]>
Commit | Line | Data |
---|---|---|
b2441318 | 1 | /* SPDX-License-Identifier: GPL-2.0 */ |
5b99cd0e HC |
2 | #ifndef _LINUX_MM_TYPES_H |
3 | #define _LINUX_MM_TYPES_H | |
4 | ||
2e58f173 IM |
5 | #include <linux/mm_types_task.h> |
6 | ||
4f9a58d7 | 7 | #include <linux/auxvec.h> |
5b99cd0e HC |
8 | #include <linux/list.h> |
9 | #include <linux/spinlock.h> | |
c92ff1bd MS |
10 | #include <linux/rbtree.h> |
11 | #include <linux/rwsem.h> | |
12 | #include <linux/completion.h> | |
cddb8a5c | 13 | #include <linux/cpumask.h> |
d4b3b638 | 14 | #include <linux/uprobes.h> |
bbeae5b0 | 15 | #include <linux/page-flags-layout.h> |
ec8d7c14 | 16 | #include <linux/workqueue.h> |
57efa1fe | 17 | #include <linux/seqlock.h> |
2e58f173 | 18 | |
c92ff1bd | 19 | #include <asm/mmu.h> |
5b99cd0e | 20 | |
4f9a58d7 OH |
21 | #ifndef AT_VECTOR_SIZE_ARCH |
22 | #define AT_VECTOR_SIZE_ARCH 0 | |
23 | #endif | |
24 | #define AT_VECTOR_SIZE (2*(AT_VECTOR_SIZE_ARCH + AT_VECTOR_SIZE_BASE + 1)) | |
25 | ||
82e69a12 | 26 | #define INIT_PASID 0 |
1c8f4220 | 27 | |
5b99cd0e | 28 | struct address_space; |
1306a85a | 29 | struct mem_cgroup; |
5b99cd0e HC |
30 | |
31 | /* | |
32 | * Each physical page in the system has a struct page associated with | |
33 | * it to keep track of whatever it is we are using the page for at the | |
34 | * moment. Note that we have no way to track which tasks are using | |
35 | * a page, though if it is a pagecache page, rmap structures can tell us | |
97b4a671 | 36 | * who is mapping it. |
be50015d | 37 | * |
97b4a671 MW |
38 | * If you allocate the page using alloc_pages(), you can use some of the |
39 | * space in struct page for your own purposes. The five words in the main | |
40 | * union are available, except for bit 0 of the first word which must be | |
41 | * kept clear. Many users use this word to store a pointer to an object | |
42 | * which is guaranteed to be aligned. If you use the same storage as | |
43 | * page->mapping, you must restore it to NULL before freeing the page. | |
be50015d | 44 | * |
97b4a671 MW |
45 | * If your page will not be mapped to userspace, you can also use the four |
46 | * bytes in the mapcount union, but you must call page_mapcount_reset() | |
47 | * before freeing it. | |
48 | * | |
49 | * If you want to use the refcount field, it must be used in such a way | |
50 | * that other CPUs temporarily incrementing and then decrementing the | |
51 | * refcount does not cause problems. On receiving the page from | |
52 | * alloc_pages(), the refcount will be positive. | |
53 | * | |
54 | * If you allocate pages of order > 0, you can use some of the fields | |
55 | * in each subpage, but you may need to restore some of their values | |
56 | * afterwards. | |
fc9bb8c7 | 57 | * |
4cf7c8bf MW |
58 | * SLUB uses cmpxchg_double() to atomically update its freelist and |
59 | * counters. That requires that freelist & counters be adjacent and | |
60 | * double-word aligned. We align all struct pages to double-word | |
61 | * boundaries, and ensure that 'freelist' is aligned within the | |
62 | * struct. | |
5b99cd0e | 63 | */ |
e20df2c6 MW |
64 | #ifdef CONFIG_HAVE_ALIGNED_STRUCT_PAGE |
65 | #define _struct_page_alignment __aligned(2 * sizeof(unsigned long)) | |
66 | #else | |
0dd4da5b | 67 | #define _struct_page_alignment |
7d27a04b | 68 | #endif |
e20df2c6 | 69 | |
5b99cd0e HC |
70 | struct page { |
71 | unsigned long flags; /* Atomic flags, some possibly | |
72 | * updated asynchronously */ | |
b7ccc7f8 | 73 | /* |
4da1984e MW |
74 | * Five words (20/40 bytes) are available in this union. |
75 | * WARNING: bit 0 of the first word is used for PageTail(). That | |
76 | * means the other users of this union MUST NOT use the bit to | |
b7ccc7f8 MW |
77 | * avoid collision and false-positive PageTail(). |
78 | */ | |
8456a648 | 79 | union { |
66a6ffd2 | 80 | struct { /* Page cache and anonymous pages */ |
4da1984e MW |
81 | /** |
82 | * @lru: Pageout list, eg. active_list protected by | |
15b44736 | 83 | * lruvec->lru_lock. Sometimes used as a generic list |
4da1984e MW |
84 | * by the page owner. |
85 | */ | |
86 | struct list_head lru; | |
66a6ffd2 MW |
87 | /* See page-flags.h for PAGE_MAPPING_FLAGS */ |
88 | struct address_space *mapping; | |
89 | pgoff_t index; /* Our offset within mapping. */ | |
90 | /** | |
91 | * @private: Mapping-private opaque data. | |
