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/rwsem.h> |
13 | #include <linux/completion.h> |
14 | #include <linux/cpumask.h> |
15 | #include <linux/uprobes.h> |
16 | #include <linux/rcupdate.h> |
17 | #include <linux/page-flags-layout.h> |
18 | #include <linux/workqueue.h> |
19 | #include <linux/seqlock.h> |
20 | |
21 | #include <asm/mmu.h> |
22 | |
23 | #ifndef AT_VECTOR_SIZE_ARCH |
24 | #define AT_VECTOR_SIZE_ARCH 0 |
25 | #endif |
26 | #define AT_VECTOR_SIZE (2*(AT_VECTOR_SIZE_ARCH + AT_VECTOR_SIZE_BASE + 1)) |
27 | |
28 | #define INIT_PASID 0 |
29 | |
30 | struct address_space; |
31 | struct mem_cgroup; |
32 | |
33 | /* |
34 | * Each physical page in the system has a struct page associated with |
35 | * it to keep track of whatever it is we are using the page for at the |
36 | * moment. Note that we have no way to track which tasks are using |
37 | * a page, though if it is a pagecache page, rmap structures can tell us |
38 | * who is mapping it. |
39 | * |
40 | * If you allocate the page using alloc_pages(), you can use some of the |
41 | * space in struct page for your own purposes. The five words in the main |
42 | * union are available, except for bit 0 of the first word which must be |
43 | * kept clear. Many users use this word to store a pointer to an object |
44 | * which is guaranteed to be aligned. If you use the same storage as |
45 | * page->mapping, you must restore it to NULL before freeing the page. |
46 | * |
47 | * If your page will not be mapped to userspace, you can also use the four |
48 | * bytes in the mapcount union, but you must call page_mapcount_reset() |
49 | * before freeing it. |
50 | * |
51 | * If you want to use the refcount field, it must be used in such a way |
52 | * that other CPUs temporarily incrementing and then decrementing the |
53 | * refcount does not cause problems. On receiving the page from |
54 | * alloc_pages(), the refcount will be positive. |
55 | * |
56 | * If you allocate pages of order > 0, you can use some of the fields |
57 | * in each subpage, but you may need to restore some of their values |
58 | * afterwards. |
59 | * |
60 | * SLUB uses cmpxchg_double() to atomically update its freelist and counters. |
61 | * That requires that freelist & counters in struct slab be adjacent and |
62 | * double-word aligned. Because struct slab currently just reinterprets the |
63 | * bits of struct page, we align all struct pages to double-word boundaries, |
64 | * and ensure that 'freelist' is aligned within struct slab. |
65 | */ |
66 | #ifdef CONFIG_HAVE_ALIGNED_STRUCT_PAGE |
67 | #define _struct_page_alignment __aligned(2 * sizeof(unsigned long)) |
68 | #else |
69 | #define _struct_page_alignment |
70 | #endif |
71 | |
72 | struct page { |
73 | unsigned long flags; /* Atomic flags, some possibly |
74 | * updated asynchronously */ |
75 | /* |
76 | * Five words (20/40 bytes) are available in this union. |
77 | * WARNING: bit 0 of the first word is used for PageTail(). That |
78 | * means the other users of this union MUST NOT use the bit to |
79 | * avoid collision and false-positive PageTail(). |
80 | */ |
81 | union { |
82 | struct { /* Page cache and anonymous pages */ |
83 | /** |
84 | * @lru: Pageout list, eg. active_list protected by |
85 | * lruvec->lru_lock. Sometimes used as a generic list |
86 | * by the page owner. |
87 | */ |
88 | union { |
89 | struct list_head lru; |
90 | |
91 | /* Or, for the Unevictable "LRU list" slot */ |
92 | struct { |
93 | /* Always even, to negate PageTail */ |
94 | void *__filler; |
95 | /* Count page's or folio's mlocks */ |
96 | unsigned int mlock_count; |
97 | }; |
98 | |
99 | /* Or, free page */ |
100 | struct list_head buddy_list; |
101 | struct list_head pcp_list; |
102 | }; |
103 | /* See page-flags.h for PAGE_MAPPING_FLAGS */ |
104 | struct address_space *mapping; |
105 | pgoff_t index; /* Our offset within mapping. */ |
106 | /** |
107 | * @private: Mapping-private opaque data. |
108 | * Usually used for buffer_heads if PagePrivate. |
109 | * Used for swp_entry_t if PageSwapCache. |
110 | * Indicates order in the buddy system if PageBuddy. |
111 | */ |
112 | unsigned long private; |
113 | }; |
114 | struct { /* page_pool used by netstack */ |
