3396static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
3397{
3398 mm->hiwater_rss = get_mm_rss(mm);
3399}
3400
3401static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
3402 struct mm_struct *mm)
3403{
3404 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
3405
3406 if (*maxrss < hiwater_rss)
3407 *maxrss = hiwater_rss;
3408}
3409
3410#ifndef CONFIG_ARCH_HAS_PTE_SPECIAL
3411static inline int pte_special(pte_t pte)
3412{
3413 return 0;
3414}
3415
3416static inline pte_t pte_mkspecial(pte_t pte)
3417{
3418 return pte;
3419}
3420#endif
3421
3422#ifndef CONFIG_ARCH_SUPPORTS_PMD_PFNMAP
3423static inline bool pmd_special(pmd_t pmd)
3424{
3425 return false;
3426}
3427
3428static inline pmd_t pmd_mkspecial(pmd_t pmd)
3429{
3430 return pmd;
3431}
3432#endif /* CONFIG_ARCH_SUPPORTS_PMD_PFNMAP */
3433
3434#ifndef CONFIG_ARCH_SUPPORTS_PUD_PFNMAP
3435static inline bool pud_special(pud_t pud)
3436{
3437 return false;
3438}
3439
3440static inline pud_t pud_mkspecial(pud_t pud)
3441{
3442 return pud;
3443}
3444#endif /* CONFIG_ARCH_SUPPORTS_PUD_PFNMAP */
3445
3446extern pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
3447 spinlock_t **ptl);
3448
3449#ifdef __PAGETABLE_P4D_FOLDED
3450static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
3451 unsigned long address)
3452{
3453 return 0;
3454}
3455#else
3456int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
3457#endif
3458
3459#if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
3460static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
3461 unsigned long address)
3462{
3463 return 0;
3464}
3465static inline void mm_inc_nr_puds(struct mm_struct *mm) {}
3466static inline void mm_dec_nr_puds(struct mm_struct *mm) {}
3467
3468#else
3469int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
3470
3471static inline void mm_inc_nr_puds(struct mm_struct *mm)
3472{
3473 if (mm_pud_folded(mm))
3474 return;
3475 atomic_long_add(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
3476}
3477
3478static inline void mm_dec_nr_puds(struct mm_struct *mm)
3479{
3480 if (mm_pud_folded(mm))
3481 return;
3482 atomic_long_sub(PTRS_PER_PUD * sizeof(pud_t), &mm->pgtables_bytes);
3483}
3484#endif
3485
3486#if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
3487static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
3488 unsigned long address)
3489{
3490 return 0;
3491}
3492
3493static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
3494static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
3495
3496#else
3497int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
3498
3499static inline void mm_inc_nr_pmds(struct mm_struct *mm)
3500{
3501 if (mm_pmd_folded(mm))
3502 return;
3503 atomic_long_add(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
3504}
3505
3506static inline void mm_dec_nr_pmds(struct mm_struct *mm)
3507{
3508 if (mm_pmd_folded(mm))
3509 return;
3510 atomic_long_sub(PTRS_PER_PMD * sizeof(pmd_t), &mm->pgtables_bytes);
3511}
3512#endif
3513
3514#ifdef CONFIG_MMU
3515static inline void mm_pgtables_bytes_init(struct mm_struct *mm)
3516{
3517 atomic_long_set(&mm->pgtables_bytes, 0);
3518}
3519
3520static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
3521{
3522 return atomic_long_read(&mm->pgtables_bytes);
3523}
3524
3525static inline void mm_inc_nr_ptes(struct mm_struct *mm)
3526{
3527 atomic_long_add(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
3528}
3529
3530static inline void mm_dec_nr_ptes(struct mm_struct *mm)
3531{
3532 atomic_long_sub(PTRS_PER_PTE * sizeof(pte_t), &mm->pgtables_bytes);
3533}
3534#else
3535
3536static inline void mm_pgtables_bytes_init(struct mm_struct *mm) {}
3537static inline unsigned long mm_pgtables_bytes(const struct mm_struct *mm)
3538{
3539 return 0;
3540}
3541
3542static inline void mm_inc_nr_ptes(struct mm_struct *mm) {}
3543static inline void mm_dec_nr_ptes(struct mm_struct *mm) {}
3544#endif
3545
3546int __pte_alloc(struct mm_struct *mm, pmd_t *pmd);
3547int __pte_alloc_kernel(pmd_t *pmd);
3548
3549#if defined(CONFIG_MMU)
3550
3551static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
3552 unsigned long address)
3553{
3554 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
3555 NULL : p4d_offset(pgd, address);
3556}
3557
3558static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
3559 unsigned long address)
3560{
3561 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
3562 NULL : pud_offset(p4d, address);
3563}
3564
3565static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
3566{
3567 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
3568 NULL: pmd_offset(pud, address);
3569}
3570#endif /* CONFIG_MMU */
3571
3572enum pt_flags {
3573 PT_kernel = PG_referenced,
3574 PT_reserved = PG_reserved,
3575 /* High bits are used for zone/node/section */
3576};
3577
3578static inline struct ptdesc *virt_to_ptdesc(const void *x)
3579{
3580 return page_ptdesc(virt_to_page(x));
3581}
3582
3583/**
3584 * ptdesc_address - Virtual address of page table.
3585 * @pt: Page table descriptor.
3586 *
3587 * Return: The first byte of the page table described by @pt.
3588 */
3589static inline void *ptdesc_address(const struct ptdesc *pt)
3590{
3591 return folio_address(ptdesc_folio(pt));
3592}
3593
3594static inline bool pagetable_is_reserved(struct ptdesc *pt)
3595{
3596 return test_bit(PT_reserved, &pt->pt_flags.f);
3597}
3598
3599/**
3600 * ptdesc_set_kernel - Mark a ptdesc used to map the kernel
3601 * @ptdesc: The ptdesc to be marked
3602 *
3603 * Kernel page tables often need special handling. Set a flag so that
3604 * the handling code knows this ptdesc will not be used for userspace.
3605 */
3606static inline void ptdesc_set_kernel(struct ptdesc *ptdesc)
3607{
3608 set_bit(PT_kernel, &ptdesc->pt_flags.f);
3609}
3610
3611/**
3612 * ptdesc_clear_kernel - Mark a ptdesc as no longer used to map the kernel
3613 * @ptdesc: The ptdesc to be unmarked
3614 *
3615 * Use when the ptdesc is no longer used to map the kernel and no longer
3616 * needs special handling.
3617 */
3618static inline void ptdesc_clear_kernel(struct ptdesc *ptdesc)
3619{
3620 /*
3621 * Note: the 'PG_referenced' bit does not strictly need to be
3622 * cleared before freeing the page. But this is nice for
3623 * symmetry.
3624 */
3625 clear_bit(PT_kernel, &ptdesc->pt_flags.f);
3626}
3627
3628/**
3629 * ptdesc_test_kernel - Check if a ptdesc is used to map the kernel
3630 * @ptdesc: The ptdesc being tested
3631 *
3632 * Call to tell if the ptdesc used to map the kernel.
3633 */
3634static inline bool ptdesc_test_kernel(const struct ptdesc *ptdesc)
3635{
3636 return test_bit(PT_kernel, &ptdesc->pt_flags.f);
3637}
3638
3639/**
3640 * pagetable_alloc - Allocate pagetables
3641 * @gfp: GFP flags
3642 * @order: desired pagetable order
3643 *
3644 * pagetable_alloc allocates memory for page tables as well as a page table
3645 * descriptor to describe that memory.
3646 *
3647 * Return: The ptdesc describing the allocated page tables.
3648 */
3649static inline struct ptdesc *pagetable_alloc_noprof(gfp_t gfp, unsigned int order)
3650{
3651 struct page *page = alloc_pages_noprof(gfp | __GFP_COMP, order);
3652
3653 return page_ptdesc(page);
3654}
3655#define pagetable_alloc(...) alloc_hooks(pagetable_alloc_noprof(__VA_ARGS__))
3656
3657static inline void __pagetable_free(struct ptdesc *pt)
3658{
3659 struct page *page = ptdesc_page(pt);
3660
3661 __free_pages(page, compound_order(page));
3662}
3663
3664#ifdef CONFIG_ASYNC_KERNEL_PGTABLE_FREE
3665void pagetable_free_kernel(struct ptdesc *pt);
3666#else
3667static inline void pagetable_free_kernel(struct ptdesc *pt)
3668{
3669 __pagetable_free(pt);
3670}
3671#endif
3672/**
3673 * pagetable_free - Free pagetables
3674 * @pt: The page table descriptor
3675 *
3676 * pagetable_free frees the memory of all page tables described by a page
3677 * table descriptor and the memory for the descriptor itself.
