]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blame - include/linux/mm_types.h
MAINTAINERS: add keyboard driver to Hyper-V file list
[mirror_ubuntu-zesty-kernel.git] / include / linux / mm_types.h
CommitLineData
5b99cd0e
HC
1#ifndef _LINUX_MM_TYPES_H
2#define _LINUX_MM_TYPES_H
3
4f9a58d7 4#include <linux/auxvec.h>
5b99cd0e
HC
5#include <linux/types.h>
6#include <linux/threads.h>
7#include <linux/list.h>
8#include <linux/spinlock.h>
c92ff1bd
MS
9#include <linux/rbtree.h>
10#include <linux/rwsem.h>
11#include <linux/completion.h>
cddb8a5c 12#include <linux/cpumask.h>
6a11f75b 13#include <linux/page-debug-flags.h>
d4b3b638 14#include <linux/uprobes.h>
bbeae5b0 15#include <linux/page-flags-layout.h>
c92ff1bd
MS
16#include <asm/page.h>
17#include <asm/mmu.h>
5b99cd0e 18
4f9a58d7
OH
19#ifndef AT_VECTOR_SIZE_ARCH
20#define AT_VECTOR_SIZE_ARCH 0
21#endif
22#define AT_VECTOR_SIZE (2*(AT_VECTOR_SIZE_ARCH + AT_VECTOR_SIZE_BASE + 1))
23
5b99cd0e
HC
24struct address_space;
25
57c1ffce 26#define USE_SPLIT_PTE_PTLOCKS (NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS)
e009bb30
KS
27#define USE_SPLIT_PMD_PTLOCKS (USE_SPLIT_PTE_PTLOCKS && \
28 IS_ENABLED(CONFIG_ARCH_ENABLE_SPLIT_PMD_PTLOCK))
f7d0b926 29
5b99cd0e
HC
30/*
31 * Each physical page in the system has a struct page associated with
32 * it to keep track of whatever it is we are using the page for at the
33 * moment. Note that we have no way to track which tasks are using
34 * a page, though if it is a pagecache page, rmap structures can tell us
35 * who is mapping it.
fc9bb8c7
CL
36 *
37 * The objects in struct page are organized in double word blocks in
38 * order to allows us to use atomic double word operations on portions
39 * of struct page. That is currently only used by slub but the arrangement
40 * allows the use of atomic double word operations on the flags/mapping
41 * and lru list pointers also.
5b99cd0e
HC
42 */
43struct page {
fc9bb8c7 44 /* First double word block */
5b99cd0e
HC
45 unsigned long flags; /* Atomic flags, some possibly
46 * updated asynchronously */
fc9bb8c7
CL
47 struct address_space *mapping; /* If low bit clear, points to
48 * inode address_space, or NULL.
49 * If page mapped as anonymous
50 * memory, low bit is set, and
51 * it points to anon_vma object:
52 * see PAGE_MAPPING_ANON below.
53 */
54 /* Second double word */
013e8963
CL
55 struct {
56 union {
fc9bb8c7 57 pgoff_t index; /* Our offset within mapping. */
b8c24c4a 58 void *freelist; /* slub/slob first free object */
072bb0aa 59 bool pfmemalloc; /* If set by the page allocator,
b37f1dd0 60 * ALLOC_NO_WATERMARKS was set
072bb0aa
MG
61 * and the low watermark was not
62 * met implying that the system
63 * is under some pressure. The
64 * caller should try ensure
65 * this page is only used to
66 * free other pages.
67 */
e009bb30
KS
68#if defined(CONFIG_TRANSPARENT_HUGEPAGE) && USE_SPLIT_PMD_PTLOCKS
69 pgtable_t pmd_huge_pte; /* protected by page->ptl */
70#endif
013e8963
CL
71 };
72
73 union {
abca7c49
PS
74#if defined(CONFIG_HAVE_CMPXCHG_DOUBLE) && \
75 defined(CONFIG_HAVE_ALIGNED_STRUCT_PAGE)
013e8963
CL
76 /* Used for cmpxchg_double in slub */
77 unsigned long counters;
abca7c49
PS
78#else
79 /*
80 * Keep _count separate from slub cmpxchg_double data.