115 | /** |
116 | * @pp_magic: magic value to avoid recycling non |
117 | * page_pool allocated pages. |
118 | */ |
119 | unsigned long pp_magic; |
120 | struct page_pool *pp; |
121 | unsigned long _pp_mapping_pad; |
122 | unsigned long dma_addr; |
123 | union { |
124 | /** |
125 | * dma_addr_upper: might require a 64-bit |
126 | * value on 32-bit architectures. |
127 | */ |
128 | unsigned long dma_addr_upper; |
129 | /** |
130 | * For frag page support, not supported in |
131 | * 32-bit architectures with 64-bit DMA. |
132 | */ |
133 | atomic_long_t pp_frag_count; |
134 | }; |
135 | }; |
136 | struct { /* Tail pages of compound page */ |
137 | unsigned long compound_head; /* Bit zero is set */ |
138 | |
139 | /* First tail page only */ |
140 | unsigned char compound_dtor; |
141 | unsigned char compound_order; |
142 | atomic_t compound_mapcount; |
143 | atomic_t compound_pincount; |
144 | #ifdef CONFIG_64BIT |
145 | unsigned int compound_nr; /* 1 << compound_order */ |
146 | #endif |
147 | }; |
148 | struct { /* Second tail page of compound page */ |
149 | unsigned long _compound_pad_1; /* compound_head */ |
150 | unsigned long _compound_pad_2; |
151 | /* For both global and memcg */ |
152 | struct list_head deferred_list; |
153 | }; |
154 | struct { /* Page table pages */ |
155 | unsigned long _pt_pad_1; /* compound_head */ |
156 | pgtable_t pmd_huge_pte; /* protected by page->ptl */ |
157 | unsigned long _pt_pad_2; /* mapping */ |
158 | union { |
159 | struct mm_struct *pt_mm; /* x86 pgds only */ |
160 | atomic_t pt_frag_refcount; /* powerpc */ |
161 | }; |
162 | #if ALLOC_SPLIT_PTLOCKS |
163 | spinlock_t *ptl; |
164 | #else |
165 | spinlock_t ptl; |
166 | #endif |
167 | }; |
168 | struct { /* ZONE_DEVICE pages */ |
169 | /** @pgmap: Points to the hosting device page map. */ |
170 | struct dev_pagemap *pgmap; |
171 | void *zone_device_data; |
172 | /* |
173 | * ZONE_DEVICE private pages are counted as being |
174 | * mapped so the next 3 words hold the mapping, index, |
175 | * and private fields from the source anonymous or |
176 | * page cache page while the page is migrated to device |
177 | * private memory. |
178 | * ZONE_DEVICE MEMORY_DEVICE_FS_DAX pages also |
179 | * use the mapping, index, and private fields when |
180 | * pmem backed DAX files are mapped. |
181 | */ |
182 | }; |
183 | |
184 | /** @rcu_head: You can use this to free a page by RCU. */ |
185 | struct rcu_head rcu_head; |
186 | }; |
187 | |
188 | union { /* This union is 4 bytes in size. */ |
189 | /* |
190 | * If the page can be mapped to userspace, encodes the number |
191 | * of times this page is referenced by a page table. |
192 | */ |
193 | atomic_t _mapcount; |
194 | |
195 | /* |
196 | * If the page is neither PageSlab nor mappable to userspace, |
197 | * the value stored here may help determine what this page |
198 | * is used for. See page-flags.h for a list of page types |
199 | * which are currently stored here. |
200 | */ |
201 | unsigned int page_type; |
202 | }; |
203 | |
204 | /* Usage count. *DO NOT USE DIRECTLY*. See page_ref.h */ |
205 | atomic_t _refcount; |
206 | |
207 | #ifdef CONFIG_MEMCG |
208 | unsigned long memcg_data; |
209 | #endif |
210 | |
211 | /* |
212 | * On machines where all RAM is mapped into kernel address space, |
213 | * we can simply calculate the virtual address. On machines with |
214 | * highmem some memory is mapped into kernel virtual memory |
215 | * dynamically, so we need a place to store that address. |
216 | * Note that this field could be 16 bits on x86 ... ;) |
217 | * |
218 | * Architectures with slow multiplication can define |
219 | * WANT_PAGE_VIRTUAL in asm/page.h |
220 | */ |
221 | #if defined(WANT_PAGE_VIRTUAL) |
222 | void *virtual; /* Kernel virtual address (NULL if |
223 | not kmapped, ie. highmem) */ |
224 | #endif /* WANT_PAGE_VIRTUAL */ |
225 | |
226 | #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS |
227 | int _last_cpupid; |
228 | #endif |
229 | } _struct_page_alignment; |
230 | |
231 | /** |
232 | * struct folio - Represents a contiguous set of bytes. |
233 | * @flags: Identical to the page flags. |
234 | * @lru: Least Recently Used list; tracks how recently this folio was used. |