3678 */
3679static inline void pagetable_free(struct ptdesc *pt)
3680{
3681 if (ptdesc_test_kernel(pt)) {
3682 ptdesc_clear_kernel(pt);
3683 pagetable_free_kernel(pt);
3684 } else {
3685 __pagetable_free(pt);
3686 }
3687}
3688
3689#if defined(CONFIG_SPLIT_PTE_PTLOCKS)
3690#if ALLOC_SPLIT_PTLOCKS
3691void __init ptlock_cache_init(void);
3692bool ptlock_alloc(struct ptdesc *ptdesc);
3693void ptlock_free(struct ptdesc *ptdesc);
3694
3695static inline spinlock_t *ptlock_ptr(struct ptdesc *ptdesc)
3696{
3697 return ptdesc->ptl;
3698}
3699#else /* ALLOC_SPLIT_PTLOCKS */
3700static inline void ptlock_cache_init(void)
3701{
3702}
3703
3704static inline bool ptlock_alloc(struct ptdesc *ptdesc)
3705{
3706 return true;
3707}
3708
3709static inline void ptlock_free(struct ptdesc *ptdesc)
3710{
3711}
3712
3713static inline spinlock_t *ptlock_ptr(struct ptdesc *ptdesc)
3714{
3715 return &ptdesc->ptl;
3716}
3717#endif /* ALLOC_SPLIT_PTLOCKS */
3718
3719static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
3720{
3721 return ptlock_ptr(page_ptdesc(pmd_page(*pmd)));
3722}
3723
3724static inline spinlock_t *ptep_lockptr(struct mm_struct *mm, pte_t *pte)
3725{
3726 BUILD_BUG_ON(IS_ENABLED(CONFIG_HIGHPTE));
3727 BUILD_BUG_ON(MAX_PTRS_PER_PTE * sizeof(pte_t) > PAGE_SIZE);
3728 return ptlock_ptr(virt_to_ptdesc(pte));
3729}
3730
3731static inline bool ptlock_init(struct ptdesc *ptdesc)
3732{
3733 /*
3734 * prep_new_page() initialize page->private (and therefore page->ptl)
3735 * with 0. Make sure nobody took it in use in between.
3736 *
3737 * It can happen if arch try to use slab for page table allocation:
3738 * slab code uses page->slab_cache, which share storage with page->ptl.
3739 */
3740 VM_BUG_ON_PAGE(*(unsigned long *)&ptdesc->ptl, ptdesc_page(ptdesc));
3741 if (!ptlock_alloc(ptdesc))
3742 return false;
3743 spin_lock_init(ptlock_ptr(ptdesc));
3744 return true;
3745}
3746
3747#else /* !defined(CONFIG_SPLIT_PTE_PTLOCKS) */
3748/*
3749 * We use mm->page_table_lock to guard all pagetable pages of the mm.
3750 */
3751static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
3752{
3753 return &mm->page_table_lock;
3754}
3755static inline spinlock_t *ptep_lockptr(struct mm_struct *mm, pte_t *pte)
3756{
3757 return &mm->page_table_lock;
3758}
3759static inline void ptlock_cache_init(void) {}
3760static inline bool ptlock_init(struct ptdesc *ptdesc) { return true; }
3761static inline void ptlock_free(struct ptdesc *ptdesc) {}
3762#endif /* defined(CONFIG_SPLIT_PTE_PTLOCKS) */
3763
3764static inline void __pagetable_ctor(struct ptdesc *ptdesc)
3765{
3766 struct folio *folio = ptdesc_folio(ptdesc);
3767
3768 __folio_set_pgtable(folio);
3769 lruvec_stat_add_folio(folio, NR_PAGETABLE);
3770}
3771
3772static inline void pagetable_dtor(struct ptdesc *ptdesc)
3773{
3774 struct folio *folio = ptdesc_folio(ptdesc);
3775
3776 ptlock_free(ptdesc);
3777 __folio_clear_pgtable(folio);
3778 lruvec_stat_sub_folio(folio, NR_PAGETABLE);
3779}
3780
3781static inline void pagetable_dtor_free(struct ptdesc *ptdesc)
3782{
3783 pagetable_dtor(ptdesc);
3784 pagetable_free(ptdesc);
3785}
3786
3787static inline bool pagetable_pte_ctor(struct mm_struct *mm,
3788 struct ptdesc *ptdesc)
3789{
3790 if (mm != &init_mm && !ptlock_init(ptdesc))
3791 return false;
3792 __pagetable_ctor(ptdesc);
3793 return true;
3794}
3795
3796pte_t *__pte_offset_map(pmd_t *pmd, unsigned long addr, pmd_t *pmdvalp);
3797
3798static inline pte_t *pte_offset_map(pmd_t *pmd, unsigned long addr)
3799{
3800 return __pte_offset_map(pmd, addr, NULL);
3801}
3802
3803pte_t *pte_offset_map_lock(struct mm_struct *mm, pmd_t *pmd,
3804 unsigned long addr, spinlock_t **ptlp);
3805
3806pte_t *pte_offset_map_ro_nolock(struct mm_struct *mm, pmd_t *pmd,
3807 unsigned long addr, spinlock_t **ptlp);
3808pte_t *pte_offset_map_rw_nolock(struct mm_struct *mm, pmd_t *pmd,
3809 unsigned long addr, pmd_t *pmdvalp,
3810 spinlock_t **ptlp);
3811
3812#define pte_unmap_unlock(pte, ptl) do { \
3813 spin_unlock(ptl); \
3814 pte_unmap(pte); \
3815} while (0)
3816
3817#define pte_alloc(mm, pmd) (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd))
3818
3819#define pte_alloc_map(mm, pmd, address) \
3820 (pte_alloc(mm, pmd) ? NULL : pte_offset_map(pmd, address))
3821
3822#define pte_alloc_map_lock(mm, pmd, address, ptlp) \
3823 (pte_alloc(mm, pmd) ? \
3824 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
3825
3826#define pte_alloc_kernel(pmd, address) \
3827 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
3828 NULL: pte_offset_kernel(pmd, address))
3829
3830#if defined(CONFIG_SPLIT_PMD_PTLOCKS)
3831
3832static inline struct page *pmd_pgtable_page(pmd_t *pmd)
3833{
3834 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
3835 return virt_to_page((void *)((unsigned long) pmd & mask));
3836}
3837
3838static inline struct ptdesc *pmd_ptdesc(pmd_t *pmd)
3839{
3840 return page_ptdesc(pmd_pgtable_page(pmd));
3841}
3842
3843static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
3844{
3845 return ptlock_ptr(pmd_ptdesc(pmd));
3846}
3847
3848static inline bool pmd_ptlock_init(struct ptdesc *ptdesc)
3849{
3850#ifdef CONFIG_TRANSPARENT_HUGEPAGE
3851 ptdesc->pmd_huge_pte = NULL;
3852#endif
3853 return ptlock_init(ptdesc);
3854}
3855
3856#define pmd_huge_pte(mm, pmd) (pmd_ptdesc(pmd)->pmd_huge_pte)
3857
3858#else
3859
3860static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
3861{
3862 return &mm->page_table_lock;
3863}
3864
3865static inline bool pmd_ptlock_init(struct ptdesc *ptdesc) { return true; }
3866
3867#define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
3868
3869#endif
3870
3871static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
3872{
3873 spinlock_t *ptl = pmd_lockptr(mm, pmd);
3874 spin_lock(ptl);
3875 return ptl;
3876}
3877
3878static inline bool pagetable_pmd_ctor(struct mm_struct *mm,
3879 struct ptdesc *ptdesc)
3880{
3881 if (mm != &init_mm && !pmd_ptlock_init(ptdesc))
3882 return false;
3883 ptdesc_pmd_pts_init(ptdesc);
3884 __pagetable_ctor(ptdesc);
3885 return true;
3886}
3887
3888/*
3889 * No scalability reason to split PUD locks yet, but follow the same pattern
3890 * as the PMD locks to make it easier if we decide to. The VM should not be
3891 * considered ready to switch to split PUD locks yet; there may be places
3892 * which need to be converted from page_table_lock.
3893 */
3894static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
3895{
3896 return &mm->page_table_lock;
3897}
3898
3899static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
3900{
3901 spinlock_t *ptl = pud_lockptr(mm, pud);
3902
3903 spin_lock(ptl);
3904 return ptl;
3905}
3906
3907static inline void pagetable_pud_ctor(struct ptdesc *ptdesc)
3908{
3909 __pagetable_ctor(ptdesc);
3910}
3911
3912static inline void pagetable_p4d_ctor(struct ptdesc *ptdesc)
3913{
3914 __pagetable_ctor(ptdesc);
3915}
3916
3917static inline void pagetable_pgd_ctor(struct ptdesc *ptdesc)
3918{
3919 __pagetable_ctor(ptdesc);
3920}
3921
3922extern void __init pagecache_init(void);
3923extern void free_initmem(void);
3924
3925/*
3926 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
3927 * into the buddy system. The freed pages will be poisoned with pattern
3928 * "poison" if it's within range [0, UCHAR_MAX].
