1 #ifndef _LINUX_MM_TYPES_H
2 #define _LINUX_MM_TYPES_H
3 
4 #include <linux/auxvec.h>
5 #include <linux/types.h>
6 #include <linux/threads.h>
7 #include <linux/list.h>
8 #include <linux/spinlock.h>
9 #include <linux/prio_tree.h>
10 #include <linux/rbtree.h>
11 #include <linux/rwsem.h>
12 #include <linux/completion.h>
13 #include <linux/cpumask.h>
14 #include <linux/page-debug-flags.h>
15 #include <asm/page.h>
16 #include <asm/mmu.h>
17 
18 #ifndef AT_VECTOR_SIZE_ARCH
19 #define AT_VECTOR_SIZE_ARCH 0
20 #endif
21 #define AT_VECTOR_SIZE (2*(AT_VECTOR_SIZE_ARCH + AT_VECTOR_SIZE_BASE + 1))
22 
23 struct address_space;
24 
25 #define USE_SPLIT_PTLOCKS	(NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS)
26 
27 /*
28  * Each physical page in the system has a struct page associated with
29  * it to keep track of whatever it is we are using the page for at the
30  * moment. Note that we have no way to track which tasks are using
31  * a page, though if it is a pagecache page, rmap structures can tell us
32  * who is mapping it.
33  *
34  * The objects in struct page are organized in double word blocks in
35  * order to allows us to use atomic double word operations on portions
36  * of struct page. That is currently only used by slub but the arrangement
37  * allows the use of atomic double word operations on the flags/mapping
38  * and lru list pointers also.
39  */
40 struct page {
41 	/* First double word block */
42 	unsigned long flags;		/* Atomic flags, some possibly
43 					 * updated asynchronously */
44 	struct address_space *mapping;	/* If low bit clear, points to
45 					 * inode address_space, or NULL.
46 					 * If page mapped as anonymous
47 					 * memory, low bit is set, and
48 					 * it points to anon_vma object:
49 					 * see PAGE_MAPPING_ANON below.
50 					 */
51 	/* Second double word */
52 	struct {
53 		union {
54 			pgoff_t index;		/* Our offset within mapping. */
55 			void *freelist;		/* slub first free object */
56 		};
57 
58 		union {
59 			/* Used for cmpxchg_double in slub */
60 			unsigned long counters;
61 
62 			struct {
63 
64 				union {
65 					/*
66 					 * Count of ptes mapped in
67 					 * mms, to show when page is
68 					 * mapped & limit reverse map
69 					 * searches.
70 					 *
71 					 * Used also for tail pages
72 					 * refcounting instead of
73 					 * _count. Tail pages cannot
74 					 * be mapped and keeping the
75 					 * tail page _count zero at
76 					 * all times guarantees
77 					 * get_page_unless_zero() will
78 					 * never succeed on tail
79 					 * pages.
80 					 */
81 					atomic_t _mapcount;
82 
83 					struct {
84 						unsigned inuse:16;
85 						unsigned objects:15;
86 						unsigned frozen:1;
87 					};
88 				};
89 				atomic_t _count;		/* Usage count, see below. */
90 			};
91 		};
92 	};
93 
94 	/* Third double word block */
95 	union {
96 		struct list_head lru;	/* Pageout list, eg. active_list
97 					 * protected by zone->lru_lock !
98 					 */
99 		struct {		/* slub per cpu partial pages */
100 			struct page *next;	/* Next partial slab */
101 #ifdef CONFIG_64BIT
102 			int pages;	/* Nr of partial slabs left */
103 			int pobjects;	/* Approximate # of objects */
104 #else
105 			short int pages;
106 			short int pobjects;
107 #endif
108 		};
109 	};
110 
111 	/* Remainder is not double word aligned */
112 	union {
113 		unsigned long private;		/* Mapping-private opaque data:
114 					 	 * usually used for buffer_heads
115 						 * if PagePrivate set; used for
116 						 * swp_entry_t if PageSwapCache;
117 						 * indicates order in the buddy
118 						 * system if PG_buddy is set.