92 | * Usually used for buffer_heads if PagePrivate. | |
93 | * Used for swp_entry_t if PageSwapCache. | |
94 | * Indicates order in the buddy system if PageBuddy. | |
95 | */ | |
96 | unsigned long private; | |
97 | }; | |
c25fff71 JDB |
98 | struct { /* page_pool used by netstack */ |
99 | /** | |
9ddb3c14 | 100 | * @dma_addr: might require a 64-bit value on |
c25fff71 JDB |
101 | * 32-bit architectures. |
102 | */ | |
9ddb3c14 | 103 | unsigned long dma_addr[2]; |
c25fff71 | 104 | }; |
66a6ffd2 | 105 | struct { /* slab, slob and slub */ |
4da1984e | 106 | union { |
3e05617c | 107 | struct list_head slab_list; |
4da1984e MW |
108 | struct { /* Partial pages */ |
109 | struct page *next; | |
110 | #ifdef CONFIG_64BIT | |
111 | int pages; /* Nr of pages left */ | |
112 | int pobjects; /* Approximate count */ | |
113 | #else | |
114 | short int pages; | |
115 | short int pobjects; | |
116 | #endif | |
117 | }; | |
118 | }; | |
66a6ffd2 MW |
119 | struct kmem_cache *slab_cache; /* not slob */ |
120 | /* Double-word boundary */ | |
121 | void *freelist; /* first free object */ | |
122 | union { | |
123 | void *s_mem; /* slab: first object */ | |
124 | unsigned long counters; /* SLUB */ | |
125 | struct { /* SLUB */ | |
126 | unsigned inuse:16; | |
127 | unsigned objects:15; | |
128 | unsigned frozen:1; | |
129 | }; | |
130 | }; | |
131 | }; | |
4da1984e MW |
132 | struct { /* Tail pages of compound page */ |
133 | unsigned long compound_head; /* Bit zero is set */ | |
134 | ||
135 | /* First tail page only */ | |
136 | unsigned char compound_dtor; | |
137 | unsigned char compound_order; | |
138 | atomic_t compound_mapcount; | |
1378a5ee | 139 | unsigned int compound_nr; /* 1 << compound_order */ |
4da1984e MW |
140 | }; |
141 | struct { /* Second tail page of compound page */ | |
142 | unsigned long _compound_pad_1; /* compound_head */ | |
47e29d32 | 143 | atomic_t hpage_pinned_refcount; |
87eaceb3 | 144 | /* For both global and memcg */ |
4da1984e MW |
145 | struct list_head deferred_list; |
146 | }; | |
66a6ffd2 | 147 | struct { /* Page table pages */ |
4da1984e MW |
148 | unsigned long _pt_pad_1; /* compound_head */ |
149 | pgtable_t pmd_huge_pte; /* protected by page->ptl */ | |
66a6ffd2 | 150 | unsigned long _pt_pad_2; /* mapping */ |
4231aba0 NP |
151 | union { |
152 | struct mm_struct *pt_mm; /* x86 pgds only */ | |
153 | atomic_t pt_frag_refcount; /* powerpc */ | |
154 | }; | |
7d27a04b | 155 | #if ALLOC_SPLIT_PTLOCKS |
66a6ffd2 | 156 | spinlock_t *ptl; |
7d27a04b | 157 | #else |
66a6ffd2 | 158 | spinlock_t ptl; |
7d27a04b | 159 | #endif |
7d27a04b | 160 | }; |
50e7fbc3 MW |
161 | struct { /* ZONE_DEVICE pages */ |
162 | /** @pgmap: Points to the hosting device page map. */ | |
163 | struct dev_pagemap *pgmap; | |
8a164fef | 164 | void *zone_device_data; |
76470ccd RC |
165 | /* |
166 | * ZONE_DEVICE private pages are counted as being | |
167 | * mapped so the next 3 words hold the mapping, index, | |
168 | * and private fields from the source anonymous or | |
169 | * page cache page while the page is migrated to device | |
170 | * private memory. | |
171 | * ZONE_DEVICE MEMORY_DEVICE_FS_DAX pages also | |
172 | * use the mapping, index, and private fields when | |
173 | * pmem backed DAX files are mapped. | |
174 | */ | |
50e7fbc3 | 175 | }; |
4da1984e MW |
176 | |
177 | /** @rcu_head: You can use this to free a page by RCU. */ | |
178 | struct rcu_head rcu_head; | |
7d27a04b MW |
179 | }; |
180 | ||
b21999da MW |
181 | union { /* This union is 4 bytes in size. */ |
182 | /* | |
183 | * If the page can be mapped to userspace, encodes the number | |
184 | * of times this page is referenced by a page table. | |
185 | */ | |
186 | atomic_t _mapcount; | |
187 | ||
6e292b9b MW |
188 | /* |
189 | * If the page is neither PageSlab nor mappable to userspace, | |
190 | * the value stored here may help determine what this page | |
191 | * is used for. See page-flags.h for a list of page types | |
192 | * which are currently stored here. | |
193 | */ | |
194 | unsigned int page_type; | |
195 | ||
ca9c88c7 | 196 | unsigned int active; /* SLAB */ |
ca9c88c7 | 197 | int units; /* SLOB */ |
81819f0f | 198 | }; |
fc9bb8c7 | 199 | |
b21999da MW |
200 | /* Usage count. *DO NOT USE DIRECTLY*. See page_ref.h */ |
201 | atomic_t _refcount; | |
202 | ||
1306a85a | 203 | #ifdef CONFIG_MEMCG |