3929 * Return pages freed into the buddy system.
3930 */
3931extern unsigned long free_reserved_area(void *start, void *end,
3932 int poison, const char *s);
3933
3934extern void adjust_managed_page_count(struct page *page, long count);
3935
3936/* Free the reserved page into the buddy system, so it gets managed. */
3937void free_reserved_page(struct page *page);
3938
3939static inline void mark_page_reserved(struct page *page)
3940{
3941 SetPageReserved(page);
3942 adjust_managed_page_count(page, -1);
3943}
3944
3945static inline void free_reserved_ptdesc(struct ptdesc *pt)
3946{
3947 free_reserved_page(ptdesc_page(pt));
3948}
3949
3950/*
3951 * Default method to free all the __init memory into the buddy system.
3952 * The freed pages will be poisoned with pattern "poison" if it's within
3953 * range [0, UCHAR_MAX].
3954 * Return pages freed into the buddy system.
3955 */
3956static inline unsigned long free_initmem_default(int poison)
3957{
3958 extern char __init_begin[], __init_end[];
3959
3960 return free_reserved_area(&__init_begin, &__init_end,
3961 poison, "unused kernel image (initmem)");
3962}
3963
3964static inline unsigned long get_num_physpages(void)
3965{
3966 int nid;
3967 unsigned long phys_pages = 0;
3968
3969 for_each_online_node(nid)
3970 phys_pages += node_present_pages(nid);
3971
3972 return phys_pages;
3973}
3974
3975/*
3976 * FIXME: Using memblock node mappings, an architecture may initialise its
3977 * zones, allocate the backing mem_map and account for memory holes in an
3978 * architecture independent manner.
3979 *
3980 * An architecture is expected to register range of page frames backed by
3981 * physical memory with memblock_add[_node]() before calling
3982 * free_area_init() passing in the PFN each zone ends at. At a basic
3983 * usage, an architecture is expected to do something like
3984 *
3985 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
3986 * max_highmem_pfn};
3987 * for_each_valid_physical_page_range()
3988 * memblock_add_node(base, size, nid, MEMBLOCK_NONE)
3989 * free_area_init(max_zone_pfns);
3990 */
3991void arch_zone_limits_init(unsigned long *max_zone_pfn);
3992unsigned long node_map_pfn_alignment(void);
3993extern unsigned long absent_pages_in_range(unsigned long start_pfn,
3994 unsigned long end_pfn);
3995extern void get_pfn_range_for_nid(unsigned int nid,
3996 unsigned long *start_pfn, unsigned long *end_pfn);
3997
3998#ifndef CONFIG_NUMA
3999static inline int early_pfn_to_nid(unsigned long pfn)
4000{
4001 return 0;
4002}
4003#else
4004/* please see mm/page_alloc.c */
4005extern int __meminit early_pfn_to_nid(unsigned long pfn);
4006#endif
4007
4008extern void mem_init(void);
4009extern void __init mmap_init(void);
4010
4011extern void __show_mem(unsigned int flags, nodemask_t *nodemask, int max_zone_idx);
4012static inline void show_mem(void)
4013{
4014 __show_mem(0, NULL, MAX_NR_ZONES - 1);
4015}
4016extern long si_mem_available(void);
4017extern void si_meminfo(struct sysinfo * val);
4018extern void si_meminfo_node(struct sysinfo *val, int nid);
4019
4020extern __printf(3, 4)
4021void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
4022
4023extern void setup_per_cpu_pageset(void);
4024
4025/* nommu.c */
4026extern atomic_long_t mmap_pages_allocated;
4027extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
4028
4029/* interval_tree.c */
4030void vma_interval_tree_insert(struct vm_area_struct *node,
4031 struct rb_root_cached *root);
4032void vma_interval_tree_insert_after(struct vm_area_struct *node,
4033 struct vm_area_struct *prev,
4034 struct rb_root_cached *root);
4035void vma_interval_tree_remove(struct vm_area_struct *node,
4036 struct rb_root_cached *root);
4037struct vm_area_struct *vma_interval_tree_subtree_search(struct vm_area_struct *node,
4038 unsigned long start, unsigned long last);
4039struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
4040 unsigned long start, unsigned long last);
4041struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
4042 unsigned long start, unsigned long last);
4043
4044#define vma_interval_tree_foreach(vma, root, start, last) \
4045 for (vma = vma_interval_tree_iter_first(root, start, last); \
4046 vma; vma = vma_interval_tree_iter_next(vma, start, last))
4047
4048void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
4049 struct rb_root_cached *root);
4050void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
4051 struct rb_root_cached *root);
4052struct anon_vma_chain *
4053anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
4054 unsigned long start, unsigned long last);
4055struct anon_vma_chain *anon_vma_interval_tree_iter_next(
4056 struct anon_vma_chain *node, unsigned long start, unsigned long last);
4057#ifdef CONFIG_DEBUG_VM_RB
4058void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
4059#endif
4060
4061#define anon_vma_interval_tree_foreach(avc, root, start, last) \
4062 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
4063 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
4064
4065/* mmap.c */
4066extern int __vm_enough_memory(const struct mm_struct *mm, long pages, int cap_sys_admin);
4067extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
4068extern void exit_mmap(struct mm_struct *);
4069bool mmap_read_lock_maybe_expand(struct mm_struct *mm, struct vm_area_struct *vma,
4070 unsigned long addr, bool write);
4071
4072static inline int check_data_rlimit(unsigned long rlim,
4073 unsigned long new,
4074 unsigned long start,
4075 unsigned long end_data,
4076 unsigned long start_data)
4077{
4078 if (rlim < RLIM_INFINITY) {
4079 if (((new - start) + (end_data - start_data)) > rlim)
4080 return -ENOSPC;
4081 }
4082
4083 return 0;
4084}
4085
4086extern int mm_take_all_locks(struct mm_struct *mm);
4087extern void mm_drop_all_locks(struct mm_struct *mm);
4088
4089extern int set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
4090extern int replace_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
4091extern struct file *get_mm_exe_file(struct mm_struct *mm);
4092extern struct file *get_task_exe_file(struct task_struct *task);
4093
4094extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
4095
4096extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
4097 const struct vm_special_mapping *sm);
4098struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
4099 unsigned long addr, unsigned long len,
4100 vm_flags_t vm_flags,
4101 const struct vm_special_mapping *spec);
4102
4103unsigned long randomize_stack_top(unsigned long stack_top);
4104unsigned long randomize_page(unsigned long start, unsigned long range);
4105
4106unsigned long
4107__get_unmapped_area(struct file *file, unsigned long addr, unsigned long len,
4108 unsigned long pgoff, unsigned long flags, vm_flags_t vm_flags);
4109
4110static inline unsigned long
4111get_unmapped_area(struct file *file, unsigned long addr, unsigned long len,
4112 unsigned long pgoff, unsigned long flags)
4113{
4114 return __get_unmapped_area(file, addr, len, pgoff, flags, 0);
4115}
4116
4117extern unsigned long do_mmap(struct file *file, unsigned long addr,
4118 unsigned long len, unsigned long prot, unsigned long flags,
4119 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate,
4120 struct list_head *uf);
4121extern int do_vmi_munmap(struct vma_iterator *vmi, struct mm_struct *mm,
4122 unsigned long start, size_t len, struct list_head *uf,
4123 bool unlock);
4124int do_vmi_align_munmap(struct vma_iterator *vmi, struct vm_area_struct *vma,
4125 struct mm_struct *mm, unsigned long start,
4126 unsigned long end, struct list_head *uf, bool unlock);
4127extern int do_munmap(struct mm_struct *, unsigned long, size_t,
4128 struct list_head *uf);
4129extern int do_madvise(struct mm_struct *mm, unsigned long start, size_t len_in, int behavior);
4130
4131#ifdef CONFIG_MMU
4132extern int __mm_populate(unsigned long addr, unsigned long len,
4133 int ignore_errors);
4134static inline void mm_populate(unsigned long addr, unsigned long len)
4135{
4136 /* Ignore errors */
4137 (void) __mm_populate(addr, len, 1);
4138}
4139#else
4140static inline void mm_populate(unsigned long addr, unsigned long len) {}
4141#endif
4142
4143/* This takes the mm semaphore itself */
4144int __must_check vm_brk_flags(unsigned long addr, unsigned long request, bool is_exec);
4145int vm_munmap(unsigned long start, size_t len);
4146unsigned long __must_check vm_mmap(struct file *file, unsigned long addr,
4147 unsigned long len, unsigned long prot,
4148 unsigned long flag, unsigned long offset);
4149unsigned long __must_check vm_mmap_shadow_stack(unsigned long addr,
4150 unsigned long len, unsigned long flags);
4151
4152struct vm_unmapped_area_info {
4153#define VM_UNMAPPED_AREA_TOPDOWN 1
4154 unsigned long flags;
4155 unsigned long length;
4156 unsigned long low_limit;
4157 unsigned long high_limit;
4158 unsigned long align_mask;
4159 unsigned long align_offset;
4160 unsigned long start_gap;
4161};
4162
4163extern unsigned long vm_unmapped_area(struct vm_unmapped_area_info *info);
4164
4165/* truncate.c */
4166void truncate_inode_pages(struct address_space *mapping, loff_t lstart);
4167void truncate_inode_pages_range(struct address_space *mapping, loff_t lstart,
4168 uoff_t lend);
4169void truncate_inode_pages_final(struct address_space *mapping);
4170
4171/* generic vm_area_ops exported for stackable file systems */
4172extern vm_fault_t filemap_fault(struct vm_fault *vmf);
4173extern vm_fault_t filemap_map_pages(struct vm_fault *vmf,
4174 pgoff_t start_pgoff, pgoff_t end_pgoff);
4175extern vm_fault_t filemap_page_mkwrite(struct vm_fault *vmf);
4176
4177extern unsigned long stack_guard_gap;
4178/* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
4179int expand_stack_locked(struct vm_area_struct *vma, unsigned long address);
4180struct vm_area_struct *expand_stack(struct mm_struct * mm, unsigned long addr);
4181
4182/* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
4183extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
4184extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
4185 struct vm_area_struct **pprev);
4186
4187/*
4188 * Look up the first VMA which intersects the interval [start_addr, end_addr)
4189 * NULL if none. Assume start_addr < end_addr.
