xref: /linux/include/linux/mm_types.h (revision beace86e61e465dba204a268ab3f3377153a4973)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _LINUX_MM_TYPES_H
3 #define _LINUX_MM_TYPES_H
4 
5 #include <linux/mm_types_task.h>
6 
7 #include <linux/auxvec.h>
8 #include <linux/kref.h>
9 #include <linux/list.h>
10 #include <linux/spinlock.h>
11 #include <linux/rbtree.h>
12 #include <linux/maple_tree.h>
13 #include <linux/rwsem.h>
14 #include <linux/completion.h>
15 #include <linux/cpumask.h>
16 #include <linux/uprobes.h>
17 #include <linux/rcupdate.h>
18 #include <linux/page-flags-layout.h>
19 #include <linux/workqueue.h>
20 #include <linux/seqlock.h>
21 #include <linux/percpu_counter.h>
22 #include <linux/types.h>
23 
24 #include <asm/mmu.h>
25 
26 #ifndef AT_VECTOR_SIZE_ARCH
27 #define AT_VECTOR_SIZE_ARCH 0
28 #endif
29 #define AT_VECTOR_SIZE (2*(AT_VECTOR_SIZE_ARCH + AT_VECTOR_SIZE_BASE + 1))
30 
31 
32 struct address_space;
33 struct futex_private_hash;
34 struct mem_cgroup;
35 
36 /*
37  * Each physical page in the system has a struct page associated with
38  * it to keep track of whatever it is we are using the page for at the
39  * moment. Note that we have no way to track which tasks are using
40  * a page, though if it is a pagecache page, rmap structures can tell us
41  * who is mapping it.
42  *
43  * If you allocate the page using alloc_pages(), you can use some of the
44  * space in struct page for your own purposes.  The five words in the main
45  * union are available, except for bit 0 of the first word which must be
46  * kept clear.  Many users use this word to store a pointer to an object
47  * which is guaranteed to be aligned.  If you use the same storage as
48  * page->mapping, you must restore it to NULL before freeing the page.
49  *
50  * The mapcount field must not be used for own purposes.
51  *
52  * If you want to use the refcount field, it must be used in such a way
53  * that other CPUs temporarily incrementing and then decrementing the
54  * refcount does not cause problems.  On receiving the page from
55  * alloc_pages(), the refcount will be positive.
56  *
57  * If you allocate pages of order > 0, you can use some of the fields
58  * in each subpage, but you may need to restore some of their values
59  * afterwards.
60  *
61  * SLUB uses cmpxchg_double() to atomically update its freelist and counters.
62  * That requires that freelist & counters in struct slab be adjacent and
63  * double-word aligned. Because struct slab currently just reinterprets the
64  * bits of struct page, we align all struct pages to double-word boundaries,
65  * and ensure that 'freelist' is aligned within struct slab.
66  */
67 #ifdef CONFIG_HAVE_ALIGNED_STRUCT_PAGE
68 #define _struct_page_alignment	__aligned(2 * sizeof(unsigned long))
69 #else
70 #define _struct_page_alignment	__aligned(sizeof(unsigned long))
71 #endif
72 
73 struct page {
74 	unsigned long flags;		/* Atomic flags, some possibly
75 					 * updated asynchronously */
76 	/*
77 	 * Five words (20/40 bytes) are available in this union.
78 	 * WARNING: bit 0 of the first word is used for PageTail(). That
79 	 * means the other users of this union MUST NOT use the bit to
80 	 * avoid collision and false-positive PageTail().
81 	 */
82 	union {
83 		struct {	/* Page cache and anonymous pages */
84 			/**
85 			 * @lru: Pageout list, eg. active_list protected by
86 			 * lruvec->lru_lock.  Sometimes used as a generic list
87 			 * by the page owner.
88 			 */
89 			union {
90 				struct list_head lru;
91 
92 				/* Or, for the Unevictable "LRU list" slot */
93 				struct {
94 					/* Always even, to negate PageTail */
95 					void *__filler;
96 					/* Count page's or folio's mlocks */
97 					unsigned int mlock_count;
98 				};
99 
100 				/* Or, free page */
101 				struct list_head buddy_list;
102 				struct list_head pcp_list;
103 				struct {
104 					struct llist_node pcp_llist;
105 					unsigned int order;
106 				};
107 			};
108 			struct address_space *mapping;
109 			union {
110 				pgoff_t __folio_index;		/* Our offset within mapping. */
111 				unsigned long share;	/* share count for fsdax */
112 			};
113 			/**
114 			 * @private: Mapping-private opaque data.
115 			 * Usually used for buffer_heads if PagePrivate.
116 			 * Used for swp_entry_t if swapcache flag set.
117 			 * Indicates order in the buddy system if PageBuddy.
118 			 */
119 			unsigned long private;
120 		};
121 		struct {	/* page_pool used by netstack */
122 			/**
123 			 * @pp_magic: magic value to avoid recycling non
124 			 * page_pool allocated pages.
125 			 */
126 			unsigned long pp_magic;
127 			struct page_pool *pp;
128 			unsigned long _pp_mapping_pad;
129 			unsigned long dma_addr;
130 			atomic_long_t pp_ref_count;
131 		};
132 		struct {	/* Tail pages of compound page */
133 			unsigned long compound_head;	/* Bit zero is set */
134 		};
135 		struct {	/* ZONE_DEVICE pages */
136 			/*
137 			 * The first word is used for compound_head or folio
138 			 * pgmap
139 			 */
140 			void *_unused_pgmap_compound_head;
141 			void *zone_device_data;
142 			/*
143 			 * ZONE_DEVICE private pages are counted as being
144 			 * mapped so the next 3 words hold the mapping, index,
145 			 * and private fields from the source anonymous or
146 			 * page cache page while the page is migrated to device
147 			 * private memory.
148 			 * ZONE_DEVICE MEMORY_DEVICE_FS_DAX pages also
149 			 * use the mapping, index, and private fields when
150 			 * pmem backed DAX files are mapped.
151 			 */
152 		};
153 
154 		/** @rcu_head: You can use this to free a page by RCU. */
155 		struct rcu_head rcu_head;
156 	};
157 
158 	union {		/* This union is 4 bytes in size. */
159 		/*
160 		 * For head pages of typed folios, the value stored here
161 		 * allows for determining what this page is used for. The
162 		 * tail pages of typed folios will not store a type
163 		 * (page_type == _mapcount == -1).
164 		 *
165 		 * See page-flags.h for a list of page types which are currently
166 		 * stored here.
167 		 *
168 		 * Owners of typed folios may reuse the lower 16 bit of the
169 		 * head page page_type field after setting the page type,
170 		 * but must reset these 16 bit to -1 before clearing the
171 		 * page type.
172 		 */
173 		unsigned int page_type;
174 
175 		/*
176 		 * For pages that are part of non-typed folios for which mappings
177 		 * are tracked via the RMAP, encodes the number of times this page
178 		 * is directly referenced by a page table.
179 		 *
180 		 * Note that the mapcount is always initialized to -1, so that
181 		 * transitions both from it and to it can be tracked, using
182 		 * atomic_inc_and_test() and atomic_add_negative(-1).
183 		 */
184 		atomic_t _mapcount;
185 	};
186 
187 	/* Usage count. *DO NOT USE DIRECTLY*. See page_ref.h */
188 	atomic_t _refcount;
189 
190 #ifdef CONFIG_MEMCG
191 	unsigned long memcg_data;
192 #elif defined(CONFIG_SLAB_OBJ_EXT)
193 	unsigned long _unused_slab_obj_exts;
194 #endif
195 
196 	/*
197 	 * On machines where all RAM is mapped into kernel address space,
198 	 * we can simply calculate the virtual address. On machines with
199 	 * highmem some memory is mapped into kernel virtual memory
200 	 * dynamically, so we need a place to store that address.
201 	 * Note that this field could be 16 bits on x86 ... ;)
202 	 *
203 	 * Architectures with slow multiplication can define
204 	 * WANT_PAGE_VIRTUAL in asm/page.h
205 	 */
206 #if defined(WANT_PAGE_VIRTUAL)
207 	void *virtual;			/* Kernel virtual address (NULL if
208 					   not kmapped, ie. highmem) */
209 #endif /* WANT_PAGE_VIRTUAL */
210 
211 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
212 	int _last_cpupid;
213 #endif
214 
215 #ifdef CONFIG_KMSAN
216 	/*
217 	 * KMSAN metadata for this page:
218 	 *  - shadow page: every bit indicates whether the corresponding
219 	 *    bit of the original page is initialized (0) or not (1);
220 	 *  - origin page: every 4 bytes contain an id of the stack trace
221 	 *    where the uninitialized value was created.
222 	 */
223 	struct page *kmsan_shadow;
224 	struct page *kmsan_origin;
225 #endif
226 } _struct_page_alignment;
227 
228 /*
229  * struct encoded_page - a nonexistent type marking this pointer
230  *
231  * An 'encoded_page' pointer is a pointer to a regular 'struct page', but
232  * with the low bits of the pointer indicating extra context-dependent
233  * information. Only used in mmu_gather handling, and this acts as a type
234  * system check on that use.
235  *
236  * We only really have two guaranteed bits in general, although you could
237  * play with 'struct page' alignment (see CONFIG_HAVE_ALIGNED_STRUCT_PAGE)
238  * for more.
239  *
240  * Use the supplied helper functions to endcode/decode the pointer and bits.
241  */
242 struct encoded_page;
243 
244 #define ENCODED_PAGE_BITS			3ul
245 
246 /* Perform rmap removal after we have flushed the TLB. */
247 #define ENCODED_PAGE_BIT_DELAY_RMAP		1ul
248 
249 /*
250  * The next item in an encoded_page array is the "nr_pages" argument, specifying
251  * the number of consecutive pages starting from this page, that all belong to
252  * the same folio. For example, "nr_pages" corresponds to the number of folio
253  * references that must be dropped. If this bit is not set, "nr_pages" is
254  * implicitly 1.