81 * As the rest of the double word is protected by
82 * slab_lock but _count is not.
83 */
84 unsigned counters;
85#endif
013e8963
CL
86
87 struct {
88
89 union {
70b50f94
AA
90 /*
91 * Count of ptes mapped in
92 * mms, to show when page is
93 * mapped & limit reverse map
94 * searches.
95 *
96 * Used also for tail pages
97 * refcounting instead of
98 * _count. Tail pages cannot
99 * be mapped and keeping the
100 * tail page _count zero at
101 * all times guarantees
102 * get_page_unless_zero() will
103 * never succeed on tail
104 * pages.
105 */
106 atomic_t _mapcount;
fc9bb8c7 107
b8c24c4a 108 struct { /* SLUB */
013e8963
CL
109 unsigned inuse:16;
110 unsigned objects:15;
111 unsigned frozen:1;
112 };
b8c24c4a 113 int units; /* SLOB */
3adf004d 114 };
013e8963 115 atomic_t _count; /* Usage count, see below. */
fc9bb8c7 116 };
39b26464 117 };
81819f0f 118 };
fc9bb8c7
CL
119
120 /* Third double word block */
49e22585
CL
121 union {
122 struct list_head lru; /* Pageout list, eg. active_list
fc9bb8c7
CL
123 * protected by zone->lru_lock !
124 */
49e22585
CL
125 struct { /* slub per cpu partial pages */
126 struct page *next; /* Next partial slab */
127#ifdef CONFIG_64BIT
128 int pages; /* Nr of partial slabs left */
129 int pobjects; /* Approximate # of objects */
130#else
131 short int pages;
132 short int pobjects;
133#endif
134 };
b8c24c4a
CL
135
136 struct list_head list; /* slobs list of pages */
1b4f59e3 137 struct slab *slab_page; /* slab fields */
49e22585 138 };
fc9bb8c7
CL
139
140 /* Remainder is not double word aligned */
5b99cd0e 141 union {
5b99cd0e
HC
142 unsigned long private; /* Mapping-private opaque data:
143 * usually used for buffer_heads
144 * if PagePrivate set; used for
145 * swp_entry_t if PageSwapCache;
146 * indicates order in the buddy
147 * system if PG_buddy is set.
148 */
57c1ffce 149#if USE_SPLIT_PTE_PTLOCKS
539edb58
PZ
150#if BLOATED_SPINLOCKS
151 spinlock_t *ptl;
152#else
153 spinlock_t ptl;
154#endif
5b99cd0e 155#endif
1b4f59e3 156 struct kmem_cache *slab_cache; /* SL[AU]B: Pointer to slab */
fc9bb8c7 157 struct page *first_page; /* Compound tail pages */
81819f0f 158 };
fc9bb8c7 159
5b99cd0e
HC
160 /*
161 * On machines where all RAM is mapped into kernel address space,
162 * we can simply calculate the virtual address. On machines with
163 * highmem some memory is mapped into kernel virtual memory
164 * dynamically, so we need a place to store that address.
165 * Note that this field could be 16 bits on x86 ... ;)
166 *
167 * Architectures with slow multiplication can define
168 * WANT_PAGE_VIRTUAL in asm/page.h
169 */
170#if defined(WANT_PAGE_VIRTUAL)
171 void *virtual; /* Kernel virtual address (NULL if
172 not kmapped, ie. highmem) */
173#endif /* WANT_PAGE_VIRTUAL */
ee3b4290
AM
174#ifdef CONFIG_WANT_PAGE_DEBUG_FLAGS
175 unsigned long debug_flags; /* Use atomic bitops on this */
176#endif
dfec072e
VN
177
178#ifdef CONFIG_KMEMCHECK
179 /*
180 * kmemcheck wants to track the status of each byte in a page; this
181 * is a pointer to such a status block. NULL if not tracked.