119 						 */
120 #if USE_SPLIT_PTLOCKS
121 		spinlock_t ptl;
122 #endif
123 		struct kmem_cache *slab;	/* SLUB: Pointer to slab */
124 		struct page *first_page;	/* Compound tail pages */
125 	};
126 
127 	/*
128 	 * On machines where all RAM is mapped into kernel address space,
129 	 * we can simply calculate the virtual address. On machines with
130 	 * highmem some memory is mapped into kernel virtual memory
131 	 * dynamically, so we need a place to store that address.
132 	 * Note that this field could be 16 bits on x86 ... ;)
133 	 *
134 	 * Architectures with slow multiplication can define
135 	 * WANT_PAGE_VIRTUAL in asm/page.h
136 	 */
137 #if defined(WANT_PAGE_VIRTUAL)
138 	void *virtual;			/* Kernel virtual address (NULL if
139 					   not kmapped, ie. highmem) */
140 #endif /* WANT_PAGE_VIRTUAL */
141 #ifdef CONFIG_WANT_PAGE_DEBUG_FLAGS
142 	unsigned long debug_flags;	/* Use atomic bitops on this */
143 #endif
144 
145 #ifdef CONFIG_KMEMCHECK
146 	/*
147 	 * kmemcheck wants to track the status of each byte in a page; this
148 	 * is a pointer to such a status block. NULL if not tracked.
149 	 */
150 	void *shadow;
151 #endif
152 }
153 /*
154  * The struct page can be forced to be double word aligned so that atomic ops
155  * on double words work. The SLUB allocator can make use of such a feature.
156  */
157 #ifdef CONFIG_HAVE_ALIGNED_STRUCT_PAGE
158 	__aligned(2 * sizeof(unsigned long))
159 #endif
160 ;
161 
162 struct page_frag {
163 	struct page *page;
164 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
165 	__u32 offset;
166 	__u32 size;
167 #else
168 	__u16 offset;
169 	__u16 size;
170 #endif
171 };
172 
173 typedef unsigned long __nocast vm_flags_t;
174 
175 /*
176  * A region containing a mapping of a non-memory backed file under NOMMU
177  * conditions.  These are held in a global tree and are pinned by the VMAs that
178  * map parts of them.
179  */
180 struct vm_region {
181 	struct rb_node	vm_rb;		/* link in global region tree */
182 	vm_flags_t	vm_flags;	/* VMA vm_flags */
183 	unsigned long	vm_start;	/* start address of region */
184 	unsigned long	vm_end;		/* region initialised to here */
185 	unsigned long	vm_top;		/* region allocated to here */
186 	unsigned long	vm_pgoff;	/* the offset in vm_file corresponding to vm_start */
187 	struct file	*vm_file;	/* the backing file or NULL */
188 
189 	int		vm_usage;	/* region usage count (access under nommu_region_sem) */
190 	bool		vm_icache_flushed : 1; /* true if the icache has been flushed for
191 						* this region */
192 };
193 
194 /*
195  * This struct defines a memory VMM memory area. There is one of these
196  * per VM-area/task.  A VM area is any part of the process virtual memory
197  * space that has a special rule for the page-fault handlers (ie a shared
198  * library, the executable area etc).
199  */
200 struct vm_area_struct {
201 	struct mm_struct * vm_mm;	/* The address space we belong to. */
202 	unsigned long vm_start;		/* Our start address within vm_mm. */
203 	unsigned long vm_end;		/* The first byte after our end address
204 					   within vm_mm. */
205 
206 	/* linked list of VM areas per task, sorted by address */
207 	struct vm_area_struct *vm_next, *vm_prev;
208 
209 	pgprot_t vm_page_prot;		/* Access permissions of this VMA. */
210 	unsigned long vm_flags;		/* Flags, see mm.h. */
211 
212 	struct rb_node vm_rb;
213 
214 	/*
215 	 * For areas with an address space and backing store,
216 	 * linkage into the address_space->i_mmap prio tree, or
217 	 * linkage to the list of like vmas hanging off its node, or
218 	 * linkage of vma in the address_space->i_mmap_nonlinear list.