4190 */
4191struct vm_area_struct *find_vma_intersection(struct mm_struct *mm,
4192 unsigned long start_addr, unsigned long end_addr);
4193
4194/**
4195 * vma_lookup() - Find a VMA at a specific address
4196 * @mm: The process address space.
4197 * @addr: The user address.
4198 *
4199 * Return: The vm_area_struct at the given address, %NULL otherwise.
4200 */
4201static inline
4202struct vm_area_struct *vma_lookup(struct mm_struct *mm, unsigned long addr)
4203{
4204 return mtree_load(&mm->mm_mt, addr);
4205}
4206
4207static inline unsigned long stack_guard_start_gap(const struct vm_area_struct *vma)
4208{
4209 if (vma->vm_flags & VM_GROWSDOWN)
4210 return stack_guard_gap;
4211
4212 /* See reasoning around the VM_SHADOW_STACK definition */
4213 if (vma->vm_flags & VM_SHADOW_STACK)
4214 return PAGE_SIZE;
4215
4216 return 0;
4217}
4218
4219static inline unsigned long vm_start_gap(const struct vm_area_struct *vma)
4220{
4221 unsigned long gap = stack_guard_start_gap(vma);
4222 unsigned long vm_start = vma->vm_start;
4223
4224 vm_start -= gap;
4225 if (vm_start > vma->vm_start)
4226 vm_start = 0;
4227 return vm_start;
4228}
4229
4230static inline unsigned long vm_end_gap(const struct vm_area_struct *vma)
4231{
4232 unsigned long vm_end = vma->vm_end;
4233
4234 if (vma->vm_flags & VM_GROWSUP) {
4235 vm_end += stack_guard_gap;
4236 if (vm_end < vma->vm_end)
4237 vm_end = -PAGE_SIZE;
4238 }
4239 return vm_end;
4240}
4241
4242static inline unsigned long vma_pages(const struct vm_area_struct *vma)
4243{
4244 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
4245}
4246
4247static inline unsigned long vma_last_pgoff(struct vm_area_struct *vma)
4248{
4249 return vma->vm_pgoff + vma_pages(vma) - 1;
4250}
4251
4252static inline unsigned long vma_desc_size(const struct vm_area_desc *desc)
4253{
4254 return desc->end - desc->start;
4255}
4256
4257static inline unsigned long vma_desc_pages(const struct vm_area_desc *desc)
4258{
4259 return vma_desc_size(desc) >> PAGE_SHIFT;
4260}
4261
4262/**
4263 * mmap_action_remap - helper for mmap_prepare hook to specify that a pure PFN
4264 * remap is required.
4265 * @desc: The VMA descriptor for the VMA requiring remap.
4266 * @start: The virtual address to start the remap from, must be within the VMA.
4267 * @start_pfn: The first PFN in the range to remap.
4268 * @size: The size of the range to remap, in bytes, at most spanning to the end
4269 * of the VMA.
4270 */
4271static inline void mmap_action_remap(struct vm_area_desc *desc,
4272 unsigned long start,
4273 unsigned long start_pfn,
4274 unsigned long size)
4275{
4276 struct mmap_action *action = &desc->action;
4277
4278 /* [start, start + size) must be within the VMA. */
4279 WARN_ON_ONCE(start < desc->start || start >= desc->end);
4280 WARN_ON_ONCE(start + size > desc->end);
4281
4282 action->type = MMAP_REMAP_PFN;
4283 action->remap.start = start;
4284 action->remap.start_pfn = start_pfn;
4285 action->remap.size = size;
4286 action->remap.pgprot = desc->page_prot;
4287}
4288
4289/**
4290 * mmap_action_remap_full - helper for mmap_prepare hook to specify that the
4291 * entirety of a VMA should be PFN remapped.
4292 * @desc: The VMA descriptor for the VMA requiring remap.
4293 * @start_pfn: The first PFN in the range to remap.
4294 */
4295static inline void mmap_action_remap_full(struct vm_area_desc *desc,
4296 unsigned long start_pfn)
4297{
4298 mmap_action_remap(desc, desc->start, start_pfn, vma_desc_size(desc));
4299}
4300
4301/**
4302 * mmap_action_ioremap - helper for mmap_prepare hook to specify that a pure PFN
4303 * I/O remap is required.
4304 * @desc: The VMA descriptor for the VMA requiring remap.
4305 * @start: The virtual address to start the remap from, must be within the VMA.
4306 * @start_pfn: The first PFN in the range to remap.
4307 * @size: The size of the range to remap, in bytes, at most spanning to the end
4308 * of the VMA.
4309 */
4310static inline void mmap_action_ioremap(struct vm_area_desc *desc,
4311 unsigned long start,
4312 unsigned long start_pfn,
4313 unsigned long size)
4314{
4315 mmap_action_remap(desc, start, start_pfn, size);
4316 desc->action.type = MMAP_IO_REMAP_PFN;
4317}
4318
4319/**
4320 * mmap_action_ioremap_full - helper for mmap_prepare hook to specify that the
4321 * entirety of a VMA should be PFN I/O remapped.
4322 * @desc: The VMA descriptor for the VMA requiring remap.
4323 * @start_pfn: The first PFN in the range to remap.
4324 */
4325static inline void mmap_action_ioremap_full(struct vm_area_desc *desc,
4326 unsigned long start_pfn)
4327{
4328 mmap_action_ioremap(desc, desc->start, start_pfn, vma_desc_size(desc));
4329}
4330
4331/**
4332 * mmap_action_simple_ioremap - helper for mmap_prepare hook to specify that the
4333 * physical range in [start_phys_addr, start_phys_addr + size) should be I/O
4334 * remapped.
4335 * @desc: The VMA descriptor for the VMA requiring remap.
4336 * @start_phys_addr: Start of the physical memory to be mapped.
4337 * @size: Size of the area to map.
4338 *
4339 * NOTE: Some drivers might want to tweak desc->page_prot for purposes of
4340 * write-combine or similar.
4341 */
4342static inline void mmap_action_simple_ioremap(struct vm_area_desc *desc,
4343 phys_addr_t start_phys_addr,
4344 unsigned long size)
4345{
4346 struct mmap_action *action = &desc->action;
4347
4348 action->simple_ioremap.start_phys_addr = start_phys_addr;
4349 action->simple_ioremap.size = size;
4350 action->type = MMAP_SIMPLE_IO_REMAP;
4351}
4352
4353/**
4354 * mmap_action_map_kernel_pages - helper for mmap_prepare hook to specify that
4355 * @num kernel pages contained in the @pages array should be mapped to userland
4356 * starting at virtual address @start.
4357 * @desc: The VMA descriptor for the VMA requiring kernel pags to be mapped.
4358 * @start: The virtual address from which to map them.
4359 * @pages: An array of struct page pointers describing the memory to map.
4360 * @nr_pages: The number of entries in the @pages aray.