235 | * @mlock_count: Number of times this folio has been pinned by mlock(). |
236 | * @mapping: The file this page belongs to, or refers to the anon_vma for |
237 | * anonymous memory. |
238 | * @index: Offset within the file, in units of pages. For anonymous memory, |
239 | * this is the index from the beginning of the mmap. |
240 | * @private: Filesystem per-folio data (see folio_attach_private()). |
241 | * Used for swp_entry_t if folio_test_swapcache(). |
242 | * @_mapcount: Do not access this member directly. Use folio_mapcount() to |
243 | * find out how many times this folio is mapped by userspace. |
244 | * @_refcount: Do not access this member directly. Use folio_ref_count() |
245 | * to find how many references there are to this folio. |
246 | * @memcg_data: Memory Control Group data. |
247 | * |
248 | * A folio is a physically, virtually and logically contiguous set |
249 | * of bytes. It is a power-of-two in size, and it is aligned to that |
250 | * same power-of-two. It is at least as large as %PAGE_SIZE. If it is |
251 | * in the page cache, it is at a file offset which is a multiple of that |
252 | * power-of-two. It may be mapped into userspace at an address which is |
253 | * at an arbitrary page offset, but its kernel virtual address is aligned |
254 | * to its size. |
255 | */ |
256 | struct folio { |
257 | /* private: don't document the anon union */ |
258 | union { |
259 | struct { |
260 | /* public: */ |
261 | unsigned long flags; |
262 | union { |
263 | struct list_head lru; |
264 | /* private: avoid cluttering the output */ |
265 | struct { |
266 | void *__filler; |
267 | /* public: */ |
268 | unsigned int mlock_count; |
269 | /* private: */ |
270 | }; |
271 | /* public: */ |
272 | }; |
273 | struct address_space *mapping; |
274 | pgoff_t index; |
275 | void *private; |
276 | atomic_t _mapcount; |
277 | atomic_t _refcount; |
278 | #ifdef CONFIG_MEMCG |
279 | unsigned long memcg_data; |
280 | #endif |
281 | /* private: the union with struct page is transitional */ |
282 | }; |
283 | struct page page; |
284 | }; |
285 | }; |
286 | |
287 | static_assert(sizeof(struct page) == sizeof(struct folio)); |
288 | #define FOLIO_MATCH(pg, fl) \ |
289 | static_assert(offsetof(struct page, pg) == offsetof(struct folio, fl)) |
290 | FOLIO_MATCH(flags, flags); |
291 | FOLIO_MATCH(lru, lru); |
292 | FOLIO_MATCH(mapping, mapping); |
293 | FOLIO_MATCH(compound_head, lru); |
294 | FOLIO_MATCH(index, index); |
295 | FOLIO_MATCH(private, private); |
296 | FOLIO_MATCH(_mapcount, _mapcount); |
297 | FOLIO_MATCH(_refcount, _refcount); |
298 | #ifdef CONFIG_MEMCG |
299 | FOLIO_MATCH(memcg_data, memcg_data); |
300 | #endif |
301 | #undef FOLIO_MATCH |
302 | |
303 | static inline atomic_t *folio_mapcount_ptr(struct folio *folio) |
304 | { |
305 | struct page *tail = &folio->page + 1; |
306 | return &tail->compound_mapcount; |
307 | } |
308 | |
309 | static inline atomic_t *compound_mapcount_ptr(struct page *page) |
310 | { |
311 | return &page[1].compound_mapcount; |
312 | } |
313 | |
314 | static inline atomic_t *compound_pincount_ptr(struct page *page) |
315 | { |
316 | return &page[1].compound_pincount; |
317 | } |
318 | |
319 | /* |
320 | * Used for sizing the vmemmap region on some architectures |
321 | */ |
322 | #define STRUCT_PAGE_MAX_SHIFT (order_base_2(sizeof(struct page))) |
323 | |
324 | #define PAGE_FRAG_CACHE_MAX_SIZE __ALIGN_MASK(32768, ~PAGE_MASK) |
325 | #define PAGE_FRAG_CACHE_MAX_ORDER get_order(PAGE_FRAG_CACHE_MAX_SIZE) |
326 | |
327 | /* |
328 | * page_private can be used on tail pages. However, PagePrivate is only |
329 | * checked by the VM on the head page. So page_private on the tail pages |
330 | * should be used for data that's ancillary to the head page (eg attaching |
331 | * buffer heads to tail pages after attaching buffer heads to the head page) |
332 | */ |
333 | #define page_private(page) ((page)->private) |
334 | |
335 | static inline void set_page_private(struct page *page, unsigned long private) |
336 | { |
337 | page->private = private; |
338 | } |
339 | |
340 | static inline void *folio_get_private(struct folio *folio) |
341 | { |
342 | return folio->private; |
343 | } |
344 | |
345 | struct page_frag_cache { |
346 | void * va; |