255  */
256 #define ENCODED_PAGE_BIT_NR_PAGES_NEXT		2ul
257 
258 static __always_inline struct encoded_page *encode_page(struct page *page, unsigned long flags)
259 {
260 	BUILD_BUG_ON(flags > ENCODED_PAGE_BITS);
261 	return (struct encoded_page *)(flags | (unsigned long)page);
262 }
263 
264 static inline unsigned long encoded_page_flags(struct encoded_page *page)
265 {
266 	return ENCODED_PAGE_BITS & (unsigned long)page;
267 }
268 
269 static inline struct page *encoded_page_ptr(struct encoded_page *page)
270 {
271 	return (struct page *)(~ENCODED_PAGE_BITS & (unsigned long)page);
272 }
273 
274 static __always_inline struct encoded_page *encode_nr_pages(unsigned long nr)
275 {
276 	VM_WARN_ON_ONCE((nr << 2) >> 2 != nr);
277 	return (struct encoded_page *)(nr << 2);
278 }
279 
280 static __always_inline unsigned long encoded_nr_pages(struct encoded_page *page)
281 {
282 	return ((unsigned long)page) >> 2;
283 }
284 
285 /*
286  * A swap entry has to fit into a "unsigned long", as the entry is hidden
287  * in the "index" field of the swapper address space.
288  */
289 typedef struct {
290 	unsigned long val;
291 } swp_entry_t;
292 
293 #if defined(CONFIG_MEMCG) || defined(CONFIG_SLAB_OBJ_EXT)
294 /* We have some extra room after the refcount in tail pages. */
295 #define NR_PAGES_IN_LARGE_FOLIO
296 #endif
297 
298 /*
299  * On 32bit, we can cut the required metadata in half, because:
300  * (a) PID_MAX_LIMIT implicitly limits the number of MMs we could ever have,
301  *     so we can limit MM IDs to 15 bit (32767).
302  * (b) We don't expect folios where even a single complete PTE mapping by
303  *     one MM would exceed 15 bits (order-15).
304  */
305 #ifdef CONFIG_64BIT
306 typedef int mm_id_mapcount_t;
307 #define MM_ID_MAPCOUNT_MAX		INT_MAX
308 typedef unsigned int mm_id_t;
309 #else /* !CONFIG_64BIT */
310 typedef short mm_id_mapcount_t;
311 #define MM_ID_MAPCOUNT_MAX		SHRT_MAX
312 typedef unsigned short mm_id_t;
313 #endif /* CONFIG_64BIT */
314 
315 /* We implicitly use the dummy ID for init-mm etc. where we never rmap pages. */
316 #define MM_ID_DUMMY			0
317 #define MM_ID_MIN			(MM_ID_DUMMY + 1)
318 
319 /*
320  * We leave the highest bit of each MM id unused, so we can store a flag
321  * in the highest bit of each folio->_mm_id[].
322  */
323 #define MM_ID_BITS			((sizeof(mm_id_t) * BITS_PER_BYTE) - 1)
324 #define MM_ID_MASK			((1U << MM_ID_BITS) - 1)
325 #define MM_ID_MAX			MM_ID_MASK
326 
327 /*
328  * In order to use bit_spin_lock(), which requires an unsigned long, we
329  * operate on folio->_mm_ids when working on flags.
330  */
331 #define FOLIO_MM_IDS_LOCK_BITNUM	MM_ID_BITS
332 #define FOLIO_MM_IDS_LOCK_BIT		BIT(FOLIO_MM_IDS_LOCK_BITNUM)
333 #define FOLIO_MM_IDS_SHARED_BITNUM	(2 * MM_ID_BITS + 1)
334 #define FOLIO_MM_IDS_SHARED_BIT		BIT(FOLIO_MM_IDS_SHARED_BITNUM)
335 
336 /**
337  * struct folio - Represents a contiguous set of bytes.
338  * @flags: Identical to the page flags.
339  * @lru: Least Recently Used list; tracks how recently this folio was used.
340  * @mlock_count: Number of times this folio has been pinned by mlock().
341  * @mapping: The file this page belongs to, or refers to the anon_vma for
342  *    anonymous memory.
343  * @index: Offset within the file, in units of pages.  For anonymous memory,
344  *    this is the index from the beginning of the mmap.
345  * @share: number of DAX mappings that reference this folio. See
346  *    dax_associate_entry.
347  * @private: Filesystem per-folio data (see folio_attach_private()).
348  * @swap: Used for swp_entry_t if folio_test_swapcache().
349  * @_mapcount: Do not access this member directly.  Use folio_mapcount() to
350  *    find out how many times this folio is mapped by userspace.
351  * @_refcount: Do not access this member directly.  Use folio_ref_count()
352  *    to find how many references there are to this folio.
353  * @memcg_data: Memory Control Group data.
354  * @pgmap: Metadata for ZONE_DEVICE mappings
355  * @virtual: Virtual address in the kernel direct map.
356  * @_last_cpupid: IDs of last CPU and last process that accessed the folio.
357  * @_entire_mapcount: Do not use directly, call folio_entire_mapcount().
358  * @_large_mapcount: Do not use directly, call folio_mapcount().
359  * @_nr_pages_mapped: Do not use outside of rmap and debug code.
360  * @_pincount: Do not use directly, call folio_maybe_dma_pinned().
361  * @_nr_pages: Do not use directly, call folio_nr_pages().
362  * @_mm_id: Do not use outside of rmap code.
363  * @_mm_ids: Do not use outside of rmap code.
364  * @_mm_id_mapcount: Do not use outside of rmap code.
365  * @_hugetlb_subpool: Do not use directly, use accessor in hugetlb.h.
366  * @_hugetlb_cgroup: Do not use directly, use accessor in hugetlb_cgroup.h.
367  * @_hugetlb_cgroup_rsvd: Do not use directly, use accessor in hugetlb_cgroup.h.
368  * @_hugetlb_hwpoison: Do not use directly, call raw_hwp_list_head().
369  * @_deferred_list: Folios to be split under memory pressure.
370  * @_unused_slab_obj_exts: Placeholder to match obj_exts in struct slab.
371  *
372  * A folio is a physically, virtually and logically contiguous set
373  * of bytes.  It is a power-of-two in size, and it is aligned to that
374  * same power-of-two.  It is at least as large as %PAGE_SIZE.  If it is
375  * in the page cache, it is at a file offset which is a multiple of that
376  * power-of-two.  It may be mapped into userspace at an address which is
377  * at an arbitrary page offset, but its kernel virtual address is aligned
378  * to its size.
379  */
380 struct folio {
381 	/* private: don't document the anon union */
382 	union {
383 		struct {
384 	/* public: */
385 			unsigned long flags;
386 			union {
387 				struct list_head lru;
388 	/* private: avoid cluttering the output */
389 				struct {
390 					void *__filler;
391 	/* public: */
392 					unsigned int mlock_count;
393 	/* private: */
394 				};
395 	/* public: */
396 				struct dev_pagemap *pgmap;
397 			};
398 			struct address_space *mapping;
399 			union {
400 				pgoff_t index;
401 				unsigned long share;
402 			};
403 			union {
404 				void *private;
405 				swp_entry_t swap;
406 			};
407 			atomic_t _mapcount;
408 			atomic_t _refcount;
409 #ifdef CONFIG_MEMCG
410 			unsigned long memcg_data;
411 #elif defined(CONFIG_SLAB_OBJ_EXT)
412 			unsigned long _unused_slab_obj_exts;
413 #endif
414 #if defined(WANT_PAGE_VIRTUAL)
415 			void *virtual;
416 #endif
417 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
418 			int _last_cpupid;
419 #endif
420 	/* private: the union with struct page is transitional */
421 		};
422 		struct page page;
423 	};
424 	union {
425 		struct {
426 			unsigned long _flags_1;
427 			unsigned long _head_1;
428 			union {
429 				struct {
430 	/* public: */
431 					atomic_t _large_mapcount;
432 					atomic_t _nr_pages_mapped;
433 #ifdef CONFIG_64BIT
434 					atomic_t _entire_mapcount;
435 					atomic_t _pincount;
436 #endif /* CONFIG_64BIT */
437 					mm_id_mapcount_t _mm_id_mapcount[2];
438 					union {
439 						mm_id_t _mm_id[2];
440 						unsigned long _mm_ids;
441 					};
442 	/* private: the union with struct page is transitional */
443 				};
444 				unsigned long _usable_1[4];
445 			};
446 			atomic_t _mapcount_1;
447 			atomic_t _refcount_1;
448 	/* public: */
449 #ifdef NR_PAGES_IN_LARGE_FOLIO
450 			unsigned int _nr_pages;
451 #endif /* NR_PAGES_IN_LARGE_FOLIO */
452 	/* private: the union with struct page is transitional */
453 		};
454 		struct page __page_1;
455 	};
456 	union {
457 		struct {
458 			unsigned long _flags_2;
459 			unsigned long _head_2;
460 	/* public: */
461 			struct list_head _deferred_list;
462 #ifndef CONFIG_64BIT
463 			atomic_t _entire_mapcount;
464 			atomic_t _pincount;
465 #endif /* !CONFIG_64BIT */
466 	/* private: the union with struct page is transitional */
467 		};
468 		struct page __page_2;
469 	};
470 	union {
471 		struct {
472 			unsigned long _flags_3;
473 			unsigned long _head_3;
474 	/* public: */
475 			void *_hugetlb_subpool;
476 			void *_hugetlb_cgroup;
477 			void *_hugetlb_cgroup_rsvd;
478 			void *_hugetlb_hwpoison;
479 	/* private: the union with struct page is transitional */
480 		};
481 		struct page __page_3;
482 	};
483 };
484 
485 #define FOLIO_MATCH(pg, fl)						\
486 	static_assert(offsetof(struct page, pg) == offsetof(struct folio, fl))
487 FOLIO_MATCH(flags, flags);
488 FOLIO_MATCH(lru, lru);
489 FOLIO_MATCH(mapping, mapping);
490 FOLIO_MATCH(compound_head, lru);
491 FOLIO_MATCH(__folio_index, index);
492 FOLIO_MATCH(private, private);
493 FOLIO_MATCH(_mapcount, _mapcount);
494 FOLIO_MATCH(_refcount, _refcount);
495 #ifdef CONFIG_MEMCG
496 FOLIO_MATCH(memcg_data, memcg_data);
497 #endif
498 #if defined(WANT_PAGE_VIRTUAL)
499 FOLIO_MATCH(virtual, virtual);
500 #endif
501 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
502 FOLIO_MATCH(_last_cpupid, _last_cpupid);
503 #endif
504 #undef FOLIO_MATCH
505 #define FOLIO_MATCH(pg, fl)						\
506 	static_assert(offsetof(struct folio, fl) ==			\
507 			offsetof(struct page, pg) + sizeof(struct page))
508 FOLIO_MATCH(flags, _flags_1);
509 FOLIO_MATCH(compound_head, _head_1);
510 FOLIO_MATCH(_mapcount, _mapcount_1);
511 FOLIO_MATCH(_refcount, _refcount_1);
512 #undef FOLIO_MATCH
513 #define FOLIO_MATCH(pg, fl)						\
514 	static_assert(offsetof(struct folio, fl) ==			\
515 			offsetof(struct page, pg) + 2 * sizeof(struct page))
516 FOLIO_MATCH(flags, _flags_2);
517 FOLIO_MATCH(compound_head, _head_2);
518 #undef FOLIO_MATCH
519 #define FOLIO_MATCH(pg, fl)						\
520 	static_assert(offsetof(struct folio, fl) ==			\
521 			offsetof(struct page, pg) + 3 * sizeof(struct page))
522 FOLIO_MATCH(flags, _flags_3);
523 FOLIO_MATCH(compound_head, _head_3);
524 #undef FOLIO_MATCH
525 
526 /**
527  * struct ptdesc -    Memory descriptor for page tables.
