2 * Copyright (c) 1991, 1993
3 * The Regents of the University of California. All rights reserved.
5 * This code is derived from software contributed to Berkeley by
6 * The Mach Operating System project at Carnegie-Mellon University.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of the University nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * from: @(#)vm_page.h 8.2 (Berkeley) 12/13/93
35 * Copyright (c) 1987, 1990 Carnegie-Mellon University.
36 * All rights reserved.
38 * Authors: Avadis Tevanian, Jr., Michael Wayne Young
40 * Permission to use, copy, modify and distribute this software and
41 * its documentation is hereby granted, provided that both the copyright
42 * notice and this permission notice appear in all copies of the
43 * software, derivative works or modified versions, and any portions
44 * thereof, and that both notices appear in supporting documentation.
46 * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
47 * CONDITION. CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND
48 * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
50 * Carnegie Mellon requests users of this software to return to
52 * Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU
53 * School of Computer Science
54 * Carnegie Mellon University
55 * Pittsburgh PA 15213-3890
57 * any improvements or extensions that they make and grant Carnegie the
58 * rights to redistribute these changes.
60 * $FreeBSD: src/sys/vm/vm_page.h,v 1.75.2.8 2002/03/06 01:07:09 dillon Exp $
64 * Resident memory system definitions.
67 #ifndef _VM_VM_PAGE_H_
68 #define _VM_VM_PAGE_H_
71 #include <sys/types.h>
76 #ifndef _MACHINE_PMAP_H_
77 #include <machine/pmap.h>
82 #include <machine/atomic.h>
87 #include <sys/systm.h>
89 #ifndef _SYS_THREAD2_H_
90 #include <sys/thread2.h>
94 #include <machine/vmparam.h>
99 typedef enum vm_page_event { VMEVENT_NONE, VMEVENT_COW } vm_page_event_t;
101 struct vm_page_action {
102 LIST_ENTRY(vm_page_action) entry;
104 vm_page_event_t event;
105 void (*func)(struct vm_page *,
106 struct vm_page_action *);
110 typedef struct vm_page_action *vm_page_action_t;
113 * Management of resident (logical) pages.
115 * A small structure is kept for each resident
116 * page, indexed by page number. Each structure
117 * is an element of several lists:
119 * A hash table bucket used to quickly
120 * perform object/offset lookups
122 * A list of all pages for a given object,
123 * so they can be quickly deactivated at
124 * time of deallocation.
126 * An ordered list of pages due for pageout.
128 * In addition, the structure contains the object
129 * and offset to which this page belongs (for pageout),
130 * and sundry status bits.
132 * Fields in this structure are locked either by the lock on the
133 * object that the page belongs to (O) or by the lock on the page
136 * The 'valid' and 'dirty' fields are distinct. A page may have dirty
137 * bits set without having associated valid bits set. This is used by
138 * NFS to implement piecemeal writes.
141 TAILQ_HEAD(pglist, vm_page);
145 int rb_vm_page_compare(struct vm_page *, struct vm_page *);
147 struct vm_page_rb_tree;
148 RB_PROTOTYPE2(vm_page_rb_tree, vm_page, rb_entry, rb_vm_page_compare, vm_pindex_t);
151 TAILQ_ENTRY(vm_page) pageq; /* vm_page_queues[] list (P) */
152 RB_ENTRY(vm_page) rb_entry; /* Red-Black tree based at object */
154 struct vm_object *object; /* which object am I in (O,P)*/
155 vm_pindex_t pindex; /* offset into object (O,P) */
156 vm_paddr_t phys_addr; /* physical address of page */
157 struct md_page md; /* machine dependant stuff */
158 u_short queue; /* page queue index */
159 u_short pc; /* page color */
160 u_char act_count; /* page usage count */
161 u_char busy; /* page busy count */
162 u_char pat_mode; /* hardware page attribute */
164 u_int32_t flags; /* see below */
165 u_int wire_count; /* wired down maps refs (P) */
166 int hold_count; /* page hold count */
169 * NOTE that these must support one bit per DEV_BSIZE in a page!!!
170 * so, on normal X86 kernels, they must be at least 8 bits wide.