347 | #if (PAGE_SIZE < PAGE_FRAG_CACHE_MAX_SIZE) |
348 | __u16 offset; |
349 | __u16 size; |
350 | #else |
351 | __u32 offset; |
352 | #endif |
353 | /* we maintain a pagecount bias, so that we dont dirty cache line |
354 | * containing page->_refcount every time we allocate a fragment. |
355 | */ |
356 | unsigned int pagecnt_bias; |
357 | bool pfmemalloc; |
358 | }; |
359 | |
360 | typedef unsigned long vm_flags_t; |
361 | |
362 | /* |
363 | * A region containing a mapping of a non-memory backed file under NOMMU |
364 | * conditions. These are held in a global tree and are pinned by the VMAs that |
365 | * map parts of them. |
366 | */ |
367 | struct vm_region { |
368 | struct rb_node vm_rb; /* link in global region tree */ |
369 | vm_flags_t vm_flags; /* VMA vm_flags */ |
370 | unsigned long vm_start; /* start address of region */ |
371 | unsigned long vm_end; /* region initialised to here */ |
372 | unsigned long vm_top; /* region allocated to here */ |
373 | unsigned long vm_pgoff; /* the offset in vm_file corresponding to vm_start */ |
374 | struct file *vm_file; /* the backing file or NULL */ |
375 | |
376 | int vm_usage; /* region usage count (access under nommu_region_sem) */ |
377 | bool vm_icache_flushed : 1; /* true if the icache has been flushed for |
378 | * this region */ |
379 | }; |
380 | |
381 | #ifdef CONFIG_USERFAULTFD |
382 | #define NULL_VM_UFFD_CTX ((struct vm_userfaultfd_ctx) { NULL, }) |
383 | struct vm_userfaultfd_ctx { |
384 | struct userfaultfd_ctx *ctx; |
385 | }; |
386 | #else /* CONFIG_USERFAULTFD */ |
387 | #define NULL_VM_UFFD_CTX ((struct vm_userfaultfd_ctx) {}) |
388 | struct vm_userfaultfd_ctx {}; |
389 | #endif /* CONFIG_USERFAULTFD */ |
390 | |
391 | struct anon_vma_name { |
392 | struct kref kref; |
393 | /* The name needs to be at the end because it is dynamically sized. */ |
394 | char name[]; |
395 | }; |
396 | |
397 | /* |
398 | * This struct describes a virtual memory area. There is one of these |
399 | * per VM-area/task. A VM area is any part of the process virtual memory |
400 | * space that has a special rule for the page-fault handlers (ie a shared |
401 | * library, the executable area etc). |
402 | */ |
403 | struct vm_area_struct { |
404 | /* The first cache line has the info for VMA tree walking. */ |
405 | |
406 | unsigned long vm_start; /* Our start address within vm_mm. */ |
407 | unsigned long vm_end; /* The first byte after our end address |
408 | within vm_mm. */ |
409 | |
410 | /* linked list of VM areas per task, sorted by address */ |
411 | struct vm_area_struct *vm_next, *vm_prev; |
412 | |
413 | struct rb_node vm_rb; |
414 | |
415 | /* |
416 | * Largest free memory gap in bytes to the left of this VMA. |
417 | * Either between this VMA and vma->vm_prev, or between one of the |
418 | * VMAs below us in the VMA rbtree and its ->vm_prev. This helps |
419 | * get_unmapped_area find a free area of the right size. |
420 | */ |
421 | unsigned long rb_subtree_gap; |
422 | |
423 | /* Second cache line starts here. */ |
424 | |
425 | struct mm_struct *vm_mm; /* The address space we belong to. */ |
426 | |
427 | /* |
428 | * Access permissions of this VMA. |
429 | * See vmf_insert_mixed_prot() for discussion. |
430 | */ |
431 | pgprot_t vm_page_prot; |
432 | unsigned long vm_flags; /* Flags, see mm.h. */ |
433 | |
434 | /* |
435 | * For areas with an address space and backing store, |
436 | * linkage into the address_space->i_mmap interval tree. |
437 | * |
438 | * For private anonymous mappings, a pointer to a null terminated string |
439 | * containing the name given to the vma, or NULL if unnamed. |
440 | */ |
441 | |
442 | union { |
443 | struct { |
444 | struct rb_node rb; |
445 | unsigned long rb_subtree_last; |
446 | } shared; |
447 | /* |
448 | * Serialized by mmap_sem. Never use directly because it is |
449 | * valid only when vm_file is NULL. Use anon_vma_name instead. |
450 | */ |
451 | struct anon_vma_name *anon_name; |
452 | }; |
453 | |
454 | /* |
455 | * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma |
456 | * list, after a COW of one of the file pages. A MAP_SHARED vma |
457 | * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack |
458 | * or brk vma (with NULL file) can only be in an anon_vma list. |
459 | */ |
460 | struct list_head anon_vma_chain; /* Serialized by mmap_lock & |
461 | * page_table_lock */ |
462 | struct anon_vma *anon_vma; /* Serialized by page_table_lock */ |
463 | |
464 | /* Function pointers to deal with this struct. */ |
465 | const struct vm_operations_struct *vm_ops; |
466 | |
467 | /* Information about our backing store: */ |
468 | unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE |
469 | units */ |
470 | struct file * vm_file; /* File we map to (can be NULL). */ |
471 | void * vm_private_data; /* was vm_pte (shared mem) */ |
472 | |
473 | #ifdef CONFIG_SWAP |
474 | atomic_long_t swap_readahead_info; |
475 | #endif |
476 | #ifndef CONFIG_MMU |
477 | struct vm_region *vm_region; /* NOMMU mapping region */ |
478 | #endif |
479 | #ifdef CONFIG_NUMA |
480 | struct mempolicy *vm_policy; /* NUMA policy for the VMA */ |
481 | #endif |
482 | struct vm_userfaultfd_ctx vm_userfaultfd_ctx; |
483 | } __randomize_layout; |
484 | |
485 | struct kioctx_table; |
486 | struct mm_struct { |
487 | struct { |
488 | struct vm_area_struct *mmap; /* list of VMAs */ |
489 | struct rb_root mm_rb; |
490 | u64 vmacache_seqnum; /* per-thread vmacache */ |
491 | #ifdef CONFIG_MMU |
492 | unsigned long (*get_unmapped_area) (struct file *filp, |
493 | unsigned long addr, unsigned long len, |
494 | unsigned long pgoff, unsigned long flags); |
495 | #endif |
496 | unsigned long mmap_base; /* base of mmap area */ |
497 | unsigned long mmap_legacy_base; /* base of mmap area in bottom-up allocations */ |
498 | #ifdef CONFIG_HAVE_ARCH_COMPAT_MMAP_BASES |
499 | /* Base addresses for compatible mmap() */ |
500 | unsigned long mmap_compat_base; |
501 | unsigned long mmap_compat_legacy_base; |
502 | #endif |
503 | unsigned long task_size; /* size of task vm space */ |
504 | unsigned long highest_vm_end; /* highest vma end address */ |
505 | pgd_t * pgd; |
506 | |
507 | #ifdef CONFIG_MEMBARRIER |
508 | /** |
509 | * @membarrier_state: Flags controlling membarrier behavior. |
510 | * |
511 | * This field is close to @pgd to hopefully fit in the same |
512 | * cache-line, which needs to be touched by switch_mm(). |
513 | */ |
514 | atomic_t membarrier_state; |
515 | #endif |
516 | |
517 | /** |
518 | * @mm_users: The number of users including userspace. |
519 | * |
520 | * Use mmget()/mmget_not_zero()/mmput() to modify. When this |
521 | * drops to 0 (i.e. when the task exits and there are no other |
522 | * temporary reference holders), we also release a reference on |
523 | * @mm_count (which may then free the &struct mm_struct if |
524 | * @mm_count also drops to 0). |
525 | */ |
526 | atomic_t mm_users; |
527 | |
528 | /** |
529 | * @mm_count: The number of references to &struct mm_struct |
530 | * (@mm_users count as 1). |
531 | * |
532 | * Use mmgrab()/mmdrop() to modify. When this drops to 0, the |
533 | * &struct mm_struct is freed. |
534 | */ |
535 | atomic_t mm_count; |
536 | |
537 | #ifdef CONFIG_MMU |
538 | atomic_long_t pgtables_bytes; /* PTE page table pages */ |
539 | #endif |
540 | int map_count; /* number of VMAs */ |
541 | |
542 | spinlock_t page_table_lock; /* Protects page tables and some |
543 | * counters |
544 | */ |
545 | /* |
546 | * With some kernel config, the current mmap_lock's offset |
547 | * inside 'mm_struct' is at 0x120, which is very optimal, as |
548 | * its two hot fields 'count' and 'owner' sit in 2 different |
549 | * cachelines, and when mmap_lock is highly contended, both |
550 | * of the 2 fields will be accessed frequently, current layout |
551 | * will help to reduce cache bouncing. |
552 | * |
553 | * So please be careful with adding new fields before |
554 | * mmap_lock, which can easily push the 2 fields into one |
555 | * cacheline. |
556 | */ |
557 | struct rw_semaphore mmap_lock; |
558 | |
559 | struct list_head mmlist; /* List of maybe swapped mm's. These |
560 | * are globally strung together off |
561 | * init_mm.mmlist, and are protected |
562 | * by mmlist_lock |
563 | */ |
564 | |
565 | |
566 | unsigned long ; /* High-watermark of RSS usage */ |
567 | unsigned long hiwater_vm; /* High-water virtual memory usage */ |
568 | |
569 | unsigned long total_vm; /* Total pages mapped */ |