528  * @__page_flags:     Same as page flags. Powerpc only.
529  * @pt_rcu_head:      For freeing page table pages.
530  * @pt_list:          List of used page tables. Used for s390 gmap shadow pages
531  *                    (which are not linked into the user page tables) and x86
532  *                    pgds.
533  * @_pt_pad_1:        Padding that aliases with page's compound head.
534  * @pmd_huge_pte:     Protected by ptdesc->ptl, used for THPs.
535  * @__page_mapping:   Aliases with page->mapping. Unused for page tables.
536  * @pt_index:         Used for s390 gmap.
537  * @pt_mm:            Used for x86 pgds.
538  * @pt_frag_refcount: For fragmented page table tracking. Powerpc only.
539  * @pt_share_count:   Used for HugeTLB PMD page table share count.
540  * @_pt_pad_2:        Padding to ensure proper alignment.
541  * @ptl:              Lock for the page table.
542  * @__page_type:      Same as page->page_type. Unused for page tables.
543  * @__page_refcount:  Same as page refcount.
544  * @pt_memcg_data:    Memcg data. Tracked for page tables here.
545  *
546  * This struct overlays struct page for now. Do not modify without a good
547  * understanding of the issues.
548  */
549 struct ptdesc {
550 	unsigned long __page_flags;
551 
552 	union {
553 		struct rcu_head pt_rcu_head;
554 		struct list_head pt_list;
555 		struct {
556 			unsigned long _pt_pad_1;
557 			pgtable_t pmd_huge_pte;
558 		};
559 	};
560 	unsigned long __page_mapping;
561 
562 	union {
563 		pgoff_t pt_index;
564 		struct mm_struct *pt_mm;
565 		atomic_t pt_frag_refcount;
566 #ifdef CONFIG_HUGETLB_PMD_PAGE_TABLE_SHARING
567 		atomic_t pt_share_count;
568 #endif
569 	};
570 
571 	union {
572 		unsigned long _pt_pad_2;
573 #if ALLOC_SPLIT_PTLOCKS
574 		spinlock_t *ptl;
575 #else
576 		spinlock_t ptl;
577 #endif
578 	};
579 	unsigned int __page_type;
580 	atomic_t __page_refcount;
581 #ifdef CONFIG_MEMCG
582 	unsigned long pt_memcg_data;
583 #endif
584 };
585 
586 #define TABLE_MATCH(pg, pt)						\
587 	static_assert(offsetof(struct page, pg) == offsetof(struct ptdesc, pt))
588 TABLE_MATCH(flags, __page_flags);
589 TABLE_MATCH(compound_head, pt_list);
590 TABLE_MATCH(compound_head, _pt_pad_1);
591 TABLE_MATCH(mapping, __page_mapping);
592 TABLE_MATCH(__folio_index, pt_index);
593 TABLE_MATCH(rcu_head, pt_rcu_head);
594 TABLE_MATCH(page_type, __page_type);
595 TABLE_MATCH(_refcount, __page_refcount);
596 #ifdef CONFIG_MEMCG
597 TABLE_MATCH(memcg_data, pt_memcg_data);
598 #endif
599 #undef TABLE_MATCH
600 static_assert(sizeof(struct ptdesc) <= sizeof(struct page));
601 
602 #define ptdesc_page(pt)			(_Generic((pt),			\
603 	const struct ptdesc *:		(const struct page *)(pt),	\
604 	struct ptdesc *:		(struct page *)(pt)))
605 
606 #define ptdesc_folio(pt)		(_Generic((pt),			\
607 	const struct ptdesc *:		(const struct folio *)(pt),	\
608 	struct ptdesc *:		(struct folio *)(pt)))
609 
610 #define page_ptdesc(p)			(_Generic((p),			\
611 	const struct page *:		(const struct ptdesc *)(p),	\
612 	struct page *:			(struct ptdesc *)(p)))
613 
614 #ifdef CONFIG_HUGETLB_PMD_PAGE_TABLE_SHARING
615 static inline void ptdesc_pmd_pts_init(struct ptdesc *ptdesc)
616 {
617 	atomic_set(&ptdesc->pt_share_count, 0);
618 }
619 
620 static inline void ptdesc_pmd_pts_inc(struct ptdesc *ptdesc)
621 {
622 	atomic_inc(&ptdesc->pt_share_count);
623 }
624 
625 static inline void ptdesc_pmd_pts_dec(struct ptdesc *ptdesc)
626 {
627 	atomic_dec(&ptdesc->pt_share_count);
628 }
629 
630 static inline int ptdesc_pmd_pts_count(struct ptdesc *ptdesc)
631 {
632 	return atomic_read(&ptdesc->pt_share_count);
633 }
634 #else
635 static inline void ptdesc_pmd_pts_init(struct ptdesc *ptdesc)
636 {
637 }
638 #endif
639 
640 /*
641  * Used for sizing the vmemmap region on some architectures
642  */
643 #define STRUCT_PAGE_MAX_SHIFT	(order_base_2(sizeof(struct page)))
644 
645 /*
646  * page_private can be used on tail pages.  However, PagePrivate is only
647  * checked by the VM on the head page.  So page_private on the tail pages
648  * should be used for data that's ancillary to the head page (eg attaching
649  * buffer heads to tail pages after attaching buffer heads to the head page)
650  */
651 #define page_private(page)		((page)->private)
652 
653 static inline void set_page_private(struct page *page, unsigned long private)
654 {
655 	page->private = private;
656 }
657 
658 static inline void *folio_get_private(struct folio *folio)
659 {
660 	return folio->private;
661 }
662 
663 typedef unsigned long vm_flags_t;
664 
665 /*
666  * freeptr_t represents a SLUB freelist pointer, which might be encoded
667  * and not dereferenceable if CONFIG_SLAB_FREELIST_HARDENED is enabled.
668  */
669 typedef struct { unsigned long v; } freeptr_t;
670 
671 /*
672  * A region containing a mapping of a non-memory backed file under NOMMU
673  * conditions.  These are held in a global tree and are pinned by the VMAs that
674  * map parts of them.
675  */
676 struct vm_region {
677 	struct rb_node	vm_rb;		/* link in global region tree */
678 	vm_flags_t	vm_flags;	/* VMA vm_flags */
679 	unsigned long	vm_start;	/* start address of region */
680 	unsigned long	vm_end;		/* region initialised to here */
681 	unsigned long	vm_top;		/* region allocated to here */
682 	unsigned long	vm_pgoff;	/* the offset in vm_file corresponding to vm_start */
683 	struct file	*vm_file;	/* the backing file or NULL */
684 
685 	int		vm_usage;	/* region usage count (access under nommu_region_sem) */
686 	bool		vm_icache_flushed : 1; /* true if the icache has been flushed for
687 						* this region */
688 };
689 
690 #ifdef CONFIG_USERFAULTFD
691 #define NULL_VM_UFFD_CTX ((struct vm_userfaultfd_ctx) { NULL, })
692 struct vm_userfaultfd_ctx {
693 	struct userfaultfd_ctx *ctx;
694 };
695 #else /* CONFIG_USERFAULTFD */
696 #define NULL_VM_UFFD_CTX ((struct vm_userfaultfd_ctx) {})
697 struct vm_userfaultfd_ctx {};
698 #endif /* CONFIG_USERFAULTFD */
699 
700 struct anon_vma_name {
701 	struct kref kref;
702 	/* The name needs to be at the end because it is dynamically sized. */
703 	char name[];
704 };
705 
706 #ifdef CONFIG_ANON_VMA_NAME
707 /*
708  * mmap_lock should be read-locked when calling anon_vma_name(). Caller should
709  * either keep holding the lock while using the returned pointer or it should
710  * raise anon_vma_name refcount before releasing the lock.