bcfe06bf | 204 | unsigned long memcg_data; |
1306a85a JW |
205 | #endif |
206 | ||
5b99cd0e HC |
207 | /* |
208 | * On machines where all RAM is mapped into kernel address space, | |
209 | * we can simply calculate the virtual address. On machines with | |
210 | * highmem some memory is mapped into kernel virtual memory | |
211 | * dynamically, so we need a place to store that address. | |
212 | * Note that this field could be 16 bits on x86 ... ;) | |
213 | * | |
214 | * Architectures with slow multiplication can define | |
215 | * WANT_PAGE_VIRTUAL in asm/page.h | |
216 | */ | |
217 | #if defined(WANT_PAGE_VIRTUAL) | |
218 | void *virtual; /* Kernel virtual address (NULL if | |
219 | not kmapped, ie. highmem) */ | |
220 | #endif /* WANT_PAGE_VIRTUAL */ | |
dfec072e | 221 | |
90572890 PZ |
222 | #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS |
223 | int _last_cpupid; | |
57e0a030 | 224 | #endif |
e20df2c6 | 225 | } _struct_page_alignment; |
5b99cd0e | 226 | |
169226f7 YS |
227 | static inline atomic_t *compound_mapcount_ptr(struct page *page) |
228 | { | |
229 | return &page[1].compound_mapcount; | |
230 | } | |
231 | ||
47e29d32 JH |
232 | static inline atomic_t *compound_pincount_ptr(struct page *page) |
233 | { | |
234 | return &page[2].hpage_pinned_refcount; | |
235 | } | |
236 | ||
d1402fc7 LG |
237 | /* |
238 | * Used for sizing the vmemmap region on some architectures | |
239 | */ | |
240 | #define STRUCT_PAGE_MAX_SHIFT (order_base_2(sizeof(struct page))) | |
241 | ||
b63ae8ca AD |
242 | #define PAGE_FRAG_CACHE_MAX_SIZE __ALIGN_MASK(32768, ~PAGE_MASK) |
243 | #define PAGE_FRAG_CACHE_MAX_ORDER get_order(PAGE_FRAG_CACHE_MAX_SIZE) | |
244 | ||
b03641af | 245 | #define page_private(page) ((page)->private) |
60e65a6f GJ |
246 | |
247 | static inline void set_page_private(struct page *page, unsigned long private) | |
248 | { | |
249 | page->private = private; | |
250 | } | |
b03641af | 251 | |
b63ae8ca AD |
252 | struct page_frag_cache { |
253 | void * va; | |
254 | #if (PAGE_SIZE < PAGE_FRAG_CACHE_MAX_SIZE) | |
255 | __u16 offset; | |
256 | __u16 size; | |
257 | #else | |
258 | __u32 offset; | |
259 | #endif | |
260 | /* we maintain a pagecount bias, so that we dont dirty cache line | |
0139aa7b | 261 | * containing page->_refcount every time we allocate a fragment. |
b63ae8ca AD |
262 | */ |
263 | unsigned int pagecnt_bias; | |
264 | bool pfmemalloc; | |
265 | }; | |
266 | ||
64b990d2 | 267 | typedef unsigned long vm_flags_t; |
ca16d140 | 268 | |
8feae131 DH |
269 | /* |
270 | * A region containing a mapping of a non-memory backed file under NOMMU | |
271 | * conditions. These are held in a global tree and are pinned by the VMAs that | |
272 | * map parts of them. | |
273 | */ | |
274 | struct vm_region { | |
275 | struct rb_node vm_rb; /* link in global region tree */ | |
ca16d140 | 276 | vm_flags_t vm_flags; /* VMA vm_flags */ |
8feae131 DH |
277 | unsigned long vm_start; /* start address of region */ |
278 | unsigned long vm_end; /* region initialised to here */ | |
dd8632a1 | 279 | unsigned long vm_top; /* region allocated to here */ |
8feae131 DH |
280 | unsigned long vm_pgoff; /* the offset in vm_file corresponding to vm_start */ |
281 | struct file *vm_file; /* the backing file or NULL */ | |
282 | ||
1e2ae599 | 283 | int vm_usage; /* region usage count (access under nommu_region_sem) */ |
cfe79c00 MF |
284 | bool vm_icache_flushed : 1; /* true if the icache has been flushed for |
285 | * this region */ | |
8feae131 DH |
286 | }; |
287 | ||
745f234b AA |
288 | #ifdef CONFIG_USERFAULTFD |
289 | #define NULL_VM_UFFD_CTX ((struct vm_userfaultfd_ctx) { NULL, }) | |
290 | struct vm_userfaultfd_ctx { | |
291 | struct userfaultfd_ctx *ctx; | |
292 | }; | |
293 | #else /* CONFIG_USERFAULTFD */ | |
294 | #define NULL_VM_UFFD_CTX ((struct vm_userfaultfd_ctx) {}) | |
295 | struct vm_userfaultfd_ctx {}; | |
296 | #endif /* CONFIG_USERFAULTFD */ | |
297 | ||
c92ff1bd | 298 | /* |
552657b7 | 299 | * This struct describes a virtual memory area. There is one of these |
300 | * per VM-area/task. A VM area is any part of the process virtual memory | |
c92ff1bd MS |
301 | * space that has a special rule for the page-fault handlers (ie a shared |
302 | * library, the executable area etc). | |
303 | */ | |
304 | struct vm_area_struct { | |
e4c6bfd2 RR |