219 	 */
220 	union {
221 		struct {
222 			struct list_head list;
223 			void *parent;	/* aligns with prio_tree_node parent */
224 			struct vm_area_struct *head;
225 		} vm_set;
226 
227 		struct raw_prio_tree_node prio_tree_node;
228 	} shared;
229 
230 	/*
231 	 * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma
232 	 * list, after a COW of one of the file pages.	A MAP_SHARED vma
233 	 * can only be in the i_mmap tree.  An anonymous MAP_PRIVATE, stack
234 	 * or brk vma (with NULL file) can only be in an anon_vma list.
235 	 */
236 	struct list_head anon_vma_chain; /* Serialized by mmap_sem &
237 					  * page_table_lock */
238 	struct anon_vma *anon_vma;	/* Serialized by page_table_lock */
239 
240 	/* Function pointers to deal with this struct. */
241 	const struct vm_operations_struct *vm_ops;
242 
243 	/* Information about our backing store: */
244 	unsigned long vm_pgoff;		/* Offset (within vm_file) in PAGE_SIZE
245 					   units, *not* PAGE_CACHE_SIZE */
246 	struct file * vm_file;		/* File we map to (can be NULL). */
247 	void * vm_private_data;		/* was vm_pte (shared mem) */
248 
249 #ifndef CONFIG_MMU
250 	struct vm_region *vm_region;	/* NOMMU mapping region */
251 #endif
252 #ifdef CONFIG_NUMA
253 	struct mempolicy *vm_policy;	/* NUMA policy for the VMA */
254 #endif
255 };
256 
257 struct core_thread {
258 	struct task_struct *task;
259 	struct core_thread *next;
260 };
261 
262 struct core_state {
263 	atomic_t nr_threads;
264 	struct core_thread dumper;
265 	struct completion startup;
266 };
267 
268 enum {
269 	MM_FILEPAGES,
270 	MM_ANONPAGES,
271 	MM_SWAPENTS,
272 	NR_MM_COUNTERS
273 };
274 
275 #if USE_SPLIT_PTLOCKS && defined(CONFIG_MMU)
276 #define SPLIT_RSS_COUNTING
277 /* per-thread cached information, */
278 struct task_rss_stat {
279 	int events;	/* for synchronization threshold */
280 	int count[NR_MM_COUNTERS];
281 };
282 #endif /* USE_SPLIT_PTLOCKS */
283 
284 struct mm_rss_stat {
285 	atomic_long_t count[NR_MM_COUNTERS];
286 };
287 
288 struct mm_struct {
289 	struct vm_area_struct * mmap;		/* list of VMAs */
290 	struct rb_root mm_rb;
291 	struct vm_area_struct * mmap_cache;	/* last find_vma result */
292 #ifdef CONFIG_MMU
293 	unsigned long (*get_unmapped_area) (struct file *filp,
294 				unsigned long addr, unsigned long len,
295 				unsigned long pgoff, unsigned long flags);
296 	void (*unmap_area) (struct mm_struct *mm, unsigned long addr);
297 #endif
298 	unsigned long mmap_base;		/* base of mmap area */
299 	unsigned long task_size;		/* size of task vm space */
300 	unsigned long cached_hole_size; 	/* if non-zero, the largest hole below free_area_cache */
301 	unsigned long free_area_cache;		/* first hole of size cached_hole_size or larger */
302 	pgd_t * pgd;
303 	atomic_t mm_users;			/* How many users with user space? */
304 	atomic_t mm_count;			/* How many references to "struct mm_struct" (users count as 1) */
305 	int map_count;				/* number of VMAs */
306 
307 	spinlock_t page_table_lock;		/* Protects page tables and some counters */
308 	struct rw_semaphore mmap_sem;
309 
310 	struct list_head mmlist;		/* List of maybe swapped mm's.	These are globally strung
311 						 * together off init_mm.mmlist, and are protected