570 | unsigned long locked_vm; /* Pages that have PG_mlocked set */ |
571 | atomic64_t pinned_vm; /* Refcount permanently increased */ |
572 | unsigned long data_vm; /* VM_WRITE & ~VM_SHARED & ~VM_STACK */ |
573 | unsigned long exec_vm; /* VM_EXEC & ~VM_WRITE & ~VM_STACK */ |
574 | unsigned long stack_vm; /* VM_STACK */ |
575 | unsigned long def_flags; |
576 | |
577 | /** |
578 | * @write_protect_seq: Locked when any thread is write |
579 | * protecting pages mapped by this mm to enforce a later COW, |
580 | * for instance during page table copying for fork(). |
581 | */ |
582 | seqcount_t write_protect_seq; |
583 | |
584 | spinlock_t arg_lock; /* protect the below fields */ |
585 | |
586 | unsigned long start_code, end_code, start_data, end_data; |
587 | unsigned long start_brk, brk, start_stack; |
588 | unsigned long arg_start, arg_end, env_start, env_end; |
589 | |
590 | unsigned long saved_auxv[AT_VECTOR_SIZE]; /* for /proc/PID/auxv */ |
591 | |
592 | /* |
593 | * Special counters, in some configurations protected by the |
594 | * page_table_lock, in other configurations by being atomic. |
595 | */ |
596 | struct mm_rss_stat ; |
597 | |
598 | struct linux_binfmt *binfmt; |
599 | |
600 | /* Architecture-specific MM context */ |
601 | mm_context_t context; |
602 | |
603 | unsigned long flags; /* Must use atomic bitops to access */ |
604 | |
605 | #ifdef CONFIG_AIO |
606 | spinlock_t ioctx_lock; |
607 | struct kioctx_table __rcu *ioctx_table; |
608 | #endif |
609 | #ifdef CONFIG_MEMCG |
610 | /* |
611 | * "owner" points to a task that is regarded as the canonical |
612 | * user/owner of this mm. All of the following must be true in |
613 | * order for it to be changed: |
614 | * |
615 | * current == mm->owner |
616 | * current->mm != mm |
617 | * new_owner->mm == mm |
618 | * new_owner->alloc_lock is held |
619 | */ |
620 | struct task_struct __rcu *owner; |
621 | #endif |
622 | struct user_namespace *user_ns; |
623 | |
624 | /* store ref to file /proc/<pid>/exe symlink points to */ |
625 | struct file __rcu *exe_file; |
626 | #ifdef CONFIG_MMU_NOTIFIER |
627 | struct mmu_notifier_subscriptions *notifier_subscriptions; |
628 | #endif |
629 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS |
630 | pgtable_t pmd_huge_pte; /* protected by page_table_lock */ |
631 | #endif |
632 | #ifdef CONFIG_NUMA_BALANCING |
633 | /* |
634 | * numa_next_scan is the next time that the PTEs will be marked |
635 | * pte_numa. NUMA hinting faults will gather statistics and |
636 | * migrate pages to new nodes if necessary. |
637 | */ |
638 | unsigned long numa_next_scan; |
639 | |
640 | /* Restart point for scanning and setting pte_numa */ |
641 | unsigned long numa_scan_offset; |
642 | |
643 | /* numa_scan_seq prevents two threads setting pte_numa */ |
644 | int numa_scan_seq; |
645 | #endif |
646 | /* |
647 | * An operation with batched TLB flushing is going on. Anything |
648 | * that can move process memory needs to flush the TLB when |
649 | * moving a PROT_NONE or PROT_NUMA mapped page. |
650 | */ |
651 | atomic_t tlb_flush_pending; |
652 | #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH |
653 | /* See flush_tlb_batched_pending() */ |
654 | atomic_t tlb_flush_batched; |
655 | #endif |
656 | struct uprobes_state uprobes_state; |
657 | #ifdef CONFIG_PREEMPT_RT |
658 | struct rcu_head delayed_drop; |
659 | #endif |
660 | #ifdef CONFIG_HUGETLB_PAGE |
661 | atomic_long_t hugetlb_usage; |
662 | #endif |
663 | struct work_struct async_put_work; |
664 | |
665 | #ifdef CONFIG_IOMMU_SVA |
666 | u32 pasid; |
667 | #endif |
668 | #ifdef CONFIG_KSM |
669 | /* |
670 | * Represent how many pages of this process are involved in KSM |
671 | * merging. |
672 | */ |
673 | unsigned long ksm_merging_pages; |
674 | #endif |
675 | } __randomize_layout; |
676 | |
677 | /* |
678 | * The mm_cpumask needs to be at the end of mm_struct, because it |
679 | * is dynamically sized based on nr_cpu_ids. |
680 | */ |
681 | unsigned long cpu_bitmap[]; |
682 | }; |
683 | |
684 | extern struct mm_struct init_mm; |
685 | |
686 | /* Pointer magic because the dynamic array size confuses some compilers. */ |
687 | static inline void mm_init_cpumask(struct mm_struct *mm) |