711  */
712 struct anon_vma_name *anon_vma_name(struct vm_area_struct *vma);
713 struct anon_vma_name *anon_vma_name_alloc(const char *name);
714 void anon_vma_name_free(struct kref *kref);
715 #else /* CONFIG_ANON_VMA_NAME */
716 static inline struct anon_vma_name *anon_vma_name(struct vm_area_struct *vma)
717 {
718 	return NULL;
719 }
720 
721 static inline struct anon_vma_name *anon_vma_name_alloc(const char *name)
722 {
723 	return NULL;
724 }
725 #endif
726 
727 #define VMA_LOCK_OFFSET	0x40000000
728 #define VMA_REF_LIMIT	(VMA_LOCK_OFFSET - 1)
729 
730 struct vma_numab_state {
731 	/*
732 	 * Initialised as time in 'jiffies' after which VMA
733 	 * should be scanned.  Delays first scan of new VMA by at
734 	 * least sysctl_numa_balancing_scan_delay:
735 	 */
736 	unsigned long next_scan;
737 
738 	/*
739 	 * Time in jiffies when pids_active[] is reset to
740 	 * detect phase change behaviour:
741 	 */
742 	unsigned long pids_active_reset;
743 
744 	/*
745 	 * Approximate tracking of PIDs that trapped a NUMA hinting
746 	 * fault. May produce false positives due to hash collisions.
747 	 *
748 	 *   [0] Previous PID tracking
749 	 *   [1] Current PID tracking
750 	 *
751 	 * Window moves after next_pid_reset has expired approximately
752 	 * every VMA_PID_RESET_PERIOD jiffies:
753 	 */
754 	unsigned long pids_active[2];
755 
756 	/* MM scan sequence ID when scan first started after VMA creation */
757 	int start_scan_seq;
758 
759 	/*
760 	 * MM scan sequence ID when the VMA was last completely scanned.
761 	 * A VMA is not eligible for scanning if prev_scan_seq == numa_scan_seq
762 	 */
763 	int prev_scan_seq;
764 };
765 
766 #ifdef __HAVE_PFNMAP_TRACKING
767 struct pfnmap_track_ctx {
768 	struct kref kref;
769 	unsigned long pfn;
770 	unsigned long size;	/* in bytes */
771 };
772 #endif
773 
774 /*
775  * Describes a VMA that is about to be mmap()'ed. Drivers may choose to
776  * manipulate mutable fields which will cause those fields to be updated in the
777  * resultant VMA.
778  *
779  * Helper functions are not required for manipulating any field.
780  */
781 struct vm_area_desc {
782 	/* Immutable state. */
783 	struct mm_struct *mm;
784 	unsigned long start;
785 	unsigned long end;
786 
787 	/* Mutable fields. Populated with initial state. */
788 	pgoff_t pgoff;
789 	struct file *file;
790 	vm_flags_t vm_flags;
791 	pgprot_t page_prot;
792 
793 	/* Write-only fields. */
794 	const struct vm_operations_struct *vm_ops;
795 	void *private_data;
796 };
797 
798 /*
799  * This struct describes a virtual memory area. There is one of these
800  * per VM-area/task. A VM area is any part of the process virtual memory
801  * space that has a special rule for the page-fault handlers (ie a shared
802  * library, the executable area etc).
803  *
804  * Only explicitly marked struct members may be accessed by RCU readers before
805  * getting a stable reference.
806  *
807  * WARNING: when adding new members, please update vm_area_init_from() to copy
808  * them during vm_area_struct content duplication.
809  */
810 struct vm_area_struct {
811 	/* The first cache line has the info for VMA tree walking. */
812 
813 	union {
814 		struct {
815 			/* VMA covers [vm_start; vm_end) addresses within mm */
816 			unsigned long vm_start;
817 			unsigned long vm_end;
818 		};
819 		freeptr_t vm_freeptr; /* Pointer used by SLAB_TYPESAFE_BY_RCU */
820 	};
821 
822 	/*
823 	 * The address space we belong to.
824 	 * Unstable RCU readers are allowed to read this.
825 	 */
826 	struct mm_struct *vm_mm;
827 	pgprot_t vm_page_prot;          /* Access permissions of this VMA. */
828 
829 	/*
830 	 * Flags, see mm.h.
831 	 * To modify use vm_flags_{init|reset|set|clear|mod} functions.
832 	 */
833 	union {
834 		const vm_flags_t vm_flags;
835 		vm_flags_t __private __vm_flags;
836 	};
837 
838 #ifdef CONFIG_PER_VMA_LOCK
839 	/*
840 	 * Can only be written (using WRITE_ONCE()) while holding both:
841 	 *  - mmap_lock (in write mode)
842 	 *  - vm_refcnt bit at VMA_LOCK_OFFSET is set
843 	 * Can be read reliably while holding one of:
844 	 *  - mmap_lock (in read or write mode)
845 	 *  - vm_refcnt bit at VMA_LOCK_OFFSET is set or vm_refcnt > 1
846 	 * Can be read unreliably (using READ_ONCE()) for pessimistic bailout
847 	 * while holding nothing (except RCU to keep the VMA struct allocated).
848 	 *
849 	 * This sequence counter is explicitly allowed to overflow; sequence
850 	 * counter reuse can only lead to occasional unnecessary use of the
851 	 * slowpath.
852 	 */
853 	unsigned int vm_lock_seq;
854 #endif
855 	/*
856 	 * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma
857 	 * list, after a COW of one of the file pages.	A MAP_SHARED vma
858 	 * can only be in the i_mmap tree.  An anonymous MAP_PRIVATE, stack
859 	 * or brk vma (with NULL file) can only be in an anon_vma list.
860 	 */
861 	struct list_head anon_vma_chain; /* Serialized by mmap_lock &
862 					  * page_table_lock */
863 	struct anon_vma *anon_vma;	/* Serialized by page_table_lock */
864 
865 	/* Function pointers to deal with this struct. */
866 	const struct vm_operations_struct *vm_ops;
867 
868 	/* Information about our backing store: */
869 	unsigned long vm_pgoff;		/* Offset (within vm_file) in PAGE_SIZE
870 					   units */
871 	struct file * vm_file;		/* File we map to (can be NULL). */
872 	void * vm_private_data;		/* was vm_pte (shared mem) */
873 
874 #ifdef CONFIG_SWAP
875 	atomic_long_t swap_readahead_info;
876 #endif
877 #ifndef CONFIG_MMU
878 	struct vm_region *vm_region;	/* NOMMU mapping region */
879 #endif
880 #ifdef CONFIG_NUMA
881 	struct mempolicy *vm_policy;	/* NUMA policy for the VMA */
882 #endif
883 #ifdef CONFIG_NUMA_BALANCING
884 	struct vma_numab_state *numab_state;	/* NUMA Balancing state */
885 #endif
886 #ifdef CONFIG_PER_VMA_LOCK
887 	/* Unstable RCU readers are allowed to read this. */
888 	refcount_t vm_refcnt ____cacheline_aligned_in_smp;
889 #ifdef CONFIG_DEBUG_LOCK_ALLOC
890 	struct lockdep_map vmlock_dep_map;
891 #endif
892 #endif
893 	/*
894 	 * For areas with an address space and backing store,
895 	 * linkage into the address_space->i_mmap interval tree.
896 	 *
897 	 */
898 	struct {
899 		struct rb_node rb;
900 		unsigned long rb_subtree_last;
901 	} shared;
902 #ifdef CONFIG_ANON_VMA_NAME
903 	/*
904 	 * For private and shared anonymous mappings, a pointer to a null
905 	 * terminated string containing the name given to the vma, or NULL if
906 	 * unnamed. Serialized by mmap_lock. Use anon_vma_name to access.
907 	 */
908 	struct anon_vma_name *anon_name;
909 #endif
910 	struct vm_userfaultfd_ctx vm_userfaultfd_ctx;
911 #ifdef __HAVE_PFNMAP_TRACKING
912 	struct pfnmap_track_ctx *pfnmap_track_ctx;
913 #endif
914 } __randomize_layout;
915 
916 #ifdef CONFIG_NUMA
917 #define vma_policy(vma) ((vma)->vm_policy)
918 #else
919 #define vma_policy(vma) NULL
920 #endif
921 
922 #ifdef CONFIG_SCHED_MM_CID
923 struct mm_cid {
924 	u64 time;
925 	int cid;
926 	int recent_cid;
927 };
928 #endif
929 
930 struct kioctx_table;
931 struct iommu_mm_data;
932 struct mm_struct {
933 	struct {
934 		/*
935 		 * Fields which are often written to are placed in a separate
936 		 * cache line.
937 		 */
938 		struct {
939 			/**
940 			 * @mm_count: The number of references to &struct
941 			 * mm_struct (@mm_users count as 1).
942 			 *
943 			 * Use mmgrab()/mmdrop() to modify. When this drops to
944 			 * 0, the &struct mm_struct is freed.
945 			 */
946 			atomic_t mm_count;
947 		} ____cacheline_aligned_in_smp;
948 
949 		struct maple_tree mm_mt;
950 
951 		unsigned long mmap_base;	/* base of mmap area */
952 		unsigned long mmap_legacy_base;	/* base of mmap area in bottom-up allocations */
953 #ifdef CONFIG_HAVE_ARCH_COMPAT_MMAP_BASES
954 		/* Base addresses for compatible mmap() */
955 		unsigned long mmap_compat_base;
956 		unsigned long mmap_compat_legacy_base;
957 #endif
958 		unsigned long task_size;	/* size of task vm space */
959 		pgd_t * pgd;
960 
961 #ifdef CONFIG_MEMBARRIER
962 		/**
963 		 * @membarrier_state: Flags controlling membarrier behavior.
964 		 *
965 		 * This field is close to @pgd to hopefully fit in the same
966 		 * cache-line, which needs to be touched by switch_mm().
967 		 */
968 		atomic_t membarrier_state;
969 #endif
970 
971 		/**
972 		 * @mm_users: The number of users including userspace.
973 		 *
974 		 * Use mmget()/mmget_not_zero()/mmput() to modify. When this
975 		 * drops to 0 (i.e. when the task exits and there are no other
976 		 * temporary reference holders), we also release a reference on
977 		 * @mm_count (which may then free the &struct mm_struct if
978 		 * @mm_count also drops to 0).
979 		 */
980 		atomic_t mm_users;
981 
982 #ifdef CONFIG_SCHED_MM_CID
983 		/**
984 		 * @pcpu_cid: Per-cpu current cid.
985 		 *
986 		 * Keep track of the currently allocated mm_cid for each cpu.
