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_PTE_PTLOCKS (NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS)
27 #define USE_SPLIT_PMD_PTLOCKS (USE_SPLIT_PTE_PTLOCKS && \
28 IS_ENABLED(CONFIG_ARCH_ENABLE_SPLIT_PMD_PTLOCK))
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
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.
44 /* First double word block */
45 unsigned long flags; /* Atomic flags, some possibly
46 * updated asynchronously */
48 struct address_space *mapping; /* If low bit clear, points to
49 * inode address_space, or NULL.
50 * If page mapped as anonymous
51 * memory, low bit is set, and
52 * it points to anon_vma object:
53 * see PAGE_MAPPING_ANON below.
55 void *s_mem; /* slab first object */
58 /* Second double word */
61 pgoff_t index; /* Our offset within mapping. */
62 void *freelist; /* sl[aou]b first free object */
63 bool pfmemalloc; /* If set by the page allocator,
64 * ALLOC_NO_WATERMARKS was set
65 * and the low watermark was not
66 * met implying that the system
67 * is under some pressure. The
68 * caller should try ensure
69 * this page is only used to
75 #if defined(CONFIG_HAVE_CMPXCHG_DOUBLE) && \
76 defined(CONFIG_HAVE_ALIGNED_STRUCT_PAGE)
77 /* Used for cmpxchg_double in slub */
78 unsigned long counters;
81 * Keep _count separate from slub cmpxchg_double data.
82 * As the rest of the double word is protected by
83 * slab_lock but _count is not.
92 * Count of ptes mapped in
93 * mms, to show when page is
94 * mapped & limit reverse map
97 * Used also for tail pages
98 * refcounting instead of
99 * _count. Tail pages cannot
100 * be mapped and keeping the
101 * tail page _count zero at
102 * all times guarantees
103 * get_page_unless_zero() will
104 * never succeed on tail
114 int units; /* SLOB */
116 atomic_t _count; /* Usage count, see below. */
118 unsigned int active; /* SLAB */
122 /* Third double word block */
124 struct list_head lru; /* Pageout list, eg. active_list
125 * protected by zone->lru_lock !
127 struct { /* slub per cpu partial pages */
128 struct page *next; /* Next partial slab */
130 int pages; /* Nr of partial slabs left */
131 int pobjects; /* Approximate # of objects */
138 struct list_head list; /* slobs list of pages */
139 struct slab *slab_page; /* slab fields */
140 struct rcu_head rcu_head; /* Used by SLAB
141 * when destroying via RCU
143 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && USE_SPLIT_PMD_PTLOCKS
144 pgtable_t pmd_huge_pte; /* protected by page->ptl */
148 /* Remainder is not double word aligned */
150 unsigned long private; /* Mapping-private opaque data:
151 * usually used for buffer_heads
152 * if PagePrivate set; used for
153 * swp_entry_t if PageSwapCache;
154 * indicates order in the buddy
155 * system if PG_buddy is set.
157 #if USE_SPLIT_PTE_PTLOCKS
158 #if BLOATED_SPINLOCKS
164 struct kmem_cache *slab_cache; /* SL[AU]B: Pointer to slab */
165 struct page *first_page; /* Compound tail pages */
169 * On machines where all RAM is mapped into kernel address space,
170 * we can simply calculate the virtual address. On machines with
171 * highmem some memory is mapped into kernel virtual memory
172 * dynamically, so we need a place to store that address.
173 * Note that this field could be 16 bits on x86 ... ;)
175 * Architectures with slow multiplication can define
176 * WANT_PAGE_VIRTUAL in asm/page.h
178 #if defined(WANT_PAGE_VIRTUAL)
179 void *virtual; /* Kernel virtual address (NULL if
180 not kmapped, ie. highmem) */
181 #endif /* WANT_PAGE_VIRTUAL */
182 #ifdef CONFIG_WANT_PAGE_DEBUG_FLAGS
183 unsigned long debug_flags; /* Use atomic bitops on this */
186 #ifdef CONFIG_KMEMCHECK
188 * kmemcheck wants to track the status of each byte in a page; this
189 * is a pointer to such a status block. NULL if not tracked.
194 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
199 * The struct page can be forced to be double word aligned so that atomic ops
200 * on double words work. The SLUB allocator can make use of such a feature.