182 */
183 void *shadow;
184#endif
57e0a030 185
90572890
PZ
186#ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
187 int _last_cpupid;
57e0a030 188#endif
fc9bb8c7
CL
189}
190/*
43570fd2
HC
191 * The struct page can be forced to be double word aligned so that atomic ops
192 * on double words work. The SLUB allocator can make use of such a feature.
fc9bb8c7 193 */
43570fd2
HC
194#ifdef CONFIG_HAVE_ALIGNED_STRUCT_PAGE
195 __aligned(2 * sizeof(unsigned long))
fc9bb8c7
CL
196#endif
197;
5b99cd0e 198
30d3c128
IC
199struct page_frag {
200 struct page *page;
201#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
202 __u32 offset;
203 __u32 size;
204#else
205 __u16 offset;
206 __u16 size;
207#endif
208};
209
ca16d140
KM
210typedef unsigned long __nocast vm_flags_t;
211
8feae131
DH
212/*
213 * A region containing a mapping of a non-memory backed file under NOMMU
214 * conditions. These are held in a global tree and are pinned by the VMAs that
215 * map parts of them.
216 */
217struct vm_region {
218 struct rb_node vm_rb; /* link in global region tree */
ca16d140 219 vm_flags_t vm_flags; /* VMA vm_flags */
8feae131
DH
220 unsigned long vm_start; /* start address of region */
221 unsigned long vm_end; /* region initialised to here */
dd8632a1 222 unsigned long vm_top; /* region allocated to here */
8feae131
DH
223 unsigned long vm_pgoff; /* the offset in vm_file corresponding to vm_start */
224 struct file *vm_file; /* the backing file or NULL */
225
1e2ae599 226 int vm_usage; /* region usage count (access under nommu_region_sem) */
cfe79c00
MF
227 bool vm_icache_flushed : 1; /* true if the icache has been flushed for
228 * this region */
8feae131
DH
229};
230
c92ff1bd
MS
231/*
232 * This struct defines a memory VMM memory area. There is one of these
233 * per VM-area/task. A VM area is any part of the process virtual memory
234 * space that has a special rule for the page-fault handlers (ie a shared
235 * library, the executable area etc).
236 */
237struct vm_area_struct {
e4c6bfd2
RR
238 /* The first cache line has the info for VMA tree walking. */
239
c92ff1bd
MS
240 unsigned long vm_start; /* Our start address within vm_mm. */
241 unsigned long vm_end; /* The first byte after our end address
242 within vm_mm. */
243
244 /* linked list of VM areas per task, sorted by address */
297c5eee 245 struct vm_area_struct *vm_next, *vm_prev;
c92ff1bd 246
c92ff1bd
MS
247 struct rb_node vm_rb;
248
d3737187
ML
249 /*
250 * Largest free memory gap in bytes to the left of this VMA.
251 * Either between this VMA and vma->vm_prev, or between one of the
252 * VMAs below us in the VMA rbtree and its ->vm_prev. This helps
253 * get_unmapped_area find a free area of the right size.
254 */
255 unsigned long rb_subtree_gap;
256
e4c6bfd2
RR
257 /* Second cache line starts here. */
258
259 struct mm_struct *vm_mm; /* The address space we belong to. */
260 pgprot_t vm_page_prot; /* Access permissions of this VMA. */
261 unsigned long vm_flags; /* Flags, see mm.h. */
262
c92ff1bd
MS
263 /*
264 * For areas with an address space and backing store,
6b2dbba8 265 * linkage into the address_space->i_mmap interval tree, or
c92ff1bd
MS
266 * linkage of vma in the address_space->i_mmap_nonlinear list.
267 */
268 union {
269 struct {
6b2dbba8
ML
270 struct rb_node rb;
271 unsigned long rb_subtree_last;
272 } linear;
273 struct list_head nonlinear;
c92ff1bd
MS
274 } shared;
275
276 /*
277 * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma
278 * list, after a COW of one of the file pages. A MAP_SHARED vma
279 * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack
280 * or brk vma (with NULL file) can only be in an anon_vma list.
