1 #ifndef _LINUX_MM_TYPES_H
2 #define _LINUX_MM_TYPES_H
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/rbtree.h>
10 #include <linux/rwsem.h>
11 #include <linux/completion.h>
12 #include <linux/cpumask.h>
13 #include <linux/page-debug-flags.h>
14 #include <linux/uprobes.h>
15 #include <linux/page-flags-layout.h>
19 #ifndef AT_VECTOR_SIZE_ARCH
20 #define AT_VECTOR_SIZE_ARCH 0
22 #define AT_VECTOR_SIZE (2*(AT_VECTOR_SIZE_ARCH + AT_VECTOR_SIZE_BASE + 1))
26 #define USE_SPLIT_PTLOCKS (NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS)
29 * Each physical page in the system has a struct page associated with
30 * it to keep track of whatever it is we are using the page for at the
31 * moment. Note that we have no way to track which tasks are using
32 * a page, though if it is a pagecache page, rmap structures can tell us
35 * The objects in struct page are organized in double word blocks in
36 * order to allows us to use atomic double word operations on portions
37 * of struct page. That is currently only used by slub but the arrangement
38 * allows the use of atomic double word operations on the flags/mapping
39 * and lru list pointers also.
42 /* First double word block */
43 unsigned long flags; /* Atomic flags, some possibly
44 * updated asynchronously */
45 struct address_space *mapping; /* If low bit clear, points to
46 * inode address_space, or NULL.
47 * If page mapped as anonymous
48 * memory, low bit is set, and
49 * it points to anon_vma object:
50 * see PAGE_MAPPING_ANON below.
52 /* Second double word */
55 pgoff_t index; /* Our offset within mapping. */
56 void *freelist; /* slub/slob first free object */
57 bool pfmemalloc; /* If set by the page allocator,
58 * ALLOC_NO_WATERMARKS was set
59 * and the low watermark was not
60 * met implying that the system
61 * is under some pressure. The
62 * caller should try ensure
63 * this page is only used to
69 #if defined(CONFIG_HAVE_CMPXCHG_DOUBLE) && \
70 defined(CONFIG_HAVE_ALIGNED_STRUCT_PAGE)
71 /* Used for cmpxchg_double in slub */
72 unsigned long counters;
75 * Keep _count separate from slub cmpxchg_double data.
76 * As the rest of the double word is protected by
77 * slab_lock but _count is not.
86 * Count of ptes mapped in
87 * mms, to show when page is
88 * mapped & limit reverse map
91 * Used also for tail pages
92 * refcounting instead of
93 * _count. Tail pages cannot
94 * be mapped and keeping the
95 * tail page _count zero at
96 * all times guarantees
97 * get_page_unless_zero() will
98 * never succeed on tail
108 int units; /* SLOB */
110 atomic_t _count; /* Usage count, see below. */
115 /* Third double word block */
117 struct list_head lru; /* Pageout list, eg. active_list
118 * protected by zone->lru_lock !
120 struct { /* slub per cpu partial pages */
121 struct page *next; /* Next partial slab */
123 int pages; /* Nr of partial slabs left */
124 int pobjects; /* Approximate # of objects */
131 struct list_head list; /* slobs list of pages */
132 struct slab *slab_page; /* slab fields */
135 /* Remainder is not double word aligned */
137 unsigned long private; /* Mapping-private opaque data:
138 * usually used for buffer_heads
139 * if PagePrivate set; used for
140 * swp_entry_t if PageSwapCache;
141 * indicates order in the buddy
142 * system if PG_buddy is set.
144 #if USE_SPLIT_PTLOCKS
147 struct kmem_cache *slab_cache; /* SL[AU]B: Pointer to slab */
148 struct page *first_page; /* Compound tail pages */
152 * On machines where all RAM is mapped into kernel address space,
153 * we can simply calculate the virtual address. On machines with
154 * highmem some memory is mapped into kernel virtual memory
155 * dynamically, so we need a place to store that address.
156 * Note that this field could be 16 bits on x86 ... ;)
158 * Architectures with slow multiplication can define
159 * WANT_PAGE_VIRTUAL in asm/page.h
161 #if defined(WANT_PAGE_VIRTUAL)
162 void *virtual; /* Kernel virtual address (NULL if
163 not kmapped, ie. highmem) */
164 #endif /* WANT_PAGE_VIRTUAL */
165 #ifdef CONFIG_WANT_PAGE_DEBUG_FLAGS
166 unsigned long debug_flags; /* Use atomic bitops on this */
169 #ifdef CONFIG_KMEMCHECK
171 * kmemcheck wants to track the status of each byte in a page; this
172 * is a pointer to such a status block. NULL if not tracked.