305 | /* The first cache line has the info for VMA tree walking. */ |
306 | ||
c92ff1bd MS |
307 | unsigned long vm_start; /* Our start address within vm_mm. */ |
308 | unsigned long vm_end; /* The first byte after our end address | |
309 | within vm_mm. */ | |
310 | ||
311 | /* linked list of VM areas per task, sorted by address */ | |
297c5eee | 312 | struct vm_area_struct *vm_next, *vm_prev; |
c92ff1bd | 313 | |
c92ff1bd MS |
314 | struct rb_node vm_rb; |
315 | ||
d3737187 ML |
316 | /* |
317 | * Largest free memory gap in bytes to the left of this VMA. | |
318 | * Either between this VMA and vma->vm_prev, or between one of the | |
319 | * VMAs below us in the VMA rbtree and its ->vm_prev. This helps | |
320 | * get_unmapped_area find a free area of the right size. | |
321 | */ | |
322 | unsigned long rb_subtree_gap; | |
323 | ||
e4c6bfd2 RR |
324 | /* Second cache line starts here. */ |
325 | ||
326 | struct mm_struct *vm_mm; /* The address space we belong to. */ | |
574c5b3d TH |
327 | |
328 | /* | |
329 | * Access permissions of this VMA. | |
330 | * See vmf_insert_mixed_prot() for discussion. | |
331 | */ | |
332 | pgprot_t vm_page_prot; | |
e4c6bfd2 RR |
333 | unsigned long vm_flags; /* Flags, see mm.h. */ |
334 | ||
c92ff1bd MS |
335 | /* |
336 | * For areas with an address space and backing store, | |
27ba0644 | 337 | * linkage into the address_space->i_mmap interval tree. |
c92ff1bd | 338 | */ |
ac51b934 KS |
339 | struct { |
340 | struct rb_node rb; | |
341 | unsigned long rb_subtree_last; | |
c92ff1bd MS |
342 | } shared; |
343 | ||
344 | /* | |
345 | * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma | |
346 | * list, after a COW of one of the file pages. A MAP_SHARED vma | |
347 | * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack | |
348 | * or brk vma (with NULL file) can only be in an anon_vma list. | |
349 | */ | |
c1e8d7c6 | 350 | struct list_head anon_vma_chain; /* Serialized by mmap_lock & |
5beb4930 | 351 | * page_table_lock */ |
c92ff1bd MS |
352 | struct anon_vma *anon_vma; /* Serialized by page_table_lock */ |
353 | ||
354 | /* Function pointers to deal with this struct. */ | |
f0f37e2f | 355 | const struct vm_operations_struct *vm_ops; |
c92ff1bd MS |
356 | |
357 | /* Information about our backing store: */ | |
358 | unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE | |
ea1754a0 | 359 | units */ |
c92ff1bd MS |
360 | struct file * vm_file; /* File we map to (can be NULL). */ |
361 | void * vm_private_data; /* was vm_pte (shared mem) */ | |
c92ff1bd | 362 | |
219f8a2e | 363 | #ifdef CONFIG_SWAP |
ec560175 | 364 | atomic_long_t swap_readahead_info; |
219f8a2e | 365 | #endif |
c92ff1bd | 366 | #ifndef CONFIG_MMU |
8feae131 | 367 | struct vm_region *vm_region; /* NOMMU mapping region */ |
c92ff1bd MS |
368 | #endif |
369 | #ifdef CONFIG_NUMA | |
370 | struct mempolicy *vm_policy; /* NUMA policy for the VMA */ | |
371 | #endif | |
745f234b | 372 | struct vm_userfaultfd_ctx vm_userfaultfd_ctx; |
3859a271 | 373 | } __randomize_layout; |
c92ff1bd | 374 | |
b564daf8 ON |
375 | struct core_thread { |
376 | struct task_struct *task; | |
377 | struct core_thread *next; | |
378 | }; | |
379 | ||
32ecb1f2 | 380 | struct core_state { |
c5f1cc8c | 381 | atomic_t nr_threads; |
b564daf8 | 382 | struct core_thread dumper; |
32ecb1f2 ON |
383 | struct completion startup; |
384 | }; | |
385 | ||
db446a08 | 386 | struct kioctx_table; |
c92ff1bd | 387 | struct mm_struct { |
c1a2f7f0 RR |
388 | struct { |
389 | struct vm_area_struct *mmap; /* list of VMAs */ | |
390 | struct rb_root mm_rb; | |
7a9cdebd | 391 | u64 vmacache_seqnum; /* per-thread vmacache */ |
efc1a3b1 | 392 | #ifdef CONFIG_MMU |
c1a2f7f0 | 393 | unsigned long (*get_unmapped_area) (struct file *filp, |
c92ff1bd MS |
394 | unsigned long addr, unsigned long len, |
395 | unsigned long pgoff, unsigned long flags); | |
efc1a3b1 | 396 | #endif |
c1a2f7f0 RR |
397 | unsigned long mmap_base; /* base of mmap area */ |
398 | unsigned long mmap_legacy_base; /* base of mmap area in bottom-up allocations */ | |
1b028f78 | 399 | #ifdef CONFIG_HAVE_ARCH_COMPAT_MMAP_BASES |
c1a2f7f0 RR |
400 | /* Base adresses for compatible mmap() */ |
401 | unsigned long mmap_compat_base; | |
402 | unsigned long mmap_compat_legacy_base; | |
1b028f78 | 403 | #endif |
c1a2f7f0 RR |
404 | unsigned long task_size; /* size of task vm space */ |