202 #ifdef CONFIG_HAVE_ALIGNED_STRUCT_PAGE
203 __aligned(2 * sizeof(unsigned long))
209 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
218 typedef unsigned long __nocast vm_flags_t;
221 * A region containing a mapping of a non-memory backed file under NOMMU
222 * conditions. These are held in a global tree and are pinned by the VMAs that
226 struct rb_node vm_rb; /* link in global region tree */
227 vm_flags_t vm_flags; /* VMA vm_flags */
228 unsigned long vm_start; /* start address of region */
229 unsigned long vm_end; /* region initialised to here */
230 unsigned long vm_top; /* region allocated to here */
231 unsigned long vm_pgoff; /* the offset in vm_file corresponding to vm_start */
232 struct file *vm_file; /* the backing file or NULL */
234 int vm_usage; /* region usage count (access under nommu_region_sem) */
235 bool vm_icache_flushed : 1; /* true if the icache has been flushed for
240 * This struct defines a memory VMM memory area. There is one of these
241 * per VM-area/task. A VM area is any part of the process virtual memory
242 * space that has a special rule for the page-fault handlers (ie a shared
243 * library, the executable area etc).
245 struct vm_area_struct {
246 /* The first cache line has the info for VMA tree walking. */
248 unsigned long vm_start; /* Our start address within vm_mm. */
249 unsigned long vm_end; /* The first byte after our end address
252 /* linked list of VM areas per task, sorted by address */
253 struct vm_area_struct *vm_next, *vm_prev;
255 struct rb_node vm_rb;
258 * Largest free memory gap in bytes to the left of this VMA.
259 * Either between this VMA and vma->vm_prev, or between one of the
260 * VMAs below us in the VMA rbtree and its ->vm_prev. This helps
261 * get_unmapped_area find a free area of the right size.
263 unsigned long rb_subtree_gap;
265 /* Second cache line starts here. */
267 struct mm_struct *vm_mm; /* The address space we belong to. */
268 pgprot_t vm_page_prot; /* Access permissions of this VMA. */
269 unsigned long vm_flags; /* Flags, see mm.h. */
272 * For areas with an address space and backing store,
273 * linkage into the address_space->i_mmap interval tree, or
274 * linkage of vma in the address_space->i_mmap_nonlinear list.
279 unsigned long rb_subtree_last;
281 struct list_head nonlinear;
285 * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma
286 * list, after a COW of one of the file pages. A MAP_SHARED vma
287 * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack
288 * or brk vma (with NULL file) can only be in an anon_vma list.
290 struct list_head anon_vma_chain; /* Serialized by mmap_sem &
292 struct anon_vma *anon_vma; /* Serialized by page_table_lock */
294 /* Function pointers to deal with this struct. */
295 const struct vm_operations_struct *vm_ops;
297 /* Information about our backing store: */
298 unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE
299 units, *not* PAGE_CACHE_SIZE */
300 struct file * vm_file; /* File we map to (can be NULL). */
301 void * vm_private_data; /* was vm_pte (shared mem) */
304 struct vm_region *vm_region; /* NOMMU mapping region */
307 struct mempolicy *vm_policy; /* NUMA policy for the VMA */
312 struct task_struct *task;
313 struct core_thread *next;
318 struct core_thread dumper;
319 struct completion startup;
329 #if USE_SPLIT_PTE_PTLOCKS && defined(CONFIG_MMU)
330 #define SPLIT_RSS_COUNTING
331 /* per-thread cached information, */
332 struct task_rss_stat {
333 int events; /* for synchronization threshold */
334 int count[NR_MM_COUNTERS];
336 #endif /* USE_SPLIT_PTE_PTLOCKS */
339 atomic_long_t count[NR_MM_COUNTERS];
344 struct vm_area_struct * mmap; /* list of VMAs */
345 struct rb_root mm_rb;
346 struct vm_area_struct * mmap_cache; /* last find_vma result */
348 unsigned long (*get_unmapped_area) (struct file *filp,
349 unsigned long addr, unsigned long len,
350 unsigned long pgoff, unsigned long flags);
352 unsigned long mmap_base; /* base of mmap area */
353 unsigned long mmap_legacy_base; /* base of mmap area in bottom-up allocations */
354 unsigned long task_size; /* size of task vm space */
355 unsigned long highest_vm_end; /* highest vma end address */
357 atomic_t mm_users; /* How many users with user space? */
358 atomic_t mm_count; /* How many references to "struct mm_struct" (users count as 1) */
359 atomic_long_t nr_ptes; /* Page table pages */
360 int map_count; /* number of VMAs */
362 spinlock_t page_table_lock; /* Protects page tables and some counters */
363 struct rw_semaphore mmap_sem;
365 struct list_head mmlist; /* List of maybe swapped mm's. These are globally strung
366 * together off init_mm.mmlist, and are protected
371 unsigned long hiwater_rss; /* High-watermark of RSS usage */
372 unsigned long hiwater_vm; /* High-water virtual memory usage */
374 unsigned long total_vm; /* Total pages mapped */
375 unsigned long locked_vm; /* Pages that have PG_mlocked set */
376 unsigned long pinned_vm; /* Refcount permanently increased */
377 unsigned long shared_vm; /* Shared pages (files) */
378 unsigned long exec_vm; /* VM_EXEC & ~VM_WRITE */
379 unsigned long stack_vm; /* VM_GROWSUP/DOWN */
380 unsigned long def_flags;
381 unsigned long start_code, end_code, start_data, end_data;
382 unsigned long start_brk, brk, start_stack;
383 unsigned long arg_start, arg_end, env_start, env_end;
385 unsigned long saved_auxv[AT_VECTOR_SIZE]; /* for /proc/PID/auxv */
388 * Special counters, in some configurations protected by the
389 * page_table_lock, in other configurations by being atomic.