177 #ifdef LAST_NID_NOT_IN_PAGE_FLAGS
182 * The struct page can be forced to be double word aligned so that atomic ops
183 * on double words work. The SLUB allocator can make use of such a feature.
185 #ifdef CONFIG_HAVE_ALIGNED_STRUCT_PAGE
186 __aligned(2 * sizeof(unsigned long))
192 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
201 typedef unsigned long __nocast vm_flags_t;
204 * A region containing a mapping of a non-memory backed file under NOMMU
205 * conditions. These are held in a global tree and are pinned by the VMAs that
209 struct rb_node vm_rb; /* link in global region tree */
210 vm_flags_t vm_flags; /* VMA vm_flags */
211 unsigned long vm_start; /* start address of region */
212 unsigned long vm_end; /* region initialised to here */
213 unsigned long vm_top; /* region allocated to here */
214 unsigned long vm_pgoff; /* the offset in vm_file corresponding to vm_start */
215 struct file *vm_file; /* the backing file or NULL */
217 int vm_usage; /* region usage count (access under nommu_region_sem) */
218 bool vm_icache_flushed : 1; /* true if the icache has been flushed for
223 * This struct defines a memory VMM memory area. There is one of these
224 * per VM-area/task. A VM area is any part of the process virtual memory
225 * space that has a special rule for the page-fault handlers (ie a shared
226 * library, the executable area etc).
228 struct vm_area_struct {
229 /* The first cache line has the info for VMA tree walking. */
231 unsigned long vm_start; /* Our start address within vm_mm. */
232 unsigned long vm_end; /* The first byte after our end address
235 /* linked list of VM areas per task, sorted by address */
236 struct vm_area_struct *vm_next, *vm_prev;
238 struct rb_node vm_rb;
241 * Largest free memory gap in bytes to the left of this VMA.
242 * Either between this VMA and vma->vm_prev, or between one of the
243 * VMAs below us in the VMA rbtree and its ->vm_prev. This helps
244 * get_unmapped_area find a free area of the right size.
246 unsigned long rb_subtree_gap;
248 /* Second cache line starts here. */
250 struct mm_struct *vm_mm; /* The address space we belong to. */
251 pgprot_t vm_page_prot; /* Access permissions of this VMA. */
252 unsigned long vm_flags; /* Flags, see mm.h. */
255 * For areas with an address space and backing store,
256 * linkage into the address_space->i_mmap interval tree, or
257 * linkage of vma in the address_space->i_mmap_nonlinear list.
259 * For private anonymous mappings, a pointer to a null terminated string
260 * in the user process containing the name given to the vma, or NULL
266 unsigned long rb_subtree_last;
268 struct list_head nonlinear;
269 const char __user *anon_name;
273 * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma
274 * list, after a COW of one of the file pages. A MAP_SHARED vma
275 * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack
276 * or brk vma (with NULL file) can only be in an anon_vma list.
278 struct list_head anon_vma_chain; /* Serialized by mmap_sem &
280 struct anon_vma *anon_vma; /* Serialized by page_table_lock */
282 /* Function pointers to deal with this struct. */
283 const struct vm_operations_struct *vm_ops;
285 /* Information about our backing store: */
286 unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE
287 units, *not* PAGE_CACHE_SIZE */
288 struct file * vm_file; /* File we map to (can be NULL). */
289 void * vm_private_data; /* was vm_pte (shared mem) */
292 struct vm_region *vm_region; /* NOMMU mapping region */
295 struct mempolicy *vm_policy; /* NUMA policy for the VMA */
300 struct task_struct *task;
301 struct core_thread *next;
306 struct core_thread dumper;
307 struct completion startup;
317 #if USE_SPLIT_PTLOCKS && defined(CONFIG_MMU)
318 #define SPLIT_RSS_COUNTING
319 /* per-thread cached information, */
320 struct task_rss_stat {
321 int events; /* for synchronization threshold */
322 int count[NR_MM_COUNTERS];
324 #endif /* USE_SPLIT_PTLOCKS */
327 atomic_long_t count[NR_MM_COUNTERS];
331 struct vm_area_struct * mmap; /* list of VMAs */
332 struct rb_root mm_rb;
333 struct vm_area_struct * mmap_cache; /* last find_vma result */
335 unsigned long (*get_unmapped_area) (struct file *filp,
336 unsigned long addr, unsigned long len,
337 unsigned long pgoff, unsigned long flags);