405 | unsigned long highest_vm_end; /* highest vma end address */ | |
406 | pgd_t * pgd; | |
407 | ||
227a4aad MD |
408 | #ifdef CONFIG_MEMBARRIER |
409 | /** | |
410 | * @membarrier_state: Flags controlling membarrier behavior. | |
411 | * | |
412 | * This field is close to @pgd to hopefully fit in the same | |
413 | * cache-line, which needs to be touched by switch_mm(). | |
414 | */ | |
415 | atomic_t membarrier_state; | |
416 | #endif | |
417 | ||
c1a2f7f0 RR |
418 | /** |
419 | * @mm_users: The number of users including userspace. | |
420 | * | |
421 | * Use mmget()/mmget_not_zero()/mmput() to modify. When this | |
422 | * drops to 0 (i.e. when the task exits and there are no other | |
423 | * temporary reference holders), we also release a reference on | |
424 | * @mm_count (which may then free the &struct mm_struct if | |
425 | * @mm_count also drops to 0). | |
426 | */ | |
427 | atomic_t mm_users; | |
428 | ||
429 | /** | |
430 | * @mm_count: The number of references to &struct mm_struct | |
431 | * (@mm_users count as 1). | |
432 | * | |
433 | * Use mmgrab()/mmdrop() to modify. When this drops to 0, the | |
434 | * &struct mm_struct is freed. | |
435 | */ | |
436 | atomic_t mm_count; | |
b279ddc3 | 437 | |
008cfe44 PX |
438 | /** |
439 | * @has_pinned: Whether this mm has pinned any pages. This can | |
440 | * be either replaced in the future by @pinned_vm when it | |
441 | * becomes stable, or grow into a counter on its own. We're | |
442 | * aggresive on this bit now - even if the pinned pages were | |
443 | * unpinned later on, we'll still keep this bit set for the | |
444 | * lifecycle of this mm just for simplicity. | |
445 | */ | |
446 | atomic_t has_pinned; | |
447 | ||
c4812909 | 448 | #ifdef CONFIG_MMU |
c1a2f7f0 | 449 | atomic_long_t pgtables_bytes; /* PTE page table pages */ |
5a3fbef3 | 450 | #endif |
c1a2f7f0 | 451 | int map_count; /* number of VMAs */ |
481b4bb5 | 452 | |
c1a2f7f0 RR |
453 | spinlock_t page_table_lock; /* Protects page tables and some |
454 | * counters | |
455 | */ | |
2e302543 FT |
456 | /* |
457 | * With some kernel config, the current mmap_lock's offset | |
458 | * inside 'mm_struct' is at 0x120, which is very optimal, as | |
459 | * its two hot fields 'count' and 'owner' sit in 2 different | |
460 | * cachelines, and when mmap_lock is highly contended, both | |
461 | * of the 2 fields will be accessed frequently, current layout | |
462 | * will help to reduce cache bouncing. | |
463 | * | |
464 | * So please be careful with adding new fields before | |
465 | * mmap_lock, which can easily push the 2 fields into one | |
466 | * cacheline. | |
467 | */ | |
da1c55f1 | 468 | struct rw_semaphore mmap_lock; |
c92ff1bd | 469 | |
c1a2f7f0 RR |
470 | struct list_head mmlist; /* List of maybe swapped mm's. These |
471 | * are globally strung together off | |
472 | * init_mm.mmlist, and are protected | |
473 | * by mmlist_lock | |
474 | */ | |
c92ff1bd | 475 | |
c92ff1bd | 476 | |
c1a2f7f0 RR |
477 | unsigned long hiwater_rss; /* High-watermark of RSS usage */ |
478 | unsigned long hiwater_vm; /* High-water virtual memory usage */ | |
c92ff1bd | 479 | |
c1a2f7f0 RR |
480 | unsigned long total_vm; /* Total pages mapped */ |
481 | unsigned long locked_vm; /* Pages that have PG_mlocked set */ | |
70f8a3ca | 482 | atomic64_t pinned_vm; /* Refcount permanently increased */ |
c1a2f7f0 RR |
483 | unsigned long data_vm; /* VM_WRITE & ~VM_SHARED & ~VM_STACK */ |
484 | unsigned long exec_vm; /* VM_EXEC & ~VM_WRITE & ~VM_STACK */ | |
485 | unsigned long stack_vm; /* VM_STACK */ | |
486 | unsigned long def_flags; | |
88aa7cc6 | 487 | |
2e302543 FT |
488 | /** |
489 | * @write_protect_seq: Locked when any thread is write | |
490 | * protecting pages mapped by this mm to enforce a later COW, | |
491 | * for instance during page table copying for fork(). | |
492 | */ | |
493 | seqcount_t write_protect_seq; | |
494 | ||
c1a2f7f0 | 495 | spinlock_t arg_lock; /* protect the below fields */ |
2e302543 | 496 | |
c1a2f7f0 RR |
497 | unsigned long start_code, end_code, start_data, end_data; |
498 | unsigned long start_brk, brk, start_stack; | |
499 | unsigned long arg_start, arg_end, env_start, env_end; | |
c92ff1bd | 500 | |
c1a2f7f0 | 501 | unsigned long saved_auxv[AT_VECTOR_SIZE]; /* for /proc/PID/auxv */ |
c92ff1bd | 502 | |
c1a2f7f0 RR |