4361 */
4362static inline void mmap_action_map_kernel_pages(struct vm_area_desc *desc,
4363 unsigned long start, struct page **pages,
4364 unsigned long nr_pages)
4365{
4366 struct mmap_action *action = &desc->action;
4367
4368 action->type = MMAP_MAP_KERNEL_PAGES;
4369 action->map_kernel.start = start;
4370 action->map_kernel.pages = pages;
4371 action->map_kernel.nr_pages = nr_pages;
4372 action->map_kernel.pgoff = desc->pgoff;
4373}
4374
4375/**
4376 * mmap_action_map_kernel_pages_full - helper for mmap_prepare hook to specify that
4377 * kernel pages contained in the @pages array should be mapped to userland
4378 * from @desc->start to @desc->end.
4379 * @desc: The VMA descriptor for the VMA requiring kernel pags to be mapped.
4380 * @pages: An array of struct page pointers describing the memory to map.
4381 *
4382 * The caller must ensure that @pages contains sufficient entries to cover the
4383 * entire range described by @desc.
4384 */
4385static inline void mmap_action_map_kernel_pages_full(struct vm_area_desc *desc,
4386 struct page **pages)
4387{
4388 mmap_action_map_kernel_pages(desc, desc->start, pages,
4389 vma_desc_pages(desc));
4390}
4391
4392int mmap_action_prepare(struct vm_area_desc *desc);
4393int mmap_action_complete(struct vm_area_struct *vma,
4394 struct mmap_action *action);
4395
4396/* Look up the first VMA which exactly match the interval vm_start ... vm_end */
4397static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
4398 unsigned long vm_start, unsigned long vm_end)
4399{
4400 struct vm_area_struct *vma = vma_lookup(mm, vm_start);
4401
4402 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
4403 vma = NULL;
4404
4405 return vma;
4406}
4407
4408/**
4409 * range_is_subset - Is the specified inner range a subset of the outer range?
4410 * @outer_start: The start of the outer range.
4411 * @outer_end: The exclusive end of the outer range.
4412 * @inner_start: The start of the inner range.
4413 * @inner_end: The exclusive end of the inner range.
4414 *
4415 * Returns: %true if [inner_start, inner_end) is a subset of [outer_start,
4416 * outer_end), otherwise %false.
4417 */
4418static inline bool range_is_subset(unsigned long outer_start,
4419 unsigned long outer_end,
4420 unsigned long inner_start,
4421 unsigned long inner_end)
4422{
4423 return outer_start <= inner_start && inner_end <= outer_end;
4424}
4425
4426/**
4427 * range_in_vma - is the specified [@start, @end) range a subset of the VMA?
4428 * @vma: The VMA against which we want to check [@start, @end).
4429 * @start: The start of the range we wish to check.
4430 * @end: The exclusive end of the range we wish to check.
4431 *
4432 * Returns: %true if [@start, @end) is a subset of [@vma->vm_start,
4433 * @vma->vm_end), %false otherwise.
4434 */
4435static inline bool range_in_vma(const struct vm_area_struct *vma,
4436 unsigned long start, unsigned long end)
4437{
4438 if (!vma)
4439 return false;
4440
4441 return range_is_subset(vma->vm_start, vma->vm_end, start, end);
4442}
4443
4444/**
4445 * range_in_vma_desc - is the specified [@start, @end) range a subset of the VMA
4446 * described by @desc, a VMA descriptor?
4447 * @desc: The VMA descriptor against which we want to check [@start, @end).
4448 * @start: The start of the range we wish to check.
4449 * @end: The exclusive end of the range we wish to check.
4450 *
4451 * Returns: %true if [@start, @end) is a subset of [@desc->start, @desc->end),
4452 * %false otherwise.
4453 */
4454static inline bool range_in_vma_desc(const struct vm_area_desc *desc,
4455 unsigned long start, unsigned long end)
4456{
4457 if (!desc)
4458 return false;
4459
4460 return range_is_subset(desc->start, desc->end, start, end);
4461}
4462
4463#ifdef CONFIG_MMU
4464pgprot_t vm_get_page_prot(vm_flags_t vm_flags);
4465
4466static inline pgprot_t vma_get_page_prot(vma_flags_t vma_flags)
4467{
4468 const vm_flags_t vm_flags = vma_flags_to_legacy(vma_flags);
4469
4470 return vm_get_page_prot(vm_flags);
4471}
4472
4473void vma_set_page_prot(struct vm_area_struct *vma);
4474#else
4475static inline pgprot_t vm_get_page_prot(vm_flags_t vm_flags)
4476{
4477 return __pgprot(0);
4478}
4479static inline pgprot_t vma_get_page_prot(vma_flags_t vma_flags)
4480{
4481 return __pgprot(0);
4482}
4483static inline void vma_set_page_prot(struct vm_area_struct *vma)
4484{
4485 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
4486}
4487#endif
4488
4489void vma_set_file(struct vm_area_struct *vma, struct file *file);
4490
4491#ifdef CONFIG_NUMA_BALANCING
4492unsigned long change_prot_numa(struct vm_area_struct *vma,
4493 unsigned long start, unsigned long end);
4494#endif
4495
4496struct vm_area_struct *find_extend_vma_locked(struct mm_struct *,
4497 unsigned long addr);
4498int remap_pfn_range(struct vm_area_struct *vma, unsigned long addr,
4499 unsigned long pfn, unsigned long size, pgprot_t pgprot);
4500
4501int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
4502int vm_insert_pages(struct vm_area_struct *vma, unsigned long addr,
4503 struct page **pages, unsigned long *num);
4504int map_kernel_pages_prepare(struct vm_area_desc *desc);
4505int map_kernel_pages_complete(struct vm_area_struct *vma,
4506 struct mmap_action *action);
4507int vm_map_pages(struct vm_area_struct *vma, struct page **pages,
4508 unsigned long num);
4509int vm_map_pages_zero(struct vm_area_struct *vma, struct page **pages,
4510 unsigned long num);
4511vm_fault_t vmf_insert_page_mkwrite(struct vm_fault *vmf, struct page *page,
4512 bool write);
4513vm_fault_t vmf_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
4514 unsigned long pfn);
4515vm_fault_t vmf_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
4516 unsigned long pfn, pgprot_t pgprot);
4517vm_fault_t vmf_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
4518 unsigned long pfn);
4519vm_fault_t vmf_insert_mixed_mkwrite(struct vm_area_struct *vma,
4520 unsigned long addr, unsigned long pfn);
4521int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
4522
4523static inline vm_fault_t vmf_insert_page(struct vm_area_struct *vma,
4524 unsigned long addr, struct page *page)
4525{
4526 int err = vm_insert_page(vma, addr, page);
4527
4528 if (err == -ENOMEM)
4529 return VM_FAULT_OOM;
4530 if (err < 0 && err != -EBUSY)
4531 return VM_FAULT_SIGBUS;
4532
4533 return VM_FAULT_NOPAGE;
4534}
4535
4536#ifndef io_remap_pfn_range_pfn
4537static inline unsigned long io_remap_pfn_range_pfn(unsigned long pfn,
4538 unsigned long size)
4539{
4540 return pfn;
4541}
4542#endif
4543
4544static inline int io_remap_pfn_range(struct vm_area_struct *vma,
4545 unsigned long addr, unsigned long orig_pfn,
4546 unsigned long size, pgprot_t orig_prot)
4547{
4548 const unsigned long pfn = io_remap_pfn_range_pfn(orig_pfn, size);
4549 const pgprot_t prot = pgprot_decrypted(orig_prot);
4550
4551 return remap_pfn_range(vma, addr, pfn, size, prot);
4552}
4553
4554static inline vm_fault_t vmf_error(int err)
4555{
4556 if (err == -ENOMEM)
4557 return VM_FAULT_OOM;
4558 else if (err == -EHWPOISON)
4559 return VM_FAULT_HWPOISON;
4560 return VM_FAULT_SIGBUS;
4561}
4562
4563/*
4564 * Convert errno to return value for ->page_mkwrite() calls.
4565 *
4566 * This should eventually be merged with vmf_error() above, but will need a
4567 * careful audit of all vmf_error() callers.
4568 */
4569static inline vm_fault_t vmf_fs_error(int err)
4570{
4571 if (err == 0)
4572 return VM_FAULT_LOCKED;
4573 if (err == -EFAULT || err == -EAGAIN)
4574 return VM_FAULT_NOPAGE;
4575 if (err == -ENOMEM)
4576 return VM_FAULT_OOM;
4577 /* -ENOSPC, -EDQUOT, -EIO ... */
4578 return VM_FAULT_SIGBUS;
4579}
4580
4581static inline int vm_fault_to_errno(vm_fault_t vm_fault, int foll_flags)
4582{
4583 if (vm_fault & VM_FAULT_OOM)
4584 return -ENOMEM;
4585 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
4586 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
4587 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
4588 return -EFAULT;
4589 return 0;
4590}
4591
4592/*
4593 * Indicates whether GUP can follow a PROT_NONE mapped page, or whether
4594 * a (NUMA hinting) fault is required.