987 		 * The per-cpu mm_cid values are serialized by their respective
988 		 * runqueue locks.
989 		 */
990 		struct mm_cid __percpu *pcpu_cid;
991 		/*
992 		 * @mm_cid_next_scan: Next mm_cid scan (in jiffies).
993 		 *
994 		 * When the next mm_cid scan is due (in jiffies).
995 		 */
996 		unsigned long mm_cid_next_scan;
997 		/**
998 		 * @nr_cpus_allowed: Number of CPUs allowed for mm.
999 		 *
1000 		 * Number of CPUs allowed in the union of all mm's
1001 		 * threads allowed CPUs.
1002 		 */
1003 		unsigned int nr_cpus_allowed;
1004 		/**
1005 		 * @max_nr_cid: Maximum number of allowed concurrency
1006 		 *              IDs allocated.
1007 		 *
1008 		 * Track the highest number of allowed concurrency IDs
1009 		 * allocated for the mm.
1010 		 */
1011 		atomic_t max_nr_cid;
1012 		/**
1013 		 * @cpus_allowed_lock: Lock protecting mm cpus_allowed.
1014 		 *
1015 		 * Provide mutual exclusion for mm cpus_allowed and
1016 		 * mm nr_cpus_allowed updates.
1017 		 */
1018 		raw_spinlock_t cpus_allowed_lock;
1019 #endif
1020 #ifdef CONFIG_MMU
1021 		atomic_long_t pgtables_bytes;	/* size of all page tables */
1022 #endif
1023 		int map_count;			/* number of VMAs */
1024 
1025 		spinlock_t page_table_lock; /* Protects page tables and some
1026 					     * counters
1027 					     */
1028 		/*
1029 		 * With some kernel config, the current mmap_lock's offset
1030 		 * inside 'mm_struct' is at 0x120, which is very optimal, as
1031 		 * its two hot fields 'count' and 'owner' sit in 2 different
1032 		 * cachelines,  and when mmap_lock is highly contended, both
1033 		 * of the 2 fields will be accessed frequently, current layout
1034 		 * will help to reduce cache bouncing.
1035 		 *
1036 		 * So please be careful with adding new fields before
1037 		 * mmap_lock, which can easily push the 2 fields into one
1038 		 * cacheline.
1039 		 */
1040 		struct rw_semaphore mmap_lock;
1041 
1042 		struct list_head mmlist; /* List of maybe swapped mm's.	These
1043 					  * are globally strung together off
1044 					  * init_mm.mmlist, and are protected
1045 					  * by mmlist_lock
1046 					  */
1047 #ifdef CONFIG_PER_VMA_LOCK
1048 		struct rcuwait vma_writer_wait;
1049 		/*
1050 		 * This field has lock-like semantics, meaning it is sometimes
1051 		 * accessed with ACQUIRE/RELEASE semantics.
1052 		 * Roughly speaking, incrementing the sequence number is
1053 		 * equivalent to releasing locks on VMAs; reading the sequence
1054 		 * number can be part of taking a read lock on a VMA.
1055 		 * Incremented every time mmap_lock is write-locked/unlocked.
1056 		 * Initialized to 0, therefore odd values indicate mmap_lock
1057 		 * is write-locked and even values that it's released.
1058 		 *
1059 		 * Can be modified under write mmap_lock using RELEASE
1060 		 * semantics.
1061 		 * Can be read with no other protection when holding write
1062 		 * mmap_lock.
1063 		 * Can be read with ACQUIRE semantics if not holding write
1064 		 * mmap_lock.
1065 		 */
1066 		seqcount_t mm_lock_seq;
1067 #endif
1068 #ifdef CONFIG_FUTEX_PRIVATE_HASH
1069 		struct mutex			futex_hash_lock;
1070 		struct futex_private_hash	__rcu *futex_phash;
1071 		struct futex_private_hash	*futex_phash_new;
1072 		/* futex-ref */
1073 		unsigned long			futex_batches;
1074 		struct rcu_head			futex_rcu;
1075 		atomic_long_t			futex_atomic;
1076 		unsigned int			__percpu *futex_ref;
1077 #endif
1078 
1079 		unsigned long hiwater_rss; /* High-watermark of RSS usage */
1080 		unsigned long hiwater_vm;  /* High-water virtual memory usage */
1081 
1082 		unsigned long total_vm;	   /* Total pages mapped */
1083 		unsigned long locked_vm;   /* Pages that have PG_mlocked set */
1084 		atomic64_t    pinned_vm;   /* Refcount permanently increased */
1085 		unsigned long data_vm;	   /* VM_WRITE & ~VM_SHARED & ~VM_STACK */
1086 		unsigned long exec_vm;	   /* VM_EXEC & ~VM_WRITE & ~VM_STACK */
1087 		unsigned long stack_vm;	   /* VM_STACK */
1088 		vm_flags_t def_flags;
1089 
1090 		/**
1091 		 * @write_protect_seq: Locked when any thread is write
1092 		 * protecting pages mapped by this mm to enforce a later COW,
1093 		 * for instance during page table copying for fork().
1094 		 */
1095 		seqcount_t write_protect_seq;
1096 
1097 		spinlock_t arg_lock; /* protect the below fields */
1098 
1099 		unsigned long start_code, end_code, start_data, end_data;
1100 		unsigned long start_brk, brk, start_stack;
1101 		unsigned long arg_start, arg_end, env_start, env_end;
1102 
1103 		unsigned long saved_auxv[AT_VECTOR_SIZE]; /* for /proc/PID/auxv */
1104 
1105 		struct percpu_counter rss_stat[NR_MM_COUNTERS];
1106 
1107 		struct linux_binfmt *binfmt;
1108 
1109 		/* Architecture-specific MM context */
1110 		mm_context_t context;
1111 
1112 		unsigned long flags; /* Must use atomic bitops to access */
1113 
1114 #ifdef CONFIG_AIO
1115 		spinlock_t			ioctx_lock;
1116 		struct kioctx_table __rcu	*ioctx_table;
1117 #endif
1118 #ifdef CONFIG_MEMCG
1119 		/*
1120 		 * "owner" points to a task that is regarded as the canonical
1121 		 * user/owner of this mm. All of the following must be true in
1122 		 * order for it to be changed:
1123 		 *
1124 		 * current == mm->owner
1125 		 * current->mm != mm
1126 		 * new_owner->mm == mm
1127 		 * new_owner->alloc_lock is held
1128 		 */
1129 		struct task_struct __rcu *owner;
1130 #endif
1131 		struct user_namespace *user_ns;
1132 
1133 		/* store ref to file /proc/<pid>/exe symlink points to */
1134 		struct file __rcu *exe_file;
1135 #ifdef CONFIG_MMU_NOTIFIER
1136 		struct mmu_notifier_subscriptions *notifier_subscriptions;
1137 #endif
1138 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !defined(CONFIG_SPLIT_PMD_PTLOCKS)
1139 		pgtable_t pmd_huge_pte; /* protected by page_table_lock */
1140 #endif
1141 #ifdef CONFIG_NUMA_BALANCING
1142 		/*
1143 		 * numa_next_scan is the next time that PTEs will be remapped
1144 		 * PROT_NONE to trigger NUMA hinting faults; such faults gather
1145 		 * statistics and migrate pages to new nodes if necessary.
1146 		 */
1147 		unsigned long numa_next_scan;
1148 
1149 		/* Restart point for scanning and remapping PTEs. */
1150 		unsigned long numa_scan_offset;
1151 
1152 		/* numa_scan_seq prevents two threads remapping PTEs. */
1153 		int numa_scan_seq;
1154 #endif
1155 		/*
1156 		 * An operation with batched TLB flushing is going on. Anything
1157 		 * that can move process memory needs to flush the TLB when
1158 		 * moving a PROT_NONE mapped page.
1159 		 */
1160 		atomic_t tlb_flush_pending;
1161 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
1162 		/* See flush_tlb_batched_pending() */
1163 		atomic_t tlb_flush_batched;
1164 #endif
1165 		struct uprobes_state uprobes_state;
1166 #ifdef CONFIG_PREEMPT_RT
1167 		struct rcu_head delayed_drop;
1168 #endif
1169 #ifdef CONFIG_HUGETLB_PAGE
1170 		atomic_long_t hugetlb_usage;
1171 #endif
1172 		struct work_struct async_put_work;
1173 
1174 #ifdef CONFIG_IOMMU_MM_DATA
1175 		struct iommu_mm_data *iommu_mm;
1176 #endif
1177 #ifdef CONFIG_KSM
1178 		/*
1179 		 * Represent how many pages of this process are involved in KSM
1180 		 * merging (not including ksm_zero_pages).
1181 		 */
1182 		unsigned long ksm_merging_pages;
1183 		/*
1184 		 * Represent how many pages are checked for ksm merging
1185 		 * including merged and not merged.
1186 		 */
1187 		unsigned long ksm_rmap_items;
1188 		/*
1189 		 * Represent how many empty pages are merged with kernel zero
1190 		 * pages when enabling KSM use_zero_pages.
1191 		 */
1192 		atomic_long_t ksm_zero_pages;
1193 #endif /* CONFIG_KSM */
1194 #ifdef CONFIG_LRU_GEN_WALKS_MMU
1195 		struct {
1196 			/* this mm_struct is on lru_gen_mm_list */
1197 			struct list_head list;
1198 			/*
1199 			 * Set when switching to this mm_struct, as a hint of
1200 			 * whether it has been used since the last time per-node
1201 			 * page table walkers cleared the corresponding bits.
1202 			 */
1203 			unsigned long bitmap;
1204 #ifdef CONFIG_MEMCG
1205 			/* points to the memcg of "owner" above */
1206 			struct mem_cgroup *memcg;
1207 #endif
1208 		} lru_gen;
1209 #endif /* CONFIG_LRU_GEN_WALKS_MMU */
1210 #ifdef CONFIG_MM_ID
1211 		mm_id_t mm_id;
1212 #endif /* CONFIG_MM_ID */
1213 	} __randomize_layout;
1214 
1215 	/*
1216 	 * The mm_cpumask needs to be at the end of mm_struct, because it
1217 	 * is dynamically sized based on nr_cpu_ids.