688 | { |
689 | unsigned long cpu_bitmap = (unsigned long)mm; |
690 | |
691 | cpu_bitmap += offsetof(struct mm_struct, cpu_bitmap); |
692 | cpumask_clear((struct cpumask *)cpu_bitmap); |
693 | } |
694 | |
695 | /* Future-safe accessor for struct mm_struct's cpu_vm_mask. */ |
696 | static inline cpumask_t *mm_cpumask(struct mm_struct *mm) |
697 | { |
698 | return (struct cpumask *)&mm->cpu_bitmap; |
699 | } |
700 | |
701 | struct mmu_gather; |
702 | extern void tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm); |
703 | extern void tlb_gather_mmu_fullmm(struct mmu_gather *tlb, struct mm_struct *mm); |
704 | extern void tlb_finish_mmu(struct mmu_gather *tlb); |
705 | |
706 | struct vm_fault; |
707 | |
708 | /** |
709 | * typedef vm_fault_t - Return type for page fault handlers. |
710 | * |
711 | * Page fault handlers return a bitmask of %VM_FAULT values. |
712 | */ |
713 | typedef __bitwise unsigned int vm_fault_t; |
714 | |
715 | /** |
716 | * enum vm_fault_reason - Page fault handlers return a bitmask of |
717 | * these values to tell the core VM what happened when handling the |
718 | * fault. Used to decide whether a process gets delivered SIGBUS or |
719 | * just gets major/minor fault counters bumped up. |
720 | * |
721 | * @VM_FAULT_OOM: Out Of Memory |
722 | * @VM_FAULT_SIGBUS: Bad access |
723 | * @VM_FAULT_MAJOR: Page read from storage |
724 | * @VM_FAULT_WRITE: Special case for get_user_pages |
725 | * @VM_FAULT_HWPOISON: Hit poisoned small page |
726 | * @VM_FAULT_HWPOISON_LARGE: Hit poisoned large page. Index encoded |
727 | * in upper bits |
728 | * @VM_FAULT_SIGSEGV: segmentation fault |
729 | * @VM_FAULT_NOPAGE: ->fault installed the pte, not return page |
730 | * @VM_FAULT_LOCKED: ->fault locked the returned page |
731 | * @VM_FAULT_RETRY: ->fault blocked, must retry |
732 | * @VM_FAULT_FALLBACK: huge page fault failed, fall back to small |
733 | * @VM_FAULT_DONE_COW: ->fault has fully handled COW |
734 | * @VM_FAULT_NEEDDSYNC: ->fault did not modify page tables and needs |
735 | * fsync() to complete (for synchronous page faults |
736 | * in DAX) |
737 | * @VM_FAULT_COMPLETED: ->fault completed, meanwhile mmap lock released |
738 | * @VM_FAULT_HINDEX_MASK: mask HINDEX value |
739 | * |
740 | */ |
741 | enum vm_fault_reason { |
742 | VM_FAULT_OOM = (__force vm_fault_t)0x000001, |
743 | VM_FAULT_SIGBUS = (__force vm_fault_t)0x000002, |
744 | VM_FAULT_MAJOR = (__force vm_fault_t)0x000004, |
745 | VM_FAULT_WRITE = (__force vm_fault_t)0x000008, |
746 | VM_FAULT_HWPOISON = (__force vm_fault_t)0x000010, |
747 | VM_FAULT_HWPOISON_LARGE = (__force vm_fault_t)0x000020, |
748 | VM_FAULT_SIGSEGV = (__force vm_fault_t)0x000040, |
749 | VM_FAULT_NOPAGE = (__force vm_fault_t)0x000100, |
750 | VM_FAULT_LOCKED = (__force vm_fault_t)0x000200, |
751 | VM_FAULT_RETRY = (__force vm_fault_t)0x000400, |
752 | VM_FAULT_FALLBACK = (__force vm_fault_t)0x000800, |
753 | VM_FAULT_DONE_COW = (__force vm_fault_t)0x001000, |
754 | VM_FAULT_NEEDDSYNC = (__force vm_fault_t)0x002000, |
755 | VM_FAULT_COMPLETED = (__force vm_fault_t)0x004000, |
756 | VM_FAULT_HINDEX_MASK = (__force vm_fault_t)0x0f0000, |
757 | }; |
758 | |
759 | /* Encode hstate index for a hwpoisoned large page */ |
760 | #define VM_FAULT_SET_HINDEX(x) ((__force vm_fault_t)((x) << 16)) |
761 | #define VM_FAULT_GET_HINDEX(x) (((__force unsigned int)(x) >> 16) & 0xf) |
762 | |
763 | #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | \ |
764 | VM_FAULT_SIGSEGV | VM_FAULT_HWPOISON | \ |
765 | VM_FAULT_HWPOISON_LARGE | VM_FAULT_FALLBACK) |
766 | |
767 | #define VM_FAULT_RESULT_TRACE \ |
768 | { VM_FAULT_OOM, "OOM" }, \ |
769 | { VM_FAULT_SIGBUS, "SIGBUS" }, \ |
770 | { VM_FAULT_MAJOR, "MAJOR" }, \ |
771 | { VM_FAULT_WRITE, "WRITE" }, \ |
772 | { VM_FAULT_HWPOISON, "HWPOISON" }, \ |
773 | { VM_FAULT_HWPOISON_LARGE, "HWPOISON_LARGE" }, \ |
774 | { VM_FAULT_SIGSEGV, "SIGSEGV" }, \ |
775 | { VM_FAULT_NOPAGE, "NOPAGE" }, \ |
776 | { VM_FAULT_LOCKED, "LOCKED" }, \ |
777 | { VM_FAULT_RETRY, "RETRY" }, \ |
778 | { VM_FAULT_FALLBACK, "FALLBACK" }, \ |
779 | { VM_FAULT_DONE_COW, "DONE_COW" }, \ |
780 | { VM_FAULT_NEEDDSYNC, "NEEDDSYNC" } |
781 | |
782 | struct vm_special_mapping { |