338 void (*unmap_area) (struct mm_struct *mm, unsigned long addr);
340 unsigned long mmap_base; /* base of mmap area */
341 unsigned long mmap_legacy_base; /* base of mmap area in bottom-up allocations */
342 unsigned long task_size; /* size of task vm space */
343 unsigned long cached_hole_size; /* if non-zero, the largest hole below free_area_cache */
344 unsigned long free_area_cache; /* first hole of size cached_hole_size or larger */
345 unsigned long highest_vm_end; /* highest vma end address */
347 atomic_t mm_users; /* How many users with user space? */
348 atomic_t mm_count; /* How many references to "struct mm_struct" (users count as 1) */
349 int map_count; /* number of VMAs */
351 spinlock_t page_table_lock; /* Protects page tables and some counters */
352 struct rw_semaphore mmap_sem;
354 struct list_head mmlist; /* List of maybe swapped mm's. These are globally strung
355 * together off init_mm.mmlist, and are protected
360 unsigned long hiwater_rss; /* High-watermark of RSS usage */
361 unsigned long hiwater_vm; /* High-water virtual memory usage */
363 unsigned long total_vm; /* Total pages mapped */
364 unsigned long locked_vm; /* Pages that have PG_mlocked set */
365 unsigned long pinned_vm; /* Refcount permanently increased */
366 unsigned long shared_vm; /* Shared pages (files) */
367 unsigned long exec_vm; /* VM_EXEC & ~VM_WRITE */
368 unsigned long stack_vm; /* VM_GROWSUP/DOWN */
369 unsigned long def_flags;
370 unsigned long nr_ptes; /* Page table pages */
371 unsigned long start_code, end_code, start_data, end_data;
372 unsigned long start_brk, brk, start_stack;
373 unsigned long arg_start, arg_end, env_start, env_end;
375 unsigned long saved_auxv[AT_VECTOR_SIZE]; /* for /proc/PID/auxv */
378 * Special counters, in some configurations protected by the
379 * page_table_lock, in other configurations by being atomic.
381 struct mm_rss_stat rss_stat;
383 struct linux_binfmt *binfmt;
385 cpumask_var_t cpu_vm_mask_var;
387 /* Architecture-specific MM context */
388 mm_context_t context;
390 unsigned long flags; /* Must use atomic bitops to access the bits */
392 struct core_state *core_state; /* coredumping support */
394 spinlock_t ioctx_lock;
395 struct hlist_head ioctx_list;
397 #ifdef CONFIG_MM_OWNER
399 * "owner" points to a task that is regarded as the canonical
400 * user/owner of this mm. All of the following must be true in
401 * order for it to be changed:
403 * current == mm->owner
405 * new_owner->mm == mm
406 * new_owner->alloc_lock is held
408 struct task_struct __rcu *owner;
411 /* store ref to file /proc/<pid>/exe symlink points to */
412 struct file *exe_file;
413 #ifdef CONFIG_MMU_NOTIFIER
414 struct mmu_notifier_mm *mmu_notifier_mm;
416 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
417 pgtable_t pmd_huge_pte; /* protected by page_table_lock */
419 #ifdef CONFIG_CPUMASK_OFFSTACK
420 struct cpumask cpumask_allocation;
422 #ifdef CONFIG_NUMA_BALANCING
424 * numa_next_scan is the next time that the PTEs will be marked
425 * pte_numa. NUMA hinting faults will gather statistics and migrate
426 * pages to new nodes if necessary.
428 unsigned long numa_next_scan;
430 /* numa_next_reset is when the PTE scanner period will be reset */
431 unsigned long numa_next_reset;
433 /* Restart point for scanning and setting pte_numa */
434 unsigned long numa_scan_offset;
436 /* numa_scan_seq prevents two threads setting pte_numa */
440 * The first node a task was scheduled on. If a task runs on
441 * a different node than Make PTE Scan Go Now.
445 struct uprobes_state uprobes_state;
448 /* first nid will either be a valid NID or one of these values */
449 #define NUMA_PTE_SCAN_INIT -1
450 #define NUMA_PTE_SCAN_ACTIVE -2
452 static inline void mm_init_cpumask(struct mm_struct *mm)
454 #ifdef CONFIG_CPUMASK_OFFSTACK
455 mm->cpu_vm_mask_var = &mm->cpumask_allocation;
459 /* Future-safe accessor for struct mm_struct's cpu_vm_mask. */
460 static inline cpumask_t *mm_cpumask(struct mm_struct *mm)
462 return mm->cpu_vm_mask_var;
466 /* Return the name for an anonymous mapping or NULL for a file-backed mapping */
467 static inline const char __user *vma_get_anon_name(struct vm_area_struct *vma)
472 return vma->shared.anon_name;
475 #endif /* _LINUX_MM_TYPES_H */