281 */
5beb4930
RR
282 struct list_head anon_vma_chain; /* Serialized by mmap_sem &
283 * page_table_lock */
c92ff1bd
MS
284 struct anon_vma *anon_vma; /* Serialized by page_table_lock */
285
286 /* Function pointers to deal with this struct. */
f0f37e2f 287 const struct vm_operations_struct *vm_ops;
c92ff1bd
MS
288
289 /* Information about our backing store: */
290 unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE
291 units, *not* PAGE_CACHE_SIZE */
292 struct file * vm_file; /* File we map to (can be NULL). */
293 void * vm_private_data; /* was vm_pte (shared mem) */
c92ff1bd
MS
294
295#ifndef CONFIG_MMU
8feae131 296 struct vm_region *vm_region; /* NOMMU mapping region */
c92ff1bd
MS
297#endif
298#ifdef CONFIG_NUMA
299 struct mempolicy *vm_policy; /* NUMA policy for the VMA */
300#endif
301};
302
b564daf8
ON
303struct core_thread {
304 struct task_struct *task;
305 struct core_thread *next;
306};
307
32ecb1f2 308struct core_state {
c5f1cc8c 309 atomic_t nr_threads;
b564daf8 310 struct core_thread dumper;
32ecb1f2
ON
311 struct completion startup;
312};
313
d559db08
KH
314enum {
315 MM_FILEPAGES,
316 MM_ANONPAGES,
b084d435 317 MM_SWAPENTS,
d559db08
KH
318 NR_MM_COUNTERS
319};
320
57c1ffce 321#if USE_SPLIT_PTE_PTLOCKS && defined(CONFIG_MMU)
34e55232 322#define SPLIT_RSS_COUNTING
34e55232
KH
323/* per-thread cached information, */
324struct task_rss_stat {
325 int events; /* for synchronization threshold */
326 int count[NR_MM_COUNTERS];
327};
57c1ffce 328#endif /* USE_SPLIT_PTE_PTLOCKS */
172703b0 329
d559db08 330struct mm_rss_stat {
172703b0 331 atomic_long_t count[NR_MM_COUNTERS];
d559db08 332};
d559db08 333
db446a08 334struct kioctx_table;
c92ff1bd
MS
335struct mm_struct {
336 struct vm_area_struct * mmap; /* list of VMAs */
337 struct rb_root mm_rb;
338 struct vm_area_struct * mmap_cache; /* last find_vma result */
efc1a3b1 339#ifdef CONFIG_MMU
c92ff1bd
MS
340 unsigned long (*get_unmapped_area) (struct file *filp,
341 unsigned long addr, unsigned long len,
342 unsigned long pgoff, unsigned long flags);
efc1a3b1 343#endif
c92ff1bd 344 unsigned long mmap_base; /* base of mmap area */
41aacc1e 345 unsigned long mmap_legacy_base; /* base of mmap area in bottom-up allocations */
c92ff1bd 346 unsigned long task_size; /* size of task vm space */
d3737187 347 unsigned long highest_vm_end; /* highest vma end address */
c92ff1bd
MS
348 pgd_t * pgd;
349 atomic_t mm_users; /* How many users with user space? */
350 atomic_t mm_count; /* How many references to "struct mm_struct" (users count as 1) */
e1f56c89 351 atomic_long_t nr_ptes; /* Page table pages */
c92ff1bd 352 int map_count; /* number of VMAs */
481b4bb5 353
c92ff1bd 354 spinlock_t page_table_lock; /* Protects page tables and some counters */
481b4bb5 355 struct rw_semaphore mmap_sem;
c92ff1bd
MS
356
357 struct list_head mmlist; /* List of maybe swapped mm's. These are globally strung
358 * together off init_mm.mmlist, and are protected
359 * by mmlist_lock
360 */
361
c92ff1bd
MS
362
363 unsigned long hiwater_rss; /* High-watermark of RSS usage */
364 unsigned long hiwater_vm; /* High-water virtual memory usage */
365
e10d59f2
CL
366 unsigned long total_vm; /* Total pages mapped */
367 unsigned long locked_vm; /* Pages that have PG_mlocked set */
368 unsigned long pinned_vm; /* Refcount permanently increased */
369 unsigned long shared_vm; /* Shared pages (files) */
370 unsigned long exec_vm; /* VM_EXEC & ~VM_WRITE */
371 unsigned long stack_vm; /* VM_GROWSUP/DOWN */
e10d59f2 372 unsigned long def_flags;
c92ff1bd
MS
373 unsigned long start_code, end_code, start_data, end_data;
374 unsigned long start_brk, brk, start_stack;
375 unsigned long arg_start, arg_end, env_start, env_end;
376
377 unsigned long saved_auxv[AT_VECTOR_SIZE]; /* for /proc/PID/auxv */
378
d559db08
KH
379 /*
380 * Special counters, in some configurations protected by the
381 * page_table_lock, in other configurations by being atomic.