783 | const char *name; /* The name, e.g. "[vdso]". */ |
784 | |
785 | /* |
786 | * If .fault is not provided, this points to a |
787 | * NULL-terminated array of pages that back the special mapping. |
788 | * |
789 | * This must not be NULL unless .fault is provided. |
790 | */ |
791 | struct page **pages; |
792 | |
793 | /* |
794 | * If non-NULL, then this is called to resolve page faults |
795 | * on the special mapping. If used, .pages is not checked. |
796 | */ |
797 | vm_fault_t (*fault)(const struct vm_special_mapping *sm, |
798 | struct vm_area_struct *vma, |
799 | struct vm_fault *vmf); |
800 | |
801 | int (*mremap)(const struct vm_special_mapping *sm, |
802 | struct vm_area_struct *new_vma); |
803 | }; |
804 | |
805 | enum tlb_flush_reason { |
806 | TLB_FLUSH_ON_TASK_SWITCH, |
807 | TLB_REMOTE_SHOOTDOWN, |
808 | TLB_LOCAL_SHOOTDOWN, |
809 | TLB_LOCAL_MM_SHOOTDOWN, |
810 | TLB_REMOTE_SEND_IPI, |
811 | NR_TLB_FLUSH_REASONS, |
812 | }; |
813 | |
814 | /* |
815 | * A swap entry has to fit into a "unsigned long", as the entry is hidden |
816 | * in the "index" field of the swapper address space. |
817 | */ |
818 | typedef struct { |
819 | unsigned long val; |
820 | } swp_entry_t; |
821 | |
822 | /** |
823 | * enum fault_flag - Fault flag definitions. |
824 | * @FAULT_FLAG_WRITE: Fault was a write fault. |
825 | * @FAULT_FLAG_MKWRITE: Fault was mkwrite of existing PTE. |
826 | * @FAULT_FLAG_ALLOW_RETRY: Allow to retry the fault if blocked. |
827 | * @FAULT_FLAG_RETRY_NOWAIT: Don't drop mmap_lock and wait when retrying. |
828 | * @FAULT_FLAG_KILLABLE: The fault task is in SIGKILL killable region. |
829 | * @FAULT_FLAG_TRIED: The fault has been tried once. |
830 | * @FAULT_FLAG_USER: The fault originated in userspace. |
831 | * @FAULT_FLAG_REMOTE: The fault is not for current task/mm. |
832 | * @FAULT_FLAG_INSTRUCTION: The fault was during an instruction fetch. |
833 | * @FAULT_FLAG_INTERRUPTIBLE: The fault can be interrupted by non-fatal signals. |
834 | * @FAULT_FLAG_UNSHARE: The fault is an unsharing request to unshare (and mark |
835 | * exclusive) a possibly shared anonymous page that is |
836 | * mapped R/O. |
837 | * @FAULT_FLAG_ORIG_PTE_VALID: whether the fault has vmf->orig_pte cached. |
838 | * We should only access orig_pte if this flag set. |
839 | * |
840 | * About @FAULT_FLAG_ALLOW_RETRY and @FAULT_FLAG_TRIED: we can specify |
841 | * whether we would allow page faults to retry by specifying these two |
842 | * fault flags correctly. Currently there can be three legal combinations: |
843 | * |
844 | * (a) ALLOW_RETRY and !TRIED: this means the page fault allows retry, and |
845 | * this is the first try |
846 | * |
847 | * (b) ALLOW_RETRY and TRIED: this means the page fault allows retry, and |
848 | * we've already tried at least once |
849 | * |
850 | * (c) !ALLOW_RETRY and !TRIED: this means the page fault does not allow retry |
851 | * |
852 | * The unlisted combination (!ALLOW_RETRY && TRIED) is illegal and should never |
853 | * be used. Note that page faults can be allowed to retry for multiple times, |
854 | * in which case we'll have an initial fault with flags (a) then later on |
855 | * continuous faults with flags (b). We should always try to detect pending |
856 | * signals before a retry to make sure the continuous page faults can still be |
857 | * interrupted if necessary. |
858 | * |
859 | * The combination FAULT_FLAG_WRITE|FAULT_FLAG_UNSHARE is illegal. |
860 | * FAULT_FLAG_UNSHARE is ignored and treated like an ordinary read fault when |
861 | * no existing R/O-mapped anonymous page is encountered. |
862 | */ |
863 | enum fault_flag { |
864 | FAULT_FLAG_WRITE = 1 << 0, |
865 | FAULT_FLAG_MKWRITE = 1 << 1, |
866 | FAULT_FLAG_ALLOW_RETRY = 1 << 2, |
867 | FAULT_FLAG_RETRY_NOWAIT = 1 << 3, |
868 | FAULT_FLAG_KILLABLE = 1 << 4, |
869 | FAULT_FLAG_TRIED = 1 << 5, |
870 | FAULT_FLAG_USER = 1 << 6, |
871 | FAULT_FLAG_REMOTE = 1 << 7, |
872 | FAULT_FLAG_INSTRUCTION = 1 << 8, |
873 | FAULT_FLAG_INTERRUPTIBLE = 1 << 9, |
874 | FAULT_FLAG_UNSHARE = 1 << 10, |
875 | FAULT_FLAG_ORIG_PTE_VALID = 1 << 11, |
876 | }; |
877 | |
878 | typedef unsigned int __bitwise zap_flags_t; |
879 | |
880 | #endif /* _LINUX_MM_TYPES_H */ |
881 | |