172 u_char valid; /* map of valid DEV_BSIZE chunks */
173 u_char dirty; /* map of dirty DEV_BSIZE chunks */
175 int ku_pagecnt; /* kmalloc helper */
177 const char *busy_func;
183 #define VM_PAGE_DEBUG_EXT(name) name ## _debug
184 #define VM_PAGE_DEBUG_ARGS , const char *func, int lineno
186 #define VM_PAGE_DEBUG_EXT(name) name
187 #define VM_PAGE_DEBUG_ARGS
190 #ifndef __VM_PAGE_T_DEFINED__
191 #define __VM_PAGE_T_DEFINED__
192 typedef struct vm_page *vm_page_t;
196 * Page coloring parameters. We use generous parameters designed to
197 * statistically spread pages over available cpu cache space. This has
198 * become less important over time as cache associativity is higher
199 * in modern times but we still use the core algorithm to help reduce
200 * lock contention between cpus.
202 * Page coloring cannot be disabled.
205 #define PQ_PRIME1 31 /* Prime number somewhat less than PQ_HASH_SIZE */
206 #define PQ_PRIME2 23 /* Prime number somewhat less than PQ_HASH_SIZE */
207 #define PQ_L2_SIZE 256 /* A number of colors opt for 1M cache */
208 #define PQ_L2_MASK (PQ_L2_SIZE - 1)
211 #define PQ_FREE (1 + 0*PQ_L2_SIZE)
212 #define PQ_INACTIVE (1 + 1*PQ_L2_SIZE)
213 #define PQ_ACTIVE (1 + 2*PQ_L2_SIZE)
214 #define PQ_CACHE (1 + 3*PQ_L2_SIZE)
215 #define PQ_HOLD (1 + 4*PQ_L2_SIZE)
216 #define PQ_COUNT (1 + 5*PQ_L2_SIZE)
223 struct rb_vm_page_scan_info {
224 vm_pindex_t start_pindex;
225 vm_pindex_t end_pindex;
233 vm_pindex_t backing_offset_index;
234 struct vm_object *object;
235 struct vm_object *backing_object;
236 struct vm_page *mpte;
241 int rb_vm_page_scancmp(struct vm_page *, void *);
247 int flipflop; /* probably not the best place */
248 struct spinlock spin;
249 int zero_count; /* only applies to PQ_FREE queues */
251 char unused[64 - sizeof(struct pglist) -
252 sizeof(int *) - sizeof(int) * 4];
255 extern struct vpgqueues vm_page_queues[PQ_COUNT];
258 * These are the flags defined for vm_page.
260 * PG_UNMANAGED (used by OBJT_PHYS) indicates that the page is
261 * not under PV management but otherwise should be treated as a
262 * normal page. Pages not under PV management cannot be paged out
263 * via the object/vm_page_t because there is no knowledge of their
264 * pte mappings, nor can they be removed from their objects via
265 * the object, and such pages are also not on any PQ queue. The
266 * PG_MAPPED and PG_WRITEABLE flags are not applicable.
268 * PG_MAPPED only applies to managed pages, indicating whether the page
269 * is mapped onto one or more pmaps. A page might still be mapped to
270 * special pmaps in an unmanaged fashion, for example when mapped into a
271 * buffer cache buffer, without setting PG_MAPPED.
273 * PG_WRITEABLE indicates that there may be a writeable managed pmap entry
274 * somewhere, and that the page can be dirtied by hardware at any time
275 * and may have to be tested for that. The modified bit in unmanaged
276 * mappings or in the special clean map is not tested.
278 * PG_SWAPPED indicates that the page is backed by a swap block. Any
279 * VM object type other than OBJT_DEFAULT can have swap-backed pages now.
281 * PG_SBUSY is set when m->busy != 0. PG_SBUSY and m->busy are only modified
282 * when the page is PG_BUSY.