503 | /* |
504 | * Special counters, in some configurations protected by the | |
505 | * page_table_lock, in other configurations by being atomic. | |
506 | */ | |
507 | struct mm_rss_stat rss_stat; | |
801460d0 | 508 | |
c1a2f7f0 | 509 | struct linux_binfmt *binfmt; |
6345d24d | 510 | |
c1a2f7f0 RR |
511 | /* Architecture-specific MM context */ |
512 | mm_context_t context; | |
c92ff1bd | 513 | |
c1a2f7f0 | 514 | unsigned long flags; /* Must use atomic bitops to access */ |
c92ff1bd | 515 | |
c1a2f7f0 | 516 | struct core_state *core_state; /* coredumping support */ |
227a4aad | 517 | |
858f0993 | 518 | #ifdef CONFIG_AIO |
c1a2f7f0 RR |
519 | spinlock_t ioctx_lock; |
520 | struct kioctx_table __rcu *ioctx_table; | |
858f0993 | 521 | #endif |
f98bafa0 | 522 | #ifdef CONFIG_MEMCG |
c1a2f7f0 RR |
523 | /* |
524 | * "owner" points to a task that is regarded as the canonical | |
525 | * user/owner of this mm. All of the following must be true in | |
526 | * order for it to be changed: | |
527 | * | |
528 | * current == mm->owner | |
529 | * current->mm != mm | |
530 | * new_owner->mm == mm | |
531 | * new_owner->alloc_lock is held | |
532 | */ | |
533 | struct task_struct __rcu *owner; | |
78fb7466 | 534 | #endif |
c1a2f7f0 | 535 | struct user_namespace *user_ns; |
925d1c40 | 536 | |
c1a2f7f0 RR |
537 | /* store ref to file /proc/<pid>/exe symlink points to */ |
538 | struct file __rcu *exe_file; | |
cddb8a5c | 539 | #ifdef CONFIG_MMU_NOTIFIER |
984cfe4e | 540 | struct mmu_notifier_subscriptions *notifier_subscriptions; |
e7a00c45 | 541 | #endif |
e009bb30 | 542 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS |
c1a2f7f0 | 543 | pgtable_t pmd_huge_pte; /* protected by page_table_lock */ |
cbee9f88 PZ |
544 | #endif |
545 | #ifdef CONFIG_NUMA_BALANCING | |
c1a2f7f0 RR |
546 | /* |
547 | * numa_next_scan is the next time that the PTEs will be marked | |
548 | * pte_numa. NUMA hinting faults will gather statistics and | |
549 | * migrate pages to new nodes if necessary. | |
550 | */ | |
551 | unsigned long numa_next_scan; | |
cbee9f88 | 552 | |
c1a2f7f0 RR |
553 | /* Restart point for scanning and setting pte_numa */ |
554 | unsigned long numa_scan_offset; | |
6e5fb223 | 555 | |
c1a2f7f0 RR |
556 | /* numa_scan_seq prevents two threads setting pte_numa */ |
557 | int numa_scan_seq; | |
20841405 | 558 | #endif |
c1a2f7f0 RR |
559 | /* |
560 | * An operation with batched TLB flushing is going on. Anything | |
561 | * that can move process memory needs to flush the TLB when | |
562 | * moving a PROT_NONE or PROT_NUMA mapped page. | |
563 | */ | |
564 | atomic_t tlb_flush_pending; | |
3ea27719 | 565 | #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH |
c1a2f7f0 RR |
566 | /* See flush_tlb_batched_pending() */ |
567 | bool tlb_flush_batched; | |
6345d24d | 568 | #endif |
c1a2f7f0 | 569 | struct uprobes_state uprobes_state; |
5d317b2b | 570 | #ifdef CONFIG_HUGETLB_PAGE |
c1a2f7f0 | 571 | atomic_long_t hugetlb_usage; |
5d317b2b | 572 | #endif |
c1a2f7f0 | 573 | struct work_struct async_put_work; |
52ad9bc6 FY |
574 | |
575 | #ifdef CONFIG_IOMMU_SUPPORT | |
576 | u32 pasid; | |
577 | #endif | |
c1a2f7f0 RR |
578 | } __randomize_layout; |
579 | ||
580 | /* | |
581 | * The mm_cpumask needs to be at the end of mm_struct, because it | |
582 | * is dynamically sized based on nr_cpu_ids. | |
583 | */ | |
584 | unsigned long cpu_bitmap[]; | |
585 | }; | |
c92ff1bd | 586 | |
abe722a1 IM |
587 | extern struct mm_struct init_mm; |
588 | ||
c1a2f7f0 | 589 | /* Pointer magic because the dynamic array size confuses some compilers. */ |
6345d24d LT |
590 | static inline void mm_init_cpumask(struct mm_struct *mm) |
591 | { | |
c1a2f7f0 RR |
592 | unsigned long cpu_bitmap = (unsigned long)mm; |
593 | ||
594 | cpu_bitmap += offsetof(struct mm_struct, cpu_bitmap); | |
595 | cpumask_clear((struct cpumask *)cpu_bitmap); | |
6345d24d LT |
596 | } |
597 | ||
45e575ab | 598 | /* Future-safe accessor for struct mm_struct's cpu_vm_mask. */ |
de03c72c KM |
599 | static inline cpumask_t *mm_cpumask(struct mm_struct *mm) |
600 | { | |
c1a2f7f0 | 601 | return (struct cpumask *)&mm->cpu_bitmap; |
de03c72c | 602 | } |
45e575ab | 603 | |
56236a59 | 604 | struct mmu_gather; |
a72afd87 | 605 | extern void tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm); |