4595 */
4596static inline bool gup_can_follow_protnone(const struct vm_area_struct *vma,
4597 unsigned int flags)
4598{
4599 /*
4600 * If callers don't want to honor NUMA hinting faults, no need to
4601 * determine if we would actually have to trigger a NUMA hinting fault.
4602 */
4603 if (!(flags & FOLL_HONOR_NUMA_FAULT))
4604 return true;
4605
4606 /*
4607 * NUMA hinting faults don't apply in inaccessible (PROT_NONE) VMAs.
4608 *
4609 * Requiring a fault here even for inaccessible VMAs would mean that
4610 * FOLL_FORCE cannot make any progress, because handle_mm_fault()
4611 * refuses to process NUMA hinting faults in inaccessible VMAs.
4612 */
4613 return !vma_is_accessible(vma);
4614}
4615
4616typedef int (*pte_fn_t)(pte_t *pte, unsigned long addr, void *data);
4617extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
4618 unsigned long size, pte_fn_t fn, void *data);
4619extern int apply_to_existing_page_range(struct mm_struct *mm,
4620 unsigned long address, unsigned long size,
4621 pte_fn_t fn, void *data);
4622
4623#ifdef CONFIG_PAGE_POISONING
4624extern void __kernel_poison_pages(struct page *page, int numpages);
4625extern void __kernel_unpoison_pages(struct page *page, int numpages);
4626extern bool _page_poisoning_enabled_early;
4627DECLARE_STATIC_KEY_FALSE(_page_poisoning_enabled);
4628static inline bool page_poisoning_enabled(void)
4629{
4630 return _page_poisoning_enabled_early;
4631}
4632/*
4633 * For use in fast paths after init_mem_debugging() has run, or when a
4634 * false negative result is not harmful when called too early.
4635 */
4636static inline bool page_poisoning_enabled_static(void)
4637{
4638 return static_branch_unlikely(&_page_poisoning_enabled);
4639}
4640static inline void kernel_poison_pages(struct page *page, int numpages)
4641{
4642 if (page_poisoning_enabled_static())
4643 __kernel_poison_pages(page, numpages);
4644}
4645static inline void kernel_unpoison_pages(struct page *page, int numpages)
4646{
4647 if (page_poisoning_enabled_static())
4648 __kernel_unpoison_pages(page, numpages);
4649}
4650#else
4651static inline bool page_poisoning_enabled(void) { return false; }
4652static inline bool page_poisoning_enabled_static(void) { return false; }
4653static inline void __kernel_poison_pages(struct page *page, int nunmpages) { }
4654static inline void kernel_poison_pages(struct page *page, int numpages) { }
4655static inline void kernel_unpoison_pages(struct page *page, int numpages) { }
4656#endif
4657
4658DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_ALLOC_DEFAULT_ON, init_on_alloc);
4659static inline bool want_init_on_alloc(gfp_t flags)
4660{
4661 if (static_branch_maybe(CONFIG_INIT_ON_ALLOC_DEFAULT_ON,
4662 &init_on_alloc))
4663 return true;
4664 return flags & __GFP_ZERO;
4665}
4666
4667DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_FREE_DEFAULT_ON, init_on_free);
4668static inline bool want_init_on_free(void)
4669{
4670 return static_branch_maybe(CONFIG_INIT_ON_FREE_DEFAULT_ON,
4671 &init_on_free);
4672}
4673
4674extern bool _debug_pagealloc_enabled_early;
4675DECLARE_STATIC_KEY_FALSE(_debug_pagealloc_enabled);
4676
4677static inline bool debug_pagealloc_enabled(void)
4678{
4679 return IS_ENABLED(CONFIG_DEBUG_PAGEALLOC) &&
4680 _debug_pagealloc_enabled_early;
4681}
4682
4683/*
4684 * For use in fast paths after mem_debugging_and_hardening_init() has run,
4685 * or when a false negative result is not harmful when called too early.
4686 */
4687static inline bool debug_pagealloc_enabled_static(void)
4688{
4689 if (!IS_ENABLED(CONFIG_DEBUG_PAGEALLOC))
4690 return false;
4691
4692 return static_branch_unlikely(&_debug_pagealloc_enabled);
4693}
4694
4695/*
4696 * To support DEBUG_PAGEALLOC architecture must ensure that
4697 * __kernel_map_pages() never fails
4698 */
4699extern void __kernel_map_pages(struct page *page, int numpages, int enable);
4700#ifdef CONFIG_DEBUG_PAGEALLOC
4701static inline void debug_pagealloc_map_pages(struct page *page, int numpages)
4702{
4703 iommu_debug_check_unmapped(page, numpages);
4704
4705 if (debug_pagealloc_enabled_static())
4706 __kernel_map_pages(page, numpages, 1);
4707}
4708
4709static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages)
4710{
4711 iommu_debug_check_unmapped(page, numpages);
4712
4713 if (debug_pagealloc_enabled_static())
4714 __kernel_map_pages(page, numpages, 0);
4715}
4716
4717extern unsigned int _debug_guardpage_minorder;
4718DECLARE_STATIC_KEY_FALSE(_debug_guardpage_enabled);
4719
4720static inline unsigned int debug_guardpage_minorder(void)
4721{
4722 return _debug_guardpage_minorder;
4723}
4724
4725static inline bool debug_guardpage_enabled(void)
4726{
4727 return static_branch_unlikely(&_debug_guardpage_enabled);
4728}
4729
4730static inline bool page_is_guard(const struct page *page)
4731{
4732 if (!debug_guardpage_enabled())
4733 return false;
4734
4735 return PageGuard(page);
4736}
4737
4738bool __set_page_guard(struct zone *zone, struct page *page, unsigned int order);
4739static inline bool set_page_guard(struct zone *zone, struct page *page,
4740 unsigned int order)
4741{
4742 if (!debug_guardpage_enabled())
4743 return false;
4744 return __set_page_guard(zone, page, order);
4745}
4746
4747void __clear_page_guard(struct zone *zone, struct page *page, unsigned int order);
4748static inline void clear_page_guard(struct zone *zone, struct page *page,
4749 unsigned int order)
4750{
4751 if (!debug_guardpage_enabled())
4752 return;
4753 __clear_page_guard(zone, page, order);
4754}
4755
4756#else /* CONFIG_DEBUG_PAGEALLOC */
4757static inline void debug_pagealloc_map_pages(struct page *page, int numpages) {}
4758static inline void debug_pagealloc_unmap_pages(struct page *page, int numpages) {}
4759static inline unsigned int debug_guardpage_minorder(void) { return 0; }
4760static inline bool debug_guardpage_enabled(void) { return false; }
4761static inline bool page_is_guard(const struct page *page) { return false; }
4762static inline bool set_page_guard(struct zone *zone, struct page *page,
4763 unsigned int order) { return false; }
4764static inline void clear_page_guard(struct zone *zone, struct page *page,
4765 unsigned int order) {}
4766#endif /* CONFIG_DEBUG_PAGEALLOC */
4767
4768#ifndef clear_pages
4769/**
4770 * clear_pages() - clear a page range for kernel-internal use.
4771 * @addr: start address
4772 * @npages: number of pages
4773 *
4774 * Use clear_user_pages() instead when clearing a page range to be
4775 * mapped to user space.
4776 *
4777 * Does absolutely no exception handling.
4778 *
4779 * Note that even though the clearing operation is preemptible, clear_pages()
4780 * does not (and on architectures where it reduces to a few long-running
4781 * instructions, might not be able to) call cond_resched() to check if
4782 * rescheduling is required.
4783 *
4784 * When running under preemptible models this is not a problem. Under
4785 * cooperatively scheduled models, however, the caller is expected to
4786 * limit @npages to no more than PROCESS_PAGES_NON_PREEMPT_BATCH.
4787 */
4788static inline void clear_pages(void *addr, unsigned int npages)
4789{
4790 do {
4791 clear_page(addr);
4792 addr += PAGE_SIZE;
4793 } while (--npages);
4794}
4795#endif
4796
4797#ifndef PROCESS_PAGES_NON_PREEMPT_BATCH
4798#ifdef clear_pages
4799/*
4800 * The architecture defines clear_pages(), and we assume that it is
4801 * generally "fast". So choose a batch size large enough to allow the processor
4802 * headroom for optimizing the operation and yet small enough that we see
4803 * reasonable preemption latency for when this optimization is not possible
4804 * (ex. slow microarchitectures, memory bandwidth saturation.)
4805 *
4806 * With a value of 32MB and assuming a memory bandwidth of ~10GBps, this should
4807 * result in worst case preemption latency of around 3ms when clearing pages.
4808 *
4809 * (See comment above clear_pages() for why preemption latency is a concern
4810 * here.)