1218 	 */
1219 	unsigned long cpu_bitmap[];
1220 };
1221 
1222 #define MM_MT_FLAGS	(MT_FLAGS_ALLOC_RANGE | MT_FLAGS_LOCK_EXTERN | \
1223 			 MT_FLAGS_USE_RCU)
1224 extern struct mm_struct init_mm;
1225 
1226 /* Pointer magic because the dynamic array size confuses some compilers. */
1227 static inline void mm_init_cpumask(struct mm_struct *mm)
1228 {
1229 	unsigned long cpu_bitmap = (unsigned long)mm;
1230 
1231 	cpu_bitmap += offsetof(struct mm_struct, cpu_bitmap);
1232 	cpumask_clear((struct cpumask *)cpu_bitmap);
1233 }
1234 
1235 /* Future-safe accessor for struct mm_struct's cpu_vm_mask. */
1236 static inline cpumask_t *mm_cpumask(struct mm_struct *mm)
1237 {
1238 	return (struct cpumask *)&mm->cpu_bitmap;
1239 }
1240 
1241 #ifdef CONFIG_LRU_GEN
1242 
1243 struct lru_gen_mm_list {
1244 	/* mm_struct list for page table walkers */
1245 	struct list_head fifo;
1246 	/* protects the list above */
1247 	spinlock_t lock;
1248 };
1249 
1250 #endif /* CONFIG_LRU_GEN */
1251 
1252 #ifdef CONFIG_LRU_GEN_WALKS_MMU
1253 
1254 void lru_gen_add_mm(struct mm_struct *mm);
1255 void lru_gen_del_mm(struct mm_struct *mm);
1256 void lru_gen_migrate_mm(struct mm_struct *mm);
1257 
1258 static inline void lru_gen_init_mm(struct mm_struct *mm)
1259 {
1260 	INIT_LIST_HEAD(&mm->lru_gen.list);
1261 	mm->lru_gen.bitmap = 0;
1262 #ifdef CONFIG_MEMCG
1263 	mm->lru_gen.memcg = NULL;
1264 #endif
1265 }
1266 
1267 static inline void lru_gen_use_mm(struct mm_struct *mm)
1268 {
1269 	/*
1270 	 * When the bitmap is set, page reclaim knows this mm_struct has been
1271 	 * used since the last time it cleared the bitmap. So it might be worth
1272 	 * walking the page tables of this mm_struct to clear the accessed bit.
1273 	 */
1274 	WRITE_ONCE(mm->lru_gen.bitmap, -1);
1275 }
1276 
1277 #else /* !CONFIG_LRU_GEN_WALKS_MMU */
1278 
1279 static inline void lru_gen_add_mm(struct mm_struct *mm)
1280 {
1281 }
1282 
1283 static inline void lru_gen_del_mm(struct mm_struct *mm)
1284 {
1285 }
1286 
1287 static inline void lru_gen_migrate_mm(struct mm_struct *mm)
1288 {
1289 }
1290 
1291 static inline void lru_gen_init_mm(struct mm_struct *mm)
1292 {
1293 }
1294 
1295 static inline void lru_gen_use_mm(struct mm_struct *mm)
1296 {
1297 }
1298 
1299 #endif /* CONFIG_LRU_GEN_WALKS_MMU */
1300 
1301 struct vma_iterator {
1302 	struct ma_state mas;
1303 };
1304 
1305 #define VMA_ITERATOR(name, __mm, __addr)				\
1306 	struct vma_iterator name = {					\
1307 		.mas = {						\
1308 			.tree = &(__mm)->mm_mt,				\
1309 			.index = __addr,				\
1310 			.node = NULL,					\
1311 			.status = ma_start,				\
1312 		},							\
1313 	}
1314 
1315 static inline void vma_iter_init(struct vma_iterator *vmi,
1316 		struct mm_struct *mm, unsigned long addr)
1317 {
1318 	mas_init(&vmi->mas, &mm->mm_mt, addr);
1319 }
1320 
1321 #ifdef CONFIG_SCHED_MM_CID
1322 
1323 enum mm_cid_state {
1324 	MM_CID_UNSET = -1U,		/* Unset state has lazy_put flag set. */
1325 	MM_CID_LAZY_PUT = (1U << 31),
1326 };
1327 
1328 static inline bool mm_cid_is_unset(int cid)
1329 {
1330 	return cid == MM_CID_UNSET;
1331 }
1332 
1333 static inline bool mm_cid_is_lazy_put(int cid)
1334 {
1335 	return !mm_cid_is_unset(cid) && (cid & MM_CID_LAZY_PUT);
1336 }
1337 
1338 static inline bool mm_cid_is_valid(int cid)
1339 {
1340 	return !(cid & MM_CID_LAZY_PUT);
1341 }
1342 
1343 static inline int mm_cid_set_lazy_put(int cid)
1344 {
1345 	return cid | MM_CID_LAZY_PUT;
1346 }
1347 
1348 static inline int mm_cid_clear_lazy_put(int cid)
1349 {
1350 	return cid & ~MM_CID_LAZY_PUT;
1351 }
1352 
1353 /*
1354  * mm_cpus_allowed: Union of all mm's threads allowed CPUs.
1355  */
1356 static inline cpumask_t *mm_cpus_allowed(struct mm_struct *mm)
1357 {
1358 	unsigned long bitmap = (unsigned long)mm;
1359 
1360 	bitmap += offsetof(struct mm_struct, cpu_bitmap);
1361 	/* Skip cpu_bitmap */
1362 	bitmap += cpumask_size();
1363 	return (struct cpumask *)bitmap;
1364 }
1365 
1366 /* Accessor for struct mm_struct's cidmask. */
1367 static inline cpumask_t *mm_cidmask(struct mm_struct *mm)
1368 {
1369 	unsigned long cid_bitmap = (unsigned long)mm_cpus_allowed(mm);
1370 
1371 	/* Skip mm_cpus_allowed */
1372 	cid_bitmap += cpumask_size();
1373 	return (struct cpumask *)cid_bitmap;
1374 }
1375 
1376 static inline void mm_init_cid(struct mm_struct *mm, struct task_struct *p)
1377 {
1378 	int i;
1379 
1380 	for_each_possible_cpu(i) {
1381 		struct mm_cid *pcpu_cid = per_cpu_ptr(mm->pcpu_cid, i);
1382 
1383 		pcpu_cid->cid = MM_CID_UNSET;
1384 		pcpu_cid->recent_cid = MM_CID_UNSET;
1385 		pcpu_cid->time = 0;
1386 	}
1387 	mm->nr_cpus_allowed = p->nr_cpus_allowed;
1388 	atomic_set(&mm->max_nr_cid, 0);
1389 	raw_spin_lock_init(&mm->cpus_allowed_lock);
1390 	cpumask_copy(mm_cpus_allowed(mm), &p->cpus_mask);
1391 	cpumask_clear(mm_cidmask(mm));
1392 }
1393 
1394 static inline int mm_alloc_cid_noprof(struct mm_struct *mm, struct task_struct *p)
1395 {
1396 	mm->pcpu_cid = alloc_percpu_noprof(struct mm_cid);
1397 	if (!mm->pcpu_cid)
1398 		return -ENOMEM;
1399 	mm_init_cid(mm, p);
1400 	return 0;
1401 }
1402 #define mm_alloc_cid(...)	alloc_hooks(mm_alloc_cid_noprof(__VA_ARGS__))
1403 
1404 static inline void mm_destroy_cid(struct mm_struct *mm)
1405 {
1406 	free_percpu(mm->pcpu_cid);
1407 	mm->pcpu_cid = NULL;
1408 }
1409 
1410 static inline unsigned int mm_cid_size(void)
1411 {
1412 	return 2 * cpumask_size();	/* mm_cpus_allowed(), mm_cidmask(). */
1413 }
1414 
1415 static inline void mm_set_cpus_allowed(struct mm_struct *mm, const struct cpumask *cpumask)
1416 {
1417 	struct cpumask *mm_allowed = mm_cpus_allowed(mm);
1418 
1419 	if (!mm)
1420 		return;
1421 	/* The mm_cpus_allowed is the union of each thread allowed CPUs masks. */
1422 	raw_spin_lock(&mm->cpus_allowed_lock);
1423 	cpumask_or(mm_allowed, mm_allowed, cpumask);
1424 	WRITE_ONCE(mm->nr_cpus_allowed, cpumask_weight(mm_allowed));
1425 	raw_spin_unlock(&mm->cpus_allowed_lock);
1426 }
1427 #else /* CONFIG_SCHED_MM_CID */
1428 static inline void mm_init_cid(struct mm_struct *mm, struct task_struct *p) { }
1429 static inline int mm_alloc_cid(struct mm_struct *mm, struct task_struct *p) { return 0; }
1430 static inline void mm_destroy_cid(struct mm_struct *mm) { }
1431 
1432 static inline unsigned int mm_cid_size(void)
1433 {
1434 	return 0;
1435 }
1436 static inline void mm_set_cpus_allowed(struct mm_struct *mm, const struct cpumask *cpumask) { }
1437 #endif /* CONFIG_SCHED_MM_CID */
1438 
1439 struct mmu_gather;
1440 extern void tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm);
1441 extern void tlb_gather_mmu_fullmm(struct mmu_gather *tlb, struct mm_struct *mm);
1442 extern void tlb_finish_mmu(struct mmu_gather *tlb);
1443 
1444 struct vm_fault;
1445 
1446 /**
1447  * typedef vm_fault_t - Return type for page fault handlers.
1448  *
1449  * Page fault handlers return a bitmask of %VM_FAULT values.
1450  */
1451 typedef __bitwise unsigned int vm_fault_t;
1452 
1453 /**
1454  * enum vm_fault_reason - Page fault handlers return a bitmask of
1455  * these values to tell the core VM what happened when handling the
1456  * fault. Used to decide whether a process gets delivered SIGBUS or
1457  * just gets major/minor fault counters bumped up.