d8b45053 | 606 | extern void tlb_gather_mmu_fullmm(struct mmu_gather *tlb, struct mm_struct *mm); |
ae8eba8b | 607 | extern void tlb_finish_mmu(struct mmu_gather *tlb); |
56236a59 | 608 | |
16af97dc | 609 | static inline void init_tlb_flush_pending(struct mm_struct *mm) |
20841405 | 610 | { |
16af97dc | 611 | atomic_set(&mm->tlb_flush_pending, 0); |
20841405 | 612 | } |
16af97dc NA |
613 | |
614 | static inline void inc_tlb_flush_pending(struct mm_struct *mm) | |
20841405 | 615 | { |
16af97dc | 616 | atomic_inc(&mm->tlb_flush_pending); |
af2c1401 | 617 | /* |
8b1b436d PZ |
618 | * The only time this value is relevant is when there are indeed pages |
619 | * to flush. And we'll only flush pages after changing them, which | |
620 | * requires the PTL. | |
621 | * | |
622 | * So the ordering here is: | |
623 | * | |
040cca3a | 624 | * atomic_inc(&mm->tlb_flush_pending); |
8b1b436d PZ |
625 | * spin_lock(&ptl); |
626 | * ... | |
627 | * set_pte_at(); | |
628 | * spin_unlock(&ptl); | |
629 | * | |
630 | * spin_lock(&ptl) | |
631 | * mm_tlb_flush_pending(); | |
632 | * .... | |
633 | * spin_unlock(&ptl); | |
634 | * | |
635 | * flush_tlb_range(); | |
040cca3a | 636 | * atomic_dec(&mm->tlb_flush_pending); |
8b1b436d | 637 | * |
0e709703 PZ |
638 | * Where the increment if constrained by the PTL unlock, it thus |
639 | * ensures that the increment is visible if the PTE modification is | |
640 | * visible. After all, if there is no PTE modification, nobody cares | |
641 | * about TLB flushes either. | |
642 | * | |
643 | * This very much relies on users (mm_tlb_flush_pending() and | |
644 | * mm_tlb_flush_nested()) only caring about _specific_ PTEs (and | |
645 | * therefore specific PTLs), because with SPLIT_PTE_PTLOCKS and RCpc | |
646 | * locks (PPC) the unlock of one doesn't order against the lock of | |
647 | * another PTL. | |
648 | * | |
649 | * The decrement is ordered by the flush_tlb_range(), such that | |
650 | * mm_tlb_flush_pending() will not return false unless all flushes have | |
651 | * completed. | |
af2c1401 | 652 | */ |
20841405 | 653 | } |
16af97dc | 654 | |
16af97dc | 655 | static inline void dec_tlb_flush_pending(struct mm_struct *mm) |
20841405 | 656 | { |
0a2c4048 | 657 | /* |
0e709703 PZ |
658 | * See inc_tlb_flush_pending(). |
659 | * | |
660 | * This cannot be smp_mb__before_atomic() because smp_mb() simply does | |
661 | * not order against TLB invalidate completion, which is what we need. | |
662 | * | |
663 | * Therefore we must rely on tlb_flush_*() to guarantee order. | |
0a2c4048 | 664 | */ |
16af97dc | 665 | atomic_dec(&mm->tlb_flush_pending); |
20841405 | 666 | } |
20841405 | 667 | |
0e709703 PZ |
668 | static inline bool mm_tlb_flush_pending(struct mm_struct *mm) |
669 | { | |
670 | /* | |
671 | * Must be called after having acquired the PTL; orders against that | |
672 | * PTLs release and therefore ensures that if we observe the modified | |
673 | * PTE we must also observe the increment from inc_tlb_flush_pending(). | |
674 | * | |
675 | * That is, it only guarantees to return true if there is a flush | |
676 | * pending for _this_ PTL. | |
677 | */ | |
678 | return atomic_read(&mm->tlb_flush_pending); | |
679 | } | |
680 | ||
681 | static inline bool mm_tlb_flush_nested(struct mm_struct *mm) | |
682 | { | |
683 | /* | |
684 | * Similar to mm_tlb_flush_pending(), we must have acquired the PTL | |
685 | * for which there is a TLB flush pending in order to guarantee | |
686 | * we've seen both that PTE modification and the increment. | |
687 | * | |
688 | * (no requirement on actually still holding the PTL, that is irrelevant) | |
689 | */ | |
690 | return atomic_read(&mm->tlb_flush_pending) > 1; | |
691 | } | |
692 | ||
f872f540 AL |
693 | struct vm_fault; |
694 | ||
3d353901 SJ |
695 | /** |
696 | * typedef vm_fault_t - Return type for page fault handlers. | |
697 | * | |
698 | * Page fault handlers return a bitmask of %VM_FAULT values. | |
699 | */ | |
700 | typedef __bitwise unsigned int vm_fault_t; | |
701 | ||
702 | /** | |
703 | * enum vm_fault_reason - Page fault handlers return a bitmask of | |
704 | * these values to tell the core VM what happened when handling the | |
705 | * fault. Used to decide whether a process gets delivered SIGBUS or | |
706 | * just gets major/minor fault counters bumped up. | |
707 | * | |
708 | * @VM_FAULT_OOM: Out Of Memory | |