391 struct mm_rss_stat rss_stat;
393 struct linux_binfmt *binfmt;
395 cpumask_var_t cpu_vm_mask_var;
397 /* Architecture-specific MM context */
398 mm_context_t context;
400 unsigned long flags; /* Must use atomic bitops to access the bits */
402 struct core_state *core_state; /* coredumping support */
404 spinlock_t ioctx_lock;
405 struct kioctx_table __rcu *ioctx_table;
407 #ifdef CONFIG_MM_OWNER
409 * "owner" points to a task that is regarded as the canonical
410 * user/owner of this mm. All of the following must be true in
411 * order for it to be changed:
413 * current == mm->owner
415 * new_owner->mm == mm
416 * new_owner->alloc_lock is held
418 struct task_struct __rcu *owner;
421 /* store ref to file /proc/<pid>/exe symlink points to */
422 struct file *exe_file;
423 #ifdef CONFIG_MMU_NOTIFIER
424 struct mmu_notifier_mm *mmu_notifier_mm;
426 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
427 pgtable_t pmd_huge_pte; /* protected by page_table_lock */
429 #ifdef CONFIG_CPUMASK_OFFSTACK
430 struct cpumask cpumask_allocation;
432 #ifdef CONFIG_NUMA_BALANCING
434 * numa_next_scan is the next time that the PTEs will be marked
435 * pte_numa. NUMA hinting faults will gather statistics and migrate
436 * pages to new nodes if necessary.
438 unsigned long numa_next_scan;
440 /* Restart point for scanning and setting pte_numa */
441 unsigned long numa_scan_offset;
443 /* numa_scan_seq prevents two threads setting pte_numa */
446 #if defined(CONFIG_NUMA_BALANCING) || defined(CONFIG_COMPACTION)
448 * An operation with batched TLB flushing is going on. Anything that
449 * can move process memory needs to flush the TLB when moving a
450 * PROT_NONE or PROT_NUMA mapped page.
452 bool tlb_flush_pending;
454 struct uprobes_state uprobes_state;
457 static inline void mm_init_cpumask(struct mm_struct *mm)
459 #ifdef CONFIG_CPUMASK_OFFSTACK
460 mm->cpu_vm_mask_var = &mm->cpumask_allocation;
464 /* Future-safe accessor for struct mm_struct's cpu_vm_mask. */
465 static inline cpumask_t *mm_cpumask(struct mm_struct *mm)
467 return mm->cpu_vm_mask_var;
470 #if defined(CONFIG_NUMA_BALANCING) || defined(CONFIG_COMPACTION)
472 * Memory barriers to keep this state in sync are graciously provided by
473 * the page table locks, outside of which no page table modifications happen.
474 * The barriers below prevent the compiler from re-ordering the instructions
475 * around the memory barriers that are already present in the code.
477 static inline bool mm_tlb_flush_pending(struct mm_struct *mm)
480 return mm->tlb_flush_pending;
482 static inline void set_tlb_flush_pending(struct mm_struct *mm)
484 mm->tlb_flush_pending = true;
487 * Guarantee that the tlb_flush_pending store does not leak into the
488 * critical section updating the page tables
490 smp_mb__before_spinlock();
492 /* Clearing is done after a TLB flush, which also provides a barrier. */
493 static inline void clear_tlb_flush_pending(struct mm_struct *mm)
496 mm->tlb_flush_pending = false;
499 static inline bool mm_tlb_flush_pending(struct mm_struct *mm)
503 static inline void set_tlb_flush_pending(struct mm_struct *mm)
506 static inline void clear_tlb_flush_pending(struct mm_struct *mm)
511 #endif /* _LINUX_MM_TYPES_H */