1458  *
1459  * @VM_FAULT_OOM:		Out Of Memory
1460  * @VM_FAULT_SIGBUS:		Bad access
1461  * @VM_FAULT_MAJOR:		Page read from storage
1462  * @VM_FAULT_HWPOISON:		Hit poisoned small page
1463  * @VM_FAULT_HWPOISON_LARGE:	Hit poisoned large page. Index encoded
1464  *				in upper bits
1465  * @VM_FAULT_SIGSEGV:		segmentation fault
1466  * @VM_FAULT_NOPAGE:		->fault installed the pte, not return page
1467  * @VM_FAULT_LOCKED:		->fault locked the returned page
1468  * @VM_FAULT_RETRY:		->fault blocked, must retry
1469  * @VM_FAULT_FALLBACK:		huge page fault failed, fall back to small
1470  * @VM_FAULT_DONE_COW:		->fault has fully handled COW
1471  * @VM_FAULT_NEEDDSYNC:		->fault did not modify page tables and needs
1472  *				fsync() to complete (for synchronous page faults
1473  *				in DAX)
1474  * @VM_FAULT_COMPLETED:		->fault completed, meanwhile mmap lock released
1475  * @VM_FAULT_HINDEX_MASK:	mask HINDEX value
1476  *
1477  */
1478 enum vm_fault_reason {
1479 	VM_FAULT_OOM            = (__force vm_fault_t)0x000001,
1480 	VM_FAULT_SIGBUS         = (__force vm_fault_t)0x000002,
1481 	VM_FAULT_MAJOR          = (__force vm_fault_t)0x000004,
1482 	VM_FAULT_HWPOISON       = (__force vm_fault_t)0x000010,
1483 	VM_FAULT_HWPOISON_LARGE = (__force vm_fault_t)0x000020,
1484 	VM_FAULT_SIGSEGV        = (__force vm_fault_t)0x000040,
1485 	VM_FAULT_NOPAGE         = (__force vm_fault_t)0x000100,
1486 	VM_FAULT_LOCKED         = (__force vm_fault_t)0x000200,
1487 	VM_FAULT_RETRY          = (__force vm_fault_t)0x000400,
1488 	VM_FAULT_FALLBACK       = (__force vm_fault_t)0x000800,
1489 	VM_FAULT_DONE_COW       = (__force vm_fault_t)0x001000,
1490 	VM_FAULT_NEEDDSYNC      = (__force vm_fault_t)0x002000,
1491 	VM_FAULT_COMPLETED      = (__force vm_fault_t)0x004000,
1492 	VM_FAULT_HINDEX_MASK    = (__force vm_fault_t)0x0f0000,
1493 };
1494 
1495 /* Encode hstate index for a hwpoisoned large page */
1496 #define VM_FAULT_SET_HINDEX(x) ((__force vm_fault_t)((x) << 16))
1497 #define VM_FAULT_GET_HINDEX(x) (((__force unsigned int)(x) >> 16) & 0xf)
1498 
1499 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS |	\
1500 			VM_FAULT_SIGSEGV | VM_FAULT_HWPOISON |	\
1501 			VM_FAULT_HWPOISON_LARGE | VM_FAULT_FALLBACK)
1502 
1503 #define VM_FAULT_RESULT_TRACE \
1504 	{ VM_FAULT_OOM,                 "OOM" },	\
1505 	{ VM_FAULT_SIGBUS,              "SIGBUS" },	\
1506 	{ VM_FAULT_MAJOR,               "MAJOR" },	\
1507 	{ VM_FAULT_HWPOISON,            "HWPOISON" },	\
1508 	{ VM_FAULT_HWPOISON_LARGE,      "HWPOISON_LARGE" },	\
1509 	{ VM_FAULT_SIGSEGV,             "SIGSEGV" },	\
1510 	{ VM_FAULT_NOPAGE,              "NOPAGE" },	\
1511 	{ VM_FAULT_LOCKED,              "LOCKED" },	\
1512 	{ VM_FAULT_RETRY,               "RETRY" },	\
1513 	{ VM_FAULT_FALLBACK,            "FALLBACK" },	\
1514 	{ VM_FAULT_DONE_COW,            "DONE_COW" },	\
1515 	{ VM_FAULT_NEEDDSYNC,           "NEEDDSYNC" },	\
1516 	{ VM_FAULT_COMPLETED,           "COMPLETED" }
1517 
1518 struct vm_special_mapping {
1519 	const char *name;	/* The name, e.g. "[vdso]". */
1520 
1521 	/*
1522 	 * If .fault is not provided, this points to a
1523 	 * NULL-terminated array of pages that back the special mapping.
1524 	 *
1525 	 * This must not be NULL unless .fault is provided.
1526 	 */
1527 	struct page **pages;
1528 
1529 	/*
1530 	 * If non-NULL, then this is called to resolve page faults
1531 	 * on the special mapping.  If used, .pages is not checked.
1532 	 */
1533 	vm_fault_t (*fault)(const struct vm_special_mapping *sm,
1534 				struct vm_area_struct *vma,
1535 				struct vm_fault *vmf);
1536 
1537 	int (*mremap)(const struct vm_special_mapping *sm,
1538 		     struct vm_area_struct *new_vma);
1539 
1540 	void (*close)(const struct vm_special_mapping *sm,
1541 		      struct vm_area_struct *vma);
1542 };
1543 
1544 enum tlb_flush_reason {
1545 	TLB_FLUSH_ON_TASK_SWITCH,
1546 	TLB_REMOTE_SHOOTDOWN,
1547 	TLB_LOCAL_SHOOTDOWN,
1548 	TLB_LOCAL_MM_SHOOTDOWN,
1549 	TLB_REMOTE_SEND_IPI,
1550 	TLB_REMOTE_WRONG_CPU,
1551 	NR_TLB_FLUSH_REASONS,
1552 };
1553 
1554 /**
1555  * enum fault_flag - Fault flag definitions.
1556  * @FAULT_FLAG_WRITE: Fault was a write fault.
1557  * @FAULT_FLAG_MKWRITE: Fault was mkwrite of existing PTE.
1558  * @FAULT_FLAG_ALLOW_RETRY: Allow to retry the fault if blocked.
1559  * @FAULT_FLAG_RETRY_NOWAIT: Don't drop mmap_lock and wait when retrying.
1560  * @FAULT_FLAG_KILLABLE: The fault task is in SIGKILL killable region.
1561  * @FAULT_FLAG_TRIED: The fault has been tried once.
1562  * @FAULT_FLAG_USER: The fault originated in userspace.
1563  * @FAULT_FLAG_REMOTE: The fault is not for current task/mm.
1564  * @FAULT_FLAG_INSTRUCTION: The fault was during an instruction fetch.
1565  * @FAULT_FLAG_INTERRUPTIBLE: The fault can be interrupted by non-fatal signals.
1566  * @FAULT_FLAG_UNSHARE: The fault is an unsharing request to break COW in a
1567  *                      COW mapping, making sure that an exclusive anon page is
1568  *                      mapped after the fault.
1569  * @FAULT_FLAG_ORIG_PTE_VALID: whether the fault has vmf->orig_pte cached.
1570  *                        We should only access orig_pte if this flag set.
1571  * @FAULT_FLAG_VMA_LOCK: The fault is handled under VMA lock.
1572  *
1573  * About @FAULT_FLAG_ALLOW_RETRY and @FAULT_FLAG_TRIED: we can specify
1574  * whether we would allow page faults to retry by specifying these two
1575  * fault flags correctly.  Currently there can be three legal combinations:
1576  *
1577  * (a) ALLOW_RETRY and !TRIED:  this means the page fault allows retry, and
1578  *                              this is the first try
1579  *
1580  * (b) ALLOW_RETRY and TRIED:   this means the page fault allows retry, and
1581  *                              we've already tried at least once
1582  *
1583  * (c) !ALLOW_RETRY and !TRIED: this means the page fault does not allow retry
1584  *
1585  * The unlisted combination (!ALLOW_RETRY && TRIED) is illegal and should never
1586  * be used.  Note that page faults can be allowed to retry for multiple times,
1587  * in which case we'll have an initial fault with flags (a) then later on
1588  * continuous faults with flags (b).  We should always try to detect pending
1589  * signals before a retry to make sure the continuous page faults can still be
1590  * interrupted if necessary.
1591  *
1592  * The combination FAULT_FLAG_WRITE|FAULT_FLAG_UNSHARE is illegal.
1593  * FAULT_FLAG_UNSHARE is ignored and treated like an ordinary read fault when
1594  * applied to mappings that are not COW mappings.
1595  */
1596 enum fault_flag {
1597 	FAULT_FLAG_WRITE =		1 << 0,
1598 	FAULT_FLAG_MKWRITE =		1 << 1,
1599 	FAULT_FLAG_ALLOW_RETRY =	1 << 2,
1600 	FAULT_FLAG_RETRY_NOWAIT = 	1 << 3,
1601 	FAULT_FLAG_KILLABLE =		1 << 4,
1602 	FAULT_FLAG_TRIED = 		1 << 5,
1603 	FAULT_FLAG_USER =		1 << 6,
1604 	FAULT_FLAG_REMOTE =		1 << 7,
1605 	FAULT_FLAG_INSTRUCTION =	1 << 8,
1606 	FAULT_FLAG_INTERRUPTIBLE =	1 << 9,
1607 	FAULT_FLAG_UNSHARE =		1 << 10,
1608 	FAULT_FLAG_ORIG_PTE_VALID =	1 << 11,
1609 	FAULT_FLAG_VMA_LOCK =		1 << 12,
1610 };
1611 
1612 typedef unsigned int __bitwise zap_flags_t;
1613 
1614 /* Flags for clear_young_dirty_ptes(). */
1615 typedef int __bitwise cydp_t;
1616 
1617 /* Clear the access bit */
1618 #define CYDP_CLEAR_YOUNG		((__force cydp_t)BIT(0))
1619 
1620 /* Clear the dirty bit */
1621 #define CYDP_CLEAR_DIRTY		((__force cydp_t)BIT(1))
1622 
1623 /*
1624  * FOLL_PIN and FOLL_LONGTERM may be used in various combinations with each
1625  * other. Here is what they mean, and how to use them:
1626  *
1627  *
1628  * FIXME: For pages which are part of a filesystem, mappings are subject to the
1629  * lifetime enforced by the filesystem and we need guarantees that longterm
1630  * users like RDMA and V4L2 only establish mappings which coordinate usage with
1631  * the filesystem.  Ideas for this coordination include revoking the longterm
1632  * pin, delaying writeback, bounce buffer page writeback, etc.  As FS DAX was
1633  * added after the problem with filesystems was found FS DAX VMAs are
1634  * specifically failed.  Filesystem pages are still subject to bugs and use of
1635  * FOLL_LONGTERM should be avoided on those pages.