312 						 * by mmlist_lock
313 						 */
314 
315 
316 	unsigned long hiwater_rss;	/* High-watermark of RSS usage */
317 	unsigned long hiwater_vm;	/* High-water virtual memory usage */
318 
319 	unsigned long total_vm;		/* Total pages mapped */
320 	unsigned long locked_vm;	/* Pages that have PG_mlocked set */
321 	unsigned long pinned_vm;	/* Refcount permanently increased */
322 	unsigned long shared_vm;	/* Shared pages (files) */
323 	unsigned long exec_vm;		/* VM_EXEC & ~VM_WRITE */
324 	unsigned long stack_vm;		/* VM_GROWSUP/DOWN */
325 	unsigned long reserved_vm;	/* VM_RESERVED|VM_IO pages */
326 	unsigned long def_flags;
327 	unsigned long nr_ptes;		/* Page table pages */
328 	unsigned long start_code, end_code, start_data, end_data;
329 	unsigned long start_brk, brk, start_stack;
330 	unsigned long arg_start, arg_end, env_start, env_end;
331 
332 	unsigned long saved_auxv[AT_VECTOR_SIZE]; /* for /proc/PID/auxv */
333 
334 	/*
335 	 * Special counters, in some configurations protected by the
336 	 * page_table_lock, in other configurations by being atomic.
337 	 */
338 	struct mm_rss_stat rss_stat;
339 
340 	struct linux_binfmt *binfmt;
341 
342 	cpumask_var_t cpu_vm_mask_var;
343 
344 	/* Architecture-specific MM context */
345 	mm_context_t context;
346 
347 	/* Swap token stuff */
348 	/*
349 	 * Last value of global fault stamp as seen by this process.
350 	 * In other words, this value gives an indication of how long
351 	 * it has been since this task got the token.
352 	 * Look at mm/thrash.c
353 	 */
354 	unsigned int faultstamp;
355 	unsigned int token_priority;
356 	unsigned int last_interval;
357 
358 	unsigned long flags; /* Must use atomic bitops to access the bits */
359 
360 	struct core_state *core_state; /* coredumping support */
361 #ifdef CONFIG_AIO
362 	spinlock_t		ioctx_lock;
363 	struct hlist_head	ioctx_list;
364 #endif
365 #ifdef CONFIG_MM_OWNER
366 	/*
367 	 * "owner" points to a task that is regarded as the canonical
368 	 * user/owner of this mm. All of the following must be true in
369 	 * order for it to be changed:
370 	 *
371 	 * current == mm->owner
372 	 * current->mm != mm
373 	 * new_owner->mm == mm
374 	 * new_owner->alloc_lock is held
375 	 */
376 	struct task_struct __rcu *owner;
377 #endif
378 
379 	/* store ref to file /proc/<pid>/exe symlink points to */
380 	struct file *exe_file;
381 	unsigned long num_exe_file_vmas;
382 #ifdef CONFIG_MMU_NOTIFIER
383 	struct mmu_notifier_mm *mmu_notifier_mm;
384 #endif
385 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
386 	pgtable_t pmd_huge_pte; /* protected by page_table_lock */
387 #endif
388 #ifdef CONFIG_CPUMASK_OFFSTACK
389 	struct cpumask cpumask_allocation;
390 #endif
391 };
392 
mm_init_cpumask(struct mm_struct * mm)393 static inline void mm_init_cpumask(struct mm_struct *mm)
394 {
395 #ifdef CONFIG_CPUMASK_OFFSTACK
396 	mm->cpu_vm_mask_var = &mm->cpumask_allocation;
397 #endif
398 }
399 
400 /* Future-safe accessor for struct mm_struct's cpu_vm_mask. */
mm_cpumask(struct mm_struct * mm)401 static inline cpumask_t *mm_cpumask(struct mm_struct *mm)
402 {
403 	return mm->cpu_vm_mask_var;
404 }
405 
406 #endif /* _LINUX_MM_TYPES_H */
407