382 */
383 struct mm_rss_stat rss_stat;
384
801460d0
HS
385 struct linux_binfmt *binfmt;
386
6345d24d
LT
387 cpumask_var_t cpu_vm_mask_var;
388
c92ff1bd
MS
389 /* Architecture-specific MM context */
390 mm_context_t context;
391
c92ff1bd
MS
392 unsigned long flags; /* Must use atomic bitops to access the bits */
393
a94e2d40 394 struct core_state *core_state; /* coredumping support */
858f0993 395#ifdef CONFIG_AIO
db446a08
BL
396 spinlock_t ioctx_lock;
397 struct kioctx_table __rcu *ioctx_table;
858f0993 398#endif
cf475ad2 399#ifdef CONFIG_MM_OWNER
4cd1a8fc
KM
400 /*
401 * "owner" points to a task that is regarded as the canonical
402 * user/owner of this mm. All of the following must be true in
403 * order for it to be changed:
404 *
405 * current == mm->owner
406 * current->mm != mm
407 * new_owner->mm == mm
408 * new_owner->alloc_lock is held
409 */
4d2deb40 410 struct task_struct __rcu *owner;
78fb7466 411#endif
925d1c40 412
925d1c40
MH
413 /* store ref to file /proc/<pid>/exe symlink points to */
414 struct file *exe_file;
cddb8a5c
AA
415#ifdef CONFIG_MMU_NOTIFIER
416 struct mmu_notifier_mm *mmu_notifier_mm;
e7a00c45 417#endif
e009bb30 418#if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
e7a00c45 419 pgtable_t pmd_huge_pte; /* protected by page_table_lock */
cddb8a5c 420#endif
6345d24d
LT
421#ifdef CONFIG_CPUMASK_OFFSTACK
422 struct cpumask cpumask_allocation;
cbee9f88
PZ
423#endif
424#ifdef CONFIG_NUMA_BALANCING
425 /*
34f0315a
MG
426 * numa_next_scan is the next time that the PTEs will be marked
427 * pte_numa. NUMA hinting faults will gather statistics and migrate
428 * pages to new nodes if necessary.
cbee9f88
PZ
429 */
430 unsigned long numa_next_scan;
431
6e5fb223
PZ
432 /* Restart point for scanning and setting pte_numa */
433 unsigned long numa_scan_offset;
434
cbee9f88
PZ
435 /* numa_scan_seq prevents two threads setting pte_numa */
436 int numa_scan_seq;
6345d24d 437#endif
d4b3b638 438 struct uprobes_state uprobes_state;
c92ff1bd
MS
439};
440
6345d24d
LT
441static inline void mm_init_cpumask(struct mm_struct *mm)
442{
443#ifdef CONFIG_CPUMASK_OFFSTACK
444 mm->cpu_vm_mask_var = &mm->cpumask_allocation;
445#endif
446}
447
45e575ab 448/* Future-safe accessor for struct mm_struct's cpu_vm_mask. */
de03c72c
KM
449static inline cpumask_t *mm_cpumask(struct mm_struct *mm)
450{
451 return mm->cpu_vm_mask_var;
452}
45e575ab 453
5b99cd0e 454#endif /* _LINUX_MM_TYPES_H */