284 #define PG_BUSY 0x00000001 /* page is in transit (O) */
285 #define PG_WANTED 0x00000002 /* someone is waiting for page (O) */
286 #define PG_WINATCFLS 0x00000004 /* flush dirty page on inactive q */
287 #define PG_FICTITIOUS 0x00000008 /* physical page doesn't exist (O) */
288 #define PG_WRITEABLE 0x00000010 /* page is writeable */
289 #define PG_MAPPED 0x00000020 /* page is mapped (managed) */
290 #define PG_UNUSED0040 0x00000040
291 #define PG_REFERENCED 0x00000080 /* page has been referenced */
292 #define PG_CLEANCHK 0x00000100 /* page will be checked for cleaning */
293 #define PG_SWAPINPROG 0x00000200 /* swap I/O in progress on page */
294 #define PG_NOSYNC 0x00000400 /* do not collect for syncer */
295 #define PG_UNMANAGED 0x00000800 /* No PV management for page */
296 #define PG_MARKER 0x00001000 /* special queue marker page */
297 #define PG_RAM 0x00002000 /* read ahead mark */
298 #define PG_SWAPPED 0x00004000 /* backed by swap */
299 #define PG_NOTMETA 0x00008000 /* do not back with swap */
300 #define PG_ACTIONLIST 0x00010000 /* lookaside action list present */
301 #define PG_SBUSY 0x00020000 /* soft-busy also set */
302 #define PG_NEED_COMMIT 0x00040000 /* clean page requires commit */
308 #define ACT_DECLINE 1
309 #define ACT_ADVANCE 3
315 * Each pageable resident page falls into one of four lists:
318 * Available for allocation now.
320 * The following are all LRU sorted:
323 * Almost available for allocation. Still in an
324 * object, but clean and immediately freeable at
325 * non-interrupt times.
328 * Low activity, candidates for reclamation.
329 * This is the list of pages that should be
333 * Pages that are "active" i.e. they have been
334 * recently referenced.
337 * Pages that are really free and have been pre-zeroed
341 extern struct vm_page *vm_page_array; /* First resident page in table */
342 extern int vm_page_array_size; /* number of vm_page_t's */
343 extern long first_page; /* first physical page number */
345 #define VM_PAGE_TO_PHYS(entry) \
348 #define PHYS_TO_VM_PAGE(pa) \
349 (&vm_page_array[atop(pa) - first_page])
352 #if PAGE_SIZE == 4096
353 #define VM_PAGE_BITS_ALL 0xff
357 * Note: the code will always use nominally free pages from the free list
358 * before trying other flag-specified sources.
360 * At least one of VM_ALLOC_NORMAL|VM_ALLOC_SYSTEM|VM_ALLOC_INTERRUPT
361 * must be specified. VM_ALLOC_RETRY may only be specified if VM_ALLOC_NORMAL
364 #define VM_ALLOC_NORMAL 0x0001 /* ok to use cache pages */
365 #define VM_ALLOC_SYSTEM 0x0002 /* ok to exhaust most of free list */
366 #define VM_ALLOC_INTERRUPT 0x0004 /* ok to exhaust entire free list */
367 #define VM_ALLOC_ZERO 0x0008 /* req pre-zero'd memory if avail */
368 #define VM_ALLOC_QUICK 0x0010 /* like NORMAL but do not use cache */
369 #define VM_ALLOC_FORCE_ZERO 0x0020 /* zero page even if already valid */
370 #define VM_ALLOC_NULL_OK 0x0040 /* ok to return NULL on collision */
371 #define VM_ALLOC_RETRY 0x0080 /* indefinite block (vm_page_grab()) */
372 #define VM_ALLOC_USE_GD 0x0100 /* use per-gd cache */
373 #define VM_ALLOC_CPU_SPEC 0x0200
375 #define VM_ALLOC_CPU_SHIFT 16
376 #define VM_ALLOC_CPU(n) (((n) << VM_ALLOC_CPU_SHIFT) | \
378 #define VM_ALLOC_GETCPU(flags) ((flags) >> VM_ALLOC_CPU_SHIFT)
380 void vm_page_queue_spin_lock(vm_page_t);
381 void vm_page_queues_spin_lock(u_short);
382 void vm_page_and_queue_spin_lock(vm_page_t);
384 void vm_page_queue_spin_unlock(vm_page_t);
385 void vm_page_queues_spin_unlock(u_short);
386 void vm_page_and_queue_spin_unlock(vm_page_t m);
388 void vm_page_init(vm_page_t m);
389 void vm_page_io_finish(vm_page_t m);
390 void vm_page_io_start(vm_page_t m);