1636  *
1637  * In the CMA case: long term pins in a CMA region would unnecessarily fragment
1638  * that region.  And so, CMA attempts to migrate the page before pinning, when
1639  * FOLL_LONGTERM is specified.
1640  *
1641  * FOLL_PIN indicates that a special kind of tracking (not just page->_refcount,
1642  * but an additional pin counting system) will be invoked. This is intended for
1643  * anything that gets a page reference and then touches page data (for example,
1644  * Direct IO). This lets the filesystem know that some non-file-system entity is
1645  * potentially changing the pages' data. In contrast to FOLL_GET (whose pages
1646  * are released via put_page()), FOLL_PIN pages must be released, ultimately, by
1647  * a call to unpin_user_page().
1648  *
1649  * FOLL_PIN is similar to FOLL_GET: both of these pin pages. They use different
1650  * and separate refcounting mechanisms, however, and that means that each has
1651  * its own acquire and release mechanisms:
1652  *
1653  *     FOLL_GET: get_user_pages*() to acquire, and put_page() to release.
1654  *
1655  *     FOLL_PIN: pin_user_pages*() to acquire, and unpin_user_pages to release.
1656  *
1657  * FOLL_PIN and FOLL_GET are mutually exclusive for a given function call.
1658  * (The underlying pages may experience both FOLL_GET-based and FOLL_PIN-based
1659  * calls applied to them, and that's perfectly OK. This is a constraint on the
1660  * callers, not on the pages.)
1661  *
1662  * FOLL_PIN should be set internally by the pin_user_pages*() APIs, never
1663  * directly by the caller. That's in order to help avoid mismatches when
1664  * releasing pages: get_user_pages*() pages must be released via put_page(),
1665  * while pin_user_pages*() pages must be released via unpin_user_page().
1666  *
1667  * Please see Documentation/core-api/pin_user_pages.rst for more information.
1668  */
1669 
1670 enum {
1671 	/* check pte is writable */
1672 	FOLL_WRITE = 1 << 0,
1673 	/* do get_page on page */
1674 	FOLL_GET = 1 << 1,
1675 	/* give error on hole if it would be zero */
1676 	FOLL_DUMP = 1 << 2,
1677 	/* get_user_pages read/write w/o permission */
1678 	FOLL_FORCE = 1 << 3,
1679 	/*
1680 	 * if a disk transfer is needed, start the IO and return without waiting
1681 	 * upon it
1682 	 */
1683 	FOLL_NOWAIT = 1 << 4,
1684 	/* do not fault in pages */
1685 	FOLL_NOFAULT = 1 << 5,
1686 	/* check page is hwpoisoned */
1687 	FOLL_HWPOISON = 1 << 6,
1688 	/* don't do file mappings */
1689 	FOLL_ANON = 1 << 7,
1690 	/*
1691 	 * FOLL_LONGTERM indicates that the page will be held for an indefinite
1692 	 * time period _often_ under userspace control.  This is in contrast to
1693 	 * iov_iter_get_pages(), whose usages are transient.
1694 	 */
1695 	FOLL_LONGTERM = 1 << 8,
1696 	/* split huge pmd before returning */
1697 	FOLL_SPLIT_PMD = 1 << 9,
1698 	/* allow returning PCI P2PDMA pages */
1699 	FOLL_PCI_P2PDMA = 1 << 10,
1700 	/* allow interrupts from generic signals */
1701 	FOLL_INTERRUPTIBLE = 1 << 11,
1702 	/*
1703 	 * Always honor (trigger) NUMA hinting faults.
1704 	 *
1705 	 * FOLL_WRITE implicitly honors NUMA hinting faults because a
1706 	 * PROT_NONE-mapped page is not writable (exceptions with FOLL_FORCE
1707 	 * apply). get_user_pages_fast_only() always implicitly honors NUMA
1708 	 * hinting faults.
1709 	 */
1710 	FOLL_HONOR_NUMA_FAULT = 1 << 12,
1711 
1712 	/* See also internal only FOLL flags in mm/internal.h */
1713 };
1714 
1715 /* mm flags */
1716 
1717 /*
1718  * The first two bits represent core dump modes for set-user-ID,
1719  * the modes are SUID_DUMP_* defined in linux/sched/coredump.h
1720  */
1721 #define MMF_DUMPABLE_BITS 2
1722 #define MMF_DUMPABLE_MASK ((1 << MMF_DUMPABLE_BITS) - 1)
1723 /* coredump filter bits */
1724 #define MMF_DUMP_ANON_PRIVATE	2
1725 #define MMF_DUMP_ANON_SHARED	3
1726 #define MMF_DUMP_MAPPED_PRIVATE	4
1727 #define MMF_DUMP_MAPPED_SHARED	5
1728 #define MMF_DUMP_ELF_HEADERS	6
1729 #define MMF_DUMP_HUGETLB_PRIVATE 7
1730 #define MMF_DUMP_HUGETLB_SHARED  8
1731 #define MMF_DUMP_DAX_PRIVATE	9
1732 #define MMF_DUMP_DAX_SHARED	10
1733 
1734 #define MMF_DUMP_FILTER_SHIFT	MMF_DUMPABLE_BITS
1735 #define MMF_DUMP_FILTER_BITS	9
1736 #define MMF_DUMP_FILTER_MASK \
1737 	(((1 << MMF_DUMP_FILTER_BITS) - 1) << MMF_DUMP_FILTER_SHIFT)
1738 #define MMF_DUMP_FILTER_DEFAULT \
1739 	((1 << MMF_DUMP_ANON_PRIVATE) |	(1 << MMF_DUMP_ANON_SHARED) |\
1740 	 (1 << MMF_DUMP_HUGETLB_PRIVATE) | MMF_DUMP_MASK_DEFAULT_ELF)
1741 
1742 #ifdef CONFIG_CORE_DUMP_DEFAULT_ELF_HEADERS
1743 # define MMF_DUMP_MASK_DEFAULT_ELF	(1 << MMF_DUMP_ELF_HEADERS)
1744 #else
1745 # define MMF_DUMP_MASK_DEFAULT_ELF	0
1746 #endif
1747 					/* leave room for more dump flags */
1748 #define MMF_VM_MERGEABLE	16	/* KSM may merge identical pages */
1749 #define MMF_VM_HUGEPAGE		17	/* set when mm is available for khugepaged */
1750 
1751 /*
1752  * This one-shot flag is dropped due to necessity of changing exe once again
1753  * on NFS restore
1754  */
1755 //#define MMF_EXE_FILE_CHANGED	18	/* see prctl_set_mm_exe_file() */
1756 
1757 #define MMF_HAS_UPROBES		19	/* has uprobes */
1758 #define MMF_RECALC_UPROBES	20	/* MMF_HAS_UPROBES can be wrong */
1759 #define MMF_OOM_SKIP		21	/* mm is of no interest for the OOM killer */
1760 #define MMF_UNSTABLE		22	/* mm is unstable for copy_from_user */
1761 #define MMF_HUGE_ZERO_PAGE	23      /* mm has ever used the global huge zero page */
1762 #define MMF_DISABLE_THP		24	/* disable THP for all VMAs */
1763 #define MMF_DISABLE_THP_MASK	(1 << MMF_DISABLE_THP)
1764 #define MMF_OOM_REAP_QUEUED	25	/* mm was queued for oom_reaper */
1765 #define MMF_MULTIPROCESS	26	/* mm is shared between processes */
1766 /*
1767  * MMF_HAS_PINNED: Whether this mm has pinned any pages.  This can be either
1768  * replaced in the future by mm.pinned_vm when it becomes stable, or grow into
1769  * a counter on its own. We're aggresive on this bit for now: even if the
1770  * pinned pages were unpinned later on, we'll still keep this bit set for the
1771  * lifecycle of this mm, just for simplicity.
1772  */
1773 #define MMF_HAS_PINNED		27	/* FOLL_PIN has run, never cleared */
1774 
1775 #define MMF_HAS_MDWE		28
1776 #define MMF_HAS_MDWE_MASK	(1 << MMF_HAS_MDWE)
1777 
1778 
1779 #define MMF_HAS_MDWE_NO_INHERIT	29
1780 
1781 #define MMF_VM_MERGE_ANY	30
1782 #define MMF_VM_MERGE_ANY_MASK	(1 << MMF_VM_MERGE_ANY)
1783 
1784 #define MMF_TOPDOWN		31	/* mm searches top down by default */
1785 #define MMF_TOPDOWN_MASK	(1 << MMF_TOPDOWN)
1786 
1787 #define MMF_INIT_MASK		(MMF_DUMPABLE_MASK | MMF_DUMP_FILTER_MASK |\
1788 				 MMF_DISABLE_THP_MASK | MMF_HAS_MDWE_MASK |\
1789 				 MMF_VM_MERGE_ANY_MASK | MMF_TOPDOWN_MASK)
1790 
1791 static inline unsigned long mmf_init_flags(unsigned long flags)
1792 {
1793 	if (flags & (1UL << MMF_HAS_MDWE_NO_INHERIT))
1794 		flags &= ~((1UL << MMF_HAS_MDWE) |
1795 			   (1UL << MMF_HAS_MDWE_NO_INHERIT));
1796 	return flags & MMF_INIT_MASK;
1797 }
1798 
1799 #endif /* _LINUX_MM_TYPES_H */
1800