709 | * @VM_FAULT_SIGBUS: Bad access | |
710 | * @VM_FAULT_MAJOR: Page read from storage | |
711 | * @VM_FAULT_WRITE: Special case for get_user_pages | |
712 | * @VM_FAULT_HWPOISON: Hit poisoned small page | |
713 | * @VM_FAULT_HWPOISON_LARGE: Hit poisoned large page. Index encoded | |
714 | * in upper bits | |
715 | * @VM_FAULT_SIGSEGV: segmentation fault | |
716 | * @VM_FAULT_NOPAGE: ->fault installed the pte, not return page | |
717 | * @VM_FAULT_LOCKED: ->fault locked the returned page | |
718 | * @VM_FAULT_RETRY: ->fault blocked, must retry | |
719 | * @VM_FAULT_FALLBACK: huge page fault failed, fall back to small | |
720 | * @VM_FAULT_DONE_COW: ->fault has fully handled COW | |
721 | * @VM_FAULT_NEEDDSYNC: ->fault did not modify page tables and needs | |
722 | * fsync() to complete (for synchronous page faults | |
723 | * in DAX) | |
724 | * @VM_FAULT_HINDEX_MASK: mask HINDEX value | |
725 | * | |
726 | */ | |
727 | enum vm_fault_reason { | |
728 | VM_FAULT_OOM = (__force vm_fault_t)0x000001, | |
729 | VM_FAULT_SIGBUS = (__force vm_fault_t)0x000002, | |
730 | VM_FAULT_MAJOR = (__force vm_fault_t)0x000004, | |
731 | VM_FAULT_WRITE = (__force vm_fault_t)0x000008, | |
732 | VM_FAULT_HWPOISON = (__force vm_fault_t)0x000010, | |
733 | VM_FAULT_HWPOISON_LARGE = (__force vm_fault_t)0x000020, | |
734 | VM_FAULT_SIGSEGV = (__force vm_fault_t)0x000040, | |
735 | VM_FAULT_NOPAGE = (__force vm_fault_t)0x000100, | |
736 | VM_FAULT_LOCKED = (__force vm_fault_t)0x000200, | |
737 | VM_FAULT_RETRY = (__force vm_fault_t)0x000400, | |
738 | VM_FAULT_FALLBACK = (__force vm_fault_t)0x000800, | |
739 | VM_FAULT_DONE_COW = (__force vm_fault_t)0x001000, | |
740 | VM_FAULT_NEEDDSYNC = (__force vm_fault_t)0x002000, | |
741 | VM_FAULT_HINDEX_MASK = (__force vm_fault_t)0x0f0000, | |
742 | }; | |
743 | ||
744 | /* Encode hstate index for a hwpoisoned large page */ | |
745 | #define VM_FAULT_SET_HINDEX(x) ((__force vm_fault_t)((x) << 16)) | |
fcae96ff | 746 | #define VM_FAULT_GET_HINDEX(x) (((__force unsigned int)(x) >> 16) & 0xf) |
3d353901 SJ |
747 | |
748 | #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | \ | |
749 | VM_FAULT_SIGSEGV | VM_FAULT_HWPOISON | \ | |
750 | VM_FAULT_HWPOISON_LARGE | VM_FAULT_FALLBACK) | |
751 | ||
752 | #define VM_FAULT_RESULT_TRACE \ | |
753 | { VM_FAULT_OOM, "OOM" }, \ | |
754 | { VM_FAULT_SIGBUS, "SIGBUS" }, \ | |
755 | { VM_FAULT_MAJOR, "MAJOR" }, \ | |
756 | { VM_FAULT_WRITE, "WRITE" }, \ | |
757 | { VM_FAULT_HWPOISON, "HWPOISON" }, \ | |
758 | { VM_FAULT_HWPOISON_LARGE, "HWPOISON_LARGE" }, \ | |
759 | { VM_FAULT_SIGSEGV, "SIGSEGV" }, \ | |
760 | { VM_FAULT_NOPAGE, "NOPAGE" }, \ | |
761 | { VM_FAULT_LOCKED, "LOCKED" }, \ | |
762 | { VM_FAULT_RETRY, "RETRY" }, \ | |
763 | { VM_FAULT_FALLBACK, "FALLBACK" }, \ | |
764 | { VM_FAULT_DONE_COW, "DONE_COW" }, \ | |
765 | { VM_FAULT_NEEDDSYNC, "NEEDDSYNC" } | |
766 | ||
f872f540 AL |
767 | struct vm_special_mapping { |
768 | const char *name; /* The name, e.g. "[vdso]". */ | |
769 | ||
770 | /* | |
771 | * If .fault is not provided, this points to a | |
772 | * NULL-terminated array of pages that back the special mapping. | |
773 | * | |
774 | * This must not be NULL unless .fault is provided. | |
775 | */ | |
a62c34bd | 776 | struct page **pages; |
f872f540 AL |
777 | |
778 | /* | |
779 | * If non-NULL, then this is called to resolve page faults | |
780 | * on the special mapping. If used, .pages is not checked. | |
781 | */ | |
b3ec9f33 SJ |
782 | vm_fault_t (*fault)(const struct vm_special_mapping *sm, |
783 | struct vm_area_struct *vma, | |
784 | struct vm_fault *vmf); | |
b059a453 DS |
785 | |
786 | int (*mremap)(const struct vm_special_mapping *sm, | |
787 | struct vm_area_struct *new_vma); | |
a62c34bd AL |
788 | }; |
789 | ||
d17d8f9d DH |
790 | enum tlb_flush_reason { |
791 | TLB_FLUSH_ON_TASK_SWITCH, | |
792 | TLB_REMOTE_SHOOTDOWN, | |
793 | TLB_LOCAL_SHOOTDOWN, | |
794 | TLB_LOCAL_MM_SHOOTDOWN, | |
5b74283a | 795 | TLB_REMOTE_SEND_IPI, |
d17d8f9d DH |
796 | NR_TLB_FLUSH_REASONS, |
797 | }; | |
798 | ||
bd6dace7 TH |
799 | /* |
800 | * A swap entry has to fit into a "unsigned long", as the entry is hidden | |
801 | * in the "index" field of the swapper address space. | |
802 | */ | |
803 | typedef struct { | |
804 | unsigned long val; | |
805 | } swp_entry_t; | |
806 | ||
5b99cd0e | 807 | #endif /* _LINUX_MM_TYPES_H */ |