4811 */
4812#define PROCESS_PAGES_NON_PREEMPT_BATCH (SZ_32M >> PAGE_SHIFT)
4813#else /* !clear_pages */
4814/*
4815 * The architecture does not provide a clear_pages() implementation. Assume
4816 * that clear_page() -- which clear_pages() will fallback to -- is relatively
4817 * slow and choose a small value for PROCESS_PAGES_NON_PREEMPT_BATCH.
4818 */
4819#define PROCESS_PAGES_NON_PREEMPT_BATCH 1
4820#endif
4821#endif
4822
4823#ifdef __HAVE_ARCH_GATE_AREA
4824extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
4825extern int in_gate_area_no_mm(unsigned long addr);
4826extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
4827#else
4828static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
4829{
4830 return NULL;
4831}
4832static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
4833static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
4834{
4835 return 0;
4836}
4837#endif /* __HAVE_ARCH_GATE_AREA */
4838
4839bool process_shares_mm(const struct task_struct *p, const struct mm_struct *mm);
4840
4841void drop_slab(void);
4842
4843#ifndef CONFIG_MMU
4844#define randomize_va_space 0
4845#else
4846extern int randomize_va_space;
4847#endif
4848
4849const char * arch_vma_name(struct vm_area_struct *vma);
4850#ifdef CONFIG_MMU
4851void print_vma_addr(char *prefix, unsigned long rip);
4852#else
4853static inline void print_vma_addr(char *prefix, unsigned long rip)
4854{
4855}
4856#endif
4857
4858void *sparse_buffer_alloc(unsigned long size);
4859unsigned long section_map_size(void);
4860struct page * __populate_section_memmap(unsigned long pfn,
4861 unsigned long nr_pages, int nid, struct vmem_altmap *altmap,
4862 struct dev_pagemap *pgmap);
4863pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
4864p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
4865pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
4866pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
4867pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node,
4868 struct vmem_altmap *altmap, unsigned long ptpfn,
4869 unsigned long flags);
4870void *vmemmap_alloc_block(unsigned long size, int node);
4871struct vmem_altmap;
4872void *vmemmap_alloc_block_buf(unsigned long size, int node,
4873 struct vmem_altmap *altmap);
4874void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
4875void vmemmap_set_pmd(pmd_t *pmd, void *p, int node,
4876 unsigned long addr, unsigned long next);
4877int vmemmap_check_pmd(pmd_t *pmd, int node,
4878 unsigned long addr, unsigned long next);
4879int vmemmap_populate_basepages(unsigned long start, unsigned long end,
4880 int node, struct vmem_altmap *altmap);
4881int vmemmap_populate_hugepages(unsigned long start, unsigned long end,
4882 int node, struct vmem_altmap *altmap);
4883int vmemmap_populate(unsigned long start, unsigned long end, int node,
4884 struct vmem_altmap *altmap);
4885int vmemmap_populate_hvo(unsigned long start, unsigned long end,
4886 unsigned int order, struct zone *zone,
4887 unsigned long headsize);
4888void vmemmap_wrprotect_hvo(unsigned long start, unsigned long end, int node,
4889 unsigned long headsize);
4890void vmemmap_populate_print_last(void);
4891#ifdef CONFIG_MEMORY_HOTPLUG
4892void vmemmap_free(unsigned long start, unsigned long end,
4893 struct vmem_altmap *altmap);
4894#endif
4895
4896#ifdef CONFIG_SPARSEMEM_VMEMMAP
4897static inline unsigned long vmem_altmap_offset(const struct vmem_altmap *altmap)
4898{
4899 /* number of pfns from base where pfn_to_page() is valid */
4900 if (altmap)
4901 return altmap->reserve + altmap->free;
4902 return 0;
4903}
4904
4905static inline void vmem_altmap_free(struct vmem_altmap *altmap,
4906 unsigned long nr_pfns)
4907{
4908 altmap->alloc -= nr_pfns;
4909}
4910#else
4911static inline unsigned long vmem_altmap_offset(const struct vmem_altmap *altmap)
4912{
4913 return 0;
4914}
4915
4916static inline void vmem_altmap_free(struct vmem_altmap *altmap,
4917 unsigned long nr_pfns)
4918{
4919}
4920#endif
4921
4922#define VMEMMAP_RESERVE_NR 2
4923#ifdef CONFIG_ARCH_WANT_OPTIMIZE_DAX_VMEMMAP
4924static inline bool __vmemmap_can_optimize(struct vmem_altmap *altmap,
4925 struct dev_pagemap *pgmap)
4926{
4927 unsigned long nr_pages;
4928 unsigned long nr_vmemmap_pages;
4929
4930 if (!pgmap || !is_power_of_2(sizeof(struct page)))
4931 return false;
4932
4933 nr_pages = pgmap_vmemmap_nr(pgmap);
4934 nr_vmemmap_pages = ((nr_pages * sizeof(struct page)) >> PAGE_SHIFT);
4935 /*
4936 * For vmemmap optimization with DAX we need minimum 2 vmemmap
4937 * pages. See layout diagram in Documentation/mm/vmemmap_dedup.rst
4938 */
4939 return !altmap && (nr_vmemmap_pages > VMEMMAP_RESERVE_NR);
4940}
4941/*
4942 * If we don't have an architecture override, use the generic rule
4943 */
4944#ifndef vmemmap_can_optimize
4945#define vmemmap_can_optimize __vmemmap_can_optimize
4946#endif
4947
4948#else
4949static inline bool vmemmap_can_optimize(struct vmem_altmap *altmap,
4950 struct dev_pagemap *pgmap)
4951{
4952 return false;
4953}
4954#endif
4955
4956enum mf_flags {
4957 MF_COUNT_INCREASED = 1 << 0,
4958 MF_ACTION_REQUIRED = 1 << 1,
4959 MF_MUST_KILL = 1 << 2,
4960 MF_SOFT_OFFLINE = 1 << 3,
4961 MF_UNPOISON = 1 << 4,
4962 MF_SW_SIMULATED = 1 << 5,
4963 MF_NO_RETRY = 1 << 6,
4964 MF_MEM_PRE_REMOVE = 1 << 7,
4965};
4966int mf_dax_kill_procs(struct address_space *mapping, pgoff_t index,
4967 unsigned long count, int mf_flags);
4968extern int memory_failure(unsigned long pfn, int flags);
4969extern int unpoison_memory(unsigned long pfn);
4970extern atomic_long_t num_poisoned_pages __read_mostly;
4971extern int soft_offline_page(unsigned long pfn, int flags);
4972#ifdef CONFIG_MEMORY_FAILURE
4973/*
4974 * Sysfs entries for memory failure handling statistics.
4975 */
4976extern const struct attribute_group memory_failure_attr_group;
4977extern void memory_failure_queue(unsigned long pfn, int flags);
4978extern int __get_huge_page_for_hwpoison(unsigned long pfn, int flags,
4979 bool *migratable_cleared);
4980void num_poisoned_pages_inc(unsigned long pfn);
4981void num_poisoned_pages_sub(unsigned long pfn, long i);
4982#else
4983static inline void memory_failure_queue(unsigned long pfn, int flags)
4984{
4985}
4986
4987static inline int __get_huge_page_for_hwpoison(unsigned long pfn, int flags,
4988 bool *migratable_cleared)
4989{
4990 return 0;
4991}
4992
4993static inline void num_poisoned_pages_inc(unsigned long pfn)
4994{
4995}
4996
4997static inline void num_poisoned_pages_sub(unsigned long pfn, long i)
4998{
4999}
5000#endif
5001
5002#if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_MEMORY_HOTPLUG)
5003extern void memblk_nr_poison_inc(unsigned long pfn);
5004extern void memblk_nr_poison_sub(unsigned long pfn, long i);
5005#else
5006static inline void memblk_nr_poison_inc(unsigned long pfn)
5007{
5008}
5009
5010static inline void memblk_nr_poison_sub(unsigned long pfn, long i)
5011{
5012}
5013#endif
5014
5015#ifndef arch_memory_failure
5016static inline int arch_memory_failure(unsigned long pfn, int flags)
5017{
5018 return -ENXIO;
5019}
5020#endif
5021
5022#ifndef arch_is_platform_page
5023static inline bool arch_is_platform_page(u64 paddr)
5024{
5025 return false;
5026}
5027#endif
5028
5029/*
5030 * Error handlers for various types of pages.