391 void vm_page_need_commit(vm_page_t m);
392 void vm_page_clear_commit(vm_page_t m);
393 void vm_page_wakeup(vm_page_t m);
394 void vm_page_hold(vm_page_t);
395 void vm_page_unhold(vm_page_t);
396 void vm_page_activate (vm_page_t);
398 vm_size_t vm_contig_avail_pages(void);
399 vm_page_t vm_page_alloc (struct vm_object *, vm_pindex_t, int);
400 vm_page_t vm_page_alloc_contig(vm_paddr_t low, vm_paddr_t high,
401 unsigned long alignment, unsigned long boundary,
402 unsigned long size, vm_memattr_t memattr);
404 vm_page_t vm_page_grab (struct vm_object *, vm_pindex_t, int);
405 void vm_page_cache (vm_page_t);
406 int vm_page_try_to_cache (vm_page_t);
407 int vm_page_try_to_free (vm_page_t);
408 void vm_page_dontneed (vm_page_t);
409 void vm_page_deactivate (vm_page_t);
410 void vm_page_deactivate_locked (vm_page_t);
411 void vm_page_initfake(vm_page_t m, vm_paddr_t paddr, vm_memattr_t memattr);
412 int vm_page_insert (vm_page_t, struct vm_object *, vm_pindex_t);
413 vm_page_t vm_page_lookup (struct vm_object *, vm_pindex_t);
414 vm_page_t VM_PAGE_DEBUG_EXT(vm_page_lookup_busy_wait)(
415 struct vm_object *, vm_pindex_t, int, const char *
417 vm_page_t VM_PAGE_DEBUG_EXT(vm_page_lookup_busy_try)(
418 struct vm_object *, vm_pindex_t, int, int *
420 vm_page_t vm_page_repurpose(struct vm_object *, vm_pindex_t, int, int *,
421 vm_page_t, int *, int *);
422 void vm_page_remove (vm_page_t);
423 void vm_page_rename (vm_page_t, struct vm_object *, vm_pindex_t);
424 void vm_page_startup (void);
425 void vm_numa_organize(vm_paddr_t ran_beg, vm_paddr_t bytes, int physid);
426 void vm_page_unmanage (vm_page_t);
427 void vm_page_unwire (vm_page_t, int);
428 void vm_page_wire (vm_page_t);
429 void vm_page_unqueue (vm_page_t);
430 void vm_page_unqueue_nowakeup (vm_page_t);
431 vm_page_t vm_page_next (vm_page_t);
432 void vm_page_set_validclean (vm_page_t, int, int);
433 void vm_page_set_validdirty (vm_page_t, int, int);
434 void vm_page_set_valid (vm_page_t, int, int);
435 void vm_page_set_dirty (vm_page_t, int, int);
436 void vm_page_clear_dirty (vm_page_t, int, int);
437 void vm_page_set_invalid (vm_page_t, int, int);
438 int vm_page_is_valid (vm_page_t, int, int);
439 void vm_page_test_dirty (vm_page_t);
440 int vm_page_bits (int, int);
441 vm_page_t vm_page_list_find(int basequeue, int index, boolean_t prefer_zero);
442 void vm_page_zero_invalid(vm_page_t m, boolean_t setvalid);
443 void vm_page_free_toq(vm_page_t m);
444 void vm_page_free_contig(vm_page_t m, unsigned long size);
445 vm_page_t vm_page_free_fromq_fast(void);
446 void vm_page_event_internal(vm_page_t, vm_page_event_t);
447 void vm_page_dirty(vm_page_t m);
448 void vm_page_register_action(vm_page_action_t action, vm_page_event_t event);
449 void vm_page_unregister_action(vm_page_action_t action);
450 void vm_page_sleep_busy(vm_page_t m, int also_m_busy, const char *msg);
451 void VM_PAGE_DEBUG_EXT(vm_page_busy_wait)(vm_page_t m,
452 int also_m_busy, const char *wmsg VM_PAGE_DEBUG_ARGS);
453 int VM_PAGE_DEBUG_EXT(vm_page_busy_try)(vm_page_t m,
454 int also_m_busy VM_PAGE_DEBUG_ARGS);
455 u_short vm_get_pg_color(int cpuid, vm_object_t object, vm_pindex_t pindex);
459 #define vm_page_lookup_busy_wait(object, pindex, alsob, msg) \
460 vm_page_lookup_busy_wait_debug(object, pindex, alsob, msg, \
463 #define vm_page_lookup_busy_try(object, pindex, alsob, errorp) \
464 vm_page_lookup_busy_try_debug(object, pindex, alsob, errorp, \
467 #define vm_page_busy_wait(m, alsob, msg) \
468 vm_page_busy_wait_debug(m, alsob, msg, __func__, __LINE__)
470 #define vm_page_busy_try(m, alsob) \
471 vm_page_busy_try_debug(m, alsob, __func__, __LINE__)
476 #endif /* !_VM_VM_PAGE_H_ */