5031 */
5032enum mf_result {
5033 MF_IGNORED, /* Error: cannot be handled */
5034 MF_FAILED, /* Error: handling failed */
5035 MF_DELAYED, /* Will be handled later */
5036 MF_RECOVERED, /* Successfully recovered */
5037};
5038
5039enum mf_action_page_type {
5040 MF_MSG_KERNEL,
5041 MF_MSG_KERNEL_HIGH_ORDER,
5042 MF_MSG_DIFFERENT_COMPOUND,
5043 MF_MSG_HUGE,
5044 MF_MSG_FREE_HUGE,
5045 MF_MSG_GET_HWPOISON,
5046 MF_MSG_UNMAP_FAILED,
5047 MF_MSG_DIRTY_SWAPCACHE,
5048 MF_MSG_CLEAN_SWAPCACHE,
5049 MF_MSG_DIRTY_MLOCKED_LRU,
5050 MF_MSG_CLEAN_MLOCKED_LRU,
5051 MF_MSG_DIRTY_UNEVICTABLE_LRU,
5052 MF_MSG_CLEAN_UNEVICTABLE_LRU,
5053 MF_MSG_DIRTY_LRU,
5054 MF_MSG_CLEAN_LRU,
5055 MF_MSG_TRUNCATED_LRU,
5056 MF_MSG_BUDDY,
5057 MF_MSG_DAX,
5058 MF_MSG_UNSPLIT_THP,
5059 MF_MSG_ALREADY_POISONED,
5060 MF_MSG_PFN_MAP,
5061 MF_MSG_UNKNOWN,
5062};
5063
5064#if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
5065void folio_zero_user(struct folio *folio, unsigned long addr_hint);
5066int copy_user_large_folio(struct folio *dst, struct folio *src,
5067 unsigned long addr_hint,
5068 struct vm_area_struct *vma);
5069long copy_folio_from_user(struct folio *dst_folio,
5070 const void __user *usr_src,
5071 bool allow_pagefault);
5072
5073#endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
5074
5075#if MAX_NUMNODES > 1
5076void __init setup_nr_node_ids(void);
5077#else
5078static inline void setup_nr_node_ids(void) {}
5079#endif
5080
5081extern int memcmp_pages(struct page *page1, struct page *page2);
5082
5083static inline int pages_identical(struct page *page1, struct page *page2)
5084{
5085 return !memcmp_pages(page1, page2);
5086}
5087
5088#ifdef CONFIG_MAPPING_DIRTY_HELPERS
5089unsigned long clean_record_shared_mapping_range(struct address_space *mapping,
5090 pgoff_t first_index, pgoff_t nr,
5091 pgoff_t bitmap_pgoff,
5092 unsigned long *bitmap,
5093 pgoff_t *start,
5094 pgoff_t *end);
5095
5096unsigned long wp_shared_mapping_range(struct address_space *mapping,
5097 pgoff_t first_index, pgoff_t nr);
5098#endif
5099
5100#ifdef CONFIG_ANON_VMA_NAME
5101int set_anon_vma_name(unsigned long addr, unsigned long size,
5102 const char __user *uname);
5103#else
5104static inline
5105int set_anon_vma_name(unsigned long addr, unsigned long size,
5106 const char __user *uname)
5107{
5108 return -EINVAL;
5109}
5110#endif
5111
5112#ifdef CONFIG_UNACCEPTED_MEMORY
5113
5114bool range_contains_unaccepted_memory(phys_addr_t start, unsigned long size);
5115void accept_memory(phys_addr_t start, unsigned long size);
5116
5117#else
5118
5119static inline bool range_contains_unaccepted_memory(phys_addr_t start,
5120 unsigned long size)
5121{
5122 return false;
5123}
5124
5125static inline void accept_memory(phys_addr_t start, unsigned long size)
5126{
5127}
5128
5129#endif
5130
5131static inline bool pfn_is_unaccepted_memory(unsigned long pfn)
5132{
5133 return range_contains_unaccepted_memory(pfn << PAGE_SHIFT, PAGE_SIZE);
5134}
5135
5136void vma_pgtable_walk_begin(struct vm_area_struct *vma);
5137void vma_pgtable_walk_end(struct vm_area_struct *vma);
5138
5139int reserve_mem_find_by_name(const char *name, phys_addr_t *start, phys_addr_t *size);
5140int reserve_mem_release_by_name(const char *name);
5141
5142#ifdef CONFIG_64BIT
5143int do_mseal(unsigned long start, size_t len_in, unsigned long flags);
5144#else
5145static inline int do_mseal(unsigned long start, size_t len_in, unsigned long flags)
5146{
5147 /* noop on 32 bit */
5148 return 0;
5149}
5150#endif
5151
5152/*
5153 * user_alloc_needs_zeroing checks if a user folio from page allocator needs to
5154 * be zeroed or not.
5155 */
5156static inline bool user_alloc_needs_zeroing(void)
5157{
5158 /*
5159 * for user folios, arch with cache aliasing requires cache flush and
5160 * arc changes folio->flags to make icache coherent with dcache, so
5161 * always return false to make caller use
5162 * clear_user_page()/clear_user_highpage().
5163 */
5164 return cpu_dcache_is_aliasing() || cpu_icache_is_aliasing() ||
5165 !static_branch_maybe(CONFIG_INIT_ON_ALLOC_DEFAULT_ON,
5166 &init_on_alloc);
5167}
5168
5169int arch_get_shadow_stack_status(struct task_struct *t, unsigned long __user *status);
5170int arch_set_shadow_stack_status(struct task_struct *t, unsigned long status);
5171int arch_lock_shadow_stack_status(struct task_struct *t, unsigned long status);
5172
5173/*
5174 * DMA mapping IDs for page_pool
5175 *
5176 * When DMA-mapping a page, page_pool allocates an ID (from an xarray) and
5177 * stashes it in the upper bits of page->pp_magic. Non-PP pages can have
5178 * arbitrary kernel pointers stored in the same field as pp_magic (since
5179 * it overlaps with page->lru.next), so we must ensure that we cannot
5180 * mistake a valid kernel pointer with any of the values we write into this
5181 * field.
5182 *
5183 * On architectures that set POISON_POINTER_DELTA, this is already ensured,
5184 * since this value becomes part of PP_SIGNATURE; meaning we can just use the
5185 * space between the PP_SIGNATURE value (without POISON_POINTER_DELTA), and the
5186 * lowest bits of POISON_POINTER_DELTA. On arches where POISON_POINTER_DELTA is
5187 * 0, we use the lowest bit of PAGE_OFFSET as the boundary if that value is
5188 * known at compile-time.
5189 *
5190 * If the value of PAGE_OFFSET is not known at compile time, or if it is too
5191 * small to leave at least 8 bits available above PP_SIGNATURE, we define the
5192 * number of bits to be 0, which turns off the DMA index tracking altogether
5193 * (see page_pool_register_dma_index()).
5194 */
5195#define PP_DMA_INDEX_SHIFT (1 + __fls(PP_SIGNATURE - POISON_POINTER_DELTA))
5196#if POISON_POINTER_DELTA > 0
5197/* PP_SIGNATURE includes POISON_POINTER_DELTA, so limit the size of the DMA
5198 * index to not overlap with that if set
5199 */
5200#define PP_DMA_INDEX_BITS MIN(32, __ffs(POISON_POINTER_DELTA) - PP_DMA_INDEX_SHIFT)
5201#else
5202/* Use the lowest bit of PAGE_OFFSET if there's at least 8 bits available; see above */
5203#define PP_DMA_INDEX_MIN_OFFSET (1 << (PP_DMA_INDEX_SHIFT + 8))
5204#define PP_DMA_INDEX_BITS ((__builtin_constant_p(PAGE_OFFSET) && \
5205 PAGE_OFFSET >= PP_DMA_INDEX_MIN_OFFSET && \
5206 !(PAGE_OFFSET & (PP_DMA_INDEX_MIN_OFFSET - 1))) ? \
5207 MIN(32, __ffs(PAGE_OFFSET) - PP_DMA_INDEX_SHIFT) : 0)
5208
5209#endif
5210
5211#define PP_DMA_INDEX_MASK GENMASK(PP_DMA_INDEX_BITS + PP_DMA_INDEX_SHIFT - 1, \
5212 PP_DMA_INDEX_SHIFT)
5213
5214#define PAGE_SNAPSHOT_FAITHFUL (1 << 0)
5215#define PAGE_SNAPSHOT_PG_BUDDY (1 << 1)
5216#define PAGE_SNAPSHOT_PG_IDLE (1 << 2)
5217
5218struct page_snapshot {
5219 struct folio folio_snapshot;
5220 struct page page_snapshot;
5221 unsigned long pfn;
5222 unsigned long idx;
5223 unsigned long flags;
5224};
5225
5226static inline bool snapshot_page_is_faithful(const struct page_snapshot *ps)
5227{
5228 return ps->flags & PAGE_SNAPSHOT_FAITHFUL;
5229}
5230
5231void snapshot_page(struct page_snapshot *ps, const struct page *page);
5232
5233void map_anon_folio_pte_nopf(struct folio *folio, pte_t *pte,
5234 struct vm_area_struct *vma, unsigned long addr,
5235 bool uffd_wp);
5236
5237#endif /* _LINUX_MM_H */